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/video_frame.h"
  22 #include "media/base/video_util.h"
  23 #include "media/ffmpeg/ffmpeg_common.h"
  24 #include "media/filters/ffmpeg_glue.h"
  25
  26 namespace media {
  27
  28 // Always try to use three threads for video decoding.  There is little reason
  29 // not to since current day CPUs tend to be multi-core and we measured
  30 // performance benefits on older machines such as P4s with hyperthreading.
  31 //
  32 // Handling decoding on separate threads also frees up the pipeline thread to
  33 // continue processing. Although it'd be nice to have the option of a single
  34 // decoding thread, FFmpeg treats having one thread the same as having zero
  35 // threads (i.e., avcodec_decode_video() will execute on the calling thread).
  36 // Yet another reason for having two threads :)
  37 static const int kDecodeThreads = 2;
  38 static const int kMaxDecodeThreads = 16;
  39
  40 // Returns the number of threads given the FFmpeg CodecID. Also inspects the
  41 // command line for a valid --video-threads flag.
  42 static int GetThreadCount(AVCodecID codec_id) {
  43   // Refer to http://crbug.com/93932 for tsan suppressions on decoding.
  44   int decode_threads = kDecodeThreads;
  45
  46   const base::CommandLine* cmd_line = base::CommandLine::ForCurrentProcess();
  47   std::string threads(cmd_line->GetSwitchValueASCII(switches::kVideoThreads));
  48   if (threads.empty() || !base::StringToInt(threads, &decode_threads))
  49     return decode_threads;
  50
  51   decode_threads = std::max(decode_threads, 0);
  52   decode_threads = std::min(decode_threads, kMaxDecodeThreads);
  53   return decode_threads;
  54 }
  55
  56 static int GetVideoBufferImpl(struct AVCodecContext* s,
  57                               AVFrame* frame,
  58                               int flags) {
  59   FFmpegVideoDecoder* decoder = static_cast<FFmpegVideoDecoder*>(s->opaque);
  60   return decoder->GetVideoBuffer(s, frame, flags);
  61 }
  62
  63 static void ReleaseVideoBufferImpl(void* opaque, uint8* data) {
  64   scoped_refptr<VideoFrame> video_frame;
  65   video_frame.swap(reinterpret_cast<VideoFrame**>(&opaque));
  66 }
  67
  68 static size_t RoundUp(size_t value, size_t alignment) {
  69   // Check that |alignment| is a power of 2.
  70   DCHECK((alignment + (alignment - 1)) == (alignment | (alignment - 1)));
  71   return ((value + (alignment - 1)) & ~(alignment - 1));
  72 }
  73
  74 FFmpegVideoDecoder::FFmpegVideoDecoder(
  75     const scoped_refptr<base::SingleThreadTaskRunner>& task_runner)
  76     : task_runner_(task_runner), state_(kUninitialized),
  77       decode_nalus_(false) {}
  78
  79 int FFmpegVideoDecoder::GetVideoBuffer(struct AVCodecContext* codec_context,
  80                                        AVFrame* frame,
  81                                        int flags) {
  82   // Don't use |codec_context_| here! With threaded decoding,
  83   // it will contain unsynchronized width/height/pix_fmt values,
  84   // whereas |codec_context| contains the current threads's
  85   // updated width/height/pix_fmt, which can change for adaptive
  86   // content.
