git.sesse.net Git - nageru/blob - nageru/video_encoder.cpp

   1 #include "video_encoder.h"
   2
   3 #include <assert.h>
   4 #include <stdio.h>
   5 #include <time.h>
   6 #include <unistd.h>
   7 #include <sys/types.h>
   8 #include <sys/socket.h>
   9 #include <netdb.h>
  10 #include <string>
  11 #include <thread>
  12
  13 extern "C" {
  14 #include <libavutil/mem.h>
  15 }
  16
  17 #include "audio_encoder.h"
  18 #ifdef HAVE_AV1
  19 #include "av1_encoder.h"
  20 #endif
  21 #include "defs.h"
  22 #include "shared/ffmpeg_raii.h"
  23 #include "flags.h"
  24 #include "shared/httpd.h"
  25 #include "shared/mux.h"
  26 #include "quicksync_encoder.h"
  27 #include "shared/timebase.h"
  28 #include "x264_encoder.h"
  29
  30 class RefCountedFrame;
  31
  32 using namespace std;
  33 using namespace movit;
  34
  35 namespace {
  36
  37 string generate_local_dump_filename(int frame)
  38 {
  39         time_t now = time(NULL);
  40         tm now_tm;
  41         localtime_r(&now, &now_tm);
  42
  43         char timestamp[64];
  44         strftime(timestamp, sizeof(timestamp), "%F-%H%M%S%z", &now_tm);
  45
  46         // Use the frame number to disambiguate between two cuts starting
  47         // on the same second.
  48         char filename[256];
  49         snprintf(filename, sizeof(filename), "%s/%s%s-f%02d%s",
  50                 global_flags.recording_dir.c_str(),
  51                 LOCAL_DUMP_PREFIX, timestamp, frame % 100, LOCAL_DUMP_SUFFIX);
  52         return filename;
  53 }
  54
  55 }  // namespace
  56
  57 VideoEncoder::VideoEncoder(ResourcePool *resource_pool, QSurface *surface, const std::string &va_display, int width, int height, HTTPD *httpd, DiskSpaceEstimator *disk_space_estimator)
  58         : resource_pool(resource_pool), surface(surface), va_display(va_display), width(width), height(height), httpd(httpd), disk_space_estimator(disk_space_estimator)
  59 {
  60         // TODO: If we're outputting AV1, we can't use MPEG-TS currently.
  61         srt_oformat = av_guess_format("mpegts", nullptr, nullptr);
  62         assert(srt_oformat != nullptr);
  63
  64         oformat = av_guess_format(global_flags.stream_mux_name.c_str(), nullptr, nullptr);
  65         assert(oformat != nullptr);
  66         if (global_flags.stream_audio_codec_name.empty()) {
  67                 stream_audio_encoder.reset(new AudioEncoder(AUDIO_OUTPUT_CODEC_NAME, DEFAULT_AUDIO_OUTPUT_BIT_RATE, oformat));
  68         } else {
  69                 stream_audio_encoder.reset(new AudioEncoder(global_flags.stream_audio_codec_name, global_flags.stream_audio_codec_bitrate, oformat));
  70         }
  71         if (global_flags.x264_video_to_http || global_flags.x264_video_to_disk) {
  72                 x264_encoder.reset(new X264Encoder(oformat, /*use_separate_disk_params=*/false));
  73         }
  74         VideoCodecInterface *http_encoder = x264_encoder.get();
  75         VideoCodecInterface *disk_encoder = x264_encoder.get();
  76 #ifdef HAVE_AV1
  77         if (global_flags.av1_video_to_http) {
  78                 av1_encoder.reset(new AV1Encoder(oformat));
  79                 http_encoder = av1_encoder.get();
  80         }
  81 #endif
  82         if (global_flags.x264_separate_disk_encode) {
  83                 x264_disk_encoder.reset(new X264Encoder(oformat, /*use_separate_disk_params=*/true));
  84                 disk_encoder = x264_disk_encoder.get();
  85         }
  86
