Skip to content
GitLab
Commit 5df65e4b authored by Xaver Hugl's avatar Xaver Hugl
Browse files

backends/drm: refactor egl/gbm render surface into its own class 

This allows sharing most code with overlay plane layers
parent c30339cc
Hide whitespace changes
Inline Side-by-side
src/backends/drm/CMakeLists.txt
View file @ 5df65e4b
......@@ -28,6 +28,7 @@ set(DRM_SOURCES
placeholderinputeventfilter.cpp
virtual_egl_gbm_layer.cpp
drm_lease_egl_gbm_layer.cpp
egl_gbm_layer_surface.cpp
)
add_library(KWinWaylandDrmBackend MODULE ${DRM_SOURCES})
......
src/backends/drm/egl_gbm_layer.cpp
View file @ 5df65e4b
......@@ -9,17 +9,13 @@
#include "egl_gbm_layer.h"
#include "drm_abstract_output.h"
#include "drm_backend.h"
#include "drm_buffer_gbm.h"
#include "drm_gpu.h"
#include "drm_output.h"
#include "drm_pipeline.h"
#include "dumb_swapchain.h"
#include "egl_dmabuf.h"
#include "egl_gbm_backend.h"
#include "gbm_surface.h"
#include "kwineglimagetexture.h"
#include "kwineglutils_p.h"
#include "logging.h"
#include "shadowbuffer.h"
#include "surfaceitem_wayland.h"
#include "KWaylandServer/linuxdmabufv1clientbuffer.h"
......@@ -36,7 +32,7 @@ namespace KWin
EglGbmLayer::EglGbmLayer(EglGbmBackend *eglBackend, DrmPipeline *pipeline)
: DrmPipelineLayer(pipeline)
, m_eglBackend(eglBackend)
, m_surface(pipeline->gpu(), eglBackend)
{
connect(eglBackend, &EglGbmBackend::aboutToBeDestroyed, this, &EglGbmLayer::destroyResources);
}
......@@ -48,17 +44,7 @@ EglGbmLayer::~EglGbmLayer()
void EglGbmLayer::destroyResources()
{
if (m_shadowBuffer || m_oldShadowBuffer) {
if (m_gbmSurface) {
m_gbmSurface->makeContextCurrent();
} else if (m_oldGbmSurface) {
m_oldGbmSurface->makeContextCurrent();
}
}
m_shadowBuffer.reset();
m_oldShadowBuffer.reset();
m_gbmSurface.reset();
m_oldGbmSurface.reset();
m_surface.destroyResources();
}
OutputLayerBeginFrameInfo EglGbmLayer::beginFrame()
......@@ -73,88 +59,20 @@ OutputLayerBeginFrameInfo EglGbmLayer::beginFrame()
}
m_scanoutCandidate.attemptedThisFrame = false;
// gbm surface
if (doesGbmSurfaceFit(m_gbmSurface.data())) {
m_oldGbmSurface.reset();
} else {
if (doesGbmSurfaceFit(m_oldGbmSurface.data())) {
m_gbmSurface = m_oldGbmSurface;
} else {
if (!createGbmSurface()) {
return {};
}
// dmabuf might work with the new surface
m_importMode = MultiGpuImportMode::Dmabuf;
}
}
if (!m_gbmSurface->makeContextCurrent()) {
return {};
}
auto repaintRegion = m_gbmSurface->repaintRegion();
// shadow buffer
if (doesShadowBufferFit(m_shadowBuffer.data())) {
m_oldShadowBuffer.reset();
} else {
if (doesShadowBufferFit(m_oldShadowBuffer.data())) {
m_shadowBuffer = m_oldShadowBuffer;
} else {
if (m_pipeline->pending.bufferTransformation != m_pipeline->pending.sourceTransformation) {
const auto format = m_eglBackend->gbmFormatForDrmFormat(m_gbmSurface->format()).value();
m_shadowBuffer = QSharedPointer<ShadowBuffer>::create(m_pipeline->sourceSize(), format);
if (!m_shadowBuffer->isComplete()) {
return {};
}
} else {
m_shadowBuffer.reset();
}
}
}
GLFramebuffer::pushFramebuffer(m_gbmSurface->fbo());
if (m_shadowBuffer) {
// the blit after rendering will completely overwrite the back buffer anyways
repaintRegion = QRegion();
GLFramebuffer::pushFramebuffer(m_shadowBuffer->fbo());
