summaryrefslogtreecommitdiff
path: root/libavcodec/h264dec.c (plain)
blob: cb2b4416d8461eb2c02ba32cab99056816e005cd
1/*
2 * H.26L/H.264/AVC/JVT/14496-10/... decoder
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 codec.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
27
28#define UNCHECKED_BITSTREAM_READER 1
29
30#include "libavutil/avassert.h"
31#include "libavutil/display.h"
32#include "libavutil/imgutils.h"
33#include "libavutil/opt.h"
34#include "libavutil/stereo3d.h"
35#include "libavutil/timer.h"
36#include "internal.h"
37#include "bytestream.h"
38#include "cabac.h"
39#include "cabac_functions.h"
40#include "error_resilience.h"
41#include "avcodec.h"
42#include "h264.h"
43#include "h264dec.h"
44#include "h2645_parse.h"
45#include "h264data.h"
46#include "h264chroma.h"
47#include "h264_mvpred.h"
48#include "h264_ps.h"
49#include "golomb.h"
50#include "mathops.h"
51#include "me_cmp.h"
52#include "mpegutils.h"
53#include "profiles.h"
54#include "rectangle.h"
55#include "thread.h"
56#include "vdpau_compat.h"
57
58static int h264_decode_end(AVCodecContext *avctx);
59
60const uint16_t ff_h264_mb_sizes[4] = { 256, 384, 512, 768 };
61
62int avpriv_h264_has_num_reorder_frames(AVCodecContext *avctx)
63{
64 H264Context *h = avctx->priv_data;
65 return h && h->ps.sps ? h->ps.sps->num_reorder_frames : 0;
66}
67
68static void h264_er_decode_mb(void *opaque, int ref, int mv_dir, int mv_type,
69 int (*mv)[2][4][2],
70 int mb_x, int mb_y, int mb_intra, int mb_skipped)
71{
72 H264Context *h = opaque;
73 H264SliceContext *sl = &h->slice_ctx[0];
74
75 sl->mb_x = mb_x;
76 sl->mb_y = mb_y;
77 sl->mb_xy = mb_x + mb_y * h->mb_stride;
78 memset(sl->non_zero_count_cache, 0, sizeof(sl->non_zero_count_cache));
79 av_assert1(ref >= 0);
80 /* FIXME: It is possible albeit uncommon that slice references
81 * differ between slices. We take the easy approach and ignore
82 * it for now. If this turns out to have any relevance in
83 * practice then correct remapping should be added. */
84 if (ref >= sl->ref_count[0])
85 ref = 0;
86 if (!sl->ref_list[0][ref].data[0]) {
87 av_log(h->avctx, AV_LOG_DEBUG, "Reference not available for error concealing\n");
88 ref = 0;
89 }
90 if ((sl->ref_list[0][ref].reference&3) != 3) {
91 av_log(h->avctx, AV_LOG_DEBUG, "Reference invalid\n");
92 return;
93 }
94 fill_rectangle(&h->cur_pic.ref_index[0][4 * sl->mb_xy],
95 2, 2, 2, ref, 1);
96 fill_rectangle(&sl->ref_cache[0][scan8[0]], 4, 4, 8, ref, 1);
97 fill_rectangle(sl->mv_cache[0][scan8[0]], 4, 4, 8,
98 pack16to32((*mv)[0][0][0], (*mv)[0][0][1]), 4);
99 sl->mb_mbaff =
100 sl->mb_field_decoding_flag = 0;
101 ff_h264_hl_decode_mb(h, &h->slice_ctx[0]);
102}
103
104void ff_h264_draw_horiz_band(const H264Context *h, H264SliceContext *sl,
105 int y, int height)
106{
107 AVCodecContext *avctx = h->avctx;
108 const AVFrame *src = h->cur_pic.f;
109 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(avctx->pix_fmt);
110 int vshift = desc->log2_chroma_h;
111 const int field_pic = h->picture_structure != PICT_FRAME;
112 if (field_pic) {
113 height <<= 1;
114 y <<= 1;
115 }
116
117 height = FFMIN(height, avctx->height - y);
118
119 if (field_pic && h->first_field && !(avctx->slice_flags & SLICE_FLAG_ALLOW_FIELD))
120 return;
121
122 if (avctx->draw_horiz_band) {
123 int offset[AV_NUM_DATA_POINTERS];
124 int i;
125
126 offset[0] = y * src->linesize[0];
127 offset[1] =
128 offset[2] = (y >> vshift) * src->linesize[1];
129 for (i = 3; i < AV_NUM_DATA_POINTERS; i++)
130 offset[i] = 0;
131
132 emms_c();
133
134 avctx->draw_horiz_band(avctx, src, offset,
135 y, h->picture_structure, height);
136 }
137}
138
139void ff_h264_free_tables(H264Context *h)
140{
141 int i;
142
143 av_freep(&h->intra4x4_pred_mode);
144 av_freep(&h->chroma_pred_mode_table);
145 av_freep(&h->cbp_table);
146 av_freep(&h->mvd_table[0]);
147 av_freep(&h->mvd_table[1]);
148 av_freep(&h->direct_table);
149 av_freep(&h->non_zero_count);
150 av_freep(&h->slice_table_base);
151 h->slice_table = NULL;
152 av_freep(&h->list_counts);
153
154 av_freep(&h->mb2b_xy);
155 av_freep(&h->mb2br_xy);
156
157 av_buffer_pool_uninit(&h->qscale_table_pool);
158 av_buffer_pool_uninit(&h->mb_type_pool);
159 av_buffer_pool_uninit(&h->motion_val_pool);
160 av_buffer_pool_uninit(&h->ref_index_pool);
161
162 for (i = 0; i < h->nb_slice_ctx; i++) {
163 H264SliceContext *sl = &h->slice_ctx[i];
164
165 av_freep(&sl->dc_val_base);
