diff options
Diffstat (limited to 'toxav/audio.c')
-rw-r--r-- | toxav/audio.c | 439 |
1 files changed, 439 insertions, 0 deletions
diff --git a/toxav/audio.c b/toxav/audio.c new file mode 100644 index 00000000..ad543502 --- /dev/null +++ b/toxav/audio.c | |||
@@ -0,0 +1,439 @@ | |||
1 | /** audio.c | ||
2 | * | ||
3 | * Copyright (C) 2013-2015 Tox project All Rights Reserved. | ||
4 | * | ||
5 | * This file is part of Tox. | ||
6 | * | ||
7 | * Tox is free software: you can redistribute it and/or modify | ||
8 | * it under the terms of the GNU General Public License as published by | ||
9 | * the Free Software Foundation, either version 3 of the License, or | ||
10 | * (at your option) any later version. | ||
11 | * | ||
12 | * Tox 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 | ||
15 | * GNU General Public License for more details. | ||
16 | * | ||
17 | * You should have received a copy of the GNU General Public License | ||
18 | * along with Tox. If not, see <http://www.gnu.org/licenses/>. | ||
19 | * | ||
20 | */ | ||
21 | |||
22 | #ifdef HAVE_CONFIG_H | ||
23 | #include "config.h" | ||
24 | #endif /* HAVE_CONFIG_H */ | ||
25 | |||
26 | #include <stdlib.h> | ||
27 | |||
28 | #include "audio.h" | ||
29 | #include "rtp.h" | ||
30 | |||
31 | #include "../toxcore/logger.h" | ||
32 | |||
33 | static struct JitterBuffer *jbuf_new(uint32_t capacity); | ||
34 | static void jbuf_clear(struct JitterBuffer *q); | ||
35 | static void jbuf_free(struct JitterBuffer *q); | ||
36 | static int jbuf_write(struct JitterBuffer *q, struct RTPMessage *m); | ||
37 | static struct RTPMessage *jbuf_read(struct JitterBuffer *q, int32_t *success); | ||
38 | OpusEncoder *create_audio_encoder (int32_t bit_rate, int32_t sampling_rate, int32_t channel_count); | ||
39 | bool reconfigure_audio_encoder(OpusEncoder **e, int32_t new_br, int32_t new_sr, uint8_t new_ch, | ||
40 | int32_t *old_br, int32_t *old_sr, int32_t *old_ch); | ||
41 | bool reconfigure_audio_decoder(ACSession *ac, int32_t sampling_rate, int8_t channels); | ||
42 | |||
43 | |||
44 | |||
45 | ACSession *ac_new(ToxAV *av, uint32_t friend_number, toxav_audio_receive_frame_cb *cb, void *cb_data) | ||
46 | { | ||
47 | ACSession *ac = calloc(sizeof(ACSession), 1); | ||
48 | |||
49 | if (!ac) { | ||
50 | LOGGER_WARNING("Allocation failed! Application might misbehave!"); | ||
51 | return NULL; | ||
52 | } | ||
53 | |||
54 | if (create_recursive_mutex(ac->queue_mutex) != 0) { | ||
55 | LOGGER_WARNING("Failed to create recursive mutex!"); | ||
56 | free(ac); | ||
57 | return NULL; | ||
58 | } | ||
59 | |||
60 | int status; | ||
61 | ac->decoder = opus_decoder_create(48000, 2, &status); | ||
62 | |||
63 | if (status != OPUS_OK) { | ||
64 | LOGGER_ERROR("Error while starting audio decoder: %s", opus_strerror(status)); | ||
65 | goto BASE_CLEANUP; | ||
66 | } | ||
67 | |||
68 | if (!(ac->j_buf = jbuf_new(3))) { | ||
69 | LOGGER_WARNING("Jitter buffer creaton failed!"); | ||
70 | opus_decoder_destroy(ac->decoder); | ||
71 | goto BASE_CLEANUP; | ||
