FreeRDP
Loading...
Searching...
No Matches
dsp_ffmpeg.c
1
21#include <freerdp/config.h>
22
23#include <freerdp/log.h>
24
25#include <libavcodec/avcodec.h>
26#include <libavutil/avutil.h>
27#include <libavutil/opt.h>
28#if defined(SWRESAMPLE_FOUND)
29#include <libswresample/swresample.h>
30#elif defined(AVRESAMPLE_FOUND)
31#include <libavresample/avresample.h>
32#else
33#error "libswresample or libavresample required"
34#endif
35
36#include "dsp.h"
37#include "dsp_ffmpeg.h"
38
39#define TAG FREERDP_TAG("dsp.ffmpeg")
40
41struct S_FREERDP_DSP_CONTEXT
42{
44
45 BOOL isOpen;
46
47 UINT32 bufferedSamples;
48
49 enum AVCodecID id;
50 const AVCodec* codec;
51 AVCodecContext* context;
52 AVFrame* frame;
53 AVFrame* resampled;
54 AVFrame* buffered;
55 AVPacket* packet;
56#if defined(SWRESAMPLE_FOUND)
57 SwrContext* rcontext;
58#else
59 AVAudioResampleContext* rcontext;
60#endif
61};
62
63static BOOL ffmpeg_codec_is_filtered(enum AVCodecID id, WINPR_ATTR_UNUSED BOOL encoder)
64{
65 switch (id)
66 {
67#if !defined(WITH_DSP_EXPERIMENTAL)
68
69 case AV_CODEC_ID_ADPCM_IMA_OKI:
70 case AV_CODEC_ID_MP3:
71 case AV_CODEC_ID_ADPCM_MS:
72 case AV_CODEC_ID_G723_1:
73 case AV_CODEC_ID_GSM_MS:
74 case AV_CODEC_ID_PCM_ALAW:
75 case AV_CODEC_ID_PCM_MULAW:
76 return TRUE;
77#endif
78
79 case AV_CODEC_ID_NONE:
80 return TRUE;
81
82 case AV_CODEC_ID_AAC:
83 case AV_CODEC_ID_AAC_LATM:
84 return FALSE;
85
86 default:
87 return FALSE;
88 }
89}
90
91static enum AVCodecID ffmpeg_get_avcodec(const AUDIO_FORMAT* WINPR_RESTRICT format)
92{
93 if (!format)
94 return AV_CODEC_ID_NONE;
95
96 switch (format->wFormatTag)
97 {
98 case WAVE_FORMAT_UNKNOWN:
99 return AV_CODEC_ID_NONE;
100
101 case WAVE_FORMAT_PCM:
102 switch (format->wBitsPerSample)
103 {
104 case 16:
105 return AV_CODEC_ID_PCM_U16LE;
106
107 case 8:
108 return AV_CODEC_ID_PCM_U8;
109
110 default:
111 return AV_CODEC_ID_NONE;
112 }
113
114 case WAVE_FORMAT_DVI_ADPCM:
115 return AV_CODEC_ID_ADPCM_IMA_OKI;
116
117 case WAVE_FORMAT_ADPCM:
118 return AV_CODEC_ID_ADPCM_MS;
119
120 case WAVE_FORMAT_ALAW:
121 return AV_CODEC_ID_PCM_ALAW;
122
123 case WAVE_FORMAT_MULAW:
124 return AV_CODEC_ID_PCM_MULAW;
125
126 case WAVE_FORMAT_GSM610:
127 return AV_CODEC_ID_GSM_MS;
128
129 case WAVE_FORMAT_MSG723:
130 return AV_CODEC_ID_G723_1;
131
132 case WAVE_FORMAT_AAC_MS:
133 return AV_CODEC_ID_AAC;
134
135 case WAVE_FORMAT_OPUS:
136 return AV_CODEC_ID_OPUS;
137
138 default:
139 return AV_CODEC_ID_NONE;
140 }
141}
142
143static int ffmpeg_sample_format(const AUDIO_FORMAT* WINPR_RESTRICT format)
144{
145 switch (format->wFormatTag)
146 {
147 case WAVE_FORMAT_PCM:
148 switch (format->wBitsPerSample)
149 {
150 case 8:
151 return AV_SAMPLE_FMT_U8;
152
153 case 16:
154 return AV_SAMPLE_FMT_S16;
155
156 default:
157 return FALSE;
158 }
159
160 case WAVE_FORMAT_DVI_ADPCM:
161 case WAVE_FORMAT_ADPCM:
162 return AV_SAMPLE_FMT_S16P;
163
164 case WAVE_FORMAT_MPEGLAYER3:
165 case WAVE_FORMAT_AAC_MS:
166 return AV_SAMPLE_FMT_FLTP;
167
168 case WAVE_FORMAT_OPUS:
169 return AV_SAMPLE_FMT_S16;
170
171 case WAVE_FORMAT_MSG723:
172 case WAVE_FORMAT_GSM610:
173 return AV_SAMPLE_FMT_S16P;
174
175 case WAVE_FORMAT_ALAW:
176 return AV_SAMPLE_FMT_S16;
177
178 default:
179 return FALSE;
180 }
181}
182
183static void ffmpeg_close_context(FREERDP_DSP_CONTEXT* WINPR_RESTRICT context)
