1 /*
2 * NotchLC decoder
3 * Copyright (c) 2020 Paul B Mahol
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 #include <stdio.h>
23 #include <stdlib.h>
24 #include <string.h>
25
26 #define BITSTREAM_READER_LE
27 #include "libavutil/intreadwrite.h"
28 #include "avcodec.h"
29 #include "bytestream.h"
30 #include "get_bits.h"
31 #include "internal.h"
32 #include "lzf.h"
33 #include "thread.h"
34
35 typedef struct NotchLCContext {
36 unsigned compressed_size;
37 unsigned format;
38
39 uint8_t *uncompressed_buffer;
40 unsigned uncompressed_size;
41
42 uint8_t *lzf_buffer;
43 int64_t lzf_size;
44
45 unsigned texture_size_x;
46 unsigned texture_size_y;
47 unsigned y_data_row_offsets;
48 unsigned uv_offset_data_offset;
49 unsigned y_control_data_offset;
50 unsigned a_control_word_offset;
51 unsigned y_data_offset;
52 unsigned uv_data_offset;
53 unsigned y_data_size;
54 unsigned a_data_offset;
55 unsigned uv_count_offset;
56 unsigned a_count_size;
57 unsigned data_end;
58
59 GetByteContext gb;
60 PutByteContext pb;
61 } NotchLCContext;
62
decode_init(AVCodecContext * avctx)63 static av_cold int decode_init(AVCodecContext *avctx)
64 {
65 avctx->pix_fmt = AV_PIX_FMT_YUVA444P12;
66 avctx->color_range = AVCOL_RANGE_JPEG;
67 avctx->colorspace = AVCOL_SPC_RGB;
68 avctx->color_primaries = AVCOL_PRI_BT709;
69 avctx->color_trc = AVCOL_TRC_IEC61966_2_1;
70
71 return 0;
72 }
73
74 #define HISTORY_SIZE (64 * 1024)
75
lz4_decompress(AVCodecContext * avctx,GetByteContext * gb,PutByteContext * pb)76 static int lz4_decompress(AVCodecContext *avctx,
77 GetByteContext *gb,
78 PutByteContext *pb)
79 {
80 unsigned reference_pos, match_length, delta, pos = 0;
81 uint8_t history[64 * 1024];
82
83 while (bytestream2_get_bytes_left(gb) > 0) {
84 uint8_t token = bytestream2_get_byte(gb);
85 unsigned num_literals = token >> 4;
86
87 if (num_literals == 15) {
88 unsigned char current;
89 do {
90 current = bytestream2_get_byte(gb);
91 num_literals += current;
92 } while (current == 255);
93 }
94
95 if (pos + num_literals < HISTORY_SIZE) {
96 bytestream2_get_buffer(gb, history + pos, num_literals);
97 pos += num_literals;
98 } else {
99 while (num_literals-- > 0) {
100 history[pos++] = bytestream2_get_byte(gb);
101 if (pos == HISTORY_SIZE) {
102 bytestream2_put_buffer(pb, history, HISTORY_SIZE);
103 pos = 0;
104 }
105 }
106 }
107
108 if (bytestream2_get_bytes_left(gb) <= 0)
109 break;
110
111 delta = bytestream2_get_le16(gb);
112 if (delta == 0)
113 return 0;
114 match_length = 4 + (token & 0x0F);
115 if (match_length == 4 + 0x0F) {
116 uint8_t current;
117
118 do {
119 current = bytestream2_get_byte(gb);
120 match_length += current;
121 } while (current == 255);
122 }
123 reference_pos = (pos >= delta) ? (pos - delta) : (HISTORY_SIZE + pos - delta);
124 if (pos + match_length < HISTORY_SIZE && reference_pos + match_length < HISTORY_SIZE) {
125 if (pos >= reference_pos + match_length || reference_pos >= pos + match_length) {
126 memcpy(history + pos, history + reference_pos, match_length);
127 pos += match_length;
128 } else {
129 while (match_length-- > 0)
130 history[pos++] = history[reference_pos++];
