// Copyright 2014 The Chromium Authors
// 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_encode_accelerator.h"
#include <fcntl.h>
#include <linux/videodev2.h>
#include <poll.h>
#include <string.h>
#include <sys/eventfd.h>
#include <sys/ioctl.h>
#include <sys/mman.h>
#include <memory>
#include <numeric>
#include <optional>
#include <utility>
#include "base/bits.h"
#include "base/command_line.h"
#include "base/functional/bind.h"
#include "base/functional/callback.h"
#include "base/functional/callback_helpers.h"
#include "base/memory/shared_memory_mapping.h"
#include "base/memory/unsafe_shared_memory_region.h"
#include "base/numerics/safe_conversions.h"
#include "base/strings/strcat.h"
#include "base/task/sequenced_task_runner.h"
#include "base/task/single_thread_task_runner.h"
#include "base/task/task_traits.h"
#include "base/task/thread_pool.h"
#include "base/trace_event/trace_event.h"
#include "media/base/bitstream_buffer.h"
#include "media/base/color_plane_layout.h"
#include "media/base/media_log.h"
#include "media/base/media_switches.h"
#include "media/base/media_util.h"
#include "media/base/video_frame_layout.h"
#include "media/base/video_types.h"
#include "media/gpu/chromeos/fourcc.h"
#include "media/gpu/chromeos/image_processor_factory.h"
#include "media/gpu/chromeos/platform_video_frame_utils.h"
#include "media/gpu/gpu_video_encode_accelerator_helpers.h"
#include "media/gpu/macros.h"
#include "media/gpu/v4l2/v4l2_utils.h"
#include "media/parsers/h264_level_limits.h"
#include "media/parsers/h264_parser.h"
#ifndef V4L2_CID_MPEG_VIDEO_H264_HIER_CODING_L0_BR
#define V4L2_CID_MPEG_VIDEO_H264_HIER_CODING_L0_BR (V4L2_CID_CODEC_BASE + 391)
#endif
#ifndef V4L2_CID_MPEG_VIDEO_H264_HIER_CODING_L1_BR
#define V4L2_CID_MPEG_VIDEO_H264_HIER_CODING_L1_BR (V4L2_CID_CODEC_BASE + 392)
#endif
namespace {
const uint8_t kH264StartCode[] = {0, 0, 0, 1};
const size_t kH264StartCodeSize = sizeof(kH264StartCode);
// Copy a H.264 NALU of size |src_size| (without start code), located at |src|,
// into a buffer starting at |dst| of size |dst_size|, prepending it with
// a H.264 start code (as long as both fit). After copying, update |dst| to
// point to the address immediately after the copied data, and update |dst_size|
// to contain remaining destination buffer size.
static void CopyNALUPrependingStartCode(const uint8_t* src,
size_t src_size,
uint8_t** dst,
size_t* dst_size) {
size_t size_to_copy = kH264StartCodeSize + src_size;
if (size_to_copy > *dst_size) {
VLOGF(1) << "Could not copy a NALU, not enough space in destination buffer";
return;
}
memcpy(*dst, kH264StartCode, kH264StartCodeSize);
memcpy(*dst + kH264StartCodeSize, src, src_size);
*dst += size_to_copy;
*dst_size -= size_to_copy;
}
} // namespace
namespace media {
namespace {
// Convert VideoFrameLayout to ImageProcessor::PortConfig.
std::optional<ImageProcessor::PortConfig> VideoFrameLayoutToPortConfig(
const VideoFrameLayout& layout,
const gfx::Rect& visible_rect,
VideoFrame::StorageType storage_type) {
auto fourcc =
Fourcc::FromVideoPixelFormat(layout.format(), !layout.is_multi_planar());
if (!fourcc) {
DVLOGF(1) << "Failed to create Fourcc from video pixel format "
<< VideoPixelFormatToString(layout.format());
return std::nullopt;
}
return ImageProcessor::PortConfig(*fourcc, layout.coded_size(),
layout.planes(), visible_rect,
storage_type);
}
// Create Layout from |layout| with is_multi_planar = true.
std::optional<VideoFrameLayout> AsMultiPlanarLayout(
const VideoFrameLayout& layout) {
if (layout.is_multi_planar())
return std::make_optional<VideoFrameLayout>(layout);
return VideoFrameLayout::CreateMultiPlanar(
layout.format(), layout.coded_size(), layout.planes());
}
scoped_refptr<base::SequencedTaskRunner> CreateEncoderTaskRunner() {
if (base::FeatureList::IsEnabled(kUSeSequencedTaskRunnerForVEA)) {
return base::ThreadPool::CreateSequencedTaskRunner(
{base::WithBaseSyncPrimitives(), base::TaskPriority::USER_VISIBLE,
base::MayBlock()});
} else {
return base::ThreadPool::CreateSingleThreadTaskRunner(
{base::WithBaseSyncPrimitives(), base::MayBlock(),
base::TaskPriority::USER_VISIBLE},
base::SingleThreadTaskRunnerThreadMode::DEDICATED);
}
}
} // namespace
struct V4L2VideoEncodeAccelerator::BitstreamBufferRef {
BitstreamBufferRef(int32_t id, base::WritableSharedMemoryMapping shm_mapping)
: id(id), shm_mapping(std::move(shm_mapping)) {}
const int32_t id;
base::WritableSharedMemoryMapping shm_mapping;
};
V4L2VideoEncodeAccelerator::InputRecord::InputRecord() = default;
V4L2VideoEncodeAccelerator::InputRecord::InputRecord(const InputRecord&) =
default;
V4L2VideoEncodeAccelerator::InputRecord::~InputRecord() = default;
V4L2VideoEncodeAccelerator::InputFrameInfo::InputFrameInfo()
: InputFrameInfo(nullptr, false) {}
V4L2VideoEncodeAccelerator::InputFrameInfo::InputFrameInfo(
scoped_refptr<VideoFrame> frame,
bool force_keyframe)
: frame(frame), force_keyframe(force_keyframe) {}
V4L2VideoEncodeAccelerator::InputFrameInfo::InputFrameInfo(
scoped_refptr<VideoFrame> frame,
bool force_keyframe,
size_t index)
: frame(std::move(frame)),
force_keyframe(force_keyframe),
ip_output_buffer_index(index) {}
V4L2VideoEncodeAccelerator::InputFrameInfo::InputFrameInfo(
const InputFrameInfo&) = default;
V4L2VideoEncodeAccelerator::InputFrameInfo::~InputFrameInfo() = default;
// static
base::AtomicRefCount V4L2VideoEncodeAccelerator::num_instances_(0);
V4L2VideoEncodeAccelerator::V4L2VideoEncodeAccelerator(
scoped_refptr<V4L2Device> device)
: can_use_encoder_(num_instances_.Increment() < kMaxNumOfInstances),
child_task_runner_(base::SequencedTaskRunner::GetCurrentDefault()),
native_input_mode_(false),
output_buffer_byte_size_(0),
output_format_fourcc_(0),
current_framerate_(0),
encoder_state_(kUninitialized),
device_(std::move(device)),
input_memory_type_(V4L2_MEMORY_USERPTR),
is_flush_supported_(false),
// TODO(akahuang): Remove WithBaseSyncPrimitives() after replacing poll
// thread by V4L2DevicePoller.
encoder_task_runner_(CreateEncoderTaskRunner()),
device_poll_thread_("V4L2EncoderDevicePollThread") {
DCHECK_CALLED_ON_VALID_SEQUENCE(child_sequence_checker_);
DETACH_FROM_SEQUENCE(encoder_sequence_checker_);
weak_this_ = weak_this_factory_.GetWeakPtr();
}
V4L2VideoEncodeAccelerator::~V4L2VideoEncodeAccelerator() {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK(!device_poll_thread_.IsRunning());
VLOGF(2);
num_instances_.Decrement();
}
bool V4L2VideoEncodeAccelerator::Initialize(
const Config& config,
Client* client,
std::unique_ptr<MediaLog> media_log) {
DCHECK_CALLED_ON_VALID_SEQUENCE(child_sequence_checker_);
DCHECK_EQ(encoder_state_, kUninitialized);
TRACE_EVENT0("media,gpu", "V4L2VEA::Initialize");
VLOGF(2) << ": " << config.AsHumanReadableString();
if (!can_use_encoder_) {
MEDIA_LOG(ERROR, media_log.get()) << "Too many encoders are allocated";
return false;
}
if (config.HasSpatialLayer()) {
MEDIA_LOG(ERROR, media_log.get())
<< "Spatial layer encoding is not yet supported";
return false;
}
// Currently only Qualcomm (SC7180) supports temporal layers, MTK drivers
// do not. There is no check here to determine if the driver supports temporal
// layering. It is expected that clients will first call
// GetSupportedScalabilityModesForV4L2Codec() to get the capabilities.
if (config.HasTemporalLayer()) {
if (VideoCodecProfileToVideoCodec(config.output_profile) ==
VideoCodec::kH264) {
constexpr uint8_t kNumSupportedH264TemporalLayers = 2;
const uint8_t num_temporal_layers =
config.spatial_layers[0].num_of_temporal_layers;
if (num_temporal_layers == kNumSupportedH264TemporalLayers) {
h264_l1t2_enabled_ = true;
} else {
MEDIA_LOG(ERROR, media_log.get())
<< "Unsupported number of temporal layers: "
<< base::strict_cast<size_t>(num_temporal_layers);
return false;
}
} else {
MEDIA_LOG(WARNING, media_log.get())
<< GetProfileName(config.output_profile)
<< " does not support temporal scalability. L1T1 will be produced.";
}
}
encoder_input_visible_rect_ = gfx::Rect(config.input_visible_size);
client_ptr_factory_ = std::make_unique<base::WeakPtrFactory<Client>>(client);
client_ = client_ptr_factory_->GetWeakPtr();
output_format_fourcc_ =
VideoCodecProfileToV4L2PixFmt(config.output_profile, false);
if (output_format_fourcc_ == V4L2_PIX_FMT_INVALID) {
MEDIA_LOG(ERROR, media_log.get())
<< "invalid output_profile=" << GetProfileName(config.output_profile);
return false;
}
if (!device_->Open(V4L2Device::Type::kEncoder, output_format_fourcc_)) {
MEDIA_LOG(ERROR, media_log.get())
<< "Failed to open device for profile="
<< GetProfileName(config.output_profile)
<< ", fourcc=" << FourccToString(output_format_fourcc_);
return false;
}
gfx::Size min_resolution;
gfx::Size max_resolution;
GetSupportedResolution(base::BindRepeating(&V4L2Device::Ioctl, device_),
output_format_fourcc_, &min_resolution,
&max_resolution);
if (config.input_visible_size.width() < min_resolution.width() ||
config.input_visible_size.height() < min_resolution.height() ||
config.input_visible_size.width() > max_resolution.width() ||
config.input_visible_size.height() > max_resolution.height()) {
MEDIA_LOG(ERROR, media_log.get())
<< "Unsupported resolution: " << config.input_visible_size.ToString()
<< ", min=" << min_resolution.ToString()
<< ", max=" << max_resolution.ToString();
return false;
}
// Ask if V4L2_ENC_CMD_STOP (Flush) is supported.
