| // Copyright 2019 The Chromium Authors. All rights reserved. |
| // Use of this source code is governed by a BSD-style license that can be |
| // found in the LICENSE file. |
| |
| #include "media/gpu/v4l2/v4l2_video_decoder_backend_stateless.h" |
| |
| #include <fcntl.h> |
| #include <linux/media.h> |
| #include <sys/ioctl.h> |
| |
| #include <memory> |
| |
| #include "base/bind.h" |
| #include "base/callback_helpers.h" |
| #include "base/containers/contains.h" |
| #include "base/metrics/histogram_macros.h" |
| #include "base/numerics/safe_conversions.h" |
| #include "base/posix/eintr_wrapper.h" |
| #include "base/sequenced_task_runner.h" |
| #include "media/base/decode_status.h" |
| #include "media/base/video_codecs.h" |
| #include "media/base/video_frame.h" |
| #include "media/gpu/accelerated_video_decoder.h" |
| #include "media/gpu/chromeos/dmabuf_video_frame_pool.h" |
| #include "media/gpu/macros.h" |
| #include "media/gpu/v4l2/v4l2_device.h" |
| #include "media/gpu/v4l2/v4l2_video_decoder_delegate_h264.h" |
| #include "media/gpu/v4l2/v4l2_video_decoder_delegate_h264_legacy.h" |
| #include "media/gpu/v4l2/v4l2_video_decoder_delegate_vp8.h" |
| #include "media/gpu/v4l2/v4l2_video_decoder_delegate_vp8_legacy.h" |
| #include "media/gpu/v4l2/v4l2_video_decoder_delegate_vp9_chromium.h" |
| #include "media/gpu/v4l2/v4l2_video_decoder_delegate_vp9_legacy.h" |
| |
| namespace media { |
| |
| namespace { |
| |
| // Size of the timestamp cache, needs to be large enough for frame-reordering. |
| constexpr size_t kTimestampCacheSize = 128; |
| // Number of requests to allocate for submitting input buffers, if requests |
| // are used. |
| |
| } // namespace |
| |
| struct V4L2StatelessVideoDecoderBackend::OutputRequest { |
| static OutputRequest Surface(scoped_refptr<V4L2DecodeSurface> s, |
| base::TimeDelta t) { |
| return OutputRequest(std::move(s), t); |
| } |
| |
| static OutputRequest FlushFence() { return OutputRequest(kFlushFence); } |
| |
| static OutputRequest ChangeResolutionFence() { |
| return OutputRequest(kChangeResolutionFence); |
| } |
| |
| bool IsReady() const { |
| return (type != OutputRequestType::kSurface) || surface->decoded(); |
| } |
| |
| // Allow move, but not copy. |
| OutputRequest(OutputRequest&&) = default; |
| |
| enum OutputRequestType { |
| // The surface to be outputted. |
| kSurface, |
| // The fence to indicate the flush request. |
| kFlushFence, |
| // The fence to indicate resolution change request. |
| kChangeResolutionFence, |
| }; |
| |
| // The type of the request. |
| const OutputRequestType type; |
| // The surface to be outputted. |
| scoped_refptr<V4L2DecodeSurface> surface; |
| // The timestamp of the output frame. Because a surface might be outputted |
| // multiple times with different timestamp, we need to store timestamp out of |
| // surface. |
| base::TimeDelta timestamp; |
| |
| private: |
| OutputRequest(scoped_refptr<V4L2DecodeSurface> s, base::TimeDelta t) |
| : type(kSurface), surface(std::move(s)), timestamp(t) {} |
| explicit OutputRequest(OutputRequestType t) : type(t) {} |
| |
| DISALLOW_COPY_AND_ASSIGN(OutputRequest); |
| }; |
| |
| V4L2StatelessVideoDecoderBackend::DecodeRequest::DecodeRequest( |
| scoped_refptr<DecoderBuffer> buf, |
| VideoDecoder::DecodeCB cb, |
| int32_t id) |
| : buffer(std::move(buf)), decode_cb(std::move(cb)), bitstream_id(id) {} |
| |
| V4L2StatelessVideoDecoderBackend::DecodeRequest::DecodeRequest( |
| DecodeRequest&&) = default; |
