diff --git a/libs/renderengine/Android.bp b/libs/renderengine/Android.bp
index ec39137..f395ab4 100644
--- a/libs/renderengine/Android.bp
+++ b/libs/renderengine/Android.bp
@@ -48,6 +48,7 @@
     name: "librenderengine_sources",
     srcs: [
         "Description.cpp",
+        "ExternalTexture.cpp",
         "Mesh.cpp",
         "RenderEngine.cpp",
         "Texture.cpp",
diff --git a/libs/renderengine/ExternalTexture.cpp b/libs/renderengine/ExternalTexture.cpp
new file mode 100644
index 0000000..eabff58
--- /dev/null
+++ b/libs/renderengine/ExternalTexture.cpp
@@ -0,0 +1,43 @@
+/*
+ * Copyright 2021 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.
+ */
+
+#include <renderengine/ExternalTexture.h>
+#include <renderengine/RenderEngine.h>
+#include <ui/GraphicBuffer.h>
+
+#include "log/log_main.h"
+
+namespace android::renderengine {
+
+ExternalTexture::ExternalTexture(const sp<GraphicBuffer>& buffer, RenderEngine& renderEngine,
+                                 uint32_t usage)
+      : mBuffer(buffer), mRenderEngine(renderEngine) {
+    LOG_ALWAYS_FATAL_IF(buffer == nullptr,
+                        "Attempted to bind a null buffer to an external texture!");
+    // GLESRenderEngine has a separate texture cache for output buffers,
+    if (usage == Usage::WRITEABLE &&
+        (mRenderEngine.getRenderEngineType() == RenderEngine::RenderEngineType::GLES ||
+         mRenderEngine.getRenderEngineType() == RenderEngine::RenderEngineType::THREADED)) {
+        return;
+    }
+    mRenderEngine.mapExternalTextureBuffer(mBuffer, usage & Usage::WRITEABLE);
+}
+
+ExternalTexture::~ExternalTexture() {
+    mRenderEngine.unmapExternalTextureBuffer(mBuffer);
+}
+
+} // namespace android::renderengine
diff --git a/libs/renderengine/gl/GLESRenderEngine.cpp b/libs/renderengine/gl/GLESRenderEngine.cpp
index a2963a7..d87315f 100644
--- a/libs/renderengine/gl/GLESRenderEngine.cpp
+++ b/libs/renderengine/gl/GLESRenderEngine.cpp
@@ -746,7 +746,8 @@
     return;
 }
 
-void GLESRenderEngine::cacheExternalTextureBuffer(const sp<GraphicBuffer>& buffer) {
+void GLESRenderEngine::mapExternalTextureBuffer(const sp<GraphicBuffer>& buffer,
+                                                bool /*isRenderable*/) {
     ATRACE_CALL();
     mImageManager->cacheAsync(buffer, nullptr);
 }
@@ -797,8 +798,8 @@
     return NO_ERROR;
 }
 
-void GLESRenderEngine::unbindExternalTextureBuffer(uint64_t bufferId) {
-    mImageManager->releaseAsync(bufferId, nullptr);
+void GLESRenderEngine::unmapExternalTextureBuffer(const sp<GraphicBuffer>& buffer) {
+    mImageManager->releaseAsync(buffer->getId(), nullptr);
 }
 
 std::shared_ptr<ImageManager::Barrier> GLESRenderEngine::unbindExternalTextureBufferForTesting(
@@ -1102,7 +1103,7 @@
 
 status_t GLESRenderEngine::drawLayers(const DisplaySettings& display,
                                       const std::vector<const LayerSettings*>& layers,
-                                      const sp<GraphicBuffer>& buffer,
+                                      const std::shared_ptr<ExternalTexture>& buffer,
                                       const bool useFramebufferCache, base::unique_fd&& bufferFence,
                                       base::unique_fd* drawFence) {
     ATRACE_CALL();
@@ -1125,7 +1126,7 @@
         return BAD_VALUE;
     }
 
-    validateOutputBufferUsage(buffer);
+    validateOutputBufferUsage(buffer->getBuffer());
 
     std::unique_ptr<BindNativeBufferAsFramebuffer> fbo;
     // Gathering layers that requested blur, we'll need them to decide when to render to an
@@ -1142,11 +1143,13 @@
 
     if (blurLayersSize == 0) {
         fbo = std::make_unique<BindNativeBufferAsFramebuffer>(*this,
-                                                              buffer.get()->getNativeBuffer(),
+                                                              buffer->getBuffer()
+                                                                      .get()
+                                                                      ->getNativeBuffer(),
                                                               useFramebufferCache);
         if (fbo->getStatus() != NO_ERROR) {
             ALOGE("Failed to bind framebuffer! Aborting GPU composition for buffer (%p).",
-                  buffer->handle);
+                  buffer->getBuffer()->handle);
             checkErrors();
             return fbo->getStatus();
         }
@@ -1157,7 +1160,7 @@
                 mBlurFilter->setAsDrawTarget(display, blurLayers.front()->backgroundBlurRadius);
         if (status != NO_ERROR) {
             ALOGE("Failed to prepare blur filter! Aborting GPU composition for buffer (%p).",
-                  buffer->handle);
+                  buffer->getBuffer()->handle);
             checkErrors();
             return status;
         }
@@ -1194,7 +1197,7 @@
             auto status = mBlurFilter->prepare();
             if (status != NO_ERROR) {
                 ALOGE("Failed to render blur effect! Aborting GPU composition for buffer (%p).",
-                      buffer->handle);
+                      buffer->getBuffer()->handle);
                 checkErrors("Can't render first blur pass");
                 return status;
             }
@@ -1203,6 +1206,7 @@
                 // Done blurring, time to bind the native FBO and render our blur onto it.
                 fbo = std::make_unique<BindNativeBufferAsFramebuffer>(*this,
                                                                       buffer.get()
+                                                                              ->getBuffer()
                                                                               ->getNativeBuffer(),
                                                                       useFramebufferCache);
                 status = fbo->getStatus();
@@ -1215,7 +1219,7 @@
             }
             if (status != NO_ERROR) {
                 ALOGE("Failed to bind framebuffer! Aborting GPU composition for buffer (%p).",
-                      buffer->handle);
+                      buffer->getBuffer()->handle);
                 checkErrors("Can't bind native framebuffer");
                 return status;
             }
@@ -1223,7 +1227,7 @@
             status = mBlurFilter->render(blurLayersSize > 1);
             if (status != NO_ERROR) {
                 ALOGE("Failed to render blur effect! Aborting GPU composition for buffer (%p).",
-                      buffer->handle);
+                      buffer->getBuffer()->handle);
                 checkErrors("Can't render blur filter");
                 return status;
             }
@@ -1250,7 +1254,7 @@
             disableTexture = false;
             isOpaque = layer->source.buffer.isOpaque;
 
-            sp<GraphicBuffer> gBuf = layer->source.buffer.buffer;
+            sp<GraphicBuffer> gBuf = layer->source.buffer.buffer->getBuffer();
             validateInputBufferUsage(gBuf);
             bindExternalTextureBuffer(layer->source.buffer.textureName, gBuf,
                                       layer->source.buffer.fence);
@@ -1274,7 +1278,7 @@
 
             // Do not cache protected EGLImage, protected memory is limited.
             if (gBuf->getUsage() & GRALLOC_USAGE_PROTECTED) {
-                unbindExternalTextureBuffer(gBuf->getId());
+                unmapExternalTextureBuffer(gBuf);
             }
         }
 
diff --git a/libs/renderengine/gl/GLESRenderEngine.h b/libs/renderengine/gl/GLESRenderEngine.h
index cd7a86b..e7ed9c0 100644
--- a/libs/renderengine/gl/GLESRenderEngine.h
+++ b/libs/renderengine/gl/GLESRenderEngine.h
@@ -60,16 +60,14 @@
     void primeCache() override;
     void genTextures(size_t count, uint32_t* names) override;
     void deleteTextures(size_t count, uint32_t const* names) override;
-    void cacheExternalTextureBuffer(const sp<GraphicBuffer>& buffer) EXCLUDES(mRenderingMutex);
-    void unbindExternalTextureBuffer(uint64_t bufferId) EXCLUDES(mRenderingMutex);
-
     bool isProtected() const override { return mInProtectedContext; }
     bool supportsProtectedContent() const override;
     bool useProtectedContext(bool useProtectedContext) override;
     status_t drawLayers(const DisplaySettings& display,
                         const std::vector<const LayerSettings*>& layers,
-                        const sp<GraphicBuffer>& buffer, const bool useFramebufferCache,
-                        base::unique_fd&& bufferFence, base::unique_fd* drawFence) override;
+                        const std::shared_ptr<ExternalTexture>& buffer,
+                        const bool useFramebufferCache, base::unique_fd&& bufferFence,
+                        base::unique_fd* drawFence) override;
     bool cleanupPostRender(CleanupMode mode) override;
     int getContextPriority() override;
     bool supportsBackgroundBlur() override { return mBlurFilter != nullptr; }
@@ -105,6 +103,9 @@
             EXCLUDES(mFramebufferImageCacheMutex);
     size_t getMaxTextureSize() const override;
     size_t getMaxViewportDims() const override;
+    void mapExternalTextureBuffer(const sp<GraphicBuffer>& buffer, bool isRenderable)
+            EXCLUDES(mRenderingMutex);
+    void unmapExternalTextureBuffer(const sp<GraphicBuffer>& buffer) EXCLUDES(mRenderingMutex);
 
 private:
     friend class BindNativeBufferAsFramebuffer;
diff --git a/libs/renderengine/include/renderengine/ExternalTexture.h b/libs/renderengine/include/renderengine/ExternalTexture.h
new file mode 100644
index 0000000..07f0833
--- /dev/null
+++ b/libs/renderengine/include/renderengine/ExternalTexture.h
@@ -0,0 +1,61 @@
+/*
+ * Copyright 2021 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 <android-base/macros.h>
+#include <ui/GraphicBuffer.h>
+
+namespace android::renderengine {
+
+class RenderEngine;
+
+/**
+ * Manages GPU image resources on behalf of clients using RenderEngine.
+ *
+ * Clients of RenderEngine are required to wrap their GraphicBuffer objects as an ExternalTexture,
+ * which is then mapped into GPU resources required by RenderEngine. When a client no longer needs
+ * to use the GraphicBuffer as input into RenderEngine::drawLayers, then the client should delete
+ * their ExternalTexture so that resources may be freed.
+ */
+class ExternalTexture {
+public:
+    // Usage specifies the rendering intent for the buffer.
+    enum Usage : uint32_t {
+        // When a buffer is not READABLE but is WRITEABLE, then GLESRenderEngine will use that as a
+        // hint to load the buffer into a separate cache
+        READABLE = 1 << 0,
+
+        // The buffer needs to be mapped as a 2D texture if set, otherwise must be mapped as an
+        // external texture
+        WRITEABLE = 1 << 1,
+    };
+    // Creates an ExternalTexture for the provided buffer and RenderEngine instance, with the given
+    // usage hint of type Usage.
+    ExternalTexture(const sp<GraphicBuffer>& buffer, RenderEngine& renderEngine, uint32_t usage);
+
+    ~ExternalTexture();
+
+    // Retrieves the buffer that is bound to this texture.
+    const sp<GraphicBuffer>& getBuffer() const { return mBuffer; }
+
+private:
+    sp<GraphicBuffer> mBuffer;
+    RenderEngine& mRenderEngine;
+    DISALLOW_COPY_AND_ASSIGN(ExternalTexture);
+};
+
+} // namespace android::renderengine
diff --git a/libs/renderengine/include/renderengine/LayerSettings.h b/libs/renderengine/include/renderengine/LayerSettings.h
index 7661233..c54c5ba 100644
--- a/libs/renderengine/include/renderengine/LayerSettings.h
+++ b/libs/renderengine/include/renderengine/LayerSettings.h
@@ -16,11 +16,9 @@
 
 #pragma once
 
-#include <iosfwd>
-
 #include <math/mat4.h>
 #include <math/vec3.h>
-#include <renderengine/Texture.h>
+#include <renderengine/ExternalTexture.h>
 #include <ui/BlurRegion.h>
 #include <ui/Fence.h>
 #include <ui/FloatRect.h>
@@ -31,6 +29,8 @@
 #include <ui/StretchEffect.h>
 #include <ui/Transform.h>
 
+#include <iosfwd>
+
 namespace android {
 namespace renderengine {
 
@@ -39,7 +39,7 @@
     // Buffer containing the image that we will render.
     // If buffer == nullptr, then the rest of the fields in this struct will be
     // ignored.
-    sp<GraphicBuffer> buffer = nullptr;
+    std::shared_ptr<ExternalTexture> buffer = nullptr;
 
     // Fence that will fire when the buffer is ready to be bound.
     sp<Fence> fence = nullptr;
diff --git a/libs/renderengine/include/renderengine/RenderEngine.h b/libs/renderengine/include/renderengine/RenderEngine.h
index 8dd98c3..c8a0f0a 100644
--- a/libs/renderengine/include/renderengine/RenderEngine.h
+++ b/libs/renderengine/include/renderengine/RenderEngine.h
@@ -17,19 +17,20 @@
 #ifndef SF_RENDERENGINE_H_
 #define SF_RENDERENGINE_H_
 
-#include <stdint.h>
-#include <sys/types.h>
-#include <memory>
-
 #include <android-base/unique_fd.h>
 #include <math/mat4.h>
 #include <renderengine/DisplaySettings.h>
+#include <renderengine/ExternalTexture.h>
 #include <renderengine/Framebuffer.h>
 #include <renderengine/Image.h>
 #include <renderengine/LayerSettings.h>
+#include <stdint.h>
+#include <sys/types.h>
 #include <ui/GraphicTypes.h>
 #include <ui/Transform.h>
 
+#include <memory>
+
 /**
  * Allows to set RenderEngine backend to GLES (default) or SkiaGL (NOT yet supported).
  */
@@ -51,6 +52,7 @@
 
 namespace renderengine {
 
+class ExternalTexture;
 class Image;
 class Mesh;
 class Texture;
@@ -104,23 +106,6 @@
 
     virtual void genTextures(size_t count, uint32_t* names) = 0;
     virtual void deleteTextures(size_t count, uint32_t const* names) = 0;
-    // Caches Image resources for this buffer, but does not bind the buffer to
-    // a particular texture.
-    // Note that work is deferred to an additional thread, i.e. this call
-    // is made asynchronously, but the caller can expect that cache/unbind calls
-    // are performed in a manner that's conflict serializable, i.e. unbinding
-    // a buffer should never occur before binding the buffer if the caller
-    // called {bind, cache}ExternalTextureBuffer before calling unbind.
-    virtual void cacheExternalTextureBuffer(const sp<GraphicBuffer>& buffer) = 0;
-    // Removes internal resources referenced by the bufferId. This method should be
-    // invoked when the caller will no longer hold a reference to a GraphicBuffer
-    // and needs to clean up its resources.
-    // Note that work is deferred to an additional thread, i.e. this call
-    // is made asynchronously, but the caller can expect that cache/unbind calls
-    // are performed in a manner that's conflict serializable, i.e. unbinding
-    // a buffer should never occur before binding the buffer if the caller
-    // called {bind, cache}ExternalTextureBuffer before calling unbind.
-    virtual void unbindExternalTextureBuffer(uint64_t bufferId) = 0;
 
     enum class CleanupMode {
         CLEAN_OUTPUT_RESOURCES,
@@ -191,8 +176,9 @@
     // now, this always returns NO_ERROR.
     virtual status_t drawLayers(const DisplaySettings& display,
                                 const std::vector<const LayerSettings*>& layers,
-                                const sp<GraphicBuffer>& buffer, const bool useFramebufferCache,
-                                base::unique_fd&& bufferFence, base::unique_fd* drawFence) = 0;
+                                const std::shared_ptr<ExternalTexture>& buffer,
+                                const bool useFramebufferCache, base::unique_fd&& bufferFence,
+                                base::unique_fd* drawFence) = 0;
     virtual void cleanFramebufferCache() = 0;
     // Returns the priority this context was actually created with. Note: this may not be
     // the same as specified at context creation time, due to implementation limits on the
@@ -213,6 +199,31 @@
     static void validateOutputBufferUsage(const sp<GraphicBuffer>&);
 
 protected:
+    // Maps GPU resources for this buffer.
+    // Note that work may be deferred to an additional thread, i.e. this call
+    // is made asynchronously, but the caller can expect that map/unmap calls
+    // are performed in a manner that's conflict serializable, i.e. unmapping
+    // a buffer should never occur before binding the buffer if the caller
+    // called mapExternalTextureBuffer before calling unmap.
+    // Note also that if the buffer contains protected content, then mapping those GPU resources may
+    // be deferred until the buffer is really used for drawing. This is because typical SoCs that
+    // support protected memory only support a limited amount, so optimisitically mapping protected
+    // memory may be too burdensome. If a buffer contains protected content and the RenderEngine
+    // implementation supports protected context, then GPU resources may be mapped into both the
+    // protected and unprotected contexts.
+    // If the buffer may ever be written to by RenderEngine, then isRenderable must be true.
+    virtual void mapExternalTextureBuffer(const sp<GraphicBuffer>& buffer, bool isRenderable) = 0;
+    // Unmaps GPU resources used by this buffer. This method should be
+    // invoked when the caller will no longer hold a reference to a GraphicBuffer
+    // and needs to clean up its resources.
+    // Note that if there are multiple callers holding onto the same buffer, then the buffer's
+    // resources may be internally ref-counted to guard against use-after-free errors. Note that
+    // work may be deferred to an additional thread, i.e. this call is expected to be made
+    // asynchronously, but the caller can expect that map/unmap calls are performed in a manner
+    // that's conflict serializable, i.e. unmap a buffer should never occur before binding the
+    // buffer if the caller called mapExternalTextureBuffer before calling unmap.
+    virtual void unmapExternalTextureBuffer(const sp<GraphicBuffer>& buffer) = 0;
+    friend class ExternalTexture;
     friend class threaded::RenderEngineThreaded;
     const RenderEngineType mRenderEngineType;
 };
diff --git a/libs/renderengine/include/renderengine/mock/RenderEngine.h b/libs/renderengine/include/renderengine/mock/RenderEngine.h
index 228553d..27dbd1e 100644
--- a/libs/renderengine/include/renderengine/mock/RenderEngine.h
+++ b/libs/renderengine/include/renderengine/mock/RenderEngine.h
@@ -39,8 +39,6 @@
     MOCK_METHOD1(dump, void(std::string&));
     MOCK_METHOD2(genTextures, void(size_t, uint32_t*));
     MOCK_METHOD2(deleteTextures, void(size_t, uint32_t const*));
-    MOCK_METHOD1(cacheExternalTextureBuffer, void(const sp<GraphicBuffer>&));
-    MOCK_METHOD1(unbindExternalTextureBuffer, void(uint64_t));
     MOCK_METHOD1(drawMesh, void(const renderengine::Mesh&));
     MOCK_CONST_METHOD0(getMaxTextureSize, size_t());
     MOCK_CONST_METHOD0(getMaxViewportDims, size_t());
@@ -50,12 +48,17 @@
     MOCK_METHOD1(cleanupPostRender, bool(CleanupMode mode));
     MOCK_METHOD6(drawLayers,
                  status_t(const DisplaySettings&, const std::vector<const LayerSettings*>&,
-                          const sp<GraphicBuffer>&, const bool, base::unique_fd&&,
+                          const std::shared_ptr<ExternalTexture>&, const bool, base::unique_fd&&,
                           base::unique_fd*));
     MOCK_METHOD0(cleanFramebufferCache, void());
     MOCK_METHOD0(getContextPriority, int());
     MOCK_METHOD0(supportsBackgroundBlur, bool());
     MOCK_METHOD1(onPrimaryDisplaySizeChanged, void(ui::Size));
+
+protected:
+    // mock renderengine still needs to implement these, but callers should never need to call them.
+    void mapExternalTextureBuffer(const sp<GraphicBuffer>&, bool) {}
+    void unmapExternalTextureBuffer(const sp<GraphicBuffer>&) {}
 };
 
 } // namespace mock
diff --git a/libs/renderengine/skia/AutoBackendTexture.h b/libs/renderengine/skia/AutoBackendTexture.h
index bb75878..2d61cf8 100644
--- a/libs/renderengine/skia/AutoBackendTexture.h
+++ b/libs/renderengine/skia/AutoBackendTexture.h
@@ -21,9 +21,9 @@
 #include <SkImage.h>
 #include <SkSurface.h>
 #include <sys/types.h>
+#include <ui/GraphicTypes.h>
 
 #include "android-base/macros.h"
-#include "ui/GraphicTypes.h"
 
 namespace android {
 namespace renderengine {
@@ -41,13 +41,18 @@
     // of shared ownership with Skia objects, so we wrap it here instead.
     class LocalRef {
     public:
-        LocalRef() {}
+        LocalRef(AutoBackendTexture* texture) { setTexture(texture); }
 
         ~LocalRef() {
             // Destroying the texture is the same as setting it to null
             setTexture(nullptr);
         }
 