166 av_freep(&sl->er.mb_index2xy);
167 av_freep(&sl->er.error_status_table);
168 av_freep(&sl->er.er_temp_buffer);
169
170 av_freep(&sl->bipred_scratchpad);
171 av_freep(&sl->edge_emu_buffer);
172 av_freep(&sl->top_borders[0]);
173 av_freep(&sl->top_borders[1]);
174
175 sl->bipred_scratchpad_allocated = 0;
176 sl->edge_emu_buffer_allocated = 0;
177 sl->top_borders_allocated[0] = 0;
178 sl->top_borders_allocated[1] = 0;
179 }
180}
181
182int ff_h264_alloc_tables(H264Context *h)
183{
184 const int big_mb_num = h->mb_stride * (h->mb_height + 1);
185 const int row_mb_num = 2*h->mb_stride*FFMAX(h->nb_slice_ctx, 1);
186 int x, y;
187
188 FF_ALLOCZ_ARRAY_OR_GOTO(h->avctx, h->intra4x4_pred_mode,
189 row_mb_num, 8 * sizeof(uint8_t), fail)
190 h->slice_ctx[0].intra4x4_pred_mode = h->intra4x4_pred_mode;
191
192 FF_ALLOCZ_OR_GOTO(h->avctx, h->non_zero_count,
193 big_mb_num * 48 * sizeof(uint8_t), fail)
194 FF_ALLOCZ_OR_GOTO(h->avctx, h->slice_table_base,
195 (big_mb_num + h->mb_stride) * sizeof(*h->slice_table_base), fail)
196 FF_ALLOCZ_OR_GOTO(h->avctx, h->cbp_table,
197 big_mb_num * sizeof(uint16_t), fail)
198 FF_ALLOCZ_OR_GOTO(h->avctx, h->chroma_pred_mode_table,
199 big_mb_num * sizeof(uint8_t), fail)
200 FF_ALLOCZ_ARRAY_OR_GOTO(h->avctx, h->mvd_table[0],
201 row_mb_num, 16 * sizeof(uint8_t), fail);
202 FF_ALLOCZ_ARRAY_OR_GOTO(h->avctx, h->mvd_table[1],
203 row_mb_num, 16 * sizeof(uint8_t), fail);
204 h->slice_ctx[0].mvd_table[0] = h->mvd_table[0];
205 h->slice_ctx[0].mvd_table[1] = h->mvd_table[1];
206
207 FF_ALLOCZ_OR_GOTO(h->avctx, h->direct_table,
208 4 * big_mb_num * sizeof(uint8_t), fail);
209 FF_ALLOCZ_OR_GOTO(h->avctx, h->list_counts,
210 big_mb_num * sizeof(uint8_t), fail)
211
212 memset(h->slice_table_base, -1,
213 (big_mb_num + h->mb_stride) * sizeof(*h->slice_table_base));
214 h->slice_table = h->slice_table_base + h->mb_stride * 2 + 1;
215
216 FF_ALLOCZ_OR_GOTO(h->avctx, h->mb2b_xy,
217 big_mb_num * sizeof(uint32_t), fail);
218 FF_ALLOCZ_OR_GOTO(h->avctx, h->mb2br_xy,
219 big_mb_num * sizeof(uint32_t), fail);
220 for (y = 0; y < h->mb_height; y++)
221 for (x = 0; x < h->mb_width; x++) {
222 const int mb_xy = x + y * h->mb_stride;
223 const int b_xy = 4 * x + 4 * y * h->b_stride;
224
225 h->mb2b_xy[mb_xy] = b_xy;
226 h->mb2br_xy[mb_xy] = 8 * (FMO ? mb_xy : (mb_xy % (2 * h->mb_stride)));
227 }
228
229 return 0;
230
231fail:
232 ff_h264_free_tables(h);
233 return AVERROR(ENOMEM);
234}
235
236/**
237 * Init context
238 * Allocate buffers which are not shared amongst multiple threads.
239 */
240int ff_h264_slice_context_init(H264Context *h, H264SliceContext *sl)
241{
242 ERContext *er = &sl->er;
243 int mb_array_size = h->mb_height * h->mb_stride;
244 int y_size = (2 * h->mb_width + 1) * (2 * h->mb_height + 1);
245 int c_size = h->mb_stride * (h->mb_height + 1);
246 int yc_size = y_size + 2 * c_size;
247 int x, y, i;
248
249 sl->ref_cache[0][scan8[5] + 1] =
250 sl->ref_cache[0][scan8[7] + 1] =
251 sl->ref_cache[0][scan8[13] + 1] =
252 sl->ref_cache[1][scan8[5] + 1] =
253 sl->ref_cache[1][scan8[7] + 1] =
254 sl->ref_cache[1][scan8[13] + 1] = PART_NOT_AVAILABLE;
255
256 if (sl != h->slice_ctx) {
257 memset(er, 0, sizeof(*er));
258 } else
259 if (CONFIG_ERROR_RESILIENCE) {
260
261 /* init ER */
262 er->avctx = h->avctx;
263 er->decode_mb = h264_er_decode_mb;
264 er->opaque = h;
265 er->quarter_sample = 1;
266
267 er->mb_num = h->mb_num;
268 er->mb_width = h->mb_width;
269 er->mb_height = h->mb_height;
270 er->mb_stride = h->mb_stride;
271 er->b8_stride = h->mb_width * 2 + 1;
272
273 // error resilience code looks cleaner with this
274 FF_ALLOCZ_OR_GOTO(h->avctx, er->mb_index2xy,
275 (h->mb_num + 1) * sizeof(int), fail);
276
277 for (y = 0; y < h->mb_height; y++)
278 for (x = 0; x < h->mb_width; x++)
279 er->mb_index2xy[x + y * h->mb_width] = x + y * h->mb_stride;
280
281 er->mb_index2xy[h->mb_height * h->mb_width] = (h->mb_height - 1) *
282 h->mb_stride + h->mb_width;
283
284 FF_ALLOCZ_OR_GOTO(h->avctx, er->error_status_table,
285 mb_array_size * sizeof(uint8_t), fail);
286
287 FF_ALLOC_OR_GOTO(h->avctx, er->er_temp_buffer,
288 h->mb_height * h->mb_stride * (4*sizeof(int) + 1), fail);
289
290 FF_ALLOCZ_OR_GOTO(h->avctx, sl->dc_val_base,
291 yc_size * sizeof(int16_t), fail);
292 er->dc_val[0] = sl->dc_val_base + h->mb_width * 2 + 2;
293 er->dc_val[1] = sl->dc_val_base + y_size + h->mb_stride + 1;
294 er->dc_val[2] = er->dc_val[1] + c_size;