131 }
132 } else {
133 while (match_length-- > 0) {
134 history[pos++] = history[reference_pos++];
135 if (pos == HISTORY_SIZE) {
136 bytestream2_put_buffer(pb, history, HISTORY_SIZE);
137 pos = 0;
138 }
139 reference_pos %= HISTORY_SIZE;
140 }
141 }
142 }
143
144 bytestream2_put_buffer(pb, history, pos);
145
146 return bytestream2_tell_p(pb);
147 }
148
decode_blocks(AVCodecContext * avctx,AVFrame * p,ThreadFrame * frame,unsigned uncompressed_size)149 static int decode_blocks(AVCodecContext *avctx, AVFrame *p, ThreadFrame *frame,
150 unsigned uncompressed_size)
151 {
152 NotchLCContext *s = avctx->priv_data;
153 GetByteContext rgb, dgb, *gb = &s->gb;
154 GetBitContext bit;
155 int ylinesize, ulinesize, vlinesize, alinesize;
156 uint16_t *dsty, *dstu, *dstv, *dsta;
157 int ret;
158
159 s->texture_size_x = bytestream2_get_le32(gb);
160 s->texture_size_y = bytestream2_get_le32(gb);
161
162 ret = ff_set_dimensions(avctx, s->texture_size_x, s->texture_size_y);
163 if (ret < 0)
164 return ret;
165
166 s->uv_offset_data_offset = bytestream2_get_le32(gb);
167 if (s->uv_offset_data_offset >= UINT_MAX / 4)
168 return AVERROR_INVALIDDATA;
169 s->uv_offset_data_offset *= 4;
170 if (s->uv_offset_data_offset >= uncompressed_size)
171 return AVERROR_INVALIDDATA;
172
173 s->y_control_data_offset = bytestream2_get_le32(gb);
174 if (s->y_control_data_offset >= UINT_MAX / 4)
175 return AVERROR_INVALIDDATA;
176 s->y_control_data_offset *= 4;
177 if (s->y_control_data_offset >= uncompressed_size)
178 return AVERROR_INVALIDDATA;
179
180 s->a_control_word_offset = bytestream2_get_le32(gb);
181 if (s->a_control_word_offset >= UINT_MAX / 4)
182 return AVERROR_INVALIDDATA;
183 s->a_control_word_offset *= 4;
184 if (s->a_control_word_offset >= uncompressed_size)
185 return AVERROR_INVALIDDATA;
186
187 s->uv_data_offset = bytestream2_get_le32(gb);
188 if (s->uv_data_offset >= UINT_MAX / 4)
189 return AVERROR_INVALIDDATA;
190 s->uv_data_offset *= 4;
191 if (s->uv_data_offset >= uncompressed_size)
192 return AVERROR_INVALIDDATA;
193
194 s->y_data_size = bytestream2_get_le32(gb);
195 if (s->y_data_size >= UINT_MAX / 4)
196 return AVERROR_INVALIDDATA;
197
198 s->a_data_offset = bytestream2_get_le32(gb);
199 if (s->a_data_offset >= UINT_MAX / 4)
200 return AVERROR_INVALIDDATA;
201 s->a_data_offset *= 4;
202 if (s->a_data_offset >= uncompressed_size)
203 return AVERROR_INVALIDDATA;
204
205 s->a_count_size = bytestream2_get_le32(gb);
206 if (s->a_count_size >= UINT_MAX / 4)
207 return AVERROR_INVALIDDATA;
208 s->a_count_size *= 4;
209 if (s->a_count_size >= uncompressed_size)
210 return AVERROR_INVALIDDATA;
211
212 s->data_end = bytestream2_get_le32(gb);
213 if (s->data_end > uncompressed_size)
214 return AVERROR_INVALIDDATA;
215
216 s->y_data_row_offsets = bytestream2_tell(gb);
217 if (s->data_end <= s->y_data_size)
218 return AVERROR_INVALIDDATA;
219 s->y_data_offset = s->data_end - s->y_data_size;
220 if (s->y_data_offset <= s->a_data_offset)
221 return AVERROR_INVALIDDATA;
222 s->uv_count_offset = s->y_data_offset - s->a_data_offset;
223
224 if ((ret = ff_thread_get_buffer(avctx, frame, 0)) < 0)
