Coverage Report

Created: 2025-10-08 19:34

/work/toxav/video.c
Line
Count
Source (jump to first uncovered line)
1
/* SPDX-License-Identifier: GPL-3.0-or-later
2
 * Copyright © 2016-2025 The TokTok team.
3
 * Copyright © 2013-2015 Tox project.
4
 */
5
#include "video.h"
6
7
#include <assert.h>
8
#include <stdlib.h>
9
#include <string.h>
10
11
#include "ring_buffer.h"
12
#include "rtp.h"
13
14
#include "../toxcore/ccompat.h"
15
#include "../toxcore/logger.h"
16
#include "../toxcore/mono_time.h"
17
18
/**
19
 * Codec control function to set encoder internal speed settings. Changes in
20
 * this value influences, among others, the encoder's selection of motion
21
 * estimation methods. Values greater than 0 will increase encoder speed at the
22
 * expense of quality.
23
 *
24
 * Note Valid range for VP8: `-16..16`
25
 */
26
18
#define VP8E_SET_CPUUSED_VALUE 16
27
28
/**
29
 * Initialize encoder with this value.
30
 *
31
 * Target bandwidth to use for this stream, in kilobits per second.
32
 */
33
18
#define VIDEO_BITRATE_INITIAL_VALUE 5000
34
18
#define VIDEO_DECODE_BUFFER_SIZE 5 // this buffer has normally max. 1 entry
35
36
static vpx_codec_iface_t *video_codec_decoder_interface(void)
37
18
{
38
18
    return vpx_codec_vp8_dx();
39
18
}
40
static vpx_codec_iface_t *video_codec_encoder_interface(void)
41
36
{
42
36
    return vpx_codec_vp8_cx();
43
36
}
44
45
36
#define VIDEO_CODEC_DECODER_MAX_WIDTH  800 // its a dummy value, because the struct needs a value there
46
36
#define VIDEO_CODEC_DECODER_MAX_HEIGHT 600 // its a dummy value, because the struct needs a value there
47
48
18
#define VPX_MAX_DIST_START 40
49
50
18
#define VPX_MAX_ENCODER_THREADS 4
51
18
#define VPX_MAX_DECODER_THREADS 4
52
18
#define VIDEO_VP8_DECODER_POST_PROCESSING_ENABLED 0
53
54
static void vc_init_encoder_cfg(const Logger *log, vpx_codec_enc_cfg_t *cfg, int16_t kf_max_dist)
55
18
{
56
18
    const vpx_codec_err_t rc = vpx_codec_enc_config_default(video_codec_encoder_interface(), cfg, 0);
57
58
18
    if (rc != VPX_CODEC_OK) {
59
0
        LOGGER_ERROR(log, "vc_init_encoder_cfg:Failed to get config: %s", vpx_codec_err_to_string(rc));
60
0
    }
61
62
    /* Target bandwidth to use for this stream, in kilobits per second */
63
18
    cfg->rc_target_bitrate = VIDEO_BITRATE_INITIAL_VALUE;
64
18
    cfg->g_w = VIDEO_CODEC_DECODER_MAX_WIDTH;
65
18
    cfg->g_h = VIDEO_CODEC_DECODER_MAX_HEIGHT;
66
18
    cfg->g_pass = VPX_RC_ONE_PASS;
67
18
    cfg->g_error_resilient = VPX_ERROR_RESILIENT_DEFAULT | VPX_ERROR_RESILIENT_PARTITIONS;
68
18
    cfg->g_lag_in_frames = 0;
69
70
    /* Allow lagged encoding
71
     *
72
     * If set, this value allows the encoder to consume a number of input
73
     * frames before producing output frames. This allows the encoder to
74
     * base decisions for the current frame on future frames. This does
75
     * increase the latency of the encoding pipeline, so it is not appropriate
76
     * in all situations (ex: realtime encoding).
77
     *
78
     * Note that this is a maximum value -- the encoder may produce frames
79
     * sooner than the given limit. Set this value to 0 to disable this
80
     * feature.
81
     */
82
18
    cfg->kf_min_dist = 0;
83
18
    cfg->kf_mode = VPX_KF_AUTO; // Encoder determines optimal placement automatically
84
18
    cfg->rc_end_usage = VPX_VBR; // what quality mode?
