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

   1 #include "ffmpeg_capture.h"
   2
   3 #include <assert.h>
   4 #include <pthread.h>
   5 #include <stdint.h>
   6 #include <stdio.h>
   7 #include <stdlib.h>
   8 #include <string.h>
   9 #include <sys/stat.h>
  10 #include <unistd.h>
  11
  12 extern "C" {
  13 #include <libavcodec/avcodec.h>
  14 #include <libavformat/avformat.h>
  15 #include <libavutil/avutil.h>
  16 #include <libavutil/error.h>
  17 #include <libavutil/frame.h>
  18 #include <libavutil/imgutils.h>
  19 #include <libavutil/mem.h>
  20 #include <libavutil/pixfmt.h>
  21 #include <libavutil/opt.h>
  22 #include <libswscale/swscale.h>
  23 }
  24
  25 #include <chrono>
  26 #include <cstdint>
  27 #include <utility>
  28 #include <vector>
  29
  30 #include "bmusb/bmusb.h"
  31 #include "ffmpeg_raii.h"
  32 #include "ffmpeg_util.h"
  33 #include "flags.h"
  34 #include "image_input.h"
  35 #include "ref_counted_frame.h"
  36 #include "timebase.h"
  37
  38 #define FRAME_SIZE (8 << 20)  // 8 MB.
  39
  40 using namespace std;
  41 using namespace std::chrono;
  42 using namespace bmusb;
  43 using namespace movit;
  44
  45 namespace {
  46
  47 steady_clock::time_point compute_frame_start(int64_t frame_pts, int64_t pts_origin, const AVRational &video_timebase, const steady_clock::time_point &origin, double rate)
  48 {
  49         const duration<double> pts((frame_pts - pts_origin) * double(video_timebase.num) / double(video_timebase.den));
  50         return origin + duration_cast<steady_clock::duration>(pts / rate);
  51 }
  52
  53 bool changed_since(const std::string &pathname, const timespec &ts)
  54 {
  55         if (ts.tv_sec < 0) {
  56                 return false;
  57         }
  58         struct stat buf;
  59         if (stat(pathname.c_str(), &buf) != 0) {
  60                 fprintf(stderr, "%s: Couldn't check for new version, leaving the old in place.\n", pathname.c_str());
  61                 return false;
  62         }
  63         return (buf.st_mtim.tv_sec != ts.tv_sec || buf.st_mtim.tv_nsec != ts.tv_nsec);
  64 }
  65
  66 bool is_full_range(const AVPixFmtDescriptor *desc)
  67 {
  68         // This is horrible, but there's no better way that I know of.
  69         return (strchr(desc->name, 'j') != nullptr);
  70 }
  71
  72 AVPixelFormat decide_dst_format(AVPixelFormat src_format, bmusb::PixelFormat dst_format_type)
  73 {
  74         if (dst_format_type == bmusb::PixelFormat_8BitBGRA) {
  75                 return AV_PIX_FMT_BGRA;
  76         }
  77         if (dst_format_type == FFmpegCapture::PixelFormat_NV12) {
  78                 return AV_PIX_FMT_NV12;
  79         }
  80
  81         assert(dst_format_type == bmusb::PixelFormat_8BitYCbCrPlanar);
  82
  83         // If this is a non-Y'CbCr format, just convert to 4:4:4 Y'CbCr
  84         // and be done with it. It's too strange to spend a lot of time on.
  85         // (Let's hope there's no alpha.)
  86         const AVPixFmtDescriptor *src_desc = av_pix_fmt_desc_get(src_format);
  87         if (src_desc == nullptr ||
  88             src_desc->nb_components != 3 ||
  89             (src_desc->flags & AV_PIX_FMT_FLAG_RGB)) {
  90                 return AV_PIX_FMT_YUV444P;
  91         }
  92
  93         // The best for us would be Cb and Cr together if possible,
  94         // but FFmpeg doesn't support that except in the special case of
  95         // NV12, so we need to go to planar even for the case of NV12.
  96         // Thus, look for the closest (but no worse) 8-bit planar Y'CbCr format
  97         // that matches in color range. (This will also include the case of
  98         // the source format already being acceptable.)
  99         bool src_full_range = is_full_range(src_desc);
 100         const char *best_format = "yuv444p";
 101         unsigned best_score = numeric_limits<unsigned>::max();
 102         for (const AVPixFmtDescriptor *desc = av_pix_fmt_desc_next(nullptr);
 103              desc;
 104              desc = av_pix_fmt_desc_next(desc)) {
 105                 // Find planar Y'CbCr formats only.
 106                 if (desc->nb_components != 3) continue;
 107                 if (desc->flags & AV_PIX_FMT_FLAG_RGB) continue;
 108                 if (!(desc->flags & AV_PIX_FMT_FLAG_PLANAR)) continue;
 109                 if (desc->comp[0].plane != 0 ||
 110                     desc->comp[1].plane != 1 ||
 111                     desc->comp[2].plane != 2) continue;
 112
 113                 // 8-bit formats only.