}
return OutputLayerBeginFrameInfo{
.renderTarget = RenderTarget(GLFramebuffer::currentFramebuffer()),
.repaint = repaintRegion,
};
return m_surface.startRendering(m_pipeline->bufferSize(), m_pipeline->pending.sourceTransformation, m_pipeline->pending.bufferTransformation, m_pipeline->formats());
}
void EglGbmLayer::aboutToStartPainting(const QRegion &damagedRegion)
{
if (m_gbmSurface && m_gbmSurface->bufferAge() > 0 && !damagedRegion.isEmpty() && m_eglBackend->supportsPartialUpdate()) {
QVector<EGLint> rects = m_pipeline->output()->regionToRects(damagedRegion);
const bool correct = eglSetDamageRegionKHR(m_eglBackend->eglDisplay(), m_gbmSurface->eglSurface(), rects.data(), rects.count() / 4);
if (!correct) {
qCWarning(KWIN_DRM) << "eglSetDamageRegionKHR failed:" << getEglErrorString();
}
}
m_surface.aboutToStartPainting(m_pipeline->output(), damagedRegion);
}
void EglGbmLayer::endFrame(const QRegion &renderedRegion, const QRegion &damagedRegion)
{
Q_UNUSED(renderedRegion)
if (m_shadowBuffer) {
GLFramebuffer::popFramebuffer();
// TODO handle m_pipeline->pending.bufferTransformation != Rotate0
m_shadowBuffer->render(m_pipeline->pending.sourceTransformation);
}
GLFramebuffer::popFramebuffer();
QSharedPointer<DrmBuffer> buffer;
if (m_pipeline->gpu() == m_eglBackend->gpu()) {
buffer = m_gbmSurface->swapBuffersForDrm(damagedRegion);
} else {
if (m_gbmSurface->swapBuffers(damagedRegion)) {
buffer = importBuffer();
}
}
if (buffer) {
m_currentBuffer = buffer;
m_currentDamage = damagedRegion;
const auto ret = m_surface.endRendering(m_pipeline->pending.bufferTransformation, damagedRegion);
if (ret.has_value()) {
std::tie(m_currentBuffer, m_currentDamage) = ret.value();
}
}
......@@ -165,19 +83,8 @@ QRegion EglGbmLayer::currentDamage() const
QSharedPointer<DrmBuffer> EglGbmLayer::testBuffer()
{
if (!m_currentBuffer || !doesGbmSurfaceFit(m_gbmSurface.data())) {
if (doesGbmSurfaceFit(m_oldGbmSurface.data())) {
// re-use old surface and buffer without rendering
m_gbmSurface = m_oldGbmSurface;
if (m_gbmSurface->currentBuffer()) {
m_currentBuffer = m_gbmSurface->currentDrmBuffer();
return m_currentBuffer;
}
}
if (!renderTestBuffer() && m_importMode == MultiGpuImportMode::DumbBufferXrgb8888) {
// try multi-gpu import again, this time with DRM_FORMAT_XRGB8888
renderTestBuffer();
}
if (!m_surface.doesSurfaceFit(m_pipeline->sourceSize(), m_pipeline->formats())) {
renderTestBuffer();
}
return m_currentBuffer;
}
......@@ -191,228 +98,13 @@ bool EglGbmLayer::renderTestBuffer()
return m_currentBuffer != oldBuffer;
}
bool EglGbmLayer::createGbmSurface(uint32_t format, const QVector<uint64_t> &modifiers)
{
static bool modifiersEnvSet = false;
static const bool modifiersEnv = qEnvironmentVariableIntValue("KWIN_DRM_USE_MODIFIERS", &modifiersEnvSet) != 0;
const bool allowModifiers = m_eglBackend->gpu()->addFB2ModifiersSupported() && m_pipeline->gpu()->addFB2ModifiersSupported()
&& ((m_eglBackend->gpu()->isNVidia() && !modifiersEnvSet) || (modifiersEnvSet && modifiersEnv));
const auto size = m_pipeline->bufferSize();
const auto config = m_eglBackend->config(format);
if (config == EGL_NO_CONFIG_KHR) {
return false;
}
QSharedPointer<GbmSurface> gbmSurface;