295 for (i = 0; i < yc_size; i++)
296 sl->dc_val_base[i] = 1024;
297 }
298
299 return 0;
300
301fail:
302 return AVERROR(ENOMEM); // ff_h264_free_tables will clean up for us
303}
304
305static int h264_init_context(AVCodecContext *avctx, H264Context *h)
306{
307 int i;
308
309 h->avctx = avctx;
310 h->cur_chroma_format_idc = -1;
311
312 h->picture_structure = PICT_FRAME;
313 h->workaround_bugs = avctx->workaround_bugs;
314 h->flags = avctx->flags;
315 h->poc.prev_poc_msb = 1 << 16;
316 h->recovery_frame = -1;
317 h->frame_recovered = 0;
318 h->poc.prev_frame_num = -1;
319 h->sei.frame_packing.frame_packing_arrangement_cancel_flag = -1;
320 h->sei.unregistered.x264_build = -1;
321
322 h->next_outputed_poc = INT_MIN;
323 for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++)
324 h->last_pocs[i] = INT_MIN;
325
326 ff_h264_sei_uninit(&h->sei);
327
328 avctx->chroma_sample_location = AVCHROMA_LOC_LEFT;
329
330 h->nb_slice_ctx = (avctx->active_thread_type & FF_THREAD_SLICE) ? avctx->thread_count : 1;
331 h->slice_ctx = av_mallocz_array(h->nb_slice_ctx, sizeof(*h->slice_ctx));
332 if (!h->slice_ctx) {
333 h->nb_slice_ctx = 0;
334 return AVERROR(ENOMEM);
335 }
336
337 for (i = 0; i < H264_MAX_PICTURE_COUNT; i++) {
338 h->DPB[i].f = av_frame_alloc();
339 if (!h->DPB[i].f)
340 return AVERROR(ENOMEM);
341 }
342
343 h->cur_pic.f = av_frame_alloc();
344 if (!h->cur_pic.f)
345 return AVERROR(ENOMEM);
346
347 h->last_pic_for_ec.f = av_frame_alloc();
348 if (!h->last_pic_for_ec.f)
349 return AVERROR(ENOMEM);
350
351 for (i = 0; i < h->nb_slice_ctx; i++)
352 h->slice_ctx[i].h264 = h;
353
354 return 0;
355}
356
357static av_cold int h264_decode_end(AVCodecContext *avctx)
358{
359 H264Context *h = avctx->priv_data;
360 int i;
361
362 ff_h264_remove_all_refs(h);
363 ff_h264_free_tables(h);
364
365 for (i = 0; i < H264_MAX_PICTURE_COUNT; i++) {
366 ff_h264_unref_picture(h, &h->DPB[i]);
367 av_frame_free(&h->DPB[i].f);
368 }
369 memset(h->delayed_pic, 0, sizeof(h->delayed_pic));
370
371 h->cur_pic_ptr = NULL;
372
373 av_freep(&h->slice_ctx);
374 h->nb_slice_ctx = 0;
375
376 ff_h264_sei_uninit(&h->sei);
377 ff_h264_ps_uninit(&h->ps);
378
379 ff_h2645_packet_uninit(&h->pkt);
380
381 ff_h264_unref_picture(h, &h->cur_pic);
382 av_frame_free(&h->cur_pic.f);
383 ff_h264_unref_picture(h, &h->last_pic_for_ec);
384 av_frame_free(&h->last_pic_for_ec.f);
385
386 return 0;
387}
388
389static AVOnce h264_vlc_init = AV_ONCE_INIT;
390
391av_cold int ff_h264_decode_init(AVCodecContext *avctx)
392{
393 H264Context *h = avctx->priv_data;
394 int ret;
395
396 ret = h264_init_context(avctx, h);
397 if (ret < 0)
398 return ret;
399
400 ret = ff_thread_once(&h264_vlc_init, ff_h264_decode_init_vlc);
401 if (ret != 0) {
402 av_log(avctx, AV_LOG_ERROR, "pthread_once has failed.");
403 return AVERROR_UNKNOWN;
404 }
405
406 if (avctx->ticks_per_frame == 1) {
407 if(h->avctx->time_base.den < INT_MAX/2) {
408 h->avctx->time_base.den *= 2;
409 } else
410 h->avctx->time_base.num /= 2;
411 }
412 avctx->ticks_per_frame = 2;
413
414 if (avctx->extradata_size > 0 && avctx->extradata) {
415 ret = ff_h264_decode_extradata(avctx->extradata, avctx->extradata_size,
416 &h->ps, &h->is_avc, &h->nal_length_size,
417 avctx->err_recognition, avctx);
418 if (ret < 0) {
419 h264_decode_end(avctx);
420 return ret;
421 }
422 }
423
424 if (h->ps.sps && h->ps.sps->bitstream_restriction_flag &&
425 h->avctx->has_b_frames < h->ps.sps->num_reorder_frames) {
426 h->avctx->has_b_frames = h->ps.sps->num_reorder_frames;
427 }
428
429 avctx->internal->allocate_progress = 1;
430
431 ff_h264_flush_change(h);
432
433 if (h->enable_er < 0 && (avctx->active_thread_type & FF_THREAD_SLICE))
434 h->enable_er = 0;
435
436 if (h->enable_er && (avctx->active_thread_type & FF_THREAD_SLICE)) {
437 av_log(avctx, AV_LOG_WARNING,
438 "Error resilience with slice threads is enabled. It is unsafe and unsupported and may crash. "
439 "Use it at your own risk\n");
440 }
441
442 return 0;
443}
444
445#if HAVE_THREADS
446static int decode_init_thread_copy(AVCodecContext *avctx)
447{
448 H264Context *h = avctx->priv_data;
449 int ret;
450
451 if (!avctx->internal->is_copy)
452 return 0;
453
454 memset(h, 0, sizeof(*h));
455
456 ret = h264_init_context(avctx, h);
457 if (ret < 0)
458 return ret;
459
460 h->context_initialized = 0;
461
462 return 0;
463}
464#endif
465
466/**
467 * instantaneous decoder refresh.