  87         string filename = generate_local_dump_filename(/*frame=*/0);
  88         quicksync_encoder.reset(new QuickSyncEncoder(filename, resource_pool, surface, va_display, width, height, oformat, http_encoder, disk_encoder, disk_space_estimator));
  89
  90         open_output_streams();
  91         stream_audio_encoder->add_mux(http_mux.get());
  92         if (srt_mux != nullptr) {
  93                 stream_audio_encoder->add_mux(srt_mux.get());
  94         }
  95         quicksync_encoder->set_http_mux(http_mux.get());
  96         if (srt_mux != nullptr) {
  97                 quicksync_encoder->set_srt_mux(srt_mux.get());
  98         }
  99         if (global_flags.x264_video_to_http) {
 100                 x264_encoder->add_mux(http_mux.get());
 101                 if (srt_mux != nullptr) {
 102                         x264_encoder->add_mux(srt_mux.get());
 103                 }
 104         }
 105 #ifdef HAVE_AV1
 106         if (global_flags.av1_video_to_http) {
 107                 av1_encoder->add_mux(http_mux.get());
 108                 if (srt_mux != nullptr) {
 109                         av1_encoder->add_mux(srt_mux.get());
 110                 }
 111         }
 112 #endif
 113 }
 114
 115 VideoEncoder::~VideoEncoder()
 116 {
 117         should_quit = true;
 118         quicksync_encoder->shutdown();
 119         x264_encoder.reset(nullptr);
 120         x264_disk_encoder.reset(nullptr);
 121         quicksync_encoder->close_file();
 122         quicksync_encoder.reset(nullptr);
 123         while (quicksync_encoders_in_shutdown.load() > 0) {
 124                 usleep(10000);
 125         }
 126 }
 127
 128 void VideoEncoder::do_cut(int frame)
 129 {
 130         string filename = generate_local_dump_filename(frame);
 131         printf("Starting new recording: %s\n", filename.c_str());
 132
 133         // Do the shutdown of the old encoder in a separate thread, since it can
 134         // take some time (it needs to wait for all the frames in the queue to be
 135         // done encoding, for one) and we are running on the main mixer thread.
 136         // However, since this means both encoders could be sending packets at
 137         // the same time, it means pts could come out of order to the stream mux,
 138         // and we need to plug it until the shutdown is complete.
 139         http_mux->plug();
 140         lock(qs_mu, qs_audio_mu);
 141         lock_guard<mutex> lock1(qs_mu, adopt_lock), lock2(qs_audio_mu, adopt_lock);
 142         QuickSyncEncoder *old_encoder = quicksync_encoder.release();  // When we go C++14, we can use move capture instead.
 143         X264Encoder *old_x264_encoder = nullptr;
 144         X264Encoder *old_x264_disk_encoder = nullptr;
 145         if (global_flags.x264_video_to_disk) {
 146                 old_x264_encoder = x264_encoder.release();
 147         }
 148         if (global_flags.x264_separate_disk_encode) {
 149                 old_x264_disk_encoder = x264_disk_encoder.release();
 150         }
 151         thread([old_encoder, old_x264_encoder, old_x264_disk_encoder, this]{
 152                 old_encoder->shutdown();
 153                 delete old_x264_encoder;
 154                 delete old_x264_disk_encoder;
 155                 old_encoder->close_file();
 156                 http_mux->unplug();
 157
 158                 // We cannot delete the encoder here, as this thread has no OpenGL context.
 159                 // We'll deal with it in begin_frame().
