blob: 56f508b60b968f584bb1236aca3127b1c5a4c18f [file] [log] [blame]
* 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
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* See the License for the specific language governing permissions and
* limitations under the License.
#include <stdint.h>
#include <sys/types.h>
#include <EGL/egl.h>
#include <EGL/eglext.h>
#include <utils/RefBase.h>
#include <utils/String8.h>
#include <utils/Timers.h>
#include <ui/FloatRect.h>
#include <ui/FrameStats.h>
#include <ui/GraphicBuffer.h>
#include <ui/PixelFormat.h>
#include <ui/Region.h>
#include <gui/ISurfaceComposerClient.h>
#include <private/gui/LayerState.h>
#include <list>
#include "FrameTracker.h"
#include "Client.h"
#include "LayerVector.h"
#include "MonitoredProducer.h"
#include "SurfaceFlinger.h"
#include "SurfaceFlingerConsumer.h"
#include "Transform.h"
#include "DisplayHardware/HWComposer.h"
#include "DisplayHardware/HWComposerBufferCache.h"
#include "RenderEngine/Mesh.h"
#include "RenderEngine/Texture.h"
namespace android {
// ---------------------------------------------------------------------------
class Client;
class Colorizer;
class DisplayDevice;
class GraphicBuffer;
class SurfaceFlinger;
class LayerDebugInfo;
// ---------------------------------------------------------------------------
* A new BufferQueue and a new SurfaceFlingerConsumer are created when the
* Layer is first referenced.
* This also implements onFrameAvailable(), which notifies SurfaceFlinger
* that new data has arrived.
class Layer : public SurfaceFlingerConsumer::ContentsChangedListener {
static int32_t sSequence;
mutable bool contentDirty;
// regions below are in window-manager space
Region visibleRegion;
Region coveredRegion;
Region visibleNonTransparentRegion;
Region surfaceDamageRegion;
// Layer serial number. This gives layers an explicit ordering, so we
// have a stable sort order when their layer stack and Z-order are
// the same.
int32_t sequence;
enum { // flags for doTransaction()
eDontUpdateGeometryState = 0x00000001,
eVisibleRegion = 0x00000002,
struct Geometry {
uint32_t w;
uint32_t h;
Transform transform;
inline bool operator ==(const Geometry& rhs) const {
return (w == rhs.w && h == rhs.h) &&
(transform.tx() == rhs.transform.tx()) &&
(transform.ty() == rhs.transform.ty());
inline bool operator !=(const Geometry& rhs) const {
return !operator ==(rhs);
struct State {
Geometry active;
Geometry requested;
int32_t z;
// The identifier of the layer stack this layer belongs to. A layer can
// only be associated to a single layer stack. A layer stack is a
// z-ordered group of layers which can be associated to one or more
// displays. Using the same layer stack on different displays is a way
// to achieve mirroring.
uint32_t layerStack;
#ifdef USE_HWC2
float alpha;
uint8_t alpha;
uint8_t flags;
uint8_t mask;
uint8_t reserved[2];
int32_t sequence; // changes when visible regions can change
bool modified;
// Crop is expressed in layer space coordinate.
Rect crop;
Rect requestedCrop;
// finalCrop is expressed in display space coordinate.
Rect finalCrop;
Rect requestedFinalCrop;
// If set, defers this state update until the identified Layer
// receives a frame with the given frameNumber
wp<Layer> barrierLayer;
uint64_t frameNumber;
// the transparentRegion hint is a bit special, it's latched only
// when we receive a buffer -- this is because it's "content"
// dependent.
Region activeTransparentRegion;
Region requestedTransparentRegion;
android_dataspace dataSpace;
uint32_t appId;
uint32_t type;
// If non-null, a Surface this Surface's Z-order is interpreted relative to.
wp<Layer> zOrderRelativeOf;
// A list of surfaces whose Z-order is interpreted relative to ours.
SortedVector<wp<Layer>> zOrderRelatives;
// -----------------------------------------------------------------------
Layer(SurfaceFlinger* flinger, const sp<Client>& client,
const String8& name, uint32_t w, uint32_t h, uint32_t flags);
virtual ~Layer();
void setPrimaryDisplayOnly() { mPrimaryDisplayOnly = true; }
// the this layer's size and format
status_t setBuffers(uint32_t w, uint32_t h, PixelFormat format, uint32_t flags);
// ------------------------------------------------------------------------
// Geometry setting functions.