85
86
    /*
87
     * VPX_VBR    Variable Bit Rate (VBR) mode
88
     * VPX_CBR    Constant Bit Rate (CBR) mode
89
     * VPX_CQ     Constrained Quality (CQ) mode -> give codec a hint that we may be on low bandwidth connection
90
     * VPX_Q    Constant Quality (Q) mode
91
     */
92
18
    if (kf_max_dist > 1) {
93
0
        cfg->kf_max_dist = kf_max_dist; // a full frame every x frames minimum (can be more often, codec decides automatically)
94
0
        LOGGER_DEBUG(log, "kf_max_dist=%u (1)", cfg->kf_max_dist);
95
18
    } else {
96
18
        cfg->kf_max_dist = VPX_MAX_DIST_START;
97
18
        LOGGER_DEBUG(log, "kf_max_dist=%u (2)", cfg->kf_max_dist);
98
18
    }
99
100
18
    cfg->g_threads = VPX_MAX_ENCODER_THREADS; // Maximum number of threads to use
101
    /* TODO: set these to something reasonable */
102
    // cfg->g_timebase.num = 1;
103
    // cfg->g_timebase.den = 60; // 60 fps
104
18
    cfg->rc_resize_allowed = 1; // allow encoder to resize to smaller resolution
105
18
    cfg->rc_resize_up_thresh = 40;
106
18
    cfg->rc_resize_down_thresh = 5;
107
108
    /* TODO: make quality setting an API call, but start with normal quality */
109
#if 0
110
    /* Highest-resolution encoder settings */
111
    cfg->rc_dropframe_thresh = 0;
112
    cfg->rc_resize_allowed = 0;
113
    cfg->rc_min_quantizer = 2;
114
    cfg->rc_max_quantizer = 56;
115
    cfg->rc_undershoot_pct = 100;
116
    cfg->rc_overshoot_pct = 15;
117
    cfg->rc_buf_initial_sz = 500;
118
    cfg->rc_buf_optimal_sz = 600;
119
    cfg->rc_buf_sz = 1000;
120
#endif /* 0 */
121
18
}
122
123
VCSession *vc_new(const Logger *log, const Mono_Time *mono_time, ToxAV *av, uint32_t friend_number,
124
                  toxav_video_receive_frame_cb *cb, void *cb_data)
125
18
{
126
18
    VCSession *vc = (VCSession *)calloc(1, sizeof(VCSession));
127
18
    vpx_codec_err_t rc;
128
129
18
    if (vc == nullptr) {
130
0
        LOGGER_WARNING(log, "Allocation failed! Application might misbehave!");
131
0
        return nullptr;
132
0
    }
133
134
18
    if (create_recursive_mutex(vc->queue_mutex) != 0) {
135
0
        LOGGER_WARNING(log, "Failed to create recursive mutex!");
136
0
        free(vc);
137
0
        return nullptr;
138
0
    }
139
140
18
    const int cpu_used_value = VP8E_SET_CPUUSED_VALUE;
141
142
18
    vc->vbuf_raw = rb_new(VIDEO_DECODE_BUFFER_SIZE);
143
144
18
    if (vc->vbuf_raw == nullptr) {
145
0
        goto BASE_CLEANUP;
146
0
    }
147
148
    /*
149
     * VPX_CODEC_USE_FRAME_THREADING
150
     *    Enable frame-based multi-threading
151
     *
152
     * VPX_CODEC_USE_ERROR_CONCEALMENT
153
     *    Conceal errors in decoded frames
154
     */
155
18
    vpx_codec_dec_cfg_t  dec_cfg;
156
18
    dec_cfg.threads = VPX_MAX_DECODER_THREADS; // Maximum number of threads to use
157
18
    dec_cfg.w = VIDEO_CODEC_DECODER_MAX_WIDTH;
158
18
    dec_cfg.h = VIDEO_CODEC_DECODER_MAX_HEIGHT;
159
160
18
    LOGGER_DEBUG(log, "Using VP8 codec for decoder (0)");
161
18
    rc = vpx_codec_dec_init(vc->decoder, video_codec_decoder_interface(), &dec_cfg,
162
18