#if HAVE_GBM_BO_GET_FD_FOR_PLANE
if (!allowModifiers) {
#else
// modifiers have to be disabled with multi-gpu if gbm_bo_get_fd_for_plane is not available
if (!allowModifiers || m_pipeline->gpu() != m_eglBackend->gpu()) {
#endif
int flags = GBM_BO_USE_RENDERING;
if (m_pipeline->gpu() == m_eglBackend->gpu()) {
flags |= GBM_BO_USE_SCANOUT;
} else {
flags |= GBM_BO_USE_LINEAR;
}
gbmSurface = QSharedPointer<GbmSurface>::create(m_eglBackend->gpu(), size, format, flags, config);
} else {
gbmSurface = QSharedPointer<GbmSurface>::create(m_eglBackend->gpu(), size, format, modifiers, config);
if (!gbmSurface->isValid()) {
// the egl / gbm implementation may reject the modifier list from another gpu
// as a fallback use linear, to at least make CPU copy more efficient
const QVector<uint64_t> linear = {DRM_FORMAT_MOD_LINEAR};
gbmSurface = QSharedPointer<GbmSurface>::create(m_eglBackend->gpu(), size, format, linear, config);
}
}
if (gbmSurface->isValid()) {
m_oldGbmSurface = m_gbmSurface;
m_gbmSurface = gbmSurface;
return true;
} else {
return false;
}
}
bool EglGbmLayer::createGbmSurface()
{
const auto formats = m_pipeline->formats();
QVector<GbmFormat> sortedFormats;
for (auto it = formats.begin(); it != formats.end(); it++) {
const auto format = m_eglBackend->gbmFormatForDrmFormat(it.key());
if (format.has_value()) {
sortedFormats << format.value();
}
}
std::sort(sortedFormats.begin(), sortedFormats.end(), [this](const auto &lhs, const auto &rhs) {
if (lhs.drmFormat == rhs.drmFormat) {
// prefer having an alpha channel
return lhs.alphaSize > rhs.alphaSize;
} else if (m_eglBackend->prefer10bpc() && ((lhs.bpp == 30) != (rhs.bpp == 30))) {
// prefer 10bpc / 30bpp formats
return lhs.bpp == 30 ? true : false;
} else {
// fallback
return lhs.drmFormat < rhs.drmFormat;
}
});
for (const auto &format : qAsConst(sortedFormats)) {
if (m_importMode == MultiGpuImportMode::DumbBufferXrgb8888 && format.drmFormat != DRM_FORMAT_XRGB8888) {
continue;
}
if (formats.contains(format.drmFormat) && createGbmSurface(format.drmFormat, formats[format.drmFormat])) {
return true;
}
}
qCCritical(KWIN_DRM, "Failed to create a gbm surface!");
return false;
}
bool EglGbmLayer::doesGbmSurfaceFit(GbmSurface *surf) const
{
return surf && surf->size() == m_pipeline->bufferSize()
&& m_pipeline->formats().contains(surf->format())
&& (m_importMode != MultiGpuImportMode::DumbBufferXrgb8888 || surf->format() == DRM_FORMAT_XRGB8888)
&& (surf->modifiers().isEmpty() || m_pipeline->formats().value(surf->format()) == surf->modifiers());
}
bool EglGbmLayer::doesShadowBufferFit(ShadowBuffer *buffer) const
{
if (m_pipeline->pending.bufferTransformation != m_pipeline->pending.sourceTransformation) {
return buffer && buffer->texture()->size() == m_pipeline->sourceSize() && buffer->drmFormat() == m_gbmSurface->format();
} else {
return buffer == nullptr;
}
}
bool EglGbmLayer::doesSwapchainFit(DumbSwapchain *swapchain) const
{
return swapchain && swapchain->size() == m_pipeline->sourceSize() && swapchain->drmFormat() == m_gbmSurface->format();
}
QSharedPointer<GLTexture> EglGbmLayer::texture() const
{
if (m_scanoutBuffer) {
return m_scanoutBuffer->createTexture(m_eglBackend->eglDisplay());