// The following group of functions are used to specify the layers
// bounds, and the mapping of the texture on to those bounds. According
// to various settings changes to them may apply immediately, or be delayed until
// a pending resize is completed by the producer submitting a buffer. For example
// if we were to change the buffer size, and update the matrix ahead of the
// new buffer arriving, then we would be stretching the buffer to a different
// aspect before and after the buffer arriving, which probably isn't what we wanted.
// The first set of geometry functions are controlled by the scaling mode, described
// in window.h. The scaling mode may be set by the client, as it submits buffers.
// This value may be overriden through SurfaceControl, with setOverrideScalingMode.
// Put simply, if our scaling mode is SCALING_MODE_FREEZE, then
// matrix updates will not be applied while a resize is pending
// and the size and transform will remain in their previous state
// until a new buffer is submitted. If the scaling mode is another value
// then the old-buffer will immediately be scaled to the pending size
// and the new matrix will be immediately applied following this scaling
// transformation.
// Set the default buffer size for the assosciated Producer, in pixels. This is
// also the rendered size of the layer prior to any transformations. Parent
// or local matrix transformations will not affect the size of the buffer,
// but may affect it's on-screen size or clipping.
bool setSize(uint32_t w, uint32_t h);
// Set a 2x2 transformation matrix on the layer. This transform
// will be applied after parent transforms, but before any final
// producer specified transform.
bool setMatrix(const layer_state_t::matrix22_t& matrix);
// This second set of geometry attributes are controlled by
// setGeometryAppliesWithResize, and their default mode is to be
// immediate. If setGeometryAppliesWithResize is specified
// while a resize is pending, then update of these attributes will
// be delayed until the resize completes.
// setPosition operates in parent buffer space (pre parent-transform) or display
// space for top-level layers.
bool setPosition(float x, float y, bool immediate);
// Buffer space
bool setCrop(const Rect& crop, bool immediate);
// Parent buffer space/display space
bool setFinalCrop(const Rect& crop, bool immediate);
// TODO(b/38182121): Could we eliminate the various latching modes by
// using the layer hierarchy?
// -----------------------------------------------------------------------
bool setLayer(int32_t z);
bool setRelativeLayer(const sp<IBinder>& relativeToHandle, int32_t relativeZ);
#ifdef USE_HWC2
bool setAlpha(float alpha);
bool setAlpha(uint8_t alpha);
bool setTransparentRegionHint(const Region& transparent);
bool setFlags(uint8_t flags, uint8_t mask);
bool setLayerStack(uint32_t layerStack);
bool setDataSpace(android_dataspace dataSpace);
android_dataspace getDataSpace() const;
uint32_t getLayerStack() const;
void deferTransactionUntil(const sp<IBinder>& barrierHandle, uint64_t frameNumber);
void deferTransactionUntil(const sp<Layer>& barrierLayer, uint64_t frameNumber);
bool setOverrideScalingMode(int32_t overrideScalingMode);
void setInfo(uint32_t type, uint32_t appId);
bool reparentChildren(const sp<IBinder>& layer);
bool detachChildren();
// If we have received a new buffer this frame, we will pass its surface
// damage down to hardware composer. Otherwise, we must send a region with
// one empty rect.
void useSurfaceDamage();
void useEmptyDamage();
uint32_t getTransactionFlags(uint32_t flags);
uint32_t setTransactionFlags(uint32_t flags);
bool belongsToDisplay(uint32_t layerStack, bool isPrimaryDisplay) const {
return getLayerStack() == layerStack && (!mPrimaryDisplayOnly || isPrimaryDisplay);
void computeGeometry(const sp<const DisplayDevice>& hw, Mesh& mesh,
bool useIdentityTransform) const;
FloatRect computeBounds(const Region& activeTransparentRegion) const;
FloatRect computeBounds() const;
int32_t getSequence() const { return sequence; }
// -----------------------------------------------------------------------
// Virtuals
virtual const char* getTypeId() const { return "Layer"; }
* isOpaque - true if this surface is opaque
* This takes into account the buffer format (i.e. whether or not the
* pixel format includes an alpha channel) and the "opaque" flag set
* on the layer. It does not examine the current plane alpha value.
