blob: e0b592e826c318a1095e90df4f5366f5c739ff7f [file] [log] [blame]
#include "hardware_composer.h"
#include <log/log.h>
#include <cutils/properties.h>
#include <cutils/sched_policy.h>
#include <fcntl.h>
#include <poll.h>
#include <sync/sync.h>
#include <sys/eventfd.h>
#include <sys/prctl.h>
#include <sys/resource.h>
#include <sys/system_properties.h>
#include <sys/timerfd.h>
#include <unistd.h>
#include <utils/Trace.h>
#include <algorithm>
#include <functional>
#include <map>
#include <dvr/performance_client_api.h>
#include <private/dvr/clock_ns.h>
#include <private/dvr/display_types.h>
#include <private/dvr/pose_client_internal.h>
#include <private/dvr/sync_util.h>
#include "debug_hud_data.h"
#include "screenshot_service.h"
using android::pdx::LocalHandle;
namespace android {
namespace dvr {
namespace {
// If the number of pending fences goes over this count at the point when we
// are about to submit a new frame to HWC, we will drop the frame. This should
// be a signal that the display driver has begun queuing frames. Note that with
// smart displays (with RAM), the fence is signaled earlier than the next vsync,
// at the point when the DMA to the display completes. Currently we use a smart
// display and the EDS timing coincides with zero pending fences, so this is 0.
constexpr int kAllowedPendingFenceCount = 0;
// If we think we're going to miss vsync by more than this amount, skip the
// frame.
constexpr int64_t kFrameSkipThresholdNs = 4000000; // 4ms
// Counter PostLayers() deficiency by requiring apps to produce a frame at least
// 2.5ms before vsync. See b/28881672.
constexpr int64_t kFrameTimeEstimateMin = 2500000; // 2.5ms
constexpr size_t kDefaultDisplayConfigCount = 32;
constexpr float kMetersPerInch = 0.0254f;
const char kBacklightBrightnessSysFile[] =
"/sys/class/leds/lcd-backlight/brightness";
const char kPrimaryDisplayVSyncEventFile[] =
"/sys/class/graphics/fb0/vsync_event";
const char kPrimaryDisplayWaitPPEventFile[] = "/sys/class/graphics/fb0/wait_pp";
const char kDvrPerformanceProperty[] = "sys.dvr.performance";
const char kRightEyeOffsetProperty[] = "dreamos.right_eye_offset_ns";
// Returns our best guess for the time the compositor will spend rendering the
// next frame.
int64_t GuessFrameTime(int compositor_visible_layer_count) {
// The cost of asynchronous EDS and lens warp is currently measured at 2.5ms
// for one layer and 7ms for two layers, but guess a higher frame time to
// account for CPU overhead. This guess is only used before we've measured the
// actual time to render a frame for the current compositor configuration.
switch (compositor_visible_layer_count) {
case 0:
return 500000; // .5ms
case 1:
return 5000000; // 5ms
default:
return 10500000; // 10.5ms
}
}
// Get time offset from a vsync to when the pose for that vsync should be
// predicted out to. For example, if scanout gets halfway through the frame
// at the halfway point between vsyncs, then this could be half the period.
// With global shutter displays, this should be changed to the offset to when
// illumination begins. Low persistence adds a frame of latency, so we predict
// to the center of the next frame.
inline int64_t GetPosePredictionTimeOffset(int64_t vsync_period_ns) {
return (vsync_period_ns * 150) / 100;
}
} // anonymous namespace
HardwareComposer::HardwareComposer()
: HardwareComposer(nullptr) {
}
HardwareComposer::HardwareComposer(Hwc2::Composer* hwc2_hidl)
: hwc2_hidl_(hwc2_hidl),
display_transform_(HWC_TRANSFORM_NONE),
display_surfaces_updated_(false),
hardware_layers_need_update_(false),
display_on_(false),
active_layer_count_(0),
gpu_layer_(nullptr),
terminate_post_thread_event_fd_(-1),
pause_post_thread_(true),
backlight_brightness_fd_(-1),
primary_display_vsync_event_fd_(-1),
primary_display_wait_pp_fd_(-1),
vsync_sleep_timer_fd_(-1),
last_vsync_timestamp_(0),
vsync_count_(0),
frame_skip_count_(0),
pose_client_(nullptr) {
std::transform(layer_storage_.begin(), layer_storage_.end(), layers_.begin(),
[](auto& layer) { return &layer; });
callbacks_ = new ComposerCallback;
}
HardwareComposer::~HardwareComposer(void) {
if (!IsSuspended()) {
Suspend();
}
}
bool HardwareComposer::Resume() {
std::lock_guard<std::mutex> autolock(layer_mutex_);
if (!IsSuspended()) {
ALOGE("HardwareComposer::Resume: HardwareComposer is already running.");
return false;
}
int32_t ret = HWC2_ERROR_NONE;
static const uint32_t attributes[] = {
HWC_DISPLAY_WIDTH, HWC_DISPLAY_HEIGHT, HWC_DISPLAY_VSYNC_PERIOD,
HWC_DISPLAY_DPI_X, HWC_DISPLAY_DPI_Y, HWC_DISPLAY_NO_ATTRIBUTE,
};
std::vector<Hwc2::Config> configs;
ret = (int32_t)hwc2_hidl_->getDisplayConfigs(HWC_DISPLAY_PRIMARY, &configs);
if (ret != HWC2_ERROR_NONE) {
ALOGE("HardwareComposer: Failed to get display configs");
return false;
}
uint32_t num_configs = configs.size();
for (size_t i = 0; i < num_configs; i++) {
ALOGI("HardwareComposer: cfg[%zd/%zd] = 0x%08x", i, num_configs,
configs[i]);
ret = GetDisplayMetrics(HWC_DISPLAY_PRIMARY, configs[i],
&native_display_metrics_);
if (ret != HWC2_ERROR_NONE) {
ALOGE("HardwareComposer: Failed to get display attributes %d", ret);
continue;
} else {
ret =
(int32_t)hwc2_hidl_->setActiveConfig(HWC_DISPLAY_PRIMARY, configs[i]);
if (ret != HWC2_ERROR_NONE) {
ALOGE("HardwareComposer: Failed to set display configuration; ret=%d",
ret);
continue;
}
break;
}
}
if (ret != HWC2_ERROR_NONE) {
ALOGE("HardwareComposer: Could not set a valid display configuration.");
return false;
}
// Set the display metrics but never use rotation to avoid the long latency of
// rotation processing in hwc.
display_transform_ = HWC_TRANSFORM_NONE;
display_metrics_ = native_display_metrics_;
ALOGI(
"HardwareComposer: primary display attributes: width=%d height=%d "
"vsync_period_ns=%d DPI=%dx%d",
native_display_metrics_.width, native_display_metrics_.height,
native_display_metrics_.vsync_period_ns, native_display_metrics_.dpi.x,
native_display_metrics_.dpi.y);
// Always turn off vsync when we start.