} else if (m_shadowBuffer) {
return m_shadowBuffer->texture();
} else if (GbmBuffer *gbmBuffer = m_gbmSurface->currentBuffer().data()) {
return gbmBuffer->createTexture(m_eglBackend->eglDisplay());
} else {
qCWarning(KWIN_DRM) << "Failed to record frame: No gbm buffer!";
return nullptr;
}
}
QSharedPointer<DrmBuffer> EglGbmLayer::importBuffer()
{
if (m_importMode == MultiGpuImportMode::Dmabuf) {
if (const auto buffer = importDmabuf()) {
return buffer;
} else {
// don't bother trying again, it will most likely fail every time
m_importMode = MultiGpuImportMode::DumbBuffer;
}
}
if (const auto buffer = importWithCpu()) {
return buffer;
} else if (m_importMode == MultiGpuImportMode::DumbBuffer) {
m_importMode = MultiGpuImportMode::DumbBufferXrgb8888;
return nullptr;
}
if (m_importMode != MultiGpuImportMode::Failed) {
qCCritical(KWIN_DRM, "All multi gpu imports failed!");
m_importMode = MultiGpuImportMode::Failed;
}
return nullptr;
}
QSharedPointer<DrmBuffer> EglGbmLayer::importDmabuf()
{
const auto bo = m_gbmSurface->currentBuffer()->getBo();
gbm_bo *importedBuffer;
#if HAVE_GBM_BO_GET_FD_FOR_PLANE
if (gbm_bo_get_handle_for_plane(bo, 0).s32 != -1) {
gbm_import_fd_modifier_data data = {
.width = gbm_bo_get_width(bo),
.height = gbm_bo_get_height(bo),
.format = gbm_bo_get_format(bo),
.num_fds = static_cast<uint32_t>(gbm_bo_get_plane_count(bo)),
.fds = {},
.strides = {},
.offsets = {},
.modifier = gbm_bo_get_modifier(bo),
};
for (uint32_t i = 0; i < data.num_fds; i++) {
data.fds[i] = gbm_bo_get_fd_for_plane(bo, i);
if (data.fds[i] < 0) {
qCWarning(KWIN_DRM, "failed to export gbm_bo plane %d as dma-buf: %s", i, strerror(errno));
for (uint32_t f = 0; f < i; f++) {
close(data.fds[f]);
}
return nullptr;
}
data.strides[i] = gbm_bo_get_stride_for_plane(bo, i);
data.offsets[i] = gbm_bo_get_offset(bo, i);
}
importedBuffer = gbm_bo_import(m_pipeline->gpu()->gbmDevice(), GBM_BO_IMPORT_FD_MODIFIER, &data, GBM_BO_USE_SCANOUT);
} else {
#endif
gbm_import_fd_data data = {
.fd = gbm_bo_get_fd(bo),
.width = gbm_bo_get_width(bo),
.height = gbm_bo_get_height(bo),
.stride = gbm_bo_get_stride(bo),
.format = gbm_bo_get_format(bo),
};
if (data.fd < 0) {
qCWarning(KWIN_DRM, "failed to export gbm_bo as dma-buf: %s", strerror(errno));
return nullptr;
}
importedBuffer = gbm_bo_import(m_pipeline->gpu()->gbmDevice(), GBM_BO_IMPORT_FD_MODIFIER, &data, GBM_BO_USE_SCANOUT);
#if HAVE_GBM_BO_GET_FD_FOR_PLANE
}
#endif
if (!importedBuffer) {
qCWarning(KWIN_DRM, "failed to import gbm_bo for multi-gpu usage: %s", strerror(errno));
return nullptr;
}
const auto buffer = QSharedPointer<DrmGbmBuffer>::create(m_pipeline->gpu(), nullptr, importedBuffer);
return buffer->bufferId() ? buffer : nullptr;
}
QSharedPointer<DrmBuffer> EglGbmLayer::importWithCpu()
{
if (doesSwapchainFit(m_importSwapchain.data())) {
m_oldImportSwapchain.reset();
return m_scanoutBuffer->createTexture(m_surface.eglBackend()->eglDisplay());
} else {
if (doesSwapchainFit(m_oldImportSwapchain.data())) {
m_importSwapchain = m_oldImportSwapchain;
} else {
const auto swapchain = QSharedPointer<DumbSwapchain>::create(m_pipeline->gpu(), m_pipeline->sourceSize(), m_gbmSurface->format());
if (swapchain->isEmpty()) {
return nullptr;
}