72 | } | ||
73 | |||
74 | /* Initialize encoders with default values */ | ||
75 | ac->encoder = create_audio_encoder(48000, 48000, 2); | ||
76 | |||
77 | if (ac->encoder == NULL) | ||
78 | goto DECODER_CLEANUP; | ||
79 | |||
80 | ac->le_bit_rate = 48000; | ||
81 | ac->le_sample_rate = 48000; | ||
82 | ac->le_channel_count = 2; | ||
83 | |||
84 | ac->ld_channel_count = 2; | ||
85 | ac->ld_sample_rate = 48000; | ||
86 | ac->ldrts = 0; /* Make it possible to reconfigure straight away */ | ||
87 | |||
88 | /* These need to be set in order to properly | ||
89 | * do error correction with opus */ | ||
90 | ac->lp_frame_duration = 120; | ||
91 | ac->lp_sampling_rate = 48000; | ||
92 | ac->lp_channel_count = 1; | ||
93 | |||
94 | ac->av = av; | ||
95 | ac->friend_number = friend_number; | ||
96 | ac->acb.first = cb; | ||
97 | ac->acb.second = cb_data; | ||
98 | |||
99 | return ac; | ||
100 | |||
101 | DECODER_CLEANUP: | ||
102 | opus_decoder_destroy(ac->decoder); | ||
103 | jbuf_free(ac->j_buf); | ||
104 | BASE_CLEANUP: | ||
105 | pthread_mutex_destroy(ac->queue_mutex); | ||
106 | free(ac); | ||
107 | return NULL; | ||
108 | } | ||
109 | void ac_kill(ACSession *ac) | ||
110 | { | ||
111 | if (!ac) | ||
112 | return; | ||
113 | |||
114 | opus_encoder_destroy(ac->encoder); | ||
115 | opus_decoder_destroy(ac->decoder); | ||
116 | jbuf_free(ac->j_buf); | ||
117 | |||
118 | pthread_mutex_destroy(ac->queue_mutex); | ||
119 | |||
120 | LOGGER_DEBUG("Terminated audio handler: %p", ac); | ||
121 | free(ac); | ||
122 | } | ||
123 | void ac_iterate(ACSession *ac) | ||
124 | { | ||
125 | if (!ac) | ||
126 | return; | ||
127 | |||
128 | /* TODO fix this and jitter buffering */ | ||
129 | |||
130 | /* Enough space for the maximum frame size (120 ms 48 KHz stereo audio) */ | ||
131 | int16_t tmp[5760 * 2]; | ||
132 | |||
133 | struct RTPMessage *msg; | ||
134 | int rc = 0; | ||
135 | |||
136 | pthread_mutex_lock(ac->queue_mutex); | ||
137 | |||
138 | while ((msg = jbuf_read(ac->j_buf, &rc)) || rc == 2) { | ||
139 | pthread_mutex_unlock(ac->queue_mutex); | ||
140 | |||
141 | if (rc == 2) { | ||
142 | LOGGER_DEBUG("OPUS correction"); | ||
143 | int fs = (ac->lp_sampling_rate * ac->lp_frame_duration) / 1000; | ||
144 | rc = opus_decode(ac->decoder, NULL, 0, tmp, fs, 1); | ||
145 | } else { | ||
146 | /* Get values from packet and decode. */ | ||
147 | /* NOTE: This didn't work very well | ||
148 | rc = convert_bw_to_sampling_rate(opus_packet_get_bandwidth(msg->data)); | ||
149 | if (rc != -1) { | ||
150 | cs->last_packet_sampling_rate = rc; | ||
151 | } else { | ||
152 | LOGGER_WARNING("Failed to load packet values!"); | ||
153 | rtp_free_msg(msg); | ||
154 | continue; | ||
155 | }*/ | ||
156 | |||
157 | |||
158 | /* Pick up sampling rate from packet */ | ||
159 | memcpy(&ac->lp_sampling_rate, msg->data, 4); | ||
160 | ac->lp_sampling_rate = ntohl(ac->lp_sampling_rate); | ||
161 | |||
162 | ac->lp_channel_count = opus_packet_get_nb_channels(msg->data + 4); | ||
163 | |||