184{
185 if (context)
186 {
187 if (context->context)
188 avcodec_free_context(&context->context);
189
190 if (context->frame)
191 av_frame_free(&context->frame);
192
193 if (context->resampled)
194 av_frame_free(&context->resampled);
195
196 if (context->buffered)
197 av_frame_free(&context->buffered);
198
199 if (context->packet)
200 av_packet_free(&context->packet);
201
202 if (context->rcontext)
203 {
204#if defined(SWRESAMPLE_FOUND)
205 swr_free(&context->rcontext);
206#else
207 avresample_free(&context->rcontext);
208#endif
209 }
210
211 context->id = AV_CODEC_ID_NONE;
212 context->codec = NULL;
213 context->isOpen = FALSE;
214 context->context = NULL;
215 context->frame = NULL;
216 context->resampled = NULL;
217 context->packet = NULL;
218 context->rcontext = NULL;
219 }
220}
221
222static BOOL ffmpeg_open_context(FREERDP_DSP_CONTEXT* WINPR_RESTRICT context)
223{
224 int ret = 0;
225
226 if (!context || context->isOpen)
227 return FALSE;
228
229 const AUDIO_FORMAT* format = &context->common.format;
230
231 if (!format)
232 return FALSE;
233 context->id = ffmpeg_get_avcodec(format);
234
235 if (ffmpeg_codec_is_filtered(context->id, context->common.encoder))
236 goto fail;
237
238 if (context->common.encoder)
239 context->codec = avcodec_find_encoder(context->id);
240 else
241 context->codec = avcodec_find_decoder(context->id);
242
243 if (!context->codec)
244 goto fail;
245
246 context->context = avcodec_alloc_context3(context->codec);
247
248 if (!context->context)
249 goto fail;
250
251 switch (context->id)
252 {
253 /* We need support for multichannel and sample rates != 8000 */
254 case AV_CODEC_ID_GSM_MS:
255 context->context->strict_std_compliance = FF_COMPLIANCE_UNOFFICIAL;
256 break;
257
258 case AV_CODEC_ID_AAC:
259 context->context->profile = FF_PROFILE_AAC_MAIN;
260 break;
261
262 default:
263 break;
264 }
265
266 context->context->max_b_frames = 1;
267 context->context->delay = 0;
268
269#if LIBAVUTIL_VERSION_INT >= AV_VERSION_INT(57, 28, 100)
270 av_channel_layout_default(&context->context->ch_layout, format->nChannels);
271#else
272 context->context->channels = format->nChannels;
273 const int64_t layout = av_get_default_channel_layout(format->nChannels);
274 context->context->channel_layout = layout;
275#endif
276 context->context->sample_rate = (int)format->nSamplesPerSec;
277 context->context->block_align = format->nBlockAlign;
278 context->context->bit_rate = format->nAvgBytesPerSec * 8LL;
279 context->context->sample_fmt = ffmpeg_sample_format(format);
280 context->context->time_base = av_make_q(1, context->context->sample_rate);
281
282 if ((ret = avcodec_open2(context->context, context->codec, NULL)) < 0)
283 {
284 const char* err = av_err2str(ret);
285 WLog_ERR(TAG, "Error avcodec_open2 %s [%d]", err, ret);
286 goto fail;
287 }
288
289 context->packet = av_packet_alloc();
290
291 if (!context->packet)
292 goto fail;
293
294 context->frame = av_frame_alloc();
295
296 if (!context->frame)
297 goto fail;
298
299 context->resampled = av_frame_alloc();
300
301 if (!context->resampled)
302 goto fail;
303
304 context->buffered = av_frame_alloc();
305
306 if (!context->buffered)
307 goto fail;
308
309#if defined(SWRESAMPLE_FOUND)
310 context->rcontext = swr_alloc();
311#else