m_importSwapchain = swapchain;
}
}
const auto bo = m_gbmSurface->currentBuffer();
if (!bo->map(GBM_BO_TRANSFER_READ)) {
qCWarning(KWIN_DRM, "mapping a gbm_bo failed: %s", strerror(errno));
return nullptr;
}
const auto importBuffer = m_importSwapchain->acquireBuffer();
if (bo->stride() != importBuffer->stride()) {
qCCritical(KWIN_DRM, "stride of gbm_bo (%d) and dumb buffer (%d) don't match!", bo->stride(), importBuffer->stride());
return nullptr;
}
if (!memcpy(importBuffer->data(), bo->mappedData(), importBuffer->size().height() * importBuffer->stride())) {
return nullptr;
return m_surface.texture();
}
return importBuffer;
}
bool EglGbmLayer::scanout(SurfaceItem *surfaceItem)
......@@ -468,7 +160,7 @@ void EglGbmLayer::sendDmabufFeedback(KWaylandServer::LinuxDmaBufV1ClientBuffer *
if (m_scanoutCandidate.surface->dmabufFeedbackV1()) {
QVector<KWaylandServer::LinuxDmaBufV1Feedback::Tranche> scanoutTranches;
const auto drmFormats = m_pipeline->formats();
const auto tranches = m_eglBackend->dmabuf()->tranches();
const auto tranches = m_surface.eglBackend()->dmabuf()->tranches();
for (const auto &tranche : tranches) {
KWaylandServer::LinuxDmaBufV1Feedback::Tranche scanoutTranche;
for (auto it = tranche.formatTable.constBegin(); it != tranche.formatTable.constEnd(); it++) {
......
src/backends/drm/egl_gbm_layer.h
View file @ 5df65e4b
......@@ -9,6 +9,8 @@
#pragma once
#include "drm_layer.h"
#include "egl_gbm_layer_surface.h"
#include <QMap>
#include <QPointer>
#include <QRegion>
......@@ -25,10 +27,8 @@ class LinuxDmaBufV1ClientBuffer;
namespace KWin
{
class GbmSurface;
class DumbSwapchain;
class ShadowBuffer;
class EglGbmBackend;
class EglGbmLayerSurface;
class DrmGbmBuffer;
class EglGbmLayer : public DrmPipelineLayer
......@@ -48,27 +48,10 @@ public:
QSharedPointer<GLTexture> texture() const override;
private:
bool createGbmSurface(uint32_t format, const QVector<uint64_t> &modifiers);
bool createGbmSurface();
bool doesGbmSurfaceFit(GbmSurface *surf) const;
bool doesShadowBufferFit(ShadowBuffer *buffer) const;
bool doesSwapchainFit(DumbSwapchain *swapchain) const;
void sendDmabufFeedback(KWaylandServer::LinuxDmaBufV1ClientBuffer *failedBuffer);
bool renderTestBuffer();
void destroyResources();
QSharedPointer<DrmBuffer> importBuffer();
QSharedPointer<DrmBuffer> importDmabuf();
QSharedPointer<DrmBuffer> importWithCpu();
enum class MultiGpuImportMode {
Dmabuf,
DumbBuffer,
DumbBufferXrgb8888,
Failed
};
MultiGpuImportMode m_importMode = MultiGpuImportMode::Dmabuf;
struct
{
QPointer<KWaylandServer::SurfaceInterface> surface;
......@@ -79,14 +62,8 @@ private:
QSharedPointer<DrmGbmBuffer> m_scanoutBuffer;
QSharedPointer<DrmBuffer> m_currentBuffer;
QRegion m_currentDamage;
QSharedPointer<GbmSurface> m_gbmSurface;
QSharedPointer<GbmSurface> m_oldGbmSurface;
QSharedPointer<ShadowBuffer> m_shadowBuffer;
QSharedPointer<ShadowBuffer> m_oldShadowBuffer;
QSharedPointer<DumbSwapchain> m_importSwapchain;
QSharedPointer<DumbSwapchain> m_oldImportSwapchain;
EglGbmBackend *const m_eglBackend;
EglGbmLayerSurface m_surface;
};
}
src/backends/drm/egl_gbm_layer_surface.cpp 0 → 100644
View file @ 5df65e4b
/*
KWin - the KDE window manager
This file is part of the KDE project.