164 | /** NOTE: even though OPUS supports decoding mono frames with stereo decoder and vice versa, | ||
165 | * it didn't work quite well. | ||
166 | */ | ||
167 | if (!reconfigure_audio_decoder(ac, ac->lp_sampling_rate, ac->lp_channel_count)) { | ||
168 | LOGGER_WARNING("Failed to reconfigure decoder!"); | ||
169 | free(msg); | ||
170 | continue; | ||
171 | } | ||
172 | |||
173 | rc = opus_decode(ac->decoder, msg->data + 4, msg->len - 4, tmp, 5760, 0); | ||
174 | free(msg); | ||
175 | } | ||
176 | |||
177 | if (rc < 0) { | ||
178 | LOGGER_WARNING("Decoding error: %s", opus_strerror(rc)); | ||
179 | } else if (ac->acb.first) { | ||
180 | ac->lp_frame_duration = (rc * 1000) / ac->lp_sampling_rate; | ||
181 | |||
182 | ac->acb.first(ac->av, ac->friend_number, tmp, rc, ac->lp_channel_count, | ||
183 | ac->lp_sampling_rate, ac->acb.second); | ||
184 | } | ||
185 | |||
186 | return; | ||
187 | } | ||
188 | |||
189 | pthread_mutex_unlock(ac->queue_mutex); | ||
190 | } | ||
191 | int ac_queue_message(void *acp, struct RTPMessage *msg) | ||
192 | { | ||
193 | if (!acp || !msg) | ||
194 | return -1; | ||
195 | |||
196 | if ((msg->header.pt & 0x7f) == (rtp_TypeAudio + 2) % 128) { | ||
197 | LOGGER_WARNING("Got dummy!"); | ||
198 | free(msg); | ||
199 | return 0; | ||
200 | } | ||
201 | |||
202 | if ((msg->header.pt & 0x7f) != rtp_TypeAudio % 128) { | ||
203 | LOGGER_WARNING("Invalid payload type!"); | ||
204 | free(msg); | ||
205 | return -1; | ||
206 | } | ||
207 | |||
208 | ACSession *ac = acp; | ||
209 | |||
210 | pthread_mutex_lock(ac->queue_mutex); | ||
211 | int rc = jbuf_write(ac->j_buf, msg); | ||
212 | pthread_mutex_unlock(ac->queue_mutex); | ||
213 | |||
214 | if (rc == -1) { | ||
215 | LOGGER_WARNING("Could not queue the message!"); | ||
216 | free(msg); | ||
217 | return -1; | ||
218 | } | ||
219 | |||
220 | return 0; | ||
221 | } | ||
222 | int ac_reconfigure_encoder(ACSession *ac, int32_t bit_rate, int32_t sampling_rate, uint8_t channels) | ||
223 | { | ||
224 | if (!ac || !reconfigure_audio_encoder(&ac->encoder, bit_rate, | ||
225 | sampling_rate, channels, | ||
226 | &ac->le_bit_rate, | ||
227 | &ac->le_sample_rate, | ||
228 | &ac->le_channel_count)) | ||
229 | return -1; | ||
230 | |||
231 | return 0; | ||
232 | } | ||
233 | |||
234 | |||
235 | |||
236 | struct JitterBuffer { | ||
237 | struct RTPMessage **queue; | ||
238 | uint32_t size; | ||
239 | uint32_t capacity; | ||
240 | uint16_t bottom; | ||
241 | uint16_t top; | ||
242 | }; | ||
243 | |||
244 | static struct JitterBuffer *jbuf_new(uint32_t capacity) | ||
245 | { | ||
246 | unsigned int size = 1; | ||
247 | |||
248 | while (size <= (capacity * 4)) { | ||
249 | size *= 2; | ||
250 | } | ||
251 | |||
252 | struct JitterBuffer *q; | ||
253 | |||
254 | if (!(q = calloc(sizeof(struct JitterBuffer), 1))) return NULL; | ||
255 | |||
256 | if (!(q->queue = calloc(sizeof(struct RTPMessage *), size))) { | ||
257 | free(q); | ||
258 | return NULL; | ||
259 | } | ||
260 | |||
261 | q->size = size; | ||
262 | q->capacity = capacity; | ||
263 | return q; | ||