 160                 lock_guard<mutex> lock(qs_mu);
 161                 qs_needing_cleanup.emplace_back(old_encoder);
 162         }).detach();
 163
 164         if (global_flags.x264_video_to_disk) {
 165                 x264_encoder.reset(new X264Encoder(oformat, /*use_separate_disk_params=*/false));
 166                 assert(global_flags.x264_video_to_http);
 167                 if (global_flags.x264_video_to_http) {
 168                         x264_encoder->add_mux(http_mux.get());
 169                 }
 170                 if (overriding_bitrate != 0) {
 171                         x264_encoder->change_bitrate(overriding_bitrate);
 172                 }
 173         }
 174         X264Encoder *http_encoder = x264_encoder.get();
 175         X264Encoder *disk_encoder = x264_encoder.get();
 176         if (global_flags.x264_separate_disk_encode) {
 177                 x264_disk_encoder.reset(new X264Encoder(oformat, /*use_separate_disk_params=*/true));
 178                 disk_encoder = x264_disk_encoder.get();
 179         }
 180
 181         quicksync_encoder.reset(new QuickSyncEncoder(filename, resource_pool, surface, va_display, width, height, oformat, http_encoder, disk_encoder, disk_space_estimator));
 182         quicksync_encoder->set_http_mux(http_mux.get());
 183 }
 184
 185 void VideoEncoder::change_x264_bitrate(unsigned rate_kbit)
 186 {
 187         overriding_bitrate = rate_kbit;
 188         x264_encoder->change_bitrate(rate_kbit);
 189 }
 190
 191 void VideoEncoder::add_audio(int64_t pts, std::vector<float> audio)
 192 {
 193         // Take only qs_audio_mu, since add_audio() is thread safe
 194         // (we can only conflict with do_cut(), which takes qs_audio_mu)
 195         // and we don't want to contend with begin_frame().
 196         {
 197                 lock_guard<mutex> lock(qs_audio_mu);
 198                 quicksync_encoder->add_audio(pts, audio);
 199         }
 200         stream_audio_encoder->encode_audio(audio, pts + quicksync_encoder->global_delay());
 201 }
 202
 203 bool VideoEncoder::is_zerocopy() const
 204 {
 205         // Explicitly do _not_ take qs_mu; this is called from the mixer,
 206         // and qs_mu might be contended. is_zerocopy() is thread safe
 207         // and never called in parallel with do_cut() (both happen only
 208         // from the mixer thread).
 209         return quicksync_encoder->is_zerocopy();
 210 }
 211
 212 bool VideoEncoder::begin_frame(int64_t pts, int64_t duration, movit::YCbCrLumaCoefficients ycbcr_coefficients, const std::vector<RefCountedFrame> &input_frames, GLuint *y_tex, GLuint *cbcr_tex)
 213 {
 214         lock_guard<mutex> lock(qs_mu);
 215         qs_needing_cleanup.clear();  // Since we have an OpenGL context here, and are called regularly.
 216         return quicksync_encoder->begin_frame(pts, duration, ycbcr_coefficients, input_frames, y_tex, cbcr_tex);
 217 }
 218
 219 RefCountedGLsync VideoEncoder::end_frame()
 220 {
 221         want_srt_metric_update = true;
 222         lock_guard<mutex> lock(qs_mu);
 223         return quicksync_encoder->end_frame();
 224 }
 225
 226 void VideoEncoder::open_output_streams()
 227 {
 228         for (bool is_srt : {false, true}) {
 229                 if (is_srt && global_flags.srt_destination_host.empty()) {
 230                         continue;
 231                 }
 232
 233                 AVFormatContext *avctx = avformat_alloc_context();
 234                 avctx->oformat = is_srt ? srt_oformat : oformat;
 235
 236                 uint8_t *buf = (uint8_t *)av_malloc(MUX_BUFFER_SIZE);
 237                 avctx->pb = avio_alloc_context(buf, MUX_BUFFER_SIZE, 1, this, nullptr, nullptr, nullptr);
 238                 if (is_srt) {
 239                         avctx->pb->write_packet = &VideoEncoder::write_srt_packet_thunk;
 240                 } else {