EnableVsync(false);
constexpr int format = HAL_PIXEL_FORMAT_RGBA_8888;
constexpr int usage =
GRALLOC_USAGE_HW_FB | GRALLOC_USAGE_HW_COMPOSER | GRALLOC_USAGE_HW_RENDER;
framebuffer_target_ = std::make_shared<IonBuffer>(
native_display_metrics_.width, native_display_metrics_.height, format,
usage);
// Associate each Layer instance with a hardware composer layer.
for (auto layer : layers_) {
layer->Initialize(hwc2_hidl_.get(), &native_display_metrics_);
}
// Open the backlight brightness control sysfs node.
backlight_brightness_fd_ = LocalHandle(kBacklightBrightnessSysFile, O_RDWR);
ALOGW_IF(!backlight_brightness_fd_,
"HardwareComposer: Failed to open backlight brightness control: %s",
strerror(errno));
// Open the vsync event node for the primary display.
// TODO(eieio): Move this into a platform-specific class.
primary_display_vsync_event_fd_ =
LocalHandle(kPrimaryDisplayVSyncEventFile, O_RDONLY);
ALOGE_IF(!primary_display_vsync_event_fd_,
"HardwareComposer: Failed to open vsync event node for primary "
"display: %s",
strerror(errno));
// Open the wait pingpong status node for the primary display.
// TODO(eieio): Move this into a platform-specific class.
primary_display_wait_pp_fd_ =
LocalHandle(kPrimaryDisplayWaitPPEventFile, O_RDONLY);
ALOGE_IF(
!primary_display_wait_pp_fd_,
"HardwareComposer: Failed to open wait_pp node for primary display: %s",
strerror(errno));
// Create a timerfd based on CLOCK_MONOTINIC.
vsync_sleep_timer_fd_.Reset(timerfd_create(CLOCK_MONOTONIC, 0));
LOG_ALWAYS_FATAL_IF(
!vsync_sleep_timer_fd_,
"HardwareComposer: Failed to create vsync sleep timerfd: %s",
strerror(errno));
// Connect to pose service.
pose_client_ = dvrPoseCreate();
ALOGE_IF(!pose_client_, "HardwareComposer: Failed to create pose client");
// Variables used to control the post thread state
pause_post_thread_ = false;
terminate_post_thread_event_fd_.Reset(eventfd(0, EFD_CLOEXEC | EFD_NONBLOCK));
LOG_ALWAYS_FATAL_IF(
!terminate_post_thread_event_fd_,
"HardwareComposer: Failed to create terminate PostThread event fd : %s",
strerror(errno));
// If get_id() is the default thread::id object, it has not been created yet
if (post_thread_.get_id() == std::thread::id()) {
post_thread_ = std::thread(&HardwareComposer::PostThread, this);
} else {
UpdateDisplayState();
thread_pause_semaphore_.notify_one();
}
return true;
}
bool HardwareComposer::Suspend() {
// Wait for any pending layer operations to finish
std::unique_lock<std::mutex> layer_lock(layer_mutex_);
if (IsSuspended()) {
ALOGE("HardwareComposer::Suspend: HardwareComposer is already suspended.");
return false;
}
PausePostThread();
EnableVsync(false);
SetPowerMode(HWC_DISPLAY_PRIMARY, HWC2_POWER_MODE_OFF);
backlight_brightness_fd_.Close();
primary_display_vsync_event_fd_.Close();
primary_display_wait_pp_fd_.Close();
vsync_sleep_timer_fd_.Close();
retire_fence_fds_.clear();
gpu_layer_ = nullptr;
// We have to destroy the layers before we close the hwc device
for (size_t i = 0; i < kMaxHardwareLayers; ++i) {
layers_[i]->Reset();
}
active_layer_count_ = 0;
framebuffer_target_.reset();
//hwc2_hidl_.reset();
if (pose_client_)
dvrPoseDestroy(pose_client_);
return true;
}
void HardwareComposer::PausePostThread() {
pause_post_thread_ = true;
int error = eventfd_write(terminate_post_thread_event_fd_.Get(), 1);
ALOGE_IF(error,
"HardwareComposer::PausePostThread: could not write post "
"thread termination event fd : %d",
error);
std::unique_lock<std::mutex> wait_for_thread(thread_pause_mutex_);
terminate_post_thread_event_fd_.Close();
}
DisplayMetrics HardwareComposer::GetHmdDisplayMetrics() const {
vec2i screen_size(display_metrics_.width, display_metrics_.height);
DisplayOrientation orientation =
(display_metrics_.width > display_metrics_.height
? DisplayOrientation::kLandscape
: DisplayOrientation::kPortrait);
float dpi_x = static_cast<float>(display_metrics_.dpi.x) / 1000.0f;
float dpi_y = static_cast<float>(display_metrics_.dpi.y) / 1000.0f;
float meters_per_pixel_x = kMetersPerInch / dpi_x;
float meters_per_pixel_y = kMetersPerInch / dpi_y;
vec2 meters_per_pixel(meters_per_pixel_x, meters_per_pixel_y);
double frame_duration_s =
static_cast<double>(display_metrics_.vsync_period_ns) / 1000000000.0;
// TODO(hendrikw): Hard coding to 3mm. The Pixel is actually 4mm, but it
// seems that their tray to lens distance is wrong too, which
// offsets this, at least for the pixel.
float border_size = 0.003f;
return DisplayMetrics(screen_size, meters_per_pixel, border_size,
static_cast<float>(frame_duration_s), orientation);
}
int32_t HardwareComposer::Validate(hwc2_display_t display) {
uint32_t num_types;
uint32_t num_requests;
int32_t error =
(int32_t)hwc2_hidl_->validateDisplay(display, &num_types, &num_requests);
if (error == HWC2_ERROR_HAS_CHANGES) {
// TODO(skiazyk): We might need to inspect the requested changes first, but
// so far it seems like we shouldn't ever hit a bad state.
// error = hwc2_funcs_.accept_display_changes_fn_(hardware_composer_device_,
// display);
error = (int32_t)hwc2_hidl_->acceptDisplayChanges(display);
}
return error;
}
int32_t HardwareComposer::EnableVsync(bool enabled) {
return (int32_t)hwc2_hidl_->setVsyncEnabled(
HWC_DISPLAY_PRIMARY,
(Hwc2::IComposerClient::Vsync)(enabled ? HWC2_VSYNC_ENABLE
: HWC2_VSYNC_DISABLE));
}
int32_t HardwareComposer::Present(hwc2_display_t display) {
int32_t present_fence;
int32_t error = (int32_t)hwc2_hidl_->presentDisplay(display, &present_fence);
// According to the documentation, this fence is signaled at the time of
// vsync/DMA for physical displays.
