Android #124

.github/workflows/android-build.yml

@@ -40,11 +40,11 @@ jobs:
           sudo apt-get install -y ccache apksigner glslang-dev glslang-tools
       - name: Build
         run: ./.ci/scripts/android/build.sh
+      - name: Copy and sign artifacts
         env:
           ANDROID_KEYSTORE_B64: ${{ secrets.ANDROID_KEYSTORE_B64 }}
           ANDROID_KEY_ALIAS: ${{ secrets.ANDROID_KEY_ALIAS }}
           ANDROID_KEYSTORE_PASS: ${{ secrets.ANDROID_KEYSTORE_PASS }}
-      - name: Copy artifacts
         run: ./.ci/scripts/android/upload.sh
       - name: Upload
         uses: actions/upload-artifact@v3

src/CMakeLists.txt

@@ -21,7 +21,7 @@ if (MSVC)
     # Avoid windows.h from including some usually unused libs like winsocks.h, since this might cause some redefinition errors.
     add_definitions(-DWIN32_LEAN_AND_MEAN)
 
-    # Ensure that projects are built with Unicode support.
+    # Ensure that projects build with Unicode support.
     add_definitions(-DUNICODE -D_UNICODE)
 
     # /W4                 - Level 4 warnings
@@ -54,11 +54,11 @@ if (MSVC)
         /GT
 
         # Modules
-        /experimental:module- # Explicitly disable module support due to conflicts with precompiled headers.
+        /experimental:module- # Disable module support explicitly due to conflicts with precompiled headers
 
         # External headers diagnostics
         /external:anglebrackets # Treats all headers included by #include <header>, where the header file is enclosed in angle brackets (< >), as external headers
-        /external:W0            # Sets the default warning level to 0 for external headers, effectively disabling warnings for them.
+        /external:W0            # Sets the default warning level to 0 for external headers, effectively turning off warnings for external headers
 
         # Warnings
         /W4

src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/EmulationFragment.kt

@@ -10,6 +10,7 @@ import android.content.DialogInterface
 import android.content.SharedPreferences
 import android.content.pm.ActivityInfo
 import android.content.res.Configuration
+import android.graphics.Color
 import android.net.Uri
 import android.os.Bundle
 import android.os.Handler
@@ -154,6 +155,7 @@ class EmulationFragment : Fragment(), SurfaceHolder.Callback {
         }
 
         binding.surfaceEmulation.holder.addCallback(this)
+        binding.showFpsText.setTextColor(Color.YELLOW)
         binding.doneControlConfig.setOnClickListener { stopConfiguringControls() }
 
         binding.drawerLayout.addDrawerListener(object : DrawerListener {
@@ -412,12 +414,12 @@ class EmulationFragment : Fragment(), SurfaceHolder.Callback {
             val FRAMETIME = 2
             val SPEED = 3
             perfStatsUpdater = {
-                if (emulationViewModel.emulationStarted.value) {
+                if (emulationViewModel.emulationStarted.value == true) {
                     val perfStats = NativeLibrary.getPerfStats()
-                    if (_binding != null) {
+                    if (perfStats[FPS] > 0 && _binding != null) {
                         binding.showFpsText.text = String.format("FPS: %.1f", perfStats[FPS])
                     }
-                    perfStatsUpdateHandler.postDelayed(perfStatsUpdater!!, 800)
+                    perfStatsUpdateHandler.postDelayed(perfStatsUpdater!!, 100)
                 }
             }
             perfStatsUpdateHandler.post(perfStatsUpdater!!)
@@ -462,6 +464,7 @@ class EmulationFragment : Fragment(), SurfaceHolder.Callback {
                     if (it.orientation == FoldingFeature.Orientation.HORIZONTAL) {
                         // Restrict emulation and overlays to the top of the screen
                         binding.emulationContainer.layoutParams.height = it.bounds.top
+                        binding.overlayContainer.layoutParams.height = it.bounds.top
                         // Restrict input and menu drawer to the bottom of the screen
                         binding.inputContainer.layoutParams.height = it.bounds.bottom
                         binding.inGameMenu.layoutParams.height = it.bounds.bottom
@@ -475,6 +478,7 @@ class EmulationFragment : Fragment(), SurfaceHolder.Callback {
         if (!isFolding) {
             binding.emulationContainer.layoutParams.height = ViewGroup.LayoutParams.MATCH_PARENT
             binding.inputContainer.layoutParams.height = ViewGroup.LayoutParams.MATCH_PARENT
+            binding.overlayContainer.layoutParams.height = ViewGroup.LayoutParams.MATCH_PARENT
             binding.inGameMenu.layoutParams.height = ViewGroup.LayoutParams.MATCH_PARENT
             isInFoldableLayout = false
             updateOrientation()
@@ -482,6 +486,7 @@ class EmulationFragment : Fragment(), SurfaceHolder.Callback {
         }
         binding.emulationContainer.requestLayout()
         binding.inputContainer.requestLayout()
+        binding.overlayContainer.requestLayout()
         binding.inGameMenu.requestLayout()
     }
 
@@ -707,6 +712,24 @@ class EmulationFragment : Fragment(), SurfaceHolder.Callback {
             }
 
             v.setPadding(left, cutInsets.top, right, 0)
+
+            // Ensure FPS text doesn't get cut off by rounded display corners
+            val sidePadding = resources.getDimensionPixelSize(R.dimen.spacing_xtralarge)
+            if (cutInsets.left == 0) {
+                binding.showFpsText.setPadding(
+                    sidePadding,
+                    cutInsets.top,
+                    cutInsets.right,
+                    cutInsets.bottom
+                )
+            } else {
+                binding.showFpsText.setPadding(
+                    cutInsets.left,
+                    cutInsets.top,
+                    cutInsets.right,
+                    cutInsets.bottom
+                )
+            }
             windowInsets
         }
     }

src/android/app/src/main/jni/native.cpp

@@ -199,8 +199,8 @@ bool EmulationSession::IsPaused() const {
     return m_is_running && m_is_paused;
 }
 
-const Core::PerfStatsResults& EmulationSession::PerfStats() {
-    m_perf_stats = m_system.GetAndResetPerfStats();
+const Core::PerfStatsResults& EmulationSession::PerfStats() const {
+    std::scoped_lock m_perf_stats_lock(m_perf_stats_mutex);
     return m_perf_stats;
 }
 
@@ -383,6 +383,11 @@ void EmulationSession::RunEmulation() {
                 break;
             }
         }
+        {
+            // Refresh performance stats.
+            std::scoped_lock m_perf_stats_lock(m_perf_stats_mutex);
+            m_perf_stats = m_system.GetAndResetPerfStats();
+        }
     }
 }
 

src/android/app/src/main/jni/native.h

@@ -41,7 +41,7 @@ public:
     void RunEmulation();
     void ShutdownEmulation();
 
-    const Core::PerfStatsResults& PerfStats();
+    const Core::PerfStatsResults& PerfStats() const;
     void ConfigureFilesystemProvider(const std::string& filepath);
     void InitializeSystem(bool reload);
     Core::SystemResultStatus InitializeEmulation(const std::string& filepath);
@@ -80,5 +80,6 @@ private:
 
     // Synchronization
     std::condition_variable_any m_cv;
+    mutable std::mutex m_perf_stats_mutex;
     mutable std::mutex m_mutex;
 };

src/android/app/src/main/res/layout/fragment_emulation.xml

@@ -134,18 +134,16 @@
         <FrameLayout
             android:id="@+id/overlay_container"
             android:layout_width="match_parent"
-            android:layout_height="match_parent"
-            android:fitsSystemWindows="true">
+            android:layout_height="match_parent">
 
-            <com.google.android.material.textview.MaterialTextView
+            <TextView
                 android:id="@+id/show_fps_text"
-                style="@style/TextAppearance.Material3.BodyMedium"
                 android:layout_width="wrap_content"
                 android:layout_height="wrap_content"
                 android:layout_gravity="left"
                 android:clickable="false"
                 android:focusable="false"
-                android:paddingHorizontal="20dp"
+                android:shadowColor="@android:color/black"
                 android:textColor="@android:color/white"
                 android:textSize="12sp"
                 tools:ignore="RtlHardcoded" />