 241                         avctx->pb->write_data_type = &VideoEncoder::write_packet2_thunk;
 242                         avctx->pb->ignore_boundary_point = 1;
 243                 }
 244
 245                 Mux::Codec video_codec;
 246                 if (global_flags.av1_video_to_http) {
 247                         video_codec = Mux::CODEC_AV1;
 248                 } else {
 249                         video_codec = Mux::CODEC_H264;
 250                 }
 251
 252                 avctx->flags = AVFMT_FLAG_CUSTOM_IO;
 253
 254                 string video_extradata;
 255                 if (global_flags.x264_video_to_http) {
 256                         video_extradata = x264_encoder->get_global_headers();
 257 #ifdef HAVE_AV1
 258                 } else if (global_flags.av1_video_to_http) {
 259                         video_extradata = av1_encoder->get_global_headers();
 260 #endif
 261                 }
 262
 263                 Mux *mux = new Mux(avctx, width, height, video_codec, video_extradata, stream_audio_encoder->get_codec_parameters().get(),
 264                         get_color_space(global_flags.ycbcr_rec709_coefficients), COARSE_TIMEBASE,
 265                         /*write_callback=*/nullptr, is_srt ? Mux::WRITE_BACKGROUND : Mux::WRITE_FOREGROUND, { is_srt ? &srt_mux_metrics : &http_mux_metrics });
 266                 if (is_srt) {
 267                         srt_mux.reset(mux);
 268                         srt_mux_metrics.init({{ "destination", "srt" }});
 269                         srt_metrics.init({{ "cardtype", "output" }});
 270                         global_metrics.add("srt_num_connection_attempts", {{ "cardtype", "output" }}, &metric_srt_num_connection_attempts);
 271                 } else {
 272                         http_mux.reset(mux);
 273                         http_mux_metrics.init({{ "destination", "http" }});
 274                 }
 275         }
 276 }
 277
 278 int VideoEncoder::write_packet2_thunk(void *opaque, uint8_t *buf, int buf_size, AVIODataMarkerType type, int64_t time)
 279 {
 280         VideoEncoder *video_encoder = (VideoEncoder *)opaque;
 281         return video_encoder->write_packet2(buf, buf_size, type, time);
 282 }
 283
 284 int VideoEncoder::write_packet2(uint8_t *buf, int buf_size, AVIODataMarkerType type, int64_t time)
 285 {
 286         if (type == AVIO_DATA_MARKER_SYNC_POINT || type == AVIO_DATA_MARKER_BOUNDARY_POINT) {
 287                 seen_sync_markers = true;
 288         } else if (type == AVIO_DATA_MARKER_UNKNOWN && !seen_sync_markers) {
 289                 // We don't know if this is a keyframe or not (the muxer could
 290                 // avoid marking it), so we just have to make the best of it.
 291                 type = AVIO_DATA_MARKER_SYNC_POINT;
 292         }
 293
 294         if (type == AVIO_DATA_MARKER_HEADER) {
 295                 http_mux_header.append((char *)buf, buf_size);
 296                 httpd->set_header(HTTPD::StreamID{ HTTPD::MAIN_STREAM, 0 }, http_mux_header);
 297         } else {
 298                 httpd->add_data(HTTPD::StreamID{ HTTPD::MAIN_STREAM, 0 }, (char *)buf, buf_size, type == AVIO_DATA_MARKER_SYNC_POINT, time, AVRational{ AV_TIME_BASE, 1 });
 299         }
 300         return buf_size;
 301 }
 302
 303 int VideoEncoder::write_srt_packet_thunk(void *opaque, uint8_t *buf, int buf_size)
 304 {
 305         VideoEncoder *video_encoder = (VideoEncoder *)opaque;
 306         return video_encoder->write_srt_packet(buf, buf_size);
 307 }
 308
 309 static string print_addrinfo(const addrinfo *ai)
 310 {
 311         char hoststr[NI_MAXHOST], portstr[NI_MAXSERV];
 312         if (getnameinfo(ai->ai_addr, ai->ai_addrlen, hoststr, sizeof(hoststr), portstr, sizeof(portstr), NI_DGRAM | NI_NUMERICHOST | NI_NUMERICSERV) != 0) {
 313                 return "<unknown address>";  // Should basically never happen, since we're not doing DNS lookups.