312 context->rcontext = avresample_alloc_context();
313#endif
314
315 if (!context->rcontext)
316 goto fail;
317
318#if LIBAVUTIL_VERSION_INT >= AV_VERSION_INT(57, 28, 100)
319 av_channel_layout_default(&context->frame->ch_layout, format->nChannels);
320#else
321 context->frame->channel_layout = layout;
322 context->frame->channels = format->nChannels;
323#endif
324 WINPR_ASSERT(format->nSamplesPerSec <= INT_MAX);
325 context->frame->sample_rate = (int)format->nSamplesPerSec;
326 context->frame->format = AV_SAMPLE_FMT_S16;
327
328 if (context->common.encoder)
329 {
330 context->resampled->format = context->context->sample_fmt;
331 context->resampled->sample_rate = context->context->sample_rate;
332 }
333 else
334 {
335 context->resampled->format = AV_SAMPLE_FMT_S16;
336
337 WINPR_ASSERT(format->nSamplesPerSec <= INT_MAX);
338 context->resampled->sample_rate = (int)format->nSamplesPerSec;
339 }
340
341#if LIBAVUTIL_VERSION_INT >= AV_VERSION_INT(57, 28, 100)
342 av_channel_layout_default(&context->resampled->ch_layout, format->nChannels);
343#else
344 context->resampled->channel_layout = layout;
345 context->resampled->channels = format->nChannels;
346#endif
347
348 if (context->context->frame_size > 0)
349 {
350#if LIBAVUTIL_VERSION_INT >= AV_VERSION_INT(57, 28, 100)
351 ret = av_channel_layout_copy(&context->buffered->ch_layout, &context->resampled->ch_layout);
352 if (ret != 0)
353 goto fail;
354#else
355 context->buffered->channel_layout = context->resampled->channel_layout;
356 context->buffered->channels = context->resampled->channels;
357#endif
358 context->buffered->format = context->resampled->format;
359 context->buffered->nb_samples = context->context->frame_size;
360
361 ret = av_frame_get_buffer(context->buffered, 1);
362 if (ret < 0)
363 goto fail;
364 }
365
366 context->isOpen = TRUE;
367 return TRUE;
368fail:
369 ffmpeg_close_context(context);
370 return FALSE;
371}
372
373#if defined(SWRESAMPLE_FOUND)
374static BOOL ffmpeg_resample_frame(SwrContext* WINPR_RESTRICT context, AVFrame* WINPR_RESTRICT in,
375 AVFrame* WINPR_RESTRICT out)
376{
377 int ret = 0;
378
379 if (!swr_is_initialized(context))
380 {
381 if ((ret = swr_config_frame(context, out, in)) < 0)
382 {
383 const char* err = av_err2str(ret);
384 WLog_ERR(TAG, "Error during resampling %s [%d]", err, ret);
385 return FALSE;
386 }
387
388 if ((ret = (swr_init(context))) < 0)
389 {
390 const char* err = av_err2str(ret);
391 WLog_ERR(TAG, "Error during resampling %s [%d]", err, ret);
392 return FALSE;
393 }
394 }
395
396 if ((ret = swr_convert_frame(context, out, in)) < 0)
397 {
398 const char* err = av_err2str(ret);
399 WLog_ERR(TAG, "Error during resampling %s [%d]", err, ret);
400 return FALSE;
401 }
402
403 return TRUE;
404}
405#else
406static BOOL ffmpeg_resample_frame(AVAudioResampleContext* WINPR_RESTRICT context,
407 AVFrame* WINPR_RESTRICT in, AVFrame* WINPR_RESTRICT out)
408{
409 int ret;
410
411 if (!avresample_is_open(context))
412 {
413 if ((ret = avresample_config(context, out, in)) < 0)
414 {
415 const char* err = av_err2str(ret);
416 WLog_ERR(TAG, "Error during resampling %s [%d]", err, ret);
417 return FALSE;
418 }
419
420 if ((ret = (avresample_open(context))) < 0)
421 {
422 const char* err = av_err2str(ret);