 114                 if (desc->flags & AV_PIX_FMT_FLAG_BE) continue;
 115                 if (desc->comp[0].depth != 8) continue;
 116
 117                 // Same or better chroma resolution only.
 118                 int chroma_w_diff = desc->log2_chroma_w - src_desc->log2_chroma_w;
 119                 int chroma_h_diff = desc->log2_chroma_h - src_desc->log2_chroma_h;
 120                 if (chroma_w_diff < 0 || chroma_h_diff < 0)
 121                         continue;
 122
 123                 // Matching full/limited range only.
 124                 if (is_full_range(desc) != src_full_range)
 125                         continue;
 126
 127                 // Pick something with as little excess chroma resolution as possible.
 128                 unsigned score = (1 << (chroma_w_diff)) << chroma_h_diff;
 129                 if (score < best_score) {
 130                         best_score = score;
 131                         best_format = desc->name;
 132                 }
 133         }
 134         return av_get_pix_fmt(best_format);
 135 }
 136
 137 YCbCrFormat decode_ycbcr_format(const AVPixFmtDescriptor *desc, const AVFrame *frame)
 138 {
 139         YCbCrFormat format;
 140         AVColorSpace colorspace = av_frame_get_colorspace(frame);
 141         switch (colorspace) {
 142         case AVCOL_SPC_BT709:
 143                 format.luma_coefficients = YCBCR_REC_709;
 144                 break;
 145         case AVCOL_SPC_BT470BG:
 146         case AVCOL_SPC_SMPTE170M:
 147         case AVCOL_SPC_SMPTE240M:
 148                 format.luma_coefficients = YCBCR_REC_601;
 149                 break;
 150         case AVCOL_SPC_BT2020_NCL:
 151                 format.luma_coefficients = YCBCR_REC_2020;
 152                 break;
 153         case AVCOL_SPC_UNSPECIFIED:
 154                 format.luma_coefficients = (frame->height >= 720 ? YCBCR_REC_709 : YCBCR_REC_601);
 155                 break;
 156         default:
 157                 fprintf(stderr, "Unknown Y'CbCr coefficient enum %d from FFmpeg; choosing Rec. 709.\n",
 158                         colorspace);
 159                 format.luma_coefficients = YCBCR_REC_709;
 160                 break;
 161         }
 162
 163         format.full_range = is_full_range(desc);
 164         format.num_levels = 1 << desc->comp[0].depth;
 165         format.chroma_subsampling_x = 1 << desc->log2_chroma_w;
 166         format.chroma_subsampling_y = 1 << desc->log2_chroma_h;
 167
 168         switch (frame->chroma_location) {
 169         case AVCHROMA_LOC_LEFT:
 170                 format.cb_x_position = 0.0;
 171                 format.cb_y_position = 0.5;
 172                 break;
 173         case AVCHROMA_LOC_CENTER:
 174                 format.cb_x_position = 0.5;
 175                 format.cb_y_position = 0.5;
 176                 break;
 177         case AVCHROMA_LOC_TOPLEFT:
 178                 format.cb_x_position = 0.0;
 179                 format.cb_y_position = 0.0;
 180                 break;
 181         case AVCHROMA_LOC_TOP:
 182                 format.cb_x_position = 0.5;
 183                 format.cb_y_position = 0.0;
 184                 break;
 185         case AVCHROMA_LOC_BOTTOMLEFT:
 186                 format.cb_x_position = 0.0;
 187                 format.cb_y_position = 1.0;
 188                 break;
 189         case AVCHROMA_LOC_BOTTOM:
 190                 format.cb_x_position = 0.5;
 191                 format.cb_y_position = 1.0;
 192                 break;
 193         default:
 194                 fprintf(stderr, "Unknown chroma location coefficient enum %d from FFmpeg; choosing Rec. 709.\n",
 195                         frame->chroma_location);
 196                 format.cb_x_position = 0.5;
 197                 format.cb_y_position = 0.5;
 198                 break;
 199         }
 200
 201         format.cr_x_position = format.cb_x_position;
 202         format.cr_y_position = format.cb_y_position;
 203         return format;
 204 }
 205
 206 }  // namespace
 207
 208 FFmpegCapture::FFmpegCapture(const string &filename, unsigned width, unsigned height)
 209         : filename(filename), width(width), height(height), video_timebase{1, 1}
 210 {
 211         // Not really used for anything.
 212         description = "Video: " + filename;
 213
 214         last_frame = steady_clock::now();
 215
 216         avformat_network_init();  // In case someone wants this.