| V4L2StatelessVideoDecoderBackend::DecodeRequest& |
| V4L2StatelessVideoDecoderBackend::DecodeRequest::operator=(DecodeRequest&&) = |
| default; |
| |
| V4L2StatelessVideoDecoderBackend::DecodeRequest::~DecodeRequest() = default; |
| |
| V4L2StatelessVideoDecoderBackend::V4L2StatelessVideoDecoderBackend( |
| Client* const client, |
| scoped_refptr<V4L2Device> device, |
| VideoCodecProfile profile, |
| scoped_refptr<base::SequencedTaskRunner> task_runner) |
| : V4L2VideoDecoderBackend(client, std::move(device)), |
| profile_(profile), |
| bitstream_id_to_timestamp_(kTimestampCacheSize), |
| task_runner_(task_runner) { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| |
| weak_this_ = weak_this_factory_.GetWeakPtr(); |
| } |
| |
| V4L2StatelessVideoDecoderBackend::~V4L2StatelessVideoDecoderBackend() { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DCHECK(surfaces_at_device_.empty()); |
| |
| if (!output_request_queue_.empty() || flush_cb_ || current_decode_request_ || |
| !decode_request_queue_.empty()) { |
| VLOGF(1) << "Should not destroy backend during pending decode!"; |
| } |
| } |
| |
| bool V4L2StatelessVideoDecoderBackend::Initialize() { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| |
| if (!IsSupportedProfile(profile_)) { |
| VLOGF(1) << "Unsupported profile " << GetProfileName(profile_); |
| return false; |
| } |
| |
| if (!CreateAvd()) |
| return false; |
| |
| if (input_queue_->SupportsRequests()) { |
| requests_queue_ = device_->GetRequestsQueue(); |
| if (requests_queue_ == nullptr) |
| return false; |
| } |
| |
| return true; |
| } |
| |
| // static |
| void V4L2StatelessVideoDecoderBackend::ReuseOutputBufferThunk( |
| scoped_refptr<base::SequencedTaskRunner> task_runner, |
| absl::optional<base::WeakPtr<V4L2StatelessVideoDecoderBackend>> weak_this, |
| V4L2ReadableBufferRef buffer) { |
| DVLOGF(3); |
| DCHECK(weak_this); |
| |
| if (task_runner->RunsTasksInCurrentSequence()) { |
| if (*weak_this) { |
| (*weak_this)->ReuseOutputBuffer(std::move(buffer)); |
| } |
| } else { |
| task_runner->PostTask( |
| FROM_HERE, |
| base::BindOnce(&V4L2StatelessVideoDecoderBackend::ReuseOutputBuffer, |
| *weak_this, std::move(buffer))); |
| } |
| } |
| |
| void V4L2StatelessVideoDecoderBackend::ReuseOutputBuffer( |
| V4L2ReadableBufferRef buffer) { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DVLOGF(3) << "Reuse output surface #" << buffer->BufferId(); |
| |
| // Resume decoding in case of ran out of surface. |
| if (pause_reason_ == PauseReason::kRanOutOfSurfaces) { |
| pause_reason_ = PauseReason::kNone; |
| task_runner_->PostTask( |
| FROM_HERE, |
| base::BindOnce(&V4L2StatelessVideoDecoderBackend::DoDecodeWork, |
| weak_this_)); |
| } |
| } |
| |
| void V4L2StatelessVideoDecoderBackend::OnOutputBufferDequeued( |
| V4L2ReadableBufferRef dequeued_buffer) { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| |
| // Mark the output buffer decoded, and try to output surface. |
| DCHECK(!surfaces_at_device_.empty()); |
| auto surface = std::move(surfaces_at_device_.front()); |
| DCHECK_EQ(static_cast<size_t>(surface->output_record()), |
| dequeued_buffer->BufferId()); |
| surfaces_at_device_.pop(); |
| |
| surface->SetDecoded(); |
| |
| auto reuse_buffer_cb = |
| base::BindOnce(&V4L2StatelessVideoDecoderBackend::ReuseOutputBufferThunk, |
| task_runner_, weak_this_, std::move(dequeued_buffer)); |
| if (output_queue_->GetMemoryType() == V4L2_MEMORY_MMAP) { |
| // Keep a reference to the V4L2 buffer until the frame is reused, because |