if (error == HWC2_ERROR_NONE) {
ATRACE_INT("HardwareComposer: VsyncFence", present_fence);
retire_fence_fds_.emplace_back(present_fence);
} else {
ATRACE_INT("HardwareComposer: PresentResult", error);
}
return error;
}
int32_t HardwareComposer::SetPowerMode(hwc2_display_t display,
hwc2_power_mode_t mode) {
if (mode == HWC2_POWER_MODE_OFF) {
EnableVsync(false);
}
display_on_ = mode != HWC2_POWER_MODE_OFF;
return (int32_t)hwc2_hidl_->setPowerMode(
display, (Hwc2::IComposerClient::PowerMode)mode);
}
int32_t HardwareComposer::GetDisplayAttribute(hwc2_display_t display,
hwc2_config_t config,
hwc2_attribute_t attribute,
int32_t* out_value) const {
return (int32_t)hwc2_hidl_->getDisplayAttribute(
display, config, (Hwc2::IComposerClient::Attribute)attribute, out_value);
}
int32_t HardwareComposer::GetDisplayMetrics(
hwc2_display_t display, hwc2_config_t config,
HWCDisplayMetrics* out_metrics) const {
int32_t ret = HWC2_ERROR_NONE;
ret = GetDisplayAttribute(display, config, HWC2_ATTRIBUTE_WIDTH,
&out_metrics->width);
if (ret != HWC2_ERROR_NONE) {
ALOGE("HardwareComposer: Failed to get display width");
return ret;
}
ret = GetDisplayAttribute(display, config, HWC2_ATTRIBUTE_HEIGHT,
&out_metrics->height);
if (ret != HWC2_ERROR_NONE) {
ALOGE("HardwareComposer: Failed to get display height");
return ret;
}
ret = GetDisplayAttribute(display, config, HWC2_ATTRIBUTE_VSYNC_PERIOD,
&out_metrics->vsync_period_ns);
if (ret != HWC2_ERROR_NONE) {
ALOGE("HardwareComposer: Failed to get display height");
return ret;
}
ret = GetDisplayAttribute(display, config, HWC2_ATTRIBUTE_DPI_X,
&out_metrics->dpi.x);
if (ret != HWC2_ERROR_NONE) {
ALOGE("HardwareComposer: Failed to get display DPI X");
return ret;
}
ret = GetDisplayAttribute(display, config, HWC2_ATTRIBUTE_DPI_Y,
&out_metrics->dpi.y);
if (ret != HWC2_ERROR_NONE) {
ALOGE("HardwareComposer: Failed to get display DPI Y");
return ret;
}
return HWC2_ERROR_NONE;
}
void HardwareComposer::Dump(char* buffer, uint32_t* out_size) {
std::string debug_str = hwc2_hidl_->dumpDebugInfo();
ALOGI("%s", debug_str.c_str());
if (buffer == nullptr) {
*out_size = debug_str.size();
} else {
std::copy(debug_str.begin(), debug_str.begin() + *out_size, buffer);
}
}
// TODO(skiazyk): Figure out what to do with `is_geometry_changed`. There does
// not seem to be any equivalent in the HWC2 API, but that doesn't mean its not
// there.
void HardwareComposer::PostLayers(bool /*is_geometry_changed*/) {
ATRACE_NAME("HardwareComposer::PostLayers");
// Setup the hardware composer layers with current buffers.
for (size_t i = 0; i < active_layer_count_; i++) {
layers_[i]->Prepare();
}
// Now that we have taken in a frame from the application, we have a chance
// to drop the frame before passing the frame along to HWC.
// If the display driver has become backed up, we detect it here and then
// react by skipping this frame to catch up latency.
while (!retire_fence_fds_.empty() &&
(!retire_fence_fds_.front() ||
sync_wait(retire_fence_fds_.front().Get(), 0) == 0)) {
// There are only 2 fences in here, no performance problem to shift the
// array of ints.
retire_fence_fds_.erase(retire_fence_fds_.begin());
}
const bool is_frame_pending = IsFramePendingInDriver();
const bool is_fence_pending =
retire_fence_fds_.size() > kAllowedPendingFenceCount;
if (is_fence_pending || is_frame_pending) {
ATRACE_INT("frame_skip_count", ++frame_skip_count_);
ALOGW_IF(is_frame_pending, "Warning: frame already queued, dropping frame");
ALOGW_IF(is_fence_pending,
"Warning: dropping a frame to catch up with HWC (pending = %zd)",
retire_fence_fds_.size());
for (size_t i = 0; i < active_layer_count_; i++) {
layers_[i]->Drop();
}
return;
} else {
// Make the transition more obvious in systrace when the frame skip happens
// above.
ATRACE_INT("frame_skip_count", 0);
}
#if TRACE
for (size_t i = 0; i < active_layer_count_; i++)
ALOGI("HardwareComposer::PostLayers: dl[%zu] ctype=0x%08x", i,
layers_[i]->GetCompositionType());
#endif
int32_t ret = HWC2_ERROR_NONE;
std::vector<Hwc2::IComposerClient::Rect> full_region(1);
full_region[0].left = 0;
full_region[0].top = 0;
full_region[0].right = framebuffer_target_->width();
full_region[0].bottom = framebuffer_target_->height();
ALOGE_IF(ret, "Error setting client target : %d", ret);
ret = Validate(HWC_DISPLAY_PRIMARY);
if (ret) {
ALOGE("HardwareComposer::Validate failed; ret=%d", ret);
return;
}
ret = Present(HWC_DISPLAY_PRIMARY);
if (ret) {
ALOGE("HardwareComposer::Present failed; ret=%d", ret);
return;
}
std::vector<Hwc2::Layer> out_layers;
std::vector<int> out_fences;
ret = (int32_t)hwc2_hidl_->getReleaseFences(HWC_DISPLAY_PRIMARY, &out_layers,
&out_fences);
uint32_t num_elements = out_layers.size();
ALOGE_IF(ret, "HardwareComposer: GetReleaseFences failed; ret=%d", ret);
// Perform post-frame bookkeeping. Unused layers are a no-op.
for (size_t i = 0; i < num_elements; ++i) {
for (size_t j = 0; j < active_layer_count_; ++j) {
if (layers_[j]->GetLayerHandle() == out_layers[i]) {
layers_[j]->Finish(out_fences[i]);
}
}
}
}
// TODO(skiazyk): This is a work-around for the fact that we currently do not
// handle the case when new surfaces are introduced when displayd is not
// in an active state. A proper-solution will require re-structuring
// displayd a little, but hopefully this is sufficient for now.
// For example, could this be handled in |UpdateLayerSettings| instead?
void HardwareComposer::UpdateDisplayState() {
const bool has_display_surfaces = display_surfaces_.size() > 0;
if (has_display_surfaces) {
int32_t ret = SetPowerMode(HWC_DISPLAY_PRIMARY, HWC2_POWER_MODE_ON);
ALOGE_IF(ret, "HardwareComposer: Could not set power mode; ret=%d", ret);
EnableVsync(true);
}
// TODO(skiazyk): We need to do something about accessing this directly,
// supposedly there is a backlight service on the way.
SetBacklightBrightness(255);
if (!display_on_ && has_display_surfaces) {
const int error = ReadVSyncTimestamp(&last_vsync_timestamp_);
ALOGE_IF(error < 0,
"HardwareComposer::SetDisplaySurfaces: Failed to read vsync "
"timestamp: %s",
strerror(-error));
}
// Trigger target-specific performance mode change.
property_set(kDvrPerformanceProperty, display_on_ ? "performance" : "idle");
}
int HardwareComposer::SetDisplaySurfaces(
std::vector<std::shared_ptr<DisplaySurface>> surfaces) {
std::lock_guard<std::mutex> autolock(layer_mutex_);
ALOGI("HardwareComposer::SetDisplaySurfaces: surface count=%zd",
surfaces.size());
// Figure out whether we need to update hardware layers. If this surface
// change does not add or remove hardware layers we can avoid display hiccups
// by gracefully updating only the GPU compositor layers.
