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

1 files changed, 35 insertions, 64 deletions

diff --git a/libs/surfaceflinger/SurfaceFlinger.cpp b/libs/surfaceflinger/SurfaceFlinger.cpp
index c78921a..b368db6 100644
--- a/libs/surfaceflinger/SurfaceFlinger.cpp
+++ b/libs/surfaceflinger/SurfaceFlinger.cpp
@@ -44,12 +44,12 @@
 #include <GLES/gl.h>
 
 #include "clz.h"
+#include "Buffer.h"
 #include "BufferAllocator.h"
 #include "Layer.h"
 #include "LayerBlur.h"
 #include "LayerBuffer.h"
 #include "LayerDim.h"
-#include "LayerBitmap.h"
 #include "SurfaceFlinger.h"
 
 #include "DisplayHardware/DisplayHardware.h"
@@ -173,12 +173,12 @@ SurfaceFlinger::SurfaceFlinger()
     :   BnSurfaceComposer(), Thread(false),
         mTransactionFlags(0),
         mTransactionCount(0),
+        mResizeTransationPending(false),
         mLayersRemoved(false),
         mBootTime(systemTime()),
         mHardwareTest("android.permission.HARDWARE_TEST"),
         mAccessSurfaceFlinger("android.permission.ACCESS_SURFACE_FLINGER"),
         mDump("android.permission.DUMP"),
-        mLastScheduledBroadcast(NULL),
         mVisibleRegionsDirty(false),
         mDeferReleaseConsole(false),
         mFreezeDisplay(false),
@@ -497,13 +497,11 @@ bool SurfaceFlinger::threadLoop()
 
         // release the clients before we flip ('cause flip might block)
         unlockClients();
-        executeScheduledBroadcasts();
 
         postFramebuffer();
     } else {
         // pretend we did the post
         unlockClients();
-        executeScheduledBroadcasts();
         usleep(16667); // 60 fps period
     }
     return true;
@@ -773,7 +771,8 @@ void SurfaceFlinger::computeVisibleRegions(
 void SurfaceFlinger::commitTransaction()
 {
     mDrawingState = mCurrentState;
-    mTransactionCV.signal();
+    mResizeTransationPending = false;
+    mTransactionCV.broadcast();
 }
 
 void SurfaceFlinger::handlePageFlip()
@@ -910,37 +909,6 @@ void SurfaceFlinger::unlockClients()
     }
 }
 
-void SurfaceFlinger::scheduleBroadcast(const sp<Client>& client)
-{
-    if (mLastScheduledBroadcast != client) {
-        mLastScheduledBroadcast = client;
-        mScheduledBroadcasts.add(client);
-    }
-}
-
-void SurfaceFlinger::executeScheduledBroadcasts()
-{
-    SortedVector< wp<Client> >& list(mScheduledBroadcasts);
-    size_t count = list.size();
-    while (count--) {
-        sp<Client> client = list[count].promote();
-        if (client != 0) {
-            per_client_cblk_t* const cblk = client->ctrlblk;
-            if (cblk->lock.tryLock() == NO_ERROR) {
-                cblk->cv.broadcast();
-                list.removeAt(count);
-                cblk->lock.unlock();
-            } else {
-                // schedule another round
-                LOGW("executeScheduledBroadcasts() skipped, "
-                        "contention on the client. We'll try again later...");
-                signalDelayedEvent(ms2ns(4));
-            }
-        }
-    }
-    mLastScheduledBroadcast = 0;
-}
-
 void SurfaceFlinger::debugFlashRegions()
 {
      const DisplayHardware& hw(graphicPlane(0).displayHardware());
@@ -1129,18 +1097,10 @@ void SurfaceFlinger::free_resources_l()
     mLayersRemoved = false;
     
     // free resources associated with disconnected clients
-    SortedVector< wp<Client> >& scheduledBroadcasts(mScheduledBroadcasts);
     Vector< sp<Client> >& disconnectedClients(mDisconnectedClients);
     const size_t count = disconnectedClients.size();
     for (size_t i=0 ; i<count ; i++) {
         sp<Client> client = disconnectedClients[i];
-        // if this client is the scheduled broadcast list,
-        // remove it from there (and we don't need to signal it
-        // since it is dead).
-        int32_t index = scheduledBroadcasts.indexOf(client);
-        if (index >= 0) {
-            scheduledBroadcasts.removeItemsAt(index);
-        }
         mTokens.release(client->cid);
     }
     disconnectedClients.clear();
@@ -1173,6 +1133,13 @@ void SurfaceFlinger::closeGlobalTransaction()
 {
     if (android_atomic_dec(&mTransactionCount) == 1) {
         signalEvent();
+
+        // if there is a transaction with a resize, wait for it to 
+        // take effect before returning.
+        Mutex::Autolock _l(mStateLock);
+        while (mResizeTransationPending) {
+            mTransactionCV.wait(mStateLock);
+        }
     }
 }
 