| // the frame is backed up by the V4L2 buffer's memory. |
| surface->video_frame()->AddDestructionObserver(std::move(reuse_buffer_cb)); |
| } else { |
| // Keep a reference to the V4L2 buffer until the buffer is reused. The |
| // reason for this is that the config store uses V4L2 buffer IDs to |
| // reference frames, therefore we cannot reuse the same V4L2 buffer ID for |
| // another decode operation until all references to that frame are gone. |
| // Request API does not have this limitation, so we can probably remove this |
| // after config store is gone. |
| surface->SetReleaseCallback(std::move(reuse_buffer_cb)); |
| } |
| |
| PumpOutputSurfaces(); |
| |
| // If we were waiting for an output buffer to be available, schedule a |
| // decode task. |
| if (pause_reason_ == PauseReason::kWaitSubFrameDecoded) { |
| pause_reason_ = PauseReason::kNone; |
| task_runner_->PostTask( |
| FROM_HERE, |
| base::BindOnce(&V4L2StatelessVideoDecoderBackend::DoDecodeWork, |
| weak_this_)); |
| } |
| } |
| |
| scoped_refptr<V4L2DecodeSurface> |
| V4L2StatelessVideoDecoderBackend::CreateSurface() { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DVLOGF(4); |
| |
| // Request V4L2 input and output buffers. |
| auto input_buf = input_queue_->GetFreeBuffer(); |
| auto output_buf = output_queue_->GetFreeBuffer(); |
| if (!input_buf || !output_buf) { |
| DVLOGF(3) << "There is no free V4L2 buffer."; |
| return nullptr; |
| } |
| |
| DmabufVideoFramePool* pool = client_->GetVideoFramePool(); |
| scoped_refptr<VideoFrame> frame; |
| if (!pool) { |
| // Get VideoFrame from the V4L2 buffer because now we allocate from V4L2 |
| // driver via MMAP. The VideoFrame received from V4L2 buffer will remain |
| // until deallocating V4L2Queue. But we need to know when the buffer is not |
| // used by the client. So we wrap the frame here. |
| scoped_refptr<VideoFrame> origin_frame = output_buf->GetVideoFrame(); |
| frame = VideoFrame::WrapVideoFrame(origin_frame, origin_frame->format(), |
| origin_frame->visible_rect(), |
| origin_frame->natural_size()); |
| } else { |
| // Try to get VideoFrame from the pool. |
| frame = pool->GetFrame(); |
| if (!frame) { |
| // We allocate the same number of output buffer slot in V4L2 device and |
| // the output VideoFrame. If there is free output buffer slot but no free |
| // VideoFrame, it means the VideoFrame is not released at client |
| // side. Post DoDecodeWork when the pool has available frames. |
| DVLOGF(3) << "There is no available VideoFrame."; |
| pool->NotifyWhenFrameAvailable(base::BindOnce( |
| base::IgnoreResult(&base::SequencedTaskRunner::PostTask), |
| task_runner_, FROM_HERE, |
| base::BindOnce(&V4L2StatelessVideoDecoderBackend::DoDecodeWork, |
| weak_this_))); |
| return nullptr; |
| } |
| } |
| |
| scoped_refptr<V4L2DecodeSurface> dec_surface; |
| if (input_queue_->SupportsRequests()) { |
| absl::optional<V4L2RequestRef> request_ref = |
| requests_queue_->GetFreeRequest(); |
| if (!request_ref) { |
| DVLOGF(1) << "Could not get free request."; |
| return nullptr; |
| } |
| |
| dec_surface = new V4L2RequestDecodeSurface( |
| std::move(*input_buf), std::move(*output_buf), std::move(frame), |
| std::move(*request_ref)); |
| } else { |
| dec_surface = new V4L2ConfigStoreDecodeSurface( |
| std::move(*input_buf), std::move(*output_buf), std::move(frame)); |
| } |
| |
| return dec_surface; |
| } |
| |
| bool V4L2StatelessVideoDecoderBackend::SubmitSlice( |