 217 }
 218
 219 FFmpegCapture::~FFmpegCapture()
 220 {
 221         if (has_dequeue_callbacks) {
 222                 dequeue_cleanup_callback();
 223         }
 224         avresample_free(&resampler);
 225 }
 226
 227 void FFmpegCapture::configure_card()
 228 {
 229         if (video_frame_allocator == nullptr) {
 230                 owned_video_frame_allocator.reset(new MallocFrameAllocator(FRAME_SIZE, NUM_QUEUED_VIDEO_FRAMES));
 231                 set_video_frame_allocator(owned_video_frame_allocator.get());
 232         }
 233         if (audio_frame_allocator == nullptr) {
 234                 // Audio can come out in pretty large chunks, so increase from the default 1 MB.
 235                 owned_audio_frame_allocator.reset(new MallocFrameAllocator(1 << 20, NUM_QUEUED_AUDIO_FRAMES));
 236                 set_audio_frame_allocator(owned_audio_frame_allocator.get());
 237         }
 238 }
 239
 240 void FFmpegCapture::start_bm_capture()
 241 {
 242         if (running) {
 243                 return;
 244         }
 245         running = true;
 246         producer_thread_should_quit.unquit();
 247         producer_thread = thread(&FFmpegCapture::producer_thread_func, this);
 248 }
 249
 250 void FFmpegCapture::stop_dequeue_thread()
 251 {
 252         if (!running) {
 253                 return;
 254         }
 255         running = false;
 256         producer_thread_should_quit.quit();
 257         producer_thread.join();
 258 }
 259
 260 std::map<uint32_t, VideoMode> FFmpegCapture::get_available_video_modes() const
 261 {
 262         // Note: This will never really be shown in the UI.
 263         VideoMode mode;
 264
 265         char buf[256];
 266         snprintf(buf, sizeof(buf), "%ux%u", width, height);
 267         mode.name = buf;
 268
 269         mode.autodetect = false;
 270         mode.width = width;
 271         mode.height = height;
 272         mode.frame_rate_num = 60;
 273         mode.frame_rate_den = 1;
 274         mode.interlaced = false;
 275
 276         return {{ 0, mode }};
 277 }
 278
 279 void FFmpegCapture::producer_thread_func()
 280 {
 281         char thread_name[16];
 282         snprintf(thread_name, sizeof(thread_name), "FFmpeg_C_%d", card_index);
 283         pthread_setname_np(pthread_self(), thread_name);
 284
 285         while (!producer_thread_should_quit.should_quit()) {
 286                 string pathname = search_for_file(filename);
 287                 if (filename.empty()) {
 288                         fprintf(stderr, "%s not found, sleeping one second and trying again...\n", filename.c_str());
 289                         send_disconnected_frame();
 290                         producer_thread_should_quit.sleep_for(seconds(1));
 291                         continue;
 292                 }
 293                 should_interrupt = false;
 294                 if (!play_video(pathname)) {
 295                         // Error.
 296                         fprintf(stderr, "Error when playing %s, sleeping one second and trying again...\n", pathname.c_str());
 297                         send_disconnected_frame();
 298                         producer_thread_should_quit.sleep_for(seconds(1));
 299                         continue;
 300                 }
 301
 302                 // Probably just EOF, will exit the loop above on next test.
 303         }
 304
 305         if (has_dequeue_callbacks) {
 306                 dequeue_cleanup_callback();
 307                 has_dequeue_callbacks = false;
 308         }
 309 }
 310
 311 void FFmpegCapture::send_disconnected_frame()
 312 {
 313         // Send an empty frame to signal that we have no signal anymore.
 314         FrameAllocator::Frame video_frame = video_frame_allocator->alloc_frame();
 315         if (video_frame.data) {
 316                 VideoFormat video_format;
 317                 video_format.width = width;
 318                 video_format.height = height;
 319                 video_format.frame_rate_nom = 60;
 320                 video_format.frame_rate_den = 1;
 321                 video_format.is_connected = false;
 322                 if (pixel_format == bmusb::PixelFormat_8BitBGRA) {
 323                         video_format.stride = width * 4;
 324                         video_frame.len = width * height * 4;
 325                 } else {
 326                         video_format.stride = width;
 327                         current_frame_ycbcr_format.luma_coefficients = YCBCR_REC_709;
 328                         current_frame_ycbcr_format.full_range = true;
 329                         current_frame_ycbcr_format.num_levels = 256;
 330                         current_frame_ycbcr_format.chroma_subsampling_x = 2;
 331                         current_frame_ycbcr_format.chroma_subsampling_y = 2;
 332                         current_frame_ycbcr_format.cb_x_position = 0.0f;
 333                         current_frame_ycbcr_format.cb_y_position = 0.0f;
 334                         current_frame_ycbcr_format.cr_x_position = 0.0f;
 335                         current_frame_ycbcr_format.cr_y_position = 0.0f;
 336                         video_frame.len = width * height * 2;
 337                 }
 338                 memset(video_frame.data, 0, video_frame.len);
 339
 340                 frame_callback(-1, AVRational{1, TIMEBASE}, -1, AVRational{1, TIMEBASE}, timecode++,
 341                         video_frame, /*video_offset=*/0, video_format,
 342                         FrameAllocator::Frame(), /*audio_offset=*/0, AudioFormat());
 343                 last_frame_was_connected = false;
 344         }
 345 }
 346
 347 bool FFmpegCapture::play_video(const string &pathname)
 348 {
 349         // Note: Call before open, not after; otherwise, there's a race.
