services/surfaceflinger/SurfaceFlinger.h

/*
 * Copyright (C) 2007 The Android Open Source Project
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *      http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#pragma once

#include <sys/types.h>

/*
 * NOTE: Make sure this file doesn't include  anything from <gl/ > or <gl2/ >
 */

#include <android-base/thread_annotations.h>
#include <compositionengine/OutputColorSetting.h>
#include <cutils/atomic.h>
#include <cutils/compiler.h>
#include <gui/BufferQueue.h>
#include <gui/FrameTimestamps.h>
#include <gui/ISurfaceComposer.h>
#include <gui/ISurfaceComposerClient.h>
#include <gui/ITransactionCompletedListener.h>
#include <gui/LayerState.h>
#include <layerproto/LayerProtoHeader.h>
#include <math/mat4.h>
#include <renderengine/LayerSettings.h>
#include <serviceutils/PriorityDumper.h>
#include <system/graphics.h>
#include <ui/FenceTime.h>
#include <ui/PixelFormat.h>
#include <ui/Size.h>
#include <utils/Errors.h>
#include <utils/KeyedVector.h>
#include <utils/RefBase.h>
#include <utils/SortedVector.h>
#include <utils/Trace.h>
#include <utils/threads.h>

#include "ClientCache.h"
#include "DisplayDevice.h"
#include "DisplayHardware/HWC2.h"
#include "DisplayHardware/PowerAdvisor.h"
#include "DisplayIdGenerator.h"
#include "Effects/Daltonizer.h"
#include "Fps.h"
#include "FrameTracker.h"
#include "LayerVector.h"
#include "Scheduler/MessageQueue.h"
#include "Scheduler/RefreshRateConfigs.h"
#include "Scheduler/RefreshRateStats.h"
#include "Scheduler/Scheduler.h"
#include "Scheduler/VsyncModulator.h"
#include "SurfaceFlingerFactory.h"
#include "SurfaceTracing.h"
#include "TracedOrdinal.h"
#include "TransactionCallbackInvoker.h"
#include "TransactionState.h"

#include <atomic>
#include <cstdint>
#include <functional>
#include <future>
#include <map>
#include <memory>
#include <mutex>
#include <optional>
#include <queue>
#include <set>
#include <string>
#include <thread>
#include <type_traits>
#include <unordered_map>
#include <unordered_set>
#include <utility>

using namespace android::surfaceflinger;

namespace android {

class Client;
class EventThread;
class FlagManager;
class FpsReporter;
class TunnelModeEnabledReporter;
class HdrLayerInfoReporter;
class HWComposer;
class IGraphicBufferProducer;
class Layer;
class MessageBase;
class RefreshRateOverlay;
class RegionSamplingThread;
class RenderArea;
class TimeStats;
class FrameTracer;
class WindowInfosListenerInvoker;

using gui::ScreenCaptureResults;

namespace frametimeline {
class FrameTimeline;
}

namespace os {
    class IInputFlinger;
}

namespace compositionengine {
class DisplaySurface;
class OutputLayer;

struct CompositionRefreshArgs;
} // namespace compositionengine

namespace renderengine {
class RenderEngine;
} // namespace renderengine

enum {
    eTransactionNeeded = 0x01,
    eTraversalNeeded = 0x02,
    eDisplayTransactionNeeded = 0x04,
    eTransformHintUpdateNeeded = 0x08,
    eTransactionFlushNeeded = 0x10,
    eTransactionMask = 0x1f,
};

using DisplayColorSetting = compositionengine::OutputColorSetting;

struct SurfaceFlingerBE {
    FenceTimeline mGlCompositionDoneTimeline;
    FenceTimeline mDisplayTimeline;

    // protected by mCompositorTimingLock;
    mutable std::mutex mCompositorTimingLock;
    CompositorTiming mCompositorTiming;

    // Only accessed from the main thread.
    struct CompositePresentTime {
        nsecs_t composite = -1;
        std::shared_ptr<FenceTime> display = FenceTime::NO_FENCE;
    };
    std::queue<CompositePresentTime> mCompositePresentTimes;

    static const size_t NUM_BUCKETS = 8; // < 1-7, 7+
    nsecs_t mFrameBuckets[NUM_BUCKETS] = {};
    nsecs_t mTotalTime = 0;
    std::atomic<nsecs_t> mLastSwapTime = 0;
};

class SurfaceFlinger : public BnSurfaceComposer,
                       public PriorityDumper,
                       private IBinder::DeathRecipient,
                       private HWC2::ComposerCallback,
                       private ICompositor,
                       private ISchedulerCallback {
public:
    struct SkipInitializationTag {};

    SurfaceFlinger(surfaceflinger::Factory&, SkipInitializationTag) ANDROID_API;
    explicit SurfaceFlinger(surfaceflinger::Factory&) ANDROID_API;

    // set main thread scheduling policy
    static status_t setSchedFifo(bool enabled) ANDROID_API;

    // set main thread scheduling attributes
    static status_t setSchedAttr(bool enabled);

    static char const* getServiceName() ANDROID_API { return "SurfaceFlinger"; }

    // This is the phase offset in nanoseconds of the software vsync event
    // relative to the vsync event reported by HWComposer.  The software vsync
    // event is when SurfaceFlinger and Choreographer-based applications run each
    // frame.
    //
    // This phase offset allows adjustment of the minimum latency from application
    // wake-up time (by Choreographer) to the time at which the resulting window
    // image is displayed.  This value may be either positive (after the HW vsync)
    // or negative (before the HW vsync). Setting it to 0 will result in a lower
    // latency bound of two vsync periods because the app and SurfaceFlinger
    // will run just after the HW vsync.  Setting it to a positive number will
    // result in the minimum latency being:
    //
    //     (2 * VSYNC_PERIOD - (vsyncPhaseOffsetNs % VSYNC_PERIOD))
    //
    // Note that reducing this latency makes it more likely for the applications
    // to not have their window content image ready in time.  When this happens
    // the latency will end up being an additional vsync period, and animations
    // will hiccup.  Therefore, this latency should be tuned somewhat
    // conservatively (or at least with awareness of the trade-off being made).
    static int64_t vsyncPhaseOffsetNs;
    static int64_t sfVsyncPhaseOffsetNs;

    // If fences from sync Framework are supported.
    static bool hasSyncFramework;

    // The offset in nanoseconds to use when VsyncController timestamps present fence
    // signaling time.
    static int64_t dispSyncPresentTimeOffset;

    // Some hardware can do RGB->YUV conversion more efficiently in hardware
    // controlled by HWC than in hardware controlled by the video encoder.
    // This instruct VirtualDisplaySurface to use HWC for such conversion on
    // GL composition.
    static bool useHwcForRgbToYuv;

    // Controls the number of buffers SurfaceFlinger will allocate for use in
    // FramebufferSurface
    static int64_t maxFrameBufferAcquiredBuffers;

    // Controls the maximum width and height in pixels that the graphics pipeline can support for
    // GPU fallback composition. For example, 8k devices with 4k GPUs, or 4k devices with 2k GPUs.
    static uint32_t maxGraphicsWidth;
    static uint32_t maxGraphicsHeight;

    // Indicate if a device has wide color gamut display. This is typically
    // found on devices with wide color gamut (e.g. Display-P3) display.
    static bool hasWideColorDisplay;

    static ui::Rotation internalDisplayOrientation;

    // Indicate if device wants color management on its display.
    static const constexpr bool useColorManagement = true;

    static bool useContextPriority;

    // The data space and pixel format that SurfaceFlinger expects hardware composer
    // to composite efficiently. Meaning under most scenarios, hardware composer
    // will accept layers with the data space and pixel format.
    static ui::Dataspace defaultCompositionDataspace;
    static ui::PixelFormat defaultCompositionPixelFormat;

