/*
 * Copyright 2022 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 "GuestFrameComposer.h"

#include <android-base/parseint.h>
#include <android-base/properties.h>
#include <android-base/strings.h>
#include <android/hardware/graphics/common/1.0/types.h>
#include <drm_fourcc.h>
#include <libyuv.h>
#include <sync/sync.h>
#include <ui/GraphicBuffer.h>
#include <ui/GraphicBufferAllocator.h>
#include <ui/GraphicBufferMapper.h>

#include "Display.h"
#include "DisplayFinder.h"
#include "Drm.h"
#include "Layer.h"

namespace aidl::android::hardware::graphics::composer3::impl {
namespace {

using ::android::hardware::graphics::common::V1_0::ColorTransform;

uint32_t AlignToPower2(uint32_t val, uint8_t align_log) {
    uint32_t align = 1 << align_log;
    return ((val + (align - 1)) / align) * align;
}

bool LayerNeedsScaling(const Layer& layer) {
    common::Rect crop = layer.getSourceCropInt();
    common::Rect frame = layer.getDisplayFrame();

    int fromW = crop.right - crop.left;
    int fromH = crop.bottom - crop.top;
    int toW = frame.right - frame.left;
    int toH = frame.bottom - frame.top;

    bool not_rot_scale = fromW != toW || fromH != toH;
    bool rot_scale = fromW != toH || fromH != toW;

    bool needs_rot = static_cast<int32_t>(layer.getTransform()) &
                     static_cast<int32_t>(common::Transform::ROT_90);

    return needs_rot ? rot_scale : not_rot_scale;
}

bool LayerNeedsBlending(const Layer& layer) {
    return layer.getBlendMode() != common::BlendMode::NONE;
}

bool LayerNeedsAttenuation(const Layer& layer) {
    return layer.getBlendMode() == common::BlendMode::COVERAGE;
}

struct BufferSpec;
typedef int (*ConverterFunction)(const BufferSpec& src, const BufferSpec& dst, bool v_flip);
int DoCopy(const BufferSpec& src, const BufferSpec& dst, bool vFlip);
int ConvertFromRGB565(const BufferSpec& src, const BufferSpec& dst, bool vFlip);
int ConvertFromYV12(const BufferSpec& src, const BufferSpec& dst, bool vFlip);

ConverterFunction GetConverterForDrmFormat(uint32_t drmFormat) {
    switch (drmFormat) {
        case DRM_FORMAT_ABGR8888:
        case DRM_FORMAT_XBGR8888:
            return &DoCopy;
        case DRM_FORMAT_RGB565:
            return &ConvertFromRGB565;
        case DRM_FORMAT_YVU420:
            return &ConvertFromYV12;
    }
    DEBUG_LOG("Unsupported drm format: %d(%s), returning null converter", drmFormat,
              GetDrmFormatString(drmFormat));
    return nullptr;
}

bool IsDrmFormatSupported(uint32_t drmFormat) {
    return GetConverterForDrmFormat(drmFormat) != nullptr;
}

// Libyuv's convert functions only allow the combination of any rotation
// (multiple of 90 degrees) and a vertical flip, but not horizontal flips.
// Surfaceflinger's transformations are expressed in terms of a vertical flip,
// a horizontal flip and/or a single 90 degrees clockwise rotation (see
// NATIVE_WINDOW_TRANSFORM_HINT documentation on system/window.h for more
// insight). The following code allows to turn a horizontal flip into a 180
// degrees rotation and a vertical flip.
libyuv::RotationMode GetRotationFromTransform(common::Transform transform) {
    uint32_t rotation = 0;
    rotation += (static_cast<int32_t>(transform) & static_cast<int32_t>(common::Transform::ROT_90))
                    ? 1
                    : 0;  // 1 * ROT90 bit
    rotation += (static_cast<int32_t>(transform) & static_cast<int32_t>(common::Transform::FLIP_H))
                    ? 2
                    : 0;  // 2 * VFLIP bit
    return static_cast<libyuv::RotationMode>(90 * rotation);
}

bool GetVFlipFromTransform(common::Transform transform) {
    // vertical flip xor horizontal flip
    bool hasVFlip =
        static_cast<int32_t>(transform) & static_cast<int32_t>(common::Transform::FLIP_V);
    bool hasHFlip =
        static_cast<int32_t>(transform) & static_cast<int32_t>(common::Transform::FLIP_H);
    return hasVFlip ^ hasHFlip;
}

struct BufferSpec {
    uint8_t* buffer;
    std::optional<android_ycbcr> buffer_ycbcr;
    uint32_t width;
    uint32_t height;
    uint32_t cropX;
    uint32_t cropY;
    uint32_t cropWidth;
    uint32_t cropHeight;
    uint32_t drmFormat;
    uint32_t strideBytes;
    uint32_t sampleBytes;

    BufferSpec() = default;

    BufferSpec(uint8_t* buffer, std::optional<android_ycbcr> buffer_ycbcr, uint32_t width,
               uint32_t height, uint32_t cropX, uint32_t cropY, uint32_t cropWidth,
               uint32_t cropHeight, uint32_t drmFormat, uint32_t strideBytes, uint32_t sampleBytes)
        : buffer(buffer),
          buffer_ycbcr(buffer_ycbcr),
          width(width),
          height(height),
          cropX(cropX),
          cropY(cropY),
          cropWidth(cropWidth),
          cropHeight(cropHeight),
          drmFormat(drmFormat),
          strideBytes(strideBytes),
          sampleBytes(sampleBytes) {}