src/core/hle/service/nvdrv/devices/ioctl_serialization.h

@@ -1,159 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
-
-#pragma once
-
-#include <span>
-#include <vector>
-
-#include "common/concepts.h"
-#include "core/hle/service/nvdrv/devices/nvdevice.h"
-
-namespace Service::Nvidia::Devices {
-
-struct IoctlOneArgTraits {
-    template <typename T, typename R, typename A, typename... B>
-    static A GetFirstArgImpl(R (T::*)(A, B...));
-};
-
-struct IoctlTwoArgTraits {
-    template <typename T, typename R, typename A, typename B, typename... C>
-    static A GetFirstArgImpl(R (T::*)(A, B, C...));
-
-    template <typename T, typename R, typename A, typename B, typename... C>
-    static B GetSecondArgImpl(R (T::*)(A, B, C...));
-};
-
-struct Null {};
-
-// clang-format off
-
-template <typename FixedArg, typename VarArg, typename InlInVarArg, typename InlOutVarArg, typename F>
-NvResult WrapGeneric(F&& callable, std::span<const u8> input, std::span<const u8> inline_input, std::span<u8> output, std::span<u8> inline_output) {
-    constexpr bool HasFixedArg     = !std::is_same_v<FixedArg, Null>;
-    constexpr bool HasVarArg       = !std::is_same_v<VarArg, Null>;
-    constexpr bool HasInlInVarArg  = !std::is_same_v<InlInVarArg, Null>;
-    constexpr bool HasInlOutVarArg = !std::is_same_v<InlOutVarArg, Null>;
-
-    // Declare the fixed-size input value.
-    FixedArg fixed{};
-    size_t var_offset = 0;
-
-    if constexpr (HasFixedArg) {
-        // Read the fixed-size input value.
-        var_offset = std::min(sizeof(FixedArg), input.size());
-        if (var_offset > 0) {
-            std::memcpy(&fixed, input.data(), var_offset);
-        }
-    }
-
-    // Read the variable-sized inputs.
-    const size_t num_var_args = HasVarArg ? ((input.size() - var_offset) / sizeof(VarArg)) : 0;
-    std::vector<VarArg> var_args(num_var_args);
-    if constexpr (HasVarArg) {
-        if (num_var_args > 0) {
-            std::memcpy(var_args.data(), input.data() + var_offset, num_var_args * sizeof(VarArg));
-        }
-    }
-
-    const size_t num_inl_in_var_args = HasInlInVarArg ? (inline_input.size() / sizeof(InlInVarArg)) : 0;
-    std::vector<InlInVarArg> inl_in_var_args(num_inl_in_var_args);
-    if constexpr (HasInlInVarArg) {
-        if (num_inl_in_var_args > 0) {
-            std::memcpy(inl_in_var_args.data(), inline_input.data(), num_inl_in_var_args * sizeof(InlInVarArg));
-        }
-    }
-
-    // Construct inline output data.
-    const size_t num_inl_out_var_args = HasInlOutVarArg ? (inline_output.size() / sizeof(InlOutVarArg)) : 0;
-    std::vector<InlOutVarArg> inl_out_var_args(num_inl_out_var_args);
-
-    // Perform the call.
-    NvResult result = callable(fixed, var_args, inl_in_var_args, inl_out_var_args);
-
-    // Copy outputs.
-    if constexpr (HasFixedArg) {
-        if (output.size() > 0) {
-            std::memcpy(output.data(), &fixed, std::min(output.size(), sizeof(FixedArg)));
-        }
-    }
-
-    if constexpr (HasVarArg) {
-        if (num_var_args > 0 && output.size() > var_offset) {
-            const size_t max_var_size = output.size() - var_offset;
-            std::memcpy(output.data() + var_offset, var_args.data(), std::min(max_var_size, num_var_args * sizeof(VarArg)));
-        }
-    }
-
-    // Copy inline outputs.
-    if constexpr (HasInlOutVarArg) {
-        if (num_inl_out_var_args > 0) {
-            std::memcpy(inline_output.data(), inl_out_var_args.data(), num_inl_out_var_args * sizeof(InlOutVarArg));
-        }
-    }
-
-    // We're done.
-    return result;
-}
-
-template <typename Self, typename F, typename... Rest>
-NvResult WrapFixed(Self* self, F&& callable, std::span<const u8> input, std::span<u8> output, Rest&&... rest) {
-    using FixedArg = typename std::remove_reference_t<decltype(IoctlOneArgTraits::GetFirstArgImpl(callable))>;
-
-    const auto Callable = [&](auto& fixed, auto& var, auto& inl_in, auto& inl_out) -> NvResult {
-        return (self->*callable)(fixed, std::forward<Rest>(rest)...);
-    };
-
-    return WrapGeneric<FixedArg, Null, Null, Null>(std::move(Callable), input, {}, output, {});
-}
-
-template <typename Self, typename F, typename... Rest>
-NvResult WrapFixedInlOut(Self* self, F&& callable, std::span<const u8> input, std::span<u8> output, std::span<u8> inline_output, Rest&&... rest) {
-    using FixedArg     = typename std::remove_reference_t<decltype(IoctlTwoArgTraits::GetFirstArgImpl(callable))>;
-    using InlOutVarArg = typename std::remove_reference_t<decltype(IoctlTwoArgTraits::GetSecondArgImpl(callable))>::value_type;
-
-    const auto Callable = [&](auto& fixed, auto& var, auto& inl_in, auto& inl_out) -> NvResult {
-        return (self->*callable)(fixed, inl_out, std::forward<Rest>(rest)...);
-    };
-
-    return WrapGeneric<FixedArg, Null, Null, InlOutVarArg>(std::move(Callable), input, {}, output, inline_output);
-}
-
-template <typename Self, typename F, typename... Rest>
-NvResult WrapVariable(Self* self, F&& callable, std::span<const u8> input, std::span<u8> output, Rest&&... rest) {
-    using VarArg = typename std::remove_reference_t<decltype(IoctlOneArgTraits::GetFirstArgImpl(callable))>::value_type;
-
-    const auto Callable = [&](auto& fixed, auto& var, auto& inl_in, auto& inl_out) -> NvResult {
-        return (self->*callable)(var, std::forward<Rest>(rest)...);
-    };
-
-    return WrapGeneric<Null, VarArg, Null, Null>(std::move(Callable), input, {}, output, {});
-}
-
-template <typename Self, typename F, typename... Rest>
-NvResult WrapFixedVariable(Self* self, F&& callable, std::span<const u8> input, std::span<u8> output, Rest&&... rest) {
-    using FixedArg = typename std::remove_reference_t<decltype(IoctlTwoArgTraits::GetFirstArgImpl(callable))>;
-    using VarArg   = typename std::remove_reference_t<decltype(IoctlTwoArgTraits::GetSecondArgImpl(callable))>::value_type;
-
-    const auto Callable = [&](auto& fixed, auto& var, auto& inl_in, auto& inl_out) -> NvResult {
-        return (self->*callable)(fixed, var, std::forward<Rest>(rest)...);
-    };
-
-    return WrapGeneric<FixedArg, VarArg, Null, Null>(std::move(Callable), input, {}, output, {});
-}
-
-template <typename Self, typename F, typename... Rest>
-NvResult WrapFixedInlIn(Self* self, F&& callable, std::span<const u8> input, std::span<const u8> inline_input, std::span<u8> output, Rest&&... rest) {
-    using FixedArg    = typename std::remove_reference_t<decltype(IoctlTwoArgTraits::GetFirstArgImpl(callable))>;
-    using InlInVarArg = typename std::remove_reference_t<decltype(IoctlTwoArgTraits::GetSecondArgImpl(callable))>::value_type;
-
-    const auto Callable = [&](auto& fixed, auto& var, auto& inl_in, auto& inl_out) -> NvResult {
-        return (self->*callable)(fixed, inl_in, std::forward<Rest>(rest)...);
-    };
-
-    return WrapGeneric<FixedArg, Null, InlInVarArg, Null>(std::move(Callable), input, inline_input, output, {});
-}
-
-// clang-format on
-
-} // namespace Service::Nvidia::Devices

src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp

@@ -11,7 +11,6 @@
 #include "core/core.h"
 #include "core/hle/service/nvdrv/core/container.h"
 #include "core/hle/service/nvdrv/core/nvmap.h"
-#include "core/hle/service/nvdrv/devices/ioctl_serialization.h"
 #include "core/hle/service/nvdrv/devices/nvhost_as_gpu.h"
 #include "core/hle/service/nvdrv/devices/nvhost_gpu.h"
 #include "core/hle/service/nvdrv/nvdrv.h"
@@ -34,21 +33,21 @@ NvResult nvhost_as_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> i
     case 'A':
         switch (command.cmd) {
         case 0x1:
-            return WrapFixed(this, &nvhost_as_gpu::BindChannel, input, output);
+            return BindChannel(input, output);
         case 0x2:
-            return WrapFixed(this, &nvhost_as_gpu::AllocateSpace, input, output);
+            return AllocateSpace(input, output);
         case 0x3:
-            return WrapFixed(this, &nvhost_as_gpu::FreeSpace, input, output);
+            return FreeSpace(input, output);
         case 0x5:
-            return WrapFixed(this, &nvhost_as_gpu::UnmapBuffer, input, output);
+            return UnmapBuffer(input, output);
         case 0x6:
-            return WrapFixed(this, &nvhost_as_gpu::MapBufferEx, input, output);
+            return MapBufferEx(input, output);
         case 0x8:
-            return WrapFixed(this, &nvhost_as_gpu::GetVARegions1, input, output);
+            return GetVARegions(input, output);
         case 0x9:
-            return WrapFixed(this, &nvhost_as_gpu::AllocAsEx, input, output);
+            return AllocAsEx(input, output);
         case 0x14:
-            return WrapVariable(this, &nvhost_as_gpu::Remap, input, output);
+            return Remap(input, output);
         default:
             break;
         }
@@ -73,8 +72,7 @@ NvResult nvhost_as_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> i
     case 'A':
         switch (command.cmd) {
         case 0x8:
-            return WrapFixedInlOut(this, &nvhost_as_gpu::GetVARegions3, input, output,
-                                   inline_output);
+            return GetVARegions(input, output, inline_output);
         default:
             break;
         }
@@ -89,7 +87,10 @@ NvResult nvhost_as_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> i
 void nvhost_as_gpu::OnOpen(DeviceFD fd) {}
 void nvhost_as_gpu::OnClose(DeviceFD fd) {}
 
-NvResult nvhost_as_gpu::AllocAsEx(IoctlAllocAsEx& params) {
+NvResult nvhost_as_gpu::AllocAsEx(std::span<const u8> input, std::span<u8> output) {
+    IoctlAllocAsEx params{};
+    std::memcpy(&params, input.data(), input.size());
+
     LOG_DEBUG(Service_NVDRV, "called, big_page_size=0x{:X}", params.big_page_size);
 
     std::scoped_lock lock(mutex);
@@ -140,7 +141,10 @@ NvResult nvhost_as_gpu::AllocAsEx(IoctlAllocAsEx& params) {
     return NvResult::Success;
 }
 
-NvResult nvhost_as_gpu::AllocateSpace(IoctlAllocSpace& params) {
+NvResult nvhost_as_gpu::AllocateSpace(std::span<const u8> input, std::span<u8> output) {
+    IoctlAllocSpace params{};
+    std::memcpy(&params, input.data(), input.size());
+
     LOG_DEBUG(Service_NVDRV, "called, pages={:X}, page_size={:X}, flags={:X}", params.pages,
               params.page_size, params.flags);
 
@@ -190,6 +194,7 @@ NvResult nvhost_as_gpu::AllocateSpace(IoctlAllocSpace& params) {
         .big_pages = params.page_size != VM::YUZU_PAGESIZE,
     };
 
