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         stream_audio_encoder->add_mux(srt_mux.get());
  93         quicksync_encoder->set_http_mux(http_mux.get());
  94         quicksync_encoder->set_srt_mux(srt_mux.get());
  95         if (global_flags.x264_video_to_http) {
  96                 x264_encoder->add_mux(http_mux.get());
  97                 x264_encoder->add_mux(srt_mux.get());
  98         }
  99 #ifdef HAVE_AV1
 100         if (global_flags.av1_video_to_http) {
 101                 av1_encoder->add_mux(http_mux.get());
 102                 av1_encoder->add_mux(srt_mux.get());
 103         }
 104 #endif
 105 }
 106
 107 VideoEncoder::~VideoEncoder()
 108 {
 109         quicksync_encoder->shutdown();
 110         x264_encoder.reset(nullptr);
 111         x264_disk_encoder.reset(nullptr);
 112         quicksync_encoder->close_file();
 113         quicksync_encoder.reset(nullptr);
 114         while (quicksync_encoders_in_shutdown.load() > 0) {
 115                 usleep(10000);
 116         }
 117 }
 118
 119 void VideoEncoder::do_cut(int frame)
 120 {
 121         string filename = generate_local_dump_filename(frame);
 122         printf("Starting new recording: %s\n", filename.c_str());
 123
 124         // Do the shutdown of the old encoder in a separate thread, since it can
 125         // take some time (it needs to wait for all the frames in the queue to be
 126         // done encoding, for one) and we are running on the main mixer thread.
 127         // However, since this means both encoders could be sending packets at
 128         // the same time, it means pts could come out of order to the stream mux,
 129         // and we need to plug it until the shutdown is complete.
 130         http_mux->plug();
 131         lock(qs_mu, qs_audio_mu);
 132         lock_guard<mutex> lock1(qs_mu, adopt_lock), lock2(qs_audio_mu, adopt_lock);
 133         QuickSyncEncoder *old_encoder = quicksync_encoder.release();  // When we go C++14, we can use move capture instead.
 134         X264Encoder *old_x264_encoder = nullptr;
 135         X264Encoder *old_x264_disk_encoder = nullptr;
 136         if (global_flags.x264_video_to_disk) {
 137                 old_x264_encoder = x264_encoder.release();
 138         }
 139         if (global_flags.x264_separate_disk_encode) {
 140                 old_x264_disk_encoder = x264_disk_encoder.release();
 141         }
 142         thread([old_encoder, old_x264_encoder, old_x264_disk_encoder, this]{
 143                 old_encoder->shutdown();
 144                 delete old_x264_encoder;
 145                 delete old_x264_disk_encoder;
 146                 old_encoder->close_file();
 147                 http_mux->unplug();
 148
 149                 // We cannot delete the encoder here, as this thread has no OpenGL context.
 150                 // We'll deal with it in begin_frame().
 151                 lock_guard<mutex> lock(qs_mu);
 152                 qs_needing_cleanup.emplace_back(old_encoder);
 153         }).detach();
 154
 155         if (global_flags.x264_video_to_disk) {
 156                 x264_encoder.reset(new X264Encoder(oformat, /*use_separate_disk_params=*/false));
 157                 assert(global_flags.x264_video_to_http);
 158                 if (global_flags.x264_video_to_http) {
 159                         x264_encoder->add_mux(http_mux.get());
 160                 }
 161                 if (overriding_bitrate != 0) {
 162                         x264_encoder->change_bitrate(overriding_bitrate);
 163                 }
 164         }
 165         X264Encoder *http_encoder = x264_encoder.get();
 166         X264Encoder *disk_encoder = x264_encoder.get();
 167         if (global_flags.x264_separate_disk_encode) {
 168                 x264_disk_encoder.reset(new X264Encoder(oformat, /*use_separate_disk_params=*/true));
 169                 disk_encoder = x264_disk_encoder.get();
 170         }
 171
 172         quicksync_encoder.reset(new QuickSyncEncoder(filename, resource_pool, surface, va_display, width, height, oformat, http_encoder, disk_encoder, disk_space_estimator));
 173         quicksync_encoder->set_http_mux(http_mux.get());
 174 }
 175
 176 void VideoEncoder::change_x264_bitrate(unsigned rate_kbit)
 177 {
 178         overriding_bitrate = rate_kbit;
 179         x264_encoder->change_bitrate(rate_kbit);
 180 }
 181
 182 void VideoEncoder::add_audio(int64_t pts, std::vector<float> audio)
 183 {
 184         // Take only qs_audio_mu, since add_audio() is thread safe
 185         // (we can only conflict with do_cut(), which takes qs_audio_mu)
 186         // and we don't want to contend with begin_frame().