    BufferSpec(uint8_t* buffer, uint32_t width, uint32_t height, uint32_t strideBytes)
        : BufferSpec(buffer,
                     /*buffer_ycbcr=*/std::nullopt, width, height,
                     /*cropX=*/0,
                     /*cropY=*/0,
                     /*cropWidth=*/width,
                     /*cropHeight=*/height,
                     /*drmFormat=*/DRM_FORMAT_ABGR8888, strideBytes,
                     /*sampleBytes=*/4) {}
};

int DoFill(const BufferSpec& dst, const Color& color) {
    ATRACE_CALL();

    const uint8_t r = static_cast<uint8_t>(color.r * 255.0f);
    const uint8_t g = static_cast<uint8_t>(color.g * 255.0f);
    const uint8_t b = static_cast<uint8_t>(color.b * 255.0f);
    const uint8_t a = static_cast<uint8_t>(color.a * 255.0f);

    const uint32_t rgba = static_cast<uint32_t>(r) | static_cast<uint32_t>(g) << 8 |
                          static_cast<uint32_t>(b) << 16 | static_cast<uint32_t>(a) << 24;

    // Point to the upper left corner of the crop rectangle.
    uint8_t* dstBuffer = dst.buffer + dst.cropY * dst.strideBytes + dst.cropX * dst.sampleBytes;

    libyuv::SetPlane(dstBuffer, static_cast<int>(dst.strideBytes), static_cast<int>(dst.cropWidth),
                     static_cast<int>(dst.cropHeight), rgba);
    return 0;
}

int ConvertFromRGB565(const BufferSpec& src, const BufferSpec& dst, bool vFlip) {
    ATRACE_CALL();

    // Point to the upper left corner of the crop rectangle
    uint8_t* srcBuffer = src.buffer + src.cropY * src.strideBytes + src.cropX * src.sampleBytes;
    const int srcStrideBytes = static_cast<int>(src.strideBytes);
    uint8_t* dstBuffer = dst.buffer + dst.cropY * dst.strideBytes + dst.cropX * dst.sampleBytes;
    const int dstStrideBytes = static_cast<int>(dst.strideBytes);

    int width = static_cast<int>(src.cropWidth);
    int height = static_cast<int>(src.cropHeight);
    if (vFlip) {
        height = -height;
    }

    return libyuv::RGB565ToARGB(srcBuffer, srcStrideBytes,  //
                                dstBuffer, dstStrideBytes,  //
                                width, height);
}

int ConvertFromYV12(const BufferSpec& src, const BufferSpec& dst, bool vFlip) {
    ATRACE_CALL();

    // The following calculation of plane offsets and alignments are based on
    // swiftshader's Sampler::setTextureLevel() implementation
    // (Renderer/Sampler.cpp:225)

    auto& srcBufferYCbCrOpt = src.buffer_ycbcr;
    if (!srcBufferYCbCrOpt) {
        ALOGE("%s called on non ycbcr buffer", __FUNCTION__);
        return -1;
    }
    auto& srcBufferYCbCr = *srcBufferYCbCrOpt;

    // The libyuv::I420ToARGB() function is for tri-planar.
    if (srcBufferYCbCr.chroma_step != 1) {
        ALOGE("%s called with bad chroma step", __FUNCTION__);
        return -1;
    }

    uint8_t* srcY = reinterpret_cast<uint8_t*>(srcBufferYCbCr.y);
    const int strideYBytes = static_cast<int>(srcBufferYCbCr.ystride);
    uint8_t* srcU = reinterpret_cast<uint8_t*>(srcBufferYCbCr.cb);
    const int strideUBytes = static_cast<int>(srcBufferYCbCr.cstride);
    uint8_t* srcV = reinterpret_cast<uint8_t*>(srcBufferYCbCr.cr);
    const int strideVBytes = static_cast<int>(srcBufferYCbCr.cstride);

    // Adjust for crop
    srcY += src.cropY * srcBufferYCbCr.ystride + src.cropX;
    srcV += (src.cropY / 2) * srcBufferYCbCr.cstride + (src.cropX / 2);
    srcU += (src.cropY / 2) * srcBufferYCbCr.cstride + (src.cropX / 2);
    uint8_t* dstBuffer = dst.buffer + dst.cropY * dst.strideBytes + dst.cropX * dst.sampleBytes;
    const int dstStrideBytes = static_cast<int>(dst.strideBytes);

    int width = static_cast<int>(dst.cropWidth);
    int height = static_cast<int>(dst.cropHeight);

    if (vFlip) {
        height = -height;
    }

    // YV12 is the same as I420, with the U and V planes swapped
    return libyuv::I420ToARGB(srcY, strideYBytes,  //
                              srcV, strideVBytes,  //
                              srcU, strideUBytes,  //
                              dstBuffer, dstStrideBytes, width, height);
}

int DoConversion(const BufferSpec& src, const BufferSpec& dst, bool v_flip) {
    ConverterFunction func = GetConverterForDrmFormat(src.drmFormat);
    if (!func) {
        // GetConverterForDrmFormat should've logged the issue for us.
        return -1;
    }
    return func(src, dst, v_flip);
}

int DoCopy(const BufferSpec& src, const BufferSpec& dst, bool v_flip) {
    ATRACE_CALL();

    // Point to the upper left corner of the crop rectangle
    uint8_t* srcBuffer = src.buffer + src.cropY * src.strideBytes + src.cropX * src.sampleBytes;
    const int srcStrideBytes = static_cast<int>(src.strideBytes);
    uint8_t* dstBuffer = dst.buffer + dst.cropY * dst.strideBytes + dst.cropX * dst.sampleBytes;
    const int dstStrideBytes = static_cast<int>(dst.strideBytes);
    int width = static_cast<int>(src.cropWidth);
    int height = static_cast<int>(src.cropHeight);

    if (v_flip) {
        height = -height;
    }

    // HAL formats are named based on the order of the pixel components on the
    // byte stream, while libyuv formats are named based on the order of those
    // pixel components in an integer written from left to right. So
    // libyuv::FOURCC_ARGB is equivalent to HAL_PIXEL_FORMAT_BGRA_8888.
    auto ret = libyuv::ARGBCopy(srcBuffer, srcStrideBytes,  //
                                dstBuffer, dstStrideBytes,  //
                                width, height);
    return ret;
}

int DoRotation(const BufferSpec& src, const BufferSpec& dst, libyuv::RotationMode rotation,
               bool v_flip) {
    ATRACE_CALL();

