rtmp-av1.c 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601
  1. /******************************************************************************
  2. Copyright (C) 2023 by Lain Bailey <[email protected]>
  3. This program is free software: you can redistribute it and/or modify
  4. it under the terms of the GNU General Public License as published by
  5. the Free Software Foundation, either version 2 of the License, or
  6. (at your option) any later version.
  7. This program is distributed in the hope that it will be useful,
  8. but WITHOUT ANY WARRANTY; without even the implied warranty of
  9. MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  10. GNU General Public License for more details.
  11. You should have received a copy of the GNU General Public License
  12. along with this program. If not, see <http://www.gnu.org/licenses/>.
  13. ******************************************************************************/
  14. #include "rtmp-av1.h"
  15. #include "utils.h"
  16. #include <obs.h>
  17. #include <util/array-serializer.h>
  18. /* Adapted from FFmpeg's libavformat/av1.c for our FLV muxer. */
  19. #define AV1_OBU_SEQUENCE_HEADER 1
  20. #define AV1_OBU_TEMPORAL_DELIMITER 2
  21. #define AV1_OBU_REDUNDANT_FRAME_HEADER 7
  22. #define AV1_OBU_TILE_LIST 8
  23. #define AV1_OBU_PADDING 15
  24. #define AV1_OBU_METADATA 5
  25. #define AV1_OBU_TILE_GROUP 4
  26. #define AV1_OBU_TILE_LIST 8
  27. #define AV1_OBU_FRAME 6
  28. #define FF_PROFILE_AV1_MAIN 0
  29. #define FF_PROFILE_AV1_HIGH 1
  30. #define FF_PROFILE_AV1_PROFESSIONAL 2
  31. typedef struct AV1SequenceParameters {
  32. uint8_t profile;
  33. uint8_t level;
  34. uint8_t tier;
  35. uint8_t bitdepth;
  36. uint8_t monochrome;
  37. uint8_t chroma_subsampling_x;
  38. uint8_t chroma_subsampling_y;
  39. uint8_t chroma_sample_position;
  40. uint8_t color_description_present_flag;
  41. uint8_t color_primaries;
  42. uint8_t transfer_characteristics;
  43. uint8_t matrix_coefficients;
  44. uint8_t color_range;
  45. } AV1SequenceParameters;
  46. #define MAX_OBU_HEADER_SIZE (2 + 8)
  47. typedef struct Av1GetBitContext {
  48. const uint8_t *buffer, *buffer_end;
  49. int index;
  50. int size_in_bits;
  51. int size_in_bits_plus8;
  52. } Av1GetBitContext;
  53. static inline int init_get_bits_xe(Av1GetBitContext *s, const uint8_t *buffer,
  54. int bit_size)
  55. {
  56. int buffer_size;
  57. int ret = 0;
  58. if (bit_size >= INT_MAX - 64 * 8 || bit_size < 0 || !buffer) {
  59. bit_size = 0;
  60. buffer = NULL;
  61. ret = -1;
  62. }
  63. buffer_size = (bit_size + 7) >> 3;
  64. s->buffer = buffer;
  65. s->size_in_bits = bit_size;
  66. s->size_in_bits_plus8 = bit_size + 8;
  67. s->buffer_end = buffer + buffer_size;
  68. s->index = 0;
  69. return ret;
  70. }
  71. static inline int init_get_bits(Av1GetBitContext *s, const uint8_t *buffer,
  72. int bit_size)
  73. {
  74. return init_get_bits_xe(s, buffer, bit_size);
  75. }
  76. static inline int init_get_bits8(Av1GetBitContext *s, const uint8_t *buffer,
  77. int byte_size)
  78. {
  79. if (byte_size > INT_MAX / 8 || byte_size < 0)
  80. byte_size = -1;
  81. return init_get_bits(s, buffer, byte_size * 8);
  82. }
  83. static inline unsigned int get_bit1(Av1GetBitContext *s)
  84. {
  85. unsigned int index = s->index;
  86. uint8_t result = s->buffer[index >> 3];
  87. result <<= index & 7;