struct v4l2_encoder_cmd cmd;
memset(&cmd, 0, sizeof(cmd));
cmd.cmd = V4L2_ENC_CMD_STOP;
is_flush_supported_ = (device_->Ioctl(VIDIOC_TRY_ENCODER_CMD, &cmd) == 0);
if (!is_flush_supported_)
VLOGF(2) << "V4L2_ENC_CMD_STOP is not supported.";
struct v4l2_capability caps;
memset(&caps, 0, sizeof(caps));
const __u32 kCapsRequired = V4L2_CAP_VIDEO_M2M_MPLANE | V4L2_CAP_STREAMING;
if (device_->Ioctl(VIDIOC_QUERYCAP, &caps) != 0) {
MEDIA_LOG(ERROR, media_log.get())
<< "ioctl() failed: VIDIOC_QUERYCAP, errno=" << errno;
return false;
}
if ((caps.capabilities & kCapsRequired) != kCapsRequired) {
MEDIA_LOG(ERROR, media_log.get())
<< "caps check failed: 0x" << std::hex << caps.capabilities;
return false;
}
driver_name_ = device_->GetDriverName();
encoder_task_runner_->PostTask(
FROM_HERE, base::BindOnce(&V4L2VideoEncodeAccelerator::InitializeTask,
weak_this_, config));
return true;
}
void V4L2VideoEncodeAccelerator::InitializeTask(const Config& config) {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
TRACE_EVENT0("media,gpu", "V4L2VEA::InitializeTask");
// Set kInitialized here so that NotifyErrorStatus() is invoked from here.
encoder_state_ = kInitialized;
native_input_mode_ =
config.storage_type == Config::StorageType::kGpuMemoryBuffer;
input_queue_ = device_->GetQueue(V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE);
output_queue_ = device_->GetQueue(V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE);
if (!input_queue_ || !output_queue_) {
SetErrorState({EncoderStatus::Codes::kEncoderInitializationError,
"Failed to get V4L2Queue"});
return;
}
if (!SetFormats(config.input_format, config.output_profile)) {
return;
}
if (config.input_format != device_input_layout_->format()) {
VLOGF(2) << "Input format: " << config.input_format << " is not supported "
<< "by the HW. Will try to convert to "
<< device_input_layout_->format();
auto input_layout = VideoFrameLayout::CreateMultiPlanar(
config.input_format, encoder_input_visible_rect_.size(),
std::vector<ColorPlaneLayout>(
VideoFrame::NumPlanes(config.input_format)));
if (!input_layout) {
SetErrorState({EncoderStatus::Codes::kUnsupportedFrameFormat,
"Invalid image processor input layout"});
return;
}
// ImageProcessor for a pixel format conversion.
if (!CreateImageProcessor(*input_layout, device_input_layout_->format(),
device_input_layout_->coded_size(),
encoder_input_visible_rect_,
encoder_input_visible_rect_)) {
SetErrorState(
{EncoderStatus::Codes::kEncoderInitializationError,
base::StrCat(
{"Failed to create image processor ", "for format conversion",
VideoPixelFormatToString(input_layout->format()), " -> ",
VideoPixelFormatToString(device_input_layout_->format())})});
return;
}
const gfx::Size ip_output_buffer_size(
image_processor_->output_config().planes[0].stride,
image_processor_->output_config().size.height());
if (!NegotiateInputFormat(device_input_layout_->format(),
ip_output_buffer_size)) {
SetErrorState(
{EncoderStatus::Codes::kUnsupportedFrameFormat,
base::StrCat({"Failed to reconfigure v4l2 encoder driver with the ",
"ImageProcessor output buffer: ",
ip_output_buffer_size.ToString()})});
return;
}
}
if (!InitInputMemoryType(config) || !InitControls(config) ||
!CreateOutputBuffers()) {
SetErrorState(EncoderStatus::Codes::kEncoderInitializationError);
return;
}
if (config.bitrate.mode() != Bitrate::Mode::kConstant &&
config.bitrate.mode() != Bitrate::Mode::kVariable) {
SetErrorState({EncoderStatus::Codes::kEncoderUnsupportedConfig,
base::StrCat({"Invalid bitrate mode: ",
base::NumberToString(base::strict_cast<int>(
config.bitrate.mode()))})});
return;
}
if (config.bitrate.mode() == Bitrate::Mode::kVariable &&
!base::FeatureList::IsEnabled(kChromeOSHWVBREncoding)) {
SetErrorState({EncoderStatus::Codes::kEncoderUnsupportedConfig,
"VBR encoding is disabled"});
return;
}
const uint32_t bitrate_mode =
config.bitrate.mode() == Bitrate::Mode::kConstant
? V4L2_MPEG_VIDEO_BITRATE_MODE_CBR
: V4L2_MPEG_VIDEO_BITRATE_MODE_VBR;
const VideoBitrateAllocation bitrate_allocation =
AllocateBitrateForDefaultEncoding(config);
if (!device_->SetExtCtrls(
V4L2_CID_MPEG_CLASS,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_BITRATE_MODE, bitrate_mode)})) {
SetErrorState({EncoderStatus::Codes::kEncoderHardwareDriverError,
base::StrCat({"Failed to configure bitrate mode: ",
base::NumberToString(base::strict_cast<int>(
bitrate_allocation.GetMode()))})});
return;
}
current_bitrate_allocation_ = VideoBitrateAllocation(config.bitrate.mode());
RequestEncodingParametersChangeTask(bitrate_allocation, config.framerate,
std::nullopt);
// input_frame_size_ is the size of input_config of |image_processor_|.
// On native_input_mode_, since the passed size in RequireBitstreamBuffers()
// is ignored by the client, we don't update the expected frame size.
if (!native_input_mode_ && image_processor_.get())
input_frame_size_ = image_processor_->input_config().size;
child_task_runner_->PostTask(
FROM_HERE, base::BindOnce(&Client::RequireBitstreamBuffers, client_,
kInputBufferCount, input_frame_size_,
output_buffer_byte_size_));
// Notify VideoEncoderInfo after initialization.
VideoEncoderInfo encoder_info;
encoder_info.implementation_name = "V4L2VideoEncodeAccelerator";
DCHECK(!encoder_info.has_trusted_rate_controller);
DCHECK(encoder_info.is_hardware_accelerated);
DCHECK(encoder_info.supports_native_handle);
DCHECK(!encoder_info.supports_simulcast);
// V4L2VideoEncodeAccelerator only supports temporal-SVC.
if (config.HasTemporalLayer()) {
CHECK(!config.spatial_layers.empty());
for (size_t i = 0; i < config.spatial_layers.size(); ++i) {
encoder_info.fps_allocation[i] =
GetFpsAllocation(config.spatial_layers[i].num_of_temporal_layers);
}
} else {
constexpr uint8_t kFullFramerate = 255;
encoder_info.fps_allocation[0] = {kFullFramerate};
}
child_task_runner_->PostTask(
FROM_HERE,
base::BindOnce(&Client::NotifyEncoderInfoChange, client_, encoder_info));
}
bool V4L2VideoEncodeAccelerator::CreateImageProcessor(
const VideoFrameLayout& input_layout,
const VideoPixelFormat output_format,
const gfx::Size& output_size,
const gfx::Rect& input_visible_rect,
const gfx::Rect& output_visible_rect) {
VLOGF(2);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
auto ip_input_layout = AsMultiPlanarLayout(input_layout);
if (!ip_input_layout) {
LOG(ERROR) << "Failed to get multi-planar input layout, input_layout="
<< input_layout;
return false;
}
VideoFrame::StorageType input_storage_type =
native_input_mode_ ? VideoFrame::STORAGE_GPU_MEMORY_BUFFER
: VideoFrame::STORAGE_SHMEM;
auto input_config = VideoFrameLayoutToPortConfig(
*ip_input_layout, input_visible_rect, input_storage_type);
if (!input_config) {
LOG(ERROR) << "Failed to create ImageProcessor input config";
return false;
}
auto platform_layout = GetPlatformVideoFrameLayout(
output_format, output_size,
gfx::BufferUsage::VEA_READ_CAMERA_AND_CPU_READ_WRITE);
if (!platform_layout) {
LOG(ERROR) << "Failed to get Platform VideoFrameLayout";
return false;
}
auto output_layout = AsMultiPlanarLayout(platform_layout.value());
if (!output_layout) {
LOG(ERROR) << "Failed to get multi-planar platform layout, platform_layout="
<< *platform_layout;
return false;
}
auto output_config =
VideoFrameLayoutToPortConfig(*output_layout, output_visible_rect,
VideoFrame::STORAGE_GPU_MEMORY_BUFFER);
if (!output_config) {
LOG(ERROR) << "Failed to create ImageProcessor output config";
return false;
}
image_processor_ = ImageProcessorFactory::Create(
*input_config, *output_config, kImageProcBufferCount,
base::BindRepeating(&V4L2VideoEncodeAccelerator::ImageProcessorError,
weak_this_),
encoder_task_runner_);
if (!image_processor_) {
LOG(ERROR) << "Failed initializing image processor";
return false;
}
VLOGF(2) << "ImageProcessor is created: " << image_processor_->backend_type();
num_frames_in_image_processor_ = 0;
// The output of image processor is the input of encoder. Output coded
// width of processor must be the same as input coded width of encoder.