                            VPX_CODEC_USE_FRAME_THREADING | VPX_CODEC_USE_POSTPROC);
163
164
18
    if (rc == VPX_CODEC_INCAPABLE) {
165
0
        LOGGER_WARNING(log, "Postproc not supported by this decoder (0)");
166
0
        rc = vpx_codec_dec_init(vc->decoder, video_codec_decoder_interface(), &dec_cfg, VPX_CODEC_USE_FRAME_THREADING);
167
0
    }
168
169
18
    if (rc != VPX_CODEC_OK) {
170
0
        LOGGER_ERROR(log, "Init video_decoder failed: %s", vpx_codec_err_to_string(rc));
171
0
        goto BASE_CLEANUP;
172
0
    }
173
174
18
    if (VIDEO_VP8_DECODER_POST_PROCESSING_ENABLED == 1) {
175
0
        vp8_postproc_cfg_t pp = {VP8_DEBLOCK, 1, 0};
176
0
        const vpx_codec_err_t cc_res = vpx_codec_control(vc->decoder, VP8_SET_POSTPROC, &pp);
177
178
0
        if (cc_res != VPX_CODEC_OK) {
179
0
            LOGGER_WARNING(log, "Failed to turn on postproc");
180
0
        } else {
181
0
            LOGGER_DEBUG(log, "turn on postproc: OK");
182
0
        }
183
18
    } else {
184
18
        vp8_postproc_cfg_t pp = {0, 0, 0};
185
18
        vpx_codec_err_t cc_res = vpx_codec_control(vc->decoder, VP8_SET_POSTPROC, &pp);
186
187
18
        if (cc_res != VPX_CODEC_OK) {
188
0
            LOGGER_WARNING(log, "Failed to turn OFF postproc");
189
18
        } else {
190
18
            LOGGER_DEBUG(log, "Disable postproc: OK");
191
18
        }
192
18
    }
193
194
    /* Set encoder to some initial values
195
     */
196
18
    vpx_codec_enc_cfg_t cfg;
197
18
    vc_init_encoder_cfg(log, &cfg, 1);
198
199
18
    LOGGER_DEBUG(log, "Using VP8 codec for encoder (0.1)");
200
18
    rc = vpx_codec_enc_init(vc->encoder, video_codec_encoder_interface(), &cfg, VPX_CODEC_USE_FRAME_THREADING);
201
202
18
    if (rc != VPX_CODEC_OK) {
203
0
        LOGGER_ERROR(log, "Failed to initialize encoder: %s", vpx_codec_err_to_string(rc));
204
0
        goto BASE_CLEANUP_1;
205
0
    }
206
207
18
    rc = vpx_codec_control(vc->encoder, VP8E_SET_CPUUSED, cpu_used_value);
208
209
18
    if (rc != VPX_CODEC_OK) {
210
0
        LOGGER_ERROR(log, "Failed to set encoder control setting: %s", vpx_codec_err_to_string(rc));
211
0
        vpx_codec_destroy(vc->encoder);
212
0
        goto BASE_CLEANUP_1;
213
0
    }
214
215
    /*
216
     * VPX_CTRL_USE_TYPE(VP8E_SET_NOISE_SENSITIVITY, unsigned int)
217
     * control function to set noise sensitivity
218
     *   0: off, 1: OnYOnly, 2: OnYUV, 3: OnYUVAggressive, 4: Adaptive
219
     */
220
#if 0
221
    rc = vpx_codec_control(vc->encoder, VP8E_SET_NOISE_SENSITIVITY, 2);
222
223
    if (rc != VPX_CODEC_OK) {
224
        LOGGER_ERROR(log, "Failed to set encoder control setting: %s", vpx_codec_err_to_string(rc));
225
        vpx_codec_destroy(vc->encoder);
226
        goto BASE_CLEANUP_1;
227
    }
228
229
#endif /* 0 */
230
231
18
    vc->linfts = current_time_monotonic(mono_time);
232
18
    vc->lcfd = 60;
233
18
    vc->vcb = cb;
234
18
    vc->vcb_user_data = cb_data;
235
18
    vc->friend_number = friend_number;
236
18
    vc->av = av;
237
18
    vc->log = log;
238
18
    return vc;
239
240
0
BASE_CLEANUP_1:
241
0
    vpx_codec_destroy(vc->decoder);