468 */
469static void idr(H264Context *h)
470{
471 int i;
472 ff_h264_remove_all_refs(h);
473 h->poc.prev_frame_num =
474 h->poc.prev_frame_num_offset = 0;
475 h->poc.prev_poc_msb = 1<<16;
476 h->poc.prev_poc_lsb = 0;
477 for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++)
478 h->last_pocs[i] = INT_MIN;
479}
480
481/* forget old pics after a seek */
482void ff_h264_flush_change(H264Context *h)
483{
484 int i, j;
485
486 h->next_outputed_poc = INT_MIN;
487 h->prev_interlaced_frame = 1;
488 idr(h);
489
490 h->poc.prev_frame_num = -1;
491 if (h->cur_pic_ptr) {
492 h->cur_pic_ptr->reference = 0;
493 for (j=i=0; h->delayed_pic[i]; i++)
494 if (h->delayed_pic[i] != h->cur_pic_ptr)
495 h->delayed_pic[j++] = h->delayed_pic[i];
496 h->delayed_pic[j] = NULL;
497 }
498 ff_h264_unref_picture(h, &h->last_pic_for_ec);
499
500 h->first_field = 0;
501 ff_h264_sei_uninit(&h->sei);
502 h->recovery_frame = -1;
503 h->frame_recovered = 0;
504 h->current_slice = 0;
505 h->mmco_reset = 1;
506}
507
508/* forget old pics after a seek */
509static void flush_dpb(AVCodecContext *avctx)
510{
511 H264Context *h = avctx->priv_data;
512 int i;
513
514 memset(h->delayed_pic, 0, sizeof(h->delayed_pic));
515
516 ff_h264_flush_change(h);
517
518 for (i = 0; i < H264_MAX_PICTURE_COUNT; i++)
519 ff_h264_unref_picture(h, &h->DPB[i]);
520 h->cur_pic_ptr = NULL;
521 ff_h264_unref_picture(h, &h->cur_pic);
522
523 h->mb_y = 0;
524
525 ff_h264_free_tables(h);
526 h->context_initialized = 0;
527}
528
529#if FF_API_CAP_VDPAU
530static const uint8_t start_code[] = { 0x00, 0x00, 0x01 };
531#endif
532
533static int get_last_needed_nal(H264Context *h)
534{
535 int nals_needed = 0;
536 int first_slice = 0;
537 int i, ret;
538
539 for (i = 0; i < h->pkt.nb_nals; i++) {
540 H2645NAL *nal = &h->pkt.nals[i];
541 GetBitContext gb;
542
543 /* packets can sometimes contain multiple PPS/SPS,
544 * e.g. two PAFF field pictures in one packet, or a demuxer
545 * which splits NALs strangely if so, when frame threading we
546 * can't start the next thread until we've read all of them */
547 switch (nal->type) {
548 case H264_NAL_SPS:
549 case H264_NAL_PPS:
550 nals_needed = i;
551 break;
552 case H264_NAL_DPA:
553 case H264_NAL_IDR_SLICE:
554 case H264_NAL_SLICE:
555 ret = init_get_bits8(&gb, nal->data + 1, nal->size - 1);
556 if (ret < 0) {
557 av_log(h->avctx, AV_LOG_ERROR, "Invalid zero-sized VCL NAL unit\n");
558 if (h->avctx->err_recognition & AV_EF_EXPLODE)
559 return ret;
560
561 break;
562 }
563 if (!get_ue_golomb_long(&gb) || // first_mb_in_slice
564 !first_slice ||
565 first_slice != nal->type)
566 nals_needed = i;
567 if (!first_slice)
568 first_slice = nal->type;
569 }
570 }
571
572 return nals_needed;
573}
574
575static void debug_green_metadata(const H264SEIGreenMetaData *gm, void *logctx)
576{
577 av_log(logctx, AV_LOG_DEBUG, "Green Metadata Info SEI message\n");
578 av_log(logctx, AV_LOG_DEBUG, " green_metadata_type: %d\n", gm->green_metadata_type);
579
580 if (gm->green_metadata_type == 0) {
581 av_log(logctx, AV_LOG_DEBUG, " green_metadata_period_type: %d\n", gm->period_type);
582
583 if (gm->period_type == 2)
584 av_log(logctx, AV_LOG_DEBUG, " green_metadata_num_seconds: %d\n", gm->num_seconds);
585 else if (gm->period_type == 3)
586 av_log(logctx, AV_LOG_DEBUG, " green_metadata_num_pictures: %d\n", gm->num_pictures);
587
588 av_log(logctx, AV_LOG_DEBUG, " SEI GREEN Complexity Metrics: %f %f %f %f\n",
589 (float)gm->percent_non_zero_macroblocks/255,
590 (float)gm->percent_intra_coded_macroblocks/255,
591 (float)gm->percent_six_tap_filtering/255,
592 (float)gm->percent_alpha_point_deblocking_instance/255);
593
594 } else if (gm->green_metadata_type == 1) {
595 av_log(logctx, AV_LOG_DEBUG, " xsd_metric_type: %d\n", gm->xsd_metric_type);
596
597 if (gm->xsd_metric_type == 0)
598 av_log(logctx, AV_LOG_DEBUG, " xsd_metric_value: %f\n",
599 (float)gm->xsd_metric_value/100);
600 }
601}
602
603static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size)
604{
605 AVCodecContext *const avctx = h->avctx;