    // Point to the upper left corner of the crop rectangles
    uint8_t* srcBuffer = src.buffer + src.cropY * src.strideBytes + src.cropX * src.sampleBytes;
    const int srcStrideBytes = static_cast<int>(src.strideBytes);
    uint8_t* dstBuffer = dst.buffer + dst.cropY * dst.strideBytes + dst.cropX * dst.sampleBytes;
    const int dstStrideBytes = static_cast<int>(dst.strideBytes);
    int width = static_cast<int>(src.cropWidth);
    int height = static_cast<int>(src.cropHeight);

    if (v_flip) {
        height = -height;
    }

    return libyuv::ARGBRotate(srcBuffer, srcStrideBytes,  //
                              dstBuffer, dstStrideBytes,  //
                              width, height, rotation);
}

int DoScaling(const BufferSpec& src, const BufferSpec& dst, bool v_flip) {
    ATRACE_CALL();

    // Point to the upper left corner of the crop rectangles
    uint8_t* srcBuffer = src.buffer + src.cropY * src.strideBytes + src.cropX * src.sampleBytes;
    uint8_t* dstBuffer = dst.buffer + dst.cropY * dst.strideBytes + dst.cropX * dst.sampleBytes;
    const int srcStrideBytes = static_cast<int>(src.strideBytes);
    const int dstStrideBytes = static_cast<int>(dst.strideBytes);
    const int srcWidth = static_cast<int>(src.cropWidth);
    int srcHeight = static_cast<int>(src.cropHeight);
    const int dstWidth = static_cast<int>(dst.cropWidth);
    const int dstHeight = static_cast<int>(dst.cropHeight);

    if (v_flip) {
        srcHeight = -srcHeight;
    }

    return libyuv::ARGBScale(srcBuffer, srcStrideBytes, srcWidth, srcHeight, dstBuffer,
                             dstStrideBytes, dstWidth, dstHeight, libyuv::kFilterBilinear);
}

int DoAttenuation(const BufferSpec& src, const BufferSpec& dst, bool v_flip) {
    ATRACE_CALL();

    // Point to the upper left corner of the crop rectangles
    uint8_t* srcBuffer = src.buffer + src.cropY * src.strideBytes + src.cropX * src.sampleBytes;
    uint8_t* dstBuffer = dst.buffer + dst.cropY * dst.strideBytes + dst.cropX * dst.sampleBytes;
    const int srcStrideBytes = static_cast<int>(src.strideBytes);
    const int dstStrideBytes = static_cast<int>(dst.strideBytes);
    const int width = static_cast<int>(dst.cropWidth);
    int height = static_cast<int>(dst.cropHeight);
    if (v_flip) {
        height = -height;
    }

    return libyuv::ARGBAttenuate(srcBuffer, srcStrideBytes,  //
                                 dstBuffer, dstStrideBytes,  //
                                 width, height);
}

int DoBlending(const BufferSpec& src, const BufferSpec& dst, bool v_flip) {
    ATRACE_CALL();

    // Point to the upper left corner of the crop rectangles
    uint8_t* srcBuffer = src.buffer + src.cropY * src.strideBytes + src.cropX * src.sampleBytes;
    uint8_t* dstBuffer = dst.buffer + dst.cropY * dst.strideBytes + dst.cropX * dst.sampleBytes;
    const int srcStrideBytes = static_cast<int>(src.strideBytes);
    const int dstStrideBytes = static_cast<int>(dst.strideBytes);
    const int width = static_cast<int>(dst.cropWidth);
    int height = static_cast<int>(dst.cropHeight);
    if (v_flip) {
        height = -height;
    }

    // libyuv's ARGB format is hwcomposer's BGRA format, since blending only cares
    // for the position of alpha in the pixel and not the position of the colors
    // this function is perfectly usable.
    return libyuv::ARGBBlend(srcBuffer, srcStrideBytes,  //
                             dstBuffer, dstStrideBytes,  //
                             dstBuffer, dstStrideBytes,  //
                             width, height);
}

std::optional<BufferSpec> GetBufferSpec(GrallocBuffer& buffer, GrallocBufferView& bufferView,
                                        const common::Rect& bufferCrop) {
    auto bufferFormatOpt = buffer.GetDrmFormat();
    if (!bufferFormatOpt) {
        ALOGE("Failed to get gralloc buffer format.");
        return std::nullopt;
    }
    uint32_t bufferFormat = *bufferFormatOpt;

    auto bufferWidthOpt = buffer.GetWidth();
    if (!bufferWidthOpt) {
        ALOGE("Failed to get gralloc buffer width.");
        return std::nullopt;
    }
    uint32_t bufferWidth = *bufferWidthOpt;

    auto bufferHeightOpt = buffer.GetHeight();
    if (!bufferHeightOpt) {
        ALOGE("Failed to get gralloc buffer height.");
        return std::nullopt;
    }
    uint32_t bufferHeight = *bufferHeightOpt;

    uint8_t* bufferData = nullptr;
    uint32_t bufferStrideBytes = 0;
    std::optional<android_ycbcr> bufferYCbCrData;

    if (bufferFormat == DRM_FORMAT_NV12 || bufferFormat == DRM_FORMAT_NV21 ||
        bufferFormat == DRM_FORMAT_YVU420) {
        bufferYCbCrData = bufferView.GetYCbCr();
        if (!bufferYCbCrData) {
            ALOGE("%s failed to get raw ycbcr from view.", __FUNCTION__);
            return std::nullopt;
        }
    } else {
        auto bufferDataOpt = bufferView.Get();
        if (!bufferDataOpt) {
            ALOGE("%s failed to lock gralloc buffer.", __FUNCTION__);
            return std::nullopt;
        }
        bufferData = reinterpret_cast<uint8_t*>(*bufferDataOpt);

        auto bufferStrideBytesOpt = buffer.GetMonoPlanarStrideBytes();
        if (!bufferStrideBytesOpt) {
            ALOGE("%s failed to get plane stride.", __FUNCTION__);
            return std::nullopt;
        }
        bufferStrideBytes = *bufferStrideBytesOpt;
    }

    uint32_t bufferCropX = static_cast<uint32_t>(bufferCrop.left);
    uint32_t bufferCropY = static_cast<uint32_t>(bufferCrop.top);
    uint32_t bufferCropWidth = static_cast<uint32_t>(bufferCrop.right - bufferCrop.left);
    uint32_t bufferCropHeight = static_cast<uint32_t>(bufferCrop.bottom - bufferCrop.top);

    return BufferSpec(bufferData, bufferYCbCrData, bufferWidth, bufferHeight, bufferCropX,
                      bufferCropY, bufferCropWidth, bufferCropHeight, bufferFormat,
                      bufferStrideBytes, GetDrmFormatBytesPerPixel(bufferFormat));
}