  87   const VideoPixelFormat format =
  88       PixelFormatToVideoPixelFormat(codec_context->pix_fmt);
  89
  90   if (format == PIXEL_FORMAT_UNKNOWN)
  91     return AVERROR(EINVAL);
  92   DCHECK(format == PIXEL_FORMAT_YV12 || format == PIXEL_FORMAT_YV16 ||
  93          format == PIXEL_FORMAT_YV24);
  94
  95   gfx::Size size(codec_context->width, codec_context->height);
  96   const int ret = av_image_check_size(size.width(), size.height(), 0, NULL);
  97   if (ret < 0)
  98     return ret;
  99
 100   gfx::Size natural_size;
 101   if (codec_context->sample_aspect_ratio.num > 0) {
 102     natural_size = GetNaturalSize(size,
 103                                   codec_context->sample_aspect_ratio.num,
 104                                   codec_context->sample_aspect_ratio.den);
 105   } else {
 106     natural_size = config_.natural_size();
 107   }
 108
 109   // FFmpeg has specific requirements on the allocation size of the frame.  The
 110   // following logic replicates FFmpeg's allocation strategy to ensure buffers
 111   // are not overread / overwritten.  See ff_init_buffer_info() for details.
 112   //
 113   // When lowres is non-zero, dimensions should be divided by 2^(lowres), but
 114   // since we don't use this, just DCHECK that it's zero.
 115   //
 116   // Always round up to a multiple of two to match VideoFrame restrictions on
 117   // frame alignment.
 118   DCHECK_EQ(codec_context->lowres, 0);
 119   gfx::Size coded_size(
 120       RoundUp(std::max(size.width(), codec_context->coded_width), 2),
 121       RoundUp(std::max(size.height(), codec_context->coded_height), 2));
 122
 123   if (!VideoFrame::IsValidConfig(format, VideoFrame::STORAGE_UNKNOWN,
 124                                  coded_size, gfx::Rect(size), natural_size)) {
 125     return AVERROR(EINVAL);
 126   }
 127
 128   scoped_refptr<VideoFrame> video_frame = frame_pool_.CreateFrame(
 129       format, coded_size, gfx::Rect(size), natural_size, kNoTimestamp());
 130 #if defined(MEMORY_SANITIZER)
 131   MSAN_UNPOISON(video_frame->data(0),
 132                 VideoFrame::AllocationSize(format, coded_size));
 133 #endif
 134
 135   // Prefer the color space from the codec context. If it's not specified (or is
 136   // set to an unsupported value), fall back on the value from the config.
 137   ColorSpace color_space = AVColorSpaceToColorSpace(codec_context->colorspace,
 138                                                     codec_context->color_range);
 139   if (color_space == COLOR_SPACE_UNSPECIFIED)
 140     color_space = config_.color_space();
 141   video_frame->metadata()->SetInteger(VideoFrameMetadata::COLOR_SPACE,
 142                                       color_space);
 143
 144   for (int i = 0; i < 3; i++) {
 145     frame->data[i] = video_frame->data(i);
 146     frame->linesize[i] = video_frame->stride(i);
 147   }
 148
 149   frame->width = coded_size.width();
 150   frame->height = coded_size.height();
 151   frame->format = codec_context->pix_fmt;
 152   frame->reordered_opaque = codec_context->reordered_opaque;
 153
 154   // Now create an AVBufferRef for the data just allocated. It will own the
 155   // reference to the VideoFrame object.
 156   void* opaque = NULL;
 157   video_frame.swap(reinterpret_cast<VideoFrame**>(&opaque));
 158   frame->buf[0] =
 159       av_buffer_create(frame->data[0],
 160                        VideoFrame::AllocationSize(format, coded_size),
 161                        ReleaseVideoBufferImpl,
 162                        opaque,
 163                        0);
 164   return 0;
 165 }
 166
 167 std::string FFmpegVideoDecoder::GetDisplayName() const {
 168   return "FFmpegVideoDecoder";
 169 }
 170
 171 void FFmpegVideoDecoder::Initialize(const VideoDecoderConfig& config,
 172                                     bool low_delay,
 173                                     const InitCB& init_cb,
 174                                     const OutputCB& output_cb) {
 175   DCHECK(task_runner_->BelongsToCurrentThread());
 176   DCHECK(!config.is_encrypted());
 177   DCHECK(!output_cb.is_null());
 178
 179   FFmpegGlue::InitializeFFmpeg();
 180
 181   config_ = config;
 182   InitCB bound_init_cb = BindToCurrentLoop(init_cb);
 183
 184   if (!config.IsValidConfig() || !ConfigureDecoder(low_delay)) {
 185     bound_init_cb.Run(false);
 186     return;
 187   }
 188
 189   output_cb_ = BindToCurrentLoop(output_cb);
 190
 191   // Success!