606 int nals_needed = 0; ///< number of NALs that need decoding before the next frame thread starts
607 int idr_cleared=0;
608 int i, ret = 0;
609
610 h->has_slice = 0;
611 h->nal_unit_type= 0;
612
613 if (!(avctx->flags2 & AV_CODEC_FLAG2_CHUNKS)) {
614 h->current_slice = 0;
615 if (!h->first_field)
616 h->cur_pic_ptr = NULL;
617 ff_h264_sei_uninit(&h->sei);
618 }
619
620 if (h->nal_length_size == 4) {
621 if (buf_size > 8 && AV_RB32(buf) == 1 && AV_RB32(buf+5) > (unsigned)buf_size) {
622 h->is_avc = 0;
623 }else if(buf_size > 3 && AV_RB32(buf) > 1 && AV_RB32(buf) <= (unsigned)buf_size)
624 h->is_avc = 1;
625 }
626
627 ret = ff_h2645_packet_split(&h->pkt, buf, buf_size, avctx, h->is_avc,
628 h->nal_length_size, avctx->codec_id, avctx->flags2 & AV_CODEC_FLAG2_FAST);
629 if (ret < 0) {
630 av_log(avctx, AV_LOG_ERROR,
631 "Error splitting the input into NAL units.\n");
632 return ret;
633 }
634
635 if (avctx->active_thread_type & FF_THREAD_FRAME)
636 nals_needed = get_last_needed_nal(h);
637 if (nals_needed < 0)
638 return nals_needed;
639
640 for (i = 0; i < h->pkt.nb_nals; i++) {
641 H2645NAL *nal = &h->pkt.nals[i];
642 int max_slice_ctx, err;
643
644 if (avctx->skip_frame >= AVDISCARD_NONREF &&
645 nal->ref_idc == 0 && nal->type != H264_NAL_SEI)
646 continue;
647
648 // FIXME these should stop being context-global variables
649 h->nal_ref_idc = nal->ref_idc;
650 h->nal_unit_type = nal->type;
651
652 err = 0;
653 switch (nal->type) {
654 case H264_NAL_IDR_SLICE:
655 if ((nal->data[1] & 0xFC) == 0x98) {
656 av_log(h->avctx, AV_LOG_ERROR, "Invalid inter IDR frame\n");
657 h->next_outputed_poc = INT_MIN;
658 ret = -1;
659 goto end;
660 }
661 if(!idr_cleared) {
662 if (h->current_slice && (avctx->active_thread_type & FF_THREAD_SLICE)) {
663 av_log(h, AV_LOG_ERROR, "invalid mixed IDR / non IDR frames cannot be decoded in slice multithreading mode\n");
664 ret = AVERROR_INVALIDDATA;
665 goto end;
666 }
667 idr(h); // FIXME ensure we don't lose some frames if there is reordering
668 }
669 idr_cleared = 1;
670 h->has_recovery_point = 1;
671 case H264_NAL_SLICE:
672 h->has_slice = 1;
673
674 if ((err = ff_h264_queue_decode_slice(h, nal))) {
675 H264SliceContext *sl = h->slice_ctx + h->nb_slice_ctx_queued;
676 sl->ref_count[0] = sl->ref_count[1] = 0;
677 break;
678 }
679
680 if (h->current_slice == 1) {
681 if (avctx->active_thread_type & FF_THREAD_FRAME &&
682 i >= nals_needed && !h->setup_finished && h->cur_pic_ptr) {
683 ff_thread_finish_setup(avctx);
684 h->setup_finished = 1;
685 }
686
687 if (h->avctx->hwaccel &&
688 (ret = h->avctx->hwaccel->start_frame(h->avctx, buf, buf_size)) < 0)
689 goto end;
690#if FF_API_CAP_VDPAU
691 if (CONFIG_H264_VDPAU_DECODER &&
692 h->avctx->codec->capabilities & AV_CODEC_CAP_HWACCEL_VDPAU)
693 ff_vdpau_h264_picture_start(h);
694#endif
695 }
696
697 max_slice_ctx = avctx->hwaccel ? 1 : h->nb_slice_ctx;
698 if (h->nb_slice_ctx_queued == max_slice_ctx) {
699 if (h->avctx->hwaccel) {
700 ret = avctx->hwaccel->decode_slice(avctx, nal->raw_data, nal->raw_size);
701 h->nb_slice_ctx_queued = 0;
702 } else
703#if FF_API_CAP_VDPAU
704 if (CONFIG_H264_VDPAU_DECODER &&
705 h->avctx->codec->capabilities & AV_CODEC_CAP_HWACCEL_VDPAU) {
706 ff_vdpau_add_data_chunk(h->cur_pic_ptr->f->data[0],
707 start_code,
708 sizeof(start_code));
709 ff_vdpau_add_data_chunk(h->cur_pic_ptr->f->data[0],
710 nal->raw_data,
711 nal->raw_size);
712 ret = 0;
713 } else
714#endif
715 ret = ff_h264_execute_decode_slices(h);
716 if (ret < 0 && (h->avctx->err_recognition & AV_EF_EXPLODE))
717 goto end;
718 }
719 break;
720 case H264_NAL_DPA:
721 case H264_NAL_DPB:
722 case H264_NAL_DPC:
723 avpriv_request_sample(avctx, "data partitioning");
724 break;
725 case H264_NAL_SEI:
726 ret = ff_h264_sei_decode(&h->sei, &nal->gb, &h->ps, avctx);
727 h->has_recovery_point = h->has_recovery_point || h->sei.recovery_point.recovery_frame_cnt != -1;
728 if (avctx->debug & FF_DEBUG_GREEN_MD)
729 debug_green_metadata(&h->sei.green_metadata, h->avctx);
730#if FF_API_AFD
731FF_DISABLE_DEPRECATION_WARNINGS
732 h->avctx->dtg_active_format = h->sei.afd.active_format_description;