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
@@ -217,7 +222,10 @@ void nvhost_as_gpu::FreeMappingLocked(u64 offset) {
     mapping_map.erase(offset);
 }
 
-NvResult nvhost_as_gpu::FreeSpace(IoctlFreeSpace& params) {
+NvResult nvhost_as_gpu::FreeSpace(std::span<const u8> input, std::span<u8> output) {
+    IoctlFreeSpace params{};
+    std::memcpy(&params, input.data(), input.size());
+
     LOG_DEBUG(Service_NVDRV, "called, offset={:X}, pages={:X}, page_size={:X}", params.offset,
               params.pages, params.page_size);
 
@@ -256,11 +264,18 @@ NvResult nvhost_as_gpu::FreeSpace(IoctlFreeSpace& params) {
         return NvResult::BadValue;
     }
 
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_as_gpu::Remap(std::span<IoctlRemapEntry> entries) {
-    LOG_DEBUG(Service_NVDRV, "called, num_entries=0x{:X}", entries.size());
+NvResult nvhost_as_gpu::Remap(std::span<const u8> input, std::span<u8> output) {
+    const auto num_entries = input.size() / sizeof(IoctlRemapEntry);
+
+    LOG_DEBUG(Service_NVDRV, "called, num_entries=0x{:X}", num_entries);
+
+    std::scoped_lock lock(mutex);
+    entries.resize_destructive(num_entries);
+    std::memcpy(entries.data(), input.data(), input.size());
 
     if (!vm.initialised) {
         return NvResult::BadValue;
@@ -302,10 +317,14 @@ NvResult nvhost_as_gpu::Remap(std::span<IoctlRemapEntry> entries) {
         }
     }
 
+    std::memcpy(output.data(), entries.data(), output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
+NvResult nvhost_as_gpu::MapBufferEx(std::span<const u8> input, std::span<u8> output) {
+    IoctlMapBufferEx params{};
+    std::memcpy(&params, input.data(), input.size());
+
     LOG_DEBUG(Service_NVDRV,
               "called, flags={:X}, nvmap_handle={:X}, buffer_offset={}, mapping_size={}"
               ", offset={}",
@@ -402,10 +421,14 @@ NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
         mapping_map[params.offset] = mapping;
     }
 
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_as_gpu::UnmapBuffer(IoctlUnmapBuffer& params) {
+NvResult nvhost_as_gpu::UnmapBuffer(std::span<const u8> input, std::span<u8> output) {
+    IoctlUnmapBuffer params{};
+    std::memcpy(&params, input.data(), input.size());
+
     LOG_DEBUG(Service_NVDRV, "called, offset=0x{:X}", params.offset);
 
     std::scoped_lock lock(mutex);
@@ -441,7 +464,9 @@ NvResult nvhost_as_gpu::UnmapBuffer(IoctlUnmapBuffer& params) {
     return NvResult::Success;
 }
 
-NvResult nvhost_as_gpu::BindChannel(IoctlBindChannel& params) {
+NvResult nvhost_as_gpu::BindChannel(std::span<const u8> input, std::span<u8> output) {
+    IoctlBindChannel params{};
+    std::memcpy(&params, input.data(), input.size());
     LOG_DEBUG(Service_NVDRV, "called, fd={:X}", params.fd);
 
     auto gpu_channel_device = module.GetDevice<nvhost_gpu>(params.fd);
@@ -468,7 +493,10 @@ void nvhost_as_gpu::GetVARegionsImpl(IoctlGetVaRegions& params) {
     };
 }
 
-NvResult nvhost_as_gpu::GetVARegions1(IoctlGetVaRegions& params) {
+NvResult nvhost_as_gpu::GetVARegions(std::span<const u8> input, std::span<u8> output) {
+    IoctlGetVaRegions params{};
+    std::memcpy(&params, input.data(), input.size());
+
     LOG_DEBUG(Service_NVDRV, "called, buf_addr={:X}, buf_size={:X}", params.buf_addr,
               params.buf_size);
 
@@ -480,10 +508,15 @@ NvResult nvhost_as_gpu::GetVARegions1(IoctlGetVaRegions& params) {
 
     GetVARegionsImpl(params);
 
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_as_gpu::GetVARegions3(IoctlGetVaRegions& params, std::span<VaRegion> regions) {
+NvResult nvhost_as_gpu::GetVARegions(std::span<const u8> input, std::span<u8> output,
+                                     std::span<u8> inline_output) {
+    IoctlGetVaRegions params{};
+    std::memcpy(&params, input.data(), input.size());
+
     LOG_DEBUG(Service_NVDRV, "called, buf_addr={:X}, buf_size={:X}", params.buf_addr,
               params.buf_size);
 
@@ -495,10 +528,9 @@ NvResult nvhost_as_gpu::GetVARegions3(IoctlGetVaRegions& params, std::span<VaReg
 
     GetVARegionsImpl(params);
 
-    const size_t num_regions = std::min(params.regions.size(), regions.size());
-    for (size_t i = 0; i < num_regions; i++) {
-        regions[i] = params.regions[i];
-    }
+    std::memcpy(output.data(), &params, output.size());
+    std::memcpy(inline_output.data(), &params.regions[0], sizeof(VaRegion));
+    std::memcpy(inline_output.data() + sizeof(VaRegion), &params.regions[1], sizeof(VaRegion));
 
     return NvResult::Success;
 }

src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h

@@ -139,17 +139,18 @@ private:
     static_assert(sizeof(IoctlGetVaRegions) == 16 + sizeof(VaRegion) * 2,
                   "IoctlGetVaRegions is incorrect size");
 
-    NvResult AllocAsEx(IoctlAllocAsEx& params);
-    NvResult AllocateSpace(IoctlAllocSpace& params);
-    NvResult Remap(std::span<IoctlRemapEntry> params);
-    NvResult MapBufferEx(IoctlMapBufferEx& params);
-    NvResult UnmapBuffer(IoctlUnmapBuffer& params);
-    NvResult FreeSpace(IoctlFreeSpace& params);
-    NvResult BindChannel(IoctlBindChannel& params);
+    NvResult AllocAsEx(std::span<const u8> input, std::span<u8> output);
+    NvResult AllocateSpace(std::span<const u8> input, std::span<u8> output);
+    NvResult Remap(std::span<const u8> input, std::span<u8> output);
+    NvResult MapBufferEx(std::span<const u8> input, std::span<u8> output);
+    NvResult UnmapBuffer(std::span<const u8> input, std::span<u8> output);
+    NvResult FreeSpace(std::span<const u8> input, std::span<u8> output);
+    NvResult BindChannel(std::span<const u8> input, std::span<u8> output);
 
     void GetVARegionsImpl(IoctlGetVaRegions& params);
-    NvResult GetVARegions1(IoctlGetVaRegions& params);
-    NvResult GetVARegions3(IoctlGetVaRegions& params, std::span<VaRegion> regions);
+    NvResult GetVARegions(std::span<const u8> input, std::span<u8> output);
+    NvResult GetVARegions(std::span<const u8> input, std::span<u8> output,
+                          std::span<u8> inline_output);
 
     void FreeMappingLocked(u64 offset);
 
@@ -212,6 +213,7 @@ private:
         bool initialised{};
     } vm;
     std::shared_ptr<Tegra::MemoryManager> gmmu;
+    Common::ScratchBuffer<IoctlRemapEntry> entries;
 
     // s32 channel{};
     // u32 big_page_size{VM::DEFAULT_BIG_PAGE_SIZE};

src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp

@@ -14,7 +14,6 @@
 #include "core/hle/kernel/k_event.h"
 #include "core/hle/service/nvdrv/core/container.h"
 #include "core/hle/service/nvdrv/core/syncpoint_manager.h"
-#include "core/hle/service/nvdrv/devices/ioctl_serialization.h"
 #include "core/hle/service/nvdrv/devices/nvhost_ctrl.h"
 #include "video_core/gpu.h"
 #include "video_core/host1x/host1x.h"
@@ -41,19 +40,19 @@ NvResult nvhost_ctrl::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inp
     case 0x0:
         switch (command.cmd) {
         case 0x1b:
-            return WrapFixed(this, &nvhost_ctrl::NvOsGetConfigU32, input, output);
+            return NvOsGetConfigU32(input, output);
         case 0x1c:
-            return WrapFixed(this, &nvhost_ctrl::IocCtrlClearEventWait, input, output);
+            return IocCtrlClearEventWait(input, output);
         case 0x1d:
-            return WrapFixed(this, &nvhost_ctrl::IocCtrlEventWait, input, output, true);
+            return IocCtrlEventWait(input, output, true);
         case 0x1e:
-            return WrapFixed(this, &nvhost_ctrl::IocCtrlEventWait, input, output, false);
+            return IocCtrlEventWait(input, output, false);
         case 0x1f:
-            return WrapFixed(this, &nvhost_ctrl::IocCtrlEventRegister, input, output);
+            return IocCtrlEventRegister(input, output);
         case 0x20:
-            return WrapFixed(this, &nvhost_ctrl::IocCtrlEventUnregister, input, output);
+            return IocCtrlEventUnregister(input, output);
         case 0x21:
-            return WrapFixed(this, &nvhost_ctrl::IocCtrlEventUnregisterBatch, input, output);
+            return IocCtrlEventUnregisterBatch(input, output);
         }
         break;
     default:
@@ -80,19 +79,25 @@ void nvhost_ctrl::OnOpen(DeviceFD fd) {}
 
 void nvhost_ctrl::OnClose(DeviceFD fd) {}
 
-NvResult nvhost_ctrl::NvOsGetConfigU32(IocGetConfigParams& params) {
+NvResult nvhost_ctrl::NvOsGetConfigU32(std::span<const u8> input, std::span<u8> output) {
+    IocGetConfigParams params{};
+    std::memcpy(&params, input.data(), sizeof(params));
     LOG_TRACE(Service_NVDRV, "called, setting={}!{}", params.domain_str.data(),
               params.param_str.data());
     return NvResult::ConfigVarNotFound; // Returns error on production mode
 }
 