    // The data space and pixel format that SurfaceFlinger expects hardware composer
    // to composite efficiently for wide color gamut surfaces. Meaning under most scenarios,
    // hardware composer will accept layers with the data space and pixel format.
    static ui::Dataspace wideColorGamutCompositionDataspace;
    static ui::PixelFormat wideColorGamutCompositionPixelFormat;

    // Whether to use frame rate API when deciding about the refresh rate of the display. This
    // variable is caches in SF, so that we can check it with each layer creation, and a void the
    // overhead that is caused by reading from sysprop.
    static bool useFrameRateApi;

    static constexpr SkipInitializationTag SkipInitialization;

    // Whether or not SDR layers should be dimmed to the desired SDR white point instead of
    // being treated as native display brightness
    static bool enableSdrDimming;

    static bool enableLatchUnsignaled;

    // must be called before clients can connect
    void init() ANDROID_API;

    // starts SurfaceFlinger main loop in the current thread
    void run() ANDROID_API;

    SurfaceFlingerBE& getBE() { return mBE; }
    const SurfaceFlingerBE& getBE() const { return mBE; }

    // Schedule an asynchronous or synchronous task on the main thread.
    template <typename F, typename T = std::invoke_result_t<F>>
    [[nodiscard]] std::future<T> schedule(F&&);

    // Schedule commit of transactions on the main thread ahead of the next VSYNC.
    void scheduleCommit(FrameHint);
    // As above, but also force composite regardless if transactions were committed.
    void scheduleComposite(FrameHint) override;
    // As above, but also force dirty geometry to repaint.
    void scheduleRepaint();
    // Schedule sampling independently from commit or composite.
    void scheduleSample();

    surfaceflinger::Factory& getFactory() { return mFactory; }

    // The CompositionEngine encapsulates all composition related interfaces and actions.
    compositionengine::CompositionEngine& getCompositionEngine() const;

    // Obtains a name from the texture pool, or, if the pool is empty, posts a
    // synchronous message to the main thread to obtain one on the fly
    uint32_t getNewTexture();

    // utility function to delete a texture on the main thread
    void deleteTextureAsync(uint32_t texture);

    renderengine::RenderEngine& getRenderEngine() const;

    bool authenticateSurfaceTextureLocked(
        const sp<IGraphicBufferProducer>& bufferProducer) const;

    void onLayerFirstRef(Layer*);
    void onLayerDestroyed(Layer*);
    void onLayerUpdate();

    void removeHierarchyFromOffscreenLayers(Layer* layer);
    void removeFromOffscreenLayers(Layer* layer);

    // TODO: Remove atomic if move dtor to main thread CL lands
    std::atomic<uint32_t> mNumClones;

    TransactionCallbackInvoker& getTransactionCallbackInvoker() {
        return mTransactionCallbackInvoker;
    }

    // Converts from a binder handle to a Layer
    // Returns nullptr if the handle does not point to an existing layer.
    // Otherwise, returns a weak reference so that callers off the main-thread
    // won't accidentally hold onto the last strong reference.
    wp<Layer> fromHandle(const sp<IBinder>& handle) const;

    // If set, disables reusing client composition buffers. This can be set by
    // debug.sf.disable_client_composition_cache
    bool mDisableClientCompositionCache = false;
    void windowInfosReported();

    // Disables expensive rendering for all displays
    // This is scheduled on the main thread
    void disableExpensiveRendering();

protected:
    // We're reference counted, never destroy SurfaceFlinger directly
    virtual ~SurfaceFlinger();

    virtual void processDisplayAdded(const wp<IBinder>& displayToken, const DisplayDeviceState&)
            REQUIRES(mStateLock);

    // Returns true if any display matches a `bool(const DisplayDevice&)` predicate.
    template <typename Predicate>
    bool hasDisplay(Predicate p) const REQUIRES(mStateLock) {
        return static_cast<bool>(findDisplay(p));
    }

    bool exceedsMaxRenderTargetSize(uint32_t width, uint32_t height) const {
        return width > mMaxRenderTargetSize || height > mMaxRenderTargetSize;
    }

private:
    friend class BufferLayer;
    friend class BufferQueueLayer;
    friend class BufferStateLayer;
    friend class Client;
    friend class FpsReporter;
    friend class TunnelModeEnabledReporter;
    friend class Layer;
    friend class MonitoredProducer;
    friend class RefreshRateOverlay;
    friend class RegionSamplingThread;
    friend class SurfaceTracing;

    // For unit tests
    friend class TestableSurfaceFlinger;
    friend class TransactionApplicationTest;
    friend class TunnelModeEnabledReporterTest;

    using RefreshRate = scheduler::RefreshRateConfigs::RefreshRate;
    using VsyncModulator = scheduler::VsyncModulator;
    using TransactionSchedule = scheduler::TransactionSchedule;
    using TraverseLayersFunction = std::function<void(const LayerVector::Visitor&)>;
    using RenderAreaFuture = std::future<std::unique_ptr<RenderArea>>;
    using DumpArgs = Vector<String16>;
    using Dumper = std::function<void(const DumpArgs&, bool asProto, std::string&)>;

    // This value is specified in number of frames.  Log frame stats at most
    // every half hour.
    enum { LOG_FRAME_STATS_PERIOD =  30*60*60 };

    class State {
    public:
        explicit State(LayerVector::StateSet set) : stateSet(set), layersSortedByZ(set) {}
        State& operator=(const State& other) {
            // We explicitly don't copy stateSet so that, e.g., mDrawingState
            // always uses the Drawing StateSet.
            layersSortedByZ = other.layersSortedByZ;
            displays = other.displays;
            colorMatrixChanged = other.colorMatrixChanged;
            if (colorMatrixChanged) {
                colorMatrix = other.colorMatrix;
            }
            globalShadowSettings = other.globalShadowSettings;

            return *this;
        }

        const LayerVector::StateSet stateSet = LayerVector::StateSet::Invalid;
        LayerVector layersSortedByZ;
        DefaultKeyedVector< wp<IBinder>, DisplayDeviceState> displays;

        bool colorMatrixChanged = true;
        mat4 colorMatrix;

        renderengine::ShadowSettings globalShadowSettings;

        void traverse(const LayerVector::Visitor& visitor) const;
        void traverseInZOrder(const LayerVector::Visitor& visitor) const;
        void traverseInReverseZOrder(const LayerVector::Visitor& visitor) const;
    };

    // Keeps track of pending buffers per layer handle in the transaction queue or current/drawing
    // state before the buffers are latched. The layer owns the atomic counters and decrements the
    // count in the main thread when dropping or latching a buffer.
    //
    // The binder threads increment the same counter when a new transaction containing a buffer is
    // added to the transaction queue. The map is updated with the layer handle lifecycle updates.
    // This is done to avoid lock contention with the main thread.
    class BufferCountTracker {
    public:
        void increment(BBinder* layerHandle) {
            std::lock_guard<std::mutex> lock(mLock);
            auto it = mCounterByLayerHandle.find(layerHandle);
            if (it != mCounterByLayerHandle.end()) {
                auto [name, pendingBuffers] = it->second;
                int32_t count = ++(*pendingBuffers);
                ATRACE_INT(name.c_str(), count);
            } else {
                ALOGW("Handle not found! %p", layerHandle);
            }
        }

        void add(BBinder* layerHandle, const std::string& name, std::atomic<int32_t>* counter) {
            std::lock_guard<std::mutex> lock(mLock);
            mCounterByLayerHandle[layerHandle] = std::make_pair(name, counter);
        }

        void remove(BBinder* layerHandle) {
            std::lock_guard<std::mutex> lock(mLock);
            mCounterByLayerHandle.erase(layerHandle);
        }