 314         }
 315
 316         if (ai->ai_family == AF_INET6) {
 317                 return string("[") + hoststr + "]:" + portstr;
 318         } else {
 319                 return string(hoststr) + ":" + portstr;
 320         }
 321 }
 322
 323 int VideoEncoder::open_srt_socket()
 324 {
 325         int sock = srt_create_socket();
 326         if (sock == -1) {
 327                 fprintf(stderr, "srt_create_socket(): %s\n", srt_getlasterror_str());
 328                 return -1;
 329         }
 330
 331         SRT_TRANSTYPE live = SRTT_LIVE;
 332         if (srt_setsockopt(sock, 0, SRTO_TRANSTYPE, &live, sizeof(live)) < 0) {
 333                 fprintf(stderr, "srt_setsockopt(SRTO_TRANSTYPE): %s\n", srt_getlasterror_str());
 334                 srt_close(sock);
 335                 return -1;
 336         }
 337
 338         if (srt_setsockopt(sock, 0, SRTO_LATENCY, &global_flags.srt_output_latency_ms, sizeof(global_flags.srt_output_latency_ms)) < 0) {
 339                 fprintf(stderr, "srt_setsockopt(SRTO_LATENCY): %s\n", srt_getlasterror_str());
 340                 srt_close(sock);
 341                 return -1;
 342         }
 343
 344         if (!global_flags.srt_streamid.empty()) {
 345                 if (srt_setsockopt(sock, 0, SRTO_STREAMID, global_flags.srt_streamid.data(), global_flags.srt_streamid.size()) < 0) {
 346                         fprintf(stderr, "srt_setsockopt(SRTO_STREAMID): %s\n", srt_getlasterror_str());
 347                         srt_close(sock);
 348                         return -1;
 349                 }
 350         }
 351
 352         if (!global_flags.srt_passphrase.empty()) {
 353                 if (srt_setsockopt(sock, 0, SRTO_PASSPHRASE, global_flags.srt_passphrase.data(), global_flags.srt_passphrase.size()) < 0) {
 354                         fprintf(stderr, "srt_setsockopt(SRTO_PASSPHRASE): %s\n", srt_getlasterror_str());
 355                         srt_close(sock);
 356                         return -1;
 357                 }
 358         }
 359
 360         return sock;
 361 }
 362
 363 int VideoEncoder::connect_to_srt()
 364 {
 365         // We need to specify SOCK_DGRAM as a hint, or we'll get all addresses
 366         // three times (for each of TCP, UDP, raw).
 367         addrinfo hints;
 368         memset(&hints, 0, sizeof(hints));
 369         hints.ai_flags = AI_ADDRCONFIG;
 370         hints.ai_socktype = SOCK_DGRAM;
 371
 372         addrinfo *ai;
 373         int ret = getaddrinfo(global_flags.srt_destination_host.c_str(), global_flags.srt_destination_port.c_str(), &hints, &ai);
 374         if (ret != 0) {
 375                 fprintf(stderr, "getaddrinfo(%s:%s): %s\n", global_flags.srt_destination_host.c_str(), global_flags.srt_destination_port.c_str(), gai_strerror(ret));
 376                 return -1;
 377         }
 378
 379         unique_ptr<addrinfo, decltype(freeaddrinfo) *> ai_cleanup(ai, &freeaddrinfo);
 380
 381         for (const addrinfo *cur = ai; cur != nullptr; cur = cur->ai_next) {
 382                 // Seemingly, srt_create_socket() isn't universal; once we try to connect,
 383                 // it gets locked to either IPv4 or IPv6. So we need to create a new one
 384                 // for every address we try.
 385                 int sock = open_srt_socket();
 386                 if (sock == -1) {
 387                         // Die immediately.
 388                         return sock;
 389                 }
 390                 ++metric_srt_num_connection_attempts;
 391
 392                 // We do a non-blocking connect, so that we can check should_quit
 393                 // every now and then.