// Output coded height of processor can be larger but not smaller than the
// input coded height of encoder. For example, suppose input size of encoder
// is 320x193. It is OK if the output of processor is 320x208.
const auto& ip_output_size = image_processor_->output_config().size;
if (ip_output_size.width() != output_layout->coded_size().width() ||
ip_output_size.height() < output_layout->coded_size().height()) {
LOG(ERROR) << "Invalid image processor output coded size "
<< ip_output_size.ToString()
<< ", expected output coded size is "
<< output_layout->coded_size().ToString();
return false;
}
// Initialize |free_image_processor_output_buffer_indices_|.
free_image_processor_output_buffer_indices_.resize(kImageProcBufferCount);
std::iota(free_image_processor_output_buffer_indices_.begin(),
free_image_processor_output_buffer_indices_.end(), 0);
return AllocateImageProcessorOutputBuffers(kImageProcBufferCount);
}
bool V4L2VideoEncodeAccelerator::AllocateImageProcessorOutputBuffers(
size_t count) {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK(image_processor_);
DCHECK_EQ(image_processor_->output_mode(),
ImageProcessor::OutputMode::IMPORT);
// The existing buffers in |image_processor_output_buffers_| may be alive
// until they are actually consumed by the encoder driver, after they are
// destroyed here.
image_processor_output_buffers_.clear();
image_processor_output_buffers_.resize(count);
const ImageProcessor::PortConfig& output_config =
image_processor_->output_config();
for (size_t i = 0; i < count; i++) {
switch (output_config.storage_type) {
case VideoFrame::STORAGE_GPU_MEMORY_BUFFER:
image_processor_output_buffers_[i] = CreateGpuMemoryBufferVideoFrame(
output_config.fourcc.ToVideoPixelFormat(), output_config.size,
output_config.visible_rect, output_config.visible_rect.size(),
base::TimeDelta(),
gfx::BufferUsage::VEA_READ_CAMERA_AND_CPU_READ_WRITE);
break;
default:
LOG(ERROR) << "Unsupported output storage type of image processor: "
<< output_config.storage_type;
return false;
}
if (!image_processor_output_buffers_[i]) {
LOG(ERROR) << "Failed to create VideoFrame";
return false;
}
}
return true;
}
bool V4L2VideoEncodeAccelerator::InitInputMemoryType(const Config& config) {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
if (image_processor_) {
const auto storage_type = image_processor_->output_config().storage_type;
if (storage_type == VideoFrame::STORAGE_GPU_MEMORY_BUFFER) {
input_memory_type_ = V4L2_MEMORY_DMABUF;
} else if (VideoFrame::IsStorageTypeMappable(storage_type)) {
input_memory_type_ = V4L2_MEMORY_USERPTR;
} else {
LOG(ERROR) << "Unsupported image processor's output StorageType: "
<< storage_type;
return false;
}
} else {
switch (config.storage_type) {
case Config::StorageType::kShmem:
input_memory_type_ = V4L2_MEMORY_USERPTR;
break;
case Config::StorageType::kGpuMemoryBuffer:
input_memory_type_ = V4L2_MEMORY_DMABUF;
break;
}
}
return true;
}
void V4L2VideoEncodeAccelerator::ImageProcessorError() {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
LOG(ERROR) << "Image processor error";
// TODO(b/276005687): Let image processor return a minute error status and
// convert it to EncoderStatus.
SetErrorState(EncoderStatus::Codes::kFormatConversionError);
}
void V4L2VideoEncodeAccelerator::Encode(scoped_refptr<VideoFrame> frame,
bool force_keyframe) {
DVLOGF(4) << "force_keyframe=" << force_keyframe;
DCHECK_CALLED_ON_VALID_SEQUENCE(child_sequence_checker_);
encoder_task_runner_->PostTask(
FROM_HERE, base::BindOnce(&V4L2VideoEncodeAccelerator::EncodeTask,
weak_this_, std::move(frame), force_keyframe));
}
void V4L2VideoEncodeAccelerator::UseOutputBitstreamBuffer(
BitstreamBuffer buffer) {
DVLOGF(4) << "id=" << buffer.id();
DCHECK_CALLED_ON_VALID_SEQUENCE(child_sequence_checker_);
encoder_task_runner_->PostTask(
FROM_HERE,
base::BindOnce(&V4L2VideoEncodeAccelerator::UseOutputBitstreamBufferTask,
weak_this_, std::move(buffer)));
}
void V4L2VideoEncodeAccelerator::RequestEncodingParametersChange(
const Bitrate& bitrate,
uint32_t framerate,
const std::optional<gfx::Size>& size) {
DCHECK_CALLED_ON_VALID_SEQUENCE(child_sequence_checker_);
VideoBitrateAllocation allocation(bitrate.mode());
allocation.SetBitrate(0u, 0u, bitrate.target_bps());
allocation.SetPeakBps(bitrate.peak_bps());
RequestEncodingParametersChange(allocation, framerate, size);
}
void V4L2VideoEncodeAccelerator::RequestEncodingParametersChange(
const VideoBitrateAllocation& bitrate_allocation,
uint32_t framerate,
const std::optional<gfx::Size>& size) {
DCHECK_CALLED_ON_VALID_SEQUENCE(child_sequence_checker_);
encoder_task_runner_->PostTask(
FROM_HERE,
base::BindOnce(
&V4L2VideoEncodeAccelerator::RequestEncodingParametersChangeTask,
weak_this_, bitrate_allocation, framerate, size));
}
void V4L2VideoEncodeAccelerator::Destroy() {
VLOGF(2);
DCHECK_CALLED_ON_VALID_SEQUENCE(child_sequence_checker_);
// We're destroying; cancel all callbacks.
client_ptr_factory_.reset();
encoder_task_runner_->PostTask(
FROM_HERE,
base::BindOnce(&V4L2VideoEncodeAccelerator::DestroyTask, weak_this_));
}
void V4L2VideoEncodeAccelerator::Flush(FlushCallback flush_callback) {
VLOGF(2);
DCHECK_CALLED_ON_VALID_SEQUENCE(child_sequence_checker_);
encoder_task_runner_->PostTask(
FROM_HERE, base::BindOnce(&V4L2VideoEncodeAccelerator::FlushTask,
weak_this_, std::move(flush_callback)));
}
void V4L2VideoEncodeAccelerator::FlushTask(FlushCallback flush_callback) {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
if (encoder_state_ == kInitialized) {
// Flush() is called before either Encode() or UseOutputBitstreamBuffer() is
// called. Just return as successful.
child_task_runner_->PostTask(
FROM_HERE, base::BindOnce(std::move(flush_callback), true));
return;
}
if (flush_callback_ || encoder_state_ != kEncoding) {
SetErrorState({EncoderStatus::Codes::kEncoderIllegalState,
"Flush failed: there is a pending flush, or "
"VideoEncodeAccelerator is not in kEncoding state"});
child_task_runner_->PostTask(
FROM_HERE, base::BindOnce(std::move(flush_callback), false));
return;
}
flush_callback_ = std::move(flush_callback);
// Push a null frame to indicate Flush.
EncodeTask(nullptr, false);
}
bool V4L2VideoEncodeAccelerator::IsFlushSupported() {
return is_flush_supported_;
}
VideoEncodeAccelerator::SupportedProfiles
V4L2VideoEncodeAccelerator::GetSupportedProfiles() {
auto device = base::MakeRefCounted<V4L2Device>();
return device->GetSupportedEncodeProfiles();
}
void V4L2VideoEncodeAccelerator::FrameProcessed(
bool force_keyframe,
base::TimeDelta timestamp,
size_t output_buffer_index,
scoped_refptr<VideoFrame> frame) {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DVLOGF(4) << "force_keyframe=" << force_keyframe
<< ", output_buffer_index=" << output_buffer_index;
DCHECK_GE(output_buffer_index, 0u);
TRACE_EVENT_NESTABLE_ASYNC_END2(
"media,gpu", "V4L2VEA::ImageProcessor::Process",
timestamp.InMicroseconds(), "timestamp", timestamp.InMicroseconds(),
"output_size", image_processor_->output_config().size.ToString());
encoder_input_queue_.emplace(std::move(frame), force_keyframe,
output_buffer_index);
CHECK_GT(num_frames_in_image_processor_, 0u);
num_frames_in_image_processor_--;
MaybeFlushImageProcessor();
encoder_task_runner_->PostTask(
FROM_HERE,
base::BindOnce(&V4L2VideoEncodeAccelerator::Enqueue, weak_this_));
}
void V4L2VideoEncodeAccelerator::ReuseImageProcessorOutputBuffer(
size_t output_buffer_index) {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DVLOGF(4) << "output_buffer_index=" << output_buffer_index;
free_image_processor_output_buffer_indices_.push_back(output_buffer_index);
InputImageProcessorTask();
}
BitstreamBufferMetadata V4L2VideoEncodeAccelerator::GetMetadata(
const uint8_t* data,
size_t data_size_bytes,
bool key_frame,
base::TimeDelta timestamp) {
auto buffer_metadata =
BitstreamBufferMetadata(data_size_bytes, key_frame, timestamp);
if (h264_l1t2_enabled_) {
H264Metadata h264_metadata;
h264_metadata.temporal_idx = 0;
h264_metadata.layer_sync = false;
H264Parser parser;
parser.SetStream(data, data_size_bytes);
H264NALU nalu;
while (parser.AdvanceToNextNALU(&nalu) == H264Parser::kOk) {
// L1T2 describes a bitstream where every other frame is not used as a
// reference frame. This allows those non reference frames to be discarded
// and never decoded. The V4L2 api does not provide a way to request that
// frames are not used as reference frames. It does provide a set of
// controls (V4L2_CID_MPEG_VIDEO_H264_HIERARCHICAL_CODING_*) that allows
// specifying that.
// In order to determine if a frame can be dropped the frames can be
// queried to see if they are marked as reference frames
// (i.e. nal_ref_idc == 0).
if (nalu.nal_unit_type == H264NALU::kNonIDRSlice &&
nalu.nal_ref_idc == 0) {
h264_metadata.temporal_idx = 1;
h264_metadata.layer_sync = true;
break;
}
}
buffer_metadata.h264 = h264_metadata;
}
return buffer_metadata;
}
size_t V4L2VideoEncodeAccelerator::CopyIntoOutputBuffer(
const uint8_t* bitstream_data,
size_t bitstream_size,
std::unique_ptr<BitstreamBufferRef> buffer_ref) {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
uint8_t* dst_ptr = buffer_ref->shm_mapping.GetMemoryAs<uint8_t>();
size_t remaining_dst_size = buffer_ref->shm_mapping.size();
if (!inject_sps_and_pps_) {
if (bitstream_size <= remaining_dst_size) {
memcpy(dst_ptr, bitstream_data, bitstream_size);
return bitstream_size;
} else {
SetErrorState({EncoderStatus::Codes::kEncoderFailedEncode,
"Output data did not fit in the BitstreamBuffer"});
return 0;
}
}
// Cache the newest SPS and PPS found in the stream, and inject them before
// each IDR found.