 187         {
 188                 lock_guard<mutex> lock(qs_audio_mu);
 189                 quicksync_encoder->add_audio(pts, audio);
 190         }
 191         stream_audio_encoder->encode_audio(audio, pts + quicksync_encoder->global_delay());
 192 }
 193
 194 bool VideoEncoder::is_zerocopy() const
 195 {
 196         // Explicitly do _not_ take qs_mu; this is called from the mixer,
 197         // and qs_mu might be contended. is_zerocopy() is thread safe
 198         // and never called in parallel with do_cut() (both happen only
 199         // from the mixer thread).
 200         return quicksync_encoder->is_zerocopy();
 201 }
 202
 203 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)
 204 {
 205         lock_guard<mutex> lock(qs_mu);
 206         qs_needing_cleanup.clear();  // Since we have an OpenGL context here, and are called regularly.
 207         return quicksync_encoder->begin_frame(pts, duration, ycbcr_coefficients, input_frames, y_tex, cbcr_tex);
 208 }
 209
 210 RefCountedGLsync VideoEncoder::end_frame()
 211 {
 212         lock_guard<mutex> lock(qs_mu);
 213         return quicksync_encoder->end_frame();
 214 }
 215
 216 void VideoEncoder::open_output_streams()
 217 {
 218         for (bool is_srt : {false, true}) {
 219                 if (is_srt && global_flags.srt_destination_host.empty()) {
 220                         continue;
 221                 }
 222
 223                 AVFormatContext *avctx = avformat_alloc_context();
 224                 avctx->oformat = is_srt ? srt_oformat : oformat;
 225
 226                 uint8_t *buf = (uint8_t *)av_malloc(MUX_BUFFER_SIZE);
 227                 avctx->pb = avio_alloc_context(buf, MUX_BUFFER_SIZE, 1, this, nullptr, nullptr, nullptr);
 228                 if (is_srt) {
 229                         avctx->pb->write_packet = &VideoEncoder::write_srt_packet_thunk;
 230                 } else {
 231                         avctx->pb->write_data_type = &VideoEncoder::write_packet2_thunk;
 232                         avctx->pb->ignore_boundary_point = 1;
 233                 }
 234
 235                 Mux::Codec video_codec;
 236                 if (global_flags.av1_video_to_http) {
 237                         video_codec = Mux::CODEC_AV1;
 238                 } else {
 239                         video_codec = Mux::CODEC_H264;
 240                 }
 241
 242                 avctx->flags = AVFMT_FLAG_CUSTOM_IO;
 243
 244                 string video_extradata;
 245                 if (global_flags.x264_video_to_http) {
 246                         video_extradata = x264_encoder->get_global_headers();
 247 #ifdef HAVE_AV1
 248                 } else if (global_flags.av1_video_to_http) {
 249                         video_extradata = av1_encoder->get_global_headers();
 250 #endif
 251                 }
 252
 253                 Mux *mux = new Mux(avctx, width, height, video_codec, video_extradata, stream_audio_encoder->get_codec_parameters().get(),
 254                         get_color_space(global_flags.ycbcr_rec709_coefficients), COARSE_TIMEBASE,
 255                         /*write_callback=*/nullptr, is_srt ? Mux::WRITE_BACKGROUND : Mux::WRITE_FOREGROUND, { is_srt ? &srt_mux_metrics : &http_mux_metrics });
 256                 if (is_srt) {
 257                         srt_mux.reset(mux);
 258                         srt_mux_metrics.init({{ "destination", "srt" }});
 259                 } else {
 260                         http_mux.reset(mux);
