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         quicksync_encoder->shutdown();
 118         x264_encoder.reset(nullptr);
 119         x264_disk_encoder.reset(nullptr);
 120         quicksync_encoder->close_file();
 121         quicksync_encoder.reset(nullptr);
 122         while (quicksync_encoders_in_shutdown.load() > 0) {
 123                 usleep(10000);
 124         }
 125 }
 126
 127 void VideoEncoder::do_cut(int frame)
 128 {
 129         string filename = generate_local_dump_filename(frame);
 130         printf("Starting new recording: %s\n", filename.c_str());
 131
 132         // Do the shutdown of the old encoder in a separate thread, since it can
 133         // take some time (it needs to wait for all the frames in the queue to be
 134         // done encoding, for one) and we are running on the main mixer thread.
 135         // However, since this means both encoders could be sending packets at
 136         // the same time, it means pts could come out of order to the stream mux,
 137         // and we need to plug it until the shutdown is complete.
 138         http_mux->plug();
 139         lock(qs_mu, qs_audio_mu);
 140         lock_guard<mutex> lock1(qs_mu, adopt_lock), lock2(qs_audio_mu, adopt_lock);
 141         QuickSyncEncoder *old_encoder = quicksync_encoder.release();  // When we go C++14, we can use move capture instead.
 142         X264Encoder *old_x264_encoder = nullptr;
 143         X264Encoder *old_x264_disk_encoder = nullptr;
 144         if (global_flags.x264_video_to_disk) {
 145                 old_x264_encoder = x264_encoder.release();
 146         }
 147         if (global_flags.x264_separate_disk_encode) {
 148                 old_x264_disk_encoder = x264_disk_encoder.release();
 149         }
 150         thread([old_encoder, old_x264_encoder, old_x264_disk_encoder, this]{
 151                 old_encoder->shutdown();
 152                 delete old_x264_encoder;
 153                 delete old_x264_disk_encoder;
 154                 old_encoder->close_file();
 155                 http_mux->unplug();
 156
 157                 // We cannot delete the encoder here, as this thread has no OpenGL context.
 158                 // We'll deal with it in begin_frame().
 159                 lock_guard<mutex> lock(qs_mu);
 160                 qs_needing_cleanup.emplace_back(old_encoder);
 161         }).detach();
 162
 163         if (global_flags.x264_video_to_disk) {
 164                 x264_encoder.reset(new X264Encoder(oformat, /*use_separate_disk_params=*/false));
 165                 assert(global_flags.x264_video_to_http);
 166                 if (global_flags.x264_video_to_http) {
 167                         x264_encoder->add_mux(http_mux.get());
 168                 }
 169                 if (overriding_bitrate != 0) {
 170                         x264_encoder->change_bitrate(overriding_bitrate);
 171                 }
 172         }
 173         X264Encoder *http_encoder = x264_encoder.get();
 174         X264Encoder *disk_encoder = x264_encoder.get();
 175         if (global_flags.x264_separate_disk_encode) {
 176                 x264_disk_encoder.reset(new X264Encoder(oformat, /*use_separate_disk_params=*/true));
 177                 disk_encoder = x264_disk_encoder.get();
 178         }
 179
 180         quicksync_encoder.reset(new QuickSyncEncoder(filename, resource_pool, surface, va_display, width, height, oformat, http_encoder, disk_encoder, disk_space_estimator));
 181         quicksync_encoder->set_http_mux(http_mux.get());
 182 }
 183
 184 void VideoEncoder::change_x264_bitrate(unsigned rate_kbit)
 185 {
 186         overriding_bitrate = rate_kbit;
 187         x264_encoder->change_bitrate(rate_kbit);
 188 }
 189
 190 void VideoEncoder::add_audio(int64_t pts, std::vector<float> audio)
 191 {
 192         // Take only qs_audio_mu, since add_audio() is thread safe
 193         // (we can only conflict with do_cut(), which takes qs_audio_mu)
 194         // and we don't want to contend with begin_frame().