225 return ret;
226
227 rgb = *gb;
228 dgb = *gb;
229 bytestream2_seek(&rgb, s->y_data_row_offsets, SEEK_SET);
230 bytestream2_seek(gb, s->y_control_data_offset, SEEK_SET);
231
232 if (bytestream2_get_bytes_left(gb) < (avctx->height + 3) / 4 * ((avctx->width + 3) / 4) * 4)
233 return AVERROR_INVALIDDATA;
234
235 dsty = (uint16_t *)p->data[0];
236 dsta = (uint16_t *)p->data[3];
237 ylinesize = p->linesize[0] / 2;
238 alinesize = p->linesize[3] / 2;
239
240 for (int y = 0; y < avctx->height; y += 4) {
241 const unsigned row_offset = bytestream2_get_le32(&rgb);
242
243 bytestream2_seek(&dgb, s->y_data_offset + row_offset, SEEK_SET);
244
245 init_get_bits8(&bit, dgb.buffer, bytestream2_get_bytes_left(&dgb));
246 for (int x = 0; x < avctx->width; x += 4) {
247 unsigned item = bytestream2_get_le32(gb);
248 unsigned y_min = item & 4095;
249 unsigned y_max = (item >> 12) & 4095;
250 unsigned y_diff = y_max - y_min;
251 unsigned control[4];
252
253 control[0] = (item >> 24) & 3;
254 control[1] = (item >> 26) & 3;
255 control[2] = (item >> 28) & 3;
256 control[3] = (item >> 30) & 3;
257
258 for (int i = 0; i < 4; i++) {
259 const int nb_bits = control[i] + 1;
260 const int div = (1 << nb_bits) - 1;
261 const int add = div - 1;
262
263 dsty[x + i * ylinesize + 0] = av_clip_uintp2(y_min + ((y_diff * get_bits(&bit, nb_bits) + add) / div), 12);
264 dsty[x + i * ylinesize + 1] = av_clip_uintp2(y_min + ((y_diff * get_bits(&bit, nb_bits) + add) / div), 12);
265 dsty[x + i * ylinesize + 2] = av_clip_uintp2(y_min + ((y_diff * get_bits(&bit, nb_bits) + add) / div), 12);
266 dsty[x + i * ylinesize + 3] = av_clip_uintp2(y_min + ((y_diff * get_bits(&bit, nb_bits) + add) / div), 12);
267 }
268 }
269
270 dsty += 4 * ylinesize;
271 }
272
273 rgb = *gb;
274 dgb = *gb;
275 bytestream2_seek(gb, s->a_control_word_offset, SEEK_SET);
276 if (s->uv_count_offset == s->a_control_word_offset) {
277 for (int y = 0; y < avctx->height; y++) {
278 for (int x = 0; x < avctx->width; x++)
279 dsta[x] = 4095;
280 dsta += alinesize;
281 }
282 } else {
283 if (bytestream2_get_bytes_left(gb) < (avctx->height + 15) / 16 * ((avctx->width + 15) / 16) * 8)
284 return AVERROR_INVALIDDATA;
285
286 for (int y = 0; y < avctx->height; y += 16) {
287 for (int x = 0; x < avctx->width; x += 16) {
288 unsigned m = bytestream2_get_le32(gb);
289 unsigned offset = bytestream2_get_le32(gb);
290 unsigned alpha0, alpha1;
291 uint64_t control;
292
293 if (offset >= UINT_MAX / 4)
294 return AVERROR_INVALIDDATA;
295 offset = offset * 4 + s->uv_data_offset + s->a_data_offset;
296 if (offset >= s->data_end)
297 return AVERROR_INVALIDDATA;
298
299 bytestream2_seek(&dgb, offset, SEEK_SET);
300 control = bytestream2_get_le64(&dgb);
301 alpha0 = control & 0xFF;
302 alpha1 = (control >> 8) & 0xFF;
303 control = control >> 16;
304
305 for (int by = 0; by < 4; by++) {
306 for (int bx = 0; bx < 4; bx++) {
307 switch (m & 3) {
308 case 0:
309 for (int i = 0; i < 4; i++) {
310 for (int j = 0; j < 4; j++) {
311 dsta[x + (i + by * 4) * alinesize + bx * 4 + j] = 0;
312 }
313 }
314 break;
315 case 1:
316 for (int i = 0; i < 4; i++) {
317 for (int j = 0; j < 4; j++) {