242
0
BASE_CLEANUP:
243
0
    pthread_mutex_destroy(vc->queue_mutex);
244
0
    rb_kill(vc->vbuf_raw);
245
0
    free(vc);
246
247
0
    return nullptr;
248
0
}
249
250
void vc_kill(VCSession *vc)
251
18
{
252
18
    if (vc == nullptr) {
253
0
        return;
254
0
    }
255
256
18
    vpx_codec_destroy(vc->encoder);
257
18
    vpx_codec_destroy(vc->decoder);
258
18
    void *p;
259
260
22
    while (rb_read(vc->vbuf_raw, &p)) {
261
4
        free(p);
262
4
    }
263
264
18
    rb_kill(vc->vbuf_raw);
265
18
    pthread_mutex_destroy(vc->queue_mutex);
266
18
    LOGGER_DEBUG(vc->log, "Terminated video handler: %p", (void *)vc);
267
18
    free(vc);
268
18
}
269
270
void vc_iterate(VCSession *vc)
271
210
{
272
210
    if (vc == nullptr) {
273
0
        return;
274
0
    }
275
276
210
    pthread_mutex_lock(vc->queue_mutex);
277
278
210
    struct RTPMessage *p;
279
280
210
    if (!rb_read(vc->vbuf_raw, (void **)&p)) {
281
116
        LOGGER_TRACE(vc->log, "no Video frame data available");
282
116
        pthread_mutex_unlock(vc->queue_mutex);
283
116
        return;
284
116
    }
285
286
94
    const uint16_t log_rb_size = rb_size(vc->vbuf_raw);
287
94
    pthread_mutex_unlock(vc->queue_mutex);
288
94
    const struct RTPHeader *const header = &p->header;
289
290
94
    uint32_t full_data_len;
291
292
94
    if ((header->flags & RTP_LARGE_FRAME) != 0) {
293
94
        full_data_len = header->data_length_full;
294
94
        LOGGER_DEBUG(vc->log, "vc_iterate:001:full_data_len=%d", (int)full_data_len);
295
94
    } else {
296
0
        full_data_len = p->len;
297
0
        LOGGER_DEBUG(vc->log, "vc_iterate:002");
298
0
    }
299
300
94
    LOGGER_DEBUG(vc->log, "vc_iterate: rb_read p->len=%d p->header.xe=%d", (int)full_data_len, p->header.xe);
301
94
    LOGGER_DEBUG(vc->log, "vc_iterate: rb_read rb size=%d", (int)log_rb_size);
302
94
    const vpx_codec_err_t rc = vpx_codec_decode(vc->decoder, p->data, full_data_len, nullptr, 0);
303
94
    free(p);
304
305
94
    if (rc != VPX_CODEC_OK) {
306
0
        LOGGER_ERROR(vc->log, "Error decoding video: %d %s", (int)rc, vpx_codec_err_to_string(rc));
307
0
        return;
308
0
    }
309
310
    /* Play decoded images */
311
94
    vpx_codec_iter_t iter = nullptr;
312
313
94
    for (vpx_image_t *dest = vpx_codec_get_frame(vc->decoder, &iter);
314
188
            dest != nullptr;
315
94
            dest = vpx_codec_get_frame(vc->decoder, &iter)) {
316
94
        if (vc->vcb != nullptr) {
317
94
            vc->vcb(vc->av, vc->friend_number, dest->d_w, dest->d_h,
318
94
                    dest->planes[0], dest->planes[1], dest->planes[2],
319
94
                    dest->stride[0], dest->stride[1], dest->stride[2], vc->vcb_user_data);
320
94
        }
321
322
94
        vpx_img_free(dest); // is this needed? none of the VPx examples show that
323
94
    }
324
94
}
325
326
int vc_queue_message(const Mono_Time *mono_time, void *cs, struct RTPMessage *msg)
327
101
{
328
101
    VCSession *vc = (VCSession *)cs;
329
330
    /* This function is called with complete messages
331
     * they have already been assembled.