virtual bool isOpaque(const Layer::State& s) const;
* isSecure - true if this surface is secure, that is if it prevents
* screenshots or VNC servers.
virtual bool isSecure() const;
* isProtected - true if the layer may contain protected content in the
virtual bool isProtected() const;
* isVisible - true if this layer is visible, false otherwise
virtual bool isVisible() const;
* isHiddenByPolicy - true if this layer has been forced invisible.
* just because this is false, doesn't mean isVisible() is true.
* For example if this layer has no active buffer, it may not be hidden by
* policy, but it still can not be visible.
virtual bool isHiddenByPolicy() const;
* isFixedSize - true if content has a fixed size
virtual bool isFixedSize() const;
* onDraw - draws the surface.
virtual void onDraw(const sp<const DisplayDevice>& hw, const Region& clip,
bool useIdentityTransform) const;
// -----------------------------------------------------------------------
#ifdef USE_HWC2
void setGeometry(const sp<const DisplayDevice>& displayDevice, uint32_t z);
void forceClientComposition(int32_t hwcId);
void setPerFrameData(const sp<const DisplayDevice>& displayDevice);
// callIntoHwc exists so we can update our local state and call
// acceptDisplayChanges without unnecessarily updating the device's state
void setCompositionType(int32_t hwcId, HWC2::Composition type,
bool callIntoHwc = true);
HWC2::Composition getCompositionType(int32_t hwcId) const;
void setClearClientTarget(int32_t hwcId, bool clear);
bool getClearClientTarget(int32_t hwcId) const;
void updateCursorPosition(const sp<const DisplayDevice>& hw);
void setGeometry(const sp<const DisplayDevice>& hw,
HWComposer::HWCLayerInterface& layer);
void setPerFrameData(const sp<const DisplayDevice>& hw,
HWComposer::HWCLayerInterface& layer);
void setAcquireFence(const sp<const DisplayDevice>& hw,
HWComposer::HWCLayerInterface& layer);
Rect getPosition(const sp<const DisplayDevice>& hw);
* called after page-flip
#ifdef USE_HWC2
void onLayerDisplayed(const sp<Fence>& releaseFence);
void onLayerDisplayed(const sp<const DisplayDevice>& hw,
HWComposer::HWCLayerInterface* layer);
bool shouldPresentNow(const DispSync& dispSync) const;
* called before composition.
* returns true if the layer has pending updates.
bool onPreComposition(nsecs_t refreshStartTime);
* called after composition.
* returns true if the layer latched a new buffer this frame.
bool onPostComposition(const std::shared_ptr<FenceTime>& glDoneFence,
const std::shared_ptr<FenceTime>& presentFence,
const CompositorTiming& compositorTiming);
#ifdef USE_HWC2
// If a buffer was replaced this frame, release the former buffer
void releasePendingBuffer(nsecs_t dequeueReadyTime);
* draw - performs some global clipping optimizations
* and calls onDraw().
void draw(const sp<const DisplayDevice>& hw, const Region& clip) const;
void draw(const sp<const DisplayDevice>& hw, bool useIdentityTransform) const;
void draw(const sp<const DisplayDevice>& hw) const;
* doTransaction - process the transaction. This is a good place to figure
* out which attributes of the surface have changed.
uint32_t doTransaction(uint32_t transactionFlags);
* setVisibleRegion - called to set the new visible region. This gives
* a chance to update the new visible region or record the fact it changed.
void setVisibleRegion(const Region& visibleRegion);
* setCoveredRegion - called when the covered region changes. The covered
* region corresponds to any area of the surface that is covered
* (transparently or not) by another surface.
void setCoveredRegion(const Region& coveredRegion);
* setVisibleNonTransparentRegion - called when the visible and
* non-transparent region changes.
void setVisibleNonTransparentRegion(const Region&
* latchBuffer - called each time the screen is redrawn and returns whether
* the visible regions need to be recomputed (this is a fairly heavy
* operation, so this should be set only if needed). Typically this is used
* to figure out if the content or size of a surface has changed.