}  // namespace

HWC3::Error GuestFrameComposer::init() {
    DEBUG_LOG("%s", __FUNCTION__);

    HWC3::Error error = mDrmClient.init();
    if (error != HWC3::Error::None) {
        ALOGE("%s: failed to initialize DrmClient", __FUNCTION__);
        return error;
    }

    return HWC3::Error::None;
}

HWC3::Error GuestFrameComposer::registerOnHotplugCallback(const HotplugCallback& cb) {
    return mDrmClient.registerOnHotplugCallback(cb);
    return HWC3::Error::None;
}

HWC3::Error GuestFrameComposer::unregisterOnHotplugCallback() {
    return mDrmClient.unregisterOnHotplugCallback();
}

HWC3::Error GuestFrameComposer::onDisplayCreate(Display* display) {
    const uint32_t displayId = static_cast<uint32_t>(display->getId());
    int32_t displayConfigId;
    int32_t displayWidth;
    int32_t displayHeight;

    HWC3::Error error = display->getActiveConfig(&displayConfigId);
    if (error != HWC3::Error::None) {
        ALOGE("%s: display:%" PRIu32 " has no active config", __FUNCTION__, displayId);
        return error;
    }

    error = display->getDisplayAttribute(displayConfigId, DisplayAttribute::WIDTH, &displayWidth);
    if (error != HWC3::Error::None) {
        ALOGE("%s: display:%" PRIu32 " failed to get width", __FUNCTION__, displayId);
        return error;
    }

    error = display->getDisplayAttribute(displayConfigId, DisplayAttribute::HEIGHT, &displayHeight);
    if (error != HWC3::Error::None) {
        ALOGE("%s: display:%" PRIu32 " failed to get height", __FUNCTION__, displayId);
        return error;
    }

    auto it = mDisplayInfos.find(displayId);
    if (it != mDisplayInfos.end()) {
        ALOGE("%s: display:%" PRIu32 " already created?", __FUNCTION__, displayId);
    }

    DisplayInfo& displayInfo = mDisplayInfos[displayId];

    displayInfo.swapchain = DrmSwapchain::create(static_cast<uint32_t>(displayWidth),
                                                 static_cast<uint32_t>(displayHeight),
                                                 ::android::GraphicBuffer::USAGE_HW_COMPOSER |
                                                   ::android::GraphicBuffer::USAGE_SW_READ_OFTEN |
                                                   ::android::GraphicBuffer::USAGE_SW_WRITE_OFTEN,
                                                 &mDrmClient);

    if (displayId == 0) {
        auto compositionResult = displayInfo.swapchain->getNextImage();
        auto [flushError, flushSyncFd] =
                mDrmClient.flushToDisplay(displayId, compositionResult->getDrmBuffer(), -1);
        if (flushError != HWC3::Error::None) {
            ALOGW(
                "%s: Initial display flush failed. HWComposer assuming that we are "
                "running in QEMU without a display and disabling presenting.",
                __FUNCTION__);
            mPresentDisabled = true;
        } else {
            compositionResult->markAsInUse(std::move(flushSyncFd));
        }
    }

    std::optional<std::vector<uint8_t>> edid = mDrmClient.getEdid(displayId);
    if (edid) {
        display->setEdid(*edid);
    }

    return HWC3::Error::None;
}

HWC3::Error GuestFrameComposer::onDisplayDestroy(Display* display) {
    auto displayId = display->getId();

    auto it = mDisplayInfos.find(displayId);
    if (it == mDisplayInfos.end()) {
        ALOGE("%s: display:%" PRIu64 " missing display buffers?", __FUNCTION__,
            displayId);
        return HWC3::Error::BadDisplay;
    }
    mDisplayInfos.erase(it);

    return HWC3::Error::None;
}

HWC3::Error GuestFrameComposer::onDisplayClientTargetSet(Display*) { return HWC3::Error::None; }

HWC3::Error GuestFrameComposer::onActiveConfigChange(Display* /*display*/) {
    return HWC3::Error::None;
};

HWC3::Error GuestFrameComposer::getDisplayConfigsFromSystemProp(
    std::vector<GuestFrameComposer::DisplayConfig>* configs) {
    DEBUG_LOG("%s", __FUNCTION__);

    std::vector<int> propIntParts;
    parseExternalDisplaysFromProperties(propIntParts);

    while (!propIntParts.empty()) {
        DisplayConfig display_config = {
            .width = propIntParts[1],
            .height = propIntParts[2],
            .dpiX = propIntParts[3],
            .dpiY = propIntParts[3],
            .refreshRateHz = 160,
        };

        configs->push_back(display_config);

        propIntParts.erase(propIntParts.begin(), propIntParts.begin() + 5);
    }

    return HWC3::Error::None;
}

HWC3::Error GuestFrameComposer::validateDisplay(Display* display, DisplayChanges* outChanges) {
    const auto displayId = display->getId();
    DEBUG_LOG("%s display:%" PRIu64, __FUNCTION__, displayId);

    const std::vector<Layer*>& layers = display->getOrderedLayers();

    bool fallbackToClientComposition = false;
    for (Layer* layer : layers) {
        const auto layerId = layer->getId();
        const auto layerCompositionType = layer->getCompositionType();
        const auto layerCompositionTypeString = toString(layerCompositionType);