 261                         http_mux_metrics.init({{ "destination", "http" }});
 262                 }
 263         }
 264 }
 265
 266 int VideoEncoder::write_packet2_thunk(void *opaque, uint8_t *buf, int buf_size, AVIODataMarkerType type, int64_t time)
 267 {
 268         VideoEncoder *video_encoder = (VideoEncoder *)opaque;
 269         return video_encoder->write_packet2(buf, buf_size, type, time);
 270 }
 271
 272 int VideoEncoder::write_packet2(uint8_t *buf, int buf_size, AVIODataMarkerType type, int64_t time)
 273 {
 274         if (type == AVIO_DATA_MARKER_SYNC_POINT || type == AVIO_DATA_MARKER_BOUNDARY_POINT) {
 275                 seen_sync_markers = true;
 276         } else if (type == AVIO_DATA_MARKER_UNKNOWN && !seen_sync_markers) {
 277                 // We don't know if this is a keyframe or not (the muxer could
 278                 // avoid marking it), so we just have to make the best of it.
 279                 type = AVIO_DATA_MARKER_SYNC_POINT;
 280         }
 281
 282         if (type == AVIO_DATA_MARKER_HEADER) {
 283                 http_mux_header.append((char *)buf, buf_size);
 284                 httpd->set_header(HTTPD::StreamID{ HTTPD::MAIN_STREAM, 0 }, http_mux_header);
 285         } else {
 286                 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 });
 287         }
 288         return buf_size;
 289 }
 290
 291 int VideoEncoder::write_srt_packet_thunk(void *opaque, uint8_t *buf, int buf_size)
 292 {
 293         VideoEncoder *video_encoder = (VideoEncoder *)opaque;
 294         return video_encoder->write_srt_packet(buf, buf_size);
 295 }
 296
 297 static string print_addrinfo(const addrinfo *ai)
 298 {
 299         char hoststr[NI_MAXHOST], portstr[NI_MAXSERV];
 300         if (getnameinfo(ai->ai_addr, ai->ai_addrlen, hoststr, sizeof(hoststr), portstr, sizeof(portstr), NI_DGRAM | NI_NUMERICHOST | NI_NUMERICSERV) != 0) {
 301                 return "<unknown address>";  // Should basically never happen, since we're not doing DNS lookups.
 302         }
 303
 304         if (ai->ai_family == AF_INET6) {
 305                 return string("[") + hoststr + "]:" + portstr;
 306         } else {
 307                 return string(hoststr) + ":" + portstr;
 308         }
 309 }
 310
 311 int VideoEncoder::open_srt_socket()
 312 {
 313         int sock = srt_create_socket();
 314         if (sock == -1) {
 315                 fprintf(stderr, "srt_create_socket(): %s\n", srt_getlasterror_str());
 316                 return -1;
 317         }
 318
 319         SRT_TRANSTYPE live = SRTT_LIVE;
 320         if (srt_setsockopt(sock, 0, SRTO_TRANSTYPE, &live, sizeof(live)) < 0) {
 321                 fprintf(stderr, "srt_setsockopt(SRTO_TRANSTYPE): %s\n", srt_getlasterror_str());
 322                 srt_close(sock);
 323                 return -1;
 324         }
 325
 326         if (srt_setsockopt(sock, 0, SRTO_LATENCY, &global_flags.srt_output_latency, sizeof(global_flags.srt_output_latency)) < 0) {
 327                 fprintf(stderr, "srt_setsockopt(SRTO_LATENCY): %s\n", srt_getlasterror_str());
 328                 srt_close(sock);
 329                 return -1;
 330         }
 331
 332         if (!global_flags.srt_streamid.empty()) {
 333                 if (srt_setsockopt(sock, 0, SRTO_STREAMID, global_flags.srt_streamid.data(), global_flags.srt_streamid.size()) < 0) {
 334                         fprintf(stderr, "srt_setsockopt(SRTO_STREAMID): %s\n", srt_getlasterror_str());
 335                         srt_close(sock);