H264Parser parser;
parser.SetStream(bitstream_data, bitstream_size);
H264NALU nalu;
bool inserted_sps = false;
bool inserted_pps = false;
while (parser.AdvanceToNextNALU(&nalu) == H264Parser::kOk) {
// nalu.size is always without the start code, regardless of the NALU type.
if (nalu.size + kH264StartCodeSize > remaining_dst_size) {
VLOGF(1) << "Output data did not fit in the BitstreamBuffer";
break;
}
switch (nalu.nal_unit_type) {
case H264NALU::kSPS:
cached_sps_.resize(nalu.size);
memcpy(cached_sps_.data(), nalu.data, nalu.size);
cached_h264_header_size_ =
cached_sps_.size() + cached_pps_.size() + 2 * kH264StartCodeSize;
inserted_sps = true;
break;
case H264NALU::kPPS:
cached_pps_.resize(nalu.size);
memcpy(cached_pps_.data(), nalu.data, nalu.size);
cached_h264_header_size_ =
cached_sps_.size() + cached_pps_.size() + 2 * kH264StartCodeSize;
inserted_pps = true;
break;
case H264NALU::kIDRSlice:
if (inserted_sps && inserted_pps) {
// Already inserted SPS and PPS. No need to inject.
break;
}
// Only inject if we have both headers cached, and enough space for both
// the headers and the NALU itself.
if (cached_sps_.empty() || cached_pps_.empty()) {
VLOGF(1) << "Cannot inject IDR slice without SPS and PPS";
break;
}
if (cached_h264_header_size_ + nalu.size + kH264StartCodeSize >
remaining_dst_size) {
VLOGF(1) << "Not enough space to inject a stream header before IDR";
break;
}
if (!inserted_sps) {
CopyNALUPrependingStartCode(cached_sps_.data(), cached_sps_.size(),
&dst_ptr, &remaining_dst_size);
}
if (!inserted_pps) {
CopyNALUPrependingStartCode(cached_pps_.data(), cached_pps_.size(),
&dst_ptr, &remaining_dst_size);
}
VLOGF(2) << "Stream header injected before IDR";
break;
}
CopyNALUPrependingStartCode(nalu.data, nalu.size, &dst_ptr,
&remaining_dst_size);
}
return buffer_ref->shm_mapping.size() - remaining_dst_size;
}
void V4L2VideoEncodeAccelerator::EncodeTask(scoped_refptr<VideoFrame> frame,
bool force_keyframe) {
DVLOGF(4) << "force_keyframe=" << force_keyframe;
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK_NE(encoder_state_, kUninitialized);
if (encoder_state_ == kError) {
DVLOGF(1) << "early out: kError state";
return;
}
if (frame) {
TRACE_EVENT1("media,gpu", "V4L2VEA::EncodeTask", "timestamp",
frame->timestamp().InMicroseconds());
// |frame| can be nullptr to indicate a flush.
const bool is_expected_storage_type =
native_input_mode_
? frame->storage_type() == VideoFrame::STORAGE_GPU_MEMORY_BUFFER
: frame->IsMappable();
if (!is_expected_storage_type) {
SetErrorState({EncoderStatus::Codes::kInvalidInputFrame,
base::StrCat({"Unexpected storage: ",
VideoFrame::StorageTypeToString(
frame->storage_type())})});
return;
}
if (!ReconfigureFormatIfNeeded(*frame)) {
SetErrorState({EncoderStatus::Codes::kUnsupportedFrameFormat,
base::StrCat({"Unsupported frame: ",
frame->AsHumanReadableString()})});
return;
}
// If a video frame to be encoded is fed, then call VIDIOC_REQBUFS if it has
// not been called yet.
if (input_buffer_map_.empty() && !CreateInputBuffers()) {
CHECK_EQ(encoder_state_, kError);
return;
}
if (encoder_state_ == kInitialized) {
if (!StartDevicePoll())
return;
encoder_state_ = kEncoding;
}
}
if (image_processor_) {
image_processor_input_queue_.emplace(std::move(frame), force_keyframe);
InputImageProcessorTask();
} else {
encoder_input_queue_.emplace(std::move(frame), force_keyframe);
Enqueue();
}
}
bool V4L2VideoEncodeAccelerator::ReconfigureFormatIfNeeded(
const VideoFrame& frame) {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
if (input_buffer_map_.empty()) {
// Updates |input_natural_size_| on the first VideoFrame.
// |input_natural_size_| is a dimension to be encoded (i.e.
// |encoder_input_visible_rect_.size()|), but can be different from it
// in simulcast case.
input_natural_size_ = frame.natural_size();
}
if (!native_input_mode_) {
// frame.coded_size() must be the size specified in
// RequireBitstreamBuffers() in non native-input mode.
return frame.coded_size() == input_frame_size_;
}
if (!input_buffer_map_.empty()) {
// ReconfigureFormatIfNeeded() has been called with the first VideoFrame.
// We checks here we need to (re)create ImageProcessor because the visible
// rectangle of |frame| differs from the first VideoFrame.
// |frame.natural_size()| must be unchanged during encoding in the same
// VideoEncodeAccelerator instance. When it is changed, a client has to
// recreate VideoEncodeAccelerator.
if (frame.natural_size() != input_natural_size_) {
LOG(ERROR) << "Encoder resolution is changed during encoding"
<< ", frame.natural_size()=" << frame.natural_size().ToString()
<< ", input_natural_size_=" << input_natural_size_.ToString();
return false;
}
if (frame.coded_size() == input_frame_size_) {
return true;
}
// If a dimension of the underlying VideoFrame varies during video encoding
// (i.e. frame.coded_size() != input_frame_size_), we (re)create
// ImageProcessor to crop the VideoFrame, |frame.visible_rect()| ->
// |encoder_input_visible_rect_|.
// TODO(hiroh): if |frame.coded_size()| is the same as VideoFrame::
// DetermineAlignedSize(input_format, encoder_input_visible_rect_.size())
// and don't need a pixel format conversion, image processor is not
// necessary but we should rather NegotiateInputFormat().
} else if (frame.coded_size() == input_frame_size_) {
// This path is for the first frame on Encode().
// Height and width that V4L2VEA needs to configure.
const gfx::Size buffer_size(frame.stride(0), frame.coded_size().height());
// A buffer given by client is allocated with the same dimension using
// minigbm. However, it is possible that stride and height are different
// from ones adjusted by a driver.
if (!image_processor_) {
if (device_input_layout_->coded_size().width() == buffer_size.width() &&
device_input_layout_->coded_size().height() == buffer_size.height()) {
return true;
}
return NegotiateInputFormat(device_input_layout_->format(), buffer_size)
.has_value();
}
if (image_processor_->input_config().size.height() ==
buffer_size.height() &&
image_processor_->input_config().planes[0].stride ==
buffer_size.width()) {
return true;
}
}
// The |frame| dimension is different from the resolution configured to
// V4L2VEA. This is the case that V4L2VEA needs to create ImageProcessor for
// cropping and scaling. Update |input_frame_size_| to check if succeeding
// frames' dimensions are not different from the current one.
input_frame_size_ = frame.coded_size();
if (!CreateImageProcessor(frame.layout(), device_input_layout_->format(),
device_input_layout_->coded_size(),
frame.visible_rect(),
encoder_input_visible_rect_)) {
LOG(ERROR) << "Failed to create image processor";
return false;
}
if (gfx::Size(image_processor_->output_config().planes[0].stride,
image_processor_->output_config().size.height()) !=
device_input_layout_->coded_size()) {
LOG(ERROR) << "Image Processor's output buffer's size is different from "
<< "input buffer size configure to the encoder driver. "
<< "ip's output buffer size: "
<< gfx::Size(image_processor_->output_config().planes[0].stride,
image_processor_->output_config().size.height())
.ToString()
<< ", encoder's input buffer size: "
<< device_input_layout_->coded_size().ToString();
return false;
}
return true;
}
void V4L2VideoEncodeAccelerator::MaybeFlushImageProcessor() {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK(image_processor_);
if (image_processor_input_queue_.size() == 1 &&
!image_processor_input_queue_.front().frame &&
num_frames_in_image_processor_ == 0) {
// Flush the encoder once the image processor is done with its own flush.
DVLOGF(3) << "All frames to be flush have been processed by "
<< "|image_processor_|. Move the flush request to the encoder";
image_processor_input_queue_.pop();
encoder_input_queue_.emplace(nullptr, false);
Enqueue();
}
}
void V4L2VideoEncodeAccelerator::InputImageProcessorTask() {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
MaybeFlushImageProcessor();
if (free_image_processor_output_buffer_indices_.empty())
return;
if (image_processor_input_queue_.empty())
return;
// The flush request is at the top. Waiting until all frames are processed by
// the image processor.
if (!image_processor_input_queue_.front().frame)
return;
const size_t output_buffer_index =
free_image_processor_output_buffer_indices_.back();
free_image_processor_output_buffer_indices_.pop_back();
InputFrameInfo frame_info = std::move(image_processor_input_queue_.front());
image_processor_input_queue_.pop();
auto frame = std::move(frame_info.frame);
const bool force_keyframe = frame_info.force_keyframe;
auto timestamp = frame->timestamp();
TRACE_EVENT_NESTABLE_ASYNC_BEGIN1(
"media,gpu", "V4L2VEA::ImageProcessor::Process",
timestamp.InMicroseconds(), "timestamp", timestamp.InMicroseconds());
auto output_frame = image_processor_output_buffers_[output_buffer_index];
if (!image_processor_->Process(
std::move(frame), std::move(output_frame),
base::BindOnce(&V4L2VideoEncodeAccelerator::FrameProcessed,
weak_this_, force_keyframe, timestamp,
output_buffer_index))) {
SetErrorState({EncoderStatus::Codes::kFormatConversionError,
"Failed in ImageProcessor::Process"});
return;
}
num_frames_in_image_processor_++;
}
void V4L2VideoEncodeAccelerator::UseOutputBitstreamBufferTask(
BitstreamBuffer buffer) {
DVLOGF(4) << "id=" << buffer.id();
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
if (buffer.size() < output_buffer_byte_size_) {
SetErrorState({EncoderStatus::Codes::kInvalidOutputBuffer,
"Provided bitstream buffer too small"});
return;
}
base::UnsafeSharedMemoryRegion shm_region = buffer.TakeRegion();
base::WritableSharedMemoryMapping shm_mapping =
shm_region.MapAt(buffer.offset(), buffer.size());
if (!shm_mapping.IsValid()) {
SetErrorState({EncoderStatus::Codes::kSystemAPICallError,
"Failed to map a shared memory buffer"});
return;
}
bitstream_buffer_pool_.push_back(std::make_unique<BitstreamBufferRef>(
buffer.id(), std::move(shm_mapping)));
PumpBitstreamBuffers();
if (encoder_state_ == kInitialized) {
if (!StartDevicePoll())
return;
encoder_state_ = kEncoding;
}
}
void V4L2VideoEncodeAccelerator::DestroyTask() {
VLOGF(2);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
weak_this_factory_.InvalidateWeakPtrs();
// If a flush is pending, notify client that it did not finish.
if (flush_callback_) {
child_task_runner_->PostTask(
FROM_HERE, base::BindOnce(std::move(flush_callback_), false));
}
// Stop streaming and the device_poll_thread_.