        if (layerCompositionType == Composition::INVALID) {
            ALOGE("%s display:%" PRIu64 " layer:%" PRIu64 " has Invalid composition", __FUNCTION__,
                  displayId, layerId);
            continue;
        }

        if (layerCompositionType == Composition::CLIENT ||
            layerCompositionType == Composition::CURSOR ||
            layerCompositionType == Composition::SIDEBAND) {
            DEBUG_LOG("%s: display:%" PRIu64 " layer:%" PRIu64
                      " has composition type %s, falling back to client composition",
                      __FUNCTION__, displayId, layerId, layerCompositionTypeString.c_str());
            fallbackToClientComposition = true;
            break;
        }

        if (layerCompositionType == Composition::DISPLAY_DECORATION) {
            return HWC3::Error::Unsupported;
        }

        if (!canComposeLayer(layer)) {
            DEBUG_LOG("%s: display:%" PRIu64 " layer:%" PRIu64
                      " composition not supported, falling back to client composition",
                      __FUNCTION__, displayId, layerId);
            fallbackToClientComposition = true;
            break;
        }
    }

    if (fallbackToClientComposition) {
        for (Layer* layer : layers) {
            const auto layerId = layer->getId();
            const auto layerCompositionType = layer->getCompositionType();

            if (layerCompositionType == Composition::INVALID) {
                continue;
            }

            if (layerCompositionType != Composition::CLIENT) {
                DEBUG_LOG("%s display:%" PRIu64 " layer:%" PRIu64 "composition updated to Client",
                          __FUNCTION__, displayId, layerId);

                outChanges->addLayerCompositionChange(displayId, layerId, Composition::CLIENT);
            }
        }
    }

    // We can not draw below a Client (SurfaceFlinger) composed layer. Change all
    // layers below a Client composed layer to also be Client composed.
    if (layers.size() > 1) {
        for (std::size_t layerIndex = layers.size() - 1; layerIndex > 0; layerIndex--) {
            auto layer = layers[layerIndex];
            auto layerCompositionType = layer->getCompositionType();

            if (layerCompositionType == Composition::CLIENT) {
                for (std::size_t lowerLayerIndex = 0; lowerLayerIndex < layerIndex;
                     lowerLayerIndex++) {
                    auto lowerLayer = layers[lowerLayerIndex];
                    auto lowerLayerId = lowerLayer->getId();
                    auto lowerLayerCompositionType = lowerLayer->getCompositionType();

                    if (lowerLayerCompositionType != Composition::CLIENT) {
                        DEBUG_LOG("%s: display:%" PRIu64 " changing layer:%" PRIu64
                                  " to Client because"
                                  "hwcomposer can not draw below the Client composed "
                                  "layer:%" PRIu64,
                                  __FUNCTION__, displayId, lowerLayerId, layer->getId());

                        outChanges->addLayerCompositionChange(displayId, lowerLayerId,
                                                              Composition::CLIENT);
                    }
                }
            }
        }
    }

    return HWC3::Error::None;
}

HWC3::Error GuestFrameComposer::presentDisplay(
    Display* display, ::android::base::unique_fd* outDisplayFence,
    std::unordered_map<int64_t, ::android::base::unique_fd>* /*outLayerFences*/) {
    const uint32_t displayId = static_cast<uint32_t>(display->getId());
    DEBUG_LOG("%s display:%" PRIu32, __FUNCTION__, displayId);

    if (mPresentDisabled) {
        return HWC3::Error::None;
    }

    auto it = mDisplayInfos.find(displayId);
    if (it == mDisplayInfos.end()) {
        ALOGE("%s: display:%" PRIu32 " not found", __FUNCTION__, displayId);
        return HWC3::Error::NoResources;
    }

    DisplayInfo& displayInfo = it->second;

    auto compositionResult = displayInfo.swapchain->getNextImage();
    compositionResult->wait();

    if (compositionResult->getBuffer() == nullptr) {
        ALOGE("%s: display:%" PRIu32 " missing composition result buffer",
            __FUNCTION__, displayId);
        return HWC3::Error::NoResources;
    }

    if (compositionResult->getDrmBuffer() == nullptr) {
        ALOGE("%s: display:%" PRIu32 " missing composition result drm buffer",
            __FUNCTION__, displayId);
        return HWC3::Error::NoResources;
    }

    std::optional<GrallocBuffer> compositionResultBufferOpt =
        mGralloc.Import(compositionResult->getBuffer());
    if (!compositionResultBufferOpt) {
        ALOGE("%s: display:%" PRIu32 " failed to import buffer", __FUNCTION__, displayId);
        return HWC3::Error::NoResources;
    }

    std::optional<uint32_t> compositionResultBufferWidthOpt =
        compositionResultBufferOpt->GetWidth();
    if (!compositionResultBufferWidthOpt) {
        ALOGE("%s: display:%" PRIu32 " failed to query buffer width", __FUNCTION__, displayId);
        return HWC3::Error::NoResources;
    }

    std::optional<uint32_t> compositionResultBufferHeightOpt =
        compositionResultBufferOpt->GetHeight();
    if (!compositionResultBufferHeightOpt) {
        ALOGE("%s: display:%" PRIu32 " failed to query buffer height", __FUNCTION__, displayId);
        return HWC3::Error::NoResources;
    }

    std::optional<uint32_t> compositionResultBufferStrideOpt =
        compositionResultBufferOpt->GetMonoPlanarStrideBytes();
    if (!compositionResultBufferStrideOpt) {
        ALOGE("%s: display:%" PRIu32 " failed to query buffer stride", __FUNCTION__, displayId);
        return HWC3::Error::NoResources;
    }

    std::optional<GrallocBufferView> compositionResultBufferViewOpt =
        compositionResultBufferOpt->Lock();
    if (!compositionResultBufferViewOpt) {
        ALOGE("%s: display:%" PRIu32 " failed to get buffer view", __FUNCTION__, displayId);
        return HWC3::Error::NoResources;
    }