| V4L2DecodeSurface* dec_surface, |
| const uint8_t* data, |
| size_t size) { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DVLOGF(3); |
| |
| size_t plane_size = dec_surface->input_buffer().GetPlaneSize(0); |
| size_t bytes_used = dec_surface->input_buffer().GetPlaneBytesUsed(0); |
| if (size > plane_size - bytes_used) { |
| VLOGF(1) << "The size of submitted slice(" << size |
| << ") is larger than the remaining buffer size(" |
| << plane_size - bytes_used << "). Plane size is " << plane_size; |
| client_->OnBackendError(); |
| return false; |
| } |
| |
| void* mapping = dec_surface->input_buffer().GetPlaneMapping(0); |
| memcpy(reinterpret_cast<uint8_t*>(mapping) + bytes_used, data, size); |
| dec_surface->input_buffer().SetPlaneBytesUsed(0, bytes_used + size); |
| return true; |
| } |
| |
| void V4L2StatelessVideoDecoderBackend::DecodeSurface( |
| scoped_refptr<V4L2DecodeSurface> dec_surface) { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DVLOGF(3); |
| |
| if (!dec_surface->Submit()) { |
| VLOGF(1) << "Error while submitting frame for decoding!"; |
| client_->OnBackendError(); |
| return; |
| } |
| |
| surfaces_at_device_.push(std::move(dec_surface)); |
| } |
| |
| void V4L2StatelessVideoDecoderBackend::SurfaceReady( |
| scoped_refptr<V4L2DecodeSurface> dec_surface, |
| int32_t bitstream_id, |
| const gfx::Rect& visible_rect, |
| const VideoColorSpace& /* color_space */) { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DVLOGF(3); |
| |
| // Find the timestamp associated with |bitstream_id|. It's possible that a |
| // surface is output multiple times for different |bitstream_id|s (e.g. VP9 |
| // show_existing_frame feature). This means we need to output the same frame |
| // again with a different timestamp. |
| // On some rare occasions it's also possible that a single DecoderBuffer |
| // produces multiple surfaces with the same |bitstream_id|, so we shouldn't |
| // remove the timestamp from the cache. |
| const auto it = bitstream_id_to_timestamp_.Peek(bitstream_id); |
| DCHECK(it != bitstream_id_to_timestamp_.end()); |
| base::TimeDelta timestamp = it->second; |
| |
| dec_surface->SetVisibleRect(visible_rect); |
| output_request_queue_.push( |
| OutputRequest::Surface(std::move(dec_surface), timestamp)); |
| PumpOutputSurfaces(); |
| } |
| |
| void V4L2StatelessVideoDecoderBackend::EnqueueDecodeTask( |
| scoped_refptr<DecoderBuffer> buffer, |
| VideoDecoder::DecodeCB decode_cb, |
| int32_t bitstream_id) { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| |
| if (!buffer->end_of_stream()) { |
| bitstream_id_to_timestamp_.Put(bitstream_id, buffer->timestamp()); |
| } |
| |
| decode_request_queue_.push( |
| DecodeRequest(std::move(buffer), std::move(decode_cb), bitstream_id)); |
| |
| // If we are already decoding, then we don't need to pump again. |
| if (!current_decode_request_) |
| DoDecodeWork(); |
| } |
| |
| void V4L2StatelessVideoDecoderBackend::DoDecodeWork() { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| |
| if (!client_->IsDecoding()) |
| return; |
| |
| if (!PumpDecodeTask()) |
| client_->OnBackendError(); |
| } |
| |
| bool V4L2StatelessVideoDecoderBackend::PumpDecodeTask() { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DVLOGF(3) << " Number of Decode requests: " << decode_request_queue_.size(); |
| |
| pause_reason_ = PauseReason::kNone; |
| while (true) { |
| switch (avd_->Decode()) { |
| case AcceleratedVideoDecoder::kConfigChange: |
| if (avd_->GetBitDepth() != 8u) { |