423 WLog_ERR(TAG, "Error during resampling %s [%d]", err, ret);
424 return FALSE;
425 }
426 }
427
428 if ((ret = avresample_convert_frame(context, out, in)) < 0)
429 {
430 const char* err = av_err2str(ret);
431 WLog_ERR(TAG, "Error during resampling %s [%d]", err, ret);
432 return FALSE;
433 }
434
435 return TRUE;
436}
437#endif
438
439static BOOL ffmpeg_encode_frame(AVCodecContext* WINPR_RESTRICT context, AVFrame* WINPR_RESTRICT in,
440 AVPacket* WINPR_RESTRICT packet, wStream* WINPR_RESTRICT out)
441{
442 if (in->format == AV_SAMPLE_FMT_FLTP)
443 {
444 uint8_t** pp = in->extended_data;
445#if LIBAVUTIL_VERSION_INT < AV_VERSION_INT(57, 28, 100)
446 const int nr_channels = in->channels;
447#else
448 const int nr_channels = in->ch_layout.nb_channels;
449#endif
450
451 for (int y = 0; y < nr_channels; y++)
452 {
453 float* data = (float*)pp[y];
454 for (int x = 0; x < in->nb_samples; x++)
455 {
456 const float val1 = data[x];
457 if (isnan(val1))
458 data[x] = 0.0f;
459 else if (isinf(val1))
460 {
461 if (val1 < 0.0f)
462 data[x] = -1.0f;
463 else
464 data[x] = 1.0f;
465 }
466 }
467 }
468 }
469 /* send the packet with the compressed data to the encoder */
470 int ret = avcodec_send_frame(context, in);
471
472 if (ret < 0)
473 {
474 const char* err = av_err2str(ret);
475 // Ignore errors: AAC encoder sometimes returns -22
476 // The log message from ffmpeg is '[aac @ 0x7f140db753c0] Input contains (near) NaN/+-Inf'
477 if (ret == AVERROR(EINVAL))
478 {
479 WLog_DBG(TAG, "Error submitting the packet to the encoder %s [%d], ignoring", err, ret);
480 return TRUE;
481 }
482
483 WLog_ERR(TAG, "Error submitting the packet to the encoder %s [%d]", err, ret);
484
485 return FALSE;
486 }
487
488 /* read all the output frames (in general there may be any number of them */
489 while (TRUE)
490 {
491 ret = avcodec_receive_packet(context, packet);
492
493 if ((ret == AVERROR(EAGAIN)) || (ret == AVERROR_EOF))
494 break;
495
496 if (ret < 0)
497 {
498 const char* err = av_err2str(ret);
499 WLog_ERR(TAG, "Error during encoding %s [%d]", err, ret);
500 return FALSE;
501 }
502
503 WINPR_ASSERT(packet->size >= 0);
504 if (!Stream_EnsureRemainingCapacity(out, (size_t)packet->size))
505 return FALSE;
506
507 Stream_Write(out, packet->data, (size_t)packet->size);
508 av_packet_unref(packet);
509 }
510
511 return TRUE;
512}
513
514static BOOL ffmpeg_fill_frame(AVFrame* WINPR_RESTRICT frame,
515 const AUDIO_FORMAT* WINPR_RESTRICT inputFormat,
516 const BYTE* WINPR_RESTRICT data, size_t size)
517{
518 int ret = 0;
519#if LIBAVUTIL_VERSION_INT < AV_VERSION_INT(57, 28, 100)
520 frame->channels = inputFormat->nChannels;
521 frame->channel_layout = av_get_default_channel_layout(frame->channels);
522#else
523 av_channel_layout_default(&frame->ch_layout, inputFormat->nChannels);
524#endif
525 WINPR_ASSERT(inputFormat->nSamplesPerSec <= INT_MAX);
526 frame->sample_rate = (int)inputFormat->nSamplesPerSec;
527 frame->format = ffmpeg_sample_format(inputFormat);
528
529 const int bpp = av_get_bytes_per_sample(frame->format);
530 WINPR_ASSERT(bpp >= 0);
531 WINPR_ASSERT(size <= INT_MAX);
532 const size_t nb_samples = size / inputFormat->nChannels / (size_t)bpp;
533 frame->nb_samples = (int)nb_samples;
534
535 if ((ret = avcodec_fill_audio_frame(frame, inputFormat->nChannels, frame->format, data,
