fix [2068105] implement queueBuffer/lockBuffer/dequeueBuffer properly
Rewrote SurfaceFlinger's buffer management from the ground-up.
The design now support an arbitrary number of buffers per surface, however the current implementation is limited to four. Currently only 2 buffers are used in practice.
The main new feature is to be able to dequeue all buffers at once (very important when there are only two).
A client can dequeue all buffers until there are none available, it can lock all buffers except the last one that is used for composition. The client will block then, until a new buffer is enqueued.
The current implementation requires that buffers are locked in the same order they are dequeued and enqueued in the same order they are locked. Only one buffer can be locked at a time.
eg. Allowed sequence: DQ, DQ, LOCK, Q, LOCK, Q
eg. Forbidden sequence: DQ, DQ, LOCK, LOCK, Q, Q
diff --git a/include/ui/Surface.h b/include/ui/Surface.h
index 30ab82f..118fb83 100644
--- a/include/ui/Surface.h
+++ b/include/ui/Surface.h
@@ -39,8 +39,8 @@
class Rect;
class Surface;
class SurfaceComposerClient;
-struct per_client_cblk_t;
-struct layer_cblk_t;
+class SharedClient;
+class SharedBufferClient;
// ---------------------------------------------------------------------------
@@ -109,7 +109,7 @@
~SurfaceControl();
- status_t validate(per_client_cblk_t const* cblk) const;
+ status_t validate(SharedClient const* cblk) const;
void destroy();
sp<SurfaceComposerClient> mClient;
@@ -190,8 +190,7 @@
status_t getBufferLocked(int index, int usage);
- status_t validate(per_client_cblk_t const* cblk) const;
- static void _send_dirty_region(layer_cblk_t* lcblk, const Region& dirty);
+ status_t validate(SharedClient const* cblk) const;
inline const BufferMapper& getBufferMapper() const { return mBufferMapper; }
inline BufferMapper& getBufferMapper() { return mBufferMapper; }
@@ -210,11 +209,10 @@
int perform(int operation, va_list args);
status_t dequeueBuffer(sp<SurfaceBuffer>* buffer);
- status_t lockBuffer(const sp<SurfaceBuffer>& buffer);
- status_t queueBuffer(const sp<SurfaceBuffer>& buffer);
void setUsage(uint32_t reqUsage);
+ bool getUsage(uint32_t* usage);
// constants
sp<SurfaceComposerClient> mClient;
@@ -224,21 +222,23 @@
PixelFormat mFormat;
uint32_t mFlags;
BufferMapper& mBufferMapper;
+ SharedBufferClient* mSharedBufferClient;
// protected by mSurfaceLock
Rect mSwapRectangle;
uint32_t mUsage;
- bool mUsageChanged;
+ int32_t mUsageChanged;
// protected by mSurfaceLock. These are also used from lock/unlock
// but in that case, they must be called form the same thread.
sp<SurfaceBuffer> mBuffers[2];
mutable Region mDirtyRegion;
- mutable uint8_t mBackbufferIndex;
// must be used from the lock/unlock thread
sp<SurfaceBuffer> mLockedBuffer;
+ sp<SurfaceBuffer> mPostedBuffer;
mutable Region mOldDirtyRegion;
+ bool mNeedFullUpdate;
// query() must be called from dequeueBuffer() thread
uint32_t mWidth;
@@ -246,6 +246,7 @@
// Inherently thread-safe
mutable Mutex mSurfaceLock;
+ mutable Mutex mApiLock;
};
}; // namespace android