// hardware_layers_need_update_ is reset to false by the Post thread.
int old_gpu_layer_count = 0;
int new_gpu_layer_count = 0;
// Look for new hardware layers and count new GPU layers.
for (const auto& surface : surfaces) {
if (!(surface->flags() &
DVR_DISPLAY_SURFACE_FLAGS_DISABLE_SYSTEM_DISTORTION))
++new_gpu_layer_count;
else if (std::find(display_surfaces_.begin(), display_surfaces_.end(),
surface) == display_surfaces_.end())
// This is a new hardware layer, we need to update.
hardware_layers_need_update_ = true;
}
// Look for deleted hardware layers or compositor layers.
for (const auto& surface : display_surfaces_) {
if (!(surface->flags() &
DVR_DISPLAY_SURFACE_FLAGS_DISABLE_SYSTEM_DISTORTION))
++old_gpu_layer_count;
else if (std::find(surfaces.begin(), surfaces.end(), surface) ==
surfaces.end())
// This is a deleted hardware layer, we need to update.
hardware_layers_need_update_ = true;
}
// Check for compositor hardware layer transition.
if ((!old_gpu_layer_count && new_gpu_layer_count) ||
(old_gpu_layer_count && !new_gpu_layer_count))
hardware_layers_need_update_ = true;
display_surfaces_ = std::move(surfaces);
display_surfaces_updated_ = true;
// Set the chosen layer order for all surfaces.
for (size_t i = 0; i < display_surfaces_.size(); ++i) {
display_surfaces_[i]->SetLayerOrder(static_cast<int>(i));
}
// TODO(skiazyk): fix this so that it is handled seamlessly with dormant/non-
// dormant state.
if (!IsSuspended()) {
UpdateDisplayState();
}
return 0;
}
// Reads the value of the display driver wait_pingpong state. Returns 0 or 1
// (the value of the state) on success or a negative error otherwise.
// TODO(eieio): This is pretty driver specific, this should be moved to a
// separate class eventually.
int HardwareComposer::ReadWaitPPState() {
// Gracefully handle when the kernel does not support this feature.
if (!primary_display_wait_pp_fd_)
return 0;
const int wait_pp_fd = primary_display_wait_pp_fd_.Get();
int ret, error;
ret = lseek(wait_pp_fd, 0, SEEK_SET);
if (ret < 0) {
error = errno;
ALOGE("HardwareComposer::ReadWaitPPState: Failed to seek wait_pp fd: %s",
strerror(error));
return -error;
}
char data = -1;
ret = read(wait_pp_fd, &data, sizeof(data));
if (ret < 0) {
error = errno;
ALOGE("HardwareComposer::ReadWaitPPState: Failed to read wait_pp state: %s",
strerror(error));
return -error;
}
switch (data) {
case '0':
return 0;
case '1':
return 1;
default:
ALOGE(
"HardwareComposer::ReadWaitPPState: Unexpected value for wait_pp: %d",
data);
return -EINVAL;
}
}
// Reads the timestamp of the last vsync from the display driver.
// TODO(eieio): This is pretty driver specific, this should be moved to a
// separate class eventually.
int HardwareComposer::ReadVSyncTimestamp(int64_t* timestamp) {
const int event_fd = primary_display_vsync_event_fd_.Get();
int ret, error;
// The driver returns data in the form "VSYNC=<timestamp ns>".
std::array<char, 32> data;
data.fill('\0');
// Seek back to the beginning of the event file.
ret = lseek(event_fd, 0, SEEK_SET);
if (ret < 0) {
error = errno;
ALOGE(
"HardwareComposer::ReadVSyncTimestamp: Failed to seek vsync event fd: "
"%s",
strerror(error));
return -error;
}
// Read the vsync event timestamp.
ret = read(event_fd, data.data(), data.size());
if (ret < 0) {
error = errno;
ALOGE_IF(
error != EAGAIN,
"HardwareComposer::ReadVSyncTimestamp: Error while reading timestamp: "
"%s",
strerror(error));
return -error;
}
ret = sscanf(data.data(), "VSYNC=%" PRIu64,
reinterpret_cast<uint64_t*>(timestamp));
if (ret < 0) {
error = errno;
ALOGE(
"HardwareComposer::ReadVSyncTimestamp: Error while parsing timestamp: "
"%s",
strerror(error));
return -error;
}
return 0;
}
// Blocks until the next vsync event is signaled by the display driver.
// TODO(eieio): This is pretty driver specific, this should be moved to a
// separate class eventually.
int HardwareComposer::BlockUntilVSync() {
const int event_fd = primary_display_vsync_event_fd_.Get();
pollfd pfd[2] = {
{
.fd = event_fd, .events = POLLPRI, .revents = 0,
},
// This extra event fd is to ensure that we can break out of this loop to
// pause the thread even when vsync is disabled, and thus no events on the
// vsync fd are being generated.
{
.fd = terminate_post_thread_event_fd_.Get(),
.events = POLLPRI | POLLIN,
.revents = 0,
},
};
int ret, error;
do {
ret = poll(pfd, 2, -1);
error = errno;
ALOGW_IF(ret < 0,
"HardwareComposer::BlockUntilVSync: Error while waiting for vsync "
"event: %s (%d)",
strerror(error), error);
} while (ret < 0 && error == EINTR);
return ret < 0 ? -error : 0;
}
// Waits for the next vsync and returns the timestamp of the vsync event. If
// vsync already passed since the last call, returns the latest vsync timestamp
// instead of blocking. This method updates the last_vsync_timeout_ in the
// process.
//
// TODO(eieio): This is pretty driver specific, this should be moved to a
// separate class eventually.
int HardwareComposer::WaitForVSync(int64_t* timestamp) {
int error;
// Get the current timestamp and decide what to do.
while (true) {
int64_t current_vsync_timestamp;
error = ReadVSyncTimestamp(&current_vsync_timestamp);
if (error < 0 && error != -EAGAIN)
return error;
if (error == -EAGAIN) {
// Vsync was turned off, wait for the next vsync event.
error = BlockUntilVSync();
if (error < 0)
return error;
// If a request to pause the post thread was given, exit immediately
if (IsSuspended()) {
return 0;
}
// Try again to get the timestamp for this new vsync interval.
continue;
}
// Check that we advanced to a later vsync interval.
if (TimestampGT(current_vsync_timestamp, last_vsync_timestamp_)) {
*timestamp = last_vsync_timestamp_ = current_vsync_timestamp;
return 0;
}
// See how close we are to the next expected vsync. If we're within 1ms,
// sleep for 1ms and try again.