 350         // (There is now, too, but it tips the correct way. We could use fstat()
 351         // if we had the file descriptor.)
 352         timespec last_modified;
 353         struct stat buf;
 354         if (stat(pathname.c_str(), &buf) != 0) {
 355                 // Probably some sort of protocol, so can't stat.
 356                 last_modified.tv_sec = -1;
 357         } else {
 358                 last_modified = buf.st_mtim;
 359         }
 360
 361         AVDictionary *opts = nullptr;
 362         av_dict_set(&opts, "fflags", "nobuffer", 0);
 363
 364         auto format_ctx = avformat_open_input_unique(pathname.c_str(), nullptr, &opts, AVIOInterruptCB{ &FFmpegCapture::interrupt_cb_thunk, this });
 365         if (format_ctx == nullptr) {
 366                 fprintf(stderr, "%s: Error opening file\n", pathname.c_str());
 367                 return false;
 368         }
 369
 370         if (avformat_find_stream_info(format_ctx.get(), nullptr) < 0) {
 371                 fprintf(stderr, "%s: Error finding stream info\n", pathname.c_str());
 372                 return false;
 373         }
 374
 375         int video_stream_index = find_stream_index(format_ctx.get(), AVMEDIA_TYPE_VIDEO);
 376         if (video_stream_index == -1) {
 377                 fprintf(stderr, "%s: No video stream found\n", pathname.c_str());
 378                 return false;
 379         }
 380
 381         int audio_stream_index = find_stream_index(format_ctx.get(), AVMEDIA_TYPE_AUDIO);
 382
 383         // Open video decoder.
 384         const AVCodecParameters *video_codecpar = format_ctx->streams[video_stream_index]->codecpar;
 385         AVCodec *video_codec = avcodec_find_decoder(video_codecpar->codec_id);
 386         video_timebase = format_ctx->streams[video_stream_index]->time_base;
 387         AVCodecContextWithDeleter video_codec_ctx = avcodec_alloc_context3_unique(nullptr);
 388         if (avcodec_parameters_to_context(video_codec_ctx.get(), video_codecpar) < 0) {
 389                 fprintf(stderr, "%s: Cannot fill video codec parameters\n", pathname.c_str());
 390                 return false;
 391         }
 392         if (video_codec == nullptr) {
 393                 fprintf(stderr, "%s: Cannot find video decoder\n", pathname.c_str());
 394                 return false;
 395         }
 396         if (avcodec_open2(video_codec_ctx.get(), video_codec, nullptr) < 0) {
 397                 fprintf(stderr, "%s: Cannot open video decoder\n", pathname.c_str());
 398                 return false;
 399         }
 400         unique_ptr<AVCodecContext, decltype(avcodec_close)*> video_codec_ctx_cleanup(
 401                 video_codec_ctx.get(), avcodec_close);
 402
 403         // Open audio decoder, if we have audio.
 404         AVCodecContextWithDeleter audio_codec_ctx;
 405         if (audio_stream_index != -1) {
 406                 audio_codec_ctx = avcodec_alloc_context3_unique(nullptr);
 407                 const AVCodecParameters *audio_codecpar = format_ctx->streams[audio_stream_index]->codecpar;
 408                 audio_timebase = format_ctx->streams[audio_stream_index]->time_base;
 409                 if (avcodec_parameters_to_context(audio_codec_ctx.get(), audio_codecpar) < 0) {
 410                         fprintf(stderr, "%s: Cannot fill audio codec parameters\n", pathname.c_str());
 411                         return false;
 412                 }
 413                 AVCodec *audio_codec = avcodec_find_decoder(audio_codecpar->codec_id);
 414                 if (audio_codec == nullptr) {
 415                         fprintf(stderr, "%s: Cannot find audio decoder\n", pathname.c_str());
 416                         return false;
 417                 }
 418                 if (avcodec_open2(audio_codec_ctx.get(), audio_codec, nullptr) < 0) {
 419                         fprintf(stderr, "%s: Cannot open audio decoder\n", pathname.c_str());
 420                         return false;
 421                 }
 422         }
 423         unique_ptr<AVCodecContext, decltype(avcodec_close)*> audio_codec_ctx_cleanup(
 424                 audio_codec_ctx.get(), avcodec_close);
 425
 426         internal_rewind();
 427
 428         // Main loop.