 195         {
 196                 lock_guard<mutex> lock(qs_audio_mu);
 197                 quicksync_encoder->add_audio(pts, audio);
 198         }
 199         stream_audio_encoder->encode_audio(audio, pts + quicksync_encoder->global_delay());
 200 }
 201
 202 bool VideoEncoder::is_zerocopy() const
 203 {
 204         // Explicitly do _not_ take qs_mu; this is called from the mixer,
 205         // and qs_mu might be contended. is_zerocopy() is thread safe
 206         // and never called in parallel with do_cut() (both happen only
 207         // from the mixer thread).
 208         return quicksync_encoder->is_zerocopy();
 209 }
 210
 211 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)
 212 {
 213         lock_guard<mutex> lock(qs_mu);
 214         qs_needing_cleanup.clear();  // Since we have an OpenGL context here, and are called regularly.
 215         return quicksync_encoder->begin_frame(pts, duration, ycbcr_coefficients, input_frames, y_tex, cbcr_tex);
 216 }
 217
 218 RefCountedGLsync VideoEncoder::end_frame()
 219 {
 220         want_srt_metric_update = true;
 221         lock_guard<mutex> lock(qs_mu);
 222         return quicksync_encoder->end_frame();
 223 }
 224
 225 void VideoEncoder::open_output_streams()
 226 {
 227         for (bool is_srt : {false, true}) {
 228                 if (is_srt && global_flags.srt_destination_host.empty()) {
 229                         continue;
 230                 }
 231
 232                 AVFormatContext *avctx = avformat_alloc_context();
 233                 avctx->oformat = is_srt ? srt_oformat : oformat;
 234
 235                 uint8_t *buf = (uint8_t *)av_malloc(MUX_BUFFER_SIZE);
 236                 avctx->pb = avio_alloc_context(buf, MUX_BUFFER_SIZE, 1, this, nullptr, nullptr, nullptr);
 237                 if (is_srt) {
 238                         avctx->pb->write_packet = &VideoEncoder::write_srt_packet_thunk;
 239                 } else {
 240                         avctx->pb->write_data_type = &VideoEncoder::write_packet2_thunk;
 241                         avctx->pb->ignore_boundary_point = 1;
 242                 }
 243
 244                 Mux::Codec video_codec;
 245                 if (global_flags.av1_video_to_http) {
 246                         video_codec = Mux::CODEC_AV1;
 247                 } else {
 248                         video_codec = Mux::CODEC_H264;
 249                 }
 250
 251                 avctx->flags = AVFMT_FLAG_CUSTOM_IO;
 252
 253                 string video_extradata;
 254                 if (global_flags.x264_video_to_http) {
 255                         video_extradata = x264_encoder->get_global_headers();
 256 #ifdef HAVE_AV1
 257                 } else if (global_flags.av1_video_to_http) {
 258                         video_extradata = av1_encoder->get_global_headers();
 259 #endif
 260                 }
 261
 262                 Mux *mux = new Mux(avctx, width, height, video_codec, video_extradata, stream_audio_encoder->get_codec_parameters().get(),
 263                         get_color_space(global_flags.ycbcr_rec709_coefficients), COARSE_TIMEBASE,
 264                         /*write_callback=*/nullptr, is_srt ? Mux::WRITE_BACKGROUND : Mux::WRITE_FOREGROUND, { is_srt ? &srt_mux_metrics : &http_mux_metrics });
 265                 if (is_srt) {
 266                         srt_mux.reset(mux);
 267                         srt_mux_metrics.init({{ "destination", "srt" }});
 268                         srt_metrics.init({{ "cardtype", "output" }});
 269                         global_metrics.add("srt_num_connection_attempts", {{ "cardtype", "output" }}, &metric_srt_num_connection_attempts);
 270                 } else {
 271                         http_mux.reset(mux);
 272                         http_mux_metrics.init({{ "destination", "http" }});
 273                 }
 274         }
 275 }
 276
 277 int VideoEncoder::write_packet2_thunk(void *opaque, uint8_t *buf, int buf_size, AVIODataMarkerType type, int64_t time)
 278 {
 279         VideoEncoder *video_encoder = (VideoEncoder *)opaque;
 280         return video_encoder->write_packet2(buf, buf_size, type, time);
 281 }
 282
 283 int VideoEncoder::write_packet2(uint8_t *buf, int buf_size, AVIODataMarkerType type, int64_t time)
 284 {
 285         if (type == AVIO_DATA_MARKER_SYNC_POINT || type == AVIO_DATA_MARKER_BOUNDARY_POINT) {
 286                 seen_sync_markers = true;
 287         } else if (type == AVIO_DATA_MARKER_UNKNOWN && !seen_sync_markers) {
 288                 // We don't know if this is a keyframe or not (the muxer could
 289                 // avoid marking it), so we just have to make the best of it.