const int64_t ns_per_frame = display_metrics_.vsync_period_ns;
const int64_t threshold_ns = 1000000;
const int64_t next_vsync_est = last_vsync_timestamp_ + ns_per_frame;
const int64_t distance_to_vsync_est = next_vsync_est - GetSystemClockNs();
if (distance_to_vsync_est > threshold_ns) {
// Wait for vsync event notification.
error = BlockUntilVSync();
if (error < 0)
return error;
// Again, exit immediately if the thread was requested to pause
if (IsSuspended()) {
return 0;
}
} else {
// Sleep for a short time before retrying.
std::this_thread::sleep_for(std::chrono::milliseconds(1));
}
}
}
int HardwareComposer::SleepUntil(int64_t wakeup_timestamp) {
const int timer_fd = vsync_sleep_timer_fd_.Get();
const itimerspec wakeup_itimerspec = {
.it_interval = {.tv_sec = 0, .tv_nsec = 0},
.it_value = NsToTimespec(wakeup_timestamp),
};
int ret =
timerfd_settime(timer_fd, TFD_TIMER_ABSTIME, &wakeup_itimerspec, nullptr);
int error = errno;
if (ret < 0) {
ALOGE("HardwareComposer::SleepUntil: Failed to set timerfd: %s",
strerror(error));
return -error;
}
// Wait for the timer by reading the expiration count.
uint64_t expiration_count;
ret = read(timer_fd, &expiration_count, sizeof(expiration_count));
if (ret < 0) {
ALOGE("HardwareComposer::SleepUntil: Failed to wait for timerfd: %s",
strerror(error));
return -error;
}
return 0;
}
void HardwareComposer::PostThread() {
// NOLINTNEXTLINE(runtime/int)
prctl(PR_SET_NAME, reinterpret_cast<unsigned long>("PostThread"), 0, 0, 0);
std::unique_lock<std::mutex> thread_lock(thread_pause_mutex_);
// Set the scheduler to SCHED_FIFO with high priority.
int error = dvrSetSchedulerClass(0, "graphics:high");
LOG_ALWAYS_FATAL_IF(
error < 0,
"HardwareComposer::PostThread: Failed to set scheduler class: %s",
strerror(-error));
error = dvrSetCpuPartition(0, "/system/performance");
LOG_ALWAYS_FATAL_IF(
error < 0,
"HardwareComposer::PostThread: Failed to set cpu partition: %s",
strerror(-error));
// Force the layers to be setup at least once.
display_surfaces_updated_ = true;
// Initialize the GPU compositor.
LOG_ALWAYS_FATAL_IF(!compositor_.Initialize(GetHmdDisplayMetrics()),
"Failed to initialize the compositor");
const int64_t ns_per_frame = display_metrics_.vsync_period_ns;
const int64_t photon_offset_ns = GetPosePredictionTimeOffset(ns_per_frame);
// TODO(jbates) Query vblank time from device, when such an API is available.
// This value (6.3%) was measured on A00 in low persistence mode.
int64_t vblank_ns = ns_per_frame * 63 / 1000;
int64_t right_eye_photon_offset_ns = (ns_per_frame - vblank_ns) / 2;
// Check property for overriding right eye offset value.
right_eye_photon_offset_ns =
property_get_int64(kRightEyeOffsetProperty, right_eye_photon_offset_ns);
// The list of surfaces the compositor should attempt to render. This is set
// at the start of each frame.
std::vector<std::shared_ptr<DisplaySurface>> compositor_surfaces;
compositor_surfaces.reserve(2);
// Our history of frame times. This is used to get a better estimate of how
// long the next frame will take, to set a schedule for EDS.
FrameTimeHistory frame_time_history;
// The backlog is used to allow us to start rendering the next frame before
// the previous frame has finished, and still get an accurate measurement of
// frame duration.
std::vector<FrameTimeMeasurementRecord> frame_time_backlog;
constexpr int kFrameTimeBacklogMax = 2;
frame_time_backlog.reserve(kFrameTimeBacklogMax);
// Storage for retrieving fence info.
FenceInfoBuffer fence_info_buffer;
while (1) {
ATRACE_NAME("HardwareComposer::PostThread");
while (IsSuspended()) {
ALOGI("HardwareComposer::PostThread: Post thread pause requested.");
thread_pause_semaphore_.wait(thread_lock);
// The layers will need to be updated since they were deleted previously
display_surfaces_updated_ = true;
hardware_layers_need_update_ = true;
}
int64_t vsync_timestamp = 0;
{
std::array<char, 128> buf;
snprintf(buf.data(), buf.size(), "wait_vsync|vsync=%d|",
vsync_count_ + 1);
ATRACE_NAME(buf.data());
error = WaitForVSync(&vsync_timestamp);
ALOGE_IF(
error < 0,
"HardwareComposer::PostThread: Failed to wait for vsync event: %s",
strerror(-error));
// Don't bother processing this frame if a pause was requested
if (IsSuspended()) {
continue;
}
}
++vsync_count_;
static double last_print_time = -1;
double current_time = GetSystemClockSec();
if (last_print_time < 0 || current_time - last_print_time > 3) {
last_print_time = current_time;
}
if (pose_client_) {
// Signal the pose service with vsync info.
// Display timestamp is in the middle of scanout.
privateDvrPoseNotifyVsync(pose_client_, vsync_count_,
vsync_timestamp + photon_offset_ns,
ns_per_frame, right_eye_photon_offset_ns);
}
bool layer_config_changed = UpdateLayerConfig(&compositor_surfaces);
if (layer_config_changed) {
frame_time_history.ResetWithSeed(
GuessFrameTime(compositor_surfaces.size()));
frame_time_backlog.clear();
} else {
UpdateFrameTimeHistory(&frame_time_backlog, kFrameTimeBacklogMax,
&fence_info_buffer, &frame_time_history);
}
// Get our current best estimate at how long the next frame will take to
// render, based on how long previous frames took to render. Use this
// estimate to decide when to wake up for EDS.
int64_t frame_time_estimate =
frame_time_history.GetSampleCount() == 0
? GuessFrameTime(compositor_surfaces.size())
: frame_time_history.GetAverage();
frame_time_estimate = std::max(frame_time_estimate, kFrameTimeEstimateMin);
DebugHudData::data.hwc_latency = frame_time_estimate;
// Signal all of the vsync clients. Because absolute time is used for the
// wakeup time below, this can take a little time if necessary.
if (vsync_callback_)
vsync_callback_(HWC_DISPLAY_PRIMARY, vsync_timestamp, frame_time_estimate,
vsync_count_);
{
// Sleep until async EDS wakeup time.
ATRACE_NAME("sleep");
int64_t display_time_est = vsync_timestamp + ns_per_frame;
int64_t now = GetSystemClockNs();
int64_t frame_finish_time_est = now + frame_time_estimate;
int64_t sleep_time_ns = display_time_est - now - frame_time_estimate;
ATRACE_INT64("sleep_time_ns", sleep_time_ns);
if (frame_finish_time_est - display_time_est >= kFrameSkipThresholdNs) {
ATRACE_INT("frame_skip_count", ++frame_skip_count_);
ALOGE(
"HardwareComposer::PostThread: Missed frame schedule, drop "
"frame. Expected frame miss: %.1fms",
static_cast<double>(frame_finish_time_est - display_time_est) /
1000000);
// There are several reasons we might skip a frame, but one possibility
// is we mispredicted the frame time. Clear out the frame time history.
frame_time_history.ResetWithSeed(
GuessFrameTime(compositor_surfaces.size()));
frame_time_backlog.clear();
DebugHudData::data.hwc_frame_stats.SkipFrame();
continue;
} else {
// Make the transition more obvious in systrace when the frame skip
// happens above.