 429         bool first_frame = true;
 430         while (!producer_thread_should_quit.should_quit()) {
 431                 if (process_queued_commands(format_ctx.get(), pathname, last_modified, /*rewound=*/nullptr)) {
 432                         return true;
 433                 }
 434                 UniqueFrame audio_frame = audio_frame_allocator->alloc_frame();
 435                 AudioFormat audio_format;
 436
 437                 int64_t audio_pts;
 438                 bool error;
 439                 AVFrameWithDeleter frame = decode_frame(format_ctx.get(), video_codec_ctx.get(), audio_codec_ctx.get(),
 440                         pathname, video_stream_index, audio_stream_index, audio_frame.get(), &audio_format, &audio_pts, &error);
 441                 if (error) {
 442                         return false;
 443                 }
 444                 if (frame == nullptr) {
 445                         // EOF. Loop back to the start if we can.
 446                         if (av_seek_frame(format_ctx.get(), /*stream_index=*/-1, /*timestamp=*/0, /*flags=*/0) < 0) {
 447                                 fprintf(stderr, "%s: Rewind failed, not looping.\n", pathname.c_str());
 448                                 return true;
 449                         }
 450                         if (video_codec_ctx != nullptr) {
 451                                 avcodec_flush_buffers(video_codec_ctx.get());
 452                         }
 453                         if (audio_codec_ctx != nullptr) {
 454                                 avcodec_flush_buffers(audio_codec_ctx.get());
 455                         }
 456                         // If the file has changed since last time, return to get it reloaded.
 457                         // Note that depending on how you move the file into place, you might
 458                         // end up corrupting the one you're already playing, so this path
 459                         // might not trigger.
 460                         if (changed_since(pathname, last_modified)) {
 461                                 return true;
 462                         }
 463                         internal_rewind();
 464                         continue;
 465                 }
 466
 467                 VideoFormat video_format = construct_video_format(frame.get(), video_timebase);
 468                 UniqueFrame video_frame = make_video_frame(frame.get(), pathname, &error);
 469                 if (error) {
 470                         return false;
 471                 }
 472
 473                 for ( ;; ) {
 474                         if (last_pts == 0 && pts_origin == 0) {
 475                                 pts_origin = frame->pts;
 476                         }
 477                         next_frame_start = compute_frame_start(frame->pts, pts_origin, video_timebase, start, rate);
 478                         if (first_frame && last_frame_was_connected) {
 479                                 // If reconnect took more than one second, this is probably a live feed,
 480                                 // and we should reset the resampler. (Or the rate is really, really low,
 481                                 // in which case a reset on the first frame is fine anyway.)
 482                                 if (duration<double>(next_frame_start - last_frame).count() >= 1.0) {
 483                                         last_frame_was_connected = false;
 484                                 }
 485                         }
 486                         video_frame->received_timestamp = next_frame_start;
 487
 488                         // The easiest way to get all the rate conversions etc. right is to move the
 489                         // audio PTS into the video PTS timebase and go from there. (We'll get some
 490                         // rounding issues, but they should not be a big problem.)
 491                         int64_t audio_pts_as_video_pts = av_rescale_q(audio_pts, audio_timebase, video_timebase);
 492                         audio_frame->received_timestamp = compute_frame_start(audio_pts_as_video_pts, pts_origin, video_timebase, start, rate);
 493
 494                         bool finished_wakeup = producer_thread_should_quit.sleep_until(next_frame_start);
 495                         if (finished_wakeup) {
 496                                 if (audio_frame->len > 0) {
 497                                         assert(audio_pts != -1);
 498                                 }
 499                                 if (!last_frame_was_connected) {
 500                                         // We're recovering from an error (or really slow load, see above).
 501                                         // Make sure to get the audio resampler reset. (This is a hack;
 502                                         // ideally, the frame callback should just accept a way to signal
 503                                         // audio discontinuity.)
 504                                         timecode += MAX_FPS * 2 + 1;
 505                                 }
 506                                 frame_callback(frame->pts, video_timebase, audio_pts, audio_timebase, timecode++,
 507                                         video_frame.get_and_release(), 0, video_format,
 508                                         audio_frame.get_and_release(), 0, audio_format);
 509                                 first_frame = false;
 510                                 last_frame = steady_clock::now();
 511                                 last_frame_was_connected = true;
 512                                 break;
 513                         } else {
 514                                 if (producer_thread_should_quit.should_quit()) break;
 515
 516                                 bool rewound = false;
 517                                 if (process_queued_commands(format_ctx.get(), pathname, last_modified, &rewound)) {
 518                                         return true;
 519                                 }
 520                                 // If we just rewound, drop this frame on the floor and be done.
 521                                 if (rewound) {
 522                                         break;
 523                                 }
 524                                 // OK, we didn't, so probably a rate change. Recalculate next_frame_start,
 525                                 // but if it's now in the past, we'll reset the origin, so that we don't
 526                                 // generate a huge backlog of frames that we need to run through quickly.