| VLOGF(2) << "Unsupported bit depth: " |
| << base::strict_cast<int>(avd_->GetBitDepth()); |
| return false; |
| } |
| |
| if (profile_ != avd_->GetProfile()) { |
| DVLOGF(3) << "Profile is changed: " << profile_ << " -> " |
| << avd_->GetProfile(); |
| if (!IsSupportedProfile(avd_->GetProfile())) { |
| VLOGF(2) << "Unsupported profile: " << avd_->GetProfile(); |
| return false; |
| } |
| |
| profile_ = avd_->GetProfile(); |
| } |
| |
| if (pic_size_ == avd_->GetPicSize()) { |
| // There is no need to do anything in V4L2 API when only a profile is |
| // changed. |
| DVLOGF(3) << "Only profile is changed. No need to do anything."; |
| continue; |
| } |
| |
| DVLOGF(3) << "Need to change resolution. Pause decoding."; |
| client_->InitiateFlush(); |
| |
| output_request_queue_.push(OutputRequest::ChangeResolutionFence()); |
| PumpOutputSurfaces(); |
| return true; |
| |
| case AcceleratedVideoDecoder::kRanOutOfStreamData: |
| // Current decode request is finished processing. |
| if (current_decode_request_) { |
| encoding_timestamps_[current_decode_request_->buffer->timestamp() |
| .InMilliseconds()] = base::TimeTicks::Now(); |
| |
| DCHECK(current_decode_request_->decode_cb); |
| std::move(current_decode_request_->decode_cb).Run(DecodeStatus::OK); |
| current_decode_request_ = absl::nullopt; |
| } |
| |
| // Process next decode request. |
| if (decode_request_queue_.empty()) |
| return true; |
| current_decode_request_ = std::move(decode_request_queue_.front()); |
| decode_request_queue_.pop(); |
| |
| if (current_decode_request_->buffer->end_of_stream()) { |
| if (!avd_->Flush()) { |
| VLOGF(1) << "Failed flushing the decoder."; |
| return false; |
| } |
| // Put the decoder in an idle state, ready to resume. |
| avd_->Reset(); |
| |
| client_->InitiateFlush(); |
| DCHECK(!flush_cb_); |
| flush_cb_ = std::move(current_decode_request_->decode_cb); |
| |
| output_request_queue_.push(OutputRequest::FlushFence()); |
| PumpOutputSurfaces(); |
| current_decode_request_ = absl::nullopt; |
| return true; |
| } |
| |
| avd_->SetStream(current_decode_request_->bitstream_id, |
| *current_decode_request_->buffer); |
| break; |
| |
| case AcceleratedVideoDecoder::kRanOutOfSurfaces: |
| DVLOGF(3) << "Ran out of surfaces. Resume when buffer is returned."; |
| pause_reason_ = PauseReason::kRanOutOfSurfaces; |
| return true; |
| |
| case AcceleratedVideoDecoder::kNeedContextUpdate: |
| DVLOGF(3) << "Awaiting context update"; |
| pause_reason_ = PauseReason::kWaitSubFrameDecoded; |
| return true; |
| |
| case AcceleratedVideoDecoder::kDecodeError: |
| DVLOGF(3) << "Error decoding stream"; |
| return false; |
| |
| case AcceleratedVideoDecoder::kTryAgain: |
| NOTREACHED() << "Should not reach here unless this class accepts " |
| "encrypted streams."; |
| DVLOGF(4) << "No key for decoding stream."; |
| return false; |
| } |
| } |
| } |
| |
| void V4L2StatelessVideoDecoderBackend::PumpOutputSurfaces() { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DVLOGF(3) << "Number of display surfaces: " << output_request_queue_.size(); |
| |
| bool resume_decode = false; |
| while (!output_request_queue_.empty()) { |
| if (!output_request_queue_.front().IsReady()) { |
| DVLOGF(3) << "The first surface is not ready yet."; |
| // It is possible that that V4L2 buffers for this output surface are not |
| // even queued yet. Make sure that avd_->Decode() is called to continue |
| // that work and prevent the decoding thread from starving. |
| resume_decode = true; |