ATRACE_INT("frame_skip_count", 0);
}
if (sleep_time_ns > 0) {
error = SleepUntil(display_time_est - frame_time_estimate);
ALOGE_IF(error < 0, "HardwareComposer::PostThread: Failed to sleep: %s",
strerror(-error));
}
}
DebugHudData::data.hwc_frame_stats.AddFrame();
int64_t frame_start_time = GetSystemClockNs();
// Setup the output buffer for the compositor. This needs to happen before
// you draw with the compositor.
if (gpu_layer_ != nullptr) {
gpu_layer_->UpdateDirectBuffer(compositor_.GetBuffer());
}
// Call PostLayers now before performing the GL code for the compositor to
// avoid missing the deadline that can cause the lower-level hwc to get
// permanently backed up.
PostLayers(layer_config_changed);
PostCompositorBuffers(compositor_surfaces);
if (gpu_layer_ != nullptr) {
// Note, with scanline racing, this draw is timed along with the post
// layers to finish just in time.
LocalHandle frame_fence_fd;
compositor_.DrawFrame(vsync_count_ + 1, &frame_fence_fd);
if (frame_fence_fd) {
LOG_ALWAYS_FATAL_IF(frame_time_backlog.size() >= kFrameTimeBacklogMax,
"Frame time backlog exceeds capacity");
frame_time_backlog.push_back(
{frame_start_time, std::move(frame_fence_fd)});
}
} else if (!layer_config_changed) {
frame_time_history.AddSample(GetSystemClockNs() - frame_start_time);
}
HandlePendingScreenshots();
}
// TODO(skiazyk): Currently the compositor is not fully releasing its EGL
// context, which seems to prevent the thread from exiting properly.
// This shouldn't be too hard to address, I just don't have time right now.
compositor_.Shutdown();
}
bool HardwareComposer::UpdateLayerConfig(
std::vector<std::shared_ptr<DisplaySurface>>* compositor_surfaces) {
std::lock_guard<std::mutex> autolock(layer_mutex_);
if (!display_surfaces_updated_)
return false;
display_surfaces_updated_ = false;
DebugHudData::data.ResetLayers();
// Update compositor layers.
{
ATRACE_NAME("UpdateLayerConfig_GpuLayers");
compositor_.UpdateSurfaces(display_surfaces_);
compositor_surfaces->clear();
for (size_t i = 0; i < display_surfaces_.size(); ++i) {
const auto& surface = display_surfaces_[i];
if (!(surface->flags() &
DVR_DISPLAY_SURFACE_FLAGS_DISABLE_SYSTEM_DISTORTION)) {
compositor_surfaces->push_back(surface);
}
}
}
if (!hardware_layers_need_update_)
return true;
// Update hardware layers.
ATRACE_NAME("UpdateLayerConfig_HwLayers");
hardware_layers_need_update_ = false;
// Update the display layers in a non-destructive fashion.
// Create a map from surface id to hardware layer
std::map<int, Layer*> display_surface_layers;
for (size_t i = 0; i < active_layer_count_; ++i) {
auto layer = layers_[i];
int surface_id = layer->GetSurfaceId();
auto found =
std::find_if(display_surfaces_.begin(), display_surfaces_.end(),
[surface_id](const auto& surface) {
return surface->surface_id() == surface_id;
});
if (found != display_surfaces_.end()) {
display_surface_layers[surface_id] = layer;
}
}
bool has_gpu_layer = std::any_of(
display_surfaces_.begin(), display_surfaces_.end(),
[](const auto& surface) {
return !(surface->flags() &
DVR_DISPLAY_SURFACE_FLAGS_DISABLE_SYSTEM_DISTORTION);
});
if (!has_gpu_layer) {
gpu_layer_ = nullptr;
}
auto is_layer_active = [&display_surface_layers, has_gpu_layer](auto layer) {
int surface_id = layer->GetSurfaceId();
if (surface_id >= 0) {
return display_surface_layers.count(surface_id) > 0;
} else {
return has_gpu_layer;
}
};
// Compress the in-use layers to the top of the list
auto part = std::partition(
layers_.begin(), layers_.begin() + active_layer_count_, is_layer_active);
size_t new_active_layer_count = part - layers_.begin();
// Clear any unused layers
for (size_t i = new_active_layer_count; i < active_layer_count_; ++i) {
layers_[i]->Reset();
}
active_layer_count_ = new_active_layer_count;
bool gpu_layer_applied = false;
// Create/update all of the hardware layers
for (size_t i = 0; i < display_surfaces_.size(); ++i) {
const auto& surface = display_surfaces_[i];
bool is_hw_surface =
surface->flags() & DVR_DISPLAY_SURFACE_FLAGS_DISABLE_SYSTEM_DISTORTION;
hwc2_blend_mode_t blending =
i == 0 ? HWC2_BLEND_MODE_NONE : HWC2_BLEND_MODE_COVERAGE;
DebugHudData::data.SetLayerInfo(
i, surface->width(), surface->height(),
!!(surface->flags() & DVR_DISPLAY_SURFACE_FLAGS_GEOMETRY_SEPARATE_2));
if (!is_hw_surface && gpu_layer_applied) {
continue;
}
Layer* target_layer;
bool existing_layer = false;
if (is_hw_surface) {
auto it = display_surface_layers.find(surface->surface_id());
if (it != display_surface_layers.end()) {
target_layer = it->second;
existing_layer = true;
}
} else if (gpu_layer_ != nullptr) {
target_layer = gpu_layer_;
existing_layer = true;
}
if (!existing_layer) {
if (active_layer_count_ >= kMaxHardwareLayers) {
ALOGI("HardwareComposer: More than %d hardware layers requested.",
kMaxHardwareLayers);
break;
} else {
target_layer = layers_[active_layer_count_];
++active_layer_count_;
}
ALOGD_IF(TRACE,
"HardwareComposer::UpdateLayerConfig: (new) surface_id=%d -> "
"layer=%zd",
surface->surface_id(), i);
if (is_hw_surface) {
target_layer->Setup(surface, blending, display_transform_,
HWC2_COMPOSITION_DEVICE, i);
} else {
gpu_layer_ = target_layer;
target_layer->Setup(compositor_.GetBuffer(), blending,
display_transform_, HWC2_COMPOSITION_DEVICE, i);
}
} else {
ALOGD_IF(TRACE,
"HardwareComposer::UpdateLayerConfig: (retained) surface_id=%d "
"-> layer=%zd",
surface->surface_id(), i);
target_layer->SetBlending(blending);
target_layer->SetZOrderIndex(i);
target_layer->UpdateLayerSettings();
}
gpu_layer_applied = !is_hw_surface;
}
ALOGD_IF(TRACE, "HardwareComposer::UpdateLayerConfig: %zd active layers",
active_layer_count_);
return true;
}
void HardwareComposer::PostCompositorBuffers(
const std::vector<std::shared_ptr<DisplaySurface>>& compositor_surfaces) {
ATRACE_NAME("PostCompositorBuffers");
for (const auto& surface : compositor_surfaces) {
compositor_.PostBuffer(surface);
}
}
void HardwareComposer::UpdateFrameTimeHistory(
std::vector<FrameTimeMeasurementRecord>* backlog, int backlog_max,
FenceInfoBuffer* fence_info_buffer, FrameTimeHistory* history) {
while (!backlog->empty()) {
const auto& frame_time_record = backlog->front();
int64_t end_time = 0;
bool frame_finished = CheckFrameFinished(frame_time_record.fence.Get(),
fence_info_buffer, &end_time);
if (frame_finished) {
int64_t frame_duration = end_time - frame_time_record.start_time;
history->AddSample(frame_duration);
// Our backlog is tiny (2 elements), so erasing from the front is ok
backlog->erase(backlog->begin());
} else {
break;
}
}
if (backlog->size() == static_cast<size_t>(backlog_max)) {
// Yikes, something must've gone wrong if our oldest frame hasn't finished
// yet. Give up on waiting for it.