 527                                 next_frame_start = compute_frame_start(frame->pts, pts_origin, video_timebase, start, rate);
 528                                 steady_clock::time_point now = steady_clock::now();
 529                                 if (next_frame_start < now) {
 530                                         pts_origin = frame->pts;
 531                                         start = next_frame_start = now;
 532                                 }
 533                         }
 534                 }
 535                 last_pts = frame->pts;
 536         }
 537         return true;
 538 }
 539
 540 void FFmpegCapture::internal_rewind()
 541 {
 542         pts_origin = last_pts = 0;
 543         start = next_frame_start = steady_clock::now();
 544 }
 545
 546 bool FFmpegCapture::process_queued_commands(AVFormatContext *format_ctx, const std::string &pathname, timespec last_modified, bool *rewound)
 547 {
 548         // Process any queued commands from other threads.
 549         vector<QueuedCommand> commands;
 550         {
 551                 lock_guard<mutex> lock(queue_mu);
 552                 swap(commands, command_queue);
 553         }
 554         for (const QueuedCommand &cmd : commands) {
 555                 switch (cmd.command) {
 556                 case QueuedCommand::REWIND:
 557                         if (av_seek_frame(format_ctx, /*stream_index=*/-1, /*timestamp=*/0, /*flags=*/0) < 0) {
 558                                 fprintf(stderr, "%s: Rewind failed, stopping play.\n", pathname.c_str());
 559                         }
 560                         // If the file has changed since last time, return to get it reloaded.
 561                         // Note that depending on how you move the file into place, you might
 562                         // end up corrupting the one you're already playing, so this path
 563                         // might not trigger.
 564                         if (changed_since(pathname, last_modified)) {
 565                                 return true;
 566                         }
 567                         internal_rewind();
 568                         if (rewound != nullptr) {
 569                                 *rewound = true;
 570                         }
 571                         break;
 572
 573                 case QueuedCommand::CHANGE_RATE:
 574                         // Change the origin to the last played frame.
 575                         start = compute_frame_start(last_pts, pts_origin, video_timebase, start, rate);
 576                         pts_origin = last_pts;
 577                         rate = cmd.new_rate;
 578                         break;
 579                 }
 580         }
 581         return false;
 582 }
 583
 584 namespace {
 585
 586 }  // namespace
 587
 588 AVFrameWithDeleter FFmpegCapture::decode_frame(AVFormatContext *format_ctx, AVCodecContext *video_codec_ctx, AVCodecContext *audio_codec_ctx,
 589         const std::string &pathname, int video_stream_index, int audio_stream_index,
 590         FrameAllocator::Frame *audio_frame, AudioFormat *audio_format, int64_t *audio_pts, bool *error)
 591 {
 592         *error = false;
 593
 594         // Read packets until we have a frame or there are none left.
 595         bool frame_finished = false;
 596         AVFrameWithDeleter audio_avframe = av_frame_alloc_unique();
 597         AVFrameWithDeleter video_avframe = av_frame_alloc_unique();
 598         bool eof = false;
 599         *audio_pts = -1;
 600         bool has_audio = false;
 601         do {
 602                 AVPacket pkt;
 603                 unique_ptr<AVPacket, decltype(av_packet_unref)*> pkt_cleanup(
 604                         &pkt, av_packet_unref);
 605                 av_init_packet(&pkt);
 606                 pkt.data = nullptr;
 607                 pkt.size = 0;
 608                 if (av_read_frame(format_ctx, &pkt) == 0) {
 609                         if (pkt.stream_index == audio_stream_index && audio_callback != nullptr) {
 610                                 audio_callback(&pkt, format_ctx->streams[audio_stream_index]->time_base);
 611                         }
 612                         if (pkt.stream_index == video_stream_index) {
 613                                 if (avcodec_send_packet(video_codec_ctx, &pkt) < 0) {
 614                                         fprintf(stderr, "%s: Cannot send packet to video codec.\n", pathname.c_str());
 615                                         *error = true;
 616                                         return AVFrameWithDeleter(nullptr);
 617                                 }
 618                         } else if (pkt.stream_index == audio_stream_index) {
 619                                 has_audio = true;
 620                                 if (avcodec_send_packet(audio_codec_ctx, &pkt) < 0) {
 621                                         fprintf(stderr, "%s: Cannot send packet to audio codec.\n", pathname.c_str());
 622                                         *error = true;
 623                                         return AVFrameWithDeleter(nullptr);
 624                                 }
 625                         }
 626                 } else {
 627                         eof = true;  // Or error, but ignore that for the time being.
 628                 }
 629
 630                 // Decode audio, if any.
