merge various SF fixes from gingerbread to honeycomb-mr2 (DO NOT MERGE)

Fix a race that could cause GL commands to be executed from the wrong thread.
  RefBase subclasses can now decide how they want to be destroyed.
  Fix a race in SurfaceFlinger that could cause layers to be leaked forever.
  Fix a race-condtion in SurfaceFlinger that could lead to a crash.

initial cherry-pick:
resolved conflicts for merge of b9783b49 to honeycomb-plus-aosp

Change-Id: I2a335e03fff219e35c18a7b0089b3a11d636576f

include/utils/RefBase.h

@@ -121,6 +121,13 @@ protected:
                             RefBase();
     virtual                 ~RefBase();
 
+    // called when the last reference goes away. this is responsible for
+    // calling the destructor. The default implementation just does
+    // "delete this;".
+    // Make sure to never acquire a strong reference from this function. The
+    // same restrictions than for destructors apply.
+    virtual void            destroy() const;
+
     //! Flags for extendObjectLifetime()
     enum {
         OBJECT_LIFETIME_WEAK    = 0x0001,

libs/utils/RefBase.cpp

@@ -345,6 +345,10 @@ void RefBase::incStrong(const void* id) const
     const_cast<RefBase*>(this)->onFirstRef();
 }
 
+void RefBase::destroy() const {
+    delete this;
+}
+
 void RefBase::decStrong(const void* id) const
 {
     weakref_impl* const refs = mRefs;
@@ -357,7 +361,7 @@ void RefBase::decStrong(const void* id) const
     if (c == 1) {
         const_cast<RefBase*>(this)->onLastStrongRef(id);
         if ((refs->mFlags&OBJECT_LIFETIME_WEAK) != OBJECT_LIFETIME_WEAK) {
-            delete this;
+            destroy();
         }
     }
     refs->decWeak(id);
@@ -415,7 +419,8 @@ void RefBase::weakref_type::decWeak(const void* id)
     
     if ((impl->mFlags&OBJECT_LIFETIME_WEAK) != OBJECT_LIFETIME_WEAK) {
         if (impl->mStrong == INITIAL_STRONG_VALUE)
-            delete impl->mBase;
+            if (impl->mBase)
+                impl->mBase->destroy();
         else {
             // LOGV("Freeing refs %p of old RefBase %p\n", this, impl->mBase);
             delete impl;
@@ -423,7 +428,8 @@ void RefBase::weakref_type::decWeak(const void* id)
     } else {
         impl->mBase->onLastWeakRef(id);
         if ((impl->mFlags&OBJECT_LIFETIME_FOREVER) != OBJECT_LIFETIME_FOREVER) {
-            delete impl->mBase;
+            if (impl->mBase)
+                impl->mBase->destroy();
         }
     }
 }

services/surfaceflinger/Layer.cpp

@@ -77,6 +77,10 @@ Layer::~Layer()
     }
 }
 
+void Layer::destroy() const {
+    mFlinger->destroyLayer(this);
+}
+
 status_t Layer::setToken(const sp<UserClient>& userClient,
         SharedClient* sharedClient, int32_t token)
 {
@@ -145,18 +149,6 @@ sp<LayerBaseClient::Surface> Layer::createSurface() const
     return sur;
 }
 
-status_t Layer::ditch()
-{
-    // NOTE: Called from the main UI thread
-
-    // the layer is not on screen anymore. free as much resources as possible
-    mFreezeLock.clear();
-
-    Mutex::Autolock _l(mLock);
-    mWidth = mHeight = 0;
-    return NO_ERROR;
-}
-
 status_t Layer::setBuffers( uint32_t w, uint32_t h,
                             PixelFormat format, uint32_t flags)
 {

services/surfaceflinger/Layer.h

@@ -82,7 +82,6 @@ public:
     virtual bool isSecure() const           { return mSecure; }
     virtual bool isProtected() const;
     virtual sp<Surface> createSurface() const;
-    virtual status_t ditch();
     virtual void onRemoved();
 
     // only for debugging
@@ -93,6 +92,7 @@ public:
         return mFreezeLock; }
 
 protected:
+    virtual void destroy() const;
     virtual void dump(String8& result, char* scratch, size_t size) const;
 
 private:

services/surfaceflinger/LayerBase.cpp

@@ -616,10 +616,7 @@ LayerBaseClient::Surface::~Surface()
      */
 
     // destroy client resources
-    sp<LayerBaseClient> layer = getOwner();
+    mFlinger->destroySurface(mOwner);
-    if (layer != 0) {
-        mFlinger->destroySurface(layer);
-    }
 }
 
 sp<LayerBaseClient> LayerBaseClient::Surface::getOwner() const {

services/surfaceflinger/LayerBase.h

@@ -202,10 +202,6 @@ public:
      */
     virtual bool isProtected() const   { return false; }
 