 290                 type = AVIO_DATA_MARKER_SYNC_POINT;
 291         }
 292
 293         if (type == AVIO_DATA_MARKER_HEADER) {
 294                 http_mux_header.append((char *)buf, buf_size);
 295                 httpd->set_header(HTTPD::StreamID{ HTTPD::MAIN_STREAM, 0 }, http_mux_header);
 296         } else {
 297                 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 });
 298         }
 299         return buf_size;
 300 }
 301
 302 int VideoEncoder::write_srt_packet_thunk(void *opaque, uint8_t *buf, int buf_size)
 303 {
 304         VideoEncoder *video_encoder = (VideoEncoder *)opaque;
 305         return video_encoder->write_srt_packet(buf, buf_size);
 306 }
 307
 308 static string print_addrinfo(const addrinfo *ai)
 309 {
 310         char hoststr[NI_MAXHOST], portstr[NI_MAXSERV];
 311         if (getnameinfo(ai->ai_addr, ai->ai_addrlen, hoststr, sizeof(hoststr), portstr, sizeof(portstr), NI_DGRAM | NI_NUMERICHOST | NI_NUMERICSERV) != 0) {
 312                 return "<unknown address>";  // Should basically never happen, since we're not doing DNS lookups.
 313         }
 314
 315         if (ai->ai_family == AF_INET6) {
 316                 return string("[") + hoststr + "]:" + portstr;
 317         } else {
 318                 return string(hoststr) + ":" + portstr;
 319         }
 320 }
 321
 322 int VideoEncoder::open_srt_socket()
 323 {
 324         int sock = srt_create_socket();
 325         if (sock == -1) {
 326                 fprintf(stderr, "srt_create_socket(): %s\n", srt_getlasterror_str());
 327                 return -1;
 328         }
 329
 330         SRT_TRANSTYPE live = SRTT_LIVE;
 331         if (srt_setsockopt(sock, 0, SRTO_TRANSTYPE, &live, sizeof(live)) < 0) {
 332                 fprintf(stderr, "srt_setsockopt(SRTO_TRANSTYPE): %s\n", srt_getlasterror_str());
 333                 srt_close(sock);
 334                 return -1;
 335         }
 336
 337         if (srt_setsockopt(sock, 0, SRTO_LATENCY, &global_flags.srt_output_latency, sizeof(global_flags.srt_output_latency)) < 0) {
 338                 fprintf(stderr, "srt_setsockopt(SRTO_LATENCY): %s\n", srt_getlasterror_str());
 339                 srt_close(sock);
 340                 return -1;
 341         }
 342
 343         if (!global_flags.srt_streamid.empty()) {
 344                 if (srt_setsockopt(sock, 0, SRTO_STREAMID, global_flags.srt_streamid.data(), global_flags.srt_streamid.size()) < 0) {
 345                         fprintf(stderr, "srt_setsockopt(SRTO_STREAMID): %s\n", srt_getlasterror_str());
 346                         srt_close(sock);
 347                         return -1;
 348                 }
 349         }
 350
 351         if (!global_flags.srt_passphrase.empty()) {
 352                 if (srt_setsockopt(sock, 0, SRTO_PASSPHRASE, global_flags.srt_passphrase.data(), global_flags.srt_passphrase.size()) < 0) {
 353                         fprintf(stderr, "srt_setsockopt(SRTO_PASSPHRASE): %s\n", srt_getlasterror_str());
 354                         srt_close(sock);
 355                         return -1;
 356                 }
 357         }
 358
 359         return sock;
 360 }
 361
 362 int VideoEncoder::connect_to_srt()
 363 {
 364         // We need to specify SOCK_DGRAM as a hint, or we'll get all addresses
 365         // three times (for each of TCP, UDP, raw).