 631                 if (has_audio) {
 632                         for ( ;; ) {
 633                                 int err = avcodec_receive_frame(audio_codec_ctx, audio_avframe.get());
 634                                 if (err == 0) {
 635                                         if (*audio_pts == -1) {
 636                                                 *audio_pts = audio_avframe->pts;
 637                                         }
 638                                         convert_audio(audio_avframe.get(), audio_frame, audio_format);
 639                                 } else if (err == AVERROR(EAGAIN)) {
 640                                         break;
 641                                 } else {
 642                                         fprintf(stderr, "%s: Cannot receive frame from audio codec.\n", pathname.c_str());
 643                                         *error = true;
 644                                         return AVFrameWithDeleter(nullptr);
 645                                 }
 646                         }
 647                 }
 648
 649                 // Decode video, if we have a frame.
 650                 int err = avcodec_receive_frame(video_codec_ctx, video_avframe.get());
 651                 if (err == 0) {
 652                         frame_finished = true;
 653                         break;
 654                 } else if (err != AVERROR(EAGAIN)) {
 655                         fprintf(stderr, "%s: Cannot receive frame from video codec.\n", pathname.c_str());
 656                         *error = true;
 657                         return AVFrameWithDeleter(nullptr);
 658                 }
 659         } while (!eof);
 660
 661         if (frame_finished)
 662                 return video_avframe;
 663         else
 664                 return AVFrameWithDeleter(nullptr);
 665 }
 666
 667 void FFmpegCapture::convert_audio(const AVFrame *audio_avframe, FrameAllocator::Frame *audio_frame, AudioFormat *audio_format)
 668 {
 669         // Decide on a format. If there already is one in this audio frame,
 670         // we're pretty much forced to use it. If not, we try to find an exact match.
 671         // If that still doesn't work, we default to 32-bit signed chunked
 672         // (float would be nice, but there's really no way to signal that yet).
 673         AVSampleFormat dst_format;
 674         if (audio_format->bits_per_sample == 0) {
 675                 switch (audio_avframe->format) {
 676                 case AV_SAMPLE_FMT_S16:
 677                 case AV_SAMPLE_FMT_S16P:
 678                         audio_format->bits_per_sample = 16;
 679                         dst_format = AV_SAMPLE_FMT_S16;
 680                         break;
 681                 case AV_SAMPLE_FMT_S32:
 682                 case AV_SAMPLE_FMT_S32P:
 683                 default:
 684                         audio_format->bits_per_sample = 32;
 685                         dst_format = AV_SAMPLE_FMT_S32;
 686                         break;
 687                 }
 688         } else if (audio_format->bits_per_sample == 16) {
 689                 dst_format = AV_SAMPLE_FMT_S16;
 690         } else if (audio_format->bits_per_sample == 32) {
 691                 dst_format = AV_SAMPLE_FMT_S32;
 692         } else {
 693                 assert(false);
 694         }
 695         audio_format->num_channels = 2;
 696
 697         int64_t channel_layout = audio_avframe->channel_layout;
 698         if (channel_layout == 0) {
 699                 channel_layout = av_get_default_channel_layout(audio_avframe->channels);
 700         }
 701
 702         if (resampler == nullptr ||
 703             audio_avframe->format != last_src_format ||
 704             dst_format != last_dst_format ||
 705             channel_layout != last_channel_layout ||
 706             av_frame_get_sample_rate(audio_avframe) != last_sample_rate) {
 707                 avresample_free(&resampler);
 708                 resampler = avresample_alloc_context();
 709                 if (resampler == nullptr) {
 710                         fprintf(stderr, "Allocating resampler failed.\n");
 711                         exit(1);
 712                 }
 713
 714                 av_opt_set_int(resampler, "in_channel_layout",  channel_layout,                             0);
 715                 av_opt_set_int(resampler, "out_channel_layout", AV_CH_LAYOUT_STEREO,                        0);
 716                 av_opt_set_int(resampler, "in_sample_rate",     av_frame_get_sample_rate(audio_avframe),    0);
 717                 av_opt_set_int(resampler, "out_sample_rate",    OUTPUT_FREQUENCY,                           0);
 718                 av_opt_set_int(resampler, "in_sample_fmt",      audio_avframe->format,                      0);
 719                 av_opt_set_int(resampler, "out_sample_fmt",     dst_format,                                 0);
 720
 721                 if (avresample_open(resampler) < 0) {
 722                         fprintf(stderr, "Could not open resample context.\n");
 723                         exit(1);
 724                 }
 725
 726                 last_src_format = AVSampleFormat(audio_avframe->format);
 727                 last_dst_format = dst_format;
 728                 last_channel_layout = channel_layout;
 729                 last_sample_rate = av_frame_get_sample_rate(audio_avframe);
 730         }
 731
 732         size_t bytes_per_sample = (audio_format->bits_per_sample / 8) * 2;
 733         size_t num_samples_room = (audio_frame->size - audio_frame->len) / bytes_per_sample;
 734
 735         uint8_t *data = audio_frame->data + audio_frame->len;
 736         int out_samples = avresample_convert(resampler, &data, 0, num_samples_room,
 737                 const_cast<uint8_t **>(audio_avframe->data), audio_avframe->linesize[0], audio_avframe->nb_samples);
 738         if (out_samples < 0) {
 739                 fprintf(stderr, "Audio conversion failed.\n");
 740                 exit(1);
 741         }
 742
 743         audio_frame->len += out_samples * bytes_per_sample;
 744 }
 745
 746 VideoFormat FFmpegCapture::construct_video_format(const AVFrame *frame, AVRational video_timebase)
 747 {
 748         VideoFormat video_format;
 749         video_format.width = width;
 750         video_format.height = height;
 751         if (pixel_format == bmusb::PixelFormat_8BitBGRA) {
 752                 video_format.stride = width * 4;
 753         } else if (pixel_format == FFmpegCapture::PixelFormat_NV12) {
 754                 video_format.stride = width;
 755         } else {
 756                 assert(pixel_format == bmusb::PixelFormat_8BitYCbCrPlanar);
 757                 video_format.stride = width;
 758         }
 759         video_format.frame_rate_nom = video_timebase.den;
 760         video_format.frame_rate_den = av_frame_get_pkt_duration(frame) * video_timebase.num;
 761         if (video_format.frame_rate_nom == 0 || video_format.frame_rate_den == 0) {
 762                 // Invalid frame rate.
