media/filters/ffmpeg_video_decoder.cc

   1 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include "media/filters/ffmpeg_video_decoder.h"
   6
   7 #include <algorithm>
   8 #include <string>
   9
  10 #include "base/bind.h"
  11 #include "base/callback_helpers.h"
  12 #include "base/command_line.h"
  13 #include "base/location.h"
  14 #include "base/single_thread_task_runner.h"
  15 #include "base/strings/string_number_conversions.h"
  16 #include "media/base/bind_to_current_loop.h"
  17 #include "media/base/decoder_buffer.h"
  18 #include "media/base/limits.h"
  19 #include "media/base/media_switches.h"
  20 #include "media/base/pipeline.h"
  21 #include "media/base/timestamp_constants.h"
  22 #include "media/base/video_frame.h"
  23 #include "media/base/video_util.h"
  24 #include "media/ffmpeg/ffmpeg_common.h"
  25 #include "media/filters/ffmpeg_glue.h"
  26
  27 namespace media {
  28
  29 // Always try to use three threads for video decoding.  There is little reason
  30 // not to since current day CPUs tend to be multi-core and we measured
  31 // performance benefits on older machines such as P4s with hyperthreading.
  32 //
  33 // Handling decoding on separate threads also frees up the pipeline thread to
  34 // continue processing. Although it'd be nice to have the option of a single
  35 // decoding thread, FFmpeg treats having one thread the same as having zero
  36 // threads (i.e., avcodec_decode_video() will execute on the calling thread).
  37 // Yet another reason for having two threads :)
  38 static const int kDecodeThreads = 2;
  39 static const int kMaxDecodeThreads = 16;
  40
  41 // Returns the number of threads given the FFmpeg CodecID. Also inspects the
  42 // command line for a valid --video-threads flag.
  43 static int GetThreadCount(AVCodecID codec_id) {
  44   // Refer to http://crbug.com/93932 for tsan suppressions on decoding.
  45   int decode_threads = kDecodeThreads;
  46
  47   const base::CommandLine* cmd_line = base::CommandLine::ForCurrentProcess();
  48   std::string threads(cmd_line->GetSwitchValueASCII(switches::kVideoThreads));
  49   if (threads.empty() || !base::StringToInt(threads, &decode_threads))
  50     return decode_threads;
  51
  52   decode_threads = std::max(decode_threads, 0);
  53   decode_threads = std::min(decode_threads, kMaxDecodeThreads);
  54   return decode_threads;
  55 }
  56
  57 static int GetVideoBufferImpl(struct AVCodecContext* s,
  58                               AVFrame* frame,
  59                               int flags) {
  60   FFmpegVideoDecoder* decoder = static_cast<FFmpegVideoDecoder*>(s->opaque);
  61   return decoder->GetVideoBuffer(s, frame, flags);
  62 }
  63
  64 static void ReleaseVideoBufferImpl(void* opaque, uint8* data) {
  65   scoped_refptr<VideoFrame> video_frame;
  66   video_frame.swap(reinterpret_cast<VideoFrame**>(&opaque));
  67 }
  68
  69 // static
  70 bool FFmpegVideoDecoder::IsCodecSupported(VideoCodec codec) {
  71   FFmpegGlue::InitializeFFmpeg();
  72   return avcodec_find_decoder(VideoCodecToCodecID(codec)) != nullptr;
  73 }
  74
  75 FFmpegVideoDecoder::FFmpegVideoDecoder(
  76     const scoped_refptr<base::SingleThreadTaskRunner>& task_runner)
  77     : task_runner_(task_runner), state_(kUninitialized),
  78       decode_nalus_(false) {}
  79
  80 int FFmpegVideoDecoder::GetVideoBuffer(struct AVCodecContext* codec_context,
  81                                        AVFrame* frame,
  82                                        int flags) {
  83   // Don't use |codec_context_| here! With threaded decoding,
  84   // it will contain unsynchronized width/height/pix_fmt values,
  85   // whereas |codec_context| contains the current threads's
  86   // updated width/height/pix_fmt, which can change for adaptive
  87   // content.