-    /** Called from the main thread, when the surface is removed from the
-     * draw list */
-    virtual status_t ditch() { return NO_ERROR; }
-
     /** called with the state lock when the surface is removed from the
      *  current list */
     virtual void onRemoved() { };
@@ -271,7 +267,8 @@ protected:
     volatile    int32_t         mInvalidate;
                 
 
-protected:
+public:
+    // called from class SurfaceFlinger
     virtual ~LayerBase();
 
 private:

services/surfaceflinger/SurfaceFlinger.cpp

@@ -387,6 +387,9 @@ bool SurfaceFlinger::threadLoop()
 {
     waitForEvent();
 
+    // call Layer's destructor
+    handleDestroyLayers();
+
     // check for transactions
     if (UNLIKELY(mConsoleSignals)) {
         handleConsoleEvents();
@@ -395,7 +398,7 @@ bool SurfaceFlinger::threadLoop()
     if (LIKELY(mTransactionCount == 0)) {
         // if we're in a global transaction, don't do anything.
         const uint32_t mask = eTransactionNeeded | eTraversalNeeded;
-        uint32_t transactionFlags = getTransactionFlags(mask);
+        uint32_t transactionFlags = peekTransactionFlags(mask);
         if (LIKELY(transactionFlags)) {
             handleTransaction(transactionFlags);
         }
@@ -480,39 +483,26 @@ void SurfaceFlinger::handleConsoleEvents()
 
 void SurfaceFlinger::handleTransaction(uint32_t transactionFlags)
 {
-    Vector< sp<LayerBase> > ditchedLayers;
-
-    /*
-     * Perform and commit the transaction
-     */
-
-    { // scope for the lock
     Mutex::Autolock _l(mStateLock);
     const nsecs_t now = systemTime();
     mDebugInTransaction = now;
-        handleTransactionLocked(transactionFlags, ditchedLayers);
+
+    // Here we're guaranteed that some transaction flags are set
+    // so we can call handleTransactionLocked() unconditionally.
+    // We call getTransactionFlags(), which will also clear the flags,
+    // with mStateLock held to guarantee that mCurrentState won't change
+    // until the transaction is committed.
+
+    const uint32_t mask = eTransactionNeeded | eTraversalNeeded;
+    transactionFlags = getTransactionFlags(mask);
+    handleTransactionLocked(transactionFlags);
+
     mLastTransactionTime = systemTime() - now;
     mDebugInTransaction = 0;
-        invalidateHwcGeometry();
     // here the transaction has been committed
 }
 
-    /*
+void SurfaceFlinger::handleTransactionLocked(uint32_t transactionFlags)
-     * Clean-up all layers that went away
-     * (do this without the lock held)
-     */
-
-    const size_t count = ditchedLayers.size();
-    for (size_t i=0 ; i<count ; i++) {
-        if (ditchedLayers[i] != 0) {
-            //LOGD("ditching layer %p", ditchedLayers[i].get());
-            ditchedLayers[i]->ditch();
-        }
-    }
-}
-
-void SurfaceFlinger::handleTransactionLocked(
-        uint32_t transactionFlags, Vector< sp<LayerBase> >& ditchedLayers)
 {
     const LayerVector& currentLayers(mCurrentState.layersSortedByZ);
     const size_t count = currentLayers.size();
@@ -584,7 +574,6 @@ void SurfaceFlinger::handleTransactionLocked(
                 const sp<LayerBase>& layer(previousLayers[i]);
                 if (currentLayers.indexOf( layer ) < 0) {
                     // this layer is not visible anymore
-                    ditchedLayers.add(layer);
                     mDirtyRegionRemovedLayer.orSelf(layer->visibleRegionScreen);
                 }
             }
@@ -594,6 +583,31 @@ void SurfaceFlinger::handleTransactionLocked(
     commitTransaction();
 }
 