+        AutoBackendTexture* getTexture() const { return mTexture; }
+
+        DISALLOW_COPY_AND_ASSIGN(LocalRef);
+
+    private:
         // Sets the texture to locally ref-track.
         void setTexture(AutoBackendTexture* texture) {
             if (mTexture != nullptr) {
@@ -59,12 +64,6 @@
                 mTexture->ref();
             }
         }
-
-        AutoBackendTexture* getTexture() const { return mTexture; }
-
-        DISALLOW_COPY_AND_ASSIGN(LocalRef);
-
-    private:
         AutoBackendTexture* mTexture = nullptr;
     };
 
diff --git a/libs/renderengine/skia/Cache.cpp b/libs/renderengine/skia/Cache.cpp
index 1db20c0..1c2b2fc 100644
--- a/libs/renderengine/skia/Cache.cpp
+++ b/libs/renderengine/skia/Cache.cpp
@@ -46,7 +46,7 @@
 } // namespace
 
 static void drawShadowLayers(SkiaRenderEngine* renderengine, const DisplaySettings& display,
-                             sp<GraphicBuffer> dstBuffer) {
+                             const std::shared_ptr<ExternalTexture>& dstTexture) {
     // Somewhat arbitrary dimensions, but on screen and slightly shorter, based
     // on actual use.
     FloatRect rect(0, 0, display.physicalDisplay.width(), display.physicalDisplay.height() - 30);
@@ -73,7 +73,7 @@
 
     auto layers = std::vector<const LayerSettings*>{&layer};
     // The identity matrix will generate the fast shader
-    renderengine->drawLayers(display, layers, dstBuffer, kUseFrameBufferCache, base::unique_fd(),
+    renderengine->drawLayers(display, layers, dstTexture, kUseFrameBufferCache, base::unique_fd(),
                              nullptr);
     // This matrix, which has different scales for x and y, will
     // generate the slower (more general case) version, which has variants for translucent
@@ -86,13 +86,14 @@
     // clang-format on
     for (auto translucent : {false, true}) {
         layer.shadow.casterIsTranslucent = translucent;
-        renderengine->drawLayers(display, layers, dstBuffer, kUseFrameBufferCache,
+        renderengine->drawLayers(display, layers, dstTexture, kUseFrameBufferCache,
                                  base::unique_fd(), nullptr);
     }
 }
 
 static void drawImageLayers(SkiaRenderEngine* renderengine, const DisplaySettings& display,
-                            sp<GraphicBuffer> dstBuffer, sp<GraphicBuffer> srcBuffer) {
+                            const std::shared_ptr<ExternalTexture>& dstTexture,
+                            const std::shared_ptr<ExternalTexture>& srcTexture) {
     const Rect& displayRect = display.physicalDisplay;
     FloatRect rect(0, 0, displayRect.width(), displayRect.height());
     LayerSettings layer{
@@ -103,7 +104,7 @@
                     },
             .source = PixelSource{.buffer =
                                           Buffer{
-                                                  .buffer = srcBuffer,
+                                                  .buffer = srcTexture,
                                                   .maxMasteringLuminance = 1000.f,
                                                   .maxContentLuminance = 1000.f,
                                           }},
@@ -126,7 +127,7 @@
                 layer.source.buffer.isOpaque = isOpaque;
                 for (auto alpha : {half(.23999f), half(1.0f)}) {
                     layer.alpha = alpha;
-                    renderengine->drawLayers(display, layers, dstBuffer, kUseFrameBufferCache,
+                    renderengine->drawLayers(display, layers, dstTexture, kUseFrameBufferCache,
                                              base::unique_fd(), nullptr);
                 }
             }
@@ -135,7 +136,7 @@
 }
 
 static void drawSolidLayers(SkiaRenderEngine* renderengine, const DisplaySettings& display,
-                            sp<GraphicBuffer> dstBuffer) {
+                            const std::shared_ptr<ExternalTexture>& dstTexture) {
     const Rect& displayRect = display.physicalDisplay;
     FloatRect rect(0, 0, displayRect.width(), displayRect.height());
     LayerSettings layer{
@@ -143,11 +144,11 @@
                     Geometry{
                             .boundaries = rect,
                     },
-            .alpha = 1,
             .source =
                     PixelSource{
                             .solidColor = half3(0.1f, 0.2f, 0.3f),
                     },
+            .alpha = 1,
     };
 
     auto layers = std::vector<const LayerSettings*>{&layer};
@@ -155,14 +156,14 @@
         layer.geometry.positionTransform = transform;
         for (float roundedCornersRadius : {0.0f, 0.05f, 50.f}) {
             layer.geometry.roundedCornersRadius = roundedCornersRadius;
-            renderengine->drawLayers(display, layers, dstBuffer, kUseFrameBufferCache,
+            renderengine->drawLayers(display, layers, dstTexture, kUseFrameBufferCache,
                                      base::unique_fd(), nullptr);
         }
     }
 }
 
 static void drawBlurLayers(SkiaRenderEngine* renderengine, const DisplaySettings& display,
-                           sp<GraphicBuffer> dstBuffer) {
+                           const std::shared_ptr<ExternalTexture>& dstTexture) {
     const Rect& displayRect = display.physicalDisplay;
     FloatRect rect(0, 0, displayRect.width(), displayRect.height());
     LayerSettings layer{
@@ -176,7 +177,7 @@
     auto layers = std::vector<const LayerSettings*>{&layer};
     for (int radius : {9, 60}) {
         layer.backgroundBlurRadius = radius;
-        renderengine->drawLayers(display, layers, dstBuffer, kUseFrameBufferCache,
+        renderengine->drawLayers(display, layers, dstTexture, kUseFrameBufferCache,
                                  base::unique_fd(), nullptr);
     }
 }
@@ -214,6 +215,9 @@
     sp<GraphicBuffer> dstBuffer =
             new GraphicBuffer(displayRect.width(), displayRect.height(), PIXEL_FORMAT_RGBA_8888, 1,
                               usage, "primeShaderCache_dst");
+
+    const auto dstTexture = std::make_shared<ExternalTexture>(dstBuffer, *renderengine,
+                                                              ExternalTexture::Usage::WRITEABLE);
     // This buffer will be the source for the call to drawImageLayers. Draw
     // something to it as a placeholder for what an app draws. We should draw
     // something, but the details are not important. Make use of the shadow layer drawing step
@@ -222,11 +226,16 @@
             new GraphicBuffer(displayRect.width(), displayRect.height(), PIXEL_FORMAT_RGBA_8888, 1,
                               usage, "drawImageLayer_src");
 
-    drawSolidLayers(renderengine, display, dstBuffer);
-    drawShadowLayers(renderengine, display, srcBuffer);
-    drawBlurLayers(renderengine, display, dstBuffer);
+    const auto srcTexture =
+            std::make_shared<ExternalTexture>(srcBuffer, *renderengine,
+                                              ExternalTexture::Usage::READABLE |
+                                                      ExternalTexture::Usage::WRITEABLE);
+
+    drawSolidLayers(renderengine, display, dstTexture);
+    drawShadowLayers(renderengine, display, srcTexture);
+    drawBlurLayers(renderengine, display, dstTexture);
     // The majority of shaders are related to sampling images.
-    drawImageLayers(renderengine, display, dstBuffer, srcBuffer);
+    drawImageLayers(renderengine, display, dstTexture, srcTexture);
 
     // should be the same as AHARDWAREBUFFER_USAGE_GPU_SAMPLED_IMAGE;
     const int64_t usageExternal = GRALLOC_USAGE_HW_TEXTURE;
@@ -234,12 +243,12 @@
     sp<GraphicBuffer> externalBuffer =
             new GraphicBuffer(displayRect.width(), displayRect.height(), PIXEL_FORMAT_RGBA_8888, 1,
                               usageExternal, "primeShaderCache_external");
+    const auto externalTexture =
+            std::make_shared<ExternalTexture>(externalBuffer, *renderengine,
+                                              ExternalTexture::Usage::READABLE);
     // TODO(b/184665179) doubles number of image shader compilations, but only somewhere
     // between 6 and 8 will occur in real uses.
-    drawImageLayers(renderengine, display, dstBuffer, externalBuffer);
-    renderengine->unbindExternalTextureBuffer(externalBuffer->getId());
-
-    renderengine->unbindExternalTextureBuffer(srcBuffer->getId());
+    drawImageLayers(renderengine, display, dstTexture, externalTexture);
 
     const nsecs_t timeAfter = systemTime();
     const float compileTimeMs = static_cast<float>(timeAfter - timeBefore) / 1.0E6;
diff --git a/libs/renderengine/skia/SkiaGLRenderEngine.cpp b/libs/renderengine/skia/SkiaGLRenderEngine.cpp
index df40dd9..37d98a3 100644
--- a/libs/renderengine/skia/SkiaGLRenderEngine.cpp
+++ b/libs/renderengine/skia/SkiaGLRenderEngine.cpp
@@ -329,8 +329,6 @@
 }
 
 SkiaGLRenderEngine::~SkiaGLRenderEngine() {
-    cleanFramebufferCache();
-
     std::lock_guard<std::mutex> lock(mRenderingMutex);
     if (mBlurFilter) {
         delete mBlurFilter;
@@ -484,7 +482,8 @@
             sourceTransfer != destTransfer;
 }
 
-void SkiaGLRenderEngine::cacheExternalTextureBuffer(const sp<GraphicBuffer>& buffer) {
+void SkiaGLRenderEngine::mapExternalTextureBuffer(const sp<GraphicBuffer>& buffer,
+                                                  bool isRenderable) {
     // Only run this if RE is running on its own thread. This way the access to GL
     // operations is guaranteed to be happening on the same thread.
     if (mRenderEngineType != RenderEngineType::SKIA_GL_THREADED) {
@@ -505,25 +504,41 @@
     auto& cache = mInProtectedContext ? mProtectedTextureCache : mTextureCache;
 
     std::lock_guard<std::mutex> lock(mRenderingMutex);
-    auto iter = cache.find(buffer->getId());
-    if (iter != cache.end()) {
-        ALOGV("Texture already exists in cache.");
-    } else {
+    mGraphicBufferExternalRefs[buffer->getId()]++;
+
+    if (const auto& iter = cache.find(buffer->getId()); iter == cache.end()) {
         std::shared_ptr<AutoBackendTexture::LocalRef> imageTextureRef =
-                std::make_shared<AutoBackendTexture::LocalRef>();
-        imageTextureRef->setTexture(
-                new AutoBackendTexture(grContext.get(), buffer->toAHardwareBuffer(), false));
+                std::make_shared<AutoBackendTexture::LocalRef>(
+                        new AutoBackendTexture(grContext.get(), buffer->toAHardwareBuffer(),
+                                               isRenderable));
         cache.insert({buffer->getId(), imageTextureRef});
     }
     // restore the original state of the protected context if necessary
     useProtectedContext(protectedContextState);
 }
 
-void SkiaGLRenderEngine::unbindExternalTextureBuffer(uint64_t bufferId) {
+void SkiaGLRenderEngine::unmapExternalTextureBuffer(const sp<GraphicBuffer>& buffer) {
     ATRACE_CALL();
     std::lock_guard<std::mutex> lock(mRenderingMutex);
-    mTextureCache.erase(bufferId);
-    mProtectedTextureCache.erase(bufferId);
+    if (const auto& iter = mGraphicBufferExternalRefs.find(buffer->getId());
+        iter != mGraphicBufferExternalRefs.end()) {
+        if (iter->second == 0) {
+            ALOGW("Attempted to unmap GraphicBuffer <id: %" PRId64
+                  "> from RenderEngine texture, but the "
+                  "ref count was already zero!",
+                  buffer->getId());
+            mGraphicBufferExternalRefs.erase(buffer->getId());
+            return;
+        }
+
+        iter->second--;
+
+        if (iter->second == 0) {
+            mTextureCache.erase(buffer->getId());
+            mProtectedTextureCache.erase(buffer->getId());
+            mGraphicBufferExternalRefs.erase(buffer->getId());
+        }
+    }
 }
 
 sk_sp<SkShader> SkiaGLRenderEngine::createRuntimeEffectShader(sk_sp<SkShader> shader,
@@ -621,8 +636,8 @@
 
 status_t SkiaGLRenderEngine::drawLayers(const DisplaySettings& display,
                                         const std::vector<const LayerSettings*>& layers,
-                                        const sp<GraphicBuffer>& buffer,
-                                        const bool useFramebufferCache,
+                                        const std::shared_ptr<ExternalTexture>& buffer,
+                                        const bool /*useFramebufferCache*/,
                                         base::unique_fd&& bufferFence, base::unique_fd* drawFence) {
     ATRACE_NAME("SkiaGL::drawLayers");
 
@@ -645,32 +660,18 @@
         return BAD_VALUE;
     }
 
-    validateOutputBufferUsage(buffer);
+    validateOutputBufferUsage(buffer->getBuffer());
 
     auto grContext = mInProtectedContext ? mProtectedGrContext : mGrContext;
     auto& cache = mInProtectedContext ? mProtectedTextureCache : mTextureCache;
-    AHardwareBuffer_Desc bufferDesc;
-    AHardwareBuffer_describe(buffer->toAHardwareBuffer(), &bufferDesc);
 
-    std::shared_ptr<AutoBackendTexture::LocalRef> surfaceTextureRef = nullptr;
-    if (useFramebufferCache) {
-        auto iter = cache.find(buffer->getId());
-        if (iter != cache.end()) {
-            ALOGV("Cache hit!");
-            ATRACE_NAME("Cache hit");
-            surfaceTextureRef = iter->second;
-        }
-    }
-
-    if (surfaceTextureRef == nullptr || surfaceTextureRef->getTexture() == nullptr) {
-        ATRACE_NAME("Cache miss");
-        surfaceTextureRef = std::make_shared<AutoBackendTexture::LocalRef>();
-        surfaceTextureRef->setTexture(
-                new AutoBackendTexture(grContext.get(), buffer->toAHardwareBuffer(), true));
-        if (useFramebufferCache) {
-            ALOGD("Adding to cache");
-            cache.insert({buffer->getId(), surfaceTextureRef});
-        }
+    std::shared_ptr<AutoBackendTexture::LocalRef> surfaceTextureRef;
+    if (const auto& it = cache.find(buffer->getBuffer()->getId()); it != cache.end()) {
+        surfaceTextureRef = it->second;
+    } else {
+        surfaceTextureRef = std::make_shared<AutoBackendTexture::LocalRef>(
+                new AutoBackendTexture(grContext.get(), buffer->getBuffer()->toAHardwareBuffer(),
+                                       true));
     }
 
     const ui::Dataspace dstDataspace =
@@ -876,18 +877,22 @@
         SkPaint paint;
         if (layer->source.buffer.buffer) {
             ATRACE_NAME("DrawImage");
-            validateInputBufferUsage(layer->source.buffer.buffer);
+            validateInputBufferUsage(layer->source.buffer.buffer->getBuffer());
             const auto& item = layer->source.buffer;
             std::shared_ptr<AutoBackendTexture::LocalRef> imageTextureRef = nullptr;
-            auto iter = cache.find(item.buffer->getId());
-            if (iter != cache.end()) {
+
+            if (const auto& iter = cache.find(item.buffer->getBuffer()->getId());
+                iter != cache.end()) {
                 imageTextureRef = iter->second;
             } else {
-                imageTextureRef = std::make_shared<AutoBackendTexture::LocalRef>();
-                imageTextureRef->setTexture(new AutoBackendTexture(grContext.get(),
-                                                                   item.buffer->toAHardwareBuffer(),
-                                                                   false));
-                cache.insert({item.buffer->getId(), imageTextureRef});
+                // If we didn't find the image in the cache, then create a local ref but don't cache
+                // it. If we're using skia, we're guaranteed to run on a dedicated GPU thread so if
+                // we didn't find anything in the cache then we intentionally did not cache this
+                // buffer's resources.
+                imageTextureRef = std::make_shared<AutoBackendTexture::LocalRef>(
+                        new AutoBackendTexture(grContext.get(),
+                                               item.buffer->getBuffer()->toAHardwareBuffer(),
+                                               false));
             }
 
             sk_sp<SkImage> image =
@@ -1200,15 +1205,6 @@
     return eglCreatePbufferSurface(display, placeholderConfig, attributes.data());
 }
 
-void SkiaGLRenderEngine::cleanFramebufferCache() {
-    // TODO(b/180767535) Remove this method and use b/180767535 instead, which would allow
-    // SF to control texture lifecycle more tightly rather than through custom hooks into RE.
-    std::lock_guard<std::mutex> lock(mRenderingMutex);
-    mRuntimeEffects.clear();
-    mProtectedTextureCache.clear();
-    mTextureCache.clear();
-}
-
 int SkiaGLRenderEngine::getContextPriority() {
     int value;
     eglQueryContext(mEGLDisplay, mEGLContext, EGL_CONTEXT_PRIORITY_LEVEL_IMG, &value);
@@ -1281,6 +1277,12 @@
         StringAppendF(&result, "Skia's Wrapped Objects:\n");
         gpuReporter.logOutput(result, true);
 
+        StringAppendF(&result, "RenderEngine tracked buffers: %zu\n",
+                      mGraphicBufferExternalRefs.size());
+        StringAppendF(&result, "Dumping buffer ids...\n");
+        for (const auto& [id, refCounts] : mGraphicBufferExternalRefs) {
+            StringAppendF(&result, "- 0x%" PRIx64 " - %d refs \n", id, refCounts);
+        }
         StringAppendF(&result, "RenderEngine AHB/BackendTexture cache size: %zu\n",
                       mTextureCache.size());
         StringAppendF(&result, "Dumping buffer ids...\n");
diff --git a/libs/renderengine/skia/SkiaGLRenderEngine.h b/libs/renderengine/skia/SkiaGLRenderEngine.h
index 8e77c16..e71c560 100644
--- a/libs/renderengine/skia/SkiaGLRenderEngine.h
+++ b/libs/renderengine/skia/SkiaGLRenderEngine.h
@@ -23,6 +23,7 @@
 #include <GrDirectContext.h>
 #include <SkSurface.h>
 #include <android-base/thread_annotations.h>
+#include <renderengine/ExternalTexture.h>
 #include <renderengine/RenderEngine.h>
 #include <sys/types.h>
 
@@ -52,13 +53,12 @@
     ~SkiaGLRenderEngine() override EXCLUDES(mRenderingMutex);
 
     void primeCache() override;
-    void cacheExternalTextureBuffer(const sp<GraphicBuffer>& buffer) override;
-    void unbindExternalTextureBuffer(uint64_t bufferId) override;
     status_t drawLayers(const DisplaySettings& display,
                         const std::vector<const LayerSettings*>& layers,
-                        const sp<GraphicBuffer>& buffer, const bool useFramebufferCache,
-                        base::unique_fd&& bufferFence, base::unique_fd* drawFence) override;
-    void cleanFramebufferCache() override;
+                        const std::shared_ptr<ExternalTexture>& buffer,
+                        const bool useFramebufferCache, base::unique_fd&& bufferFence,
+                        base::unique_fd* drawFence) override;
+    void cleanFramebufferCache() override {}
     int getContextPriority() override;
     bool isProtected() const override { return mInProtectedContext; }
     bool supportsProtectedContent() const override;
@@ -72,6 +72,8 @@
     void dump(std::string& result) override;
     size_t getMaxTextureSize() const override;
     size_t getMaxViewportDims() const override;
+    void mapExternalTextureBuffer(const sp<GraphicBuffer>& buffer, bool isRenderable) override;
+    void unmapExternalTextureBuffer(const sp<GraphicBuffer>& buffer) override;
 
 private:
     static EGLConfig chooseEglConfig(EGLDisplay display, int format, bool logConfig);
@@ -114,7 +116,9 @@
     const PixelFormat mDefaultPixelFormat;
     const bool mUseColorManagement;
 
-    // Cache of GL textures that we'll store per GraphicBuffer ID
+    // Number of external holders of ExternalTexture references, per GraphicBuffer ID.
+    std::unordered_map<uint64_t, int32_t> mGraphicBufferExternalRefs GUARDED_BY(mRenderingMutex);
+    // Cache of GL textures that we'll store per GraphicBuffer ID, sliced by GPU context.
     std::unordered_map<uint64_t, std::shared_ptr<AutoBackendTexture::LocalRef>> mTextureCache
             GUARDED_BY(mRenderingMutex);
     std::unordered_map<uint64_t, std::shared_ptr<AutoBackendTexture::LocalRef>>
diff --git a/libs/renderengine/skia/SkiaRenderEngine.h b/libs/renderengine/skia/SkiaRenderEngine.h
index 51ef088..308c5ff 100644
--- a/libs/renderengine/skia/SkiaRenderEngine.h
+++ b/libs/renderengine/skia/SkiaRenderEngine.h
@@ -42,15 +42,12 @@
     virtual void primeCache() override{};
     virtual void genTextures(size_t /*count*/, uint32_t* /*names*/) override{};
     virtual void deleteTextures(size_t /*count*/, uint32_t const* /*names*/) override{};
-    virtual void cacheExternalTextureBuffer(const sp<GraphicBuffer>& /*buffer*/){};
-    virtual void unbindExternalTextureBuffer(uint64_t /*bufferId*/){};
-
     virtual bool isProtected() const override { return false; } // mInProtectedContext; }
     virtual bool supportsProtectedContent() const override { return false; };
     virtual bool useProtectedContext(bool /*useProtectedContext*/) override { return false; };
     virtual status_t drawLayers(const DisplaySettings& /*display*/,
                                 const std::vector<const LayerSettings*>& /*layers*/,
-                                const sp<GraphicBuffer>& /*buffer*/,
+                                const std::shared_ptr<ExternalTexture>& /*buffer*/,
                                 const bool /*useFramebufferCache*/,
                                 base::unique_fd&& /*bufferFence*/,
                                 base::unique_fd* /*drawFence*/) override {
@@ -60,6 +57,11 @@
     virtual int getContextPriority() override { return 0; }
     virtual void assertShadersCompiled(int numShaders) {}
     virtual int reportShadersCompiled() { return 0; }
+
+protected:
+    virtual void mapExternalTextureBuffer(const sp<GraphicBuffer>& /*buffer*/,
+                                          bool /*isRenderable*/) override;
+    virtual void unmapExternalTextureBuffer(const sp<GraphicBuffer>& /*buffer*/) override;
 };
 
 } // namespace skia
diff --git a/libs/renderengine/tests/RenderEngineTest.cpp b/libs/renderengine/tests/RenderEngineTest.cpp
index 7846156..d63c88b 100644
--- a/libs/renderengine/tests/RenderEngineTest.cpp
+++ b/libs/renderengine/tests/RenderEngineTest.cpp
@@ -24,6 +24,7 @@
 
 #include <cutils/properties.h>
 #include <gtest/gtest.h>
+#include <renderengine/ExternalTexture.h>
 #include <renderengine/RenderEngine.h>
 #include <sync/sync.h>
 #include <ui/PixelFormat.h>
@@ -160,27 +161,42 @@
 
 class RenderEngineTest : public ::testing::TestWithParam<std::shared_ptr<RenderEngineFactory>> {
 public:
-    static sp<GraphicBuffer> allocateDefaultBuffer() {
-        return new GraphicBuffer(DEFAULT_DISPLAY_WIDTH, DEFAULT_DISPLAY_HEIGHT,
-                                 HAL_PIXEL_FORMAT_RGBA_8888, 1,
-                                 GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN |
-                                         GRALLOC_USAGE_HW_RENDER | GRALLOC_USAGE_HW_TEXTURE,
-                                 "output");
+    std::shared_ptr<renderengine::ExternalTexture> allocateDefaultBuffer() {
+        return std::make_shared<
+                renderengine::
+                        ExternalTexture>(new GraphicBuffer(DEFAULT_DISPLAY_WIDTH,
+                                                           DEFAULT_DISPLAY_HEIGHT,
+                                                           HAL_PIXEL_FORMAT_RGBA_8888, 1,
+                                                           GRALLOC_USAGE_SW_READ_OFTEN |
+                                                                   GRALLOC_USAGE_SW_WRITE_OFTEN |
+                                                                   GRALLOC_USAGE_HW_RENDER |
+                                                                   GRALLOC_USAGE_HW_TEXTURE,
+                                                           "output"),
+                                         *mRE,
+                                         renderengine::ExternalTexture::Usage::READABLE |
+                                                 renderengine::ExternalTexture::Usage::WRITEABLE);
     }
 