 394                 int blocking = 0;
 395                 if (srt_setsockopt(sock, 0, SRTO_RCVSYN, &blocking, sizeof(blocking)) < 0) {
 396                         fprintf(stderr, "srt_setsockopt(SRTO_SNDSYN=0): %s\n", srt_getlasterror_str());
 397                         srt_close(sock);
 398                         continue;
 399                 }
 400                 if (srt_connect(sock, cur->ai_addr, cur->ai_addrlen) < 0) {
 401                         fprintf(stderr, "srt_connect(%s): %s\n", print_addrinfo(cur).c_str(), srt_getlasterror_str());
 402                         srt_close(sock);
 403                         continue;
 404                 }
 405                 int eid = srt_epoll_create();
 406                 if (eid < 0) {
 407                         fprintf(stderr, "srt_epoll_create(): %s\n", srt_getlasterror_str());
 408                         srt_close(sock);
 409                         continue;
 410                 }
 411                 int modes = SRT_EPOLL_ERR | SRT_EPOLL_OUT;
 412                 if (srt_epoll_add_usock(eid, sock, &modes) < 0) {
 413                         fprintf(stderr, "srt_epoll_usock(): %s\n", srt_getlasterror_str());
 414                         srt_close(sock);
 415                         srt_epoll_release(eid);
 416                         continue;
 417                 }
 418                 bool ok;
 419                 while (!should_quit.load()) {
 420                         SRTSOCKET errfds[1], writefds[1];
 421                         int num_errfds = 1, num_writefds = 1;
 422                         int poll_time_ms = 100;
 423                         int ret = srt_epoll_wait(eid, errfds, &num_errfds, writefds, &num_writefds, poll_time_ms, 0, 0, 0, 0);
 424                         if (ret < 0) {
 425                                 if (srt_getlasterror(nullptr) == SRT_ETIMEOUT) {
 426                                         continue;
 427                                 } else {
 428                                         fprintf(stderr, "srt_epoll_wait(): %s\n", srt_getlasterror_str());
 429                                         srt_close(sock);
 430                                         srt_epoll_release(eid);
 431                                         return -1;
 432                                 }
 433                         } else if (ret > 0) {
 434                                 // The SRT epoll framework is pretty odd, but seemingly,
 435                                 // this is the way. Getting the same error code as srt_connect()
 436                                 // would normally return seems to be impossible, though.
 437                                 ok = (num_errfds == 0);
 438                                 break;
 439                                 fprintf(stderr, "num_errfds=%d num_writefds=%d last_err=%s\n", num_errfds, num_writefds, srt_getlasterror_str());
 440                                 break;
 441                         }
 442                 }
 443                 srt_epoll_release(eid);
 444                 if (should_quit.load()) {
 445                         srt_close(sock);
 446                         return -1;
 447                 }
 448                 if (ok) {
 449                         fprintf(stderr, "Connected to destination SRT endpoint at %s.\n", print_addrinfo(cur).c_str());
 450                         return sock;
 451                 } else {
 452                         fprintf(stderr, "srt_connect(%s): %s\n", print_addrinfo(cur).c_str(), srt_getlasterror_str());
 453                         srt_close(sock);
 454                 }
 455         }
 456
 457         // Out of candidates, so give up.
 458         return -1;
 459 }
 460
 461 int VideoEncoder::write_srt_packet(uint8_t *buf, int buf_size)
 462 {
 463         if (want_srt_metric_update.exchange(false) && srt_sock != -1) {
 464                 srt_metrics.update_srt_stats(srt_sock);
 465         }
 466         while (buf_size > 0 && !should_quit.load()) {
 467                 if (srt_sock == -1) {
 468                         srt_sock = connect_to_srt();
 469                         if (srt_sock == -1) {
 470                                 usleep(100000);
 471                                 continue;
 472                         }
 473                         srt_metrics.update_srt_stats(srt_sock);
 474                 }
 475                 int to_send = min(buf_size, SRT_LIVE_DEF_PLSIZE);
 476                 int ret = srt_send(srt_sock, (char *)buf, to_send);
 477                 if (ret < 0)  {
 478                         fprintf(stderr, "srt_send(): %s\n", srt_getlasterror_str());
 479                         srt_close(srt_sock);
 480                         srt_metrics.metric_srt_uptime_seconds = 0.0 / 0.0;
 481                         srt_sock = connect_to_srt();
 482                         continue;
 483                 }
 484                 buf += ret;
 485                 buf_size -= ret;
 486         }
 487         return buf_size;
 488 }
 489