-NvResult nvhost_ctrl::IocCtrlEventWait(IocCtrlEventWaitParams& params, bool is_allocation) {
+NvResult nvhost_ctrl::IocCtrlEventWait(std::span<const u8> input, std::span<u8> output,
+                                       bool is_allocation) {
+    IocCtrlEventWaitParams params{};
+    std::memcpy(&params, input.data(), sizeof(params));
     LOG_DEBUG(Service_NVDRV, "syncpt_id={}, threshold={}, timeout={}, is_allocation={}",
               params.fence.id, params.fence.value, params.timeout, is_allocation);
 
     bool must_unmark_fail = !is_allocation;
     const u32 event_id = params.value.raw;
     SCOPE_EXIT({
+        std::memcpy(output.data(), &params, sizeof(params));
         if (must_unmark_fail) {
             events[event_id].fails = 0;
         }
@@ -226,7 +231,9 @@ NvResult nvhost_ctrl::FreeEvent(u32 slot) {
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl::IocCtrlEventRegister(IocCtrlEventRegisterParams& params) {
+NvResult nvhost_ctrl::IocCtrlEventRegister(std::span<const u8> input, std::span<u8> output) {
+    IocCtrlEventRegisterParams params{};
+    std::memcpy(&params, input.data(), sizeof(params));
     const u32 event_id = params.user_event_id;
     LOG_DEBUG(Service_NVDRV, " called, user_event_id: {:X}", event_id);
     if (event_id >= MaxNvEvents) {
@@ -245,7 +252,9 @@ NvResult nvhost_ctrl::IocCtrlEventRegister(IocCtrlEventRegisterParams& params) {
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl::IocCtrlEventUnregister(IocCtrlEventUnregisterParams& params) {
+NvResult nvhost_ctrl::IocCtrlEventUnregister(std::span<const u8> input, std::span<u8> output) {
+    IocCtrlEventUnregisterParams params{};
+    std::memcpy(&params, input.data(), sizeof(params));
     const u32 event_id = params.user_event_id & 0x00FF;
     LOG_DEBUG(Service_NVDRV, " called, user_event_id: {:X}", event_id);
 
@@ -253,7 +262,9 @@ NvResult nvhost_ctrl::IocCtrlEventUnregister(IocCtrlEventUnregisterParams& param
     return FreeEvent(event_id);
 }
 
-NvResult nvhost_ctrl::IocCtrlEventUnregisterBatch(IocCtrlEventUnregisterBatchParams& params) {
+NvResult nvhost_ctrl::IocCtrlEventUnregisterBatch(std::span<const u8> input, std::span<u8> output) {
+    IocCtrlEventUnregisterBatchParams params{};
+    std::memcpy(&params, input.data(), sizeof(params));
     u64 event_mask = params.user_events;
     LOG_DEBUG(Service_NVDRV, " called, event_mask: {:X}", event_mask);
 
@@ -269,7 +280,10 @@ NvResult nvhost_ctrl::IocCtrlEventUnregisterBatch(IocCtrlEventUnregisterBatchPar
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl::IocCtrlClearEventWait(IocCtrlEventClearParams& params) {
+NvResult nvhost_ctrl::IocCtrlClearEventWait(std::span<const u8> input, std::span<u8> output) {
+    IocCtrlEventClearParams params{};
+    std::memcpy(&params, input.data(), sizeof(params));
+
     u32 event_id = params.event_id.slot;
     LOG_DEBUG(Service_NVDRV, "called, event_id: {:X}", event_id);
 

src/core/hle/service/nvdrv/devices/nvhost_ctrl.h

@@ -186,12 +186,12 @@ private:
     static_assert(sizeof(IocCtrlEventUnregisterBatchParams) == 8,
                   "IocCtrlEventKill is incorrect size");
 
-    NvResult NvOsGetConfigU32(IocGetConfigParams& params);
-    NvResult IocCtrlEventRegister(IocCtrlEventRegisterParams& params);
-    NvResult IocCtrlEventUnregister(IocCtrlEventUnregisterParams& params);
-    NvResult IocCtrlEventUnregisterBatch(IocCtrlEventUnregisterBatchParams& params);
-    NvResult IocCtrlEventWait(IocCtrlEventWaitParams& params, bool is_allocation);
-    NvResult IocCtrlClearEventWait(IocCtrlEventClearParams& params);
+    NvResult NvOsGetConfigU32(std::span<const u8> input, std::span<u8> output);
+    NvResult IocCtrlEventWait(std::span<const u8> input, std::span<u8> output, bool is_allocation);
+    NvResult IocCtrlEventRegister(std::span<const u8> input, std::span<u8> output);
+    NvResult IocCtrlEventUnregister(std::span<const u8> input, std::span<u8> output);
+    NvResult IocCtrlEventUnregisterBatch(std::span<const u8> input, std::span<u8> output);
+    NvResult IocCtrlClearEventWait(std::span<const u8> input, std::span<u8> output);
 
     NvResult FreeEvent(u32 slot);
 

src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp

@@ -6,7 +6,6 @@
 #include "common/logging/log.h"
 #include "core/core.h"
 #include "core/core_timing.h"
-#include "core/hle/service/nvdrv/devices/ioctl_serialization.h"
 #include "core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h"
 #include "core/hle/service/nvdrv/nvdrv.h"
 
@@ -28,23 +27,23 @@ NvResult nvhost_ctrl_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8>
     case 'G':
         switch (command.cmd) {
         case 0x1:
-            return WrapFixed(this, &nvhost_ctrl_gpu::ZCullGetCtxSize, input, output);
+            return ZCullGetCtxSize(input, output);
         case 0x2:
-            return WrapFixed(this, &nvhost_ctrl_gpu::ZCullGetInfo, input, output);
+            return ZCullGetInfo(input, output);
         case 0x3:
-            return WrapFixed(this, &nvhost_ctrl_gpu::ZBCSetTable, input, output);
+            return ZBCSetTable(input, output);
         case 0x4:
-            return WrapFixed(this, &nvhost_ctrl_gpu::ZBCQueryTable, input, output);
+            return ZBCQueryTable(input, output);
         case 0x5:
-            return WrapFixed(this, &nvhost_ctrl_gpu::GetCharacteristics1, input, output);
+            return GetCharacteristics(input, output);
         case 0x6:
-            return WrapFixed(this, &nvhost_ctrl_gpu::GetTPCMasks1, input, output);
+            return GetTPCMasks(input, output);
         case 0x7:
-            return WrapFixed(this, &nvhost_ctrl_gpu::FlushL2, input, output);
+            return FlushL2(input, output);
         case 0x14:
-            return WrapFixed(this, &nvhost_ctrl_gpu::GetActiveSlotMask, input, output);
+            return GetActiveSlotMask(input, output);
         case 0x1c:
-            return WrapFixed(this, &nvhost_ctrl_gpu::GetGpuTime, input, output);
+            return GetGpuTime(input, output);
         default:
             break;
         }
@@ -66,11 +65,9 @@ NvResult nvhost_ctrl_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8>
     case 'G':
         switch (command.cmd) {
         case 0x5:
-            return WrapFixedInlOut(this, &nvhost_ctrl_gpu::GetCharacteristics3, input, output,
-                                   inline_output);
+            return GetCharacteristics(input, output, inline_output);
         case 0x6:
-            return WrapFixedInlOut(this, &nvhost_ctrl_gpu::GetTPCMasks3, input, output,
-                                   inline_output);
+            return GetTPCMasks(input, output, inline_output);
         default:
             break;
         }
@@ -85,8 +82,10 @@ NvResult nvhost_ctrl_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8>
 void nvhost_ctrl_gpu::OnOpen(DeviceFD fd) {}
 void nvhost_ctrl_gpu::OnClose(DeviceFD fd) {}
 
-NvResult nvhost_ctrl_gpu::GetCharacteristics1(IoctlCharacteristics& params) {
+NvResult nvhost_ctrl_gpu::GetCharacteristics(std::span<const u8> input, std::span<u8> output) {
     LOG_DEBUG(Service_NVDRV, "called");
+    IoctlCharacteristics params{};
+    std::memcpy(&params, input.data(), input.size());
     params.gc.arch = 0x120;
     params.gc.impl = 0xb;
     params.gc.rev = 0xa1;
@@ -124,13 +123,15 @@ NvResult nvhost_ctrl_gpu::GetCharacteristics1(IoctlCharacteristics& params) {
     params.gc.gr_compbit_store_base_hw = 0x0;
     params.gpu_characteristics_buf_size = 0xA0;
     params.gpu_characteristics_buf_addr = 0xdeadbeef; // Cannot be 0 (UNUSED)
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl_gpu::GetCharacteristics3(
-    IoctlCharacteristics& params, std::span<IoctlGpuCharacteristics> gpu_characteristics) {
+NvResult nvhost_ctrl_gpu::GetCharacteristics(std::span<const u8> input, std::span<u8> output,
+                                             std::span<u8> inline_output) {
     LOG_DEBUG(Service_NVDRV, "called");
-
+    IoctlCharacteristics params{};
+    std::memcpy(&params, input.data(), input.size());
     params.gc.arch = 0x120;
     params.gc.impl = 0xb;
     params.gc.rev = 0xa1;
@@ -168,47 +169,70 @@ NvResult nvhost_ctrl_gpu::GetCharacteristics3(
     params.gc.gr_compbit_store_base_hw = 0x0;
     params.gpu_characteristics_buf_size = 0xA0;
     params.gpu_characteristics_buf_addr = 0xdeadbeef; // Cannot be 0 (UNUSED)
-    if (!gpu_characteristics.empty()) {
-        gpu_characteristics.front() = params.gc;
-    }
+
+    std::memcpy(output.data(), &params, output.size());
+    std::memcpy(inline_output.data(), &params.gc, inline_output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl_gpu::GetTPCMasks1(IoctlGpuGetTpcMasksArgs& params) {
+NvResult nvhost_ctrl_gpu::GetTPCMasks(std::span<const u8> input, std::span<u8> output) {
+    IoctlGpuGetTpcMasksArgs params{};
+    std::memcpy(&params, input.data(), input.size());
     LOG_DEBUG(Service_NVDRV, "called, mask_buffer_size=0x{:X}", params.mask_buffer_size);
     if (params.mask_buffer_size != 0) {
         params.tcp_mask = 3;
     }
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl_gpu::GetTPCMasks3(IoctlGpuGetTpcMasksArgs& params, std::span<u32> tpc_mask) {
+NvResult nvhost_ctrl_gpu::GetTPCMasks(std::span<const u8> input, std::span<u8> output,
+                                      std::span<u8> inline_output) {
+    IoctlGpuGetTpcMasksArgs params{};
+    std::memcpy(&params, input.data(), input.size());
     LOG_DEBUG(Service_NVDRV, "called, mask_buffer_size=0x{:X}", params.mask_buffer_size);
     if (params.mask_buffer_size != 0) {
         params.tcp_mask = 3;
     }
-    if (!tpc_mask.empty()) {
-        tpc_mask.front() = params.tcp_mask;
-    }
+    std::memcpy(output.data(), &params, output.size());
+    std::memcpy(inline_output.data(), &params.tcp_mask, inline_output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl_gpu::GetActiveSlotMask(IoctlActiveSlotMask& params) {
+NvResult nvhost_ctrl_gpu::GetActiveSlotMask(std::span<const u8> input, std::span<u8> output) {
     LOG_DEBUG(Service_NVDRV, "called");
 