LOG_IF(ERROR, !StopDevicePoll()) << "Failure in termination";
DestroyInputBuffers();
DestroyOutputBuffers();
delete this;
}
void V4L2VideoEncodeAccelerator::ServiceDeviceTask() {
DVLOGF(3);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK_NE(encoder_state_, kUninitialized);
DCHECK_NE(encoder_state_, kInitialized);
if (encoder_state_ == kError) {
DVLOGF(1) << "early out: kError state";
return;
}
Dequeue();
Enqueue();
// Clear the interrupt fd.
if (!device_->ClearDevicePollInterrupt())
return;
// Device can be polled as soon as either input or output buffers are queued.
bool poll_device = (input_queue_->QueuedBuffersCount() +
output_queue_->QueuedBuffersCount() >
0);
// ServiceDeviceTask() should only ever be scheduled from DevicePollTask(),
// so either:
// * device_poll_thread_ is running normally
// * device_poll_thread_ scheduled us, but then a DestroyTask() shut it down,
// in which case we're in kError state, and we should have early-outed
// already.
DCHECK(device_poll_thread_.task_runner());
// Queue the DevicePollTask() now.
// base::Unretained(this) is safe, because device_poll_thread_ is owned by
// *this and stops before *this destruction.
device_poll_thread_.task_runner()->PostTask(
FROM_HERE, base::BindOnce(&V4L2VideoEncodeAccelerator::DevicePollTask,
base::Unretained(this), poll_device));
DVLOGF(3) << encoder_input_queue_.size() << "] => DEVICE["
<< input_queue_->FreeBuffersCount() << "+"
<< input_queue_->QueuedBuffersCount() << "/"
<< input_buffer_map_.size() << "->"
<< output_queue_->FreeBuffersCount() << "+"
<< output_queue_->QueuedBuffersCount() << "/"
<< output_queue_->AllocatedBuffersCount() << "] => OUT["
<< bitstream_buffer_pool_.size() << "]";
}
void V4L2VideoEncodeAccelerator::Enqueue() {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK(input_queue_ && output_queue_);
TRACE_EVENT0("media,gpu", "V4L2VEA::Enqueue");
DVLOGF(4) << "free_input_buffers: " << input_queue_->FreeBuffersCount()
<< ", input_queue: " << encoder_input_queue_.size();
bool do_streamon = false;
// Enqueue all the inputs we can.
const size_t old_inputs_queued = input_queue_->QueuedBuffersCount();
while (!encoder_input_queue_.empty() &&
input_queue_->FreeBuffersCount() > 0) {
// A null frame indicates a flush.
if (encoder_input_queue_.front().frame == nullptr) {
DVLOGF(3) << "All input frames needed to be flushed are enqueued.";
encoder_input_queue_.pop();
// If we are not streaming, the device is not running and there is no need
// to call V4L2_ENC_CMD_STOP to request a flush. This also means there is
// nothing left to process, so we can return flush success back to the
// client.
if (!input_queue_->IsStreaming()) {
child_task_runner_->PostTask(
FROM_HERE, base::BindOnce(std::move(flush_callback_), true));
return;
}
struct v4l2_encoder_cmd cmd;
memset(&cmd, 0, sizeof(cmd));
cmd.cmd = V4L2_ENC_CMD_STOP;
if (device_->Ioctl(VIDIOC_ENCODER_CMD, &cmd) != 0) {
SetErrorState(
{EncoderStatus::Codes::kEncoderFailedFlush,
base::StrCat({"ioctl() failed: VIDIOC_ENCODER_CMD, errno=",
base::NumberToString(errno)})});
child_task_runner_->PostTask(
FROM_HERE, base::BindOnce(std::move(flush_callback_), false));
return;
}
encoder_state_ = kFlushing;
break;
}
std::optional<V4L2WritableBufferRef> input_buffer;
switch (input_memory_type_) {
case V4L2_MEMORY_DMABUF:
input_buffer = input_queue_->GetFreeBufferForFrame(
GetSharedMemoryId(*encoder_input_queue_.front().frame));
// We may have failed to preserve buffer affinity, fallback to any
// buffer in that case.
if (!input_buffer)
input_buffer = input_queue_->GetFreeBuffer();
break;
default:
input_buffer = input_queue_->GetFreeBuffer();
break;
}
// input_buffer cannot be std::nullopt since we checked for
// input_queue_->FreeBuffersCount() > 0 before entering the loop.
DCHECK(input_buffer);
if (!EnqueueInputRecord(std::move(*input_buffer)))
return;
}
if (old_inputs_queued == 0 && input_queue_->QueuedBuffersCount() != 0) {
// We just started up a previously empty queue.
// Queue state changed; signal interrupt.
if (!device_->SetDevicePollInterrupt())
return;
// Shall call VIDIOC_STREAMON if we haven't yet.
do_streamon = !input_queue_->IsStreaming();
}
if (!input_queue_->IsStreaming() && !do_streamon) {
// We don't have to enqueue any buffers in the output queue until we enqueue
// buffers in the input queue. This enables to call S_FMT in Encode() on
// the first frame.
return;
}
// Enqueue all the outputs we can.
const size_t old_outputs_queued = output_queue_->QueuedBuffersCount();
while (auto output_buffer = output_queue_->GetFreeBuffer()) {
if (!EnqueueOutputRecord(std::move(*output_buffer)))
return;
}
if (old_outputs_queued == 0 && output_queue_->QueuedBuffersCount() != 0) {
// We just started up a previously empty queue.
// Queue state changed; signal interrupt.
if (!device_->SetDevicePollInterrupt())
return;
}
// STREAMON in CAPTURE queue first and then OUTPUT queue.
// This is a workaround of a tegra driver bug that STREAMON in CAPTURE queue
// will never return (i.e. blocks |encoder_thread_| forever) if the STREAMON
// in CAPTURE queue is called after STREAMON in OUTPUT queue.
// Once nyan_kitty, which uses tegra driver, reaches EOL, crrev.com/c/1753982
// should be reverted.
if (do_streamon) {
DCHECK(!output_queue_->IsStreaming() && !input_queue_->IsStreaming());
// When VIDIOC_STREAMON can be executed in OUTPUT queue, it is fine to call
// STREAMON in CAPTURE queue.
if (!output_queue_->Streamon()) {
SetErrorState({EncoderStatus::Codes::kEncoderHardwareDriverError,
"Failed to turn on streaming for CAPTURE queue"});
return;
}
if (!input_queue_->Streamon()) {
SetErrorState({EncoderStatus::Codes::kEncoderHardwareDriverError,
"Failed to turn on streaming for OUTPUT queue"});
return;
}
}
}
void V4L2VideoEncodeAccelerator::Dequeue() {
DVLOGF(4);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
TRACE_EVENT0("media,gpu", "V4L2VEA::Dequeue");
// Dequeue completed input (VIDEO_OUTPUT) buffers, and recycle to the free
// list.
while (input_queue_->QueuedBuffersCount() > 0) {
DVLOGF(4) << "inputs queued: " << input_queue_->QueuedBuffersCount();
DCHECK(input_queue_->IsStreaming());
auto ret = input_queue_->DequeueBuffer();
if (!ret.first) {
SetErrorState(
{EncoderStatus::Codes::kEncoderHardwareDriverError,
base::StrCat({"Failed to dequeue buffer in OUTPUT queue, errno=",
base::NumberToString(errno)})});
return;
}
if (!ret.second) {
// We're just out of buffers to dequeue.
break;
}
InputRecord& input_record = input_buffer_map_[ret.second->BufferId()];
input_record.frame = nullptr;
if (input_record.ip_output_buffer_index)
ReuseImageProcessorOutputBuffer(*input_record.ip_output_buffer_index);
}
// Dequeue completed output (VIDEO_CAPTURE) buffers, and recycle to the
// free list. Notify the client that an output buffer is complete.
bool buffer_dequeued = false;
while (output_queue_->QueuedBuffersCount() > 0) {
DCHECK(output_queue_->IsStreaming());
auto ret = output_queue_->DequeueBuffer();
if (!ret.first) {
SetErrorState(
{EncoderStatus::Codes::kEncoderHardwareDriverError,
base::StrCat({"Failed to dequeue buffer in CAPTURE queue, errno=",
base::NumberToString(errno)})});
return;
}
if (!ret.second) {
// We're just out of buffers to dequeue.
break;
}
const uint64_t timestamp_us =
ret.second->GetTimeStamp().tv_usec +
ret.second->GetTimeStamp().tv_sec * base::Time::kMicrosecondsPerSecond;
TRACE_EVENT_NESTABLE_ASYNC_END2(
"media,gpu", "PlatformEncoding.Encode", timestamp_us, "timestamp",
timestamp_us, "size", encoder_input_visible_rect_.size().ToString());
output_buffer_queue_.push_back(std::move(ret.second));
buffer_dequeued = true;
}
if (buffer_dequeued)
PumpBitstreamBuffers();
}
void V4L2VideoEncodeAccelerator::PumpBitstreamBuffers() {
DVLOGF(4);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
while (!output_buffer_queue_.empty()) {
auto output_buf = std::move(output_buffer_queue_.front());
output_buffer_queue_.pop_front();
size_t bitstream_size = base::checked_cast<size_t>(
output_buf->GetPlaneBytesUsed(0) - output_buf->GetPlaneDataOffset(0));
if (bitstream_size > 0) {
if (bitstream_buffer_pool_.empty()) {
DVLOGF(4) << "No free bitstream buffer, skip.";
output_buffer_queue_.push_front(std::move(output_buf));
break;
}
auto buffer_ref = std::move(bitstream_buffer_pool_.back());
auto buffer_id = buffer_ref->id;
bitstream_buffer_pool_.pop_back();
const uint8_t* output_buffer =
static_cast<const uint8_t*>(output_buf->GetPlaneMapping(0)) +
output_buf->GetPlaneDataOffset(0);
size_t output_data_size = CopyIntoOutputBuffer(
output_buffer, bitstream_size, std::move(buffer_ref));
DVLOGF(4) << "returning buffer_id=" << buffer_id
<< ", size=" << output_data_size
<< ", key_frame=" << output_buf->IsKeyframe();
const int64_t timestamp_us = output_buf->GetTimeStamp().tv_usec +
output_buf->GetTimeStamp().tv_sec *
base::Time::kMicrosecondsPerSecond;
TRACE_EVENT2("media,gpu", "V4L2VEA::BitstreamBufferReady", "timestamp",
timestamp_us, "bitstream_buffer_id", buffer_id);
child_task_runner_->PostTask(
FROM_HERE,
base::BindOnce(&Client::BitstreamBufferReady, client_, buffer_id,
GetMetadata(output_buffer, output_data_size,
output_buf->IsKeyframe(),
base::Microseconds(timestamp_us))));
}
if ((encoder_state_ == kFlushing) && output_buf->IsLast()) {
// Notify client that flush has finished successfully. The flush callback
// should be called after notifying the last buffer is ready.