const auto& stale_frame_time_record = backlog->front();
int64_t frame_duration =
GetSystemClockNs() - stale_frame_time_record.start_time;
backlog->erase(backlog->begin());
history->AddSample(frame_duration);
ALOGW("Frame didn't finish after %.1fms",
static_cast<double>(frame_duration) / 1000000);
}
}
bool HardwareComposer::CheckFrameFinished(int frame_fence_fd,
FenceInfoBuffer* fence_info_buffer,
int64_t* timestamp) {
int result = -1;
int sync_result = sync_wait(frame_fence_fd, 0);
if (sync_result == 0) {
result =
GetFenceSignaledTimestamp(frame_fence_fd, fence_info_buffer, timestamp);
if (result < 0) {
ALOGE("Failed getting signaled timestamp from fence");
}
} else if (errno != ETIME) {
ALOGE("sync_wait on frame fence failed");
}
return result >= 0;
}
void HardwareComposer::HandlePendingScreenshots() {
// Take a screenshot of the requested layer, if available.
// TODO(eieio): Look into using virtual displays to composite the layer stack
// into a single output buffer that can be returned to the screenshot clients.
if (active_layer_count_ > 0) {
if (auto screenshot_service = ScreenshotService::GetInstance()) {
if (screenshot_service->IsScreenshotRequestPending()) {
ATRACE_NAME("screenshot");
screenshot_service->TakeIfNeeded(layers_, compositor_);
}
} else {
ALOGW(
"HardwareComposer::HandlePendingScreenshots: Failed to get "
"screenshot service!");
}
}
}
void HardwareComposer::SetVSyncCallback(VSyncCallback callback) {
vsync_callback_ = callback;
}
void HardwareComposer::HwcRefresh(hwc2_callback_data_t /*data*/,
hwc2_display_t /*display*/) {
// TODO(eieio): implement invalidate callbacks.
}
void HardwareComposer::HwcVSync(hwc2_callback_data_t /*data*/,
hwc2_display_t /*display*/,
int64_t /*timestamp*/) {
ATRACE_NAME(__PRETTY_FUNCTION__);
// Intentionally empty. HWC may require a callback to be set to enable vsync
// signals. We bypass this callback thread by monitoring the vsync event
// directly, but signals still need to be enabled.
}
void HardwareComposer::HwcHotplug(hwc2_callback_data_t /*callbackData*/,
hwc2_display_t /*display*/,
hwc2_connection_t /*connected*/) {
// TODO(eieio): implement display hotplug callbacks.
}
void HardwareComposer::SetBacklightBrightness(int brightness) {
if (backlight_brightness_fd_) {
std::array<char, 32> text;
const int length = snprintf(text.data(), text.size(), "%d", brightness);
write(backlight_brightness_fd_.Get(), text.data(), length);
}
}
Layer::Layer()
: hwc2_hidl_(nullptr),
surface_index_(-1),
hardware_composer_layer_(0),
display_metrics_(nullptr),
blending_(HWC2_BLEND_MODE_NONE),
transform_(HWC_TRANSFORM_NONE),
composition_type_(HWC2_COMPOSITION_DEVICE),
surface_rect_functions_applied_(false) {}
void Layer::Initialize(Hwc2::Composer* hwc2_hidl, HWCDisplayMetrics* metrics) {
hwc2_hidl_ = hwc2_hidl;
display_metrics_ = metrics;
}
void Layer::Reset() {
const int ret = acquired_buffer_.Release(std::move(release_fence_));
ALOGE_IF(ret < 0, "Layer::Reset: failed to release buffer: %s",
strerror(-ret));
if (hwc2_hidl_ != nullptr && hardware_composer_layer_) {
hwc2_hidl_->destroyLayer(HWC_DISPLAY_PRIMARY, hardware_composer_layer_);
hardware_composer_layer_ = 0;
}
surface_index_ = static_cast<size_t>(-1);
blending_ = HWC2_BLEND_MODE_NONE;
transform_ = HWC_TRANSFORM_NONE;
composition_type_ = HWC2_COMPOSITION_DEVICE;
direct_buffer_ = nullptr;
surface_ = nullptr;
acquire_fence_fd_.Close();
surface_rect_functions_applied_ = false;
}
void Layer::Setup(const std::shared_ptr<DisplaySurface>& surface,
hwc2_blend_mode_t blending, hwc_transform_t transform,
hwc2_composition_t composition_type, size_t index) {
Reset();
surface_index_ = index;
surface_ = surface;
blending_ = blending;
transform_ = transform;
composition_type_ = composition_type;
CommonLayerSetup();
}
void Layer::Setup(const std::shared_ptr<IonBuffer>& buffer,
hwc2_blend_mode_t blending, hwc_transform_t transform,
hwc2_composition_t composition_type, size_t z_order) {
Reset();
surface_index_ = z_order;
direct_buffer_ = buffer;
blending_ = blending;
transform_ = transform;
composition_type_ = composition_type;
CommonLayerSetup();
}
void Layer::UpdateDirectBuffer(const std::shared_ptr<IonBuffer>& buffer) {
direct_buffer_ = buffer;
}
void Layer::SetBlending(hwc2_blend_mode_t blending) { blending_ = blending; }
void Layer::SetZOrderIndex(int z_index) { surface_index_ = z_index; }
IonBuffer* Layer::GetBuffer() {
if (direct_buffer_)
return direct_buffer_.get();
else if (acquired_buffer_.IsAvailable())
return acquired_buffer_.buffer()->buffer();
else
return nullptr;
}
void Layer::UpdateLayerSettings() {
if (!IsLayerSetup()) {
ALOGE("HardwareComposer: Trying to update layers data on an unused layer.");
return;
}
int32_t ret = HWC2_ERROR_NONE;
hwc2_display_t display = HWC_DISPLAY_PRIMARY;
ret = (int32_t)hwc2_hidl_->setLayerCompositionType(
display, hardware_composer_layer_,
(Hwc2::IComposerClient::Composition)composition_type_);
ALOGE_IF(ret, "HardwareComposer: Error setting layer composition type : %d",
ret);
// ret = (int32_t) hwc2_hidl_->setLayerTransform(display,
// hardware_composer_layer_,
// (Hwc2::IComposerClient::Transform)
// transform_);
// ALOGE_IF(ret, "HardwareComposer: Error setting layer transform : %d", ret);