    const std::optional<void*> compositionResultBufferDataOpt =
        compositionResultBufferViewOpt->Get();
    if (!compositionResultBufferDataOpt) {
        ALOGE("%s: display:%" PRIu32 " failed to get buffer data", __FUNCTION__, displayId);
        return HWC3::Error::NoResources;
    }

    uint32_t compositionResultBufferWidth = *compositionResultBufferWidthOpt;
    uint32_t compositionResultBufferHeight = *compositionResultBufferHeightOpt;
    uint32_t compositionResultBufferStride = *compositionResultBufferStrideOpt;
    uint8_t* compositionResultBufferData =
        reinterpret_cast<uint8_t*>(*compositionResultBufferDataOpt);

    const std::vector<Layer*>& layers = display->getOrderedLayers();

    const bool noOpComposition = layers.empty();
    const bool allLayersClientComposed = std::all_of(
        layers.begin(),  //
        layers.end(),    //
        [](const Layer* layer) { return layer->getCompositionType() == Composition::CLIENT; });

    if (noOpComposition) {
        ALOGW("%s: display:%" PRIu32 " empty composition", __FUNCTION__, displayId);
    } else if (allLayersClientComposed) {
        auto clientTargetBufferOpt = mGralloc.Import(display->waitAndGetClientTargetBuffer());
        if (!clientTargetBufferOpt) {
            ALOGE("%s: failed to import client target buffer.", __FUNCTION__);
            return HWC3::Error::NoResources;
        }
        GrallocBuffer& clientTargetBuffer = *clientTargetBufferOpt;

        auto clientTargetBufferViewOpt = clientTargetBuffer.Lock();
        if (!clientTargetBufferViewOpt) {
            ALOGE("%s: failed to lock client target buffer.", __FUNCTION__);
            return HWC3::Error::NoResources;
        }
        GrallocBufferView& clientTargetBufferView = *clientTargetBufferViewOpt;

        auto clientTargetPlaneLayoutsOpt = clientTargetBuffer.GetPlaneLayouts();
        if (!clientTargetPlaneLayoutsOpt) {
            ALOGE("Failed to get client target buffer plane layouts.");
            return HWC3::Error::NoResources;
        }
        auto& clientTargetPlaneLayouts = *clientTargetPlaneLayoutsOpt;

        if (clientTargetPlaneLayouts.size() != 1) {
            ALOGE("Unexpected number of plane layouts for client target buffer.");
            return HWC3::Error::NoResources;
        }

        std::size_t clientTargetPlaneSize =
            static_cast<std::size_t>(clientTargetPlaneLayouts[0].totalSizeInBytes);

        auto clientTargetDataOpt = clientTargetBufferView.Get();
        if (!clientTargetDataOpt) {
            ALOGE("%s failed to lock gralloc buffer.", __FUNCTION__);
            return HWC3::Error::NoResources;
        }
        auto* clientTargetData = reinterpret_cast<uint8_t*>(*clientTargetDataOpt);

        std::memcpy(compositionResultBufferData, clientTargetData, clientTargetPlaneSize);
    } else {
        for (Layer* layer : layers) {
            const auto layerId = layer->getId();
            const auto layerCompositionType = layer->getCompositionType();
            if (layerCompositionType != Composition::DEVICE &&
                layerCompositionType != Composition::SOLID_COLOR) {
                continue;
            }

            HWC3::Error error = composeLayerInto(displayInfo.compositionIntermediateStorage,  //
                                                 layer,                                       //
                                                 compositionResultBufferData,                 //
                                                 compositionResultBufferWidth,                //
                                                 compositionResultBufferHeight,               //
                                                 compositionResultBufferStride,               //
                                                 4);
            if (error != HWC3::Error::None) {
                ALOGE("%s: display:%" PRIu32 " failed to compose layer:%" PRIu64, __FUNCTION__,
                      displayId, layerId);
                return error;
            }
        }
    }

    if (display->hasColorTransform()) {
        HWC3::Error error = applyColorTransformToRGBA(display->getColorTransform(),   //
                                                      compositionResultBufferData,    //
                                                      compositionResultBufferWidth,   //
                                                      compositionResultBufferHeight,  //
                                                      compositionResultBufferStride);
        if (error != HWC3::Error::None) {
            ALOGE("%s: display:%" PRIu32 " failed to apply color transform", __FUNCTION__,
                  displayId);
            return error;
        }
    }

    DEBUG_LOG("%s display:%" PRIu32 " flushing drm buffer", __FUNCTION__,
                displayId);

    auto [error, fence] = mDrmClient.flushToDisplay(displayId, compositionResult->getDrmBuffer(), -1);
    if (error != HWC3::Error::None) {
        ALOGE("%s: display:%" PRIu32 " failed to flush drm buffer" PRIu64,
            __FUNCTION__, displayId);
    }

    *outDisplayFence = std::move(fence);
    compositionResult->markAsInUse(outDisplayFence->ok()
                                        ? ::android::base::unique_fd(dup(*outDisplayFence))
                                        : ::android::base::unique_fd());
    return error;
}

bool GuestFrameComposer::canComposeLayer(Layer* layer) {
    const auto layerCompositionType = layer->getCompositionType();
    if (layerCompositionType == Composition::SOLID_COLOR) {
        return true;
    }

    if (layerCompositionType != Composition::DEVICE) {
        return false;
    }

    buffer_handle_t bufferHandle = layer->getBuffer().getBuffer();
    if (bufferHandle == nullptr) {
        ALOGW("%s received a layer with a null handle", __FUNCTION__);
        return false;
    }

    auto bufferOpt = mGralloc.Import(bufferHandle);
    if (!bufferOpt) {
        ALOGE("Failed to import layer buffer.");
        return false;
    }
    GrallocBuffer& buffer = *bufferOpt;

    auto bufferFormatOpt = buffer.GetDrmFormat();
    if (!bufferFormatOpt) {
        ALOGE("Failed to get layer buffer format.");
        return false;
    }
    uint32_t bufferFormat = *bufferFormatOpt;

    if (!IsDrmFormatSupported(bufferFormat)) {
        return false;
    }

    return true;
}

HWC3::Error GuestFrameComposer::composeLayerInto(
    AlternatingImageStorage& compositionIntermediateStorage,
    Layer* srcLayer,                     //
    std::uint8_t* dstBuffer,             //
    std::uint32_t dstBufferWidth,        //
    std::uint32_t dstBufferHeight,       //
    std::uint32_t dstBufferStrideBytes,  //
    std::uint32_t dstBufferBytesPerPixel) {
    ATRACE_CALL();

    libyuv::RotationMode rotation = GetRotationFromTransform(srcLayer->getTransform());

    common::Rect srcLayerCrop = srcLayer->getSourceCropInt();
    common::Rect srcLayerDisplayFrame = srcLayer->getDisplayFrame();