DVLOGF(3) << "Flush completed. Start the encoder again.";
encoder_state_ = kEncoding;
child_task_runner_->PostTask(
FROM_HERE, base::BindOnce(std::move(flush_callback_), true));
// Start the encoder again.
struct v4l2_encoder_cmd cmd;
memset(&cmd, 0, sizeof(cmd));
cmd.cmd = V4L2_ENC_CMD_START;
if (device_->Ioctl(VIDIOC_ENCODER_CMD, &cmd) != 0) {
SetErrorState(
{EncoderStatus::Codes::kEncoderFailedFlush,
base::StrCat({"ioctl() failed: VIDIOC_ENCODER_CMD, errno=",
base::NumberToString(errno)})});
return;
}
}
}
// We may free some V4L2 output buffers above. Enqueue them if needed.
if (output_queue_->FreeBuffersCount() > 0) {
encoder_task_runner_->PostTask(
FROM_HERE,
base::BindOnce(&V4L2VideoEncodeAccelerator::Enqueue, weak_this_));
}
}
bool V4L2VideoEncodeAccelerator::EnqueueInputRecord(
V4L2WritableBufferRef input_buf) {
DVLOGF(4);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK(!encoder_input_queue_.empty());
TRACE_EVENT0("media,gpu", "V4L2VEA::EnqueueInputRecord");
// Enqueue an input (VIDEO_OUTPUT) buffer.
InputFrameInfo frame_info = encoder_input_queue_.front();
if (frame_info.force_keyframe) {
if (!device_->SetExtCtrls(
V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_FORCE_KEY_FRAME)})) {
SetErrorState({EncoderStatus::Codes::kEncoderFailedEncode,
"Failed requesting keyframe"});
return false;
}
}
scoped_refptr<VideoFrame> frame = frame_info.frame;
size_t buffer_id = input_buf.BufferId();
struct timeval timestamp;
timestamp.tv_sec = static_cast<time_t>(frame->timestamp().InSeconds());
timestamp.tv_usec =
frame->timestamp().InMicroseconds() -
frame->timestamp().InSeconds() * base::Time::kMicrosecondsPerSecond;
input_buf.SetTimeStamp(timestamp);
DCHECK_EQ(device_input_layout_->format(), frame->format());
size_t num_planes = GetNumPlanesOfV4L2PixFmt(
Fourcc::FromVideoPixelFormat(device_input_layout_->format(),
!device_input_layout_->is_multi_planar())
->ToV4L2PixFmt());
// Create GpuMemoryBufferHandle for native_input_mode.
gfx::GpuMemoryBufferHandle gmb_handle;
if (input_buf.Memory() == V4L2_MEMORY_DMABUF) {
gmb_handle = CreateGpuMemoryBufferHandle(frame.get());
if (gmb_handle.is_null() || gmb_handle.type != gfx::NATIVE_PIXMAP) {
SetErrorState({EncoderStatus::Codes::kSystemAPICallError,
"Failed to create native GpuMemoryBufferHandle"});
return false;
}
}
for (size_t i = 0; i < num_planes; ++i) {
// Single-buffer input format may have multiple color planes, so bytesused
// of the single buffer should be sum of each color planes' size.
size_t bytesused = 0;
if (num_planes == 1) {
bytesused = VideoFrame::AllocationSize(
frame->format(), device_input_layout_->coded_size());
} else {
bytesused = base::checked_cast<size_t>(
VideoFrame::PlaneSize(frame->format(), i,
device_input_layout_->coded_size())
.GetArea());
}
switch (input_buf.Memory()) {
case V4L2_MEMORY_USERPTR:
// Use buffer_size VideoEncodeAccelerator HW requested by S_FMT.
input_buf.SetPlaneSize(i, device_input_layout_->planes()[i].size);
break;
case V4L2_MEMORY_DMABUF: {
const std::vector<gfx::NativePixmapPlane>& planes =
gmb_handle.native_pixmap_handle.planes;
// TODO(crbug.com/901264): The way to pass an offset within a DMA-buf is
// not defined in V4L2 specification, so we abuse data_offset for now.
// Fix it when we have the right interface, including any necessary
// validation and potential alignment
input_buf.SetPlaneDataOffset(i, planes[i].offset);
bytesused += planes[i].offset;
// Workaround: filling length should not be needed. This is a bug of
// videobuf2 library.
input_buf.SetPlaneSize(
i, device_input_layout_->planes()[i].size + planes[i].offset);
break;
}
default:
NOTREACHED();
}
input_buf.SetPlaneBytesUsed(i, bytesused);
}
TRACE_EVENT_NESTABLE_ASYNC_BEGIN1("media,gpu", "PlatformEncoding.Encode",
frame->timestamp().InMicroseconds(),
"timestamp",
frame->timestamp().InMicroseconds());
switch (input_buf.Memory()) {
case V4L2_MEMORY_USERPTR: {
if (frame->storage_type() != VideoFrame::STORAGE_SHMEM) {
SetErrorState({EncoderStatus::Codes::kInvalidInputFrame,
"VideoFrame doesn't have shared memory"});
return false;
}
// The frame data is readable only and the driver doesn't actually write
// the buffer. But USRPTR buffer needs void*. So const_cast<> is required.
std::vector<void*> user_ptrs(num_planes);
for (size_t i = 0; i < num_planes; ++i) {
user_ptrs[i] = const_cast<uint8_t*>(frame->data(i));
}
if (!std::move(input_buf).QueueUserPtr(std::move(user_ptrs))) {
SetErrorState(
{EncoderStatus::Codes::kEncoderHardwareDriverError,
base::StrCat(
{"Failed queue a USRPTR buffer to input queue, errno=",
base::NumberToString(errno)})});
return false;
}
break;
}
case V4L2_MEMORY_DMABUF: {
if (!std::move(input_buf).QueueDMABuf(
gmb_handle.native_pixmap_handle.planes)) {
SetErrorState(
{EncoderStatus::Codes::kEncoderHardwareDriverError,
base::StrCat(
{"Failed queue a DMABUF buffer to input queue, errno=",
base::NumberToString(errno)})});
return false;
}
// TODO(b/266443239): Remove this workaround once RK3399 boards reaches
// EOL. v4lplugin holds v4l2_buffer in QBUF without duplicating the passed
// fds and resumes the QBUF request later after VIDIOC_QBUF returns. It is
// required to keep the passed fds valid until DQBUF is complete.
if (driver_name_ == "hantro-vpu") {
frame->AddDestructionObserver(base::BindOnce(
[](gfx::GpuMemoryBufferHandle) {}, std::move(gmb_handle)));
}
break;
}
default:
NOTREACHED_IN_MIGRATION();
SetErrorState({EncoderStatus::Codes::kEncoderIllegalState,
base::StrCat({"Unknown input memory type: ",
base::NumberToString(static_cast<int>(
input_buf.Memory()))})});
return false;
}
// Keep |frame| in |input_record| so that a client doesn't use |frame| until
// a driver finishes using it, that is, VIDIOC_DQBUF is called.
InputRecord& input_record = input_buffer_map_[buffer_id];
input_record.frame = frame;
input_record.ip_output_buffer_index = frame_info.ip_output_buffer_index;
encoder_input_queue_.pop();
return true;
}
bool V4L2VideoEncodeAccelerator::EnqueueOutputRecord(
V4L2WritableBufferRef output_buf) {
DVLOGF(4);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
TRACE_EVENT0("media,gpu", "V4L2VEA::EnqueueOutputRecord");
// Enqueue an output (VIDEO_CAPTURE) buffer.
if (!std::move(output_buf).QueueMMap()) {
SetErrorState({EncoderStatus::Codes::kEncoderHardwareDriverError,
base::StrCat({"Failed to QueueMMap, errno=",
base::NumberToString(errno)})});
return false;
}
return true;
}
bool V4L2VideoEncodeAccelerator::StartDevicePoll() {
DVLOGF(3);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK(!device_poll_thread_.IsRunning());
// Start up the device poll thread and schedule its first DevicePollTask().
if (!device_poll_thread_.Start()) {
SetErrorState({EncoderStatus::Codes::kSystemAPICallError,
"StartDevicePoll(): Device thread failed to start"});
return false;
}
// Enqueue a poll task with no devices to poll on -- it will wait only on the
// interrupt fd.
// base::Unretained(this) is safe, because device_poll_thread_ is owned by
// *this and stops before *this destruction.
device_poll_thread_.task_runner()->PostTask(
FROM_HERE, base::BindOnce(&V4L2VideoEncodeAccelerator::DevicePollTask,
base::Unretained(this), false));
return true;
}
bool V4L2VideoEncodeAccelerator::StopDevicePoll() {
DVLOGF(3);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
if (device_->IsValid()) {
// Signal the DevicePollTask() to stop, and stop the device poll thread.
if (!device_->SetDevicePollInterrupt()) {
return false;
}
device_poll_thread_.Stop();
// Clear the interrupt now, to be sure.
if (!device_->ClearDevicePollInterrupt()) {
return false;
}
}
// Tegra driver cannot call Streamoff() when the stream is off, so we check
// IsStreaming() first.
if (input_queue_ && input_queue_->IsStreaming() && !input_queue_->Streamoff())
return false;
if (output_queue_ && output_queue_->IsStreaming() &&
!output_queue_->Streamoff())
return false;
// Reset all our accounting info.
while (!encoder_input_queue_.empty())
encoder_input_queue_.pop();
for (auto& [frame, ip_output_buffer_index] : input_buffer_map_) {
frame = nullptr;
}
bitstream_buffer_pool_.clear();
DVLOGF(3) << "device poll stopped";
return true;
}
void V4L2VideoEncodeAccelerator::DevicePollTask(bool poll_device) {
DVLOGF(4);
DCHECK(device_poll_thread_.task_runner()->BelongsToCurrentThread());
bool event_pending;
if (!device_->Poll(poll_device, &event_pending)) {
SetErrorState({EncoderStatus::Codes::kSystemAPICallError,
"Failed to start device polloing"});
return;
}
// All processing should happen on ServiceDeviceTask(), since we shouldn't
// touch encoder state from this thread.
encoder_task_runner_->PostTask(
FROM_HERE, base::BindOnce(&V4L2VideoEncodeAccelerator::ServiceDeviceTask,
weak_this_));
}
void V4L2VideoEncodeAccelerator::SetErrorState(EncoderStatus status) {
// We can touch encoder_state_ only if this is the encoder thread or the
// encoder thread isn't running.