// ret = hwc2_funcs_->set_layer_blend_mode_fn_(
// hardware_composer_device_, display, hardware_composer_layer_,
// blending_);
ret = (int32_t)hwc2_hidl_->setLayerBlendMode(
display, hardware_composer_layer_,
(Hwc2::IComposerClient::BlendMode)blending_);
ALOGE_IF(ret, "HardwareComposer: Error setting layer blend mode : %d", ret);
Hwc2::IComposerClient::Rect display_frame;
display_frame.left = 0;
display_frame.top = 0;
display_frame.right = display_metrics_->width;
display_frame.bottom = display_metrics_->height;
ret = (int32_t)hwc2_hidl_->setLayerDisplayFrame(
display, hardware_composer_layer_, display_frame);
ALOGE_IF(ret, "HardwareComposer: Error setting layer display frame : %d",
ret);
std::vector<Hwc2::IComposerClient::Rect> visible_region(1);
visible_region[0] = display_frame;
ret = (int32_t)hwc2_hidl_->setLayerVisibleRegion(
display, hardware_composer_layer_, visible_region);
ALOGE_IF(ret, "HardwareComposer: Error setting layer visible region : %d",
ret);
ret = (int32_t)hwc2_hidl_->setLayerPlaneAlpha(display,
hardware_composer_layer_, 1.0f);
ALOGE_IF(ret, "HardwareComposer: Error setting layer plane alpha : %d", ret);
ret = (int32_t)hwc2_hidl_->setLayerZOrder(display, hardware_composer_layer_,
surface_index_);
ALOGE_IF(ret, "HardwareComposer: Error, setting z order index : %d", ret);
}
void Layer::CommonLayerSetup() {
int32_t ret = (int32_t)hwc2_hidl_->createLayer(HWC_DISPLAY_PRIMARY,
&hardware_composer_layer_);
ALOGE_IF(ret,
"HardwareComposer: Failed to create layer on primary display : %d",
ret);
UpdateLayerSettings();
}
void Layer::Prepare() {
int right, bottom;
buffer_handle_t handle;
if (surface_) {
// Only update the acquired buffer when one is either available or this is
// the first time through.
if (surface_->IsBufferAvailable()) {
// If we previously set this to a solid color layer to stall for time,
// revert it to a device layer.
if (acquired_buffer_.IsEmpty() &&
composition_type_ != HWC2_COMPOSITION_DEVICE) {
composition_type_ = HWC2_COMPOSITION_DEVICE;
hwc2_hidl_->setLayerCompositionType(
HWC_DISPLAY_PRIMARY, hardware_composer_layer_,
(Hwc2::IComposerClient::Composition)HWC2_COMPOSITION_DEVICE);
}
DebugHudData::data.AddLayerFrame(surface_index_);
acquired_buffer_.Release(std::move(release_fence_));
acquired_buffer_ = surface_->AcquireCurrentBuffer();
// Basic latency stopgap for when the application misses a frame:
// If the application recovers on the 2nd or 3rd (etc) frame after
// missing, this code will skip a frame to catch up by checking if
// the next frame is also available.
if (surface_->IsBufferAvailable()) {
DebugHudData::data.SkipLayerFrame(surface_index_);
ATRACE_NAME("DropToCatchUp");
ATRACE_ASYNC_END("BufferPost", acquired_buffer_.buffer()->id());
acquired_buffer_ = surface_->AcquireCurrentBuffer();
}
ATRACE_ASYNC_END("BufferPost", acquired_buffer_.buffer()->id());
} else if (acquired_buffer_.IsEmpty()) {
// While we are waiting for a buffer, set this to be an empty layer
if (composition_type_ != HWC2_COMPOSITION_SOLID_COLOR) {
composition_type_ = HWC2_COMPOSITION_SOLID_COLOR;
hwc2_hidl_->setLayerCompositionType(
HWC_DISPLAY_PRIMARY, hardware_composer_layer_,
(Hwc2::IComposerClient::Composition)HWC2_COMPOSITION_SOLID_COLOR);
Hwc2::IComposerClient::Color layer_color = {
0, 0, 0, 0,
};
hwc2_hidl_->setLayerColor(HWC_DISPLAY_PRIMARY, hardware_composer_layer_,
layer_color);
}
return;
}
right = acquired_buffer_.buffer()->width();
bottom = acquired_buffer_.buffer()->height();
handle = acquired_buffer_.buffer()->native_handle();
acquire_fence_fd_.Reset(acquired_buffer_.ClaimAcquireFence().Release());
} else {
right = direct_buffer_->width();
bottom = direct_buffer_->height();
handle = direct_buffer_->handle();
acquire_fence_fd_.Close();
}
int32_t ret = HWC2_ERROR_NONE;
if (composition_type_ == HWC2_COMPOSITION_DEVICE) {
ret = (int32_t)hwc2_hidl_->setLayerBuffer(HWC_DISPLAY_PRIMARY,
hardware_composer_layer_, handle,
acquire_fence_fd_.Get());
ALOGE_IF(ret, "HardwareComposer: Error setting layer buffer : %d", ret);
}
if (!surface_rect_functions_applied_) {
Hwc2::IComposerClient::FRect crop_rect = {
0, 0, static_cast<float>(right), static_cast<float>(bottom),
};
hwc2_hidl_->setLayerSourceCrop(HWC_DISPLAY_PRIMARY,
hardware_composer_layer_, crop_rect);
ALOGE_IF(ret, "HardwareComposer: Error setting layer source crop : %d",
ret);
// TODO(skiazyk): why is this ifdef'd out. Is if a driver-specific issue where
// it must/cannot be called?
#ifdef QCOM_BSP
hwc_rect_t damage_rect = {
0, 0, right, bottom,
};
hwc_region_t damage = {
1, &damage_rect,
};
// ret = hwc2_funcs_->set_layer_surface_damage(
// hardware_composer_device_, HWC_DISPLAY_PRIMARY,
// hardware_composer_layer_, damage);
// uses a std::vector as the listing
// hwc2_hidl_->setLayerSurfaceDamage(HWC_DISPLAY_PRIMARY,
// hardware_composer_layer_, vector here);
ALOGE_IF(ret, "HardwareComposer: Error settings layer surface damage : %d",
ret);
#endif
surface_rect_functions_applied_ = true;
}
}
void Layer::Finish(int release_fence_fd) {
release_fence_.Reset(release_fence_fd);
}
void Layer::Drop() { acquire_fence_fd_.Close(); }
} // namespace dvr
} // namespace android