  88. result >>= 8 - 1;
  89. if (s->index < s->size_in_bits_plus8)
  90. index++;
  91. s->index = index;
  92. return result;
  93. }
  94. static inline unsigned int get_bits(Av1GetBitContext *s, unsigned int n)
  95. {
  96. unsigned int out = 0;
  97. for (unsigned int i = 0; i < n; i++)
  98. out = (out << 1) | get_bit1(s);
  99. return out;
  100. }
  101. #define skip_bits get_bits
  102. static inline int get_bits_count(Av1GetBitContext *s)
  103. {
  104. return s->index;
  105. }
  106. static inline int get_bits_left(Av1GetBitContext *gb)
  107. {
  108. return gb->size_in_bits - get_bits_count(gb);
  109. }
  110. #define get_bits_long get_bits
  111. #define skip_bits_long get_bits_long
  112. static inline int64_t leb128(Av1GetBitContext *gb)
  113. {
  114. int64_t ret = 0;
  115. int i;
  116. for (i = 0; i < 8; i++) {
  117. int byte = get_bits(gb, 8);
  118. ret |= (int64_t)(byte & 0x7f) << (i * 7);
  119. if (!(byte & 0x80))
  120. break;
  121. }
  122. return ret;
  123. }
  124. static inline void uvlc(Av1GetBitContext *gb)
  125. {
  126. int leading_zeros = 0;
  127. while (get_bits_left(gb)) {
  128. if (get_bits(gb, 1))
  129. break;
  130. leading_zeros++;
  131. }
  132. if (leading_zeros >= 32)
  133. return;
  134. skip_bits_long(gb, leading_zeros);
  135. }
  136. static inline int parse_obu_header(const uint8_t *buf, int buf_size,
  137. int64_t *obu_size, int *start_pos, int *type,
  138. int *temporal_id, int *spatial_id)
  139. {
  140. Av1GetBitContext gb;
  141. int ret, extension_flag, has_size_flag;
  142. size_t size;
  143. ret = init_get_bits8(&gb, buf, min_i32(buf_size, MAX_OBU_HEADER_SIZE));
  144. if (ret < 0)
  145. return ret;
  146. if (get_bits(&gb, 1) != 0) // obu_forbidden_bit
  147. return -1;
  148. *type = get_bits(&gb, 4);
  149. extension_flag = get_bits(&gb, 1);
  150. has_size_flag = get_bits(&gb, 1);
  151. skip_bits(&gb, 1); // obu_reserved_1bit
  152. if (extension_flag) {
  153. *temporal_id = get_bits(&gb, 3);
  154. *spatial_id = get_bits(&gb, 2);
  155. skip_bits(&gb, 3); // extension_header_reserved_3bits
  156. } else {
  157. *temporal_id = *spatial_id = 0;
  158. }
  159. *obu_size = has_size_flag ? leb128(&gb) : buf_size - 1 - extension_flag;
  160. if (get_bits_left(&gb) < 0)
  161. return -1;
  162. *start_pos = get_bits_count(&gb) / 8;
  163. size = (size_t)(*obu_size + *start_pos);
  164. if (size > (size_t)buf_size)
  165. return -1;
  166. assert(size <= INT_MAX);
  167. return (int)size;
  168. }
  169. static inline int get_obu_bit_length(const uint8_t *buf, int size, int type)
  170. {
  171. int v;
  172. /* There are no trailing bits on these */
  173. if (type == AV1_OBU_TILE_GROUP || type == AV1_OBU_TILE_LIST ||
  174. type == AV1_OBU_FRAME) {
  175. if (size > INT_MAX / 8)
  176. return -1;
  177. else
  178. return size * 8;
  179. }
  180. while (size > 0 && buf[size - 1] == 0)
  181. size--;
  182. if (!size)
  183. return 0;
  184. v = buf[size - 1];
  185. if (size > INT_MAX / 8)
  186. return -1;
  187. size *= 8;
  188. /* Remove the trailing_one_bit and following trailing zeros */
  189. if (v)
  190. size -= ctz32(v) + 1;
  191. return size;
  192. }
  193. static int parse_color_config(AV1SequenceParameters *seq_params,
  194. Av1GetBitContext *gb)
  195. {
  196. int twelve_bit = 0;
  197. int high_bitdepth = get_bits(gb, 1);