 192   state_ = kNormal;
 193   bound_init_cb.Run(true);
 194 }
 195
 196 void FFmpegVideoDecoder::Decode(const scoped_refptr<DecoderBuffer>& buffer,
 197                                 const DecodeCB& decode_cb) {
 198   DCHECK(task_runner_->BelongsToCurrentThread());
 199   DCHECK(buffer.get());
 200   DCHECK(!decode_cb.is_null());
 201   CHECK_NE(state_, kUninitialized);
 202
 203   DecodeCB decode_cb_bound = BindToCurrentLoop(decode_cb);
 204
 205   if (state_ == kError) {
 206     decode_cb_bound.Run(kDecodeError);
 207     return;
 208   }
 209
 210   if (state_ == kDecodeFinished) {
 211     decode_cb_bound.Run(kOk);
 212     return;
 213   }
 214
 215   DCHECK_EQ(state_, kNormal);
 216
 217   // During decode, because reads are issued asynchronously, it is possible to
 218   // receive multiple end of stream buffers since each decode is acked. When the
 219   // first end of stream buffer is read, FFmpeg may still have frames queued
 220   // up in the decoder so we need to go through the decode loop until it stops
 221   // giving sensible data.  After that, the decoder should output empty
 222   // frames.  There are three states the decoder can be in:
 223   //
 224   //   kNormal: This is the starting state. Buffers are decoded. Decode errors
 225   //            are discarded.
 226   //   kDecodeFinished: All calls return empty frames.
 227   //   kError: Unexpected error happened.
 228   //
 229   // These are the possible state transitions.
 230   //
 231   // kNormal -> kDecodeFinished:
 232   //     When EOS buffer is received and the codec has been flushed.
 233   // kNormal -> kError:
 234   //     A decoding error occurs and decoding needs to stop.
 235   // (any state) -> kNormal:
 236   //     Any time Reset() is called.
 237
 238   bool has_produced_frame;
 239   do {
 240     has_produced_frame = false;
 241     if (!FFmpegDecode(buffer, &has_produced_frame)) {
 242       state_ = kError;
 243       decode_cb_bound.Run(kDecodeError);
 244       return;
 245     }
 246     // Repeat to flush the decoder after receiving EOS buffer.
 247   } while (buffer->end_of_stream() && has_produced_frame);
 248
 249   if (buffer->end_of_stream())
 250     state_ = kDecodeFinished;
 251
 252   // VideoDecoderShim expects that |decode_cb| is called only after
 253   // |output_cb_|.
 254   decode_cb_bound.Run(kOk);
 255 }
 256
 257 void FFmpegVideoDecoder::Reset(const base::Closure& closure) {
 258   DCHECK(task_runner_->BelongsToCurrentThread());
 259
 260   avcodec_flush_buffers(codec_context_.get());
 261   state_ = kNormal;
 262   task_runner_->PostTask(FROM_HERE, closure);
 263 }
 264
 265 FFmpegVideoDecoder::~FFmpegVideoDecoder() {
 266   DCHECK(task_runner_->BelongsToCurrentThread());
 267
 268   if (state_ != kUninitialized)
 269     ReleaseFFmpegResources();
 270 }
 271
 272 bool FFmpegVideoDecoder::FFmpegDecode(
 273     const scoped_refptr<DecoderBuffer>& buffer,
 274     bool* has_produced_frame) {
 275   DCHECK(!*has_produced_frame);
 276
 277   // Create a packet for input data.
 278   // Due to FFmpeg API changes we no longer have const read-only pointers.