536 (int)size, 1)) < 0)
537 {
538 const char* err = av_err2str(ret);
539 WLog_ERR(TAG, "Error during audio frame fill %s [%d]", err, ret);
540 return FALSE;
541 }
542
543 return TRUE;
544}
545#if defined(SWRESAMPLE_FOUND)
546static BOOL ffmpeg_decode(AVCodecContext* WINPR_RESTRICT dec_ctx, AVPacket* WINPR_RESTRICT pkt,
547 AVFrame* WINPR_RESTRICT frame, SwrContext* WINPR_RESTRICT resampleContext,
548 AVFrame* WINPR_RESTRICT resampled, wStream* WINPR_RESTRICT out)
549#else
550static BOOL ffmpeg_decode(AVCodecContext* dec_ctx, AVPacket* pkt, AVFrame* frame,
551 AVAudioResampleContext* resampleContext, AVFrame* resampled, wStream* out)
552#endif
553{
554 /* send the packet with the compressed data to the decoder */
555 int ret = avcodec_send_packet(dec_ctx, pkt);
556
557 if (ret < 0)
558 {
559 const char* err = av_err2str(ret);
560 WLog_ERR(TAG, "Error submitting the packet to the decoder %s [%d]", err, ret);
561 return FALSE;
562 }
563
564 /* read all the output frames (in general there may be any number of them */
565 while (ret >= 0)
566 {
567 ret = avcodec_receive_frame(dec_ctx, frame);
568
569 if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF)
570 break;
571
572 if (ret < 0)
573 {
574 const char* err = av_err2str(ret);
575 WLog_ERR(TAG, "Error during decoding %s [%d]", err, ret);
576 return FALSE;
577 }
578
579#if defined(SWRESAMPLE_FOUND)
580 if (!swr_is_initialized(resampleContext))
581 {
582 if ((ret = swr_config_frame(resampleContext, resampled, frame)) < 0)
583 {
584#else
585 if (!avresample_is_open(resampleContext))
586 {
587 if ((ret = avresample_config(resampleContext, resampled, frame)) < 0)
588 {
589#endif
590 const char* err = av_err2str(ret);
591 WLog_ERR(TAG, "Error during resampling %s [%d]", err, ret);
592 return FALSE;
593 }
594
595#if defined(SWRESAMPLE_FOUND)
596 ret = (swr_init(resampleContext));
597#else
598 ret = (avresample_open(resampleContext));
599#endif
600 if (ret < 0)
601 {
602 const char* err = av_err2str(ret);
603 WLog_ERR(TAG, "Error during resampling %s [%d]", err, ret);
604 return FALSE;
605 }
606 }
607
608#if defined(SWRESAMPLE_FOUND)
609 ret = swr_convert_frame(resampleContext, resampled, frame);
610#else
611 ret = avresample_convert_frame(resampleContext, resampled, frame);
612#endif
613 if (ret < 0)
614 {
615 const char* err = av_err2str(ret);
616 WLog_ERR(TAG, "Error during resampling %s [%d]", err, ret);
617 return FALSE;
618 }
619
620 {
621
622#if LIBAVUTIL_VERSION_INT >= AV_VERSION_INT(57, 28, 100)
623 WINPR_ASSERT(resampled->ch_layout.nb_channels >= 0);
624 const size_t nrchannels = (size_t)resampled->ch_layout.nb_channels;
625#else
626 const size_t nrchannels = resampled->channels;
627#endif
628 WINPR_ASSERT(resampled->nb_samples >= 0);
629 const size_t data_size = nrchannels * (size_t)resampled->nb_samples * 2ull;
630 if (!Stream_EnsureRemainingCapacity(out, data_size))
631 return FALSE;
632 Stream_Write(out, resampled->data[0], data_size);
633 }
634 }
635
636 return TRUE;
637}
638
639BOOL freerdp_dsp_ffmpeg_supports_format(const AUDIO_FORMAT* WINPR_RESTRICT format, BOOL encode)
640{
641 enum AVCodecID id = ffmpeg_get_avcodec(format);
642
643 if (ffmpeg_codec_is_filtered(id, encode))
644 return FALSE;
645
646 if (encode)
647 return avcodec_find_encoder(id) != NULL;
648 else