733FF_ENABLE_DEPRECATION_WARNINGS
734#endif /* FF_API_AFD */
735 if (ret < 0 && (h->avctx->err_recognition & AV_EF_EXPLODE))
736 goto end;
737 break;
738 case H264_NAL_SPS: {
739 GetBitContext tmp_gb = nal->gb;
740 if (ff_h264_decode_seq_parameter_set(&tmp_gb, avctx, &h->ps, 0) >= 0)
741 break;
742 av_log(h->avctx, AV_LOG_DEBUG,
743 "SPS decoding failure, trying again with the complete NAL\n");
744 init_get_bits8(&tmp_gb, nal->raw_data + 1, nal->raw_size - 1);
745 if (ff_h264_decode_seq_parameter_set(&tmp_gb, avctx, &h->ps, 0) >= 0)
746 break;
747 ff_h264_decode_seq_parameter_set(&nal->gb, avctx, &h->ps, 1);
748 break;
749 }
750 case H264_NAL_PPS:
751 ret = ff_h264_decode_picture_parameter_set(&nal->gb, avctx, &h->ps,
752 nal->size_bits);
753 if (ret < 0 && (h->avctx->err_recognition & AV_EF_EXPLODE))
754 goto end;
755 break;
756 case H264_NAL_AUD:
757 case H264_NAL_END_SEQUENCE:
758 case H264_NAL_END_STREAM:
759 case H264_NAL_FILLER_DATA:
760 case H264_NAL_SPS_EXT:
761 case H264_NAL_AUXILIARY_SLICE:
762 break;
763 case H264_NAL_PREFIX:
764 case H264_NAL_SUB_SPS:
765 case H264_NAL_SLC_EXT:
766 /*0x1b is ts avc stream type set mvc once*/
767 if (avctx->codec_tag == 0x1b) {
768 avctx->codec_tag = MKTAG('M', 'V', 'C', ' ');
769 }
770 break;
771 case H264_NAL_264_DV_META:
772 /*00 00 00 01 7c 01 19 08
773 nal type = 0x7c & 0x1f;
774 any conflict?
775 */
776 avctx->has_dolby_vision_meta = 1;
777 break;
778 case H264_NAL_264_DV_EL:
779 /*00 00 00 01 7e
780 nal type = 0x7e & 0x1f;
781 any conflict?
782 */
783 avctx->has_dolby_vision_el = 1;
784 default:
785 av_log(avctx, AV_LOG_DEBUG, "Unknown NAL code: %d (%d bits)\n",
786 nal->type, nal->size_bits);
787 }
788
789 if (err < 0) {
790 av_log(h->avctx, AV_LOG_ERROR, "decode_slice_header error\n");
791 }
792 }
793
794 ret = ff_h264_execute_decode_slices(h);
795 if (ret < 0 && (h->avctx->err_recognition & AV_EF_EXPLODE))
796 goto end;
797
798 ret = 0;
799end:
800
801#if CONFIG_ERROR_RESILIENCE
802 /*
803 * FIXME: Error handling code does not seem to support interlaced
804 * when slices span multiple rows
805 * The ff_er_add_slice calls don't work right for bottom
806 * fields; they cause massive erroneous error concealing
807 * Error marking covers both fields (top and bottom).
808 * This causes a mismatched s->error_count
809 * and a bad error table. Further, the error count goes to
810 * INT_MAX when called for bottom field, because mb_y is
811 * past end by one (callers fault) and resync_mb_y != 0
812 * causes problems for the first MB line, too.
813 */
814 if (!FIELD_PICTURE(h) && h->current_slice &&
815 h->ps.sps == (const SPS*)h->ps.sps_list[h->ps.pps->sps_id]->data &&
816 h->enable_er) {
817
818 H264SliceContext *sl = h->slice_ctx;
819 int use_last_pic = h->last_pic_for_ec.f->buf[0] && !sl->ref_count[0];
820
821 ff_h264_set_erpic(&sl->er.cur_pic, h->cur_pic_ptr);
822
823 if (use_last_pic) {
824 ff_h264_set_erpic(&sl->er.last_pic, &h->last_pic_for_ec);
825 sl->ref_list[0][0].parent = &h->last_pic_for_ec;
826 memcpy(sl->ref_list[0][0].data, h->last_pic_for_ec.f->data, sizeof(sl->ref_list[0][0].data));
827 memcpy(sl->ref_list[0][0].linesize, h->last_pic_for_ec.f->linesize, sizeof(sl->ref_list[0][0].linesize));
828 sl->ref_list[0][0].reference = h->last_pic_for_ec.reference;
829 } else if (sl->ref_count[0]) {
830 ff_h264_set_erpic(&sl->er.last_pic, sl->ref_list[0][0].parent);
831 } else
832 ff_h264_set_erpic(&sl->er.last_pic, NULL);
833
834 if (sl->ref_count[1])
835 ff_h264_set_erpic(&sl->er.next_pic, sl->ref_list[1][0].parent);
836
837 sl->er.ref_count = sl->ref_count[0];
838
839 ff_er_frame_end(&sl->er);
840 if (use_last_pic)
841 memset(&sl->ref_list[0][0], 0, sizeof(sl->ref_list[0][0]));
842 }
843#endif /* CONFIG_ERROR_RESILIENCE */
844 /* clean up */
845 if (h->cur_pic_ptr && !h->droppable && h->has_slice) {
846 ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX,
847 h->picture_structure == PICT_BOTTOM_FIELD);
848 }
849
850 return (ret < 0) ? ret : buf_size;
851}
852
853/**
854 * Return the number of bytes consumed for building the current frame.