264 | } | ||
265 | static void jbuf_clear(struct JitterBuffer *q) | ||
266 | { | ||
267 | for (; q->bottom != q->top; ++q->bottom) { | ||
268 | if (q->queue[q->bottom % q->size]) { | ||
269 | free(q->queue[q->bottom % q->size]); | ||
270 | q->queue[q->bottom % q->size] = NULL; | ||
271 | } | ||
272 | } | ||
273 | } | ||
274 | static void jbuf_free(struct JitterBuffer *q) | ||
275 | { | ||
276 | if (!q) return; | ||
277 | |||
278 | jbuf_clear(q); | ||
279 | free(q->queue); | ||
280 | free(q); | ||
281 | } | ||
282 | static int jbuf_write(struct JitterBuffer *q, struct RTPMessage *m) | ||
283 | { | ||
284 | uint16_t sequnum = m->header.sequnum; | ||
285 | |||
286 | unsigned int num = sequnum % q->size; | ||
287 | |||
288 | if ((uint32_t)(sequnum - q->bottom) > q->size) { | ||
289 | LOGGER_DEBUG("Clearing filled jitter buffer: %p", q); | ||
290 | |||
291 | jbuf_clear(q); | ||
292 | q->bottom = sequnum - q->capacity; | ||
293 | q->queue[num] = m; | ||
294 | q->top = sequnum + 1; | ||
295 | return 0; | ||
296 | } | ||
297 | |||
298 | if (q->queue[num]) | ||
299 | return -1; | ||
300 | |||
301 | q->queue[num] = m; | ||
302 | |||
303 | if ((sequnum - q->bottom) >= (q->top - q->bottom)) | ||
304 | q->top = sequnum + 1; | ||
305 | |||
306 | return 0; | ||
307 | } | ||
308 | static struct RTPMessage *jbuf_read(struct JitterBuffer *q, int32_t *success) | ||
309 | { | ||
310 | if (q->top == q->bottom) { | ||
311 | *success = 0; | ||
312 | return NULL; | ||
313 | } | ||
314 | |||
315 | unsigned int num = q->bottom % q->size; | ||
316 | |||
317 | if (q->queue[num]) { | ||
318 | struct RTPMessage *ret = q->queue[num]; | ||
319 | q->queue[num] = NULL; | ||
320 | ++q->bottom; | ||
321 | *success = 1; | ||
322 | return ret; | ||
323 | } | ||
324 | |||
325 | if ((uint32_t)(q->top - q->bottom) > q->capacity) { | ||
326 | ++q->bottom; | ||
327 | *success = 2; | ||
328 | return NULL; | ||
329 | } | ||
330 | |||
331 | *success = 0; | ||
332 | return NULL; | ||
333 | } | ||
334 | OpusEncoder *create_audio_encoder (int32_t bit_rate, int32_t sampling_rate, int32_t channel_count) | ||
335 | { | ||
336 | int status = OPUS_OK; | ||
337 | OpusEncoder *rc = opus_encoder_create(sampling_rate, channel_count, OPUS_APPLICATION_VOIP, &status); | ||
338 | |||
339 | if (status != OPUS_OK) { | ||
340 | LOGGER_ERROR("Error while starting audio encoder: %s", opus_strerror(status)); | ||
341 | return NULL; | ||
342 | } | ||
343 | |||
344 | status = opus_encoder_ctl(rc, OPUS_SET_BITRATE(bit_rate)); | ||
345 | |||
346 | if (status != OPUS_OK) { | ||
347 | LOGGER_ERROR("Error while setting encoder ctl: %s", opus_strerror(status)); | ||
348 | goto FAILURE; | ||
349 | } | ||
350 | |||
351 | /* Enable in-band forward error correction in codec */ | ||
352 | status = opus_encoder_ctl(rc, OPUS_SET_INBAND_FEC(1)); | ||
353 | |||
354 | if (status != OPUS_OK) { | ||
355 | LOGGER_ERROR("Error while setting encoder ctl: %s", opus_strerror(status)); | ||
356 | goto FAILURE; | ||
357 | } | ||
358 | |||