649 return avcodec_find_decoder(id) != NULL;
650}
651
652FREERDP_DSP_CONTEXT* freerdp_dsp_ffmpeg_context_new(BOOL encode)
653{
654 FREERDP_DSP_CONTEXT* context = NULL;
655#if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(58, 10, 100)
656 avcodec_register_all();
657#endif
658 context = calloc(1, sizeof(FREERDP_DSP_CONTEXT));
659
660 if (!context)
661 goto fail;
662
663 if (!freerdp_dsp_common_context_init(&context->common, encode))
664 goto fail;
665
666 return context;
667
668fail:
669 WINPR_PRAGMA_DIAG_PUSH
670 WINPR_PRAGMA_DIAG_IGNORED_MISMATCHED_DEALLOC
671 freerdp_dsp_ffmpeg_context_free(context);
672 WINPR_PRAGMA_DIAG_POP
673 return NULL;
674}
675
676void freerdp_dsp_ffmpeg_context_free(FREERDP_DSP_CONTEXT* context)
677{
678 if (context)
679 {
680 ffmpeg_close_context(context);
681 freerdp_dsp_common_context_uninit(&context->common);
682 free(context);
683 }
684}
685
686BOOL freerdp_dsp_ffmpeg_context_reset(FREERDP_DSP_CONTEXT* WINPR_RESTRICT context,
687 const AUDIO_FORMAT* WINPR_RESTRICT targetFormat)
688{
689 if (!context || !targetFormat)
690 return FALSE;
691
692 ffmpeg_close_context(context);
693 context->common.format = *targetFormat;
694 return ffmpeg_open_context(context);
695}
696
697static BOOL freerdp_dsp_channel_mix(FREERDP_DSP_CONTEXT* WINPR_RESTRICT context,
698 const BYTE* WINPR_RESTRICT src, size_t size,
699 const AUDIO_FORMAT* WINPR_RESTRICT srcFormat,
700 const BYTE** WINPR_RESTRICT data, size_t* WINPR_RESTRICT length,
701 AUDIO_FORMAT* WINPR_RESTRICT dstFormat)
702{
703 UINT32 bpp = 0;
704 size_t samples = 0;
705
706 if (!context || !data || !length || !dstFormat)
707 return FALSE;
708
709 if (srcFormat->wFormatTag != WAVE_FORMAT_PCM)
710 return FALSE;
711
712 bpp = srcFormat->wBitsPerSample > 8 ? 2 : 1;
713 samples = size / bpp / srcFormat->nChannels;
714
715 *dstFormat = *srcFormat;
716 if (context->common.format.nChannels == srcFormat->nChannels)
717 {
718 *data = src;
719 *length = size;
720 return TRUE;
721 }
722
723 Stream_SetPosition(context->common.channelmix, 0);
724
725 /* Destination has more channels than source */
726 if (context->common.format.nChannels > srcFormat->nChannels)
727 {
728 switch (srcFormat->nChannels)
729 {
730 case 1:
731 if (!Stream_EnsureCapacity(context->common.channelmix, size * 2))
732 return FALSE;
733
734 for (size_t x = 0; x < samples; x++)
735 {
736 for (size_t y = 0; y < bpp; y++)
737 Stream_Write_UINT8(context->common.channelmix, src[x * bpp + y]);
738
739 for (size_t y = 0; y < bpp; y++)
740 Stream_Write_UINT8(context->common.channelmix, src[x * bpp + y]);
741 }
742
743 Stream_SealLength(context->common.channelmix);
744 *data = Stream_Buffer(context->common.channelmix);
745 *length = Stream_Length(context->common.channelmix);
746 dstFormat->nChannels = 2;
747 return TRUE;
748
749 case 2: /* We only support stereo, so we can not handle this case. */
750 default: /* Unsupported number of channels */
751 WLog_WARN(TAG, "[%s] unsupported source channel count %" PRIu16, __func__,
752 srcFormat->nChannels);
753 return FALSE;
754 }
755 }
756
757 /* Destination has less channels than source */
758 switch (srcFormat->nChannels)
759 {
760 case 2:
761 if (!Stream_EnsureCapacity(context->common.channelmix, size / 2))
762 return FALSE;
763
764 /* Simply drop second channel.