     // Allocates a 1x1 buffer to fill with a solid color
-    static sp<GraphicBuffer> allocateSourceBuffer(uint32_t width, uint32_t height) {
-        return new GraphicBuffer(width, height, HAL_PIXEL_FORMAT_RGBA_8888, 1,
-                                 GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN |
-                                         GRALLOC_USAGE_HW_TEXTURE,
-                                 "input");
+    std::shared_ptr<renderengine::ExternalTexture> allocateSourceBuffer(uint32_t width,
+                                                                        uint32_t height) {
+        return std::make_shared<
+                renderengine::
+                        ExternalTexture>(new GraphicBuffer(width, height,
+                                                           HAL_PIXEL_FORMAT_RGBA_8888, 1,
+                                                           GRALLOC_USAGE_SW_READ_OFTEN |
+                                                                   GRALLOC_USAGE_SW_WRITE_OFTEN |
+                                                                   GRALLOC_USAGE_HW_TEXTURE,
+                                                           "input"),
+                                         *mRE,
+                                         renderengine::ExternalTexture::Usage::READABLE |
+                                                 renderengine::ExternalTexture::Usage::WRITEABLE);
     }
 
     RenderEngineTest() {
         const ::testing::TestInfo* const test_info =
                 ::testing::UnitTest::GetInstance()->current_test_info();
         ALOGD("**** Setting up for %s.%s\n", test_info->test_case_name(), test_info->name());
-        mBuffer = allocateDefaultBuffer();
     }
 
     ~RenderEngineTest() {
@@ -211,20 +227,21 @@
         }
 
         uint8_t* pixels;
-        mBuffer->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
-                      reinterpret_cast<void**>(&pixels));
+        mBuffer->getBuffer()->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
+                                   reinterpret_cast<void**>(&pixels));
 
         file << "P6\n";
-        file << mBuffer->getWidth() << "\n";
-        file << mBuffer->getHeight() << "\n";
+        file << mBuffer->getBuffer()->getWidth() << "\n";
+        file << mBuffer->getBuffer()->getHeight() << "\n";
         file << 255 << "\n";
 
-        std::vector<uint8_t> outBuffer(mBuffer->getWidth() * mBuffer->getHeight() * 3);
+        std::vector<uint8_t> outBuffer(mBuffer->getBuffer()->getWidth() *
+                                       mBuffer->getBuffer()->getHeight() * 3);
         auto outPtr = reinterpret_cast<uint8_t*>(outBuffer.data());
 
-        for (int32_t j = 0; j < mBuffer->getHeight(); j++) {
-            const uint8_t* src = pixels + (mBuffer->getStride() * j) * 4;
-            for (int32_t i = 0; i < mBuffer->getWidth(); i++) {
+        for (int32_t j = 0; j < mBuffer->getBuffer()->getHeight(); j++) {
+            const uint8_t* src = pixels + (mBuffer->getBuffer()->getStride() * j) * 4;
+            for (int32_t i = 0; i < mBuffer->getBuffer()->getWidth(); i++) {
                 // Only copy R, G and B components
                 outPtr[0] = src[0];
                 outPtr[1] = src[1];
@@ -235,7 +252,7 @@
             }
         }
         file.write(reinterpret_cast<char*>(outBuffer.data()), outBuffer.size());
-        mBuffer->unlock();
+        mBuffer->getBuffer()->unlock();
     }
 
     void expectBufferColor(const Region& region, uint8_t r, uint8_t g, uint8_t b, uint8_t a) {
@@ -262,13 +279,13 @@
     void expectBufferColor(const Rect& region, uint8_t r, uint8_t g, uint8_t b, uint8_t a,
                            std::function<bool(const uint8_t* a, const uint8_t* b)> colorCompare) {
         uint8_t* pixels;
-        mBuffer->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
-                      reinterpret_cast<void**>(&pixels));
+        mBuffer->getBuffer()->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
+                                   reinterpret_cast<void**>(&pixels));
         int32_t maxFails = 10;
         int32_t fails = 0;
         for (int32_t j = 0; j < region.getHeight(); j++) {
-            const uint8_t* src =
-                    pixels + (mBuffer->getStride() * (region.top + j) + region.left) * 4;
+            const uint8_t* src = pixels +
+                    (mBuffer->getBuffer()->getStride() * (region.top + j) + region.left) * 4;
             for (int32_t i = 0; i < region.getWidth(); i++) {
                 const uint8_t expected[4] = {r, g, b, a};
                 bool equal = colorCompare(src, expected);
@@ -289,7 +306,7 @@
                 break;
             }
         }
-        mBuffer->unlock();
+        mBuffer->getBuffer()->unlock();
     }
 
     void expectAlpha(const Rect& rect, uint8_t a) {
@@ -387,7 +404,6 @@
         base::unique_fd fence;
         status_t status =
                 mRE->drawLayers(settings, layers, mBuffer, true, base::unique_fd(), &fence);
-        mCurrentBuffer = mBuffer;
 
         int fd = fence.release();
         if (fd >= 0) {
@@ -397,7 +413,7 @@
 
         ASSERT_EQ(NO_ERROR, status);
         if (layers.size() > 0 && mGLESRE != nullptr) {
-            ASSERT_TRUE(mGLESRE->isFramebufferImageCachedForTesting(mBuffer->getId()));
+            ASSERT_TRUE(mGLESRE->isFramebufferImageCachedForTesting(mBuffer->getBuffer()->getId()));
         }
     }
 
@@ -503,17 +519,11 @@
     void initializeRenderEngine();
 
     std::unique_ptr<renderengine::RenderEngine> mRE;
+    std::shared_ptr<renderengine::ExternalTexture> mBuffer;
     // GLESRenderEngine for testing GLES-specific behavior.
     // Owened by mRE, but this is downcasted.
     renderengine::gl::GLESRenderEngine* mGLESRE = nullptr;
 
-    // Dumb hack to avoid NPE in the EGL driver: the GraphicBuffer needs to
-    // be freed *after* RenderEngine is destroyed, so that the EGL image is
-    // destroyed first.
-    sp<GraphicBuffer> mCurrentBuffer;
-
-    sp<GraphicBuffer> mBuffer;
-
     std::vector<uint32_t> mTexNames;
 };
 
@@ -530,6 +540,7 @@
     } else {
         mRE = renderEngineFactory->createRenderEngine();
     }
+    mBuffer = allocateDefaultBuffer();
 }
 
 struct ColorSourceVariant {
@@ -566,18 +577,18 @@
 struct BufferSourceVariant {
     static void fillColor(renderengine::LayerSettings& layer, half r, half g, half b,
                           RenderEngineTest* fixture) {
-        sp<GraphicBuffer> buf = RenderEngineTest::allocateSourceBuffer(1, 1);
+        const auto buf = fixture->allocateSourceBuffer(1, 1);
         uint32_t texName;
         fixture->mRE->genTextures(1, &texName);
         fixture->mTexNames.push_back(texName);
 
         uint8_t* pixels;
-        buf->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
-                  reinterpret_cast<void**>(&pixels));
+        buf->getBuffer()->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
+                               reinterpret_cast<void**>(&pixels));
 
-        for (int32_t j = 0; j < buf->getHeight(); j++) {
-            uint8_t* iter = pixels + (buf->getStride() * j) * 4;
-            for (int32_t i = 0; i < buf->getWidth(); i++) {
+        for (int32_t j = 0; j < buf->getBuffer()->getHeight(); j++) {
+            uint8_t* iter = pixels + (buf->getBuffer()->getStride() * j) * 4;
+            for (int32_t i = 0; i < buf->getBuffer()->getWidth(); i++) {
                 iter[0] = uint8_t(r * 255);
                 iter[1] = uint8_t(g * 255);
                 iter[2] = uint8_t(b * 255);
@@ -586,7 +597,7 @@
             }
         }
 
-        buf->unlock();
+        buf->getBuffer()->unlock();
 
         layer.source.buffer.buffer = buf;
         layer.source.buffer.textureName = texName;
@@ -1012,14 +1023,14 @@
     layer.sourceDataspace = ui::Dataspace::V0_SRGB_LINEAR;
     // Here will allocate a checker board texture, but transform texture
     // coordinates so that only the upper left is applied.
-    sp<GraphicBuffer> buf = allocateSourceBuffer(2, 2);
+    const auto buf = allocateSourceBuffer(2, 2);
     uint32_t texName;
     RenderEngineTest::mRE->genTextures(1, &texName);
     this->mTexNames.push_back(texName);
 
     uint8_t* pixels;
-    buf->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
-              reinterpret_cast<void**>(&pixels));
+    buf->getBuffer()->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
+                           reinterpret_cast<void**>(&pixels));
     // Red top left, Green top right, Blue bottom left, Black bottom right
     pixels[0] = 255;
     pixels[1] = 0;
@@ -1033,7 +1044,7 @@
     pixels[9] = 0;
     pixels[10] = 255;
     pixels[11] = 255;
-    buf->unlock();
+    buf->getBuffer()->unlock();
 
     layer.source.buffer.buffer = buf;
     layer.source.buffer.textureName = texName;
@@ -1061,19 +1072,19 @@
     std::vector<const renderengine::LayerSettings*> layers;
 
     renderengine::LayerSettings layer;
-    sp<GraphicBuffer> buf = allocateSourceBuffer(1, 1);
+    const auto buf = allocateSourceBuffer(1, 1);
     uint32_t texName;
     RenderEngineTest::mRE->genTextures(1, &texName);
     this->mTexNames.push_back(texName);
 
     uint8_t* pixels;
-    buf->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
-              reinterpret_cast<void**>(&pixels));
+    buf->getBuffer()->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
+                           reinterpret_cast<void**>(&pixels));
     pixels[0] = 255;
     pixels[1] = 0;
     pixels[2] = 0;
     pixels[3] = 255;
-    buf->unlock();
+    buf->getBuffer()->unlock();
 
     layer.source.buffer.buffer = buf;
     layer.source.buffer.textureName = texName;
@@ -1100,19 +1111,19 @@
     std::vector<const renderengine::LayerSettings*> layers;
 
     renderengine::LayerSettings layer;
-    sp<GraphicBuffer> buf = allocateSourceBuffer(1, 1);
+    const auto buf = allocateSourceBuffer(1, 1);
     uint32_t texName;
     RenderEngineTest::mRE->genTextures(1, &texName);
     this->mTexNames.push_back(texName);
 
     uint8_t* pixels;
-    buf->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
-              reinterpret_cast<void**>(&pixels));
+    buf->getBuffer()->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
+                           reinterpret_cast<void**>(&pixels));
     pixels[0] = 255;
     pixels[1] = 0;
     pixels[2] = 0;
     pixels[3] = 255;
-    buf->unlock();
+    buf->getBuffer()->unlock();
 
     layer.source.buffer.buffer = buf;
     layer.source.buffer.textureName = texName;
@@ -1233,8 +1244,7 @@
 }
 
 TEST_P(RenderEngineTest, drawLayers_withoutBuffers_withColorTransform) {
-    const auto& renderEngineFactory = GetParam();
-    mRE = renderEngineFactory->createRenderEngine();
+    initializeRenderEngine();
 
     renderengine::DisplaySettings settings;
     settings.outputDataspace = ui::Dataspace::V0_SRGB_LINEAR;
@@ -1295,7 +1305,6 @@
     layers.push_back(&layer);
 
     status_t status = mRE->drawLayers(settings, layers, mBuffer, true, base::unique_fd(), nullptr);
-    mCurrentBuffer = mBuffer;
     ASSERT_EQ(NO_ERROR, status);
     expectBufferColor(fullscreenRect(), 255, 0, 0, 255);
 }
@@ -1323,9 +1332,8 @@
     layers.push_back(&layer);
 
     status_t status = mRE->drawLayers(settings, layers, mBuffer, false, base::unique_fd(), nullptr);
-    mCurrentBuffer = mBuffer;
     ASSERT_EQ(NO_ERROR, status);
-    ASSERT_FALSE(mGLESRE->isFramebufferImageCachedForTesting(mBuffer->getId()));
+    ASSERT_FALSE(mGLESRE->isFramebufferImageCachedForTesting(mBuffer->getBuffer()->getId()));
     expectBufferColor(fullscreenRect(), 255, 0, 0, 255);
 }
 
@@ -1574,98 +1582,6 @@
     clearRegion();
 }
 
-TEST_P(RenderEngineTest, drawLayers_fillsBufferAndCachesImages) {
-    const auto& renderEngineFactory = GetParam();
-
-    if (renderEngineFactory->type() != renderengine::RenderEngine::RenderEngineType::GLES) {
-        // GLES-specific test
-        return;
-    }
-
-    initializeRenderEngine();
-
-    renderengine::DisplaySettings settings;
-    settings.outputDataspace = ui::Dataspace::V0_SRGB_LINEAR;
-    settings.physicalDisplay = fullscreenRect();
-    settings.clip = fullscreenRect();
-
-    std::vector<const renderengine::LayerSettings*> layers;
-
-    renderengine::LayerSettings layer;
-    layer.geometry.boundaries = fullscreenRect().toFloatRect();
-    BufferSourceVariant<ForceOpaqueBufferVariant>::fillColor(layer, 1.0f, 0.0f, 0.0f, this);
-
-    layers.push_back(&layer);
-    invokeDraw(settings, layers);
-    uint64_t bufferId = layer.source.buffer.buffer->getId();
-    EXPECT_TRUE(mGLESRE->isImageCachedForTesting(bufferId));
-    std::shared_ptr<renderengine::gl::ImageManager::Barrier> barrier =
-            mGLESRE->unbindExternalTextureBufferForTesting(bufferId);
-    std::lock_guard<std::mutex> lock(barrier->mutex);
-    ASSERT_TRUE(barrier->condition.wait_for(barrier->mutex, std::chrono::seconds(5),
-                                            [&]() REQUIRES(barrier->mutex) {
-                                                return barrier->isOpen;
-                                            }));
-    EXPECT_FALSE(mGLESRE->isImageCachedForTesting(bufferId));
-    EXPECT_EQ(NO_ERROR, barrier->result);
-}
-
-TEST_P(RenderEngineTest, cacheExternalBuffer_withNullBuffer) {
-    const auto& renderEngineFactory = GetParam();
-
-    if (renderEngineFactory->type() != renderengine::RenderEngine::RenderEngineType::GLES) {
-        // GLES-specific test
-        return;
-    }
-
-    initializeRenderEngine();
-
-    std::shared_ptr<renderengine::gl::ImageManager::Barrier> barrier =
-            mGLESRE->cacheExternalTextureBufferForTesting(nullptr);
-    std::lock_guard<std::mutex> lock(barrier->mutex);
-    ASSERT_TRUE(barrier->condition.wait_for(barrier->mutex, std::chrono::seconds(5),
-                                            [&]() REQUIRES(barrier->mutex) {
-                                                return barrier->isOpen;
-                                            }));
-    EXPECT_TRUE(barrier->isOpen);
-    EXPECT_EQ(BAD_VALUE, barrier->result);
-}
-
-TEST_P(RenderEngineTest, cacheExternalBuffer_cachesImages) {
-    const auto& renderEngineFactory = GetParam();
-
-    if (renderEngineFactory->type() != renderengine::RenderEngine::RenderEngineType::GLES) {
-        // GLES-specific test
-        return;
-    }
-
-    initializeRenderEngine();
-
-    sp<GraphicBuffer> buf = allocateSourceBuffer(1, 1);
-    uint64_t bufferId = buf->getId();
-    std::shared_ptr<renderengine::gl::ImageManager::Barrier> barrier =
-            mGLESRE->cacheExternalTextureBufferForTesting(buf);
-    {
-        std::lock_guard<std::mutex> lock(barrier->mutex);
-        ASSERT_TRUE(barrier->condition.wait_for(barrier->mutex, std::chrono::seconds(5),
-                                                [&]() REQUIRES(barrier->mutex) {
-                                                    return barrier->isOpen;
-                                                }));
-        EXPECT_EQ(NO_ERROR, barrier->result);
-    }
-    EXPECT_TRUE(mGLESRE->isImageCachedForTesting(bufferId));
-    barrier = mGLESRE->unbindExternalTextureBufferForTesting(bufferId);
-    {
-        std::lock_guard<std::mutex> lock(barrier->mutex);
-        ASSERT_TRUE(barrier->condition.wait_for(barrier->mutex, std::chrono::seconds(5),
-                                                [&]() REQUIRES(barrier->mutex) {
-                                                    return barrier->isOpen;
-                                                }));
-        EXPECT_EQ(NO_ERROR, barrier->result);
-    }
-    EXPECT_FALSE(mGLESRE->isImageCachedForTesting(bufferId));
-}
-
 TEST_P(RenderEngineTest, drawLayers_fillShadow_castsWithoutCasterLayer) {
     initializeRenderEngine();
 
@@ -1858,7 +1774,7 @@
         sync_wait(fd, -1);
     }
 
-    uint64_t bufferId = layer.source.buffer.buffer->getId();
+    uint64_t bufferId = layer.source.buffer.buffer->getBuffer()->getId();
     uint32_t texName = layer.source.buffer.textureName;
     EXPECT_TRUE(mGLESRE->isImageCachedForTesting(bufferId));
     EXPECT_EQ(bufferId, mGLESRE->getBufferIdForTextureNameForTesting(texName));
@@ -1966,16 +1882,16 @@
 
     // The next layer will overwrite redLayer with a GraphicBuffer that is green
     // applied with a translucent alpha.
-    auto buf = allocateSourceBuffer(1, 1);
+    const auto buf = allocateSourceBuffer(1, 1);
     {
         uint8_t* pixels;
-        buf->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
-                  reinterpret_cast<void**>(&pixels));
+        buf->getBuffer()->lock(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN,
+                               reinterpret_cast<void**>(&pixels));
         pixels[0] = 0;
         pixels[1] = 255;
         pixels[2] = 0;
         pixels[3] = 255;
-        buf->unlock();
+        buf->getBuffer()->unlock();
     }
 
     const renderengine::LayerSettings greenLayer{
diff --git a/libs/renderengine/tests/RenderEngineThreadedTest.cpp b/libs/renderengine/tests/RenderEngineThreadedTest.cpp
index b093e88..e3917cc 100644
--- a/libs/renderengine/tests/RenderEngineThreadedTest.cpp
+++ b/libs/renderengine/tests/RenderEngineThreadedTest.cpp
@@ -162,15 +162,18 @@
 TEST_F(RenderEngineThreadedTest, drawLayers) {
     renderengine::DisplaySettings settings;
     std::vector<const renderengine::LayerSettings*> layers;
-    sp<GraphicBuffer> buffer = new GraphicBuffer();
+    std::shared_ptr<renderengine::ExternalTexture> buffer = std::make_shared<
+            renderengine::ExternalTexture>(new GraphicBuffer(), *mRenderEngine,
+                                           renderengine::ExternalTexture::Usage::READABLE |
+                                                   renderengine::ExternalTexture::Usage::WRITEABLE);
     base::unique_fd bufferFence;
     base::unique_fd drawFence;
 
     EXPECT_CALL(*mRenderEngine, drawLayers)
             .WillOnce([](const renderengine::DisplaySettings&,
                          const std::vector<const renderengine::LayerSettings*>&,
-                         const sp<GraphicBuffer>&, const bool, base::unique_fd&&,
-                         base::unique_fd*) -> status_t { return NO_ERROR; });
+                         const std::shared_ptr<renderengine::ExternalTexture>&, const bool,
+                         base::unique_fd&&, base::unique_fd*) -> status_t { return NO_ERROR; });
 
     status_t result = mThreadedRE->drawLayers(settings, layers, buffer, false,
                                               std::move(bufferFence), &drawFence);
diff --git a/libs/renderengine/threaded/RenderEngineThreaded.cpp b/libs/renderengine/threaded/RenderEngineThreaded.cpp
index 190662b..c87a836 100644
--- a/libs/renderengine/threaded/RenderEngineThreaded.cpp
+++ b/libs/renderengine/threaded/RenderEngineThreaded.cpp
@@ -160,27 +160,28 @@
     resultFuture.wait();
 }
 
-void RenderEngineThreaded::cacheExternalTextureBuffer(const sp<GraphicBuffer>& buffer) {
+void RenderEngineThreaded::mapExternalTextureBuffer(const sp<GraphicBuffer>& buffer,
+                                                    bool isRenderable) {
     // This function is designed so it can run asynchronously, so we do not need to wait
     // for the futures.
     {
         std::lock_guard lock(mThreadMutex);
         mFunctionCalls.push([=](renderengine::RenderEngine& instance) {
-            ATRACE_NAME("REThreaded::cacheExternalTextureBuffer");
-            instance.cacheExternalTextureBuffer(buffer);
+            ATRACE_NAME("REThreaded::mapExternalTextureBuffer");
+            instance.mapExternalTextureBuffer(buffer, isRenderable);
         });
     }
     mCondition.notify_one();
 }
 
-void RenderEngineThreaded::unbindExternalTextureBuffer(uint64_t bufferId) {
+void RenderEngineThreaded::unmapExternalTextureBuffer(const sp<GraphicBuffer>& buffer) {
     // This function is designed so it can run asynchronously, so we do not need to wait
     // for the futures.
     {
         std::lock_guard lock(mThreadMutex);
         mFunctionCalls.push([=](renderengine::RenderEngine& instance) {
-            ATRACE_NAME("REThreaded::unbindExternalTextureBuffer");
-            instance.unbindExternalTextureBuffer(bufferId);
+            ATRACE_NAME("REThreaded::unmapExternalTextureBuffer");
+            instance.unmapExternalTextureBuffer(buffer);
         });
     }
     mCondition.notify_one();
@@ -242,7 +243,7 @@
 
 status_t RenderEngineThreaded::drawLayers(const DisplaySettings& display,
                                           const std::vector<const LayerSettings*>& layers,
-                                          const sp<GraphicBuffer>& buffer,
+                                          const std::shared_ptr<ExternalTexture>& buffer,
                                           const bool useFramebufferCache,
                                           base::unique_fd&& bufferFence,
                                           base::unique_fd* drawFence) {
diff --git a/libs/renderengine/threaded/RenderEngineThreaded.h b/libs/renderengine/threaded/RenderEngineThreaded.h
index 61ae9b8..eb6098e 100644
--- a/libs/renderengine/threaded/RenderEngineThreaded.h
+++ b/libs/renderengine/threaded/RenderEngineThreaded.h
@@ -48,8 +48,6 @@
 
     void genTextures(size_t count, uint32_t* names) override;
     void deleteTextures(size_t count, uint32_t const* names) override;
-    void cacheExternalTextureBuffer(const sp<GraphicBuffer>& buffer) override;
-    void unbindExternalTextureBuffer(uint64_t bufferId) override;
     size_t getMaxTextureSize() const override;
     size_t getMaxViewportDims() const override;
 