359 | /* Make codec resistant to up to 10% packet loss | ||
360 | * NOTE This could also be adjusted on the fly, rather than hard-coded, | ||
361 | * with feedback from the receiving client. | ||
362 | */ | ||
363 | status = opus_encoder_ctl(rc, OPUS_SET_PACKET_LOSS_PERC(10)); | ||
364 | |||
365 | if (status != OPUS_OK) { | ||
366 | LOGGER_ERROR("Error while setting encoder ctl: %s", opus_strerror(status)); | ||
367 | goto FAILURE; | ||
368 | } | ||
369 | |||
370 | /* Set algorithm to the highest complexity, maximizing compression */ | ||
371 | status = opus_encoder_ctl(rc, OPUS_SET_COMPLEXITY(10)); | ||
372 | |||
373 | if (status != OPUS_OK) { | ||
374 | LOGGER_ERROR("Error while setting encoder ctl: %s", opus_strerror(status)); | ||
375 | goto FAILURE; | ||
376 | } | ||
377 | |||
378 | return rc; | ||
379 | |||
380 | FAILURE: | ||
381 | opus_encoder_destroy(rc); | ||
382 | return NULL; | ||
383 | } | ||
384 | bool reconfigure_audio_encoder(OpusEncoder **e, int32_t new_br, int32_t new_sr, uint8_t new_ch, | ||
385 | int32_t *old_br, int32_t *old_sr, int32_t *old_ch) | ||
386 | { | ||
387 | /* Values are checked in toxav.c */ | ||
388 | if (*old_sr != new_sr || *old_ch != new_ch) { | ||
389 | OpusEncoder *new_encoder = create_audio_encoder(new_br, new_sr, new_ch); | ||
390 | |||
391 | if (new_encoder == NULL) | ||
392 | return false; | ||
393 | |||
394 | opus_encoder_destroy(*e); | ||
395 | *e = new_encoder; | ||
396 | } else if (*old_br == new_br) | ||
397 | return true; /* Nothing changed */ | ||
398 | else { | ||
399 | int status = opus_encoder_ctl(*e, OPUS_SET_BITRATE(new_br)); | ||
400 | |||
401 | if (status != OPUS_OK) { | ||
402 | LOGGER_ERROR("Error while setting encoder ctl: %s", opus_strerror(status)); | ||
403 | return false; | ||
404 | } | ||
405 | } | ||
406 | |||
407 | *old_br = new_br; | ||
408 | *old_sr = new_sr; | ||
409 | *old_ch = new_ch; | ||
410 | |||
411 | LOGGER_DEBUG ("Reconfigured audio encoder br: %d sr: %d cc:%d", new_br, new_sr, new_ch); | ||
412 | return true; | ||
413 | } | ||
414 | bool reconfigure_audio_decoder(ACSession *ac, int32_t sampling_rate, int8_t channels) | ||
415 | { | ||
416 | if (sampling_rate != ac->ld_sample_rate || channels != ac->ld_channel_count) { | ||
417 | if (current_time_monotonic() - ac->ldrts < 500) | ||
418 | return false; | ||
419 | |||
420 | int status; | ||
421 | OpusDecoder *new_dec = opus_decoder_create(sampling_rate, channels, &status); | ||
422 | |||
423 | if (status != OPUS_OK) { | ||
424 | LOGGER_ERROR("Error while starting audio decoder(%d %d): %s", sampling_rate, channels, opus_strerror(status)); | ||
425 | return false; | ||
426 | } | ||
427 | |||
428 | ac->ld_sample_rate = sampling_rate; | ||
429 | ac->ld_channel_count = channels; | ||
430 | ac->ldrts = current_time_monotonic(); | ||
431 | |||
432 | opus_decoder_destroy(ac->decoder); | ||
433 | ac->decoder = new_dec; | ||
434 | |||
435 | LOGGER_DEBUG("Reconfigured audio decoder sr: %d cc: %d", sampling_rate, channels); | ||
436 | } | ||
437 | |||
438 | return true; | ||
439 | } | ||