 366         addrinfo hints;
 367         memset(&hints, 0, sizeof(hints));
 368         hints.ai_flags = AI_ADDRCONFIG;
 369         hints.ai_socktype = SOCK_DGRAM;
 370
 371         addrinfo *ai;
 372         int ret = getaddrinfo(global_flags.srt_destination_host.c_str(), global_flags.srt_destination_port.c_str(), &hints, &ai);
 373         if (ret != 0) {
 374                 fprintf(stderr, "getaddrinfo(%s:%s): %s\n", global_flags.srt_destination_host.c_str(), global_flags.srt_destination_port.c_str(), gai_strerror(ret));
 375                 return -1;
 376         }
 377
 378         for (const addrinfo *cur = ai; cur != nullptr; cur = cur->ai_next) {
 379                 // Seemingly, srt_create_socket() isn't universal; once we try to connect,
 380                 // it gets locked to either IPv4 or IPv6. So we need to create a new one
 381                 // for every address we try.
 382                 int sock = open_srt_socket();
 383                 if (sock == -1) {
 384                         // Die immediately.
 385                         return sock;
 386                 }
 387                 ++metric_srt_num_connection_attempts;
 388                 if (srt_connect(sock, cur->ai_addr, cur->ai_addrlen) < 0) {
 389                         fprintf(stderr, "srt_connect(%s): %s\n", print_addrinfo(cur).c_str(), srt_getlasterror_str());
 390                         srt_close(sock);
 391                         continue;
 392                 }
 393                 fprintf(stderr, "Connected to destination SRT endpoint at %s.\n", print_addrinfo(cur).c_str());
 394                 freeaddrinfo(ai);
 395                 return sock;
 396         }
 397
 398         // Out of candidates, so give up.
 399         freeaddrinfo(ai);
 400         return -1;
 401 }
 402
 403 int VideoEncoder::write_srt_packet(uint8_t *buf, int buf_size)
 404 {
 405         if (want_srt_metric_update.exchange(false) && srt_sock != -1) {
 406                 srt_metrics.update_srt_stats(srt_sock);
 407         }
 408         while (buf_size > 0) {
 409                 if (srt_sock == -1) {
 410                         srt_sock = connect_to_srt();
 411                         if (srt_sock == -1) {
 412                                 usleep(100000);
 413                                 continue;
 414                         }
 415                         srt_metrics.update_srt_stats(srt_sock);
 416                 }
 417                 int to_send = min(buf_size, SRT_LIVE_DEF_PLSIZE);
 418                 int ret = srt_send(srt_sock, (char *)buf, to_send);
 419                 if (ret < 0)  {
 420                         fprintf(stderr, "srt_send(): %s\n", srt_getlasterror_str());
 421                         srt_close(srt_sock);
 422                         srt_metrics.metric_srt_uptime_seconds = 0.0 / 0.0;
 423                         srt_sock = connect_to_srt();
 424                         continue;
 425                 }
 426                 buf += ret;
 427                 buf_size -= ret;
 428         }
 429         return buf_size;
 430 }
 431