Region latchBuffer(bool& recomputeVisibleRegions, nsecs_t latchTime);
bool isBufferLatched() const { return mRefreshPending; }
bool isPotentialCursor() const { return mPotentialCursor;}
* called with the state lock from a binder thread when the layer is
* removed from the current list to the pending removal list
void onRemovedFromCurrentState();
* called with the state lock from the main thread when the layer is
* removed from the pending removal list
void onRemoved();
// Updates the transform hint in our SurfaceFlingerConsumer to match
// the current orientation of the display device.
void updateTransformHint(const sp<const DisplayDevice>& hw) const;
* returns the rectangle that crops the content of the layer and scales it
* to the layer's size.
Rect getContentCrop() const;
* Returns if a frame is queued.
bool hasQueuedFrame() const { return mQueuedFrames > 0 ||
mSidebandStreamChanged || mAutoRefresh; }
int32_t getQueuedFrameCount() const { return mQueuedFrames; }
#ifdef USE_HWC2
// -----------------------------------------------------------------------
bool createHwcLayer(HWComposer* hwc, int32_t hwcId);
bool destroyHwcLayer(int32_t hwcId);
void destroyAllHwcLayers();
bool hasHwcLayer(int32_t hwcId) {
return mHwcLayers.count(hwcId) > 0;
HWC2::Layer* getHwcLayer(int32_t hwcId) {
if (mHwcLayers.count(hwcId) == 0) {
return nullptr;
return mHwcLayers[hwcId].layer;
// -----------------------------------------------------------------------
void clearWithOpenGL(const sp<const DisplayDevice>& hw) const;
void setFiltering(bool filtering);
bool getFiltering() const;
// only for debugging
inline const sp<GraphicBuffer>& getActiveBuffer() const { return mActiveBuffer; }
inline const State& getDrawingState() const { return mDrawingState; }
inline const State& getCurrentState() const { return mCurrentState; }
inline State& getCurrentState() { return mCurrentState; }
LayerDebugInfo getLayerDebugInfo() const;
/* always call base class first */
#ifdef USE_HWC2
static void miniDumpHeader(String8& result);
void miniDump(String8& result, int32_t hwcId) const;
void dumpFrameStats(String8& result) const;
void dumpFrameEvents(String8& result);
void clearFrameStats();
void logFrameStats();
void getFrameStats(FrameStats* outStats) const;
std::vector<OccupancyTracker::Segment> getOccupancyHistory(bool forceFlush);
void onDisconnect();
void addAndGetFrameTimestamps(const NewFrameEventsEntry* newEntry,
FrameEventHistoryDelta* outDelta);
bool getTransformToDisplayInverse() const;
Transform getTransform() const;
// Returns the Alpha of the Surface, accounting for the Alpha
// of parent Surfaces in the hierarchy (alpha's will be multiplied
// down the hierarchy).
#ifdef USE_HWC2
float getAlpha() const;
uint8_t getAlpha() const;
void traverseInReverseZOrder(LayerVector::StateSet stateSet,
const LayerVector::Visitor& visitor);
void traverseInZOrder(LayerVector::StateSet stateSet, const LayerVector::Visitor& visitor);
size_t getChildrenCount() const;
void addChild(const sp<Layer>& layer);
// Returns index if removed, or negative value otherwise
// for symmetry with Vector::remove
ssize_t removeChild(const sp<Layer>& layer);
sp<Layer> getParent() const { return mCurrentParent.promote(); }
bool hasParent() const { return getParent() != nullptr; }
Rect computeScreenBounds(bool reduceTransparentRegion = true) const;
bool setChildLayer(const sp<Layer>& childLayer, int32_t z);
// Copy the current list of children to the drawing state. Called by
// SurfaceFlinger to complete a transaction.
void commitChildList();
int32_t getZ() const;
// constant
sp<SurfaceFlinger> mFlinger;
* Trivial class, used to ensure that mFlinger->onLayerDestroyed(mLayer)
* is called.