    private:
        std::mutex mLock;
        std::unordered_map<BBinder*, std::pair<std::string, std::atomic<int32_t>*>>
                mCounterByLayerHandle GUARDED_BY(mLock);
    };

    using ActiveModeInfo = DisplayDevice::ActiveModeInfo;

    enum class BootStage {
        BOOTLOADER,
        BOOTANIMATION,
        FINISHED,
    };

    struct HotplugEvent {
        hal::HWDisplayId hwcDisplayId;
        hal::Connection connection = hal::Connection::INVALID;
    };

    template <typename F, std::enable_if_t<!std::is_member_function_pointer_v<F>>* = nullptr>
    static Dumper dumper(F&& dump) {
        using namespace std::placeholders;
        return std::bind(std::forward<F>(dump), _3);
    }

    template <typename F, std::enable_if_t<std::is_member_function_pointer_v<F>>* = nullptr>
    Dumper dumper(F dump) {
        using namespace std::placeholders;
        return std::bind(dump, this, _3);
    }

    template <typename F>
    Dumper argsDumper(F dump) {
        using namespace std::placeholders;
        return std::bind(dump, this, _1, _3);
    }

    template <typename F>
    Dumper protoDumper(F dump) {
        using namespace std::placeholders;
        return std::bind(dump, this, _1, _2, _3);
    }

    template <typename... Args,
              typename Handler = VsyncModulator::VsyncConfigOpt (VsyncModulator::*)(Args...)>
    void modulateVsync(Handler handler, Args... args) {
        if (const auto config = (*mVsyncModulator.*handler)(args...)) {
            const auto vsyncPeriod = mScheduler->getVsyncPeriodFromRefreshRateConfigs();
            setVsyncConfig(*config, vsyncPeriod);
        }
    }

    static const int MAX_TRACING_MEMORY = 100 * 1024 * 1024; // 100MB
    // Maximum allowed number of display frames that can be set through backdoor
    static const int MAX_ALLOWED_DISPLAY_FRAMES = 2048;

    // Implements IBinder.
    status_t onTransact(uint32_t code, const Parcel& data, Parcel* reply, uint32_t flags) override;
    status_t dump(int fd, const Vector<String16>& args) override { return priorityDump(fd, args); }
    bool callingThreadHasUnscopedSurfaceFlingerAccess(bool usePermissionCache = true)
            EXCLUDES(mStateLock);

    // Implements ISurfaceComposer
    sp<ISurfaceComposerClient> createConnection() override;
    sp<IBinder> createDisplay(const String8& displayName, bool secure) override;
    void destroyDisplay(const sp<IBinder>& displayToken) override;
    std::vector<PhysicalDisplayId> getPhysicalDisplayIds() const override EXCLUDES(mStateLock) {
        Mutex::Autolock lock(mStateLock);
        return getPhysicalDisplayIdsLocked();
    }
    status_t getPrimaryPhysicalDisplayId(PhysicalDisplayId*) const override EXCLUDES(mStateLock);

    sp<IBinder> getPhysicalDisplayToken(PhysicalDisplayId displayId) const override;
    status_t setTransactionState(const FrameTimelineInfo& frameTimelineInfo,
                                 const Vector<ComposerState>& state,
                                 const Vector<DisplayState>& displays, uint32_t flags,
                                 const sp<IBinder>& applyToken,
                                 const InputWindowCommands& inputWindowCommands,
                                 int64_t desiredPresentTime, bool isAutoTimestamp,
                                 const client_cache_t& uncacheBuffer, bool hasListenerCallbacks,
                                 const std::vector<ListenerCallbacks>& listenerCallbacks,
                                 uint64_t transactionId) override;
    void bootFinished() override;
    bool authenticateSurfaceTexture(
            const sp<IGraphicBufferProducer>& bufferProducer) const override;
    status_t getSupportedFrameTimestamps(std::vector<FrameEvent>* outSupported) const override;
    sp<IDisplayEventConnection> createDisplayEventConnection(
            ISurfaceComposer::VsyncSource vsyncSource = eVsyncSourceApp,
            ISurfaceComposer::EventRegistrationFlags eventRegistration = {}) override;

    status_t captureDisplay(const DisplayCaptureArgs&, const sp<IScreenCaptureListener>&) override;
    status_t captureDisplay(DisplayId, const sp<IScreenCaptureListener>&) override;
    status_t captureLayers(const LayerCaptureArgs&, const sp<IScreenCaptureListener>&) override;

    status_t getDisplayStats(const sp<IBinder>& displayToken, DisplayStatInfo* stats) override;
    status_t getDisplayState(const sp<IBinder>& displayToken, ui::DisplayState*)
            EXCLUDES(mStateLock) override;
    status_t getStaticDisplayInfo(const sp<IBinder>& displayToken, ui::StaticDisplayInfo*)
            EXCLUDES(mStateLock) override;
    status_t getDynamicDisplayInfo(const sp<IBinder>& displayToken, ui::DynamicDisplayInfo*)
            EXCLUDES(mStateLock) override;
    status_t getDisplayNativePrimaries(const sp<IBinder>& displayToken,
                                       ui::DisplayPrimaries&) override;
    status_t setActiveColorMode(const sp<IBinder>& displayToken, ui::ColorMode colorMode) override;
    void setAutoLowLatencyMode(const sp<IBinder>& displayToken, bool on) override;
    void setGameContentType(const sp<IBinder>& displayToken, bool on) override;
    void setPowerMode(const sp<IBinder>& displayToken, int mode) override;
    status_t clearAnimationFrameStats() override;
    status_t getAnimationFrameStats(FrameStats* outStats) const override;
    status_t overrideHdrTypes(const sp<IBinder>& displayToken,
                              const std::vector<ui::Hdr>& hdrTypes) override;
    status_t onPullAtom(const int32_t atomId, std::string* pulledData, bool* success) override;
    status_t enableVSyncInjections(bool enable) override;
    status_t injectVSync(nsecs_t when) override;
    status_t getLayerDebugInfo(std::vector<LayerDebugInfo>* outLayers) override;
    status_t getColorManagement(bool* outGetColorManagement) const override;
    status_t getCompositionPreference(ui::Dataspace* outDataspace, ui::PixelFormat* outPixelFormat,
                                      ui::Dataspace* outWideColorGamutDataspace,
                                      ui::PixelFormat* outWideColorGamutPixelFormat) const override;
    status_t getDisplayedContentSamplingAttributes(const sp<IBinder>& displayToken,
                                                   ui::PixelFormat* outFormat,
                                                   ui::Dataspace* outDataspace,
                                                   uint8_t* outComponentMask) const override;
    status_t setDisplayContentSamplingEnabled(const sp<IBinder>& displayToken, bool enable,
                                              uint8_t componentMask, uint64_t maxFrames) override;
    status_t getDisplayedContentSample(const sp<IBinder>& displayToken, uint64_t maxFrames,
                                       uint64_t timestamp,
                                       DisplayedFrameStats* outStats) const override;
    status_t getProtectedContentSupport(bool* outSupported) const override;
    status_t isWideColorDisplay(const sp<IBinder>& displayToken,
                                bool* outIsWideColorDisplay) const override;
    status_t addRegionSamplingListener(const Rect& samplingArea, const sp<IBinder>& stopLayerHandle,
                                       const sp<IRegionSamplingListener>& listener) override;
    status_t removeRegionSamplingListener(const sp<IRegionSamplingListener>& listener) override;
    status_t addFpsListener(int32_t taskId, const sp<gui::IFpsListener>& listener) override;
    status_t removeFpsListener(const sp<gui::IFpsListener>& listener) override;
    status_t addTunnelModeEnabledListener(
            const sp<gui::ITunnelModeEnabledListener>& listener) override;
    status_t removeTunnelModeEnabledListener(
            const sp<gui::ITunnelModeEnabledListener>& listener) override;
    status_t setDesiredDisplayModeSpecs(const sp<IBinder>& displayToken,
                                        ui::DisplayModeId displayModeId, bool allowGroupSwitching,
                                        float primaryRefreshRateMin, float primaryRefreshRateMax,
                                        float appRequestRefreshRateMin,
                                        float appRequestRefreshRateMax) override;
    status_t getDesiredDisplayModeSpecs(const sp<IBinder>& displayToken,
                                        ui::DisplayModeId* outDefaultMode,
                                        bool* outAllowGroupSwitching,
                                        float* outPrimaryRefreshRateMin,
                                        float* outPrimaryRefreshRateMax,
                                        float* outAppRequestRefreshRateMin,
                                        float* outAppRequestRefreshRateMax) override;
    status_t getDisplayBrightnessSupport(const sp<IBinder>& displayToken,
                                         bool* outSupport) const override;
    status_t setDisplayBrightness(const sp<IBinder>& displayToken,
                                  const gui::DisplayBrightness& brightness) override;
    status_t addHdrLayerInfoListener(const sp<IBinder>& displayToken,
                                     const sp<gui::IHdrLayerInfoListener>& listener) override;
    status_t removeHdrLayerInfoListener(const sp<IBinder>& displayToken,
                                        const sp<gui::IHdrLayerInfoListener>& listener) override;
    status_t notifyPowerBoost(int32_t boostId) override;
    status_t setGlobalShadowSettings(const half4& ambientColor, const half4& spotColor,
                                     float lightPosY, float lightPosZ, float lightRadius) override;
    status_t setFrameRate(const sp<IGraphicBufferProducer>& surface, float frameRate,
                          int8_t compatibility, int8_t changeFrameRateStrategy) override;
    status_t acquireFrameRateFlexibilityToken(sp<IBinder>* outToken) override;