@@ -1424,8 +1391,10 @@ status_t SurfaceFlinger::setClientState(
                 }
             }
             if (what & eSizeChanged) {
-                if (layer->setSize(s.w, s.h))
+                if (layer->setSize(s.w, s.h)) {
                     flags |= eTraversalNeeded;
+                    mResizeTransationPending = true;
+                }
             }
             if (what & eAlphaChanged) {
                 if (layer->setAlpha(uint8_t(255.0f*s.alpha+0.5f)))
@@ -1543,28 +1512,27 @@ status_t SurfaceFlinger::dump(int fd, const Vector<String16>& args)
                         "id=0x%08x, client=0x%08x, identity=%u\n",
                         lbc->clientIndex(), client.get() ? client->cid : 0,
                         lbc->getIdentity());
+
+                result.append(buffer);
+                buffer[0] = 0;
             }
-            result.append(buffer);
-            buffer[0] = 0;
             /*** Layer ***/
             sp<Layer> l = LayerBase::dynamicCast< Layer* >(layer.get());
             if (l != 0) {
-                const LayerBitmap& buf0(l->getBuffer(0));
-                const LayerBitmap& buf1(l->getBuffer(1));
+                result.append( l->lcblk->dump("      ") );
+                sp<const Buffer> buf0(l->getBuffer(0));
+                sp<const Buffer> buf1(l->getBuffer(1));
                 snprintf(buffer, SIZE,
                         "      "
                         "format=%2d, [%3ux%3u:%3u] [%3ux%3u:%3u],"
-                        " freezeLock=%p, swapState=0x%08x\n",
+                        " freezeLock=%p\n",
                         l->pixelFormat(),
-                        buf0.getWidth(), buf0.getHeight(), 
-                        buf0.getBuffer()->getStride(),
-                        buf1.getWidth(), buf1.getHeight(), 
-                        buf1.getBuffer()->getStride(),
-                        l->getFreezeLock().get(),
-                        l->lcblk->swapState);
+                        buf0->getWidth(), buf0->getHeight(), buf0->getStride(),
+                        buf1->getWidth(), buf1->getHeight(), buf1->getStride(),
+                        l->getFreezeLock().get());
+                result.append(buffer);
+                buffer[0] = 0;
             }
-            result.append(buffer);
-            buffer[0] = 0;
             s.transparentRegion.dump(result, "transparentRegion");
             layer->transparentRegionScreen.dump(result, "transparentRegionScreen");
             layer->visibleRegionScreen.dump(result, "visibleRegionScreen");
@@ -1657,8 +1625,12 @@ status_t SurfaceFlinger::onTransact(
                 const DisplayHardware& hw(graphicPlane(0).displayHardware());
                 mDirtyRegion.set(hw.bounds()); // careful that's not thread-safe
                 signalEvent();
+                return NO_ERROR;
+            }
+            case 1005:{ // force transaction
+                setTransactionFlags(eTransactionNeeded|eTraversalNeeded);
+                return NO_ERROR;
             }
-            return NO_ERROR;
             case 1007: // set mFreezeCount
                 mFreezeCount = data.readInt32();
                 return NO_ERROR;
@@ -1688,21 +1660,20 @@ Client::Client(ClientID clientID, const sp<SurfaceFlinger>& flinger)
     : ctrlblk(0), cid(clientID), mPid(0), mBitmap(0), mFlinger(flinger)
 {
     const int pgsize = getpagesize();
-    const int cblksize = ((sizeof(per_client_cblk_t)+(pgsize-1))&~(pgsize-1));
+    const int cblksize = ((sizeof(SharedClient)+(pgsize-1))&~(pgsize-1));
 
     mCblkHeap = new MemoryHeapBase(cblksize, 0,
             "SurfaceFlinger Client control-block");
 
-    ctrlblk = static_cast<per_client_cblk_t *>(mCblkHeap->getBase());
+    ctrlblk = static_cast<SharedClient *>(mCblkHeap->getBase());
     if (ctrlblk) { // construct the shared structure in-place.
-        new(ctrlblk) per_client_cblk_t;
+        new(ctrlblk) SharedClient;
     }
 }
 
 Client::~Client() {
     if (ctrlblk) {
-        const int pgsize = getpagesize();
-        ctrlblk->~per_client_cblk_t();  // destroy our shared-structure.
+        ctrlblk->~SharedClient();  // destroy our shared-structure.
     }
 }