| break; |
| } |
| |
| OutputRequest request = std::move(output_request_queue_.front()); |
| output_request_queue_.pop(); |
| switch (request.type) { |
| case OutputRequest::kFlushFence: |
| DCHECK(output_request_queue_.empty()); |
| DVLOGF(2) << "Flush finished."; |
| std::move(flush_cb_).Run(DecodeStatus::OK); |
| resume_decode = true; |
| client_->CompleteFlush(); |
| break; |
| |
| case OutputRequest::kChangeResolutionFence: |
| DCHECK(output_request_queue_.empty()); |
| ChangeResolution(); |
| break; |
| |
| case OutputRequest::kSurface: |
| scoped_refptr<V4L2DecodeSurface> surface = std::move(request.surface); |
| |
| DCHECK(surface->video_frame()); |
| client_->OutputFrame(surface->video_frame(), surface->visible_rect(), |
| request.timestamp); |
| |
| { |
| const int64_t flat_timestamp = request.timestamp.InMilliseconds(); |
| // TODO(b/190615065) |flat_timestamp| might be repeated with H.264 |
| // bitstreams, investigate why, and change the if() to DCHECK(). |
| if (base::Contains(encoding_timestamps_, flat_timestamp)) { |
| UMA_HISTOGRAM_TIMES( |
| "Media.PlatformVideoDecoding.Decode", |
| base::TimeTicks::Now() - encoding_timestamps_[flat_timestamp]); |
| encoding_timestamps_.erase(flat_timestamp); |
| } |
| } |
| |
| break; |
| } |
| } |
| |
| if (resume_decode) { |
| task_runner_->PostTask( |
| FROM_HERE, |
| base::BindOnce(&V4L2StatelessVideoDecoderBackend::DoDecodeWork, |
| weak_this_)); |
| } |
| } |
| |
| void V4L2StatelessVideoDecoderBackend::ChangeResolution() { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| // We change resolution after outputting all pending surfaces, there should |
| // be no V4L2DecodeSurface left. |
| DCHECK(surfaces_at_device_.empty()); |
| DCHECK(output_request_queue_.empty()); |
| |
| size_t num_output_frames = avd_->GetRequiredNumOfPictures(); |
| gfx::Rect visible_rect = avd_->GetVisibleRect(); |
| gfx::Size pic_size = avd_->GetPicSize(); |
| // Set output format with the new resolution. |
| DCHECK(!pic_size.IsEmpty()); |
| DVLOGF(3) << "Change resolution to " << pic_size.ToString(); |
| client_->ChangeResolution(pic_size, visible_rect, num_output_frames); |
| } |
| |
| bool V4L2StatelessVideoDecoderBackend::ApplyResolution( |
| const gfx::Size& pic_size, |
| const gfx::Rect& visible_rect, |
| const size_t num_output_frames) { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DCHECK_EQ(input_queue_->QueuedBuffersCount(), 0u); |
| |
| auto ret = input_queue_->GetFormat().first; |
| if (!ret) { |
| VPLOGF(1) << "Failed getting OUTPUT format"; |
| return false; |
| } |
| struct v4l2_format format = std::move(*ret); |
| |
| format.fmt.pix_mp.width = pic_size.width(); |
| format.fmt.pix_mp.height = pic_size.height(); |
| if (device_->Ioctl(VIDIOC_S_FMT, &format) != 0) { |
| VPLOGF(1) << "Failed setting OUTPUT format"; |
| return false; |
| } |
| |
| return true; |
| } |
| |
| void V4L2StatelessVideoDecoderBackend::OnChangeResolutionDone(bool success) { |
| if (!success) { |
| client_->OnBackendError(); |
| return; |
| } |
| |
| pic_size_ = avd_->GetPicSize(); |
| client_->CompleteFlush(); |
| task_runner_->PostTask( |
| FROM_HERE, base::BindOnce(&V4L2StatelessVideoDecoderBackend::DoDecodeWork, |
| weak_this_)); |
| } |
| |
| void V4L2StatelessVideoDecoderBackend::OnStreamStopped(bool stop_input_queue) { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DVLOGF(3); |
| |
| // The V4L2 stream has been stopped stopped, so all surfaces on the device |
| // have been returned to the client. |