    status_t setFrameTimelineInfo(const sp<IGraphicBufferProducer>& surface,
                                  const FrameTimelineInfo& frameTimelineInfo) override;

    status_t addTransactionTraceListener(
            const sp<gui::ITransactionTraceListener>& listener) override;

    int getGPUContextPriority() override;

    status_t getMaxAcquiredBufferCount(int* buffers) const override;

    status_t addWindowInfosListener(
            const sp<gui::IWindowInfosListener>& windowInfosListener) const override;
    status_t removeWindowInfosListener(
            const sp<gui::IWindowInfosListener>& windowInfosListener) const override;

    // Implements IBinder::DeathRecipient.
    void binderDied(const wp<IBinder>& who) override;

    // HWC2::ComposerCallback overrides:
    void onComposerHalVsync(hal::HWDisplayId, int64_t timestamp,
                            std::optional<hal::VsyncPeriodNanos>) override;
    void onComposerHalHotplug(hal::HWDisplayId, hal::Connection) override;
    void onComposerHalRefresh(hal::HWDisplayId) override;
    void onComposerHalVsyncPeriodTimingChanged(hal::HWDisplayId,
                                               const hal::VsyncPeriodChangeTimeline&) override;
    void onComposerHalSeamlessPossible(hal::HWDisplayId) override;

    // ICompositor overrides:

    // Commits transactions for layers and displays. Returns whether any state has been invalidated,
    // i.e. whether a frame should be composited for each display.
    bool commit(nsecs_t frameTime, int64_t vsyncId, nsecs_t expectedVsyncTime) override;

    // Composites a frame for each display. CompositionEngine performs GPU and/or HAL composition
    // via RenderEngine and the Composer HAL, respectively.
    void composite(nsecs_t frameTime) override;

    // Samples the composited frame via RegionSamplingThread.
    void sample() override;

    /*
     * ISchedulerCallback
     */

    // Toggles hardware VSYNC by calling into HWC.
    void setVsyncEnabled(bool) override;
    // Initiates a refresh rate change to be applied on invalidate.
    void changeRefreshRate(const Scheduler::RefreshRate&, Scheduler::ModeEvent) override;
    // Called when kernel idle timer has expired. Used to update the refresh rate overlay.
    void kernelTimerChanged(bool expired) override;
    // Called when the frame rate override list changed to trigger an event.
    void triggerOnFrameRateOverridesChanged() override;
    // Toggles the kernel idle timer on or off depending the policy decisions around refresh rates.
    void toggleKernelIdleTimer() REQUIRES(mStateLock);
    // Keeps track of whether the kernel idle timer is currently enabled, so we don't have to
    // make calls to sys prop each time.
    bool mKernelIdleTimerEnabled = false;
    // Show spinner with refresh rate overlay
    bool mRefreshRateOverlaySpinner = false;

    // Called on the main thread in response to initializeDisplays()
    void onInitializeDisplays() REQUIRES(mStateLock);
    // Sets the desired active mode bit. It obtains the lock, and sets mDesiredActiveMode.
    void setDesiredActiveMode(const ActiveModeInfo& info) REQUIRES(mStateLock);
    status_t setActiveMode(const sp<IBinder>& displayToken, int id);
    // Once HWC has returned the present fence, this sets the active mode and a new refresh
    // rate in SF.
    void setActiveModeInternal() REQUIRES(mStateLock);
    // Calls to setActiveMode on the main thread if there is a pending mode change
    // that needs to be applied.
    void performSetActiveMode() REQUIRES(mStateLock);
    void clearDesiredActiveModeState(const sp<DisplayDevice>&) REQUIRES(mStateLock);
    // Called when active mode is no longer is progress
    void desiredActiveModeChangeDone(const sp<DisplayDevice>&) REQUIRES(mStateLock);
    // Called on the main thread in response to setPowerMode()
    void setPowerModeInternal(const sp<DisplayDevice>& display, hal::PowerMode mode)
            REQUIRES(mStateLock);

    // Sets the desired display mode specs.
    status_t setDesiredDisplayModeSpecsInternal(
            const sp<DisplayDevice>& display,
            const std::optional<scheduler::RefreshRateConfigs::Policy>& policy, bool overridePolicy)
            EXCLUDES(mStateLock);

    // Returns whether transactions were committed.
    bool flushAndCommitTransactions() EXCLUDES(mStateLock);

    void commitTransactions() EXCLUDES(mStateLock);
    void commitTransactionsLocked(uint32_t transactionFlags) REQUIRES(mStateLock);
    void doCommitTransactions() REQUIRES(mStateLock);

    // Returns whether a new buffer has been latched.
    bool latchBuffers();

    void updateLayerGeometry();

    void updateInputFlinger();
    void notifyWindowInfos();
    void commitInputWindowCommands() REQUIRES(mStateLock);
    void updateCursorAsync();

    void initScheduler(const sp<DisplayDevice>& display) REQUIRES(mStateLock);
    void updatePhaseConfiguration(const Fps&) REQUIRES(mStateLock);
    void setVsyncConfig(const VsyncModulator::VsyncConfig&, nsecs_t vsyncPeriod);


    /*
     * Transactions
     */
    bool applyTransactionState(const FrameTimelineInfo& info, const Vector<ComposerState>& state,
                               const Vector<DisplayState>& displays, uint32_t flags,
                               const InputWindowCommands& inputWindowCommands,
                               const int64_t desiredPresentTime, bool isAutoTimestamp,
                               const client_cache_t& uncacheBuffer, const int64_t postTime,
                               uint32_t permissions, bool hasListenerCallbacks,
                               const std::vector<ListenerCallbacks>& listenerCallbacks,
                               int originPid, int originUid, uint64_t transactionId)
            REQUIRES(mStateLock);
    // flush pending transaction that was presented after desiredPresentTime.
    bool flushTransactionQueues();
    // Returns true if there is at least one transaction that needs to be flushed
    bool transactionFlushNeeded();

    uint32_t setClientStateLocked(const FrameTimelineInfo&, const ComposerState&,
                                  int64_t desiredPresentTime, bool isAutoTimestamp,
                                  int64_t postTime, uint32_t permissions) REQUIRES(mStateLock);

    uint32_t getTransactionFlags() const;