+    IoctlActiveSlotMask params{};
+    if (input.size() > 0) {
+        std::memcpy(&params, input.data(), input.size());
+    }
     params.slot = 0x07;
     params.mask = 0x01;
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl_gpu::ZCullGetCtxSize(IoctlZcullGetCtxSize& params) {
+NvResult nvhost_ctrl_gpu::ZCullGetCtxSize(std::span<const u8> input, std::span<u8> output) {
     LOG_DEBUG(Service_NVDRV, "called");
+
+    IoctlZcullGetCtxSize params{};
+    if (input.size() > 0) {
+        std::memcpy(&params, input.data(), input.size());
+    }
     params.size = 0x1;
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl_gpu::ZCullGetInfo(IoctlNvgpuGpuZcullGetInfoArgs& params) {
+NvResult nvhost_ctrl_gpu::ZCullGetInfo(std::span<const u8> input, std::span<u8> output) {
     LOG_DEBUG(Service_NVDRV, "called");
+
+    IoctlNvgpuGpuZcullGetInfoArgs params{};
+
+    if (input.size() > 0) {
+        std::memcpy(&params, input.data(), input.size());
+    }
+
     params.width_align_pixels = 0x20;
     params.height_align_pixels = 0x20;
     params.pixel_squares_by_aliquots = 0x400;
@@ -219,28 +243,53 @@ NvResult nvhost_ctrl_gpu::ZCullGetInfo(IoctlNvgpuGpuZcullGetInfoArgs& params) {
     params.subregion_width_align_pixels = 0x20;
     params.subregion_height_align_pixels = 0x40;
     params.subregion_count = 0x10;
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl_gpu::ZBCSetTable(IoctlZbcSetTable& params) {
+NvResult nvhost_ctrl_gpu::ZBCSetTable(std::span<const u8> input, std::span<u8> output) {
     LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+
+    IoctlZbcSetTable params{};
+    std::memcpy(&params, input.data(), input.size());
     // TODO(ogniK): What does this even actually do?
+
+    // Prevent null pointer being passed as arg 1
+    if (output.empty()) {
+        LOG_WARNING(Service_NVDRV, "Avoiding passing null pointer to memcpy");
+    } else {
+        std::memcpy(output.data(), &params, output.size());
+    }
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl_gpu::ZBCQueryTable(IoctlZbcQueryTable& params) {
+NvResult nvhost_ctrl_gpu::ZBCQueryTable(std::span<const u8> input, std::span<u8> output) {
     LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+
+    IoctlZbcQueryTable params{};
+    std::memcpy(&params, input.data(), input.size());
+    // TODO : To implement properly
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl_gpu::FlushL2(IoctlFlushL2& params) {
+NvResult nvhost_ctrl_gpu::FlushL2(std::span<const u8> input, std::span<u8> output) {
     LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+
+    IoctlFlushL2 params{};
+    std::memcpy(&params, input.data(), input.size());
+    // TODO : To implement properly
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl_gpu::GetGpuTime(IoctlGetGpuTime& params) {
+NvResult nvhost_ctrl_gpu::GetGpuTime(std::span<const u8> input, std::span<u8> output) {
     LOG_DEBUG(Service_NVDRV, "called");
+
+    IoctlGetGpuTime params{};
+    std::memcpy(&params, input.data(), input.size());
     params.gpu_time = static_cast<u64_le>(system.CoreTiming().GetGlobalTimeNs().count());
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 

src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h

@@ -151,20 +151,21 @@ private:
     };
     static_assert(sizeof(IoctlGetGpuTime) == 0x10, "IoctlGetGpuTime is incorrect size");
 
-    NvResult GetCharacteristics1(IoctlCharacteristics& params);
-    NvResult GetCharacteristics3(IoctlCharacteristics& params,
-                                 std::span<IoctlGpuCharacteristics> gpu_characteristics);
+    NvResult GetCharacteristics(std::span<const u8> input, std::span<u8> output);
+    NvResult GetCharacteristics(std::span<const u8> input, std::span<u8> output,
+                                std::span<u8> inline_output);
 
-    NvResult GetTPCMasks1(IoctlGpuGetTpcMasksArgs& params);
-    NvResult GetTPCMasks3(IoctlGpuGetTpcMasksArgs& params, std::span<u32> tpc_mask);
+    NvResult GetTPCMasks(std::span<const u8> input, std::span<u8> output);
+    NvResult GetTPCMasks(std::span<const u8> input, std::span<u8> output,
+                         std::span<u8> inline_output);
 
-    NvResult GetActiveSlotMask(IoctlActiveSlotMask& params);
-    NvResult ZCullGetCtxSize(IoctlZcullGetCtxSize& params);
-    NvResult ZCullGetInfo(IoctlNvgpuGpuZcullGetInfoArgs& params);
-    NvResult ZBCSetTable(IoctlZbcSetTable& params);
-    NvResult ZBCQueryTable(IoctlZbcQueryTable& params);
-    NvResult FlushL2(IoctlFlushL2& params);
-    NvResult GetGpuTime(IoctlGetGpuTime& params);
+    NvResult GetActiveSlotMask(std::span<const u8> input, std::span<u8> output);
+    NvResult ZCullGetCtxSize(std::span<const u8> input, std::span<u8> output);
+    NvResult ZCullGetInfo(std::span<const u8> input, std::span<u8> output);
+    NvResult ZBCSetTable(std::span<const u8> input, std::span<u8> output);
+    NvResult ZBCQueryTable(std::span<const u8> input, std::span<u8> output);
+    NvResult FlushL2(std::span<const u8> input, std::span<u8> output);
+    NvResult GetGpuTime(std::span<const u8> input, std::span<u8> output);
 
     EventInterface& events_interface;
 

src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp

@@ -8,7 +8,6 @@
 #include "core/hle/service/nvdrv/core/container.h"
 #include "core/hle/service/nvdrv/core/nvmap.h"
 #include "core/hle/service/nvdrv/core/syncpoint_manager.h"
-#include "core/hle/service/nvdrv/devices/ioctl_serialization.h"
 #include "core/hle/service/nvdrv/devices/nvhost_gpu.h"
 #include "core/hle/service/nvdrv/nvdrv.h"
 #include "core/memory.h"
@@ -53,7 +52,7 @@ NvResult nvhost_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     case 0x0:
         switch (command.cmd) {
         case 0x3:
-            return WrapFixed(this, &nvhost_gpu::GetWaitbase, input, output);
+            return GetWaitbase(input, output);
         default:
             break;
         }
@@ -61,25 +60,25 @@ NvResult nvhost_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     case 'H':
         switch (command.cmd) {
         case 0x1:
-            return WrapFixed(this, &nvhost_gpu::SetNVMAPfd, input, output);
+            return SetNVMAPfd(input, output);
         case 0x3:
-            return WrapFixed(this, &nvhost_gpu::ChannelSetTimeout, input, output);
+            return ChannelSetTimeout(input, output);
         case 0x8:
-            return WrapFixedVariable(this, &nvhost_gpu::SubmitGPFIFOBase1, input, output, false);
+            return SubmitGPFIFOBase(input, output, false);
         case 0x9:
-            return WrapFixed(this, &nvhost_gpu::AllocateObjectContext, input, output);
+            return AllocateObjectContext(input, output);
         case 0xb:
-            return WrapFixed(this, &nvhost_gpu::ZCullBind, input, output);
+            return ZCullBind(input, output);
         case 0xc:
-            return WrapFixed(this, &nvhost_gpu::SetErrorNotifier, input, output);
+            return SetErrorNotifier(input, output);
         case 0xd:
-            return WrapFixed(this, &nvhost_gpu::SetChannelPriority, input, output);
+            return SetChannelPriority(input, output);
         case 0x1a:
-            return WrapFixed(this, &nvhost_gpu::AllocGPFIFOEx2, input, output);
+            return AllocGPFIFOEx2(input, output);
         case 0x1b:
-            return WrapFixedVariable(this, &nvhost_gpu::SubmitGPFIFOBase1, input, output, true);
+            return SubmitGPFIFOBase(input, output, true);
         case 0x1d:
-            return WrapFixed(this, &nvhost_gpu::ChannelSetTimeslice, input, output);
+            return ChannelSetTimeslice(input, output);
         default:
             break;
         }
@@ -87,9 +86,9 @@ NvResult nvhost_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     case 'G':
         switch (command.cmd) {
         case 0x14:
-            return WrapFixed(this, &nvhost_gpu::SetClientData, input, output);
+            return SetClientData(input, output);
         case 0x15:
-            return WrapFixed(this, &nvhost_gpu::GetClientData, input, output);
+            return GetClientData(input, output);
         default:
             break;
         }
@@ -105,8 +104,7 @@ NvResult nvhost_gpu::Ioctl2(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     case 'H':
         switch (command.cmd) {
         case 0x1b:
-            return WrapFixedInlIn(this, &nvhost_gpu::SubmitGPFIFOBase2, input, inline_input,
-                                  output);
+            return SubmitGPFIFOBase(input, inline_input, output);
         }
         break;
     }
@@ -123,45 +121,63 @@ NvResult nvhost_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> inpu
 void nvhost_gpu::OnOpen(DeviceFD fd) {}
 void nvhost_gpu::OnClose(DeviceFD fd) {}
 