SPDX-FileCopyrightText: 2022 Xaver Hugl <xaver.hugl@gmail.com>
SPDX-License-Identifier: GPL-2.0-or-later
*/
#include "egl_gbm_layer_surface.h"
#include "config-kwin.h"
#include "drm_gpu.h"
#include "drm_output.h"
#include "dumb_swapchain.h"
#include "egl_dmabuf.h"
#include "egl_gbm_backend.h"
#include "gbm_surface.h"
#include "kwineglutils_p.h"
#include "logging.h"
#include "shadowbuffer.h"
#include "surfaceitem_wayland.h"
#include <KWaylandServer/linuxdmabufv1clientbuffer.h>
#include <KWaylandServer/surface_interface.h>
#include <drm_fourcc.h>
#include <errno.h>
#include <gbm.h>
#include <unistd.h>
namespace KWin
{
EglGbmLayerSurface::EglGbmLayerSurface(DrmGpu *gpu, EglGbmBackend *eglBackend)
: m_gpu(gpu)
, m_eglBackend(eglBackend)
{
}
EglGbmLayerSurface::~EglGbmLayerSurface()
{
destroyResources();
}
void EglGbmLayerSurface::destroyResources()
{
if (m_gbmSurface && (m_shadowBuffer || m_oldShadowBuffer)) {
m_gbmSurface->makeContextCurrent();
}
m_shadowBuffer.reset();
m_oldShadowBuffer.reset();
m_gbmSurface.reset();
m_oldGbmSurface.reset();
}
OutputLayerBeginFrameInfo EglGbmLayerSurface::startRendering(const QSize &bufferSize, DrmPlane::Transformations renderTransform, DrmPlane::Transformations bufferTransform, const QMap<uint32_t, QVector<uint64_t>> &formats)
{
// gbm surface
if (doesGbmSurfaceFit(m_gbmSurface.data(), bufferSize, formats)) {
m_oldGbmSurface.reset();
} else {
if (doesGbmSurfaceFit(m_oldGbmSurface.data(), bufferSize, formats)) {
m_gbmSurface = m_oldGbmSurface;
} else {
if (!createGbmSurface(bufferSize, formats)) {
return {};
}
// dmabuf might work with the new surface
m_importMode = MultiGpuImportMode::Dmabuf;
m_importSwapchain.reset();
m_oldImportSwapchain.reset();
}
}
if (!m_gbmSurface->makeContextCurrent()) {
return {};
}
// shadow buffer
const QSize renderSize = (renderTransform & (DrmPlane::Transformation::Rotate90 | DrmPlane::Transformation::Rotate270)) ? m_gbmSurface->size().transposed() : m_gbmSurface->size();
if (doesShadowBufferFit(m_shadowBuffer.data(), renderSize, renderTransform, bufferTransform)) {
m_oldShadowBuffer.reset();
} else {
if (doesShadowBufferFit(m_oldShadowBuffer.data(), renderSize, renderTransform, bufferTransform)) {
m_shadowBuffer = m_oldShadowBuffer;
} else {
if (renderTransform != bufferTransform) {
const auto format = m_eglBackend->gbmFormatForDrmFormat(m_gbmSurface->format());
if (!format.has_value()) {
return {};
}
m_shadowBuffer = QSharedPointer<ShadowBuffer>::create(renderSize, format.value());
if (!m_shadowBuffer->isComplete()) {
return {};
}
} else {
m_shadowBuffer.reset();
}
}
}
GLFramebuffer::pushFramebuffer(m_gbmSurface->fbo());
if (m_shadowBuffer) {
GLFramebuffer::pushFramebuffer(m_shadowBuffer->fbo());
// the blit after rendering will completely overwrite the back buffer anyways
return OutputLayerBeginFrameInfo {
.renderTarget = RenderTarget(m_shadowBuffer->fbo()),
.repaint = {},
};
} else {
return OutputLayerBeginFrameInfo {
.renderTarget = RenderTarget(m_gbmSurface->fbo()),
.repaint = m_gbmSurface->repaintRegion(),
};
}
}
void EglGbmLayerSurface::aboutToStartPainting(DrmOutput *output, const QRegion &damagedRegion)
{
if (m_gbmSurface && m_gbmSurface->bufferAge() > 0 && !damagedRegion.isEmpty() && m_eglBackend->supportsPartialUpdate()) {
QVector<EGLint> rects = output->regionToRects(damagedRegion);
const bool correct = eglSetDamageRegionKHR(m_eglBackend->eglDisplay(), m_gbmSurface->eglSurface(), rects.data(), rects.count() / 4);
if (!correct) {
qCWarning(KWIN_DRM) << "eglSetDamageRegionKHR failed:" << getEglErrorString();
}
}
}