if (!encoder_task_runner_->RunsTasksInCurrentSequence()) {
encoder_task_runner_->PostTask(
FROM_HERE, base::BindOnce(&V4L2VideoEncodeAccelerator::SetErrorState,
weak_this_, status));
return;
}
CHECK(!status.is_ok());
LOG(ERROR) << "SetErrorState: code=" << static_cast<int>(status.code())
<< ", message=" << status.message();
// Post NotifyErrorStatus() only if we are already initialized, as the API
// does not allow doing so before that.
if (encoder_state_ != kError && encoder_state_ != kUninitialized) {
LOG(ERROR) << "Call NotifyErrorStatus(): code="
<< static_cast<int>(status.code())
<< ", message=" << status.message();
CHECK(child_task_runner_);
child_task_runner_->PostTask(
FROM_HERE,
base::BindOnce(&Client::NotifyErrorStatus, client_, std::move(status)));
}
encoder_state_ = kError;
}
void V4L2VideoEncodeAccelerator::RequestEncodingParametersChangeTask(
const VideoBitrateAllocation& bitrate_allocation,
uint32_t framerate,
const std::optional<gfx::Size>& size) {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
if (size.has_value()) {
SetErrorState({EncoderStatus::Codes::kEncoderUnsupportedConfig,
"Update output frame size is not supported"});
return;
}
if (current_bitrate_allocation_ == bitrate_allocation &&
current_framerate_ == framerate) {
return;
}
DVLOGF(2) << "bitrate=" << bitrate_allocation.ToString()
<< ", framerate=" << framerate;
if (bitrate_allocation.GetMode() != current_bitrate_allocation_.GetMode()) {
SetErrorState({EncoderStatus::Codes::kEncoderUnsupportedConfig,
"Bitrate mode changed during encoding"});
return;
}
TRACE_EVENT2("media,gpu", "V4L2VEA::RequestEncodingParametersChangeTask",
"bitrate", current_bitrate_allocation_.ToString(), "framerate",
framerate);
if (current_bitrate_allocation_ != bitrate_allocation) {
switch (bitrate_allocation.GetMode()) {
case Bitrate::Mode::kVariable:
device_->SetExtCtrls(V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_BITRATE_PEAK,
bitrate_allocation.GetPeakBps())});
// Both the average and peak bitrate are to be set in VBR.
// Only the average bitrate are to be set in CBR.
[[fallthrough]];
case Bitrate::Mode::kConstant:
if (h264_l1t2_enabled_) {
if (!device_->SetExtCtrls(
V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_HIER_CODING_L0_BR,
bitrate_allocation.GetBitrateBps(0u, 0u)),
V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_HIER_CODING_L1_BR,
bitrate_allocation.GetBitrateBps(0u, 1u))})) {
SetErrorState({EncoderStatus::Codes::kEncoderHardwareDriverError,
"Failed to change average bitrate"});
return;
}
} else if (!device_->SetExtCtrls(
V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(
V4L2_CID_MPEG_VIDEO_BITRATE,
bitrate_allocation.GetBitrateBps(0u, 0u))})) {
SetErrorState({EncoderStatus::Codes::kEncoderHardwareDriverError,
"Failed to change average bitrate"});
return;
}
break;
case Bitrate::Mode::kExternal:
SetErrorState({EncoderStatus::Codes::kEncoderUnsupportedConfig,
"Unsupported rate control mode."});
return;
}
}
if (current_framerate_ != framerate) {
struct v4l2_streamparm parms;
memset(&parms, 0, sizeof(parms));
parms.type = V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE;
// Note that we are provided "frames per second" but V4L2 expects "time per
// frame"; hence we provide the reciprocal of the framerate here.
parms.parm.output.timeperframe.numerator = 1;
parms.parm.output.timeperframe.denominator = framerate;
if (device_->Ioctl(VIDIOC_S_PARM, &parms) != 0) {
SetErrorState({EncoderStatus::Codes::kEncoderHardwareDriverError,
base::StrCat({"ioctl() failed: VIDIOC_S_PARM, errno=",
base::NumberToString(errno)})});
return;
}
}
current_bitrate_allocation_ = bitrate_allocation;
current_framerate_ = framerate;
}
bool V4L2VideoEncodeAccelerator::SetOutputFormat(
VideoCodecProfile output_profile) {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK(!input_queue_->IsStreaming());
DCHECK(!output_queue_->IsStreaming());
DCHECK(!encoder_input_visible_rect_.IsEmpty());
output_buffer_byte_size_ =
GetEncodeBitstreamBufferSize(encoder_input_visible_rect_.size());
// Sets 0 to width and height in CAPTURE queue, which should be ignored by the
// driver.
std::optional<struct v4l2_format> format = output_queue_->SetFormat(
output_format_fourcc_, gfx::Size(), output_buffer_byte_size_);
if (!format) {
return false;
}
// Device might have adjusted the required output size.
size_t adjusted_output_buffer_size =
base::checked_cast<size_t>(format->fmt.pix_mp.plane_fmt[0].sizeimage);
output_buffer_byte_size_ = adjusted_output_buffer_size;
return true;
}
std::optional<struct v4l2_format>
V4L2VideoEncodeAccelerator::NegotiateInputFormat(VideoPixelFormat input_format,
const gfx::Size& size) {
VLOGF(2);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK(!input_queue_->IsStreaming());
DCHECK(!output_queue_->IsStreaming());
// First see if the device can use the provided format directly.
std::vector<uint32_t> pix_fmt_candidates;
auto input_fourcc = Fourcc::FromVideoPixelFormat(input_format, false);
if (!input_fourcc) {
LOG(ERROR) << "Invalid input format "
<< VideoPixelFormatToString(input_format);
return std::nullopt;
}
pix_fmt_candidates.push_back(input_fourcc->ToV4L2PixFmt());
// Second try preferred input formats for both single-planar and
// multi-planar.
for (auto preferred_format :
device_->PreferredInputFormat(V4L2Device::Type::kEncoder)) {
pix_fmt_candidates.push_back(preferred_format);
}
for (const auto pix_fmt : pix_fmt_candidates) {
DVLOGF(3) << "Trying S_FMT with " << FourccToString(pix_fmt);
std::optional<struct v4l2_format> format =
input_queue_->SetFormat(pix_fmt, size, 0);
if (!format)
continue;
DVLOGF(3) << "Success: S_FMT with " << FourccToString(pix_fmt);
device_input_layout_ = V4L2FormatToVideoFrameLayout(*format);
if (!device_input_layout_) {
LOG(ERROR) << "Invalid device_input_layout_";
return std::nullopt;
}
DVLOG(3) << "Negotiated device_input_layout_: " << *device_input_layout_;
if (!gfx::Rect(device_input_layout_->coded_size())
.Contains(gfx::Rect(size))) {
LOG(ERROR) << "Input size " << size.ToString()
<< " exceeds encoder capability. Size encoder can handle: "
<< device_input_layout_->coded_size().ToString();
return std::nullopt;
}
// Make sure that the crop is preserved as we have changed the input
// resolution.
if (!ApplyCrop()) {
return std::nullopt;
}
return format;
}
return std::nullopt;
}
bool V4L2VideoEncodeAccelerator::ApplyCrop() {
VLOGF(2);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
struct v4l2_rect visible_rect;
visible_rect.left = encoder_input_visible_rect_.x();
visible_rect.top = encoder_input_visible_rect_.y();
visible_rect.width = encoder_input_visible_rect_.width();
visible_rect.height = encoder_input_visible_rect_.height();
struct v4l2_selection selection_arg;
memset(&selection_arg, 0, sizeof(selection_arg));
selection_arg.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
selection_arg.target = V4L2_SEL_TGT_CROP;
selection_arg.r = visible_rect;
// The width and height might be adjusted by driver.
// Need to read it back and set to |encoder_input_visible_rect_|.
if (device_->Ioctl(VIDIOC_S_SELECTION, &selection_arg) == 0) {
DVLOGF(3) << "VIDIOC_S_SELECTION is supported";
visible_rect = selection_arg.r;
} else {
DVLOGF(3) << "Fallback to VIDIOC_S/G_CROP";
struct v4l2_crop crop;
memset(&crop, 0, sizeof(crop));
crop.type = V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE;
crop.c = visible_rect;
if (device_->Ioctl(VIDIOC_S_CROP, &crop) != 0) {
SetErrorState({EncoderStatus::Codes::kEncoderHardwareDriverError,
base::StrCat({"ioctl() failed: VIDIOC_S_CROP, errno=",
base::NumberToString(errno)})});
return false;
}
if (device_->Ioctl(VIDIOC_G_CROP, &crop) != 0) {
SetErrorState({EncoderStatus::Codes::kEncoderHardwareDriverError,
base::StrCat({"ioctl() failed: VIDIOC_G_CROP",
base::NumberToString(errno)})});
return false;
}
visible_rect = crop.c;
}
const gfx::Rect adjusted_visible_rect(visible_rect.left, visible_rect.top,
visible_rect.width,
visible_rect.height);
if (encoder_input_visible_rect_ != adjusted_visible_rect) {
LOG(ERROR) << "Unsupported visible rectangle: "
<< encoder_input_visible_rect_.ToString()
<< ", the rectangle adjusted by the driver: "
<< adjusted_visible_rect.ToString();
return false;
}
return true;
}
bool V4L2VideoEncodeAccelerator::SetFormats(VideoPixelFormat input_format,
VideoCodecProfile output_profile) {
VLOGF(2);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK(!input_queue_->IsStreaming());
DCHECK(!output_queue_->IsStreaming());
if (!SetOutputFormat(output_profile)) {
SetErrorState({EncoderStatus::Codes::kEncoderUnsupportedProfile,
base::StrCat({"Unsupported codec profile: ",
GetProfileName(output_profile)})});
return false;
}
gfx::Size input_size = encoder_input_visible_rect_.size();
if (native_input_mode_) {
auto input_layout = GetPlatformVideoFrameLayout(
input_format, encoder_input_visible_rect_.size(),
gfx::BufferUsage::VEA_READ_CAMERA_AND_CPU_READ_WRITE);
if (!input_layout)
return false;
input_size = gfx::Size(input_layout->planes()[0].stride,
input_layout->coded_size().height());
}
DCHECK(input_frame_size_.IsEmpty());
auto v4l2_format = NegotiateInputFormat(input_format, input_size);
if (!v4l2_format) {
SetErrorState({EncoderStatus::Codes::kUnsupportedFrameFormat,
base::StrCat({"Unsupported input format: ",
VideoPixelFormatToString(input_format)})});
return false;
}
if (native_input_mode_) {
input_frame_size_ = VideoFrame::DetermineAlignedSize(
input_format, encoder_input_visible_rect_.size());
} else {
input_frame_size_ = V4L2Device::AllocatedSizeFromV4L2Format(*v4l2_format);
}
return true;
}
bool V4L2VideoEncodeAccelerator::InitControls(const Config& config) {
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
// Don't expect other output formats.