855 */
856static int get_consumed_bytes(int pos, int buf_size)
857{
858 if (pos == 0)
859 pos = 1; // avoid infinite loops (I doubt that is needed but...)
860 if (pos + 10 > buf_size)
861 pos = buf_size; // oops ;)
862
863 return pos;
864}
865
866static int output_frame(H264Context *h, AVFrame *dst, H264Picture *srcp)
867{
868 AVFrame *src = srcp->f;
869 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(src->format);
870 int i;
871 int ret;
872
873 if (src->format == AV_PIX_FMT_VIDEOTOOLBOX && src->buf[0]->size == 1)
874 return AVERROR_EXTERNAL;
875
876 ret = av_frame_ref(dst, src);
877 if (ret < 0)
878 return ret;
879
880 av_dict_set(&dst->metadata, "stereo_mode", ff_h264_sei_stereo_mode(&h->sei.frame_packing), 0);
881
882 if (srcp->sei_recovery_frame_cnt == 0)
883 dst->key_frame = 1;
884 if (!srcp->crop)
885 return 0;
886
887 for (i = 0; i < desc->nb_components; i++) {
888 int hshift = (i > 0) ? desc->log2_chroma_w : 0;
889 int vshift = (i > 0) ? desc->log2_chroma_h : 0;
890 int off = ((srcp->crop_left >> hshift) << h->pixel_shift) +
891 (srcp->crop_top >> vshift) * dst->linesize[i];
892 dst->data[i] += off;
893 }
894 return 0;
895}
896
897static int is_extra(const uint8_t *buf, int buf_size)
898{
899 int cnt= buf[5]&0x1f;
900 const uint8_t *p= buf+6;
901 while(cnt--){
902 int nalsize= AV_RB16(p) + 2;
903 if(nalsize > buf_size - (p-buf) || (p[2] & 0x9F) != 7)
904 return 0;
905 p += nalsize;
906 }
907 cnt = *(p++);
908 if(!cnt)
909 return 0;
910 while(cnt--){
911 int nalsize= AV_RB16(p) + 2;
912 if(nalsize > buf_size - (p-buf) || (p[2] & 0x9F) != 8)
913 return 0;
914 p += nalsize;
915 }
916 return 1;
917}
918
919static int finalize_frame(H264Context *h, AVFrame *dst, H264Picture *out, int *got_frame)
920{
921 int ret;
922
923 if (((h->avctx->flags & AV_CODEC_FLAG_OUTPUT_CORRUPT) ||
924 (h->avctx->flags2 & AV_CODEC_FLAG2_SHOW_ALL) ||
925 out->recovered)) {
926
927 if (!h->avctx->hwaccel &&
928 (out->field_poc[0] == INT_MAX ||
929 out->field_poc[1] == INT_MAX)
930 ) {
931 int p;
932 AVFrame *f = out->f;
933 int field = out->field_poc[0] == INT_MAX;
934 uint8_t *dst_data[4];
935 int linesizes[4];
936 const uint8_t *src_data[4];
937
938 av_log(h->avctx, AV_LOG_DEBUG, "Duplicating field %d to fill missing\n", field);
939
940 for (p = 0; p<4; p++) {
941 dst_data[p] = f->data[p] + (field^1)*f->linesize[p];
942 src_data[p] = f->data[p] + field *f->linesize[p];
943 linesizes[p] = 2*f->linesize[p];
944 }
945
946 av_image_copy(dst_data, linesizes, src_data, linesizes,
947 f->format, f->width, f->height>>1);
948 }
949
950 ret = output_frame(h, dst, out);
951 if (ret < 0)
952 return ret;
953
954 *got_frame = 1;
955
956 if (CONFIG_MPEGVIDEO) {
957 ff_print_debug_info2(h->avctx, dst, NULL,
958 out->mb_type,
959 out->qscale_table,
960 out->motion_val,
961 NULL,
962 h->mb_width, h->mb_height, h->mb_stride, 1);
963 }
964 }
965
966 return 0;
967}
968
969static int send_next_delayed_frame(H264Context *h, AVFrame *dst_frame,
970 int *got_frame, int buf_index)
971{
972 int ret, i, out_idx;
973 H264Picture *out = h->delayed_pic[0];
974
975 h->cur_pic_ptr = NULL;
976 h->first_field = 0;
977
978 out_idx = 0;
979 for (i = 1;
980 h->delayed_pic[i] &&
981 !h->delayed_pic[i]->f->key_frame &&
982 !h->delayed_pic[i]->mmco_reset;
983 i++)
984 if (h->delayed_pic[i]->poc < out->poc) {
985 out = h->delayed_pic[i];
986 out_idx = i;
987 }
988
989 for (i = out_idx; h->delayed_pic[i]; i++)
990 h->delayed_pic[i] = h->delayed_pic[i + 1];
991
992 if (out) {
993 out->reference &= ~DELAYED_PIC_REF;
994 ret = finalize_frame(h, dst_frame, out, got_frame);
995 if (ret < 0)
996 return ret;
997 }
998
999 return buf_index;
1000}
1001
1002static int h264_decode_frame(AVCodecContext *avctx, void *data,
1003 int *got_frame, AVPacket *avpkt)
1004{
1005 const uint8_t *buf = avpkt->data;
1006 int buf_size = avpkt->size;
1007 H264Context *h = avctx->priv_data;
1008 AVFrame *pict = data;
1009 int buf_index;
1010 int ret;
1011
1012 h->flags = avctx->flags;
1013 h->setup_finished = 0;
1014 h->nb_slice_ctx_queued = 0;
1015
1016 ff_h264_unref_picture(h, &h->last_pic_for_ec);
1017
1018 /* end of stream, output what is still in the buffers */
1019 if (buf_size == 0)
1020 return send_next_delayed_frame(h, pict, got_frame, 0);
1021
1022 if (h->is_avc && av_packet_get_side_data(avpkt, AV_PKT_DATA_NEW_EXTRADATA, NULL)) {