@@ -60,14 +58,19 @@
 
     status_t drawLayers(const DisplaySettings& display,
                         const std::vector<const LayerSettings*>& layers,
-                        const sp<GraphicBuffer>& buffer, const bool useFramebufferCache,
-                        base::unique_fd&& bufferFence, base::unique_fd* drawFence) override;
+                        const std::shared_ptr<ExternalTexture>& buffer,
+                        const bool useFramebufferCache, base::unique_fd&& bufferFence,
+                        base::unique_fd* drawFence) override;
 
     void cleanFramebufferCache() override;
     int getContextPriority() override;
     bool supportsBackgroundBlur() override;
     void onPrimaryDisplaySizeChanged(ui::Size size) override;
 
+protected:
+    void mapExternalTextureBuffer(const sp<GraphicBuffer>& buffer, bool isRenderable) override;
+    void unmapExternalTextureBuffer(const sp<GraphicBuffer>& buffer) override;
+
 private:
     void threadMain(CreateInstanceFactory factory);
     void waitUntilInitialized() const;
diff --git a/services/surfaceflinger/BufferLayer.cpp b/services/surfaceflinger/BufferLayer.cpp
index 12f63db..d243989 100644
--- a/services/surfaceflinger/BufferLayer.cpp
+++ b/services/surfaceflinger/BufferLayer.cpp
@@ -188,7 +188,7 @@
     const bool blackOutLayer = (isProtected() && !targetSettings.supportsProtectedContent) ||
             (isSecure() && !targetSettings.isSecure);
     const bool bufferCanBeUsedAsHwTexture =
-            mBufferInfo.mBuffer->getUsage() & GraphicBuffer::USAGE_HW_TEXTURE;
+            mBufferInfo.mBuffer->getBuffer()->getUsage() & GraphicBuffer::USAGE_HW_TEXTURE;
     compositionengine::LayerFE::LayerSettings& layer = *result;
     if (blackOutLayer || !bufferCanBeUsedAsHwTexture) {
         ALOGE_IF(!bufferCanBeUsedAsHwTexture, "%s is blacked out as buffer is not gpu readable",
@@ -213,7 +213,7 @@
             ? mBufferInfo.mHdrMetadata.cta8613.maxContentLightLevel
             : defaultMaxContentLuminance;
     layer.frameNumber = mCurrentFrameNumber;
-    layer.bufferId = mBufferInfo.mBuffer ? mBufferInfo.mBuffer->getId() : 0;
+    layer.bufferId = mBufferInfo.mBuffer ? mBufferInfo.mBuffer->getBuffer()->getId() : 0;
 
     const bool useFiltering =
             targetSettings.needsFiltering || mNeedsFiltering || bufferNeedsFiltering();
@@ -314,7 +314,7 @@
                 : Hwc2::IComposerClient::Composition::DEVICE;
     }
 
-    compositionState->buffer = mBufferInfo.mBuffer;
+    compositionState->buffer = mBufferInfo.mBuffer->getBuffer();
     compositionState->bufferSlot = (mBufferInfo.mBufferSlot == BufferQueue::INVALID_BUFFER_SLOT)
             ? 0
             : mBufferInfo.mBufferSlot;
@@ -442,7 +442,7 @@
 
 void BufferLayer::gatherBufferInfo() {
     mBufferInfo.mPixelFormat =
-            !mBufferInfo.mBuffer ? PIXEL_FORMAT_NONE : mBufferInfo.mBuffer->format;
+            !mBufferInfo.mBuffer ? PIXEL_FORMAT_NONE : mBufferInfo.mBuffer->getBuffer()->format;
     mBufferInfo.mFrameLatencyNeeded = true;
 }
 
@@ -539,10 +539,10 @@
     }
 
     if (oldBufferInfo.mBuffer != nullptr) {
-        uint32_t bufWidth = mBufferInfo.mBuffer->getWidth();
-        uint32_t bufHeight = mBufferInfo.mBuffer->getHeight();
-        if (bufWidth != uint32_t(oldBufferInfo.mBuffer->width) ||
-            bufHeight != uint32_t(oldBufferInfo.mBuffer->height)) {
+        uint32_t bufWidth = mBufferInfo.mBuffer->getBuffer()->getWidth();
+        uint32_t bufHeight = mBufferInfo.mBuffer->getBuffer()->getHeight();
+        if (bufWidth != uint32_t(oldBufferInfo.mBuffer->getBuffer()->width) ||
+            bufHeight != uint32_t(oldBufferInfo.mBuffer->getBuffer()->height)) {
             recomputeVisibleRegions = true;
         }
     }
@@ -563,8 +563,8 @@
 }
 
 bool BufferLayer::isProtected() const {
-    const sp<GraphicBuffer>& buffer(mBufferInfo.mBuffer);
-    return (buffer != 0) && (buffer->getUsage() & GRALLOC_USAGE_PROTECTED);
+    return (mBufferInfo.mBuffer != nullptr) &&
+            (mBufferInfo.mBuffer->getBuffer()->getUsage() & GRALLOC_USAGE_PROTECTED);
 }
 
 // As documented in libhardware header, formats in the range
@@ -651,8 +651,8 @@
         return Rect::INVALID_RECT;
     }
 
-    uint32_t bufWidth = mBufferInfo.mBuffer->getWidth();
-    uint32_t bufHeight = mBufferInfo.mBuffer->getHeight();
+    uint32_t bufWidth = mBufferInfo.mBuffer->getBuffer()->getWidth();
+    uint32_t bufHeight = mBufferInfo.mBuffer->getBuffer()->getHeight();
 
     // Undo any transformations on the buffer and return the result.
     if (mBufferInfo.mTransform & ui::Transform::ROT_90) {
@@ -683,8 +683,8 @@
         return parentBounds;
     }
 
-    uint32_t bufWidth = mBufferInfo.mBuffer->getWidth();
-    uint32_t bufHeight = mBufferInfo.mBuffer->getHeight();
+    uint32_t bufWidth = mBufferInfo.mBuffer->getBuffer()->getWidth();
+    uint32_t bufHeight = mBufferInfo.mBuffer->getBuffer()->getHeight();
 
     // Undo any transformations on the buffer and return the result.
     if (mBufferInfo.mTransform & ui::Transform::ROT_90) {
@@ -726,7 +726,7 @@
         return mBufferInfo.mCrop;
     } else if (mBufferInfo.mBuffer != nullptr) {
         // otherwise we use the whole buffer
-        return mBufferInfo.mBuffer->getBounds();
+        return mBufferInfo.mBuffer->getBuffer()->getBounds();
     } else {
         // if we don't have a buffer yet, we use an empty/invalid crop
         return Rect();
@@ -771,12 +771,14 @@
 }
 
 sp<GraphicBuffer> BufferLayer::getBuffer() const {
-    return mBufferInfo.mBuffer;
+    return mBufferInfo.mBuffer ? mBufferInfo.mBuffer->getBuffer() : nullptr;
 }
 
 void BufferLayer::getDrawingTransformMatrix(bool filteringEnabled, float outMatrix[16]) {
-    GLConsumer::computeTransformMatrix(outMatrix, mBufferInfo.mBuffer, mBufferInfo.mCrop,
-                                       mBufferInfo.mTransform, filteringEnabled);
+    GLConsumer::computeTransformMatrix(outMatrix,
+                                       mBufferInfo.mBuffer ? mBufferInfo.mBuffer->getBuffer()
+                                                           : nullptr,
+                                       mBufferInfo.mCrop, mBufferInfo.mTransform, filteringEnabled);
 }
 
 void BufferLayer::setInitialValuesForClone(const sp<Layer>& clonedFrom) {
diff --git a/services/surfaceflinger/BufferLayer.h b/services/surfaceflinger/BufferLayer.h
index 0a5235a..cd3d80e 100644
--- a/services/surfaceflinger/BufferLayer.h
+++ b/services/surfaceflinger/BufferLayer.h
@@ -132,7 +132,7 @@
         PixelFormat mPixelFormat{PIXEL_FORMAT_NONE};
         bool mTransformToDisplayInverse{false};
 
-        sp<GraphicBuffer> mBuffer;
+        std::shared_ptr<renderengine::ExternalTexture> mBuffer;
         int mBufferSlot{BufferQueue::INVALID_BUFFER_SLOT};
 
         bool mFrameLatencyNeeded{false};
diff --git a/services/surfaceflinger/BufferLayerConsumer.cpp b/services/surfaceflinger/BufferLayerConsumer.cpp
index 69d2d11..96b2247 100644
--- a/services/surfaceflinger/BufferLayerConsumer.cpp
+++ b/services/surfaceflinger/BufferLayerConsumer.cpp
@@ -40,7 +40,6 @@
 #include <gui/ISurfaceComposer.h>
 #include <gui/SurfaceComposerClient.h>
 #include <private/gui/ComposerService.h>
-#include <renderengine/Image.h>
 #include <renderengine/RenderEngine.h>
 #include <utils/Log.h>
 #include <utils/String8.h>
@@ -167,7 +166,7 @@
     }
 
     auto buffer = mPendingRelease.isPending ? mPendingRelease.graphicBuffer
-                                            : mCurrentTextureBuffer->graphicBuffer();
+                                            : mCurrentTextureBuffer->getBuffer();
     auto err = addReleaseFence(slot, buffer, fence);
     if (err != OK) {
         BLC_LOGE("setReleaseFence: failed to add the fence: %s (%d)", strerror(-err), err);
@@ -206,9 +205,11 @@
     // before, so we need to clean up old references.
     if (item->mGraphicBuffer != nullptr) {
         std::lock_guard<std::mutex> lock(mImagesMutex);
-        if (mImages[item->mSlot] == nullptr || mImages[item->mSlot]->graphicBuffer() == nullptr ||
-            mImages[item->mSlot]->graphicBuffer()->getId() != item->mGraphicBuffer->getId()) {
-            mImages[item->mSlot] = std::make_shared<Image>(item->mGraphicBuffer, mRE);
+        if (mImages[item->mSlot] == nullptr || mImages[item->mSlot]->getBuffer() == nullptr ||
+            mImages[item->mSlot]->getBuffer()->getId() != item->mGraphicBuffer->getId()) {
+            mImages[item->mSlot] = std::make_shared<
+                    renderengine::ExternalTexture>(item->mGraphicBuffer, mRE,
+                                                   renderengine::ExternalTexture::Usage::READABLE);
         }
     }
 
@@ -222,8 +223,8 @@
     int slot = item.mSlot;
 
     BLC_LOGV("updateAndRelease: (slot=%d buf=%p) -> (slot=%d buf=%p)", mCurrentTexture,
-             (mCurrentTextureBuffer != nullptr && mCurrentTextureBuffer->graphicBuffer() != nullptr)
-                     ? mCurrentTextureBuffer->graphicBuffer()->handle
+             (mCurrentTextureBuffer != nullptr && mCurrentTextureBuffer->getBuffer() != nullptr)
+                     ? mCurrentTextureBuffer->getBuffer()->handle
                      : 0,
              slot, mSlots[slot].mGraphicBuffer->handle);
 
@@ -231,7 +232,7 @@
     // releaseBufferLocked() if we're in shared buffer mode and both buffers are
     // the same.
 
-    std::shared_ptr<Image> nextTextureBuffer;
+    std::shared_ptr<renderengine::ExternalTexture> nextTextureBuffer;
     {
         std::lock_guard<std::mutex> lock(mImagesMutex);
         nextTextureBuffer = mImages[slot];
@@ -241,7 +242,7 @@
     if (mCurrentTexture != BufferQueue::INVALID_BUFFER_SLOT) {
         if (pendingRelease == nullptr) {
             status_t status =
-                    releaseBufferLocked(mCurrentTexture, mCurrentTextureBuffer->graphicBuffer());
+                    releaseBufferLocked(mCurrentTexture, mCurrentTextureBuffer->getBuffer());
             if (status < NO_ERROR) {
                 BLC_LOGE("updateAndRelease: failed to release buffer: %s (%d)", strerror(-status),
                          status);
@@ -250,7 +251,7 @@
             }
         } else {
             pendingRelease->currentTexture = mCurrentTexture;
-            pendingRelease->graphicBuffer = mCurrentTextureBuffer->graphicBuffer();
+            pendingRelease->graphicBuffer = mCurrentTextureBuffer->getBuffer();
             pendingRelease->isPending = true;
         }
     }
@@ -301,14 +302,14 @@
 
 void BufferLayerConsumer::computeCurrentTransformMatrixLocked() {
     BLC_LOGV("computeCurrentTransformMatrixLocked");
-    if (mCurrentTextureBuffer == nullptr || mCurrentTextureBuffer->graphicBuffer() == nullptr) {
+    if (mCurrentTextureBuffer == nullptr || mCurrentTextureBuffer->getBuffer() == nullptr) {
         BLC_LOGD("computeCurrentTransformMatrixLocked: "
                  "mCurrentTextureBuffer is nullptr");
     }
     GLConsumer::computeTransformMatrix(mCurrentTransformMatrix,
                                        mCurrentTextureBuffer == nullptr
                                                ? nullptr
-                                               : mCurrentTextureBuffer->graphicBuffer(),
+                                               : mCurrentTextureBuffer->getBuffer(),
                                        getCurrentCropLocked(), mCurrentTransform,
                                        mFilteringEnabled);
 }
@@ -360,7 +361,8 @@
     return mCurrentApi;
 }
 
-sp<GraphicBuffer> BufferLayerConsumer::getCurrentBuffer(int* outSlot, sp<Fence>* outFence) const {
+std::shared_ptr<renderengine::ExternalTexture> BufferLayerConsumer::getCurrentBuffer(
+        int* outSlot, sp<Fence>* outFence) const {
     Mutex::Autolock lock(mMutex);
 
     if (outSlot != nullptr) {
@@ -371,7 +373,7 @@
         *outFence = mCurrentFence;
     }
 
-    return mCurrentTextureBuffer == nullptr ? nullptr : mCurrentTextureBuffer->graphicBuffer();
+    return mCurrentTextureBuffer == nullptr ? nullptr : mCurrentTextureBuffer;
 }
 
 Rect BufferLayerConsumer::getCurrentCrop() const {
@@ -456,10 +458,12 @@
 void BufferLayerConsumer::onBufferAvailable(const BufferItem& item) {
     if (item.mGraphicBuffer != nullptr && item.mSlot != BufferQueue::INVALID_BUFFER_SLOT) {
         std::lock_guard<std::mutex> lock(mImagesMutex);
-        const std::shared_ptr<Image>& oldImage = mImages[item.mSlot];
-        if (oldImage == nullptr || oldImage->graphicBuffer() == nullptr ||
-            oldImage->graphicBuffer()->getId() != item.mGraphicBuffer->getId()) {
-            mImages[item.mSlot] = std::make_shared<Image>(item.mGraphicBuffer, mRE);
+        const std::shared_ptr<renderengine::ExternalTexture>& oldImage = mImages[item.mSlot];
+        if (oldImage == nullptr || oldImage->getBuffer() == nullptr ||
+            oldImage->getBuffer()->getId() != item.mGraphicBuffer->getId()) {
+            mImages[item.mSlot] = std::make_shared<
+                    renderengine::ExternalTexture>(item.mGraphicBuffer, mRE,
+                                                   renderengine::ExternalTexture::Usage::READABLE);
         }
     }
 }
@@ -499,22 +503,6 @@
 
     ConsumerBase::dumpLocked(result, prefix);
 }
-
-BufferLayerConsumer::Image::Image(const sp<GraphicBuffer>& graphicBuffer,
-                                  renderengine::RenderEngine& engine)
-      : mGraphicBuffer(graphicBuffer), mRE(engine) {
-    if (graphicBuffer != nullptr && (graphicBuffer->getUsage() & GRALLOC_USAGE_PROTECTED)) {
-        return;
-    }
-    mRE.cacheExternalTextureBuffer(mGraphicBuffer);
-}
-
-BufferLayerConsumer::Image::~Image() {
-    if (mGraphicBuffer != nullptr) {
-        ALOGV("Destroying buffer: %" PRId64, mGraphicBuffer->getId());
-        mRE.unbindExternalTextureBuffer(mGraphicBuffer->getId());
-    }
-}
 }; // namespace android
 
 // TODO(b/129481165): remove the #pragma below and fix conversion issues
diff --git a/services/surfaceflinger/BufferLayerConsumer.h b/services/surfaceflinger/BufferLayerConsumer.h
index dd39214..9ed80b4 100644
--- a/services/surfaceflinger/BufferLayerConsumer.h
+++ b/services/surfaceflinger/BufferLayerConsumer.h
@@ -21,12 +21,11 @@
 #include <gui/BufferQueueDefs.h>
 #include <gui/ConsumerBase.h>
 #include <gui/HdrMetadata.h>
-
+#include <renderengine/ExternalTexture.h>
 #include <ui/FenceTime.h>
 #include <ui/GraphicBuffer.h>
 #include <ui/GraphicTypes.h>
 #include <ui/Region.h>
-
 #include <utils/String8.h>
 #include <utils/Vector.h>
 #include <utils/threads.h>
@@ -39,7 +38,6 @@
 
 namespace renderengine {
 class RenderEngine;
-class Image;
 } // namespace renderengine
 
 /*
@@ -153,7 +151,8 @@
     // When outSlot is not nullptr, the current buffer slot index is also
     // returned. Simiarly, when outFence is not nullptr, the current output
     // fence is returned.
-    sp<GraphicBuffer> getCurrentBuffer(int* outSlot = nullptr, sp<Fence>* outFence = nullptr) const;
+    std::shared_ptr<renderengine::ExternalTexture> getCurrentBuffer(
+            int* outSlot = nullptr, sp<Fence>* outFence = nullptr) const;
 
     // getCurrentCrop returns the cropping rectangle of the current buffer.
     Rect getCurrentCrop() const;
@@ -258,7 +257,7 @@
     // mCurrentTextureBuffer is the buffer containing the current texture. It's
     // possible that this buffer is not associated with any buffer slot, so we
     // must track it separately in order to support the getCurrentBuffer method.
-    std::shared_ptr<Image> mCurrentTextureBuffer;
+    std::shared_ptr<renderengine::ExternalTexture> mCurrentTextureBuffer;
 
     // mCurrentCrop is the crop rectangle that applies to the current texture.
     // It gets set each time updateTexImage is called.
@@ -337,7 +336,8 @@
     int mCurrentTexture;
 
     // Shadow buffer cache for cleaning up renderengine references.
-    std::shared_ptr<Image> mImages[BufferQueueDefs::NUM_BUFFER_SLOTS] GUARDED_BY(mImagesMutex);
+    std::shared_ptr<renderengine::ExternalTexture>
+            mImages[BufferQueueDefs::NUM_BUFFER_SLOTS] GUARDED_BY(mImagesMutex);
 
     // Separate mutex guarding the shadow buffer cache.
     // mImagesMutex can be manipulated with binder threads (e.g. onBuffersAllocated)
diff --git a/services/surfaceflinger/BufferStateLayer.cpp b/services/surfaceflinger/BufferStateLayer.cpp
index fa9cecf..ac2edbe 100644
--- a/services/surfaceflinger/BufferStateLayer.cpp
+++ b/services/surfaceflinger/BufferStateLayer.cpp
@@ -71,14 +71,8 @@
     // original layer and the clone should be removed at the same time so there shouldn't be any
     // issue with the clone layer trying to use the texture.
     if (mBufferInfo.mBuffer != nullptr && !isClone()) {
-        // Ensure that mBuffer is uncached from RenderEngine here, as
-        // RenderEngine may have been using the buffer as an external texture
-        // after the client uncached the buffer.
-        auto& engine(mFlinger->getRenderEngine());
-        const uint64_t bufferId = mBufferInfo.mBuffer->getId();
-        engine.unbindExternalTextureBuffer(bufferId);
-        callReleaseBufferCallback(mDrawingState.releaseBufferListener, mBufferInfo.mBuffer,
-                                  mBufferInfo.mFence);
+        callReleaseBufferCallback(mDrawingState.releaseBufferListener,
+                                  mBufferInfo.mBuffer->getBuffer(), mBufferInfo.mFence);
     }
 }
 
@@ -344,8 +338,9 @@
     return true;
 }
 
-bool BufferStateLayer::setBuffer(const sp<GraphicBuffer>& buffer, const sp<Fence>& acquireFence,
-                                 nsecs_t postTime, nsecs_t desiredPresentTime, bool isAutoTimestamp,
+bool BufferStateLayer::setBuffer(const std::shared_ptr<renderengine::ExternalTexture>& buffer,
+                                 const sp<Fence>& acquireFence, nsecs_t postTime,
+                                 nsecs_t desiredPresentTime, bool isAutoTimestamp,
                                  const client_cache_t& clientCacheId, uint64_t frameNumber,
                                  std::optional<nsecs_t> dequeueTime, const FrameTimelineInfo& info,
                                  const sp<ITransactionCompletedListener>& releaseBufferListener) {
@@ -353,12 +348,14 @@
 
     if (mCurrentState.buffer) {
         mReleasePreviousBuffer = true;
-        if (mCurrentState.buffer != mDrawingState.buffer) {
+        if (!mDrawingState.buffer ||
+            mCurrentState.buffer->getBuffer() != mDrawingState.buffer->getBuffer()) {
             // If mCurrentState has a buffer, and we are about to update again
             // before swapping to drawing state, then the first buffer will be
             // dropped and we should decrement the pending buffer count and
             // call any release buffer callbacks if set.
-            callReleaseBufferCallback(mCurrentState.releaseBufferListener, mCurrentState.buffer,
+            callReleaseBufferCallback(mCurrentState.releaseBufferListener,
+                                      mCurrentState.buffer->getBuffer(),
                                       mCurrentState.acquireFence);
             decrementPendingBufferCount();
             if (mCurrentState.bufferSurfaceFrameTX != nullptr) {
@@ -396,8 +393,8 @@
 
     setFrameTimelineVsyncForBufferTransaction(info, postTime);
 
-    if (dequeueTime && *dequeueTime != 0) {
-        const uint64_t bufferId = buffer->getId();
+    if (buffer && dequeueTime && *dequeueTime != 0) {
+        const uint64_t bufferId = buffer->getBuffer()->getId();
         mFlinger->mFrameTracer->traceNewLayer(layerId, getName().c_str());
         mFlinger->mFrameTracer->traceTimestamp(layerId, bufferId, frameNumber, *dequeueTime,
                                                FrameTracer::FrameEvent::DEQUEUE);
@@ -405,8 +402,8 @@
                                                FrameTracer::FrameEvent::QUEUE);
     }
 
-    mCurrentState.width = mCurrentState.buffer->width;
-    mCurrentState.height = mCurrentState.buffer->height;
+    mCurrentState.width = mCurrentState.buffer->getBuffer()->getWidth();
+    mCurrentState.height = mCurrentState.buffer->getBuffer()->getHeight();
 
     return true;
 }
@@ -655,7 +652,7 @@
     }
 
     const int32_t layerId = getSequence();
-    const uint64_t bufferId = mDrawingState.buffer->getId();
+    const uint64_t bufferId = mDrawingState.buffer->getBuffer()->getId();
     const uint64_t frameNumber = mDrawingState.frameNumber;
     const auto acquireFence = std::make_shared<FenceTime>(mDrawingState.acquireFence);
     mFlinger->mTimeStats->setAcquireFence(layerId, frameNumber, acquireFence);
@@ -689,7 +686,7 @@
         return BAD_VALUE;
     }
 