765 * TODO: Calculate average */
766 for (size_t x = 0; x < samples; x++)
767 {
768 for (size_t y = 0; y < bpp; y++)
769 Stream_Write_UINT8(context->common.channelmix, src[2 * x * bpp + y]);
770 }
771
772 Stream_SealLength(context->common.channelmix);
773 *data = Stream_Buffer(context->common.channelmix);
774 *length = Stream_Length(context->common.channelmix);
775 dstFormat->nChannels = 1;
776 return TRUE;
777
778 case 1: /* Invalid, do we want to use a 0 channel sound? */
779 default: /* Unsupported number of channels */
780 WLog_WARN(TAG, "[%s] unsupported channel count %" PRIu16, __func__,
781 srcFormat->nChannels);
782 return FALSE;
783 }
784
785 return FALSE;
786}
787
788BOOL freerdp_dsp_ffmpeg_encode(FREERDP_DSP_CONTEXT* WINPR_RESTRICT context,
789 const AUDIO_FORMAT* WINPR_RESTRICT format,
790 const BYTE* WINPR_RESTRICT sdata, size_t length,
791 wStream* WINPR_RESTRICT out)
792{
793 AUDIO_FORMAT fmt = { 0 };
794
795 if (!context || !format || !sdata || !out || !context->common.encoder)
796 return FALSE;
797
798 if (!context || !sdata || !out)
799 return FALSE;
800
801 /* https://github.com/FreeRDP/FreeRDP/issues/7607
802 *
803 * we get noisy data with channel transformation, so do it ourselves.
804 */
805 const BYTE* data = NULL;
806 if (!freerdp_dsp_channel_mix(context, sdata, length, format, &data, &length, &fmt))
807 return FALSE;
808
809 /* Create input frame */
810 if (!ffmpeg_fill_frame(context->frame, format, data, length))
811 return FALSE;
812
813 /* Resample to desired format. */
814 if (!ffmpeg_resample_frame(context->rcontext, context->frame, context->resampled))
815 return FALSE;
816
817 if (context->context->frame_size <= 0)
818 {
819 return ffmpeg_encode_frame(context->context, context->resampled, context->packet, out);
820 }
821 else
822 {
823 int copied = 0;
824 int rest = context->resampled->nb_samples;
825
826 do
827 {
828 int inSamples = rest;
829
830 if ((inSamples < 0) || (context->bufferedSamples > (UINT32)(INT_MAX - inSamples)))
831 return FALSE;
832
833 if (inSamples + (int)context->bufferedSamples > context->context->frame_size)
834 inSamples = context->context->frame_size - (int)context->bufferedSamples;
835
836#if LIBAVUTIL_VERSION_INT >= AV_VERSION_INT(57, 28, 100)
837 const int nrchannels = context->context->ch_layout.nb_channels;
838#else
839 const int nrchannels = context->context->channels;
840#endif
841 const int rc =
842 av_samples_copy(context->buffered->extended_data, context->resampled->extended_data,
843 (int)context->bufferedSamples, copied, inSamples, nrchannels,
844 context->context->sample_fmt);
845 if (rc < 0)
846 return FALSE;
847 rest -= inSamples;
848 copied += inSamples;
849 context->bufferedSamples += (UINT32)inSamples;
850
851 if (context->context->frame_size <= (int)context->bufferedSamples)
852 {
853 /* Encode in desired format. */
854 if (!ffmpeg_encode_frame(context->context, context->buffered, context->packet, out))
855 return FALSE;
856
857 context->bufferedSamples = 0;
858 }
859 } while (rest > 0);
860
861 return TRUE;
862 }
863}
864
865BOOL freerdp_dsp_ffmpeg_decode(FREERDP_DSP_CONTEXT* WINPR_RESTRICT context,
866 const AUDIO_FORMAT* WINPR_RESTRICT srcFormat,
867 const BYTE* WINPR_RESTRICT data, size_t length,
868 wStream* WINPR_RESTRICT out)
869{
870 if (!context || !srcFormat || !data || !out || context->common.encoder)
871 return FALSE;
872
873#if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(58, 133, 100)
874 av_init_packet(context->packet);
875#endif
876 context->packet->data = WINPR_CAST_CONST_PTR_AWAY(data, uint8_t*);
877
878 WINPR_ASSERT(length <= INT_MAX);
879 context->packet->size = (int)length;
880 return ffmpeg_decode(context->context, context->packet, context->frame, context->rcontext,
881 context->resampled, out);
882}