+void SurfaceFlinger::destroyLayer(LayerBase const* layer)
+{
+    Mutex::Autolock _l(mDestroyedLayerLock);
+    mDestroyedLayers.add(layer);
+    signalEvent();
+}
+
+void SurfaceFlinger::handleDestroyLayers()
+{
+    Vector<LayerBase const *> destroyedLayers;
+
+    { // scope for the lock
+        Mutex::Autolock _l(mDestroyedLayerLock);
+        destroyedLayers = mDestroyedLayers;
+        mDestroyedLayers.clear();
+    }
+
+    // call destructors without a lock held
+    const size_t count = destroyedLayers.size();
+    for (size_t i=0 ; i<count ; i++) {
+        //LOGD("destroying %s", destroyedLayers[i]->getName().string());
+        delete destroyedLayers[i];
+    }
+}
+
 sp<FreezeLock> SurfaceFlinger::getFreezeLock() const
 {
     return new FreezeLock(const_cast<SurfaceFlinger *>(this));
@@ -1096,15 +1110,15 @@ status_t SurfaceFlinger::addLayer_l(const sp<LayerBase>& layer)
 ssize_t SurfaceFlinger::addClientLayer(const sp<Client>& client,
         const sp<LayerBaseClient>& lbc)
 {
-    Mutex::Autolock _l(mStateLock);
-
     // attach this layer to the client
-    ssize_t name = client->attachLayer(lbc);
+    size_t name = client->attachLayer(lbc);
+
+    Mutex::Autolock _l(mStateLock);
 
     // add this layer to the current state list
     addLayer_l(lbc);
 
-    return name;
+    return ssize_t(name);
 }
 
 status_t SurfaceFlinger::removeLayer(const sp<LayerBase>& layer)
@@ -1155,6 +1169,11 @@ status_t SurfaceFlinger::invalidateLayerVisibility(const sp<LayerBase>& layer)
     return NO_ERROR;
 }
 
+uint32_t SurfaceFlinger::peekTransactionFlags(uint32_t flags)
+{
+    return android_atomic_release_load(&mTransactionFlags);
+}
+
 uint32_t SurfaceFlinger::getTransactionFlags(uint32_t flags)
 {
     return android_atomic_and(~flags, &mTransactionFlags) & flags;
@@ -1362,51 +1381,18 @@ status_t SurfaceFlinger::removeSurface(const sp<Client>& client, SurfaceID sid)
     return err;
 }
 
-status_t SurfaceFlinger::destroySurface(const sp<LayerBaseClient>& layer)
+status_t SurfaceFlinger::destroySurface(const wp<LayerBaseClient>& layer)
 {
     // called by ~ISurface() when all references are gone
-
+    status_t err = NO_ERROR;
-    class MessageDestroySurface : public MessageBase {
+    sp<LayerBaseClient> l(layer.promote());
-        SurfaceFlinger* flinger;
+    if (l != NULL) {
-        sp<LayerBaseClient> layer;
+        Mutex::Autolock _l(mStateLock);
-    public:
+        err = removeLayer_l(l);
-        MessageDestroySurface(
-                SurfaceFlinger* flinger, const sp<LayerBaseClient>& layer)
-            : flinger(flinger), layer(layer) { }
-        virtual bool handler() {
-            sp<LayerBaseClient> l(layer);
-            layer.clear(); // clear it outside of the lock;
-            Mutex::Autolock _l(flinger->mStateLock);
-            /*
-             * remove the layer from the current list -- chances are that it's
-             * not in the list anyway, because it should have been removed
-             * already upon request of the client (eg: window manager).
-             * However, a buggy client could have not done that.
-             * Since we know we don't have any more clients, we don't need
-             * to use the purgatory.
-             */
-            status_t err = flinger->removeLayer_l(l);
-            if (err == NAME_NOT_FOUND) {
-                // The surface wasn't in the current list, which means it was
-                // removed already, which means it is in the purgatory,
-                // and need to be removed from there.
-                // This needs to happen from the main thread since its dtor
-                // must run from there (b/c of OpenGL ES). Additionally, we
-                // can't really acquire our internal lock from
-                // destroySurface() -- see postMessage() below.
-                ssize_t idx = flinger->mLayerPurgatory.remove(l);
-                LOGE_IF(idx < 0,
-                        "layer=%p is not in the purgatory list", l.get());
-            }
-
         LOGE_IF(err<0 && err != NAME_NOT_FOUND,
                 "error removing layer=%p (%s)", l.get(), strerror(-err));
-            return true;
     }
-    };
+    return err;
-
-    postMessageAsync( new MessageDestroySurface(this, layer) );
-    return NO_ERROR;
 }
 
 status_t SurfaceFlinger::setClientState(
@@ -2381,15 +2367,17 @@ status_t Client::initCheck() const {
     return NO_ERROR;
 }
 