-    if (s.buffer != mBufferInfo.mBuffer) {
+    if (!mBufferInfo.mBuffer || s.buffer->getBuffer() != mBufferInfo.mBuffer->getBuffer()) {
         decrementPendingBufferCount();
     }
 
@@ -808,13 +805,13 @@
 
 Rect BufferStateLayer::computeCrop(const State& s) {
     if (s.crop.isEmpty() && s.buffer) {
-        return s.buffer->getBounds();
+        return s.buffer->getBuffer()->getBounds();
     } else if (s.buffer) {
         Rect crop = s.crop;
         crop.left = std::max(crop.left, 0);
         crop.top = std::max(crop.top, 0);
-        uint32_t bufferWidth = s.buffer->getWidth();
-        uint32_t bufferHeight = s.buffer->getHeight();
+        uint32_t bufferWidth = s.buffer->getBuffer()->getWidth();
+        uint32_t bufferHeight = s.buffer->getBuffer()->getHeight();
         if (bufferHeight <= std::numeric_limits<int32_t>::max() &&
             bufferWidth <= std::numeric_limits<int32_t>::max()) {
             crop.right = std::min(crop.right, static_cast<int32_t>(bufferWidth));
@@ -822,7 +819,7 @@
         }
         if (!crop.isValid()) {
             // Crop rect is out of bounds, return whole buffer
-            return s.buffer->getBounds();
+            return s.buffer->getBuffer()->getBounds();
         }
         return crop;
     }
@@ -844,8 +841,8 @@
         return false;
     }
 
-    uint32_t bufferWidth = s.buffer->width;
-    uint32_t bufferHeight = s.buffer->height;
+    uint32_t bufferWidth = s.buffer->getBuffer()->width;
+    uint32_t bufferHeight = s.buffer->getBuffer()->height;
 
     // Undo any transformations on the buffer and return the result.
     if (s.bufferTransform & ui::Transform::ROT_90) {
@@ -872,14 +869,16 @@
     ATRACE_INT(mBlastTransactionName.c_str(), pendingBuffers);
 }
 
-void BufferStateLayer::bufferMayChange(sp<GraphicBuffer>& newBuffer) {
-    if (mDrawingState.buffer != nullptr && mDrawingState.buffer != mBufferInfo.mBuffer &&
-        newBuffer != mDrawingState.buffer) {
+void BufferStateLayer::bufferMayChange(const sp<GraphicBuffer>& newBuffer) {
+    if (mDrawingState.buffer != nullptr &&
+        (!mBufferInfo.mBuffer ||
+         mDrawingState.buffer->getBuffer() != mBufferInfo.mBuffer->getBuffer()) &&
+        newBuffer != mDrawingState.buffer->getBuffer()) {
         // If we are about to update mDrawingState.buffer but it has not yet latched
         // then we will drop a buffer and should decrement the pending buffer count and
         // call any release buffer callbacks if set.
-        callReleaseBufferCallback(mDrawingState.releaseBufferListener, mDrawingState.buffer,
-                                  mDrawingState.acquireFence);
+        callReleaseBufferCallback(mDrawingState.releaseBufferListener,
+                                  mDrawingState.buffer->getBuffer(), mDrawingState.acquireFence);
         decrementPendingBufferCount();
     }
 }
diff --git a/services/surfaceflinger/BufferStateLayer.h b/services/surfaceflinger/BufferStateLayer.h
index 24e0ad2..4171092 100644
--- a/services/surfaceflinger/BufferStateLayer.h
+++ b/services/surfaceflinger/BufferStateLayer.h
@@ -62,9 +62,9 @@
     bool setTransform(uint32_t transform) override;
     bool setTransformToDisplayInverse(bool transformToDisplayInverse) override;
     bool setCrop(const Rect& crop) override;
-    bool setBuffer(const sp<GraphicBuffer>& buffer, const sp<Fence>& acquireFence, nsecs_t postTime,
-                   nsecs_t desiredPresentTime, bool isAutoTimestamp,
-                   const client_cache_t& clientCacheId, uint64_t frameNumber,
+    bool setBuffer(const std::shared_ptr<renderengine::ExternalTexture>& buffer,
+                   const sp<Fence>& acquireFence, nsecs_t postTime, nsecs_t desiredPresentTime,
+                   bool isAutoTimestamp, const client_cache_t& clientCacheId, uint64_t frameNumber,
                    std::optional<nsecs_t> dequeueTime, const FrameTimelineInfo& info,
                    const sp<ITransactionCompletedListener>& transactionListener) override;
     bool setAcquireFence(const sp<Fence>& fence) override;
@@ -100,7 +100,7 @@
 
     // See mPendingBufferTransactions
     void decrementPendingBufferCount();
-    void bufferMayChange(sp<GraphicBuffer>& newBuffer) override;
+    void bufferMayChange(const sp<GraphicBuffer>& newBuffer) override;
     std::atomic<int32_t>* getPendingBufferCounter() override { return &mPendingBufferTransactions; }
     std::string getPendingBufferCounterName() override { return mBlastTransactionName; }
 
diff --git a/services/surfaceflinger/ClientCache.cpp b/services/surfaceflinger/ClientCache.cpp
index 44b33ef..f310738 100644
--- a/services/surfaceflinger/ClientCache.cpp
+++ b/services/surfaceflinger/ClientCache.cpp
@@ -102,7 +102,12 @@
         return false;
     }
 
-    processBuffers[id].buffer = buffer;
+    LOG_ALWAYS_FATAL_IF(mRenderEngine == nullptr,
+                        "Attempted to build the ClientCache before a RenderEngine instance was "
+                        "ready!");
+    processBuffers[id].buffer = std::make_shared<
+            renderengine::ExternalTexture>(buffer, *mRenderEngine,
+                                           renderengine::ExternalTexture::Usage::READABLE);
     return true;
 }
 
@@ -132,7 +137,7 @@
     }
 }
 
-sp<GraphicBuffer> ClientCache::get(const client_cache_t& cacheId) {
+std::shared_ptr<renderengine::ExternalTexture> ClientCache::get(const client_cache_t& cacheId) {
     std::lock_guard lock(mMutex);
 
     ClientCacheBuffer* buf = nullptr;
@@ -213,8 +218,8 @@
         auto &buffers = i.second.second;
         for (auto& [id, clientCacheBuffer] : buffers) {
             StringAppendF(&result, "\t ID: %d, Width/Height: %d,%d\n", (int)id,
-                          (int)clientCacheBuffer.buffer->getWidth(),
-                          (int)clientCacheBuffer.buffer->getHeight());
+                          (int)clientCacheBuffer.buffer->getBuffer()->getWidth(),
+                          (int)clientCacheBuffer.buffer->getBuffer()->getHeight());
         }
     }
 }
diff --git a/services/surfaceflinger/ClientCache.h b/services/surfaceflinger/ClientCache.h
index 0d597c8..a9b8177 100644
--- a/services/surfaceflinger/ClientCache.h
+++ b/services/surfaceflinger/ClientCache.h
@@ -19,6 +19,7 @@
 #include <android-base/thread_annotations.h>
 #include <binder/IBinder.h>
 #include <gui/LayerState.h>
+#include <renderengine/RenderEngine.h>
 #include <ui/GraphicBuffer.h>
 #include <utils/RefBase.h>
 #include <utils/Singleton.h>
@@ -39,7 +40,11 @@
     bool add(const client_cache_t& cacheId, const sp<GraphicBuffer>& buffer);
     void erase(const client_cache_t& cacheId);
 
-    sp<GraphicBuffer> get(const client_cache_t& cacheId);
+    std::shared_ptr<renderengine::ExternalTexture> get(const client_cache_t& cacheId);
+
+    // Always called immediately after setup. Will be set to non-null, and then should never be
+    // called again.
+    void setRenderEngine(renderengine::RenderEngine* renderEngine) { mRenderEngine = renderEngine; }
 
     void removeProcess(const wp<IBinder>& processToken);
 
@@ -59,7 +64,7 @@
     std::mutex mMutex;
 
     struct ClientCacheBuffer {
-        sp<GraphicBuffer> buffer;
+        std::shared_ptr<renderengine::ExternalTexture> buffer;
         std::set<wp<ErasedRecipient>> recipients;
     };
     std::map<wp<IBinder> /*caching process*/,
@@ -73,6 +78,7 @@
     };
 
     sp<CacheDeathRecipient> mDeathRecipient;
+    renderengine::RenderEngine* mRenderEngine = nullptr;
 
     bool getBuffer(const client_cache_t& cacheId, ClientCacheBuffer** outClientCacheBuffer)
             REQUIRES(mMutex);
diff --git a/services/surfaceflinger/CompositionEngine/include/compositionengine/RenderSurface.h b/services/surfaceflinger/CompositionEngine/include/compositionengine/RenderSurface.h
index f680460..daee83b 100644
--- a/services/surfaceflinger/CompositionEngine/include/compositionengine/RenderSurface.h
+++ b/services/surfaceflinger/CompositionEngine/include/compositionengine/RenderSurface.h
@@ -16,15 +16,16 @@
 
 #pragma once
 
-#include <cstdint>
-#include <vector>
-
+#include <renderengine/ExternalTexture.h>
 #include <ui/Fence.h>
 #include <ui/GraphicTypes.h>
 #include <ui/Size.h>
 #include <utils/Errors.h>
 #include <utils/StrongPointer.h>
 
+#include <cstdint>
+#include <vector>
+
 namespace android {
 
 class GraphicBuffer;
@@ -80,7 +81,8 @@
     virtual void prepareFrame(bool usesClientComposition, bool usesDeviceComposition) = 0;
 
     // Allocates a buffer as scratch space for GPU composition
-    virtual sp<GraphicBuffer> dequeueBuffer(base::unique_fd* bufferFence) = 0;
+    virtual std::shared_ptr<renderengine::ExternalTexture> dequeueBuffer(
+            base::unique_fd* bufferFence) = 0;
 
     // Queues the drawn buffer for consumption by HWC. readyFence is the fence
     // which will fire when the buffer is ready for consumption.
diff --git a/services/surfaceflinger/CompositionEngine/include/compositionengine/RenderSurfaceCreationArgs.h b/services/surfaceflinger/CompositionEngine/include/compositionengine/RenderSurfaceCreationArgs.h
index a1230b3..a8d372c 100644
--- a/services/surfaceflinger/CompositionEngine/include/compositionengine/RenderSurfaceCreationArgs.h
+++ b/services/surfaceflinger/CompositionEngine/include/compositionengine/RenderSurfaceCreationArgs.h
@@ -45,6 +45,8 @@
 
     // The DisplaySurface for this surface
     sp<DisplaySurface> displaySurface;
+
+    size_t maxTextureCacheSize;
 };
 
 /**
@@ -81,6 +83,11 @@
         return *this;
     }
 
+    RenderSurfaceCreationArgsBuilder& setMaxTextureCacheSize(size_t maxTextureCacheSize) {
+        mArgs.maxTextureCacheSize = maxTextureCacheSize;
+        return *this;
+    }
+
 private:
     RenderSurfaceCreationArgs mArgs;
 };
diff --git a/services/surfaceflinger/CompositionEngine/include/compositionengine/impl/OutputLayerCompositionState.h b/services/surfaceflinger/CompositionEngine/include/compositionengine/impl/OutputLayerCompositionState.h
index 48a54d6..c61ec59 100644
--- a/services/surfaceflinger/CompositionEngine/include/compositionengine/impl/OutputLayerCompositionState.h
+++ b/services/surfaceflinger/CompositionEngine/include/compositionengine/impl/OutputLayerCompositionState.h
@@ -18,7 +18,7 @@
 
 #include <compositionengine/ProjectionSpace.h>
 #include <compositionengine/impl/HwcBufferCache.h>
-#include <renderengine/Mesh.h>
+#include <renderengine/ExternalTexture.h>
 #include <ui/FloatRect.h>
 #include <ui/GraphicTypes.h>
 #include <ui/Rect.h>
@@ -89,7 +89,7 @@
 
     // Overrides the buffer, acquire fence, and display frame stored in LayerFECompositionState
     struct {
-        sp<GraphicBuffer> buffer = nullptr;
+        std::shared_ptr<renderengine::ExternalTexture> buffer = nullptr;
         sp<Fence> acquireFence = nullptr;
         Rect displayFrame = {};
         ui::Dataspace dataspace{ui::Dataspace::UNKNOWN};
diff --git a/services/surfaceflinger/CompositionEngine/include/compositionengine/impl/RenderSurface.h b/services/surfaceflinger/CompositionEngine/include/compositionengine/impl/RenderSurface.h
index 5127a6f..a8a5380 100644
--- a/services/surfaceflinger/CompositionEngine/include/compositionengine/impl/RenderSurface.h
+++ b/services/surfaceflinger/CompositionEngine/include/compositionengine/impl/RenderSurface.h
@@ -16,12 +16,16 @@
 
 #pragma once
 
-#include <memory>
-
 #include <android-base/unique_fd.h>
 #include <compositionengine/RenderSurface.h>
 #include <utils/StrongPointer.h>
 
+#include <memory>
+#include <vector>
+
+#include "renderengine/ExternalTexture.h"
+#include "renderengine/RenderEngine.h"
+
 struct ANativeWindow;
 
 namespace android {
@@ -54,7 +58,8 @@
     void setProtected(bool useProtected) override;
     status_t beginFrame(bool mustRecompose) override;
     void prepareFrame(bool usesClientComposition, bool usesDeviceComposition) override;
-    sp<GraphicBuffer> dequeueBuffer(base::unique_fd* bufferFence) override;
+    std::shared_ptr<renderengine::ExternalTexture> dequeueBuffer(
+            base::unique_fd* bufferFence) override;
     void queueBuffer(base::unique_fd readyFence) override;
     void onPresentDisplayCompleted() override;
     void flip() override;
@@ -66,7 +71,7 @@
     // Testing
     void setPageFlipCountForTest(std::uint32_t);
     void setSizeForTest(const ui::Size&);
-    sp<GraphicBuffer>& mutableGraphicBufferForTest();
+    std::shared_ptr<renderengine::ExternalTexture>& mutableTextureForTest();
     base::unique_fd& mutableBufferReadyForTest();
 
 private:
@@ -75,10 +80,13 @@
 
     // ANativeWindow being rendered into
     const sp<ANativeWindow> mNativeWindow;
-    // Current buffer being rendered into
-    sp<GraphicBuffer> mGraphicBuffer;
+
+    std::vector<std::shared_ptr<renderengine::ExternalTexture>> mTextureCache;
+    // Current texture being rendered into
+    std::shared_ptr<renderengine::ExternalTexture> mTexture;
     const sp<DisplaySurface> mDisplaySurface;
     ui::Size mSize;
+    const size_t mMaxTextureCacheSize;
     bool mProtected{false};
     std::uint32_t mPageFlipCount{0};
 };
diff --git a/services/surfaceflinger/CompositionEngine/include/compositionengine/impl/planner/CachedSet.h b/services/surfaceflinger/CompositionEngine/include/compositionengine/impl/planner/CachedSet.h
index c5d03a7..53f4a30 100644
--- a/services/surfaceflinger/CompositionEngine/include/compositionengine/impl/planner/CachedSet.h
+++ b/services/surfaceflinger/CompositionEngine/include/compositionengine/impl/planner/CachedSet.h
@@ -66,7 +66,7 @@
     const Rect& getBounds() const { return mBounds; }
     const Region& getVisibleRegion() const { return mVisibleRegion; }
     size_t getAge() const { return mAge; }
-    const sp<GraphicBuffer>& getBuffer() const { return mTexture.getBuffer(); }
+    const std::shared_ptr<renderengine::ExternalTexture>& getBuffer() const { return mTexture; }
     const sp<Fence>& getDrawFence() const { return mDrawFence; }
     const ProjectionSpace& getOutputSpace() const { return mOutputSpace; }
     ui::Dataspace getOutputDataspace() const { return mOutputDataspace; }
@@ -87,7 +87,7 @@
 
     void setLastUpdate(std::chrono::steady_clock::time_point now) { mLastUpdate = now; }
     void append(const CachedSet& other) {
-        mTexture.setBuffer(nullptr, nullptr);
+        mTexture = nullptr;
         mOutputDataspace = ui::Dataspace::UNKNOWN;
         mDrawFence = nullptr;
 
@@ -115,31 +115,7 @@
     Region mVisibleRegion;
     size_t mAge = 0;
 
-    class Texture {
-    public:
-        ~Texture() { setBuffer(nullptr, nullptr); }
-
-        void setBuffer(const sp<GraphicBuffer>& buffer, renderengine::RenderEngine* re) {
-            if (mRE && mBuffer) {
-                mRE->unbindExternalTextureBuffer(mBuffer->getId());
-            }
-
-            mBuffer = buffer;
-            mRE = re;
-
-            if (mRE && mBuffer) {
-                mRE->cacheExternalTextureBuffer(mBuffer);
-            }
-        }
-
-        const sp<GraphicBuffer>& getBuffer() const { return mBuffer; }
-
-    private:
-        sp<GraphicBuffer> mBuffer = nullptr;
-        renderengine::RenderEngine* mRE = nullptr;
-    };
-
-    Texture mTexture;
+    std::shared_ptr<renderengine::ExternalTexture> mTexture;
     sp<Fence> mDrawFence;
     ProjectionSpace mOutputSpace;
     ui::Dataspace mOutputDataspace;
diff --git a/services/surfaceflinger/CompositionEngine/include/compositionengine/mock/RenderSurface.h b/services/surfaceflinger/CompositionEngine/include/compositionengine/mock/RenderSurface.h
index a0cae6f..fe858c2 100644
--- a/services/surfaceflinger/CompositionEngine/include/compositionengine/mock/RenderSurface.h
+++ b/services/surfaceflinger/CompositionEngine/include/compositionengine/mock/RenderSurface.h
@@ -39,7 +39,7 @@
     MOCK_METHOD1(setBufferPixelFormat, void(ui::PixelFormat));
     MOCK_METHOD1(beginFrame, status_t(bool mustRecompose));
     MOCK_METHOD2(prepareFrame, void(bool, bool));
-    MOCK_METHOD1(dequeueBuffer, sp<GraphicBuffer>(base::unique_fd*));
+    MOCK_METHOD1(dequeueBuffer, std::shared_ptr<renderengine::ExternalTexture>(base::unique_fd*));
     MOCK_METHOD1(queueBuffer, void(base::unique_fd));
     MOCK_METHOD0(onPresentDisplayCompleted, void());
     MOCK_METHOD0(flip, void());
diff --git a/services/surfaceflinger/CompositionEngine/src/Output.cpp b/services/surfaceflinger/CompositionEngine/src/Output.cpp
index 3ac5433..3468b20 100644
--- a/services/surfaceflinger/CompositionEngine/src/Output.cpp
+++ b/services/surfaceflinger/CompositionEngine/src/Output.cpp
@@ -14,8 +14,7 @@
  * limitations under the License.
  */
 
-#include <thread>
-
+#include <SurfaceFlingerProperties.sysprop.h>
 #include <android-base/stringprintf.h>
 #include <compositionengine/CompositionEngine.h>
 #include <compositionengine/CompositionRefreshArgs.h>
@@ -29,7 +28,9 @@
 #include <compositionengine/impl/OutputLayerCompositionState.h>
 #include <compositionengine/impl/planner/Planner.h>
 
-#include <SurfaceFlingerProperties.sysprop.h>
+#include <thread>
+
+#include "renderengine/ExternalTexture.h"
 
 // TODO(b/129481165): remove the #pragma below and fix conversion issues
 #pragma clang diagnostic push
@@ -715,11 +716,11 @@
         bool skipLayer = false;
         if (layer->getState().overrideInfo.buffer != nullptr) {
             if (previousOverride != nullptr &&
-                layer->getState().overrideInfo.buffer == previousOverride) {
+                layer->getState().overrideInfo.buffer->getBuffer() == previousOverride) {
                 ALOGV("Skipping redundant buffer");
                 skipLayer = true;
             }
-            previousOverride = layer->getState().overrideInfo.buffer;
+            previousOverride = layer->getState().overrideInfo.buffer->getBuffer();
         }
 
         const bool includeGeometry = refreshArgs.updatingGeometryThisFrame;
@@ -978,14 +979,15 @@
     }
 
     base::unique_fd fd;
-    sp<GraphicBuffer> buf;
+
+    std::shared_ptr<renderengine::ExternalTexture> tex;
 
     // If we aren't doing client composition on this output, but do have a
     // flipClientTarget request for this frame on this output, we still need to
     // dequeue a buffer.
     if (hasClientComposition || outputState.flipClientTarget) {
-        buf = mRenderSurface->dequeueBuffer(&fd);
-        if (buf == nullptr) {
+        tex = mRenderSurface->dequeueBuffer(&fd);
+        if (tex == nullptr) {
             ALOGW("Dequeuing buffer for display [%s] failed, bailing out of "
                   "client composition for this frame",
                   mName.c_str());
@@ -1030,13 +1032,14 @@
     // Check if the client composition requests were rendered into the provided graphic buffer. If
     // so, we can reuse the buffer and avoid client composition.
     if (mClientCompositionRequestCache) {
-        if (mClientCompositionRequestCache->exists(buf->getId(), clientCompositionDisplay,
+        if (mClientCompositionRequestCache->exists(tex->getBuffer()->getId(),
+                                                   clientCompositionDisplay,
                                                    clientCompositionLayers)) {
             outputCompositionState.reusedClientComposition = true;
             setExpensiveRenderingExpected(false);
             return readyFence;
         }
-        mClientCompositionRequestCache->add(buf->getId(), clientCompositionDisplay,
+        mClientCompositionRequestCache->add(tex->getBuffer()->getId(), clientCompositionDisplay,
                                             clientCompositionLayers);
     }
 