    BufferSpec srcLayerSpec;

    std::optional<GrallocBuffer> srcBufferOpt;
    std::optional<GrallocBufferView> srcBufferViewOpt;

    const auto srcLayerCompositionType = srcLayer->getCompositionType();
    if (srcLayerCompositionType == Composition::DEVICE) {
        srcBufferOpt = mGralloc.Import(srcLayer->waitAndGetBuffer());
        if (!srcBufferOpt) {
            ALOGE("%s: failed to import layer buffer.", __FUNCTION__);
            return HWC3::Error::NoResources;
        }
        GrallocBuffer& srcBuffer = *srcBufferOpt;

        srcBufferViewOpt = srcBuffer.Lock();
        if (!srcBufferViewOpt) {
            ALOGE("%s: failed to lock import layer buffer.", __FUNCTION__);
            return HWC3::Error::NoResources;
        }
        GrallocBufferView& srcBufferView = *srcBufferViewOpt;

        auto srcLayerSpecOpt = GetBufferSpec(srcBuffer, srcBufferView, srcLayerCrop);
        if (!srcLayerSpecOpt) {
            return HWC3::Error::NoResources;
        }

        srcLayerSpec = *srcLayerSpecOpt;
    } else if (srcLayerCompositionType == Composition::SOLID_COLOR) {
        // srcLayerSpec not used by `needsFill` below.
    }

    // TODO(jemoreira): Remove the hardcoded fomat.
    bool needsFill = srcLayerCompositionType == Composition::SOLID_COLOR;
    bool needsConversion = srcLayerCompositionType == Composition::DEVICE &&
                           srcLayerSpec.drmFormat != DRM_FORMAT_XBGR8888 &&
                           srcLayerSpec.drmFormat != DRM_FORMAT_ABGR8888;
    bool needsScaling = LayerNeedsScaling(*srcLayer);
    bool needsRotation = rotation != libyuv::kRotate0;
    bool needsTranspose = needsRotation && rotation != libyuv::kRotate180;
    bool needsVFlip = GetVFlipFromTransform(srcLayer->getTransform());
    bool needsAttenuation = LayerNeedsAttenuation(*srcLayer);
    bool needsBlending = LayerNeedsBlending(*srcLayer);
    bool needsCopy = !(needsConversion || needsScaling || needsRotation || needsVFlip ||
                       needsAttenuation || needsBlending);

    BufferSpec dstLayerSpec(
        dstBuffer,
        /*buffer_ycbcr=*/std::nullopt, dstBufferWidth, dstBufferHeight,
        static_cast<uint32_t>(srcLayerDisplayFrame.left),
        static_cast<uint32_t>(srcLayerDisplayFrame.top),
        static_cast<uint32_t>(srcLayerDisplayFrame.right - srcLayerDisplayFrame.left),
        static_cast<uint32_t>(srcLayerDisplayFrame.bottom - srcLayerDisplayFrame.top),
        DRM_FORMAT_XBGR8888, dstBufferStrideBytes, dstBufferBytesPerPixel);

    // Add the destination layer to the bottom of the buffer stack
    std::vector<BufferSpec> dstBufferStack(1, dstLayerSpec);

    // If more than operation is to be performed, a temporary buffer is needed for
    // each additional operation

    // N operations need N destination buffers, the destination layer (the
    // framebuffer) is one of them, so only N-1 temporary buffers are needed.
    // Vertical flip is not taken into account because it can be done together
    // with any other operation.
    int neededIntermediateImages = (needsFill ? 1 : 0) + (needsConversion ? 1 : 0) +
                                   (needsScaling ? 1 : 0) + (needsRotation ? 1 : 0) +
                                   (needsAttenuation ? 1 : 0) + (needsBlending ? 1 : 0) +
                                   (needsCopy ? 1 : 0) - 1;

    uint32_t mScratchBufferWidth =
        static_cast<uint32_t>(srcLayerDisplayFrame.right - srcLayerDisplayFrame.left);
    uint32_t mScratchBufferHeight =
        static_cast<uint32_t>(srcLayerDisplayFrame.bottom - srcLayerDisplayFrame.top);
    uint32_t mScratchBufferStrideBytes =
        AlignToPower2(mScratchBufferWidth * dstBufferBytesPerPixel, 4);
    uint32_t mScratchBufferSizeBytes = mScratchBufferHeight * mScratchBufferStrideBytes;

    for (uint32_t i = 0; i < neededIntermediateImages; i++) {
        BufferSpec mScratchBufferspec(
            compositionIntermediateStorage.getRotatingScratchBuffer(mScratchBufferSizeBytes, i),
            mScratchBufferWidth, mScratchBufferHeight, mScratchBufferStrideBytes);
        dstBufferStack.push_back(mScratchBufferspec);
    }

    // Filling, conversion, and scaling should always be the first operations, so
    // that every other operation works on equally sized frames (guaranteed to fit
    // in the scratch buffers) in a common format.