    // Sets the masked bits, and returns the old flags.
    uint32_t setTransactionFlags(uint32_t mask);

    // Clears and returns the masked bits.
    uint32_t clearTransactionFlags(uint32_t mask);

    // Indicate SF should call doTraversal on layers, but don't trigger a wakeup! We use this cases
    // where there are still pending transactions but we know they won't be ready until a frame
    // arrives from a different layer. So we need to ensure we performTransaction from invalidate
    // but there is no need to try and wake up immediately to do it. Rather we rely on
    // onFrameAvailable or another layer update to wake us up.
    void setTraversalNeeded();
    uint32_t setTransactionFlags(uint32_t mask, TransactionSchedule,
                                 const sp<IBinder>& applyToken = {});
    void commitOffscreenLayers();
    bool transactionIsReadyToBeApplied(
            const FrameTimelineInfo& info, bool isAutoTimestamp, int64_t desiredPresentTime,
            uid_t originUid, const Vector<ComposerState>& states,
            const std::unordered_set<sp<IBinder>, ISurfaceComposer::SpHash<IBinder>>&
                    bufferLayersReadyToPresent) const REQUIRES(mStateLock);
    uint32_t setDisplayStateLocked(const DisplayState& s) REQUIRES(mStateLock);
    uint32_t addInputWindowCommands(const InputWindowCommands& inputWindowCommands)
            REQUIRES(mStateLock);
    bool frameIsEarly(nsecs_t expectedPresentTime, int64_t vsyncId) const;
    /*
     * Layer management
     */
    status_t createLayer(const String8& name, const sp<Client>& client, uint32_t w, uint32_t h,
                         PixelFormat format, uint32_t flags, LayerMetadata metadata,
                         sp<IBinder>* handle, sp<IGraphicBufferProducer>* gbp,
                         const sp<IBinder>& parentHandle, int32_t* outLayerId,
                         const sp<Layer>& parentLayer = nullptr,
                         uint32_t* outTransformHint = nullptr);

    status_t createBufferQueueLayer(const sp<Client>& client, std::string name, uint32_t w,
                                    uint32_t h, uint32_t flags, LayerMetadata metadata,
                                    PixelFormat& format, sp<IBinder>* outHandle,
                                    sp<IGraphicBufferProducer>* outGbp, sp<Layer>* outLayer);

    status_t createBufferStateLayer(const sp<Client>& client, std::string name, uint32_t w,
                                    uint32_t h, uint32_t flags, LayerMetadata metadata,
                                    sp<IBinder>* outHandle, sp<Layer>* outLayer);

    status_t createEffectLayer(const sp<Client>& client, std::string name, uint32_t w, uint32_t h,
                               uint32_t flags, LayerMetadata metadata, sp<IBinder>* outHandle,
                               sp<Layer>* outLayer);

    status_t createContainerLayer(const sp<Client>& client, std::string name, uint32_t w,
                                  uint32_t h, uint32_t flags, LayerMetadata metadata,
                                  sp<IBinder>* outHandle, sp<Layer>* outLayer);

    status_t mirrorLayer(const sp<Client>& client, const sp<IBinder>& mirrorFromHandle,
                         sp<IBinder>* outHandle, int32_t* outLayerId);

    std::string getUniqueLayerName(const char* name);

    // called when all clients have released all their references to
    // this layer meaning it is entirely safe to destroy all
    // resources associated to this layer.
    void onHandleDestroyed(BBinder* handle, sp<Layer>& layer);
    void markLayerPendingRemovalLocked(const sp<Layer>& layer);

    // add a layer to SurfaceFlinger
    status_t addClientLayer(const sp<Client>& client, const sp<IBinder>& handle,
                            const sp<IGraphicBufferProducer>& gbc, const sp<Layer>& lbc,
                            const sp<IBinder>& parentHandle, const sp<Layer>& parentLayer,
                            bool addToRoot, uint32_t* outTransformHint);

    // Traverse through all the layers and compute and cache its bounds.
    void computeLayerBounds();

    // Boot animation, on/off animations and screen capture
    void startBootAnim();

    std::shared_future<renderengine::RenderEngineResult> captureScreenCommon(
            RenderAreaFuture, TraverseLayersFunction, ui::Size bufferSize, ui::PixelFormat,
            bool allowProtected, bool grayscale, const sp<IScreenCaptureListener>&);
    std::shared_future<renderengine::RenderEngineResult> captureScreenCommon(
            RenderAreaFuture, TraverseLayersFunction,
            const std::shared_ptr<renderengine::ExternalTexture>&, bool regionSampling,
            bool grayscale, const sp<IScreenCaptureListener>&);
    std::shared_future<renderengine::RenderEngineResult> renderScreenImplLocked(
            const RenderArea&, TraverseLayersFunction,
            const std::shared_ptr<renderengine::ExternalTexture>&, bool canCaptureBlackoutContent,
            bool regionSampling, bool grayscale, ScreenCaptureResults&);

    // If the uid provided is not UNSET_UID, the traverse will skip any layers that don't have a
    // matching ownerUid
    void traverseLayersInLayerStack(ui::LayerStack, const int32_t uid, const LayerVector::Visitor&);

    void readPersistentProperties();

    uint32_t getMaxAcquiredBufferCountForCurrentRefreshRate(uid_t uid) const;

    /*
     * Display and layer stack management
     */
    // called when starting, or restarting after system_server death
    void initializeDisplays();

    sp<const DisplayDevice> getDisplayDeviceLocked(const wp<IBinder>& displayToken) const
            REQUIRES(mStateLock) {
        return const_cast<SurfaceFlinger*>(this)->getDisplayDeviceLocked(displayToken);
    }

    sp<DisplayDevice> getDisplayDeviceLocked(const wp<IBinder>& displayToken) REQUIRES(mStateLock) {
        const auto it = mDisplays.find(displayToken);
        return it == mDisplays.end() ? nullptr : it->second;
    }

    sp<const DisplayDevice> getDisplayDeviceLocked(PhysicalDisplayId id) const
            REQUIRES(mStateLock) {
        return const_cast<SurfaceFlinger*>(this)->getDisplayDeviceLocked(id);
    }

    sp<DisplayDevice> getDisplayDeviceLocked(PhysicalDisplayId id) REQUIRES(mStateLock) {
        if (const auto token = getPhysicalDisplayTokenLocked(id)) {
            return getDisplayDeviceLocked(token);
        }
        return nullptr;
    }

    sp<const DisplayDevice> getDefaultDisplayDeviceLocked() const REQUIRES(mStateLock) {
        return const_cast<SurfaceFlinger*>(this)->getDefaultDisplayDeviceLocked();
    }

    sp<DisplayDevice> getDefaultDisplayDeviceLocked() REQUIRES(mStateLock) {
        if (const auto display = getDisplayDeviceLocked(mActiveDisplayToken)) {
            return display;
        }
        // The active display is outdated, fall back to the internal display
        mActiveDisplayToken.clear();
        if (const auto token = getInternalDisplayTokenLocked()) {
            return getDisplayDeviceLocked(token);
        }
        return nullptr;
    }

    sp<const DisplayDevice> getDefaultDisplayDevice() const EXCLUDES(mStateLock) {
        Mutex::Autolock lock(mStateLock);
        return getDefaultDisplayDeviceLocked();
    }