class LayerCleaner {
sp<SurfaceFlinger> mFlinger;
wp<Layer> mLayer;
~LayerCleaner() {
// destroy client resources
LayerCleaner(const sp<SurfaceFlinger>& flinger,
const sp<Layer>& layer)
: mFlinger(flinger), mLayer(layer) {
virtual void onFirstRef();
friend class SurfaceInterceptor;
// Interface implementation for SurfaceFlingerConsumer::ContentsChangedListener
virtual void onFrameAvailable(const BufferItem& item) override;
virtual void onFrameReplaced(const BufferItem& item) override;
virtual void onSidebandStreamChanged() override;
void commitTransaction(const State& stateToCommit);
// needsLinearFiltering - true if this surface's state requires filtering
bool needsFiltering(const sp<const DisplayDevice>& hw) const;
uint32_t getEffectiveUsage(uint32_t usage) const;
FloatRect computeCrop(const sp<const DisplayDevice>& hw) const;
// Compute the initial crop as specified by parent layers and the SurfaceControl
// for this layer. Does not include buffer crop from the IGraphicBufferProducer
// client, as that should not affect child clipping. Returns in screen space.
Rect computeInitialCrop(const sp<const DisplayDevice>& hw) const;
bool isCropped() const;
static bool getOpacityForFormat(uint32_t format);
// drawing
void clearWithOpenGL(const sp<const DisplayDevice>& hw,
float r, float g, float b, float alpha) const;
void drawWithOpenGL(const sp<const DisplayDevice>& hw,
bool useIdentityTransform) const;
// Temporary - Used only for LEGACY camera mode.
uint32_t getProducerStickyTransform() const;
// Loads the corresponding system property once per process
static bool latchUnsignaledBuffers();
void setParent(const sp<Layer>& layer);
LayerVector makeTraversalList(LayerVector::StateSet stateSet);
void addZOrderRelative(const wp<Layer>& relative);
void removeZOrderRelative(const wp<Layer>& relative);
// -----------------------------------------------------------------------
class SyncPoint
explicit SyncPoint(uint64_t frameNumber) : mFrameNumber(frameNumber),
mFrameIsAvailable(false), mTransactionIsApplied(false) {}
uint64_t getFrameNumber() const {
return mFrameNumber;
bool frameIsAvailable() const {
return mFrameIsAvailable;
void setFrameAvailable() {
mFrameIsAvailable = true;
bool transactionIsApplied() const {
return mTransactionIsApplied;
void setTransactionApplied() {
mTransactionIsApplied = true;
const uint64_t mFrameNumber;
std::atomic<bool> mFrameIsAvailable;
std::atomic<bool> mTransactionIsApplied;
// SyncPoints which will be signaled when the correct frame is at the head
// of the queue and dropped after the frame has been latched. Protected by
// mLocalSyncPointMutex.
Mutex mLocalSyncPointMutex;
std::list<std::shared_ptr<SyncPoint>> mLocalSyncPoints;
// SyncPoints which will be signaled and then dropped when the transaction
// is applied
std::list<std::shared_ptr<SyncPoint>> mRemoteSyncPoints;
uint64_t getHeadFrameNumber() const;
bool headFenceHasSignaled() const;
// Returns false if the relevant frame has already been latched
bool addSyncPoint(const std::shared_ptr<SyncPoint>& point);
void pushPendingState();
void popPendingState(State* stateToCommit);
bool applyPendingStates(State* stateToCommit);
void clearSyncPoints();
// Returns mCurrentScaling mode (originating from the
// Client) or mOverrideScalingMode mode (originating from
// the Surface Controller) if set.
uint32_t getEffectiveScalingMode() const;
* The layer handle is just a BBinder object passed to the client
* (remote process) -- we don't keep any reference on our side such that
* the dtor is called when the remote side let go of its reference.