    if (needsFill) {
        BufferSpec& dstBufferSpec = dstBufferStack.back();

        int retval = DoFill(dstBufferSpec, srcLayer->getColor());
        if (retval) {
            ALOGE("Got error code %d from DoFill function", retval);
        }

        srcLayerSpec = dstBufferSpec;
        dstBufferStack.pop_back();
    }

    // TODO(jemoreira): We are converting to ARGB as the first step under the
    // assumption that scaling ARGB is faster than scaling I420 (the most common).
    // This should be confirmed with testing.
    if (needsConversion) {
        BufferSpec& dstBufferSpec = dstBufferStack.back();
        if (needsScaling || needsTranspose) {
            // If a rotation or a scaling operation are needed the dimensions at the
            // top of the buffer stack are wrong (wrong sizes for scaling, swapped
            // width and height for 90 and 270 rotations).
            // Make width and height match the crop sizes on the source
            uint32_t srcWidth = srcLayerSpec.cropWidth;
            uint32_t srcHeight = srcLayerSpec.cropHeight;
            uint32_t dst_stride_bytes = AlignToPower2(srcWidth * dstBufferBytesPerPixel, 4);
            uint32_t neededSize = dst_stride_bytes * srcHeight;
            dstBufferSpec.width = srcWidth;
            dstBufferSpec.height = srcHeight;
            // Adjust the stride accordingly
            dstBufferSpec.strideBytes = dst_stride_bytes;
            // Crop sizes also need to be adjusted
            dstBufferSpec.cropWidth = srcWidth;
            dstBufferSpec.cropHeight = srcHeight;
            // cropX and y are fine at 0, format is already set to match destination

            // In case of a scale, the source frame may be bigger than the default tmp
            // buffer size
            dstBufferSpec.buffer =
                compositionIntermediateStorage.getSpecialScratchBuffer(neededSize);
        }

        int retval = DoConversion(srcLayerSpec, dstBufferSpec, needsVFlip);
        if (retval) {
            ALOGE("Got error code %d from DoConversion function", retval);
        }
        needsVFlip = false;
        srcLayerSpec = dstBufferSpec;
        dstBufferStack.pop_back();
    }

    if (needsScaling) {
        BufferSpec& dstBufferSpec = dstBufferStack.back();
        if (needsTranspose) {
            // If a rotation is needed, the temporary buffer has the correct size but
            // needs to be transposed and have its stride updated accordingly. The
            // crop sizes also needs to be transposed, but not the x and y since they
            // are both zero in a temporary buffer (and it is a temporary buffer
            // because a rotation will be performed next).
            std::swap(dstBufferSpec.width, dstBufferSpec.height);
            std::swap(dstBufferSpec.cropWidth, dstBufferSpec.cropHeight);
            // TODO (jemoreira): Aligment (To align here may cause the needed size to
            // be bigger than the buffer, so care should be taken)
            dstBufferSpec.strideBytes = dstBufferSpec.width * dstBufferBytesPerPixel;
        }
        int retval = DoScaling(srcLayerSpec, dstBufferSpec, needsVFlip);
        needsVFlip = false;
        if (retval) {
            ALOGE("Got error code %d from DoScaling function", retval);
        }
        srcLayerSpec = dstBufferSpec;
        dstBufferStack.pop_back();
    }

    if (needsRotation) {
        int retval = DoRotation(srcLayerSpec, dstBufferStack.back(), rotation, needsVFlip);
        needsVFlip = false;
        if (retval) {
            ALOGE("Got error code %d from DoTransform function", retval);
        }
        srcLayerSpec = dstBufferStack.back();
        dstBufferStack.pop_back();
    }

    if (needsAttenuation) {
        int retval = DoAttenuation(srcLayerSpec, dstBufferStack.back(), needsVFlip);
        needsVFlip = false;
        if (retval) {
            ALOGE("Got error code %d from DoBlending function", retval);
        }
        srcLayerSpec = dstBufferStack.back();
        dstBufferStack.pop_back();
    }

    if (needsCopy) {
        int retval = DoCopy(srcLayerSpec, dstBufferStack.back(), needsVFlip);
        needsVFlip = false;
        if (retval) {
            ALOGE("Got error code %d from DoBlending function", retval);
        }
        srcLayerSpec = dstBufferStack.back();
        dstBufferStack.pop_back();
    }

    // Blending (if needed) should always be the last operation, so that it reads
    // and writes in the destination layer and not some temporary buffer.
    if (needsBlending) {
        int retval = DoBlending(srcLayerSpec, dstBufferStack.back(), needsVFlip);
        needsVFlip = false;
        if (retval) {
            ALOGE("Got error code %d from DoBlending function", retval);
        }
        // Don't need to assign destination to source in the last one
        dstBufferStack.pop_back();
    }

    return HWC3::Error::None;
}

namespace {

// Returns a color matrix that can be used with libyuv by converting values
// in -1 to 1 into -64 to 64 and transposing.
std::array<std::int8_t, 16> ToLibyuvColorMatrix(const std::array<float, 16>& in) {
    std::array<std::int8_t, 16> out;

    for (size_t r = 0; r < 4; r++) {
        for (size_t c = 0; c < 4; c++) {
            size_t indexIn = (4 * r) + c;
            size_t indexOut = (4 * c) + r;

            out[indexOut] = static_cast<std::int8_t>(
                std::max(-128, std::min(127, static_cast<int>(in[indexIn] * 64.0f + 0.5f))));
        }
    }

    return out;
}

}  // namespace

HWC3::Error GuestFrameComposer::applyColorTransformToRGBA(
    const std::array<float, 16>& transfromMatrix,  //
    std::uint8_t* buffer,                          //
    std::uint32_t bufferWidth,                     //
    std::uint32_t bufferHeight,                    //
    std::uint32_t bufferStrideBytes) {
    ATRACE_CALL();

    const auto transformMatrixLibyuv = ToLibyuvColorMatrix(transfromMatrix);
    libyuv::ARGBColorMatrix(buffer, static_cast<int>(bufferStrideBytes),  //
                            buffer, static_cast<int>(bufferStrideBytes),  //
                            transformMatrixLibyuv.data(),                 //
                            static_cast<int>(bufferWidth),                //
                            static_cast<int>(bufferHeight));

    return HWC3::Error::None;
}

}  // namespace aidl::android::hardware::graphics::composer3::impl