@@ -1069,12 +1072,12 @@
     // over to RenderEngine, in which case this flag can be removed from the drawLayers interface.
     const bool useFramebufferCache = outputState.layerStackInternal;
     status_t status =
-            renderEngine.drawLayers(clientCompositionDisplay, clientCompositionLayerPointers, buf,
+            renderEngine.drawLayers(clientCompositionDisplay, clientCompositionLayerPointers, tex,
                                     useFramebufferCache, std::move(fd), &readyFence);
 
     if (status != NO_ERROR && mClientCompositionRequestCache) {
         // If rendering was not successful, remove the request from the cache.
-        mClientCompositionRequestCache->remove(buf->getId());
+        mClientCompositionRequestCache->remove(tex->getBuffer()->getId());
     }
 
     auto& timeStats = getCompositionEngine().getTimeStats();
@@ -1151,9 +1154,9 @@
 
             std::vector<LayerFE::LayerSettings> results;
             if (layer->getState().overrideInfo.buffer != nullptr) {
-                if (layer->getState().overrideInfo.buffer != previousOverrideBuffer) {
+                if (layer->getState().overrideInfo.buffer->getBuffer() != previousOverrideBuffer) {
                     results = layer->getOverrideCompositionList();
-                    previousOverrideBuffer = layer->getState().overrideInfo.buffer;
+                    previousOverrideBuffer = layer->getState().overrideInfo.buffer->getBuffer();
                     ALOGV("Replacing [%s] with override in RE", layer->getLayerFE().getDebugName());
                 } else {
                     ALOGV("Skipping redundant override buffer for [%s] in RE",
diff --git a/services/surfaceflinger/CompositionEngine/src/OutputLayer.cpp b/services/surfaceflinger/CompositionEngine/src/OutputLayer.cpp
index f640f85..9ca8914 100644
--- a/services/surfaceflinger/CompositionEngine/src/OutputLayer.cpp
+++ b/services/surfaceflinger/CompositionEngine/src/OutputLayer.cpp
@@ -537,7 +537,7 @@
     sp<GraphicBuffer> buffer = outputIndependentState.buffer;
     sp<Fence> acquireFence = outputIndependentState.acquireFence;
     if (getState().overrideInfo.buffer != nullptr) {
-        buffer = getState().overrideInfo.buffer;
+        buffer = getState().overrideInfo.buffer->getBuffer();
         acquireFence = getState().overrideInfo.acquireFence;
     }
 
@@ -699,7 +699,7 @@
     settings.geometry = renderengine::Geometry{
             .boundaries = boundaries.toFloatRect(),
     };
-    settings.bufferId = getState().overrideInfo.buffer->getId();
+    settings.bufferId = getState().overrideInfo.buffer->getBuffer()->getId();
     settings.source = renderengine::PixelSource{
             .buffer = renderengine::Buffer{
                     .buffer = getState().overrideInfo.buffer,
diff --git a/services/surfaceflinger/CompositionEngine/src/RenderSurface.cpp b/services/surfaceflinger/CompositionEngine/src/RenderSurface.cpp
index 3bef77d..ef50870 100644
--- a/services/surfaceflinger/CompositionEngine/src/RenderSurface.cpp
+++ b/services/surfaceflinger/CompositionEngine/src/RenderSurface.cpp
@@ -25,8 +25,8 @@
 #include <compositionengine/impl/DumpHelpers.h>
 #include <compositionengine/impl/OutputCompositionState.h>
 #include <compositionengine/impl/RenderSurface.h>
-
 #include <log/log.h>
+#include <renderengine/ExternalTexture.h>
 #include <renderengine/RenderEngine.h>
 #include <system/window.h>
 #include <ui/GraphicBuffer.h>
@@ -63,7 +63,8 @@
         mDisplay(display),
         mNativeWindow(args.nativeWindow),
         mDisplaySurface(args.displaySurface),
-        mSize(args.displayWidth, args.displayHeight) {
+        mSize(args.displayWidth, args.displayHeight),
+        mMaxTextureCacheSize(args.maxTextureCacheSize) {
     LOG_ALWAYS_FATAL_IF(!mNativeWindow);
 }
 
@@ -146,7 +147,8 @@
     }
 }
 
-sp<GraphicBuffer> RenderSurface::dequeueBuffer(base::unique_fd* bufferFence) {
+std::shared_ptr<renderengine::ExternalTexture> RenderSurface::dequeueBuffer(
+        base::unique_fd* bufferFence) {
     ATRACE_CALL();
     int fd = -1;
     ANativeWindowBuffer* buffer = nullptr;
@@ -158,16 +160,41 @@
               mDisplay.getName().c_str(), result);
         // Return fast here as we can't do much more - any rendering we do
         // now will just be wrong.
-        return mGraphicBuffer;
+        return mTexture;
     }
 
-    ALOGW_IF(mGraphicBuffer != nullptr, "Clobbering a non-null pointer to a buffer [%p].",
-             mGraphicBuffer->getNativeBuffer()->handle);
-    mGraphicBuffer = GraphicBuffer::from(buffer);
+    ALOGW_IF(mTexture != nullptr, "Clobbering a non-null pointer to a buffer [%p].",
+             mTexture->getBuffer()->getNativeBuffer()->handle);
+
+    sp<GraphicBuffer> newBuffer = GraphicBuffer::from(buffer);
+
+    std::shared_ptr<renderengine::ExternalTexture> texture;
+
+    for (auto it = mTextureCache.begin(); it != mTextureCache.end(); it++) {
+        const auto& cachedTexture = *it;
+        if (cachedTexture->getBuffer()->getId() == newBuffer->getId()) {
+            texture = cachedTexture;
+            mTextureCache.erase(it);
+            break;
+        }
+    }
+
+    if (texture) {
+        mTexture = texture;
+    } else {
+        mTexture = std::make_shared<
+                renderengine::ExternalTexture>(GraphicBuffer::from(buffer),
+                                               mCompositionEngine.getRenderEngine(),
+                                               renderengine::ExternalTexture::Usage::WRITEABLE);
+    }
+    mTextureCache.push_back(mTexture);
+    if (mTextureCache.size() > mMaxTextureCacheSize) {
+        mTextureCache.erase(mTextureCache.begin());
+    }
 
     *bufferFence = base::unique_fd(fd);
 
-    return mGraphicBuffer;
+    return mTexture;
 }
 
 void RenderSurface::queueBuffer(base::unique_fd readyFence) {
@@ -177,24 +204,24 @@
         // hasFlipClientTargetRequest could return true even if we haven't
         // dequeued a buffer before. Try dequeueing one if we don't have a
         // buffer ready.
-        if (mGraphicBuffer == nullptr) {
+        if (mTexture == nullptr) {
             ALOGI("Attempting to queue a client composited buffer without one "
                   "previously dequeued for display [%s]. Attempting to dequeue "
                   "a scratch buffer now",
                   mDisplay.getName().c_str());
-            // We shouldn't deadlock here, since mGraphicBuffer == nullptr only
+            // We shouldn't deadlock here, since mTexture == nullptr only
             // after a successful call to queueBuffer, or if dequeueBuffer has
             // never been called.
             base::unique_fd unused;
             dequeueBuffer(&unused);
         }
 
-        if (mGraphicBuffer == nullptr) {
+        if (mTexture == nullptr) {
             ALOGE("No buffer is ready for display [%s]", mDisplay.getName().c_str());
         } else {
-            status_t result =
-                    mNativeWindow->queueBuffer(mNativeWindow.get(),
-                                               mGraphicBuffer->getNativeBuffer(), dup(readyFence));
+            status_t result = mNativeWindow->queueBuffer(mNativeWindow.get(),
+                                                         mTexture->getBuffer()->getNativeBuffer(),
+                                                         dup(readyFence));
             if (result != NO_ERROR) {
                 ALOGE("Error when queueing buffer for display [%s]: %d", mDisplay.getName().c_str(),
                       result);
@@ -204,11 +231,12 @@
                     LOG_ALWAYS_FATAL("ANativeWindow::queueBuffer failed with error: %d", result);
                 } else {
                     mNativeWindow->cancelBuffer(mNativeWindow.get(),
-                                                mGraphicBuffer->getNativeBuffer(), dup(readyFence));
+                                                mTexture->getBuffer()->getNativeBuffer(),
+                                                dup(readyFence));
                 }
             }
 
-            mGraphicBuffer = nullptr;
+            mTexture = nullptr;
         }
     }
 
@@ -256,8 +284,8 @@
     mSize = size;
 }
 
-sp<GraphicBuffer>& RenderSurface::mutableGraphicBufferForTest() {
-    return mGraphicBuffer;
+std::shared_ptr<renderengine::ExternalTexture>& RenderSurface::mutableTextureForTest() {
+    return mTexture;
 }
 
 } // namespace impl
diff --git a/services/surfaceflinger/CompositionEngine/src/planner/CachedSet.cpp b/services/surfaceflinger/CompositionEngine/src/planner/CachedSet.cpp
index dcb7555..9955e29 100644
--- a/services/surfaceflinger/CompositionEngine/src/planner/CachedSet.cpp
+++ b/services/surfaceflinger/CompositionEngine/src/planner/CachedSet.cpp
@@ -130,7 +130,7 @@
 }
 
 bool CachedSet::hasReadyBuffer() const {
-    return mTexture.getBuffer() != nullptr && mDrawFence->getStatus() == Fence::Status::Signaled;
+    return mTexture != nullptr && mDrawFence->getStatus() == Fence::Status::Signaled;
 }
 
 std::vector<CachedSet> CachedSet::decompose() const {
@@ -217,21 +217,27 @@
     sp<GraphicBuffer> buffer = new GraphicBuffer(static_cast<uint32_t>(mBounds.getWidth()),
                                                  static_cast<uint32_t>(mBounds.getHeight()),
                                                  HAL_PIXEL_FORMAT_RGBA_8888, 1, usageFlags);
+    const auto texture = std::make_shared<
+            renderengine::ExternalTexture>(buffer, renderEngine,
+                                           renderengine::ExternalTexture::Usage::READABLE |
+                                                   renderengine::ExternalTexture::Usage::WRITEABLE);
     LOG_ALWAYS_FATAL_IF(buffer->initCheck() != OK);
     base::unique_fd drawFence;
 
-    status_t result = renderEngine.drawLayers(displaySettings, layerSettingsPointers, buffer, false,
-                                              base::unique_fd(), &drawFence);
+    status_t result = renderEngine.drawLayers(displaySettings, layerSettingsPointers, texture,
+                                              false, base::unique_fd(), &drawFence);
 
     if (result == NO_ERROR) {
-        mTexture.setBuffer(buffer, &renderEngine);
         mDrawFence = new Fence(drawFence.release());
         mOutputSpace = ProjectionSpace(ui::Size(outputState.framebufferSpace.bounds.getWidth(),
                                                 outputState.framebufferSpace.bounds.getHeight()),
                                        mBounds);
+        mTexture = std::move(texture);
         mOutputSpace.orientation = outputState.framebufferSpace.orientation;
         mOutputDataspace = outputDataspace;
         mOrientation = orientation;
+    } else {
+        mTexture = nullptr;
     }
 }
 
diff --git a/services/surfaceflinger/CompositionEngine/src/planner/Planner.cpp b/services/surfaceflinger/CompositionEngine/src/planner/Planner.cpp
index ad75557..3a2534b 100644
--- a/services/surfaceflinger/CompositionEngine/src/planner/Planner.cpp
+++ b/services/surfaceflinger/CompositionEngine/src/planner/Planner.cpp
@@ -111,7 +111,12 @@
     const GraphicBuffer* currentOverrideBuffer = nullptr;
     bool hasSkippedLayers = false;
     for (auto layer : layers) {
-        const GraphicBuffer* overrideBuffer = layer->getState().overrideInfo.buffer.get();
+        if (!layer->getState().overrideInfo.buffer) {
+            continue;
+        }
+
+        const GraphicBuffer* overrideBuffer =
+                layer->getState().overrideInfo.buffer->getBuffer().get();
         if (overrideBuffer != nullptr && overrideBuffer == currentOverrideBuffer) {
             // Skip this layer since it is part of a previous cached set
             hasSkippedLayers = true;
diff --git a/services/surfaceflinger/CompositionEngine/tests/OutputLayerTest.cpp b/services/surfaceflinger/CompositionEngine/tests/OutputLayerTest.cpp
index 8a4d161..4c3f494 100644
--- a/services/surfaceflinger/CompositionEngine/tests/OutputLayerTest.cpp
+++ b/services/surfaceflinger/CompositionEngine/tests/OutputLayerTest.cpp
@@ -26,6 +26,7 @@
 #include "MockHWC2.h"
 #include "MockHWComposer.h"
 #include "RegionMatcher.h"
+#include "renderengine/mock/RenderEngine.h"
 
 namespace android::compositionengine {
 namespace {
@@ -715,7 +716,7 @@
     static const HdrMetadata kHdrMetadata;
     static native_handle_t* kSidebandStreamHandle;
     static const sp<GraphicBuffer> kBuffer;
-    static const sp<GraphicBuffer> kOverrideBuffer;
+    std::shared_ptr<renderengine::ExternalTexture> kOverrideBuffer;
     static const sp<Fence> kFence;
     static const sp<Fence> kOverrideFence;
     static const std::string kLayerGenericMetadata1Key;
@@ -724,6 +725,11 @@
     static const std::vector<uint8_t> kLayerGenericMetadata2Value;
 
     OutputLayerWriteStateToHWCTest() {
+        kOverrideBuffer = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(), mRenderEngine,
+                                               renderengine::ExternalTexture::Usage::READABLE |
+                                                       renderengine::ExternalTexture::Usage::
+                                                               WRITEABLE);
         auto& outputLayerState = mOutputLayer.editState();
         outputLayerState.hwc = impl::OutputLayerCompositionState::Hwc(mHwcLayer);
 
@@ -839,6 +845,7 @@
 
     std::shared_ptr<HWC2::mock::Layer> mHwcLayer{std::make_shared<StrictMock<HWC2::mock::Layer>>()};
     StrictMock<mock::DisplayColorProfile> mDisplayColorProfile;
+    renderengine::mock::RenderEngine mRenderEngine;
 };
 
 const half4 OutputLayerWriteStateToHWCTest::kColor{81.f / 255.f, 82.f / 255.f, 83.f / 255.f,
@@ -858,7 +865,6 @@
 native_handle_t* OutputLayerWriteStateToHWCTest::kSidebandStreamHandle =
         reinterpret_cast<native_handle_t*>(1031);
 const sp<GraphicBuffer> OutputLayerWriteStateToHWCTest::kBuffer;
-const sp<GraphicBuffer> OutputLayerWriteStateToHWCTest::kOverrideBuffer = new GraphicBuffer();
 const sp<Fence> OutputLayerWriteStateToHWCTest::kFence;
 const sp<Fence> OutputLayerWriteStateToHWCTest::kOverrideFence = new Fence();
 const std::string OutputLayerWriteStateToHWCTest::kLayerGenericMetadata1Key =
@@ -1023,7 +1029,7 @@
                               kOverrideBufferTransform, kOverrideBlendMode, kOverrideAlpha);
     expectPerFrameCommonCalls(SimulateUnsupported::None, kOverrideDataspace, kOverrideVisibleRegion,
                               kOverrideSurfaceDamage);
-    expectSetHdrMetadataAndBufferCalls(kOverrideBuffer, kOverrideFence);
+    expectSetHdrMetadataAndBufferCalls(kOverrideBuffer->getBuffer(), kOverrideFence);
     expectSetCompositionTypeCall(Hwc2::IComposerClient::Composition::DEVICE);
 
     mOutputLayer.writeStateToHWC(/*includeGeometry*/ true, /*skipLayer*/ false);
diff --git a/services/surfaceflinger/CompositionEngine/tests/OutputTest.cpp b/services/surfaceflinger/CompositionEngine/tests/OutputTest.cpp
index 5f0b0ee..e80100c 100644
--- a/services/surfaceflinger/CompositionEngine/tests/OutputTest.cpp
+++ b/services/surfaceflinger/CompositionEngine/tests/OutputTest.cpp
@@ -14,8 +14,6 @@
  * limitations under the License.
  */
 
-#include <cmath>
-
 #include <android-base/stringprintf.h>
 #include <compositionengine/LayerFECompositionState.h>
 #include <compositionengine/impl/Output.h>
@@ -31,9 +29,12 @@
 #include <ui/Rect.h>
 #include <ui/Region.h>
 
+#include <cmath>
+
 #include "CallOrderStateMachineHelper.h"
 #include "MockHWC2.h"
 #include "RegionMatcher.h"
+#include "renderengine/ExternalTexture.h"
 
 namespace android::compositionengine {
 namespace {
@@ -2960,7 +2961,10 @@
     mock::DisplayColorProfile* mDisplayColorProfile = new StrictMock<mock::DisplayColorProfile>();
     mock::RenderSurface* mRenderSurface = new StrictMock<mock::RenderSurface>();
     StrictMock<OutputPartialMock> mOutput;
-    sp<GraphicBuffer> mOutputBuffer = new GraphicBuffer();
+    std::shared_ptr<renderengine::ExternalTexture> mOutputBuffer = std::make_shared<
+            renderengine::ExternalTexture>(new GraphicBuffer(), mRenderEngine,
+                                           renderengine::ExternalTexture::Usage::READABLE |
+                                                   renderengine::ExternalTexture::Usage::WRITEABLE);
 
     std::optional<base::unique_fd> mReadyFence;
 };
@@ -3173,7 +3177,10 @@
     EXPECT_CALL(mOutput, appendRegionFlashRequests(RegionEq(kDebugRegion), _))
             .WillRepeatedly(Return());
 
-    sp<GraphicBuffer> otherOutputBuffer = new GraphicBuffer();
+    const auto otherOutputBuffer = std::make_shared<
+            renderengine::ExternalTexture>(new GraphicBuffer(), mRenderEngine,
+                                           renderengine::ExternalTexture::Usage::READABLE |
+                                                   renderengine::ExternalTexture::Usage::WRITEABLE);
     EXPECT_CALL(*mRenderSurface, dequeueBuffer(_))
             .WillOnce(Return(mOutputBuffer))
             .WillOnce(Return(otherOutputBuffer));
diff --git a/services/surfaceflinger/CompositionEngine/tests/RenderSurfaceTest.cpp b/services/surfaceflinger/CompositionEngine/tests/RenderSurfaceTest.cpp
index 5ef5d7b..9aeb290 100644
--- a/services/surfaceflinger/CompositionEngine/tests/RenderSurfaceTest.cpp
+++ b/services/surfaceflinger/CompositionEngine/tests/RenderSurfaceTest.cpp
@@ -15,6 +15,8 @@
  */
 
 // TODO(b/129481165): remove the #pragma below and fix conversion issues
+#include "renderengine/ExternalTexture.h"
+#include "ui/GraphicBuffer.h"
 #pragma clang diagnostic push
 #pragma clang diagnostic ignored "-Wextra"
 
@@ -239,9 +241,9 @@
                     DoAll(SetArgPointee<0>(buffer.get()), SetArgPointee<1>(-1), Return(NO_ERROR)));
 
     base::unique_fd fence;
-    EXPECT_EQ(buffer.get(), mSurface.dequeueBuffer(&fence).get());
+    EXPECT_EQ(buffer.get(), mSurface.dequeueBuffer(&fence)->getBuffer().get());
 
-    EXPECT_EQ(buffer.get(), mSurface.mutableGraphicBufferForTest().get());
+    EXPECT_EQ(buffer.get(), mSurface.mutableTextureForTest()->getBuffer().get());
 }
 
 /*
@@ -249,8 +251,11 @@
  */
 
 TEST_F(RenderSurfaceTest, queueBufferHandlesNoClientComposition) {
-    sp<GraphicBuffer> buffer = new GraphicBuffer();
-    mSurface.mutableGraphicBufferForTest() = buffer;
+    const auto buffer = std::make_shared<
+            renderengine::ExternalTexture>(new GraphicBuffer(), mRenderEngine,
+                                           renderengine::ExternalTexture::Usage::READABLE |
+                                                   renderengine::ExternalTexture::Usage::WRITEABLE);
+    mSurface.mutableTextureForTest() = buffer;
 
     impl::OutputCompositionState state;
     state.usesClientComposition = false;
@@ -261,43 +266,45 @@
 
     mSurface.queueBuffer(base::unique_fd());
 
-    EXPECT_EQ(buffer.get(), mSurface.mutableGraphicBufferForTest().get());
+    EXPECT_EQ(buffer.get(), mSurface.mutableTextureForTest().get());
 }
 
 TEST_F(RenderSurfaceTest, queueBufferHandlesClientComposition) {
-    sp<GraphicBuffer> buffer = new GraphicBuffer();
-    mSurface.mutableGraphicBufferForTest() = buffer;
+    const auto buffer = std::make_shared<renderengine::ExternalTexture>(new GraphicBuffer(),
+                                                                        mRenderEngine, false);
+    mSurface.mutableTextureForTest() = buffer;
 
     impl::OutputCompositionState state;
     state.usesClientComposition = true;
     state.flipClientTarget = false;
 
     EXPECT_CALL(mDisplay, getState()).WillOnce(ReturnRef(state));
-    EXPECT_CALL(*mNativeWindow, queueBuffer(buffer->getNativeBuffer(), -1))
+    EXPECT_CALL(*mNativeWindow, queueBuffer(buffer->getBuffer()->getNativeBuffer(), -1))
             .WillOnce(Return(NO_ERROR));
     EXPECT_CALL(*mDisplaySurface, advanceFrame()).Times(1);
 
     mSurface.queueBuffer(base::unique_fd());
 
-    EXPECT_EQ(nullptr, mSurface.mutableGraphicBufferForTest().get());
+    EXPECT_EQ(nullptr, mSurface.mutableTextureForTest().get());
 }
 
 TEST_F(RenderSurfaceTest, queueBufferHandlesFlipClientTargetRequest) {
-    sp<GraphicBuffer> buffer = new GraphicBuffer();
-    mSurface.mutableGraphicBufferForTest() = buffer;
+    const auto buffer = std::make_shared<renderengine::ExternalTexture>(new GraphicBuffer(),
+                                                                        mRenderEngine, false);
+    mSurface.mutableTextureForTest() = buffer;
 
     impl::OutputCompositionState state;
     state.usesClientComposition = false;
     state.flipClientTarget = true;
 
     EXPECT_CALL(mDisplay, getState()).WillOnce(ReturnRef(state));
-    EXPECT_CALL(*mNativeWindow, queueBuffer(buffer->getNativeBuffer(), -1))
+    EXPECT_CALL(*mNativeWindow, queueBuffer(buffer->getBuffer()->getNativeBuffer(), -1))
             .WillOnce(Return(NO_ERROR));
     EXPECT_CALL(*mDisplaySurface, advanceFrame()).Times(1);
 
     mSurface.queueBuffer(base::unique_fd());
 
-    EXPECT_EQ(nullptr, mSurface.mutableGraphicBufferForTest().get());
+    EXPECT_EQ(nullptr, mSurface.mutableTextureForTest().get());
 }
 
 TEST_F(RenderSurfaceTest, queueBufferHandlesFlipClientTargetRequestWithNoBufferYetDequeued) {
@@ -317,27 +324,28 @@
 
     mSurface.queueBuffer(base::unique_fd());
 
-    EXPECT_EQ(nullptr, mSurface.mutableGraphicBufferForTest().get());
+    EXPECT_EQ(nullptr, mSurface.mutableTextureForTest().get());
 }
 