 279   AVPacket packet;
 280   av_init_packet(&packet);
 281   if (buffer->end_of_stream()) {
 282     packet.data = NULL;
 283     packet.size = 0;
 284   } else {
 285     packet.data = const_cast<uint8*>(buffer->data());
 286     packet.size = buffer->data_size();
 287
 288     // Let FFmpeg handle presentation timestamp reordering.
 289     codec_context_->reordered_opaque = buffer->timestamp().InMicroseconds();
 290   }
 291
 292   int frame_decoded = 0;
 293   int result = avcodec_decode_video2(codec_context_.get(),
 294                                      av_frame_.get(),
 295                                      &frame_decoded,
 296                                      &packet);
 297   // Log the problem if we can't decode a video frame and exit early.
 298   if (result < 0) {
 299     LOG(ERROR) << "Error decoding video: " << buffer->AsHumanReadableString();
 300     return false;
 301   }
 302
 303   // FFmpeg says some codecs might have multiple frames per packet.  Previous
 304   // discussions with rbultje@ indicate this shouldn't be true for the codecs
 305   // we use.
 306   DCHECK_EQ(result, packet.size);
 307
 308   // If no frame was produced then signal that more data is required to
 309   // produce more frames. This can happen under two circumstances:
 310   //   1) Decoder was recently initialized/flushed
 311   //   2) End of stream was reached and all internal frames have been output
 312   if (frame_decoded == 0) {
 313     return true;
 314   }
 315
 316   // TODO(fbarchard): Work around for FFmpeg http://crbug.com/27675
 317   // The decoder is in a bad state and not decoding correctly.
 318   // Checking for NULL avoids a crash in CopyPlane().
 319   if (!av_frame_->data[VideoFrame::kYPlane] ||
 320       !av_frame_->data[VideoFrame::kUPlane] ||
 321       !av_frame_->data[VideoFrame::kVPlane]) {
 322     LOG(ERROR) << "Video frame was produced yet has invalid frame data.";
 323     av_frame_unref(av_frame_.get());
 324     return false;
 325   }
 326
 327   scoped_refptr<VideoFrame> frame =
 328       reinterpret_cast<VideoFrame*>(av_buffer_get_opaque(av_frame_->buf[0]));
 329   frame->set_timestamp(
 330       base::TimeDelta::FromMicroseconds(av_frame_->reordered_opaque));
 331   *has_produced_frame = true;
 332   output_cb_.Run(frame);
 333
 334   av_frame_unref(av_frame_.get());
 335   return true;
 336 }
 337
 338 void FFmpegVideoDecoder::ReleaseFFmpegResources() {
 339   codec_context_.reset();
 340   av_frame_.reset();
 341 }
 342
 343 bool FFmpegVideoDecoder::ConfigureDecoder(bool low_delay) {
 344   // Release existing decoder resources if necessary.
 345   ReleaseFFmpegResources();
 346
 347   // Initialize AVCodecContext structure.
 348   codec_context_.reset(avcodec_alloc_context3(NULL));
 349   VideoDecoderConfigToAVCodecContext(config_, codec_context_.get());
 350
 351   codec_context_->thread_count = GetThreadCount(codec_context_->codec_id);
 352   codec_context_->thread_type = low_delay ? FF_THREAD_SLICE : FF_THREAD_FRAME;
 353   codec_context_->opaque = this;
 354   codec_context_->flags |= CODEC_FLAG_EMU_EDGE;
 355   codec_context_->get_buffer2 = GetVideoBufferImpl;
 356   codec_context_->refcounted_frames = 1;
 357
 358   if (decode_nalus_)
 359     codec_context_->flags2 |= CODEC_FLAG2_CHUNKS;
 360
 361   AVCodec* codec = avcodec_find_decoder(codec_context_->codec_id);
 362   if (!codec || avcodec_open2(codec_context_.get(), codec, NULL) < 0) {
 363     ReleaseFFmpegResources();
 364     return false;
 365   }
 366
 367   av_frame_.reset(av_frame_alloc());
 368   return true;
 369 }
 370
 371 }  // namespace media