-NvResult nvhost_gpu::SetNVMAPfd(IoctlSetNvmapFD& params) {
+NvResult nvhost_gpu::SetNVMAPfd(std::span<const u8> input, std::span<u8> output) {
+    IoctlSetNvmapFD params{};
+    std::memcpy(&params, input.data(), input.size());
     LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
 
     nvmap_fd = params.nvmap_fd;
     return NvResult::Success;
 }
 
-NvResult nvhost_gpu::SetClientData(IoctlClientData& params) {
+NvResult nvhost_gpu::SetClientData(std::span<const u8> input, std::span<u8> output) {
     LOG_DEBUG(Service_NVDRV, "called");
+
+    IoctlClientData params{};
+    std::memcpy(&params, input.data(), input.size());
     user_data = params.data;
     return NvResult::Success;
 }
 
-NvResult nvhost_gpu::GetClientData(IoctlClientData& params) {
+NvResult nvhost_gpu::GetClientData(std::span<const u8> input, std::span<u8> output) {
     LOG_DEBUG(Service_NVDRV, "called");
+
+    IoctlClientData params{};
+    std::memcpy(&params, input.data(), input.size());
     params.data = user_data;
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_gpu::ZCullBind(IoctlZCullBind& params) {
-    zcull_params = params;
+NvResult nvhost_gpu::ZCullBind(std::span<const u8> input, std::span<u8> output) {
+    std::memcpy(&zcull_params, input.data(), input.size());
     LOG_DEBUG(Service_NVDRV, "called, gpu_va={:X}, mode={:X}", zcull_params.gpu_va,
               zcull_params.mode);
+
+    std::memcpy(output.data(), &zcull_params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_gpu::SetErrorNotifier(IoctlSetErrorNotifier& params) {
+NvResult nvhost_gpu::SetErrorNotifier(std::span<const u8> input, std::span<u8> output) {
+    IoctlSetErrorNotifier params{};
+    std::memcpy(&params, input.data(), input.size());
     LOG_WARNING(Service_NVDRV, "(STUBBED) called, offset={:X}, size={:X}, mem={:X}", params.offset,
                 params.size, params.mem);
+
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_gpu::SetChannelPriority(IoctlChannelSetPriority& params) {
-    channel_priority = params.priority;
+NvResult nvhost_gpu::SetChannelPriority(std::span<const u8> input, std::span<u8> output) {
+    std::memcpy(&channel_priority, input.data(), input.size());
     LOG_DEBUG(Service_NVDRV, "(STUBBED) called, priority={:X}", channel_priority);
+
     return NvResult::Success;
 }
 
-NvResult nvhost_gpu::AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params) {
+NvResult nvhost_gpu::AllocGPFIFOEx2(std::span<const u8> input, std::span<u8> output) {
+    IoctlAllocGpfifoEx2 params{};
+    std::memcpy(&params, input.data(), input.size());
     LOG_WARNING(Service_NVDRV,
                 "(STUBBED) called, num_entries={:X}, flags={:X}, unk0={:X}, "
                 "unk1={:X}, unk2={:X}, unk3={:X}",
@@ -177,14 +193,18 @@ NvResult nvhost_gpu::AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params) {
 
     params.fence_out = syncpoint_manager.GetSyncpointFence(channel_syncpoint);
 
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_gpu::AllocateObjectContext(IoctlAllocObjCtx& params) {
+NvResult nvhost_gpu::AllocateObjectContext(std::span<const u8> input, std::span<u8> output) {
+    IoctlAllocObjCtx params{};
+    std::memcpy(&params, input.data(), input.size());
     LOG_WARNING(Service_NVDRV, "(STUBBED) called, class_num={:X}, flags={:X}", params.class_num,
                 params.flags);
 
     params.obj_id = 0x0;
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
@@ -228,7 +248,8 @@ static boost::container::small_vector<Tegra::CommandHeader, 512> BuildIncrementW
     return result;
 }
 
-NvResult nvhost_gpu::SubmitGPFIFOImpl(IoctlSubmitGpfifo& params, Tegra::CommandList&& entries) {
+NvResult nvhost_gpu::SubmitGPFIFOImpl(IoctlSubmitGpfifo& params, std::span<u8> output,
+                                      Tegra::CommandList&& entries) {
     LOG_TRACE(Service_NVDRV, "called, gpfifo={:X}, num_entries={:X}, flags={:X}", params.address,
               params.num_entries, params.flags.raw);
 
@@ -269,55 +290,65 @@ NvResult nvhost_gpu::SubmitGPFIFOImpl(IoctlSubmitGpfifo& params, Tegra::CommandL
 
     flags.raw = 0;
 
+    std::memcpy(output.data(), &params, sizeof(IoctlSubmitGpfifo));
     return NvResult::Success;
 }
 
-NvResult nvhost_gpu::SubmitGPFIFOBase1(IoctlSubmitGpfifo& params,
-                                       std::span<Tegra::CommandListHeader> commands, bool kickoff) {
-    if (params.num_entries > commands.size()) {
+NvResult nvhost_gpu::SubmitGPFIFOBase(std::span<const u8> input, std::span<u8> output,
+                                      bool kickoff) {
+    if (input.size() < sizeof(IoctlSubmitGpfifo)) {
         UNIMPLEMENTED();
         return NvResult::InvalidSize;
     }
-
+    IoctlSubmitGpfifo params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
     Tegra::CommandList entries(params.num_entries);
+
     if (kickoff) {
         system.ApplicationMemory().ReadBlock(params.address, entries.command_lists.data(),
                                              params.num_entries * sizeof(Tegra::CommandListHeader));
     } else {
-        std::memcpy(entries.command_lists.data(), commands.data(),
+        std::memcpy(entries.command_lists.data(), &input[sizeof(IoctlSubmitGpfifo)],
                     params.num_entries * sizeof(Tegra::CommandListHeader));
     }
 
-    return SubmitGPFIFOImpl(params, std::move(entries));
+    return SubmitGPFIFOImpl(params, output, std::move(entries));
 }
 
-NvResult nvhost_gpu::SubmitGPFIFOBase2(IoctlSubmitGpfifo& params,
-                                       std::span<const Tegra::CommandListHeader> commands) {
-    if (params.num_entries > commands.size()) {
+NvResult nvhost_gpu::SubmitGPFIFOBase(std::span<const u8> input, std::span<const u8> input_inline,
+                                      std::span<u8> output) {
+    if (input.size() < sizeof(IoctlSubmitGpfifo)) {
         UNIMPLEMENTED();
         return NvResult::InvalidSize;
     }
-
+    IoctlSubmitGpfifo params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
     Tegra::CommandList entries(params.num_entries);
-    std::memcpy(entries.command_lists.data(), commands.data(),
-                params.num_entries * sizeof(Tegra::CommandListHeader));
-    return SubmitGPFIFOImpl(params, std::move(entries));
+    std::memcpy(entries.command_lists.data(), input_inline.data(), input_inline.size());
+    return SubmitGPFIFOImpl(params, output, std::move(entries));
 }
 
-NvResult nvhost_gpu::GetWaitbase(IoctlGetWaitbase& params) {
+NvResult nvhost_gpu::GetWaitbase(std::span<const u8> input, std::span<u8> output) {
+    IoctlGetWaitbase params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlGetWaitbase));
     LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
 
     params.value = 0; // Seems to be hard coded at 0
+    std::memcpy(output.data(), &params, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_gpu::ChannelSetTimeout(IoctlChannelSetTimeout& params) {
+NvResult nvhost_gpu::ChannelSetTimeout(std::span<const u8> input, std::span<u8> output) {
+    IoctlChannelSetTimeout params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlChannelSetTimeout));
     LOG_INFO(Service_NVDRV, "called, timeout=0x{:X}", params.timeout);
 
     return NvResult::Success;
 }
 
-NvResult nvhost_gpu::ChannelSetTimeslice(IoctlSetTimeslice& params) {
+NvResult nvhost_gpu::ChannelSetTimeslice(std::span<const u8> input, std::span<u8> output) {
+    IoctlSetTimeslice params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSetTimeslice));
     LOG_INFO(Service_NVDRV, "called, timeslice=0x{:X}", params.timeslice);
 
     channel_timeslice = params.timeslice;

src/core/hle/service/nvdrv/devices/nvhost_gpu.h

@@ -186,24 +186,23 @@ private:
     u32_le channel_priority{};
     u32_le channel_timeslice{};
 