 TEST_F(RenderSurfaceTest, queueBufferHandlesNativeWindowQueueBufferFailureOnVirtualDisplay) {
-    sp<GraphicBuffer> buffer = new GraphicBuffer();
-    mSurface.mutableGraphicBufferForTest() = buffer;
+    const auto buffer = std::make_shared<renderengine::ExternalTexture>(new GraphicBuffer(),
+                                                                        mRenderEngine, false);
+    mSurface.mutableTextureForTest() = buffer;
 
     impl::OutputCompositionState state;
     state.usesClientComposition = true;
 
     EXPECT_CALL(mDisplay, getState()).WillOnce(ReturnRef(state));
-    EXPECT_CALL(*mNativeWindow, queueBuffer(buffer->getNativeBuffer(), -1))
+    EXPECT_CALL(*mNativeWindow, queueBuffer(buffer->getBuffer()->getNativeBuffer(), -1))
             .WillOnce(Return(INVALID_OPERATION));
     EXPECT_CALL(mDisplay, isVirtual()).WillOnce(Return(true));
-    EXPECT_CALL(*mNativeWindow, cancelBuffer(buffer->getNativeBuffer(), -1))
+    EXPECT_CALL(*mNativeWindow, cancelBuffer(buffer->getBuffer()->getNativeBuffer(), -1))
             .WillOnce(Return(NO_ERROR));
     EXPECT_CALL(*mDisplaySurface, advanceFrame()).Times(1);
 
     mSurface.queueBuffer(base::unique_fd());
 
-    EXPECT_EQ(nullptr, mSurface.mutableGraphicBufferForTest().get());
+    EXPECT_EQ(nullptr, mSurface.mutableTextureForTest().get());
 }
 
 /*
diff --git a/services/surfaceflinger/CompositionEngine/tests/planner/CachedSetTest.cpp b/services/surfaceflinger/CompositionEngine/tests/planner/CachedSetTest.cpp
index f01fe27..283c692 100644
--- a/services/surfaceflinger/CompositionEngine/tests/planner/CachedSetTest.cpp
+++ b/services/surfaceflinger/CompositionEngine/tests/planner/CachedSetTest.cpp
@@ -305,8 +305,8 @@
 
     const auto drawLayers = [&](const renderengine::DisplaySettings& displaySettings,
                                 const std::vector<const renderengine::LayerSettings*>& layers,
-                                const sp<GraphicBuffer>&, const bool, base::unique_fd&&,
-                                base::unique_fd*) -> size_t {
+                                const std::shared_ptr<renderengine::ExternalTexture>&, const bool,
+                                base::unique_fd&&, base::unique_fd*) -> size_t {
         EXPECT_EQ(Rect(0, 0, 2, 2), displaySettings.physicalDisplay);
         EXPECT_EQ(mOutputState.layerStackSpace.content, displaySettings.clip);
         EXPECT_EQ(ui::Transform::toRotationFlags(mOutputState.framebufferSpace.orientation),
@@ -321,7 +321,6 @@
     EXPECT_CALL(*layerFE1, prepareClientCompositionList(_)).WillOnce(Return(clientCompList1));
     EXPECT_CALL(*layerFE2, prepareClientCompositionList(_)).WillOnce(Return(clientCompList2));
     EXPECT_CALL(mRenderEngine, drawLayers(_, _, _, _, _, _)).WillOnce(Invoke(drawLayers));
-    EXPECT_CALL(mRenderEngine, cacheExternalTextureBuffer(_));
     cachedSet.render(mRenderEngine, mOutputState);
     expectReadyBuffer(cachedSet);
 
@@ -331,7 +330,6 @@
               cachedSet.getOutputSpace().bounds);
 
     // Now check that appending a new cached set properly cleans up RenderEngine resources.
-    EXPECT_CALL(mRenderEngine, unbindExternalTextureBuffer(_));
     CachedSet::Layer& layer3 = *mTestLayers[2]->cachedSetLayer.get();
     cachedSet.append(CachedSet(layer3));
 }
diff --git a/services/surfaceflinger/DisplayDevice.cpp b/services/surfaceflinger/DisplayDevice.cpp
index b7b2cc6..8692ee6 100644
--- a/services/surfaceflinger/DisplayDevice.cpp
+++ b/services/surfaceflinger/DisplayDevice.cpp
@@ -70,11 +70,13 @@
         mIsPrimary(args.isPrimary) {
     mCompositionDisplay->editState().isSecure = args.isSecure;
     mCompositionDisplay->createRenderSurface(
-            compositionengine::RenderSurfaceCreationArgs{ANativeWindow_getWidth(
-                                                                 args.nativeWindow.get()),
-                                                         ANativeWindow_getHeight(
-                                                                 args.nativeWindow.get()),
-                                                         args.nativeWindow, args.displaySurface});
+            compositionengine::
+                    RenderSurfaceCreationArgs{ANativeWindow_getWidth(args.nativeWindow.get()),
+                                              ANativeWindow_getHeight(args.nativeWindow.get()),
+                                              args.nativeWindow, args.displaySurface,
+                                              static_cast<size_t>(
+                                                      SurfaceFlinger::
+                                                              maxFrameBufferAcquiredBuffers)});
 
     if (!mFlinger->mDisableClientCompositionCache &&
         SurfaceFlinger::maxFrameBufferAcquiredBuffers > 0) {
diff --git a/services/surfaceflinger/Layer.cpp b/services/surfaceflinger/Layer.cpp
index b45f2bc..cf215ad 100644
--- a/services/surfaceflinger/Layer.cpp
+++ b/services/surfaceflinger/Layer.cpp
@@ -880,7 +880,7 @@
     }
 
     // Allow BufferStateLayer to release any unlatched buffers in drawing state.
-    bufferMayChange(c.buffer);
+    bufferMayChange(c.buffer->getBuffer());
 
     // Commit the transaction
     commitTransaction(c);
@@ -891,7 +891,11 @@
 
 void Layer::commitTransaction(State& stateToCommit) {
     if (auto& bufferSurfaceFrame = mDrawingState.bufferSurfaceFrameTX;
-        mDrawingState.buffer != stateToCommit.buffer && bufferSurfaceFrame != nullptr &&
+        ((mDrawingState.buffer && stateToCommit.buffer &&
+          mDrawingState.buffer->getBuffer() != stateToCommit.buffer->getBuffer()) ||
+         (mDrawingState.buffer && !stateToCommit.buffer) ||
+         (!mDrawingState.buffer && stateToCommit.buffer)) &&
+        bufferSurfaceFrame != nullptr &&
         bufferSurfaceFrame->getPresentState() != PresentState::Presented) {
         // If the previous buffer was committed but not latched (refreshPending - happens during
         // back to back invalidates), it gets silently dropped here. Mark the corresponding
diff --git a/services/surfaceflinger/Layer.h b/services/surfaceflinger/Layer.h
index 5379d74..9f3ea9a 100644
--- a/services/surfaceflinger/Layer.h
+++ b/services/surfaceflinger/Layer.h
@@ -203,7 +203,7 @@
 
         Region transparentRegionHint;
 
-        sp<GraphicBuffer> buffer;
+        std::shared_ptr<renderengine::ExternalTexture> buffer;
         client_cache_t clientCacheId;
         sp<Fence> acquireFence;
         std::shared_ptr<FenceTime> acquireFenceTime;
@@ -404,10 +404,11 @@
     // Used only to set BufferStateLayer state
     virtual bool setTransform(uint32_t /*transform*/) { return false; };
     virtual bool setTransformToDisplayInverse(bool /*transformToDisplayInverse*/) { return false; };
-    virtual bool setBuffer(const sp<GraphicBuffer>& /*buffer*/, const sp<Fence>& /*acquireFence*/,
-                           nsecs_t /*postTime*/, nsecs_t /*desiredPresentTime*/,
-                           bool /*isAutoTimestamp*/, const client_cache_t& /*clientCacheId*/,
-                           uint64_t /* frameNumber */, std::optional<nsecs_t> /* dequeueTime */,
+    virtual bool setBuffer(const std::shared_ptr<renderengine::ExternalTexture>& /*buffer*/,
+                           const sp<Fence>& /*acquireFence*/, nsecs_t /*postTime*/,
+                           nsecs_t /*desiredPresentTime*/, bool /*isAutoTimestamp*/,
+                           const client_cache_t& /*clientCacheId*/, uint64_t /* frameNumber */,
+                           std::optional<nsecs_t> /* dequeueTime */,
                            const FrameTimelineInfo& /*info*/,
                            const sp<ITransactionCompletedListener>& /* releaseBufferListener */) {
         return false;
@@ -709,7 +710,7 @@
      * Called before updating the drawing state buffer. Used by BufferStateLayer to release any
      * unlatched buffers in the drawing state.
      */
-    virtual void bufferMayChange(sp<GraphicBuffer>& /* newBuffer */){};
+    virtual void bufferMayChange(const sp<GraphicBuffer>& /* newBuffer */){};
 
     /*
      * Remove relative z for the layer if its relative parent is not part of the
diff --git a/services/surfaceflinger/RefreshRateOverlay.cpp b/services/surfaceflinger/RefreshRateOverlay.cpp
index 7a3e433..a9fd16c 100644
--- a/services/surfaceflinger/RefreshRateOverlay.cpp
+++ b/services/surfaceflinger/RefreshRateOverlay.cpp
@@ -208,7 +208,8 @@
     return true;
 }
 
-const std::vector<sp<GraphicBuffer>>& RefreshRateOverlay::getOrCreateBuffers(uint32_t fps) {
+const std::vector<std::shared_ptr<renderengine::ExternalTexture>>&
+RefreshRateOverlay::getOrCreateBuffers(uint32_t fps) {
     if (mBufferCache.find(fps) == mBufferCache.end()) {
         // Ensure the range is > 0, so we don't divide by 0.
         const auto rangeLength = std::max(1u, mHighFps - mLowFps);
@@ -222,7 +223,17 @@
         color.g = HIGH_FPS_COLOR.g * fpsScale + LOW_FPS_COLOR.g * (1 - fpsScale);
         color.b = HIGH_FPS_COLOR.b * fpsScale + LOW_FPS_COLOR.b * (1 - fpsScale);
         color.a = ALPHA;
-        mBufferCache.emplace(fps, SevenSegmentDrawer::drawNumber(fps, color, mShowSpinner));
+        auto buffers = SevenSegmentDrawer::drawNumber(fps, color, mShowSpinner);
+        std::vector<std::shared_ptr<renderengine::ExternalTexture>> textures;
+        std::transform(buffers.begin(), buffers.end(), std::back_inserter(textures),
+                       [&](const auto& buffer) -> std::shared_ptr<renderengine::ExternalTexture> {
+                           return std::make_shared<
+                                   renderengine::ExternalTexture>(buffer,
+                                                                  mFlinger.getRenderEngine(),
+                                                                  renderengine::ExternalTexture::
+                                                                          Usage::READABLE);
+                       });
+        mBufferCache.emplace(fps, textures);
     }
 
     return mBufferCache[fps];
diff --git a/services/surfaceflinger/RefreshRateOverlay.h b/services/surfaceflinger/RefreshRateOverlay.h
index c16cfa0..aa8329c 100644
--- a/services/surfaceflinger/RefreshRateOverlay.h
+++ b/services/surfaceflinger/RefreshRateOverlay.h
@@ -16,13 +16,14 @@
 
 #pragma once
 
-#include <unordered_map>
-
 #include <math/vec4.h>
+#include <renderengine/RenderEngine.h>
 #include <ui/Rect.h>
 #include <ui/Size.h>
 #include <utils/StrongPointer.h>
 
+#include <unordered_map>
+
 #include "Fps.h"
 
 namespace android {
@@ -70,7 +71,8 @@
     };
 
     bool createLayer();
-    const std::vector<sp<GraphicBuffer>>& getOrCreateBuffers(uint32_t fps);
+    const std::vector<std::shared_ptr<renderengine::ExternalTexture>>& getOrCreateBuffers(
+            uint32_t fps);
 
     SurfaceFlinger& mFlinger;
     const sp<Client> mClient;
@@ -78,7 +80,8 @@
     sp<IBinder> mIBinder;
     sp<IGraphicBufferProducer> mGbp;
 
-    std::unordered_map<int, std::vector<sp<GraphicBuffer>>> mBufferCache;
+    std::unordered_map<int, std::vector<std::shared_ptr<renderengine::ExternalTexture>>>
+            mBufferCache;
     std::optional<int> mCurrentFps;
     int mFrame = 0;
     static constexpr float ALPHA = 0.8f;
diff --git a/services/surfaceflinger/RegionSamplingThread.cpp b/services/surfaceflinger/RegionSamplingThread.cpp
index d0032ac..00090d9 100644
--- a/services/surfaceflinger/RegionSamplingThread.cpp
+++ b/services/surfaceflinger/RegionSamplingThread.cpp
@@ -438,18 +438,22 @@
         mFlinger.traverseLayersInLayerStack(layerStack, CaptureArgs::UNSET_UID, filterVisitor);
     };
 
-    sp<GraphicBuffer> buffer = nullptr;
-    if (mCachedBuffer && mCachedBuffer->getWidth() == sampledBounds.getWidth() &&
-        mCachedBuffer->getHeight() == sampledBounds.getHeight()) {
+    std::shared_ptr<renderengine::ExternalTexture> buffer = nullptr;
+    if (mCachedBuffer && mCachedBuffer->getBuffer()->getWidth() == sampledBounds.getWidth() &&
+        mCachedBuffer->getBuffer()->getHeight() == sampledBounds.getHeight()) {
         buffer = mCachedBuffer;
     } else {
         const uint32_t usage =
                 GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_HW_RENDER | GRALLOC_USAGE_HW_TEXTURE;
-        buffer = new GraphicBuffer(sampledBounds.getWidth(), sampledBounds.getHeight(),
-                                   PIXEL_FORMAT_RGBA_8888, 1, usage, "RegionSamplingThread");
-        const status_t bufferStatus = buffer->initCheck();
+        sp<GraphicBuffer> graphicBuffer =
+                new GraphicBuffer(sampledBounds.getWidth(), sampledBounds.getHeight(),
+                                  PIXEL_FORMAT_RGBA_8888, 1, usage, "RegionSamplingThread");
+        const status_t bufferStatus = graphicBuffer->initCheck();
         LOG_ALWAYS_FATAL_IF(bufferStatus != OK, "captureSample: Buffer failed to allocate: %d",
                             bufferStatus);
+        buffer = std::make_shared<
+                renderengine::ExternalTexture>(graphicBuffer, mFlinger.getRenderEngine(),
+                                               renderengine::ExternalTexture::Usage::WRITEABLE);
     }
 
     const sp<SyncScreenCaptureListener> captureListener = new SyncScreenCaptureListener();
@@ -465,8 +469,8 @@
     }
 
     ALOGV("Sampling %zu descriptors", activeDescriptors.size());
-    std::vector<float> lumas =
-            sampleBuffer(buffer, sampledBounds.leftTop(), activeDescriptors, orientation);
+    std::vector<float> lumas = sampleBuffer(buffer->getBuffer(), sampledBounds.leftTop(),
+                                            activeDescriptors, orientation);
     if (lumas.size() != activeDescriptors.size()) {
         ALOGW("collected %zu median luma values for %zu descriptors", lumas.size(),
               activeDescriptors.size());
@@ -477,16 +481,6 @@
         activeDescriptors[d].listener->onSampleCollected(lumas[d]);
     }
 
-    // Extend the lifetime of mCachedBuffer from the previous frame to here to ensure that:
-    // 1) The region sampling thread is the last owner of the buffer, and the freeing of the buffer
-    // happens in this thread, as opposed to the main thread.
-    // 2) The listener(s) receive their notifications prior to freeing the buffer.
-    if (mCachedBuffer != nullptr && mCachedBuffer != buffer) {
-        if (mFlinger.getRenderEngine().getRenderEngineType() ==
-            renderengine::RenderEngine::RenderEngineType::SKIA_GL_THREADED) {
-            mFlinger.getRenderEngine().unbindExternalTextureBuffer(mCachedBuffer->getId());
-        }
-    }
     mCachedBuffer = buffer;
     ATRACE_INT(lumaSamplingStepTag, static_cast<int>(samplingStep::noWorkNeeded));
 }
diff --git a/services/surfaceflinger/RegionSamplingThread.h b/services/surfaceflinger/RegionSamplingThread.h
index 0defdb3..86632db 100644
--- a/services/surfaceflinger/RegionSamplingThread.h
+++ b/services/surfaceflinger/RegionSamplingThread.h
@@ -16,17 +16,19 @@
 
 #pragma once
 
+#include <android-base/thread_annotations.h>
+#include <binder/IBinder.h>
+#include <renderengine/ExternalTexture.h>
+#include <ui/GraphicBuffer.h>
+#include <ui/Rect.h>
+#include <utils/StrongPointer.h>
+
 #include <chrono>
 #include <condition_variable>
 #include <mutex>
 #include <thread>
 #include <unordered_map>
 
-#include <android-base/thread_annotations.h>
-#include <binder/IBinder.h>
-#include <ui/GraphicBuffer.h>
-#include <ui/Rect.h>
-#include <utils/StrongPointer.h>
 #include "Scheduler/OneShotTimer.h"
 
 namespace android {
@@ -122,7 +124,8 @@
 
     std::mutex mSamplingMutex;
     std::unordered_map<wp<IBinder>, Descriptor, WpHash> mDescriptors GUARDED_BY(mSamplingMutex);
-    sp<GraphicBuffer> mCachedBuffer GUARDED_BY(mSamplingMutex) = nullptr;
+    std::shared_ptr<renderengine::ExternalTexture> mCachedBuffer GUARDED_BY(mSamplingMutex) =
+            nullptr;
 };
 
 } // namespace android
diff --git a/services/surfaceflinger/SurfaceFlinger.cpp b/services/surfaceflinger/SurfaceFlinger.cpp
index a4803df..21c8193 100644
--- a/services/surfaceflinger/SurfaceFlinger.cpp
+++ b/services/surfaceflinger/SurfaceFlinger.cpp
@@ -727,6 +727,7 @@
     mCompositionEngine->setTimeStats(mTimeStats);
     mCompositionEngine->setHwComposer(getFactory().createHWComposer(mHwcServiceName));
     mCompositionEngine->getHwComposer().setConfiguration(this, getBE().mComposerSequenceId);
+    ClientCache::getInstance().setRenderEngine(&getRenderEngine());
     // Process any initial hotplug and resulting display changes.
     processDisplayHotplugEventsLocked();
     const auto display = getDefaultDisplayDeviceLocked();
@@ -3692,7 +3693,6 @@
 
     if (uncacheBuffer.isValid()) {
         ClientCache::getInstance().erase(uncacheBuffer);
-        getRenderEngine().unbindExternalTextureBuffer(uncacheBuffer.id);
     }
 
     // If a synchronous transaction is explicitly requested without any changes, force a transaction
@@ -4056,23 +4056,16 @@
     }
     bool bufferChanged = what & layer_state_t::eBufferChanged;
     bool cacheIdChanged = what & layer_state_t::eCachedBufferChanged;
-    sp<GraphicBuffer> buffer;
+    std::shared_ptr<renderengine::ExternalTexture> buffer;
     if (bufferChanged && cacheIdChanged && s.buffer != nullptr) {
-        buffer = s.buffer;
-        bool success = ClientCache::getInstance().add(s.cachedBuffer, s.buffer);
-        if (success) {
-            getRenderEngine().cacheExternalTextureBuffer(s.buffer);
-            success = ClientCache::getInstance()
-                              .registerErasedRecipient(s.cachedBuffer,
-                                                       wp<ClientCache::ErasedRecipient>(this));
-            if (!success) {
-                getRenderEngine().unbindExternalTextureBuffer(s.buffer->getId());
-            }
-        }
+        ClientCache::getInstance().add(s.cachedBuffer, s.buffer);
+        buffer = ClientCache::getInstance().get(s.cachedBuffer);
     } else if (cacheIdChanged) {
         buffer = ClientCache::getInstance().get(s.cachedBuffer);
-    } else if (bufferChanged) {
-        buffer = s.buffer;
+    } else if (bufferChanged && s.buffer != nullptr) {
+        buffer = std::make_shared<
+                renderengine::ExternalTexture>(s.buffer, getRenderEngine(),
+                                               renderengine::ExternalTexture::Usage::READABLE);
     }
     if (buffer) {
         const bool frameNumberChanged = what & layer_state_t::eFrameNumberChanged;
@@ -5993,15 +5986,17 @@
     const status_t bufferStatus = buffer->initCheck();
     LOG_ALWAYS_FATAL_IF(bufferStatus != OK, "captureScreenCommon: Buffer failed to allocate: %d",
                         bufferStatus);
-    return captureScreenCommon(std::move(renderAreaFuture), traverseLayers, buffer,
+    const auto texture = std::make_shared<
+            renderengine::ExternalTexture>(buffer, getRenderEngine(),
+                                           renderengine::ExternalTexture::Usage::WRITEABLE);
+    return captureScreenCommon(std::move(renderAreaFuture), traverseLayers, texture,
                                false /* regionSampling */, grayscale, captureListener);
 }
 
-status_t SurfaceFlinger::captureScreenCommon(RenderAreaFuture renderAreaFuture,
-                                             TraverseLayersFunction traverseLayers,
-                                             sp<GraphicBuffer>& buffer, bool regionSampling,
-                                             bool grayscale,
-                                             const sp<IScreenCaptureListener>& captureListener) {
+status_t SurfaceFlinger::captureScreenCommon(
+        RenderAreaFuture renderAreaFuture, TraverseLayersFunction traverseLayers,
+        const std::shared_ptr<renderengine::ExternalTexture>& buffer, bool regionSampling,
+        bool grayscale, const sp<IScreenCaptureListener>& captureListener) {
     ATRACE_CALL();
 
     if (captureListener == nullptr) {
@@ -6034,15 +6029,6 @@
                                             regionSampling, grayscale, captureResults);
         });
 
-        // TODO(b/180767535): Remove this once we optimize buffer lifecycle for RenderEngine
-        // Only do this when we're not doing region sampling, to allow the region sampling thread to
-        // manage buffer lifecycle itself.
-        if (!regionSampling &&
-            getRenderEngine().getRenderEngineType() ==
-                    renderengine::RenderEngine::RenderEngineType::SKIA_GL_THREADED) {
-            getRenderEngine().unbindExternalTextureBuffer(buffer->getId());
-        }
-
         captureResults.result = result;
         captureListener->onScreenCaptureCompleted(captureResults);
     }));
@@ -6050,11 +6036,10 @@
     return NO_ERROR;
 }
 