  88   const VideoPixelFormat format =
  89       AVPixelFormatToVideoPixelFormat(codec_context->pix_fmt);
  90
  91   if (format == PIXEL_FORMAT_UNKNOWN)
  92     return AVERROR(EINVAL);
  93   DCHECK(format == PIXEL_FORMAT_YV12 || format == PIXEL_FORMAT_YV16 ||
  94          format == PIXEL_FORMAT_YV24);
  95
  96   gfx::Size size(codec_context->width, codec_context->height);
  97   const int ret = av_image_check_size(size.width(), size.height(), 0, NULL);
  98   if (ret < 0)
  99     return ret;
 100
 101   gfx::Size natural_size;
 102   if (codec_context->sample_aspect_ratio.num > 0) {
 103     natural_size = GetNaturalSize(size,
 104                                   codec_context->sample_aspect_ratio.num,
 105                                   codec_context->sample_aspect_ratio.den);
 106   } else {
 107     natural_size = config_.natural_size();
 108   }
 109
 110   // FFmpeg has specific requirements on the allocation size of the frame.  The
 111   // following logic replicates FFmpeg's allocation strategy to ensure buffers
 112   // are not overread / overwritten.  See ff_init_buffer_info() for details.
 113   //
 114   // When lowres is non-zero, dimensions should be divided by 2^(lowres), but
 115   // since we don't use this, just DCHECK that it's zero.
 116   DCHECK_EQ(codec_context->lowres, 0);
 117   gfx::Size coded_size(std::max(size.width(), codec_context->coded_width),
 118                        std::max(size.height(), codec_context->coded_height));
 119
 120   if (!VideoFrame::IsValidConfig(format, VideoFrame::STORAGE_UNKNOWN,
 121                                  coded_size, gfx::Rect(size), natural_size)) {
 122     return AVERROR(EINVAL);
 123   }
 124
 125   // FFmpeg expects the initialize allocation to be zero-initialized.  Failure
 126   // to do so can lead to unitialized value usage.  See http://crbug.com/390941
 127   scoped_refptr<VideoFrame> video_frame = frame_pool_.CreateFrame(
 128       format, coded_size, gfx::Rect(size), natural_size, kNoTimestamp());
 129
 130   // Prefer the color space from the codec context. If it's not specified (or is
 131   // set to an unsupported value), fall back on the value from the config.
 132   ColorSpace color_space = AVColorSpaceToColorSpace(codec_context->colorspace,
 133                                                     codec_context->color_range);
 134   if (color_space == COLOR_SPACE_UNSPECIFIED)
 135     color_space = config_.color_space();
 136   video_frame->metadata()->SetInteger(VideoFrameMetadata::COLOR_SPACE,
 137                                       color_space);
 138
 139   for (size_t i = 0; i < VideoFrame::NumPlanes(video_frame->format()); i++) {
 140     frame->data[i] = video_frame->data(i);
 141     frame->linesize[i] = video_frame->stride(i);
 142   }
 143
 144   frame->width = coded_size.width();
 145   frame->height = coded_size.height();
 146   frame->format = codec_context->pix_fmt;
 147   frame->reordered_opaque = codec_context->reordered_opaque;
 148
 149   // Now create an AVBufferRef for the data just allocated. It will own the
 150   // reference to the VideoFrame object.
 151   void* opaque = NULL;
 152   video_frame.swap(reinterpret_cast<VideoFrame**>(&opaque));
 153   frame->buf[0] =
 154       av_buffer_create(frame->data[0],
 155                        VideoFrame::AllocationSize(format, coded_size),
 156                        ReleaseVideoBufferImpl,
 157                        opaque,
 158                        0);
 159   return 0;
 160 }
 161
 162 std::string FFmpegVideoDecoder::GetDisplayName() const {
 163   return "FFmpegVideoDecoder";
 164 }
 165
 166 void FFmpegVideoDecoder::Initialize(const VideoDecoderConfig& config,
 167                                     bool low_delay,
 168                                     const InitCB& init_cb,
 169                                     const OutputCB& output_cb) {
 170   DCHECK(task_runner_->BelongsToCurrentThread());
 171   DCHECK(!config.is_encrypted());
 172   DCHECK(!output_cb.is_null());
 173
 174   FFmpegGlue::InitializeFFmpeg();
 175
 176   config_ = config;
 177   InitCB bound_init_cb = BindToCurrentLoop(init_cb);
 178
 179   if (!config.IsValidConfig() || !ConfigureDecoder(low_delay)) {
 180     bound_init_cb.Run(false);
 181     return;
 182   }
 183
 184   output_cb_ = BindToCurrentLoop(output_cb);
 185
 186   // Success!