318 dsta[x + (i + by * 4) * alinesize + bx * 4 + j] = 4095;
319 }
320 }
321 break;
322 case 2:
323 for (int i = 0; i < 4; i++) {
324 for (int j = 0; j < 4; j++) {
325 dsta[x + (i + by * 4) * alinesize + bx * 4 + j] = (alpha0 + (alpha1 - alpha0) * (control & 7)) << 4;
326 }
327 }
328 break;
329 default:
330 return AVERROR_INVALIDDATA;
331 }
332
333 control >>= 3;
334 m >>= 2;
335 }
336 }
337 }
338
339 dsta += 16 * alinesize;
340 }
341 }
342
343 bytestream2_seek(&rgb, s->uv_offset_data_offset, SEEK_SET);
344
345 dstu = (uint16_t *)p->data[1];
346 dstv = (uint16_t *)p->data[2];
347 ulinesize = p->linesize[1] / 2;
348 vlinesize = p->linesize[2] / 2;
349
350 for (int y = 0; y < avctx->height; y += 16) {
351 for (int x = 0; x < avctx->width; x += 16) {
352 unsigned offset = bytestream2_get_le32(&rgb) * 4;
353 int u[16][16] = { 0 }, v[16][16] = { 0 };
354 int u0, v0, u1, v1, udif, vdif;
355 unsigned escape, is8x8, loc;
356
357 bytestream2_seek(&dgb, s->uv_data_offset + offset, SEEK_SET);
358
359 is8x8 = bytestream2_get_le16(&dgb);
360 escape = bytestream2_get_le16(&dgb);
361
362 if (escape == 0 && is8x8 == 0) {
363 u0 = bytestream2_get_byte(&dgb);
364 v0 = bytestream2_get_byte(&dgb);
365 u1 = bytestream2_get_byte(&dgb);
366 v1 = bytestream2_get_byte(&dgb);
367 loc = bytestream2_get_le32(&dgb);
368 u0 = (u0 << 4) | (u0 & 0xF);
369 v0 = (v0 << 4) | (v0 & 0xF);
370 u1 = (u1 << 4) | (u1 & 0xF);
371 v1 = (v1 << 4) | (v1 & 0xF);
372 udif = u1 - u0;
373 vdif = v1 - v0;
374
375 for (int i = 0; i < 16; i += 4) {
376 for (int j = 0; j < 16; j += 4) {
377 for (int ii = 0; ii < 4; ii++) {
378 for (int jj = 0; jj < 4; jj++) {
379 u[i + ii][j + jj] = u0 + ((udif * (int)(loc & 3) + 2) / 3);
380 v[i + ii][j + jj] = v0 + ((vdif * (int)(loc & 3) + 2) / 3);
381 }
382 }
383
384 loc >>= 2;
385 }
386 }
387 } else {
388 for (int i = 0; i < 16; i += 8) {
389 for (int j = 0; j < 16; j += 8) {
390 if (is8x8 & 1) {
391 u0 = bytestream2_get_byte(&dgb);
392 v0 = bytestream2_get_byte(&dgb);
393 u1 = bytestream2_get_byte(&dgb);
394 v1 = bytestream2_get_byte(&dgb);
395 loc = bytestream2_get_le32(&dgb);
396 u0 = (u0 << 4) | (u0 & 0xF);
397 v0 = (v0 << 4) | (v0 & 0xF);
398 u1 = (u1 << 4) | (u1 & 0xF);
399 v1 = (v1 << 4) | (v1 & 0xF);
400 udif = u1 - u0;
401 vdif = v1 - v0;
402
403 for (int ii = 0; ii < 8; ii += 2) {
404 for (int jj = 0; jj < 8; jj += 2) {
405 for (int iii = 0; iii < 2; iii++) {
406 for (int jjj = 0; jjj < 2; jjj++) {
407 u[i + ii + iii][j + jj + jjj] = u0 + ((udif * (int)(loc & 3) + 2) / 3);
408 v[i + ii + iii][j + jj + jjj] = v0 + ((vdif * (int)(loc & 3) + 2) / 3);
409 }
410 }
411
412 loc >>= 2;
413 }
414 }
415 } else if (escape) {
416 for (int ii = 0; ii < 8; ii += 4) {
417 for (int jj = 0; jj < 8; jj += 4) {
418 u0 = bytestream2_get_byte(&dgb);
419 v0 = bytestream2_get_byte(&dgb);
420 u1 = bytestream2_get_byte(&dgb);
421 v1 = bytestream2_get_byte(&dgb);
422 loc = bytestream2_get_le32(&dgb);
423 u0 = (u0 << 4) | (u0 & 0xF);
424 v0 = (v0 << 4) | (v0 & 0xF);
425 u1 = (u1 << 4) | (u1 & 0xF);
426 v1 = (v1 << 4) | (v1 & 0xF);
427 udif = u1 - u0;
428 vdif = v1 - v0;
429