| surfaces_at_device_ = {}; |
| } |
| |
| void V4L2StatelessVideoDecoderBackend::ClearPendingRequests( |
| DecodeStatus status) { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DVLOGF(3); |
| |
| if (avd_) { |
| // If we reset during resolution change, re-create AVD. Then the new AVD |
| // will trigger resolution change again after reset. |
| if (pic_size_ != avd_->GetPicSize()) { |
| CreateAvd(); |
| } else { |
| avd_->Reset(); |
| } |
| } |
| |
| // Clear output_request_queue_. |
| while (!output_request_queue_.empty()) |
| output_request_queue_.pop(); |
| |
| if (flush_cb_) |
| std::move(flush_cb_).Run(status); |
| |
| // Clear current_decode_request_ and decode_request_queue_. |
| if (current_decode_request_) { |
| std::move(current_decode_request_->decode_cb).Run(status); |
| current_decode_request_ = absl::nullopt; |
| } |
| |
| while (!decode_request_queue_.empty()) { |
| auto request = std::move(decode_request_queue_.front()); |
| decode_request_queue_.pop(); |
| std::move(request.decode_cb).Run(status); |
| } |
| } |
| |
| bool V4L2StatelessVideoDecoderBackend::StopInputQueueOnResChange() const { |
| return true; |
| } |
| |
| bool V4L2StatelessVideoDecoderBackend::IsSupportedProfile( |
| VideoCodecProfile profile) { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DCHECK(device_); |
| if (supported_profiles_.empty()) { |
| constexpr uint32_t kSupportedInputFourccs[] = { |
| V4L2_PIX_FMT_H264_SLICE, |
| V4L2_PIX_FMT_VP8_FRAME, |
| V4L2_PIX_FMT_VP9_FRAME, |
| }; |
| scoped_refptr<V4L2Device> device = V4L2Device::Create(); |
| VideoDecodeAccelerator::SupportedProfiles profiles = |
| device->GetSupportedDecodeProfiles(base::size(kSupportedInputFourccs), |
| kSupportedInputFourccs); |
| for (const auto& profile : profiles) |
| supported_profiles_.push_back(profile.profile); |
| } |
| return std::find(supported_profiles_.begin(), supported_profiles_.end(), |
| profile) != supported_profiles_.end(); |
| } |
| |
| bool V4L2StatelessVideoDecoderBackend::CreateAvd() { |
| DCHECK_CALLED_ON_VALID_SEQUENCE(sequence_checker_); |
| DVLOGF(3); |
| |
| pic_size_ = gfx::Size(); |
| |
| if (profile_ >= H264PROFILE_MIN && profile_ <= H264PROFILE_MAX) { |
| if (input_queue_->SupportsRequests()) { |
| avd_ = std::make_unique<H264Decoder>( |
| std::make_unique<V4L2VideoDecoderDelegateH264>(this, device_.get()), |
| profile_); |
| } else { |
| avd_ = std::make_unique<H264Decoder>( |
| std::make_unique<V4L2VideoDecoderDelegateH264Legacy>(this, |
| device_.get()), |
| profile_); |
| } |
| } else if (profile_ >= VP8PROFILE_MIN && profile_ <= VP8PROFILE_MAX) { |
| if (input_queue_->SupportsRequests()) { |
| avd_ = std::make_unique<VP8Decoder>( |
| std::make_unique<V4L2VideoDecoderDelegateVP8>(this, device_.get())); |
| } else { |
| avd_ = std::make_unique<VP8Decoder>( |
| std::make_unique<V4L2VideoDecoderDelegateVP8Legacy>(this, |
| device_.get())); |
| } |
| } else if (profile_ >= VP9PROFILE_MIN && profile_ <= VP9PROFILE_MAX) { |
| if (input_queue_->SupportsRequests()) { |
| avd_ = std::make_unique<VP9Decoder>( |
| std::make_unique<V4L2VideoDecoderDelegateVP9Chromium>(this, |
| device_.get()), |
| profile_); |
| } else { |
| avd_ = std::make_unique<VP9Decoder>( |
| std::make_unique<V4L2VideoDecoderDelegateVP9Legacy>(this, |
| device_.get()), |
| profile_); |
| } |
| } else { |
| VLOGF(1) << "Unsupported profile " << GetProfileName(profile_); |
| return false; |
| } |
| return true; |
| } |
| |
| } // namespace media |