 336                         return -1;
 337                 }
 338         }
 339
 340         if (!global_flags.srt_passphrase.empty()) {
 341                 if (srt_setsockopt(sock, 0, SRTO_PASSPHRASE, global_flags.srt_passphrase.data(), global_flags.srt_passphrase.size()) < 0) {
 342                         fprintf(stderr, "srt_setsockopt(SRTO_PASSPHRASE): %s\n", srt_getlasterror_str());
 343                         srt_close(sock);
 344                         return -1;
 345                 }
 346         }
 347
 348         return sock;
 349 }
 350
 351 int VideoEncoder::connect_to_srt()
 352 {
 353         // We need to specify SOCK_DGRAM as a hint, or we'll get all addresses
 354         // three times (for each of TCP, UDP, raw).
 355         addrinfo hints;
 356         memset(&hints, 0, sizeof(hints));
 357         hints.ai_flags = AI_ADDRCONFIG;
 358         hints.ai_socktype = SOCK_DGRAM;
 359
 360         addrinfo *ai;
 361         int ret = getaddrinfo(global_flags.srt_destination_host.c_str(), global_flags.srt_destination_port.c_str(), &hints, &ai);
 362         if (ret != 0) {
 363                 fprintf(stderr, "getaddrinfo(%s:%s): %s\n", global_flags.srt_destination_host.c_str(), global_flags.srt_destination_port.c_str(), gai_strerror(ret));
 364                 return -1;
 365         }
 366
 367         for (const addrinfo *cur = ai; cur != nullptr; cur = cur->ai_next) {
 368                 // Seemingly, srt_create_socket() isn't universal; once we try to connect,
 369                 // it gets locked to either IPv4 or IPv6. So we need to create a new one
 370                 // for every address we try.
 371                 int sock = open_srt_socket();
 372                 if (sock == -1) {
 373                         // Die immediately.
 374                         return sock;
 375                 }
 376                 if (srt_connect(sock, cur->ai_addr, cur->ai_addrlen) < 0) {
 377                         fprintf(stderr, "srt_connect(%s): %s\n", print_addrinfo(cur).c_str(), srt_getlasterror_str());
 378                         srt_close(sock);
 379                         continue;
 380                 }
 381                 fprintf(stderr, "Connected to destination SRT endpoint at %s.\n", print_addrinfo(cur).c_str());
 382                 freeaddrinfo(ai);
 383                 return sock;
 384         }
 385
 386         // Out of candidates, so give up.
 387         freeaddrinfo(ai);
 388         return -1;
 389 }
 390
 391 int VideoEncoder::write_srt_packet(uint8_t *buf, int buf_size)
 392 {
 393         while (buf_size > 0) {
 394                 if (srt_sock == -1) {
 395                         srt_sock = connect_to_srt();
 396                         if (srt_sock == -1) {
 397                                 usleep(100000);
 398                                 continue;
 399                         }
 400                 }
 401                 int to_send = min(buf_size, SRT_LIVE_DEF_PLSIZE);
 402                 int ret = srt_send(srt_sock, (char *)buf, to_send);
 403                 if (ret < 0)  {
 404                         fprintf(stderr, "srt_send(): %s\n", srt_getlasterror_str());
 405                         srt_close(srt_sock);
 406                         srt_sock = connect_to_srt();
 407                         continue;
 408                 }
 409                 buf += ret;
 410                 buf_size -= ret;
 411         }
 412         return buf_size;
 413 }
 414