430 for (int iii = 0; iii < 4; iii++) {
431 for (int jjj = 0; jjj < 4; jjj++) {
432 u[i + ii + iii][j + jj + jjj] = u0 + ((udif * (int)(loc & 3) + 2) / 3);
433 v[i + ii + iii][j + jj + jjj] = v0 + ((vdif * (int)(loc & 3) + 2) / 3);
434
435 loc >>= 2;
436 }
437 }
438 }
439 }
440 }
441
442 is8x8 >>= 1;
443 }
444 }
445 }
446
447 for (int i = 0; i < 16; i++) {
448 for (int j = 0; j < 16; j++) {
449 dstu[x + i * ulinesize + j] = u[i][j];
450 dstv[x + i * vlinesize + j] = v[i][j];
451 }
452 }
453 }
454
455 dstu += 16 * ulinesize;
456 dstv += 16 * vlinesize;
457 }
458
459 return 0;
460 }
461
decode_frame(AVCodecContext * avctx,void * data,int * got_frame,AVPacket * avpkt)462 static int decode_frame(AVCodecContext *avctx,
463 void *data, int *got_frame,
464 AVPacket *avpkt)
465 {
466 NotchLCContext *s = avctx->priv_data;
467 ThreadFrame frame = { .f = data };
468 GetByteContext *gb = &s->gb;
469 PutByteContext *pb = &s->pb;
470 unsigned uncompressed_size;
471 AVFrame *p = data;
472 int ret;
473
474 if (avpkt->size <= 40)
475 return AVERROR_INVALIDDATA;
476
477 bytestream2_init(gb, avpkt->data, avpkt->size);
478
479 if (bytestream2_get_le32(gb) != MKBETAG('N','L','C','1'))
480 return AVERROR_INVALIDDATA;
481
482 uncompressed_size = bytestream2_get_le32(gb);
483 s->compressed_size = bytestream2_get_le32(gb);
484 s->format = bytestream2_get_le32(gb);
485
486 if (s->format > 2)
487 return AVERROR_PATCHWELCOME;
488
489 if (s->format == 0) {
490 ret = ff_lzf_uncompress(gb, &s->lzf_buffer, &s->lzf_size);
491 if (ret < 0)
492 return ret;
493
494 if (uncompressed_size > s->lzf_size)
495 return AVERROR_INVALIDDATA;
496
497 bytestream2_init(gb, s->lzf_buffer, uncompressed_size);
498 } else if (s->format == 1) {
499 if (bytestream2_get_bytes_left(gb) < uncompressed_size / 255)
500 return AVERROR_INVALIDDATA;
501
502 av_fast_padded_malloc(&s->uncompressed_buffer, &s->uncompressed_size,
503 uncompressed_size);
504 if (!s->uncompressed_buffer)
505 return AVERROR(ENOMEM);
506
507 bytestream2_init_writer(pb, s->uncompressed_buffer, s->uncompressed_size);
508
509 ret = lz4_decompress(avctx, gb, pb);
510 if (ret != uncompressed_size)
511 return AVERROR_INVALIDDATA;
512
513 bytestream2_init(gb, s->uncompressed_buffer, uncompressed_size);
514 }
515
516 ret = decode_blocks(avctx, p, &frame, uncompressed_size);
517 if (ret < 0)
518 return ret;
519
520 p->pict_type = AV_PICTURE_TYPE_I;
521 p->key_frame = 1;
522
523 *got_frame = 1;
524
525 return avpkt->size;
526 }
527
decode_end(AVCodecContext * avctx)528 static av_cold int decode_end(AVCodecContext *avctx)
529 {
530 NotchLCContext *s = avctx->priv_data;
531
532 av_freep(&s->uncompressed_buffer);
533 s->uncompressed_size = 0;
534 av_freep(&s->lzf_buffer);
535 s->lzf_size = 0;
536
537 return 0;
538 }
539
540 AVCodec ff_notchlc_decoder = {
541 .name = "notchlc",
542 .long_name = NULL_IF_CONFIG_SMALL("NotchLC"),
543 .type = AVMEDIA_TYPE_VIDEO,
544 .id = AV_CODEC_ID_NOTCHLC,
545 .priv_data_size = sizeof(NotchLCContext),
546 .init = decode_init,
547 .close = decode_end,
548 .decode = decode_frame,
549 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS,
550 };
551