  198. if (seq_params->profile == FF_PROFILE_AV1_PROFESSIONAL && high_bitdepth)
  199. twelve_bit = get_bits(gb, 1);
  200. seq_params->bitdepth = 8 + (high_bitdepth * 2) + (twelve_bit * 2);
  201. if (seq_params->profile == FF_PROFILE_AV1_HIGH)
  202. seq_params->monochrome = 0;
  203. else
  204. seq_params->monochrome = get_bits(gb, 1);
  205. seq_params->color_description_present_flag = get_bits(gb, 1);
  206. if (seq_params->color_description_present_flag) {
  207. seq_params->color_primaries = get_bits(gb, 8);
  208. seq_params->transfer_characteristics = get_bits(gb, 8);
  209. seq_params->matrix_coefficients = get_bits(gb, 8);
  210. } else {
  211. seq_params->color_primaries = 2;
  212. seq_params->transfer_characteristics = 2;
  213. seq_params->matrix_coefficients = 2;
  214. }
  215. if (seq_params->monochrome) {
  216. seq_params->color_range = get_bits(gb, 1);
  217. seq_params->chroma_subsampling_x = 1;
  218. seq_params->chroma_subsampling_y = 1;
  219. seq_params->chroma_sample_position = 0;
  220. return 0;
  221. } else if (seq_params->color_primaries == 1 &&
  222. seq_params->transfer_characteristics == 13 &&
  223. seq_params->matrix_coefficients == 0) {
  224. seq_params->chroma_subsampling_x = 0;
  225. seq_params->chroma_subsampling_y = 0;
  226. } else {
  227. seq_params->color_range = get_bits(gb, 1);
  228. if (seq_params->profile == FF_PROFILE_AV1_MAIN) {
  229. seq_params->chroma_subsampling_x = 1;
  230. seq_params->chroma_subsampling_y = 1;
  231. } else if (seq_params->profile == FF_PROFILE_AV1_HIGH) {
  232. seq_params->chroma_subsampling_x = 0;
  233. seq_params->chroma_subsampling_y = 0;
  234. } else {
  235. if (twelve_bit) {
  236. seq_params->chroma_subsampling_x =
  237. get_bits(gb, 1);
  238. if (seq_params->chroma_subsampling_x)
  239. seq_params->chroma_subsampling_y =
  240. get_bits(gb, 1);
  241. else
  242. seq_params->chroma_subsampling_y = 0;
  243. } else {
  244. seq_params->chroma_subsampling_x = 1;
  245. seq_params->chroma_subsampling_y = 0;
  246. }
  247. }
  248. if (seq_params->chroma_subsampling_x &&
  249. seq_params->chroma_subsampling_y)
  250. seq_params->chroma_sample_position = get_bits(gb, 2);
  251. }
  252. skip_bits(gb, 1); // separate_uv_delta_q
  253. return 0;
  254. }
  255. static int parse_sequence_header(AV1SequenceParameters *seq_params,
  256. const uint8_t *buf, int size)
  257. {
  258. Av1GetBitContext gb;
  259. int reduced_still_picture_header;
  260. int frame_width_bits_minus_1, frame_height_bits_minus_1;
  261. int size_bits, ret;
  262. size_bits = get_obu_bit_length(buf, size, AV1_OBU_SEQUENCE_HEADER);
  263. if (size_bits < 0)
  264. return size_bits;
  265. ret = init_get_bits(&gb, buf, size_bits);
  266. if (ret < 0)
  267. return ret;
  268. memset(seq_params, 0, sizeof(*seq_params));
  269. seq_params->profile = get_bits(&gb, 3);
  270. skip_bits(&gb, 1); // still_picture
  271. reduced_still_picture_header = get_bits(&gb, 1);
  272. if (reduced_still_picture_header) {
  273. seq_params->level = get_bits(&gb, 5);
  274. seq_params->tier = 0;
  275. } else {
  276. int initial_display_delay_present_flag,
  277. operating_points_cnt_minus_1;
  278. int decoder_model_info_present_flag,
  279. buffer_delay_length_minus_1;
  280. if (get_bits(&gb, 1)) { // timing_info_present_flag
  281. skip_bits_long(&gb, 32); // num_units_in_display_tick