-ssize_t Client::attachLayer(const sp<LayerBaseClient>& layer)
+size_t Client::attachLayer(const sp<LayerBaseClient>& layer)
 {
-    int32_t name = android_atomic_inc(&mNameGenerator);
+    Mutex::Autolock _l(mLock);
+    size_t name = mNameGenerator++;
     mLayers.add(name, layer);
     return name;
 }
 
 void Client::detachLayer(const LayerBaseClient* layer)
 {
+    Mutex::Autolock _l(mLock);
     // we do a linear search here, because this doesn't happen often
     const size_t count = mLayers.size();
     for (size_t i=0 ; i<count ; i++) {
@@ -2399,9 +2387,11 @@ void Client::detachLayer(const LayerBaseClient* layer)
         }
     }
 }
-sp<LayerBaseClient> Client::getLayerUser(int32_t i) const {
+sp<LayerBaseClient> Client::getLayerUser(int32_t i) const
+{
+    Mutex::Autolock _l(mLock);
     sp<LayerBaseClient> lbc;
-    const wp<LayerBaseClient>& layer(mLayers.valueFor(i));
+    wp<LayerBaseClient> layer(mLayers.valueFor(i));
     if (layer != 0) {
         lbc = layer.promote();
         LOGE_IF(lbc==0, "getLayerUser(name=%d) is dead", int(i));

services/surfaceflinger/SurfaceFlinger.h

@@ -65,7 +65,7 @@ public:
     status_t initCheck() const;
 
     // protected by SurfaceFlinger::mStateLock
-    ssize_t attachLayer(const sp<LayerBaseClient>& layer);
+    size_t attachLayer(const sp<LayerBaseClient>& layer);
     void detachLayer(const LayerBaseClient* layer);
     sp<LayerBaseClient> getLayerUser(int32_t i) const;
 
@@ -81,9 +81,15 @@ private:
     virtual status_t destroySurface(SurfaceID surfaceId);
     virtual status_t setState(int32_t count, const layer_state_t* states);
 
-    DefaultKeyedVector< size_t, wp<LayerBaseClient> > mLayers;
+    // constant
     sp<SurfaceFlinger> mFlinger;
-    int32_t mNameGenerator;
+
+    // protected by mLock
+    DefaultKeyedVector< size_t, wp<LayerBaseClient> > mLayers;
+    size_t mNameGenerator;
+
+    // thread-safe
+    mutable Mutex mLock;
 };
 
 class UserClient : public BnSurfaceComposerClient
@@ -227,6 +233,7 @@ public:
     status_t addLayer(const sp<LayerBase>& layer);
     status_t invalidateLayerVisibility(const sp<LayerBase>& layer);
     void invalidateHwcGeometry();
+    void destroyLayer(LayerBase const* layer);
 
     sp<Layer> getLayer(const sp<ISurface>& sur) const;
 
@@ -254,7 +261,7 @@ private:
             uint32_t w, uint32_t h, uint32_t flags);
 
     status_t removeSurface(const sp<Client>& client, SurfaceID sid);
-    status_t destroySurface(const sp<LayerBaseClient>& layer);
+    status_t destroySurface(const wp<LayerBaseClient>& layer);
     status_t setClientState(const sp<Client>& client,
             int32_t count, const layer_state_t* states);
 
@@ -299,9 +306,8 @@ public:     // hack to work around gcc 4.0.3 bug
 private:
             void        handleConsoleEvents();
             void        handleTransaction(uint32_t transactionFlags);
-            void        handleTransactionLocked(
+            void        handleTransactionLocked(uint32_t transactionFlags);
-                            uint32_t transactionFlags, 
+            void        handleDestroyLayers();
-                            Vector< sp<LayerBase> >& ditchedLayers);
 
             void        computeVisibleRegions(
                             LayerVector& currentLayers,
@@ -324,6 +330,7 @@ private:
             status_t    purgatorizeLayer_l(const sp<LayerBase>& layer);
 
             uint32_t    getTransactionFlags(uint32_t flags);
+            uint32_t    peekTransactionFlags(uint32_t flags);
             uint32_t    setTransactionFlags(uint32_t flags);
             void        commitTransaction();
 
@@ -422,6 +429,11 @@ private:
                 // these are thread safe
     mutable     Barrier                     mReadyToRunBarrier;
 
+
+                // protected by mDestroyedLayerLock;
+    mutable     Mutex                       mDestroyedLayerLock;
+                Vector<LayerBase const *>   mDestroyedLayers;
+
                 // atomic variables
                 enum {
                     eConsoleReleased = 1,