 187   state_ = kNormal;
 188   bound_init_cb.Run(true);
 189 }
 190
 191 void FFmpegVideoDecoder::Decode(const scoped_refptr<DecoderBuffer>& buffer,
 192                                 const DecodeCB& decode_cb) {
 193   DCHECK(task_runner_->BelongsToCurrentThread());
 194   DCHECK(buffer.get());
 195   DCHECK(!decode_cb.is_null());
 196   CHECK_NE(state_, kUninitialized);
 197
 198   DecodeCB decode_cb_bound = BindToCurrentLoop(decode_cb);
 199
 200   if (state_ == kError) {
 201     decode_cb_bound.Run(kDecodeError);
 202     return;
 203   }
 204
 205   if (state_ == kDecodeFinished) {
 206     decode_cb_bound.Run(kOk);
 207     return;
 208   }
 209
 210   DCHECK_EQ(state_, kNormal);
 211
 212   // During decode, because reads are issued asynchronously, it is possible to
 213   // receive multiple end of stream buffers since each decode is acked. When the
 214   // first end of stream buffer is read, FFmpeg may still have frames queued
 215   // up in the decoder so we need to go through the decode loop until it stops
 216   // giving sensible data.  After that, the decoder should output empty
 217   // frames.  There are three states the decoder can be in:
 218   //
 219   //   kNormal: This is the starting state. Buffers are decoded. Decode errors
 220   //            are discarded.
 221   //   kDecodeFinished: All calls return empty frames.
 222   //   kError: Unexpected error happened.
 223   //
 224   // These are the possible state transitions.
 225   //
 226   // kNormal -> kDecodeFinished:
 227   //     When EOS buffer is received and the codec has been flushed.
 228   // kNormal -> kError:
 229   //     A decoding error occurs and decoding needs to stop.
 230   // (any state) -> kNormal:
 231   //     Any time Reset() is called.
 232
 233   bool has_produced_frame;
 234   do {
 235     has_produced_frame = false;
 236     if (!FFmpegDecode(buffer, &has_produced_frame)) {
 237       state_ = kError;
 238       decode_cb_bound.Run(kDecodeError);
 239       return;
 240     }
 241     // Repeat to flush the decoder after receiving EOS buffer.
 242   } while (buffer->end_of_stream() && has_produced_frame);
 243
 244   if (buffer->end_of_stream())
 245     state_ = kDecodeFinished;
 246
 247   // VideoDecoderShim expects that |decode_cb| is called only after
 248   // |output_cb_|.
 249   decode_cb_bound.Run(kOk);
 250 }
 251
 252 void FFmpegVideoDecoder::Reset(const base::Closure& closure) {
 253   DCHECK(task_runner_->BelongsToCurrentThread());
 254
 255   avcodec_flush_buffers(codec_context_.get());
 256   state_ = kNormal;
 257   task_runner_->PostTask(FROM_HERE, closure);
 258 }
 259
 260 FFmpegVideoDecoder::~FFmpegVideoDecoder() {
 261   DCHECK(task_runner_->BelongsToCurrentThread());
 262
 263   if (state_ != kUninitialized)
 264     ReleaseFFmpegResources();
 265 }
 266
 267 bool FFmpegVideoDecoder::FFmpegDecode(
 268     const scoped_refptr<DecoderBuffer>& buffer,
 269     bool* has_produced_frame) {
 270   DCHECK(!*has_produced_frame);
 271
 272   // Create a packet for input data.
 273   // Due to FFmpeg API changes we no longer have const read-only pointers.