* LayerCleaner ensures that mFlinger->onLayerDestroyed() is called for
* this layer when the handle is destroyed.
class Handle : public BBinder, public LayerCleaner {
Handle(const sp<SurfaceFlinger>& flinger, const sp<Layer>& layer)
: LayerCleaner(flinger, layer), owner(layer) {}
wp<Layer> owner;
sp<IBinder> getHandle();
sp<IGraphicBufferProducer> getProducer() const;
const String8& getName() const;
void notifyAvailableFrames();
PixelFormat getPixelFormat() const { return mFormat; }
// -----------------------------------------------------------------------
// Check all of the local sync points to ensure that all transactions
// which need to have been applied prior to the frame which is about to
// be latched have signaled
bool allTransactionsSignaled();
// constants
sp<SurfaceFlingerConsumer> mSurfaceFlingerConsumer;
sp<IGraphicBufferProducer> mProducer;
uint32_t mTextureName; // from GLES
bool mPremultipliedAlpha;
String8 mName;
String8 mTransactionName; // A cached version of "TX - " + mName for systraces
PixelFormat mFormat;
bool mPrimaryDisplayOnly = false;
// these are protected by an external lock
State mCurrentState;
State mDrawingState;
volatile int32_t mTransactionFlags;
// Accessed from main thread and binder threads
Mutex mPendingStateMutex;
Vector<State> mPendingStates;
// thread-safe
volatile int32_t mQueuedFrames;
volatile int32_t mSidebandStreamChanged; // used like an atomic boolean
// Timestamp history for UIAutomation. Thread safe.
FrameTracker mFrameTracker;
// Timestamp history for the consumer to query.
// Accessed by both consumer and producer on main and binder threads.
Mutex mFrameEventHistoryMutex;
ConsumerFrameEventHistory mFrameEventHistory;
FenceTimeline mAcquireTimeline;
FenceTimeline mReleaseTimeline;
// main thread
int mActiveBufferSlot;
sp<GraphicBuffer> mActiveBuffer;
sp<NativeHandle> mSidebandStream;
Rect mCurrentCrop;
uint32_t mCurrentTransform;
uint32_t mCurrentScalingMode;
// We encode unset as -1.
int32_t mOverrideScalingMode;
bool mCurrentOpacity;
bool mBufferLatched = false; // TODO: Use mActiveBuffer?
std::atomic<uint64_t> mCurrentFrameNumber;
uint64_t mPreviousFrameNumber; // Only accessed on the main thread.
bool mRefreshPending;
bool mFrameLatencyNeeded;
// Whether filtering is forced on or not
bool mFiltering;
// Whether filtering is needed b/c of the drawingstate
bool mNeedsFiltering;
// The mesh used to draw the layer in GLES composition mode
mutable Mesh mMesh;
// The texture used to draw the layer in GLES composition mode
mutable Texture mTexture;
#ifdef USE_HWC2
// HWC items, accessed from the main thread
struct HWCInfo {
: hwc(nullptr),
clearClientTarget(false) {}
HWComposer* hwc;
HWC2::Layer* layer;
bool forceClientComposition;
HWC2::Composition compositionType;
bool clearClientTarget;
Rect displayFrame;
FloatRect sourceCrop;
HWComposerBufferCache bufferCache;
// A layer can be attached to multiple displays when operating in mirror mode
// (a.k.a: when several displays are attached with equal layerStack). In this
// case we need to keep track. In non-mirror mode, a layer will have only one
// HWCInfo. This map key is a display layerStack.
std::unordered_map<int32_t, HWCInfo> mHwcLayers;
bool mIsGlesComposition;
// page-flip thread (currently main thread)
bool mProtectedByApp; // application requires protected path to external sink
// protected by mLock
mutable Mutex mLock;
// Set to true once we've returned this surface's handle
mutable bool mHasSurface;
const wp<Client> mClientRef;
// This layer can be a cursor on some displays.
bool mPotentialCursor;
// Local copy of the queued contents of the incoming BufferQueue
mutable Mutex mQueueItemLock;
Condition mQueueItemCondition;
Vector<BufferItem> mQueueItems;
std::atomic<uint64_t> mLastFrameNumberReceived;
bool mUpdateTexImageFailed; // This is only accessed on the main thread.
bool mAutoRefresh;
bool mFreezeGeometryUpdates;
// Child list about to be committed/used for editing.
LayerVector mCurrentChildren;
// Child list used for rendering.
LayerVector mDrawingChildren;
wp<Layer> mCurrentParent;
wp<Layer> mDrawingParent;
// ---------------------------------------------------------------------------
}; // namespace android