    // Returns the first display that matches a `bool(const DisplayDevice&)` predicate.
    template <typename Predicate>
    sp<DisplayDevice> findDisplay(Predicate p) const REQUIRES(mStateLock) {
        const auto it = std::find_if(mDisplays.begin(), mDisplays.end(),
                                     [&](const auto& pair) { return p(*pair.second); });

        return it == mDisplays.end() ? nullptr : it->second;
    }

    sp<const DisplayDevice> getDisplayDeviceLocked(DisplayId id) const REQUIRES(mStateLock) {
        // TODO(b/182939859): Replace tokens with IDs for display lookup.
        return findDisplay([id](const auto& display) { return display.getId() == id; });
    }

    std::vector<PhysicalDisplayId> getPhysicalDisplayIdsLocked() const REQUIRES(mStateLock);

    // mark a region of a layer stack dirty. this updates the dirty
    // region of all screens presenting this layer stack.
    void invalidateLayerStack(const sp<const Layer>& layer, const Region& dirty);

    sp<DisplayDevice> getDisplayWithInputByLayer(Layer* layer) const REQUIRES(mStateLock);

    bool isDisplayActiveLocked(const sp<const DisplayDevice>& display) const REQUIRES(mStateLock) {
        return display->getDisplayToken() == mActiveDisplayToken;
    }

    /*
     * H/W composer
     */
    // The following thread safety rules apply when accessing HWComposer:
    // 1. When reading display state from HWComposer on the main thread, it's not necessary to
    //    acquire mStateLock.
    // 2. When accessing HWComposer on a thread other than the main thread, we always
    //    need to acquire mStateLock. This is because the main thread could be
    //    in the process of writing display state, e.g. creating or destroying a display.
    HWComposer& getHwComposer() const;

    /*
     * Compositing
     */
    void postComposition();
    void getCompositorTiming(CompositorTiming* compositorTiming);
    void updateCompositorTiming(const DisplayStatInfo& stats, nsecs_t compositeTime,
                                std::shared_ptr<FenceTime>& presentFenceTime);
    void setCompositorTimingSnapped(const DisplayStatInfo& stats,
                                    nsecs_t compositeToPresentLatency);

    void postFrame();

    /*
     * Display management
     */
    void loadDisplayModes(PhysicalDisplayId displayId, DisplayModes& outModes,
                          DisplayModePtr& outActiveMode) const REQUIRES(mStateLock);
    sp<DisplayDevice> setupNewDisplayDeviceInternal(
            const wp<IBinder>& displayToken,
            std::shared_ptr<compositionengine::Display> compositionDisplay,
            const DisplayDeviceState& state,
            const sp<compositionengine::DisplaySurface>& displaySurface,
            const sp<IGraphicBufferProducer>& producer) REQUIRES(mStateLock);
    void processDisplayChangesLocked() REQUIRES(mStateLock);
    void processDisplayRemoved(const wp<IBinder>& displayToken) REQUIRES(mStateLock);
    void processDisplayChanged(const wp<IBinder>& displayToken,
                               const DisplayDeviceState& currentState,
                               const DisplayDeviceState& drawingState) REQUIRES(mStateLock);
    void processDisplayHotplugEventsLocked() REQUIRES(mStateLock);

    void dispatchDisplayHotplugEvent(PhysicalDisplayId displayId, bool connected);

    /*
     * VSYNC
     */
    nsecs_t getVsyncPeriodFromHWC() const REQUIRES(mStateLock);

    void setHWCVsyncEnabled(PhysicalDisplayId id, hal::Vsync enabled) {
        mLastHWCVsyncState = enabled;
        getHwComposer().setVsyncEnabled(id, enabled);
    }

    // Sets the refresh rate by switching active configs, if they are available for
    // the desired refresh rate.
    void changeRefreshRateLocked(const RefreshRate&, Scheduler::ModeEvent) REQUIRES(mStateLock);

    struct FenceWithFenceTime {
        sp<Fence> fence = Fence::NO_FENCE;
        std::shared_ptr<FenceTime> fenceTime = FenceTime::NO_FENCE;
    };

    // Gets the fence for the previous frame.
    // Must be called on the main thread.
    FenceWithFenceTime previousFrameFence();

    // Whether the previous frame has not yet been presented to the display.
    // If graceTimeMs is positive, this method waits for at most the provided
    // grace period before reporting if the frame missed.
    // Must be called on the main thread.
    bool previousFramePending(int graceTimeMs = 0);

    // Returns the previous time that the frame was presented. If the frame has
    // not been presented yet, then returns Fence::SIGNAL_TIME_PENDING. If there
    // is no pending frame, then returns Fence::SIGNAL_TIME_INVALID.
    // Must be called on the main thread.
    nsecs_t previousFramePresentTime();

    // Calculates the expected present time for this frame. For negative offsets, performs a
    // correction using the predicted vsync for the next frame instead.

    nsecs_t calculateExpectedPresentTime(DisplayStatInfo) const;

    /*
     * Display identification
     */
    sp<IBinder> getPhysicalDisplayTokenLocked(PhysicalDisplayId displayId) const
            REQUIRES(mStateLock) {
        const auto it = mPhysicalDisplayTokens.find(displayId);
        return it != mPhysicalDisplayTokens.end() ? it->second : nullptr;
    }

    std::optional<PhysicalDisplayId> getPhysicalDisplayIdLocked(
            const sp<IBinder>& displayToken) const REQUIRES(mStateLock) {
        for (const auto& [id, token] : mPhysicalDisplayTokens) {
            if (token == displayToken) {
                return id;
            }
        }
        return {};
    }

    // TODO(b/182939859): SF conflates the primary (a.k.a. default) display with the first display
    // connected at boot, which is typically internal. (Theoretically, it must be internal because
    // SF does not support disconnecting it, though in practice HWC may circumvent this limitation.)
    //
    // SF inherits getInternalDisplayToken and getInternalDisplayId from ISurfaceComposer, so these
    // locked counterparts are named consistently. Once SF supports headless mode and can designate
    // any display as primary, the "internal" misnomer will be phased out.
    sp<IBinder> getInternalDisplayTokenLocked() const REQUIRES(mStateLock) {
        return getPhysicalDisplayTokenLocked(getInternalDisplayIdLocked());
    }

    PhysicalDisplayId getInternalDisplayIdLocked() const REQUIRES(mStateLock) {
        return getHwComposer().getPrimaryDisplayId();
    }

    // Toggles use of HAL/GPU virtual displays.
    void enableHalVirtualDisplays(bool);

    // Virtual display lifecycle for ID generation and HAL allocation.
    VirtualDisplayId acquireVirtualDisplay(ui::Size, ui::PixelFormat) REQUIRES(mStateLock);
    void releaseVirtualDisplay(VirtualDisplayId);

    void onActiveDisplayChangedLocked(const sp<DisplayDevice>& activeDisplay) REQUIRES(mStateLock);

    void onActiveDisplaySizeChanged(const sp<DisplayDevice>& activeDisplay);

    /*
     * Debugging & dumpsys
     */
    void dumpAllLocked(const DumpArgs& args, std::string& result) const REQUIRES(mStateLock);

    void appendSfConfigString(std::string& result) const;
    void listLayersLocked(std::string& result) const;
    void dumpStatsLocked(const DumpArgs& args, std::string& result) const REQUIRES(mStateLock);
    void clearStatsLocked(const DumpArgs& args, std::string& result);
    void dumpTimeStats(const DumpArgs& args, bool asProto, std::string& result) const;
    void dumpFrameTimeline(const DumpArgs& args, std::string& result) const;
    void logFrameStats();

    void dumpVSync(std::string& result) const REQUIRES(mStateLock);
    void dumpStaticScreenStats(std::string& result) const;
    // Not const because each Layer needs to query Fences and cache timestamps.
    void dumpFrameEventsLocked(std::string& result);
    void dumpDisplayIdentificationData(std::string& result) const REQUIRES(mStateLock);
    void dumpRawDisplayIdentificationData(const DumpArgs&, std::string& result) const;
    void dumpWideColorInfo(std::string& result) const REQUIRES(mStateLock);
    LayersProto dumpDrawingStateProto(uint32_t traceFlags) const;
    void dumpOffscreenLayersProto(LayersProto& layersProto,
                                  uint32_t traceFlags = SurfaceTracing::TRACE_ALL) const;
    void dumpDisplayProto(LayersTraceProto& layersTraceProto) const;