 763                 video_format.frame_rate_nom = 60;
 764                 video_format.frame_rate_den = 1;
 765         }
 766         video_format.has_signal = true;
 767         video_format.is_connected = true;
 768         return video_format;
 769 }
 770
 771 UniqueFrame FFmpegCapture::make_video_frame(const AVFrame *frame, const string &pathname, bool *error)
 772 {
 773         *error = false;
 774
 775         UniqueFrame video_frame(video_frame_allocator->alloc_frame());
 776         if (video_frame->data == nullptr) {
 777                 return video_frame;
 778         }
 779
 780         if (sws_ctx == nullptr ||
 781             sws_last_width != frame->width ||
 782             sws_last_height != frame->height ||
 783             sws_last_src_format != frame->format) {
 784                 sws_dst_format = decide_dst_format(AVPixelFormat(frame->format), pixel_format);
 785                 sws_ctx.reset(
 786                         sws_getContext(frame->width, frame->height, AVPixelFormat(frame->format),
 787                                 width, height, sws_dst_format,
 788                                 SWS_BICUBIC, nullptr, nullptr, nullptr));
 789                 sws_last_width = frame->width;
 790                 sws_last_height = frame->height;
 791                 sws_last_src_format = frame->format;
 792         }
 793         if (sws_ctx == nullptr) {
 794                 fprintf(stderr, "%s: Could not create scaler context\n", pathname.c_str());
 795                 *error = true;
 796                 return video_frame;
 797         }
 798
 799         uint8_t *pic_data[4] = { nullptr, nullptr, nullptr, nullptr };
 800         int linesizes[4] = { 0, 0, 0, 0 };
 801         if (pixel_format == bmusb::PixelFormat_8BitBGRA) {
 802                 pic_data[0] = video_frame->data;
 803                 linesizes[0] = width * 4;
 804                 video_frame->len = (width * 4) * height;
 805         } else if (pixel_format == PixelFormat_NV12) {
 806                 pic_data[0] = video_frame->data;
 807                 linesizes[0] = width;
 808
 809                 pic_data[1] = pic_data[0] + width * height;
 810                 linesizes[1] = width;
 811
 812                 video_frame->len = (width * 2) * height;
 813
 814                 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(sws_dst_format);
 815                 current_frame_ycbcr_format = decode_ycbcr_format(desc, frame);
 816         } else {
 817                 assert(pixel_format == bmusb::PixelFormat_8BitYCbCrPlanar);
 818                 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(sws_dst_format);
 819
 820                 int chroma_width = AV_CEIL_RSHIFT(int(width), desc->log2_chroma_w);
 821                 int chroma_height = AV_CEIL_RSHIFT(int(height), desc->log2_chroma_h);
 822
 823                 pic_data[0] = video_frame->data;
 824                 linesizes[0] = width;
 825
 826                 pic_data[1] = pic_data[0] + width * height;
 827                 linesizes[1] = chroma_width;
 828
 829                 pic_data[2] = pic_data[1] + chroma_width * chroma_height;
 830                 linesizes[2] = chroma_width;
 831
 832                 video_frame->len = width * height + 2 * chroma_width * chroma_height;
 833
 834                 current_frame_ycbcr_format = decode_ycbcr_format(desc, frame);
 835         }
 836         sws_scale(sws_ctx.get(), frame->data, frame->linesize, 0, frame->height, pic_data, linesizes);
 837
 838         return video_frame;
 839 }
 840
 841 int FFmpegCapture::interrupt_cb_thunk(void *unique)
 842 {
 843         return reinterpret_cast<FFmpegCapture *>(unique)->interrupt_cb();
 844 }
 845
 846 int FFmpegCapture::interrupt_cb()
 847 {
 848         return should_interrupt.load();
 849 }