-    NvResult SetNVMAPfd(IoctlSetNvmapFD& params);
-    NvResult SetClientData(IoctlClientData& params);
-    NvResult GetClientData(IoctlClientData& params);
-    NvResult ZCullBind(IoctlZCullBind& params);
-    NvResult SetErrorNotifier(IoctlSetErrorNotifier& params);
-    NvResult SetChannelPriority(IoctlChannelSetPriority& params);
-    NvResult AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params);
-    NvResult AllocateObjectContext(IoctlAllocObjCtx& params);
-
-    NvResult SubmitGPFIFOImpl(IoctlSubmitGpfifo& params, Tegra::CommandList&& entries);
-    NvResult SubmitGPFIFOBase1(IoctlSubmitGpfifo& params,
-                               std::span<Tegra::CommandListHeader> commands, bool kickoff = false);
-    NvResult SubmitGPFIFOBase2(IoctlSubmitGpfifo& params,
-                               std::span<const Tegra::CommandListHeader> commands);
-
-    NvResult GetWaitbase(IoctlGetWaitbase& params);
-    NvResult ChannelSetTimeout(IoctlChannelSetTimeout& params);
-    NvResult ChannelSetTimeslice(IoctlSetTimeslice& params);
+    NvResult SetNVMAPfd(std::span<const u8> input, std::span<u8> output);
+    NvResult SetClientData(std::span<const u8> input, std::span<u8> output);
+    NvResult GetClientData(std::span<const u8> input, std::span<u8> output);
+    NvResult ZCullBind(std::span<const u8> input, std::span<u8> output);
+    NvResult SetErrorNotifier(std::span<const u8> input, std::span<u8> output);
+    NvResult SetChannelPriority(std::span<const u8> input, std::span<u8> output);
+    NvResult AllocGPFIFOEx2(std::span<const u8> input, std::span<u8> output);
+    NvResult AllocateObjectContext(std::span<const u8> input, std::span<u8> output);
+    NvResult SubmitGPFIFOImpl(IoctlSubmitGpfifo& params, std::span<u8> output,
+                              Tegra::CommandList&& entries);
+    NvResult SubmitGPFIFOBase(std::span<const u8> input, std::span<u8> output,
+                              bool kickoff = false);
+    NvResult SubmitGPFIFOBase(std::span<const u8> input, std::span<const u8> input_inline,
+                              std::span<u8> output);
+    NvResult GetWaitbase(std::span<const u8> input, std::span<u8> output);
+    NvResult ChannelSetTimeout(std::span<const u8> input, std::span<u8> output);
+    NvResult ChannelSetTimeslice(std::span<const u8> input, std::span<u8> output);
 
     EventInterface& events_interface;
     NvCore::Container& core;

src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp

@@ -6,7 +6,6 @@
 #include "common/logging/log.h"
 #include "core/core.h"
 #include "core/hle/service/nvdrv/core/container.h"
-#include "core/hle/service/nvdrv/devices/ioctl_serialization.h"
 #include "core/hle/service/nvdrv/devices/nvhost_nvdec.h"
 #include "video_core/renderer_base.h"
 
@@ -26,18 +25,18 @@ NvResult nvhost_nvdec::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> in
             if (!host1x_file.fd_to_id.contains(fd)) {
                 host1x_file.fd_to_id[fd] = host1x_file.nvdec_next_id++;
             }
-            return WrapFixedVariable(this, &nvhost_nvdec::Submit, input, output, fd);
+            return Submit(fd, input, output);
         }
         case 0x2:
-            return WrapFixed(this, &nvhost_nvdec::GetSyncpoint, input, output);
+            return GetSyncpoint(input, output);
         case 0x3:
-            return WrapFixed(this, &nvhost_nvdec::GetWaitbase, input, output);
+            return GetWaitbase(input, output);
         case 0x7:
-            return WrapFixed(this, &nvhost_nvdec::SetSubmitTimeout, input, output);
+            return SetSubmitTimeout(input, output);
         case 0x9:
-            return WrapFixedVariable(this, &nvhost_nvdec::MapBuffer, input, output);
+            return MapBuffer(input, output);
         case 0xa:
-            return WrapFixedVariable(this, &nvhost_nvdec::UnmapBuffer, input, output);
+            return UnmapBuffer(input, output);
         default:
             break;
         }
@@ -45,7 +44,7 @@ NvResult nvhost_nvdec::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> in
     case 'H':
         switch (command.cmd) {
         case 0x1:
-            return WrapFixed(this, &nvhost_nvdec::SetNVMAPfd, input, output);
+            return SetNVMAPfd(input);
         default:
             break;
         }

src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp

@@ -29,9 +29,6 @@ std::size_t SliceVectors(std::span<const u8> input, std::vector<T>& dst, std::si
         return 0;
     }
     const size_t bytes_copied = count * sizeof(T);
-    if (input.size() < offset + bytes_copied) {
-        return 0;
-    }
     std::memcpy(dst.data(), input.data() + offset, bytes_copied);
     return bytes_copied;
 }
@@ -44,9 +41,6 @@ std::size_t WriteVectors(std::span<u8> dst, const std::vector<T>& src, std::size
         return 0;
     }
     const size_t bytes_copied = src.size() * sizeof(T);
-    if (dst.size() < offset + bytes_copied) {
-        return 0;
-    }
     std::memcpy(dst.data() + offset, src.data(), bytes_copied);
     return bytes_copied;
 }
@@ -69,14 +63,18 @@ nvhost_nvdec_common::~nvhost_nvdec_common() {
     core.Host1xDeviceFile().syncpts_accumulated.push_back(channel_syncpoint);
 }
 
-NvResult nvhost_nvdec_common::SetNVMAPfd(IoctlSetNvmapFD& params) {
+NvResult nvhost_nvdec_common::SetNVMAPfd(std::span<const u8> input) {
+    IoctlSetNvmapFD params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSetNvmapFD));
     LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
 
     nvmap_fd = params.nvmap_fd;
     return NvResult::Success;
 }
 
-NvResult nvhost_nvdec_common::Submit(IoctlSubmit& params, std::span<u8> data, DeviceFD fd) {
+NvResult nvhost_nvdec_common::Submit(DeviceFD fd, std::span<const u8> input, std::span<u8> output) {
+    IoctlSubmit params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSubmit));
     LOG_DEBUG(Service_NVDRV, "called NVDEC Submit, cmd_buffer_count={}", params.cmd_buffer_count);
 
     // Instantiate param buffers
@@ -87,12 +85,12 @@ NvResult nvhost_nvdec_common::Submit(IoctlSubmit& params, std::span<u8> data, De
     std::vector<u32> fence_thresholds(params.fence_count);
 
     // Slice input into their respective buffers
-    std::size_t offset = 0;
-    offset += SliceVectors(data, command_buffers, params.cmd_buffer_count, offset);
-    offset += SliceVectors(data, relocs, params.relocation_count, offset);
-    offset += SliceVectors(data, reloc_shifts, params.relocation_count, offset);
-    offset += SliceVectors(data, syncpt_increments, params.syncpoint_count, offset);
-    offset += SliceVectors(data, fence_thresholds, params.fence_count, offset);
+    std::size_t offset = sizeof(IoctlSubmit);
+    offset += SliceVectors(input, command_buffers, params.cmd_buffer_count, offset);
+    offset += SliceVectors(input, relocs, params.relocation_count, offset);
+    offset += SliceVectors(input, reloc_shifts, params.relocation_count, offset);
+    offset += SliceVectors(input, syncpt_increments, params.syncpoint_count, offset);
+    offset += SliceVectors(input, fence_thresholds, params.fence_count, offset);
 
     auto& gpu = system.GPU();
     if (gpu.UseNvdec()) {
@@ -110,51 +108,72 @@ NvResult nvhost_nvdec_common::Submit(IoctlSubmit& params, std::span<u8> data, De
                                              cmdlist.size() * sizeof(u32));
         gpu.PushCommandBuffer(core.Host1xDeviceFile().fd_to_id[fd], cmdlist);
     }
+    std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
     // Some games expect command_buffers to be written back
-    offset = 0;
-    offset += WriteVectors(data, command_buffers, offset);
-    offset += WriteVectors(data, relocs, offset);
-    offset += WriteVectors(data, reloc_shifts, offset);
-    offset += WriteVectors(data, syncpt_increments, offset);
-    offset += WriteVectors(data, fence_thresholds, offset);
+    offset = sizeof(IoctlSubmit);
+    offset += WriteVectors(output, command_buffers, offset);
+    offset += WriteVectors(output, relocs, offset);
+    offset += WriteVectors(output, reloc_shifts, offset);
+    offset += WriteVectors(output, syncpt_increments, offset);
+    offset += WriteVectors(output, fence_thresholds, offset);
 
     return NvResult::Success;
 }
 
-NvResult nvhost_nvdec_common::GetSyncpoint(IoctlGetSyncpoint& params) {
+NvResult nvhost_nvdec_common::GetSyncpoint(std::span<const u8> input, std::span<u8> output) {
+    IoctlGetSyncpoint params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlGetSyncpoint));
     LOG_DEBUG(Service_NVDRV, "called GetSyncpoint, id={}", params.param);
+
+    // const u32 id{NvCore::SyncpointManager::channel_syncpoints[static_cast<u32>(channel_type)]};
     params.value = channel_syncpoint;
+    std::memcpy(output.data(), &params, sizeof(IoctlGetSyncpoint));
+
     return NvResult::Success;
 }
 
-NvResult nvhost_nvdec_common::GetWaitbase(IoctlGetWaitbase& params) {
+NvResult nvhost_nvdec_common::GetWaitbase(std::span<const u8> input, std::span<u8> output) {
+    IoctlGetWaitbase params{};
     LOG_CRITICAL(Service_NVDRV, "called WAITBASE");
+    std::memcpy(&params, input.data(), sizeof(IoctlGetWaitbase));
     params.value = 0; // Seems to be hard coded at 0
+    std::memcpy(output.data(), &params, sizeof(IoctlGetWaitbase));
     return NvResult::Success;
 }
 