1023 int side_size;
1024 uint8_t *side = av_packet_get_side_data(avpkt, AV_PKT_DATA_NEW_EXTRADATA, &side_size);
1025 if (is_extra(side, side_size))
1026 ff_h264_decode_extradata(side, side_size,
1027 &h->ps, &h->is_avc, &h->nal_length_size,
1028 avctx->err_recognition, avctx);
1029 }
1030 if(h->is_avc && buf_size >= 9 && buf[0]==1 && buf[2]==0 && (buf[4]&0xFC)==0xFC && (buf[5]&0x1F) && buf[8]==0x67){
1031 if (is_extra(buf, buf_size))
1032 return ff_h264_decode_extradata(buf, buf_size,
1033 &h->ps, &h->is_avc, &h->nal_length_size,
1034 avctx->err_recognition, avctx);
1035 }
1036
1037 buf_index = decode_nal_units(h, buf, buf_size);
1038 if (buf_index < 0)
1039 return AVERROR_INVALIDDATA;
1040
1041 if (!h->cur_pic_ptr && h->nal_unit_type == H264_NAL_END_SEQUENCE) {
1042 av_assert0(buf_index <= buf_size);
1043 return send_next_delayed_frame(h, pict, got_frame, buf_index);
1044 }
1045
1046 if (!(avctx->flags2 & AV_CODEC_FLAG2_CHUNKS) && (!h->cur_pic_ptr || !h->has_slice)) {
1047 if (avctx->skip_frame >= AVDISCARD_NONREF ||
1048 buf_size >= 4 && !memcmp("Q264", buf, 4))
1049 return buf_size;
1050 av_log(avctx, AV_LOG_ERROR, "no frame!\n");
1051 return AVERROR_INVALIDDATA;
1052 }
1053
1054 if (!(avctx->flags2 & AV_CODEC_FLAG2_CHUNKS) ||
1055 (h->mb_y >= h->mb_height && h->mb_height)) {
1056 if ((ret = ff_h264_field_end(h, &h->slice_ctx[0], 0)) < 0)
1057 return ret;
1058
1059 /* Wait for second field. */
1060 if (h->next_output_pic) {
1061 ret = finalize_frame(h, pict, h->next_output_pic, got_frame);
1062 if (ret < 0)
1063 return ret;
1064 }
1065 }
1066
1067 av_assert0(pict->buf[0] || !*got_frame);
1068
1069 ff_h264_unref_picture(h, &h->last_pic_for_ec);
1070
1071 return get_consumed_bytes(buf_index, buf_size);
1072}
1073
1074#define OFFSET(x) offsetof(H264Context, x)
1075#define VD AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM
1076static const AVOption h264_options[] = {
1077 { "is_avc", "is avc", OFFSET(is_avc), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, 0 },
1078 { "nal_length_size", "nal_length_size", OFFSET(nal_length_size), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 4, 0 },
1079 { "enable_er", "Enable error resilience on damaged frames (unsafe)", OFFSET(enable_er), AV_OPT_TYPE_BOOL, { .i64 = -1 }, -1, 1, VD },
1080 { NULL },
1081};
1082
1083static const AVClass h264_class = {
1084 .class_name = "H264 Decoder",
1085 .item_name = av_default_item_name,
1086 .option = h264_options,
1087 .version = LIBAVUTIL_VERSION_INT,
1088};
1089
1090AVCodec ff_h264_decoder = {
1091 .name = "h264",
1092 .long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
1093 .type = AVMEDIA_TYPE_VIDEO,
1094 .id = AV_CODEC_ID_H264,
1095 .priv_data_size = sizeof(H264Context),
1096 .init = ff_h264_decode_init,
1097 .close = h264_decode_end,
1098 .decode = h264_decode_frame,
1099 .capabilities = /*AV_CODEC_CAP_DRAW_HORIZ_BAND |*/ AV_CODEC_CAP_DR1 |
1100 AV_CODEC_CAP_DELAY | AV_CODEC_CAP_SLICE_THREADS |
1101 AV_CODEC_CAP_FRAME_THREADS,
1102 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE,
1103 .flush = flush_dpb,
1104 .init_thread_copy = ONLY_IF_THREADS_ENABLED(decode_init_thread_copy),
1105 .update_thread_context = ONLY_IF_THREADS_ENABLED(ff_h264_update_thread_context),
1106 .profiles = NULL_IF_CONFIG_SMALL(ff_h264_profiles),
1107 .priv_class = &h264_class,
1108};
1109
1110#if CONFIG_H264_VDPAU_DECODER && FF_API_VDPAU
1111static const AVClass h264_vdpau_class = {
1112 .class_name = "H264 VDPAU Decoder",
1113 .item_name = av_default_item_name,
1114 .option = h264_options,
1115 .version = LIBAVUTIL_VERSION_INT,
1116};
1117
1118AVCodec ff_h264_vdpau_decoder = {
1119 .name = "h264_vdpau",
1120 .long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (VDPAU acceleration)"),
1121 .type = AVMEDIA_TYPE_VIDEO,
1122 .id = AV_CODEC_ID_H264,
1123 .priv_data_size = sizeof(H264Context),
1124 .init = ff_h264_decode_init,
1125 .close = h264_decode_end,
1126 .decode = h264_decode_frame,
1127 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY | AV_CODEC_CAP_HWACCEL_VDPAU,
1128 .flush = flush_dpb,
1129 .pix_fmts = (const enum AVPixelFormat[]) { AV_PIX_FMT_VDPAU_H264,
1130 AV_PIX_FMT_NONE},
1131 .profiles = NULL_IF_CONFIG_SMALL(ff_h264_profiles),
1132 .priv_class = &h264_vdpau_class,
1133};
1134#endif
1135
generated by cgit at 2024-05-10 19:21:17 (GMT)