    // Dumps state from HW Composer
    void dumpHwc(std::string& result) const;
    LayersProto dumpProtoFromMainThread(uint32_t traceFlags = SurfaceTracing::TRACE_ALL)
            EXCLUDES(mStateLock);
    void dumpOffscreenLayers(std::string& result) EXCLUDES(mStateLock);
    void dumpPlannerInfo(const DumpArgs& args, std::string& result) const REQUIRES(mStateLock);

    status_t doDump(int fd, const DumpArgs& args, bool asProto);

    status_t dumpCritical(int fd, const DumpArgs&, bool asProto);

    status_t dumpAll(int fd, const DumpArgs& args, bool asProto) override {
        return doDump(fd, args, asProto);
    }

    void onFrameRateFlexibilityTokenReleased();

    static mat4 calculateColorMatrix(float saturation);

    void updateColorMatrixLocked();

    // Verify that transaction is being called by an approved process:
    // either AID_GRAPHICS or AID_SYSTEM.
    status_t CheckTransactCodeCredentials(uint32_t code);

    // Add transaction to the Transaction Queue
    void queueTransaction(TransactionState& state) EXCLUDES(mQueueLock);
    void waitForSynchronousTransaction(const CountDownLatch& transactionCommittedSignal);
    void signalSynchronousTransactions(const uint32_t flag);

    /*
     * Generic Layer Metadata
     */
    const std::unordered_map<std::string, uint32_t>& getGenericLayerMetadataKeyMap() const;

    /*
     * Misc
     */
    std::vector<ui::ColorMode> getDisplayColorModes(const DisplayDevice&) REQUIRES(mStateLock);

    static int calculateMaxAcquiredBufferCount(Fps refreshRate,
                                               std::chrono::nanoseconds presentLatency);
    int getMaxAcquiredBufferCountForRefreshRate(Fps refreshRate) const;

    void updateInternalDisplayVsyncLocked(const sp<DisplayDevice>& activeDisplay)
            REQUIRES(mStateLock);

    sp<StartPropertySetThread> mStartPropertySetThread;
    surfaceflinger::Factory& mFactory;

    std::future<void> mRenderEnginePrimeCacheFuture;

    // access must be protected by mStateLock
    mutable Mutex mStateLock;
    State mCurrentState{LayerVector::StateSet::Current};
    std::atomic<int32_t> mTransactionFlags = 0;
    std::vector<std::shared_ptr<CountDownLatch>> mTransactionCommittedSignals;
    bool mAnimTransactionPending = false;
    SortedVector<sp<Layer>> mLayersPendingRemoval;

    // global color transform states
    Daltonizer mDaltonizer;
    float mGlobalSaturationFactor = 1.0f;
    mat4 mClientColorMatrix;

    // Can't be unordered_set because wp<> isn't hashable
    std::set<wp<IBinder>> mGraphicBufferProducerList;
    size_t mMaxGraphicBufferProducerListSize = ISurfaceComposer::MAX_LAYERS;
    // If there are more GraphicBufferProducers tracked by SurfaceFlinger than
    // this threshold, then begin logging.
    size_t mGraphicBufferProducerListSizeLogThreshold =
            static_cast<size_t>(0.95 * static_cast<double>(MAX_LAYERS));

    void removeGraphicBufferProducerAsync(const wp<IBinder>&);

    // protected by mStateLock (but we could use another lock)
    bool mLayersRemoved = false;
    bool mLayersAdded = false;

    std::atomic_bool mMustComposite = false;
    std::atomic_bool mGeometryDirty = false;

    // constant members (no synchronization needed for access)
    const nsecs_t mBootTime = systemTime();
    bool mGpuToCpuSupported = false;
    bool mIsUserBuild = true;

    // Can only accessed from the main thread, these members
    // don't need synchronization
    State mDrawingState{LayerVector::StateSet::Drawing};
    bool mVisibleRegionsDirty = false;

    // VisibleRegions dirty is already cleared by postComp, but we need to track it to prevent
    // extra work in the HDR layer info listener.
    bool mVisibleRegionsWereDirtyThisFrame = false;
    // Used to ensure we omit a callback when HDR layer info listener is newly added but the
    // scene hasn't changed
    bool mAddingHDRLayerInfoListener = false;

    // Set during transaction application stage to track if the input info or children
    // for a layer has changed.
    // TODO: Also move visibleRegions over to a boolean system.
    bool mInputInfoChanged = false;
    bool mSomeChildrenChanged;
    bool mSomeDataspaceChanged = false;
    bool mForceTransactionDisplayChange = false;

    bool mAnimCompositionPending = false;

    // Tracks layers that have pending frames which are candidates for being
    // latched.
    std::unordered_set<sp<Layer>, ISurfaceComposer::SpHash<Layer>> mLayersWithQueuedFrames;
    // Tracks layers that need to update a display's dirty region.
    std::vector<sp<Layer>> mLayersPendingRefresh;
    std::array<FenceWithFenceTime, 2> mPreviousPresentFences;
    // True if in the previous frame at least one layer was composed via the GPU.
    bool mHadClientComposition = false;
    // True if in the previous frame at least one layer was composed via HW Composer.
    // Note that it is possible for a frame to be composed via both client and device
    // composition, for example in the case of overlays.
    bool mHadDeviceComposition = false;
    // True if in the previous frame, the client composition was skipped by reusing the buffer
    // used in a previous composition. This can happed if the client composition requests
    // did not change.
    bool mReusedClientComposition = false;

    BootStage mBootStage = BootStage::BOOTLOADER;

    std::vector<HotplugEvent> mPendingHotplugEvents GUARDED_BY(mStateLock);

    // this may only be written from the main thread with mStateLock held
    // it may be read from other threads with mStateLock held
    std::map<wp<IBinder>, sp<DisplayDevice>> mDisplays GUARDED_BY(mStateLock);
    std::unordered_map<PhysicalDisplayId, sp<IBinder>> mPhysicalDisplayTokens
            GUARDED_BY(mStateLock);

    struct {
        DisplayIdGenerator<GpuVirtualDisplayId> gpu;
        std::optional<DisplayIdGenerator<HalVirtualDisplayId>> hal;
    } mVirtualDisplayIdGenerators;

    std::atomic_uint mDebugFlashDelay = 0;
    std::atomic_bool mDebugDisableHWC = false;
    std::atomic_bool mDebugDisableTransformHint = false;
    std::atomic<nsecs_t> mDebugInTransaction = 0;
    std::atomic_bool mForceFullDamage = false;

    bool mLayerCachingEnabled = false;
    bool mPropagateBackpressureClientComposition = false;
    sp<SurfaceInterceptor> mInterceptor;

    SurfaceTracing mTracing{*this};
    std::mutex mTracingLock;
    bool mTracingEnabled = false;
    bool mTracePostComposition = false;
    std::atomic<bool> mTracingEnabledChanged = false;

    const std::shared_ptr<TimeStats> mTimeStats;
    const std::unique_ptr<FrameTracer> mFrameTracer;
    const std::unique_ptr<frametimeline::FrameTimeline> mFrameTimeline;