 274   AVPacket packet;
 275   av_init_packet(&packet);
 276   if (buffer->end_of_stream()) {
 277     packet.data = NULL;
 278     packet.size = 0;
 279   } else {
 280     packet.data = const_cast<uint8*>(buffer->data());
 281     packet.size = buffer->data_size();
 282
 283     // Let FFmpeg handle presentation timestamp reordering.
 284     codec_context_->reordered_opaque = buffer->timestamp().InMicroseconds();
 285   }
 286
 287   int frame_decoded = 0;
 288   int result = avcodec_decode_video2(codec_context_.get(),
 289                                      av_frame_.get(),
 290                                      &frame_decoded,
 291                                      &packet);
 292   // Log the problem if we can't decode a video frame and exit early.
 293   if (result < 0) {
 294     LOG(ERROR) << "Error decoding video: " << buffer->AsHumanReadableString();
 295     return false;
 296   }
 297
 298   // FFmpeg says some codecs might have multiple frames per packet.  Previous
 299   // discussions with rbultje@ indicate this shouldn't be true for the codecs
 300   // we use.
 301   DCHECK_EQ(result, packet.size);
 302
 303   // If no frame was produced then signal that more data is required to
 304   // produce more frames. This can happen under two circumstances:
 305   //   1) Decoder was recently initialized/flushed
 306   //   2) End of stream was reached and all internal frames have been output
 307   if (frame_decoded == 0) {
 308     return true;
 309   }
 310
 311   // TODO(fbarchard): Work around for FFmpeg http://crbug.com/27675
 312   // The decoder is in a bad state and not decoding correctly.
 313   // Checking for NULL avoids a crash in CopyPlane().
 314   if (!av_frame_->data[VideoFrame::kYPlane] ||
 315       !av_frame_->data[VideoFrame::kUPlane] ||
 316       !av_frame_->data[VideoFrame::kVPlane]) {
 317     LOG(ERROR) << "Video frame was produced yet has invalid frame data.";
 318     av_frame_unref(av_frame_.get());
 319     return false;
 320   }
 321
 322   scoped_refptr<VideoFrame> frame =
 323       reinterpret_cast<VideoFrame*>(av_buffer_get_opaque(av_frame_->buf[0]));
 324   frame->set_timestamp(
 325       base::TimeDelta::FromMicroseconds(av_frame_->reordered_opaque));
 326   *has_produced_frame = true;
 327   output_cb_.Run(frame);
 328
 329   av_frame_unref(av_frame_.get());
 330   return true;
 331 }
 332
 333 void FFmpegVideoDecoder::ReleaseFFmpegResources() {
 334   codec_context_.reset();
 335   av_frame_.reset();
 336 }
 337
 338 bool FFmpegVideoDecoder::ConfigureDecoder(bool low_delay) {
 339   // Release existing decoder resources if necessary.
 340   ReleaseFFmpegResources();
 341
 342   // Initialize AVCodecContext structure.
 343   codec_context_.reset(avcodec_alloc_context3(NULL));
 344   VideoDecoderConfigToAVCodecContext(config_, codec_context_.get());
 345
 346   codec_context_->thread_count = GetThreadCount(codec_context_->codec_id);
 347   codec_context_->thread_type = low_delay ? FF_THREAD_SLICE : FF_THREAD_FRAME;
 348   codec_context_->opaque = this;
 349   codec_context_->flags |= CODEC_FLAG_EMU_EDGE;
 350   codec_context_->get_buffer2 = GetVideoBufferImpl;
 351   codec_context_->refcounted_frames = 1;
 352
 353   if (decode_nalus_)
 354     codec_context_->flags2 |= CODEC_FLAG2_CHUNKS;
 355
 356   AVCodec* codec = avcodec_find_decoder(codec_context_->codec_id);
 357   if (!codec || avcodec_open2(codec_context_.get(), codec, NULL) < 0) {
 358     ReleaseFFmpegResources();
 359     return false;
 360   }
 361
 362   av_frame_.reset(av_frame_alloc());
 363   return true;
 364 }
 365
 366 }  // namespace media