-status_t SurfaceFlinger::renderScreenImplLocked(const RenderArea& renderArea,
-                                                TraverseLayersFunction traverseLayers,
-                                                const sp<GraphicBuffer>& buffer, bool forSystem,
-                                                bool regionSampling, bool grayscale,
-                                                ScreenCaptureResults& captureResults) {
+status_t SurfaceFlinger::renderScreenImplLocked(
+        const RenderArea& renderArea, TraverseLayersFunction traverseLayers,
+        const std::shared_ptr<renderengine::ExternalTexture>& buffer, bool forSystem,
+        bool regionSampling, bool grayscale, ScreenCaptureResults& captureResults) {
     ATRACE_CALL();
 
     traverseLayers([&](Layer* layer) {
@@ -6062,7 +6047,7 @@
                 captureResults.capturedSecureLayers || (layer->isVisible() && layer->isSecure());
     });
 
-    const bool useProtected = buffer->getUsage() & GRALLOC_USAGE_PROTECTED;
+    const bool useProtected = buffer->getBuffer()->getUsage() & GRALLOC_USAGE_PROTECTED;
 
     // We allow the system server to take screenshots of secure layers for
     // use in situations like the Screen-rotation animation and place
@@ -6072,7 +6057,7 @@
         return PERMISSION_DENIED;
     }
 
-    captureResults.buffer = buffer;
+    captureResults.buffer = buffer->getBuffer();
     captureResults.capturedDataspace = renderArea.getReqDataSpace();
 
     const auto reqWidth = renderArea.getReqWidth();
@@ -6163,11 +6148,9 @@
     base::unique_fd drawFence;
     getRenderEngine().useProtectedContext(useProtected);
 
-    // TODO(b/180767535): Remove this once we optimize buffer lifecycle for RenderEngine
-    const bool useFramebufferCache = getRenderEngine().getRenderEngineType() ==
-            renderengine::RenderEngine::RenderEngineType::SKIA_GL_THREADED;
+    const constexpr bool kUseFramebufferCache = false;
     getRenderEngine().drawLayers(clientCompositionDisplay, clientCompositionLayerPointers, buffer,
-                                 useFramebufferCache, std::move(bufferFence), &drawFence);
+                                 kUseFramebufferCache, std::move(bufferFence), &drawFence);
 
     if (drawFence >= 0) {
         sp<Fence> releaseFence = new Fence(dup(drawFence));
@@ -6432,10 +6415,6 @@
     mOffscreenLayers.erase(layer);
 }
 
-void SurfaceFlinger::bufferErased(const client_cache_t& clientCacheId) {
-    getRenderEngine().unbindExternalTextureBuffer(clientCacheId.id);
-}
-
 status_t SurfaceFlinger::setGlobalShadowSettings(const half4& ambientColor, const half4& spotColor,
                                                  float lightPosY, float lightPosZ,
                                                  float lightRadius) {
diff --git a/services/surfaceflinger/SurfaceFlinger.h b/services/surfaceflinger/SurfaceFlinger.h
index 9135632..893b3d8 100644
--- a/services/surfaceflinger/SurfaceFlinger.h
+++ b/services/surfaceflinger/SurfaceFlinger.h
@@ -178,7 +178,6 @@
 
 class SurfaceFlinger : public BnSurfaceComposer,
                        public PriorityDumper,
-                       public ClientCache::ErasedRecipient,
                        private IBinder::DeathRecipient,
                        private HWC2::ComposerCallback,
                        private ISchedulerCallback {
@@ -326,9 +325,6 @@
     wp<Layer> fromHandle(const sp<IBinder>& handle);
     wp<Layer> fromHandleLocked(const sp<IBinder>& handle) const REQUIRES(mStateLock);
 
-    // Inherit from ClientCache::ErasedRecipient
-    void bufferErased(const client_cache_t& clientCacheId) override;
-
     // If set, disables reusing client composition buffers. This can be set by
     // debug.sf.disable_client_composition_cache
     bool mDisableClientCompositionCache = false;
@@ -901,12 +897,14 @@
     status_t captureScreenCommon(RenderAreaFuture, TraverseLayersFunction, ui::Size bufferSize,
                                  ui::PixelFormat, bool allowProtected, bool grayscale,
                                  const sp<IScreenCaptureListener>&);
-    status_t captureScreenCommon(RenderAreaFuture, TraverseLayersFunction, sp<GraphicBuffer>&,
+    status_t captureScreenCommon(RenderAreaFuture, TraverseLayersFunction,
+                                 const std::shared_ptr<renderengine::ExternalTexture>&,
                                  bool regionSampling, bool grayscale,
                                  const sp<IScreenCaptureListener>&);
     status_t renderScreenImplLocked(const RenderArea&, TraverseLayersFunction,
-                                    const sp<GraphicBuffer>&, bool forSystem, bool regionSampling,
-                                    bool grayscale, ScreenCaptureResults&);
+                                    const std::shared_ptr<renderengine::ExternalTexture>&,
+                                    bool forSystem, bool regionSampling, bool grayscale,
+                                    ScreenCaptureResults&);
 
     sp<DisplayDevice> getDisplayByIdOrLayerStack(uint64_t displayOrLayerStack) REQUIRES(mStateLock);
     sp<DisplayDevice> getDisplayById(DisplayId displayId) const REQUIRES(mStateLock);
diff --git a/services/surfaceflinger/tests/unittests/CompositionTest.cpp b/services/surfaceflinger/tests/unittests/CompositionTest.cpp
index 4e1c0c7..3042450 100644
--- a/services/surfaceflinger/tests/unittests/CompositionTest.cpp
+++ b/services/surfaceflinger/tests/unittests/CompositionTest.cpp
@@ -15,6 +15,7 @@
  */
 
 // TODO(b/129481165): remove the #pragma below and fix conversion issues
+#include "renderengine/ExternalTexture.h"
 #pragma clang diagnostic push
 #pragma clang diagnostic ignored "-Wconversion"
 #pragma clang diagnostic ignored "-Wextra"
@@ -194,7 +195,7 @@
 
     sp<Fence> mClientTargetAcquireFence = Fence::NO_FENCE;
 
-    sp<GraphicBuffer> mCaptureScreenBuffer;
+    std::shared_ptr<renderengine::ExternalTexture> mCaptureScreenBuffer;
 };
 
 template <typename LayerCase>
@@ -243,11 +244,15 @@
     // TODO: Eliminate expensive/real allocation if possible.
     const uint32_t usage = GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN |
             GRALLOC_USAGE_HW_RENDER | GRALLOC_USAGE_HW_TEXTURE;
-    mCaptureScreenBuffer = new GraphicBuffer(renderArea->getReqWidth(), renderArea->getReqHeight(),
-                                             HAL_PIXEL_FORMAT_RGBA_8888, 1, usage, "screenshot");
+    mCaptureScreenBuffer = std::make_shared<
+            renderengine::ExternalTexture>(new GraphicBuffer(renderArea->getReqWidth(),
+                                                             renderArea->getReqHeight(),
+                                                             HAL_PIXEL_FORMAT_RGBA_8888, 1, usage,
+                                                             "screenshot"),
+                                           *mRenderEngine, true);
 
     status_t result =
-            mFlinger.renderScreenImplLocked(*renderArea, traverseLayers, mCaptureScreenBuffer.get(),
+            mFlinger.renderScreenImplLocked(*renderArea, traverseLayers, mCaptureScreenBuffer,
                                             forSystem, regionSampling);
     EXPECT_EQ(NO_ERROR, result);
 
@@ -340,8 +345,8 @@
         EXPECT_CALL(*test->mRenderEngine, drawLayers)
                 .WillRepeatedly([](const renderengine::DisplaySettings& displaySettings,
                                    const std::vector<const renderengine::LayerSettings*>&,
-                                   const sp<GraphicBuffer>&, const bool, base::unique_fd&&,
-                                   base::unique_fd*) -> status_t {
+                                   const std::shared_ptr<renderengine::ExternalTexture>&,
+                                   const bool, base::unique_fd&&, base::unique_fd*) -> status_t {
                     EXPECT_EQ(DEFAULT_DISPLAY_MAX_LUMINANCE, displaySettings.maxLuminance);
                     EXPECT_EQ(Rect(DEFAULT_DISPLAY_WIDTH, DEFAULT_DISPLAY_HEIGHT),
                               displaySettings.physicalDisplay);
@@ -389,8 +394,8 @@
         EXPECT_CALL(*test->mRenderEngine, drawLayers)
                 .WillRepeatedly([](const renderengine::DisplaySettings& displaySettings,
                                    const std::vector<const renderengine::LayerSettings*>&,
-                                   const sp<GraphicBuffer>&, const bool, base::unique_fd&&,
-                                   base::unique_fd*) -> status_t {
+                                   const std::shared_ptr<renderengine::ExternalTexture>&,
+                                   const bool, base::unique_fd&&, base::unique_fd*) -> status_t {
                     EXPECT_EQ(DEFAULT_DISPLAY_MAX_LUMINANCE, displaySettings.maxLuminance);
                     EXPECT_EQ(Rect(DEFAULT_DISPLAY_WIDTH, DEFAULT_DISPLAY_HEIGHT),
                               displaySettings.physicalDisplay);
@@ -625,8 +630,8 @@
         EXPECT_CALL(*test->mRenderEngine, drawLayers)
                 .WillOnce([](const renderengine::DisplaySettings& displaySettings,
                              const std::vector<const renderengine::LayerSettings*>& layerSettings,
-                             const sp<GraphicBuffer>&, const bool, base::unique_fd&&,
-                             base::unique_fd*) -> status_t {
+                             const std::shared_ptr<renderengine::ExternalTexture>&, const bool,
+                             base::unique_fd&&, base::unique_fd*) -> status_t {
                     EXPECT_EQ(DEFAULT_DISPLAY_MAX_LUMINANCE, displaySettings.maxLuminance);
                     EXPECT_EQ(Rect(DEFAULT_DISPLAY_WIDTH, DEFAULT_DISPLAY_HEIGHT),
                               displaySettings.physicalDisplay);
@@ -674,8 +679,8 @@
         EXPECT_CALL(*test->mRenderEngine, drawLayers)
                 .WillOnce([](const renderengine::DisplaySettings& displaySettings,
                              const std::vector<const renderengine::LayerSettings*>& layerSettings,
-                             const sp<GraphicBuffer>&, const bool, base::unique_fd&&,
-                             base::unique_fd*) -> status_t {
+                             const std::shared_ptr<renderengine::ExternalTexture>&, const bool,
+                             base::unique_fd&&, base::unique_fd*) -> status_t {
                     EXPECT_EQ(DEFAULT_DISPLAY_MAX_LUMINANCE, displaySettings.maxLuminance);
                     EXPECT_EQ(Rect(DEFAULT_DISPLAY_WIDTH, DEFAULT_DISPLAY_HEIGHT),
                               displaySettings.physicalDisplay);
@@ -751,8 +756,8 @@
         EXPECT_CALL(*test->mRenderEngine, drawLayers)
                 .WillOnce([](const renderengine::DisplaySettings& displaySettings,
                              const std::vector<const renderengine::LayerSettings*>& layerSettings,
-                             const sp<GraphicBuffer>&, const bool, base::unique_fd&&,
-                             base::unique_fd*) -> status_t {
+                             const std::shared_ptr<renderengine::ExternalTexture>&, const bool,
+                             base::unique_fd&&, base::unique_fd*) -> status_t {
                     EXPECT_EQ(DEFAULT_DISPLAY_MAX_LUMINANCE, displaySettings.maxLuminance);
                     EXPECT_EQ(Rect(DEFAULT_DISPLAY_WIDTH, DEFAULT_DISPLAY_HEIGHT),
                               displaySettings.physicalDisplay);
diff --git a/services/surfaceflinger/tests/unittests/TestableSurfaceFlinger.h b/services/surfaceflinger/tests/unittests/TestableSurfaceFlinger.h
index 63baf7d..d004b9d 100644
--- a/services/surfaceflinger/tests/unittests/TestableSurfaceFlinger.h
+++ b/services/surfaceflinger/tests/unittests/TestableSurfaceFlinger.h
@@ -352,8 +352,8 @@
 
     auto renderScreenImplLocked(const RenderArea& renderArea,
                                 SurfaceFlinger::TraverseLayersFunction traverseLayers,
-                                const sp<GraphicBuffer>& buffer, bool forSystem,
-                                bool regionSampling) {
+                                const std::shared_ptr<renderengine::ExternalTexture>& buffer,
+                                bool forSystem, bool regionSampling) {
         ScreenCaptureResults captureResults;
         return mFlinger->renderScreenImplLocked(renderArea, traverseLayers, buffer, forSystem,
                                                 regionSampling, false /* grayscale */,
diff --git a/services/surfaceflinger/tests/unittests/TransactionFrameTracerTest.cpp b/services/surfaceflinger/tests/unittests/TransactionFrameTracerTest.cpp
index ea1ce47..25001d3 100644
--- a/services/surfaceflinger/tests/unittests/TransactionFrameTracerTest.cpp
+++ b/services/surfaceflinger/tests/unittests/TransactionFrameTracerTest.cpp
@@ -21,6 +21,8 @@
 #include <gtest/gtest.h>
 #include <gui/SurfaceComposerClient.h>
 #include <log/log.h>
+#include <renderengine/ExternalTexture.h>
+#include <renderengine/mock/RenderEngine.h>
 #include <utils/String8.h>
 
 #include "TestableSurfaceFlinger.h"
@@ -99,6 +101,7 @@
 
     TestableSurfaceFlinger mFlinger;
     Hwc2::mock::Composer* mComposer = nullptr;
+    renderengine::mock::RenderEngine mRenderEngine;
     FenceToFenceTimeMap fenceFactory;
     client_cache_t mClientCache;
 
@@ -106,9 +109,12 @@
         sp<BufferStateLayer> layer = createBufferStateLayer();
 
         sp<Fence> fence(new Fence());
-        sp<GraphicBuffer> buffer{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
+        const auto buffer = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888,
+                                                                 1, 0),
+                                               mRenderEngine, false);
         int32_t layerId = layer->getSequence();
-        uint64_t bufferId = buffer->getId();
+        uint64_t bufferId = buffer->getBuffer()->getId();
         uint64_t frameNumber = 5;
         nsecs_t dequeueTime = 10;
         nsecs_t postTime = 20;
diff --git a/services/surfaceflinger/tests/unittests/TransactionSurfaceFrameTest.cpp b/services/surfaceflinger/tests/unittests/TransactionSurfaceFrameTest.cpp
index 09a1c2a..b7917aa 100644
--- a/services/surfaceflinger/tests/unittests/TransactionSurfaceFrameTest.cpp
+++ b/services/surfaceflinger/tests/unittests/TransactionSurfaceFrameTest.cpp
@@ -21,6 +21,8 @@
 #include <gtest/gtest.h>
 #include <gui/SurfaceComposerClient.h>
 #include <log/log.h>
+#include <renderengine/ExternalTexture.h>
+#include <renderengine/mock/RenderEngine.h>
 #include <utils/String8.h>
 
 #include "TestableSurfaceFlinger.h"
@@ -99,6 +101,7 @@
 
     TestableSurfaceFlinger mFlinger;
     Hwc2::mock::Composer* mComposer = nullptr;
+    renderengine::mock::RenderEngine mRenderEngine;
     FenceToFenceTimeMap fenceFactory;
     client_cache_t mClientCache;
 
@@ -119,7 +122,10 @@
         sp<BufferStateLayer> layer = createBufferStateLayer();
         sp<Fence> fence(new Fence());
         auto acquireFence = fenceFactory.createFenceTimeForTest(fence);
-        sp<GraphicBuffer> buffer{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
+        const auto buffer = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888,
+                                                                 1, 0),
+                                               mRenderEngine, false);
         layer->setBuffer(buffer, fence, 10, 20, false, mClientCache, 1, std::nullopt,
                          {/*vsyncId*/ 1, /*inputEventId*/ 0}, nullptr /* releaseBufferCallback */);
         acquireFence->signalForTest(12);
@@ -144,7 +150,10 @@
 
         sp<Fence> fence1(new Fence());
         auto acquireFence1 = fenceFactory.createFenceTimeForTest(fence1);
-        sp<GraphicBuffer> buffer1{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
+        const auto buffer1 = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888,
+                                                                 1, 0),
+                                               mRenderEngine, false);
         layer->setBuffer(buffer1, fence1, 10, 20, false, mClientCache, 1, std::nullopt,
                          {/*vsyncId*/ 1, /*inputEventId*/ 0}, nullptr /* releaseBufferCallback */);
         EXPECT_EQ(0u, layer->mCurrentState.bufferlessSurfaceFramesTX.size());
@@ -153,7 +162,10 @@
 
         sp<Fence> fence2(new Fence());
         auto acquireFence2 = fenceFactory.createFenceTimeForTest(fence2);
-        sp<GraphicBuffer> buffer2{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
+        const auto buffer2 = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888,
+                                                                 1, 0),
+                                               mRenderEngine, false);
         nsecs_t start = systemTime();
         layer->setBuffer(buffer2, fence2, 10, 20, false, mClientCache, 1, std::nullopt,
                          {/*vsyncId*/ 1, /*inputEventId*/ 0}, nullptr /* releaseBufferCallback */);
@@ -191,8 +203,10 @@
 
         sp<Fence> fence(new Fence());
         auto acquireFence = fenceFactory.createFenceTimeForTest(fence);
-        sp<GraphicBuffer> buffer{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
-
+        const auto buffer = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888,
+                                                                 1, 0),
+                                               mRenderEngine, false);
         layer->setBuffer(buffer, fence, 10, 20, false, mClientCache, 1, std::nullopt,
                          {/*vsyncId*/ 1, /*inputEventId*/ 0}, nullptr /* releaseBufferCallback */);
         acquireFence->signalForTest(12);
@@ -217,8 +231,10 @@
         sp<BufferStateLayer> layer = createBufferStateLayer();
         sp<Fence> fence(new Fence());
         auto acquireFence = fenceFactory.createFenceTimeForTest(fence);
-        sp<GraphicBuffer> buffer{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
-
+        const auto buffer = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888,
+                                                                 1, 0),
+                                               mRenderEngine, false);
         layer->setBuffer(buffer, fence, 10, 20, false, mClientCache, 1, std::nullopt,
                          {/*vsyncId*/ 1, /*inputEventId*/ 0}, nullptr /* releaseBufferCallback */);
         EXPECT_EQ(0u, layer->mCurrentState.bufferlessSurfaceFramesTX.size());
@@ -247,8 +263,10 @@
 
         sp<Fence> fence(new Fence());
         auto acquireFence = fenceFactory.createFenceTimeForTest(fence);
-        sp<GraphicBuffer> buffer{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
-
+        const auto buffer = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888,
+                                                                 1, 0),
+                                               mRenderEngine, false);
         layer->setBuffer(buffer, fence, 10, 20, false, mClientCache, 1, std::nullopt,
                          {/*vsyncId*/ 3, /*inputEventId*/ 0}, nullptr /* releaseBufferCallback */);
         EXPECT_EQ(2u, layer->mCurrentState.bufferlessSurfaceFramesTX.size());
@@ -283,7 +301,10 @@
 
         sp<Fence> fence1(new Fence());
         auto acquireFence1 = fenceFactory.createFenceTimeForTest(fence1);
-        sp<GraphicBuffer> buffer1{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
+        const auto buffer1 = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888,
+                                                                 1, 0),
+                                               mRenderEngine, false);
         layer->setBuffer(buffer1, fence1, 10, 20, false, mClientCache, 1, std::nullopt,
                          {/*vsyncId*/ 1, /*inputEventId*/ 0}, nullptr /* releaseBufferCallback */);
         ASSERT_NE(nullptr, layer->mCurrentState.bufferSurfaceFrameTX);
@@ -291,7 +312,10 @@
 
         sp<Fence> fence2(new Fence());
         auto acquireFence2 = fenceFactory.createFenceTimeForTest(fence2);
-        sp<GraphicBuffer> buffer2{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
+        const auto buffer2 = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888,
+                                                                 1, 0),
+                                               mRenderEngine, false);
         layer->setBuffer(buffer2, fence2, 10, 20, false, mClientCache, 1, std::nullopt,
                          {/*vsyncId*/ 1, /*inputEventId*/ 0}, nullptr /* releaseBufferCallback */);
         acquireFence2->signalForTest(12);
@@ -318,7 +342,10 @@
 
         sp<Fence> fence1(new Fence());
         auto acquireFence1 = fenceFactory.createFenceTimeForTest(fence1);
-        sp<GraphicBuffer> buffer1{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
+        const auto buffer1 = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888,
+                                                                 1, 0),
+                                               mRenderEngine, false);
         layer->setBuffer(buffer1, fence1, 10, 20, false, mClientCache, 1, std::nullopt,
                          {/*vsyncId*/ 1, /*inputEventId*/ 0}, nullptr /* releaseBufferCallback */);
         EXPECT_EQ(0u, layer->mCurrentState.bufferlessSurfaceFramesTX.size());
@@ -327,7 +354,10 @@
 
         sp<Fence> fence2(new Fence());
         auto acquireFence2 = fenceFactory.createFenceTimeForTest(fence2);
-        sp<GraphicBuffer> buffer2{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
+        const auto buffer2 = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888,
+                                                                 1, 0),
+                                               mRenderEngine, false);
         auto dropStartTime1 = systemTime();
         layer->setBuffer(buffer2, fence2, 10, 20, false, mClientCache, 1, std::nullopt,
                          {/*vsyncId*/ FrameTimelineInfo::INVALID_VSYNC_ID, /*inputEventId*/ 0},
@@ -339,7 +369,10 @@
 
         sp<Fence> fence3(new Fence());
         auto acquireFence3 = fenceFactory.createFenceTimeForTest(fence3);
-        sp<GraphicBuffer> buffer3{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
+        const auto buffer3 = std::make_shared<
+                renderengine::ExternalTexture>(new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888,
+                                                                 1, 0),
+                                               mRenderEngine, false);
         auto dropStartTime2 = systemTime();
         layer->setBuffer(buffer3, fence3, 10, 20, false, mClientCache, 1, std::nullopt,
                          {/*vsyncId*/ 2, /*inputEventId*/ 0}, nullptr /* releaseBufferCallback */);
@@ -379,7 +412,11 @@
         std::vector<std::shared_ptr<frametimeline::SurfaceFrame>> bufferlessSurfaceFrames;
         for (int i = 0; i < 10; i += 2) {
             sp<Fence> fence1(new Fence());
-            sp<GraphicBuffer> buffer1{new GraphicBuffer(1, 1, HAL_PIXEL_FORMAT_RGBA_8888, 1, 0)};
+            const auto buffer1 = std::make_shared<
+                    renderengine::ExternalTexture>(new GraphicBuffer(1, 1,
+                                                                     HAL_PIXEL_FORMAT_RGBA_8888, 1,
+                                                                     0),
+                                                   mRenderEngine, false);
             layer->setBuffer(buffer1, fence1, 10, 20, false, mClientCache, 1, std::nullopt,
                              {/*vsyncId*/ 1, /*inputEventId*/ 0},
                              nullptr /* releaseBufferCallback */);