332
     * this function gets called from handle_rtp_packet()
333
     */
334
101
    if (vc == nullptr || msg == nullptr) {
335
0
        free(msg);
336
337
0
        return -1;
338
0
    }
339
340
101
    const struct RTPHeader *const header = &msg->header;
341
342
101
    if (msg->header.pt == (RTP_TYPE_VIDEO + 2) % 128) {
343
0
        LOGGER_WARNING(vc->log, "Got dummy!");
344
0
        free(msg);
345
0
        return 0;
346
0
    }
347
348
101
    if (msg->header.pt != RTP_TYPE_VIDEO % 128) {
349
0
        LOGGER_WARNING(vc->log, "Invalid payload type! pt=%d", (int)msg->header.pt);
350
0
        free(msg);
351
0
        return -1;
352
0
    }
353
354
101
    pthread_mutex_lock(vc->queue_mutex);
355
356
101
    if ((header->flags & RTP_LARGE_FRAME) != 0 && header->pt == RTP_TYPE_VIDEO % 128) {
357
101
        LOGGER_DEBUG(vc->log, "rb_write msg->len=%d b0=%d b1=%d", (int)msg->len, (int)msg->data[0], (int)msg->data[1]);
358
101
    }
359
360
101
    free(rb_write(vc->vbuf_raw, msg));
361
362
    /* Calculate time it took for peer to send us this frame */
363
101
    const uint32_t t_lcfd = current_time_monotonic(mono_time) - vc->linfts;
364
101
    vc->lcfd = t_lcfd > 100 ? vc->lcfd : t_lcfd;
365
101
    vc->linfts = current_time_monotonic(mono_time);
366
101
    pthread_mutex_unlock(vc->queue_mutex);
367
101
    return 0;
368
101
}
369
370
int vc_reconfigure_encoder(VCSession *vc, uint32_t bit_rate, uint16_t width, uint16_t height, int16_t kf_max_dist)
371
108
{
372
108
    if (vc == nullptr) {
373
0
        return -1;
374
0
    }
375
376
108
    vpx_codec_enc_cfg_t cfg2 = *vc->encoder->config.enc;
377
378
108
    if (cfg2.rc_target_bitrate == bit_rate && cfg2.g_w == width && cfg2.g_h == height && kf_max_dist == -1) {
379
102
        return 0; /* Nothing changed */
380
102
    }
381
382
6
    if (cfg2.g_w == width && cfg2.g_h == height && kf_max_dist == -1) {
383
        /* Only bit rate changed */
384
6
        LOGGER_INFO(vc->log, "bitrate change from: %u to: %u", (uint32_t)cfg2.rc_target_bitrate, (uint32_t)bit_rate);
385
6
        cfg2.rc_target_bitrate = bit_rate;
386
6
        const vpx_codec_err_t rc = vpx_codec_enc_config_set(vc->encoder, &cfg2);
387
388
6
        if (rc != VPX_CODEC_OK) {
389
0
            LOGGER_ERROR(vc->log, "Failed to set encoder control setting: %s", vpx_codec_err_to_string(rc));
390
0
            return -1;
391
0
        }
392
6
    } else {
393
        /* Resolution is changed, must reinitialize encoder since libvpx v1.4 doesn't support
394
         * reconfiguring encoder to use resolutions greater than initially set.
395
         */
396
0
        LOGGER_DEBUG(vc->log, "Have to reinitialize vpx encoder on session %p", (void *)vc);
397
0
        vpx_codec_ctx_t new_c;
398
0
        vpx_codec_enc_cfg_t  cfg;
399
0
        vc_init_encoder_cfg(vc->log, &cfg, kf_max_dist);
400
0
        cfg.rc_target_bitrate = bit_rate;
401
0
        cfg.g_w = width;
402
0
        cfg.g_h = height;
403
404
0
        LOGGER_DEBUG(vc->log, "Using VP8 codec for encoder");
405
0
        vpx_codec_err_t rc = vpx_codec_enc_init(&new_c, video_codec_encoder_interface(), &cfg, VPX_CODEC_USE_FRAME_THREADING);
406
407
0
        if (rc != VPX_CODEC_OK) {
408
0
            LOGGER_ERROR(vc->log, "Failed to initialize encoder: %s", vpx_codec_err_to_string(rc));
409
0
            return -1;
410
0
        }
411
412
0
        const int cpu_used_value = VP8E_SET_CPUUSED_VALUE;
413
414
0
        rc = vpx_codec_control(&new_c, VP8E_SET_CPUUSED, cpu_used_value);
415
416
0
        if (rc != VPX_CODEC_OK) {
417
0
            LOGGER_ERROR(vc->log, "Failed to set encoder control setting: %s", vpx_codec_err_to_string(rc));
418
0
            vpx_codec_destroy(&new_c);
419
0
            return -1;
420
0
        }
421
422
0
        vpx_codec_destroy(vc->encoder);
423
0
        memcpy(vc->encoder, &new_c, sizeof(new_c));
424
0
    }
425
426
6
    return 0;
427
6
}