  282. skip_bits_long(&gb, 32); // time_scale
  283. if (get_bits(&gb, 1)) // equal_picture_interval
  284. uvlc(&gb); // num_ticks_per_picture_minus_1
  285. decoder_model_info_present_flag = get_bits(&gb, 1);
  286. if (decoder_model_info_present_flag) {
  287. buffer_delay_length_minus_1 = get_bits(&gb, 5);
  288. skip_bits_long(&gb, 32);
  289. skip_bits(&gb, 10);
  290. }
  291. } else
  292. decoder_model_info_present_flag = 0;
  293. initial_display_delay_present_flag = get_bits(&gb, 1);
  294. operating_points_cnt_minus_1 = get_bits(&gb, 5);
  295. for (int i = 0; i <= operating_points_cnt_minus_1; i++) {
  296. int seq_level_idx, seq_tier;
  297. skip_bits(&gb, 12);
  298. seq_level_idx = get_bits(&gb, 5);
  299. if (seq_level_idx > 7)
  300. seq_tier = get_bits(&gb, 1);
  301. else
  302. seq_tier = 0;
  303. if (decoder_model_info_present_flag) {
  304. if (get_bits(&gb, 1)) {
  305. skip_bits_long(
  306. &gb,
  307. buffer_delay_length_minus_1 +
  308. 1);
  309. skip_bits_long(
  310. &gb,
  311. buffer_delay_length_minus_1 +
  312. 1);
  313. skip_bits(&gb, 1);
  314. }
  315. }
  316. if (initial_display_delay_present_flag) {
  317. if (get_bits(&gb, 1))
  318. skip_bits(&gb, 4);
  319. }
  320. if (i == 0) {
  321. seq_params->level = seq_level_idx;
  322. seq_params->tier = seq_tier;
  323. }
  324. }
  325. }
  326. frame_width_bits_minus_1 = get_bits(&gb, 4);
  327. frame_height_bits_minus_1 = get_bits(&gb, 4);
  328. skip_bits(&gb, frame_width_bits_minus_1 + 1); // max_frame_width_minus_1
  329. skip_bits(&gb,
  330. frame_height_bits_minus_1 + 1); // max_frame_height_minus_1
  331. if (!reduced_still_picture_header) {
  332. if (get_bits(&gb, 1)) // frame_id_numbers_present_flag
  333. skip_bits(&gb, 7);
  334. }
  335. skip_bits(
  336. &gb,
  337. 3); // use_128x128_superblock (1), enable_filter_intra (1), enable_intra_edge_filter (1)
  338. if (!reduced_still_picture_header) {
  339. int enable_order_hint, seq_force_screen_content_tools;
  340. skip_bits(&gb, 4);
  341. enable_order_hint = get_bits(&gb, 1);
  342. if (enable_order_hint)
  343. skip_bits(&gb, 2);
  344. if (get_bits(&gb, 1)) // seq_choose_screen_content_tools
  345. seq_force_screen_content_tools = 2;
  346. else
  347. seq_force_screen_content_tools = get_bits(&gb, 1);
  348. if (seq_force_screen_content_tools) {
  349. if (!get_bits(&gb, 1)) // seq_choose_integer_mv
  350. skip_bits(&gb, 1); // seq_force_integer_mv
  351. }
  352. if (enable_order_hint)
  353. skip_bits(&gb, 3); // order_hint_bits_minus_1
  354. }
  355. skip_bits(&gb, 3);
  356. parse_color_config(seq_params, &gb);
  357. skip_bits(&gb, 1); // film_grain_params_present
  358. if (get_bits_left(&gb))
  359. return -1;
  360. return 0;
  361. }
  362. size_t obs_parse_av1_header(uint8_t **header, const uint8_t *data, size_t size)
  363. {
  364. if (data[0] & 0x80) {
  365. int config_record_version = data[0] & 0x7f;
  366. if (config_record_version != 1 || size < 4)
  367. return 0;
  368. *header = bmemdup(data, size);
  369. return size;
  370. }
  371. // AV1S init
  372. AV1SequenceParameters seq_params;
  373. int nb_seq = 0, seq_size = 0, meta_size = 0;
  374. const uint8_t *seq = 0, *meta = 0;
  375. uint8_t *buf = (uint8_t *)data;
  376. while (size > 0) {
  377. int64_t obu_size;
  378. int start_pos, type, temporal_id, spatial_id;
  379. assert(size <= INT_MAX);