CHECK(output_format_fourcc_ == V4L2_PIX_FMT_H264 ||
output_format_fourcc_ == V4L2_PIX_FMT_VP8);
// Enable frame-level bitrate control. This is the only mandatory control.
if (!device_->SetExtCtrls(
V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_FRAME_RC_ENABLE, 1)})) {
SetErrorState({EncoderStatus::Codes::kEncoderHardwareDriverError,
"Failed enabling bitrate control"});
return false;
}
switch (output_format_fourcc_) {
case V4L2_PIX_FMT_H264:
if (!InitControlsH264(config)) {
return false;
}
break;
case V4L2_PIX_FMT_VP8:
InitControlsVP8(config);
break;
default:
NOTREACHED_IN_MIGRATION()
<< "Unsupported codec " << FourccToString(output_format_fourcc_);
return false;
}
// Optional controls:
// - Enable macroblock-level bitrate control.
device_->SetExtCtrls(V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_MB_RC_ENABLE, 1)});
// - Set GOP length, or default 0 to disable periodic key frames.
device_->SetGOPLength(config.gop_length.value_or(0));
return true;
}
bool V4L2VideoEncodeAccelerator::InitControlsH264(const Config& config) {
#ifndef V4L2_CID_MPEG_VIDEO_PREPEND_SPSPPS_TO_IDR
#define V4L2_CID_MPEG_VIDEO_PREPEND_SPSPPS_TO_IDR (V4L2_CID_MPEG_BASE + 644)
#endif
// Request to inject SPS and PPS before each IDR, if the device supports
// that feature. Otherwise we'll have to cache and inject ourselves.
if (device_->IsCtrlExposed(V4L2_CID_MPEG_VIDEO_PREPEND_SPSPPS_TO_IDR)) {
if (!device_->SetExtCtrls(
V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_PREPEND_SPSPPS_TO_IDR, 1)})) {
SetErrorState(
{EncoderStatus::Codes::kEncoderHardwareDriverError,
"Failed to set V4L2_CID_MPEG_VIDEO_PREPEND_SPSPPS_TO_IDR to 1"});
return false;
}
inject_sps_and_pps_ = false;
DVLOGF(2) << "Device supports injecting SPS+PPS before each IDR";
} else {
inject_sps_and_pps_ = true;
DVLOGF(2) << "Will inject SPS+PPS before each IDR, unsupported by device";
}
// No B-frames, for lowest decoding latency.
device_->SetExtCtrls(V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_B_FRAMES, 0)});
// Set H.264 profile.
int32_t profile_value =
V4L2Device::VideoCodecProfileToV4L2H264Profile(config.output_profile);
if (profile_value < 0) {
SetErrorState({EncoderStatus::Codes::kEncoderUnsupportedProfile,
base::StrCat({"Unexpected h264 profile: ",
GetProfileName(config.output_profile)})});
return false;
}
if (!device_->SetExtCtrls(
V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_PROFILE, profile_value)})) {
SetErrorState({EncoderStatus::Codes::kEncoderUnsupportedProfile,
base::StrCat({"Unsupported h264 profile: ",
GetProfileName(config.output_profile)})});
return false;
}
// Set H.264 output level from config. Use Level 4.0 as fallback default.
uint8_t h264_level = config.h264_output_level.value_or(H264SPS::kLevelIDC4p0);
constexpr int kH264MacroblockSizeInPixels = 16;
const uint32_t framerate = config.framerate;
const uint32_t mb_width =
base::bits::AlignUpDeprecatedDoNotUse(config.input_visible_size.width(),
kH264MacroblockSizeInPixels) /
kH264MacroblockSizeInPixels;
const uint32_t mb_height =
base::bits::AlignUpDeprecatedDoNotUse(config.input_visible_size.height(),
kH264MacroblockSizeInPixels) /
kH264MacroblockSizeInPixels;
const uint32_t framesize_in_mbs = mb_width * mb_height;
// Check whether the h264 level is valid.
if (!CheckH264LevelLimits(config.output_profile, h264_level,
config.bitrate.target_bps(), framerate,
framesize_in_mbs)) {
std::optional<uint8_t> valid_level =
FindValidH264Level(config.output_profile, config.bitrate.target_bps(),
framerate, framesize_in_mbs);
if (!valid_level) {
SetErrorState(
{EncoderStatus::Codes::kEncoderInitializationError,
base::StrCat({"Could not find a valid h264 level for"
" profile=",
GetProfileName(config.output_profile), " bitrate=",
base::NumberToString(config.bitrate.target_bps()),
" framerate=", base::NumberToString(framerate),
" size=", config.input_visible_size.ToString()})});
return false;
}
h264_level = *valid_level;
}
int32_t level_value = V4L2Device::H264LevelIdcToV4L2H264Level(h264_level);
device_->SetExtCtrls(
V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_LEVEL, level_value)});
// Ask not to put SPS and PPS into separate bitstream buffers.
device_->SetExtCtrls(
V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_HEADER_MODE,
V4L2_MPEG_VIDEO_HEADER_MODE_JOINED_WITH_1ST_FRAME)});
// H264 coding tools parameter. Since a driver may not support some of them,
// EXT_S_CTRLS is called to each of them to enable as many coding tools as
// possible.
// Don't produce Intra-only frame.
device_->SetExtCtrls(V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_I_PERIOD, 0)});
// Enable deblocking filter.
device_->SetExtCtrls(
V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_LOOP_FILTER_MODE,
V4L2_MPEG_VIDEO_H264_LOOP_FILTER_MODE_ENABLED)});
// Use CABAC in Main and High profiles.
if (config.output_profile == H264PROFILE_MAIN ||
config.output_profile == H264PROFILE_HIGH) {
device_->SetExtCtrls(
V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_ENTROPY_MODE,
V4L2_MPEG_VIDEO_H264_ENTROPY_MODE_CABAC)});
}
// Use 8x8 transform in High profile.
if (config.output_profile == H264PROFILE_HIGH) {
device_->SetExtCtrls(
V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_8X8_TRANSFORM, true)});
}
// Quantization parameter. The h264 qp range is 0-51.
// Note: Webrtc default values are 24 and 37 respectively, see
// h264_encoder_impl.cc.
// These values were previously copied from the VA-API encoder.
// The MAX_QP parameter needed modification to 51 due to
// b/274867782 and b/241549978.
// Ignore return values as these controls are optional.
device_->SetExtCtrls(V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_MAX_QP, 51)});
// Don't set MIN_QP with other controls since it is not supported by
// some devices and may prevent other controls from being set.
// The MIN_QP needed modification due to b/280853786
// The value 18 was tuned experimentally to let the test pass but
// to be close to the original one
device_->SetExtCtrls(V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_MIN_QP, 18)});
if (h264_l1t2_enabled_) {
if (!device_->SetExtCtrls(
V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_HIERARCHICAL_CODING, 1),
V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_HIERARCHICAL_CODING_TYPE,
V4L2_MPEG_VIDEO_H264_HIERARCHICAL_CODING_P),
V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_H264_HIERARCHICAL_CODING_LAYER,
2u)})) {
SetErrorState(
{EncoderStatus::Codes::kEncoderInitializationError,
base::StrCat(
{"h264 hierachical coding was requested, but configuration of "
"the V4L2_CID_MPEG_VIDEO_H264_HIERARCHICAL_CODING* controls "
"failed."})});
return false;
}
}
return true;
}
void V4L2VideoEncodeAccelerator::InitControlsVP8(const Config& config) {
// Quantization parameter. They are vp8 ac/dc indices and their ranges are
// 0-127. These values were copied from the VA-API encoder.
// Ignore return values as these controls are optional.
device_->SetExtCtrls(V4L2_CTRL_CLASS_MPEG,
{V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_VPX_MIN_QP, 4),
V4L2ExtCtrl(V4L2_CID_MPEG_VIDEO_VPX_MAX_QP, 117)});
}
bool V4L2VideoEncodeAccelerator::CreateInputBuffers() {
VLOGF(2);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK(!input_queue_->IsStreaming());
// If using DMABUF input, we want to reuse the same V4L2 buffer index
// for the same input buffer as much as possible. But we don't know in advance
// how many different input buffers we will get. Therefore we allocate as
// many V4L2 buffers as possible (VIDEO_MAX_FRAME == 32). Unused indexes
// won't have a tangible cost since they don't have backing memory.
size_t num_buffers;
switch (input_memory_type_) {
case V4L2_MEMORY_DMABUF:
num_buffers = VIDEO_MAX_FRAME;
break;
default:
num_buffers = kInputBufferCount;
break;
}
if (input_queue_->AllocateBuffers(num_buffers, input_memory_type_,
/*incoherent=*/false) < kInputBufferCount) {
SetErrorState({EncoderStatus::Codes::kEncoderHardwareDriverError,
"Failed to allocate V4L2 input buffers."});
return false;
}
DCHECK(input_buffer_map_.empty());
input_buffer_map_.resize(input_queue_->AllocatedBuffersCount());
return true;
}
bool V4L2VideoEncodeAccelerator::CreateOutputBuffers() {
VLOGF(2);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
DCHECK(!output_queue_->IsStreaming());
if (output_queue_->AllocateBuffers(kOutputBufferCount, V4L2_MEMORY_MMAP,
/*incoherent=*/false) <
kOutputBufferCount) {
SetErrorState(
{EncoderStatus::Codes::kEncoderInitializationError,
base::StrCat({"Failed to allocate V4L2 output buffers, errno=",
base::NumberToString(errno)})});
return false;
}
return true;
}
void V4L2VideoEncodeAccelerator::DestroyInputBuffers() {
VLOGF(2);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
if (!input_queue_ || input_queue_->AllocatedBuffersCount() == 0)
return;
DCHECK(!input_queue_->IsStreaming());
if (!input_queue_->DeallocateBuffers())
VLOGF(1) << "Failed to deallocate V4L2 input buffers";
input_buffer_map_.clear();
}
void V4L2VideoEncodeAccelerator::DestroyOutputBuffers() {
VLOGF(2);
DCHECK_CALLED_ON_VALID_SEQUENCE(encoder_sequence_checker_);
if (!output_queue_ || output_queue_->AllocatedBuffersCount() == 0)
return;
DCHECK(!output_queue_->IsStreaming());
if (!output_queue_->DeallocateBuffers())
VLOGF(1) << "Failed to deallocate V4L2 output buffers";
}
} // namespace media
Codebase Browser
chromium