    // If blurs should be enabled on this device.
    bool mSupportsBlur = false;
    // If blurs are considered expensive and should require high GPU frequency.
    bool mBlursAreExpensive = false;
    std::atomic<uint32_t> mFrameMissedCount = 0;
    std::atomic<uint32_t> mHwcFrameMissedCount = 0;
    std::atomic<uint32_t> mGpuFrameMissedCount = 0;

    TransactionCallbackInvoker mTransactionCallbackInvoker;

    // these are thread safe
    std::unique_ptr<MessageQueue> mEventQueue;
    FrameTracker mAnimFrameTracker;

    // protected by mDestroyedLayerLock;
    mutable Mutex mDestroyedLayerLock;
    Vector<Layer const *> mDestroyedLayers;

    // We maintain a pool of pre-generated texture names to hand out to avoid
    // layer creation needing to run on the main thread (which it would
    // otherwise need to do to access RenderEngine).
    std::mutex mTexturePoolMutex;
    uint32_t mTexturePoolSize = 0;
    std::vector<uint32_t> mTexturePool;

    mutable Mutex mQueueLock;
    Condition mTransactionQueueCV;
    std::unordered_map<sp<IBinder>, std::queue<TransactionState>, IListenerHash>
            mPendingTransactionQueues GUARDED_BY(mQueueLock);
    std::queue<TransactionState> mTransactionQueue GUARDED_BY(mQueueLock);
    /*
     * Feature prototyping
     */

    // Static screen stats
    bool mHasPoweredOff = false;

    std::atomic<size_t> mNumLayers = 0;

    // to linkToDeath
    sp<IBinder> mWindowManager;
    // We want to avoid multiple calls to BOOT_FINISHED as they come in on
    // different threads without a lock and could trigger unsynchronized writes to
    // to mWindowManager or mInputFlinger
    std::atomic<bool> mBootFinished = false;

    std::thread::id mMainThreadId = std::this_thread::get_id();

    DisplayColorSetting mDisplayColorSetting = DisplayColorSetting::kEnhanced;

    // Color mode forced by setting persist.sys.sf.color_mode, it must:
    //     1. not be NATIVE color mode, NATIVE color mode means no forced color mode;
    //     2. be one of the supported color modes returned by hardware composer, otherwise
    //        it will not be respected.
    // persist.sys.sf.color_mode will only take effect when persist.sys.sf.native_mode
    // is not set to 1.
    // This property can be used to force SurfaceFlinger to always pick a certain color mode.
    ui::ColorMode mForceColorMode = ui::ColorMode::NATIVE;

    ui::Dataspace mDefaultCompositionDataspace;
    ui::Dataspace mWideColorGamutCompositionDataspace;
    ui::Dataspace mColorSpaceAgnosticDataspace;

    SurfaceFlingerBE mBE;
    std::unique_ptr<compositionengine::CompositionEngine> mCompositionEngine;
    // mMaxRenderTargetSize is only set once in init() so it doesn't need to be protected by
    // any mutex.
    size_t mMaxRenderTargetSize{1};

    const std::string mHwcServiceName;

    bool hasMockHwc() const { return mHwcServiceName == "mock"; }

    /*
     * Scheduler
     */
    std::unique_ptr<Scheduler> mScheduler;
    scheduler::ConnectionHandle mAppConnectionHandle;
    scheduler::ConnectionHandle mSfConnectionHandle;

    // Stores phase offsets configured per refresh rate.
    std::unique_ptr<scheduler::VsyncConfiguration> mVsyncConfiguration;

    // Optional to defer construction until PhaseConfiguration is created.
    sp<VsyncModulator> mVsyncModulator;

    std::unique_ptr<scheduler::RefreshRateStats> mRefreshRateStats;

    std::atomic<nsecs_t> mExpectedPresentTime = 0;
    nsecs_t mScheduledPresentTime = 0;
    hal::Vsync mHWCVsyncPendingState = hal::Vsync::DISABLE;
    hal::Vsync mLastHWCVsyncState = hal::Vsync::DISABLE;

    // below flags are set by main thread only
    bool mSetActiveModePending = false;

    bool mLumaSampling = true;
    sp<RegionSamplingThread> mRegionSamplingThread;
    sp<FpsReporter> mFpsReporter;
    sp<TunnelModeEnabledReporter> mTunnelModeEnabledReporter;
    ui::DisplayPrimaries mInternalDisplayPrimaries;

    const float mInternalDisplayDensity;
    const float mEmulatedDisplayDensity;

    sp<os::IInputFlinger> mInputFlinger;
    // Should only be accessed by the main thread.
    InputWindowCommands mInputWindowCommands;

    Hwc2::impl::PowerAdvisor mPowerAdvisor;

    void enableRefreshRateOverlay(bool enable) REQUIRES(mStateLock);

    // Flag used to set override desired display mode from backdoor
    bool mDebugDisplayModeSetByBackdoor = false;

    // A set of layers that have no parent so they are not drawn on screen.
    // Should only be accessed by the main thread.
    // The Layer pointer is removed from the set when the destructor is called so there shouldn't
    // be any issues with a raw pointer referencing an invalid object.
    std::unordered_set<Layer*> mOffscreenLayers;

    int mFrameRateFlexibilityTokenCount = 0;

    sp<IBinder> mDebugFrameRateFlexibilityToken;

    BufferCountTracker mBufferCountTracker;

    std::unordered_map<DisplayId, sp<HdrLayerInfoReporter>> mHdrLayerInfoListeners
            GUARDED_BY(mStateLock);
    mutable Mutex mCreatedLayersLock;
    struct LayerCreatedState {
        LayerCreatedState(const wp<Layer>& layer, const wp<IBinder>& parent,
                          const wp<Layer> parentLayer, const wp<IBinder>& producer, bool addToRoot)
              : layer(layer),
                initialParent(parent),
                initialParentLayer(parentLayer),
                initialProducer(producer),
                addToRoot(addToRoot) {}
        wp<Layer> layer;
        // Indicates the initial parent of the created layer, only used for creating layer in
        // SurfaceFlinger. If nullptr, it may add the created layer into the current root layers.
        wp<IBinder> initialParent;
        wp<Layer> initialParentLayer;
        // Indicates the initial graphic buffer producer of the created layer, only used for
        // creating layer in SurfaceFlinger.
        wp<IBinder> initialProducer;
        // Indicates whether the layer getting created should be added at root if there's no parent
        // and has permission ACCESS_SURFACE_FLINGER. If set to false and no parent, the layer will
        // be added offscreen.
        bool addToRoot;
    };

    // A temporay pool that store the created layers and will be added to current state in main
    // thread.
    std::unordered_map<BBinder*, std::unique_ptr<LayerCreatedState>> mCreatedLayers;
    void setLayerCreatedState(const sp<IBinder>& handle, const wp<Layer>& layer,
                              const wp<IBinder>& parent, const wp<Layer> parentLayer,
                              const wp<IBinder>& producer, bool addToRoot);
    auto getLayerCreatedState(const sp<IBinder>& handle);
    sp<Layer> handleLayerCreatedLocked(const sp<IBinder>& handle) REQUIRES(mStateLock);

    std::atomic<ui::Transform::RotationFlags> mActiveDisplayTransformHint;

    bool isRefreshRateOverlayEnabled() const REQUIRES(mStateLock) {
        return std::any_of(mDisplays.begin(), mDisplays.end(),
                           [](std::pair<wp<IBinder>, sp<DisplayDevice>> display) {
                               return display.second->isRefreshRateOverlayEnabled();
                           });
    }

    wp<IBinder> mActiveDisplayToken GUARDED_BY(mStateLock);

    const sp<WindowInfosListenerInvoker> mWindowInfosListenerInvoker;

    std::unique_ptr<FlagManager> mFlagManager;
};

} // namespace android