  380. int len = parse_obu_header(buf, (int)size, &obu_size,
  381. &start_pos, &type, &temporal_id,
  382. &spatial_id);
  383. if (len < 0)
  384. return 0;
  385. switch (type) {
  386. case AV1_OBU_SEQUENCE_HEADER:
  387. nb_seq++;
  388. if (!obu_size || nb_seq > 1) {
  389. return 0;
  390. }
  391. assert(obu_size <= INT_MAX);
  392. if (parse_sequence_header(&seq_params, buf + start_pos,
  393. (int)obu_size) < 0)
  394. return 0;
  395. seq = buf;
  396. seq_size = len;
  397. break;
  398. case AV1_OBU_METADATA:
  399. if (!obu_size)
  400. return 0;
  401. meta = buf;
  402. meta_size = len;
  403. break;
  404. default:
  405. break;
  406. }
  407. size -= len;
  408. buf += len;
  409. }
  410. if (!nb_seq)
  411. return 0;
  412. uint8_t av1header[4];
  413. av1header[0] = (1 << 7) | 1; // marker and version
  414. av1header[1] = (seq_params.profile << 5) | (seq_params.level);
  415. av1header[2] = (seq_params.tier << 7) | (seq_params.bitdepth > 8) << 6 |
  416. (seq_params.bitdepth == 12) << 5 |
  417. (seq_params.monochrome) << 4 |
  418. (seq_params.chroma_subsampling_x) << 3 |
  419. (seq_params.chroma_subsampling_y) << 2 |
  420. (seq_params.chroma_sample_position);
  421. av1header[3] = 0;
  422. struct array_output_data output;
  423. struct serializer s;
  424. array_output_serializer_init(&s, &output);
  425. s_write(&s, av1header, sizeof(av1header));
  426. if (seq_size)
  427. s_write(&s, seq, seq_size);
  428. if (meta_size)
  429. s_write(&s, meta, meta_size);
  430. *header = output.bytes.array;
  431. return output.bytes.num;
  432. }
  433. static void serialize_av1_data(struct serializer *s, const uint8_t *data,
  434. size_t size, bool *is_keyframe, int *priority)
  435. {
  436. (void)is_keyframe;
  437. (void)priority;
  438. uint8_t *buf = (uint8_t *)data;
  439. uint8_t *end = (uint8_t *)data + size;
  440. enum {
  441. START_NOT_FOUND,
  442. START_FOUND,
  443. END_FOUND,
  444. OFFSET_IMPOSSIBLE,
  445. } state = START_NOT_FOUND;
  446. while (buf < end) {
  447. int64_t obu_size;
  448. int start_pos, type, temporal_id, spatial_id;
  449. assert(end - buf <= INT_MAX);
  450. int len = parse_obu_header(buf, (int)(end - buf), &obu_size,
  451. &start_pos, &type, &temporal_id,
  452. &spatial_id);
  453. if (len < 0)
  454. return;
  455. switch (type) {
  456. case AV1_OBU_TEMPORAL_DELIMITER:
  457. case AV1_OBU_REDUNDANT_FRAME_HEADER:
  458. case AV1_OBU_TILE_LIST:
  459. if (state == START_FOUND)
  460. state = END_FOUND;
  461. break;
  462. default:
  463. if (state == START_NOT_FOUND) {
  464. state = START_FOUND;
  465. } else if (state == END_FOUND) {
  466. state = OFFSET_IMPOSSIBLE;
  467. }
  468. s_write(s, buf, len);
  469. size += len;
  470. break;
  471. }
  472. buf += len;
  473. }
  474. }
  475. void obs_parse_av1_packet(struct encoder_packet *av1_packet,
  476. const struct encoder_packet *src)
  477. {
  478. struct array_output_data output;
  479. struct serializer s;
  480. long ref = 1;
  481. array_output_serializer_init(&s, &output);
  482. serialize(&s, &ref, sizeof(ref));
  483. *av1_packet = *src;
  484. serialize_av1_data(&s, src->data, src->size, &av1_packet->keyframe,
  485. &av1_packet->priority);
  486. av1_packet->data = output.bytes.array + sizeof(ref);
  487. av1_packet->size = output.bytes.num - sizeof(ref);
  488. av1_packet->drop_priority = av1_packet->priority;
  489. }