-NvResult nvhost_nvdec_common::MapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries) {
-    const size_t num_entries = std::min(params.num_entries, static_cast<u32>(entries.size()));
-    for (size_t i = 0; i < num_entries; i++) {
-        entries[i].map_address = nvmap.PinHandle(entries[i].map_handle);
+NvResult nvhost_nvdec_common::MapBuffer(std::span<const u8> input, std::span<u8> output) {
+    IoctlMapBuffer params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
+    std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries);
+
+    SliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
+
+    for (auto& cmd_buffer : cmd_buffer_handles) {
+        cmd_buffer.map_address = nvmap.PinHandle(cmd_buffer.map_handle);
+    }
+    std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
+    std::memcpy(output.data() + sizeof(IoctlMapBuffer), cmd_buffer_handles.data(),
+                cmd_buffer_handles.size() * sizeof(MapBufferEntry));
+
+    return NvResult::Success;
+}
+
+NvResult nvhost_nvdec_common::UnmapBuffer(std::span<const u8> input, std::span<u8> output) {
+    IoctlMapBuffer params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
+    std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries);
+
+    SliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
+    for (auto& cmd_buffer : cmd_buffer_handles) {
+        nvmap.UnpinHandle(cmd_buffer.map_handle);
     }
 
+    std::memset(output.data(), 0, output.size());
     return NvResult::Success;
 }
 
-NvResult nvhost_nvdec_common::UnmapBuffer(IoctlMapBuffer& params,
-                                          std::span<MapBufferEntry> entries) {
-    const size_t num_entries = std::min(params.num_entries, static_cast<u32>(entries.size()));
-    for (size_t i = 0; i < num_entries; i++) {
-        nvmap.UnpinHandle(entries[i].map_handle);
-        entries[i] = {};
-    }
-
-    params = {};
-    return NvResult::Success;
-}
-
-NvResult nvhost_nvdec_common::SetSubmitTimeout(u32 timeout) {
+NvResult nvhost_nvdec_common::SetSubmitTimeout(std::span<const u8> input, std::span<u8> output) {
+    std::memcpy(&submit_timeout, input.data(), input.size());
     LOG_WARNING(Service_NVDRV, "(STUBBED) called");
     return NvResult::Success;
 }

src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h

@@ -107,13 +107,13 @@ protected:
     static_assert(sizeof(IoctlMapBuffer) == 0x0C, "IoctlMapBuffer is incorrect size");
 
     /// Ioctl command implementations
-    NvResult SetNVMAPfd(IoctlSetNvmapFD&);
-    NvResult Submit(IoctlSubmit& params, std::span<u8> input, DeviceFD fd);
-    NvResult GetSyncpoint(IoctlGetSyncpoint& params);
-    NvResult GetWaitbase(IoctlGetWaitbase& params);
-    NvResult MapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries);
-    NvResult UnmapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries);
-    NvResult SetSubmitTimeout(u32 timeout);
+    NvResult SetNVMAPfd(std::span<const u8> input);
+    NvResult Submit(DeviceFD fd, std::span<const u8> input, std::span<u8> output);
+    NvResult GetSyncpoint(std::span<const u8> input, std::span<u8> output);
+    NvResult GetWaitbase(std::span<const u8> input, std::span<u8> output);
+    NvResult MapBuffer(std::span<const u8> input, std::span<u8> output);
+    NvResult UnmapBuffer(std::span<const u8> input, std::span<u8> output);
+    NvResult SetSubmitTimeout(std::span<const u8> input, std::span<u8> output);
 
     Kernel::KEvent* QueryEvent(u32 event_id) override;
 

src/core/hle/service/nvdrv/devices/nvhost_nvjpg.cpp

@@ -5,7 +5,6 @@
 
 #include "common/assert.h"
 #include "common/logging/log.h"
-#include "core/hle/service/nvdrv/devices/ioctl_serialization.h"
 #include "core/hle/service/nvdrv/devices/nvhost_nvjpg.h"
 
 namespace Service::Nvidia::Devices {
@@ -19,7 +18,7 @@ NvResult nvhost_nvjpg::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> in
     case 'H':
         switch (command.cmd) {
         case 0x1:
-            return WrapFixed(this, &nvhost_nvjpg::SetNVMAPfd, input, output);
+            return SetNVMAPfd(input, output);
         default:
             break;
         }
@@ -47,7 +46,9 @@ NvResult nvhost_nvjpg::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> in
 void nvhost_nvjpg::OnOpen(DeviceFD fd) {}
 void nvhost_nvjpg::OnClose(DeviceFD fd) {}
 
-NvResult nvhost_nvjpg::SetNVMAPfd(IoctlSetNvmapFD& params) {
+NvResult nvhost_nvjpg::SetNVMAPfd(std::span<const u8> input, std::span<u8> output) {
+    IoctlSetNvmapFD params{};
+    std::memcpy(&params, input.data(), input.size());
     LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
 
     nvmap_fd = params.nvmap_fd;

src/core/hle/service/nvdrv/devices/nvhost_nvjpg.h

@@ -33,7 +33,7 @@ private:
 
     s32_le nvmap_fd{};
 
-    NvResult SetNVMAPfd(IoctlSetNvmapFD& params);
+    NvResult SetNVMAPfd(std::span<const u8> input, std::span<u8> output);
 };
 
 } // namespace Service::Nvidia::Devices

src/core/hle/service/nvdrv/devices/nvhost_vic.cpp

@@ -5,7 +5,6 @@
 #include "common/logging/log.h"
 #include "core/core.h"
 #include "core/hle/service/nvdrv/core/container.h"
-#include "core/hle/service/nvdrv/devices/ioctl_serialization.h"
 #include "core/hle/service/nvdrv/devices/nvhost_vic.h"
 #include "video_core/renderer_base.h"
 
@@ -26,16 +25,16 @@ NvResult nvhost_vic::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpu
             if (!host1x_file.fd_to_id.contains(fd)) {
                 host1x_file.fd_to_id[fd] = host1x_file.vic_next_id++;
             }
-            return WrapFixedVariable(this, &nvhost_vic::Submit, input, output, fd);
+            return Submit(fd, input, output);
         }
         case 0x2:
-            return WrapFixed(this, &nvhost_vic::GetSyncpoint, input, output);
+            return GetSyncpoint(input, output);
         case 0x3:
-            return WrapFixed(this, &nvhost_vic::GetWaitbase, input, output);
+            return GetWaitbase(input, output);
         case 0x9:
-            return WrapFixedVariable(this, &nvhost_vic::MapBuffer, input, output);
+            return MapBuffer(input, output);
         case 0xa:
-            return WrapFixedVariable(this, &nvhost_vic::UnmapBuffer, input, output);
+            return UnmapBuffer(input, output);
         default:
             break;
         }
@@ -43,7 +42,7 @@ NvResult nvhost_vic::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     case 'H':
         switch (command.cmd) {
         case 0x1:
-            return WrapFixed(this, &nvhost_vic::SetNVMAPfd, input, output);
+            return SetNVMAPfd(input);
         default:
             break;
         }

src/core/hle/service/nvdrv/devices/nvmap.cpp

@@ -13,7 +13,6 @@
 #include "core/hle/kernel/k_process.h"
 #include "core/hle/service/nvdrv/core/container.h"
 #include "core/hle/service/nvdrv/core/nvmap.h"
-#include "core/hle/service/nvdrv/devices/ioctl_serialization.h"
 #include "core/hle/service/nvdrv/devices/nvmap.h"
 #include "core/memory.h"
 
@@ -32,17 +31,17 @@ NvResult nvmap::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> input,
     case 0x1:
         switch (command.cmd) {
         case 0x1:
-            return WrapFixed(this, &nvmap::IocCreate, input, output);
+            return IocCreate(input, output);
         case 0x3:
-            return WrapFixed(this, &nvmap::IocFromId, input, output);
+            return IocFromId(input, output);
         case 0x4:
-            return WrapFixed(this, &nvmap::IocAlloc, input, output);
+            return IocAlloc(input, output);
         case 0x5:
-            return WrapFixed(this, &nvmap::IocFree, input, output);
+            return IocFree(input, output);
         case 0x9:
-            return WrapFixed(this, &nvmap::IocParam, input, output);
+            return IocParam(input, output);
         case 0xe:
-            return WrapFixed(this, &nvmap::IocGetId, input, output);
+            return IocGetId(input, output);
         default:
             break;
         }
@@ -70,7 +69,9 @@ NvResult nvmap::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, st
 void nvmap::OnOpen(DeviceFD fd) {}
 void nvmap::OnClose(DeviceFD fd) {}
 
-NvResult nvmap::IocCreate(IocCreateParams& params) {
+NvResult nvmap::IocCreate(std::span<const u8> input, std::span<u8> output) {
+    IocCreateParams params;
+    std::memcpy(&params, input.data(), sizeof(params));
     LOG_DEBUG(Service_NVDRV, "called, size=0x{:08X}", params.size);
 
     std::shared_ptr<NvCore::NvMap::Handle> handle_description{};
@@ -84,10 +85,13 @@ NvResult nvmap::IocCreate(IocCreateParams& params) {
     params.handle = handle_description->id;
     LOG_DEBUG(Service_NVDRV, "handle: {}, size: 0x{:X}", handle_description->id, params.size);
 
+    std::memcpy(output.data(), &params, sizeof(params));
     return NvResult::Success;
 }
 
-NvResult nvmap::IocAlloc(IocAllocParams& params) {
+NvResult nvmap::IocAlloc(std::span<const u8> input, std::span<u8> output) {
+    IocAllocParams params;
+    std::memcpy(&params, input.data(), sizeof(params));
     LOG_DEBUG(Service_NVDRV, "called, addr={:X}", params.address);
 
     if (!params.handle) {
@@ -129,10 +133,14 @@ NvResult nvmap::IocAlloc(IocAllocParams& params) {
                                              handle_description->size,
                                              Kernel::KMemoryPermission::None, true, false)
                .IsSuccess());
+    std::memcpy(output.data(), &params, sizeof(params));
     return result;
 }
 
-NvResult nvmap::IocGetId(IocGetIdParams& params) {
+NvResult nvmap::IocGetId(std::span<const u8> input, std::span<u8> output) {
+    IocGetIdParams params;
+    std::memcpy(&params, input.data(), sizeof(params));
+
     LOG_DEBUG(Service_NVDRV, "called");
 
     // See the comment in FromId for extra info on this function
@@ -149,10 +157,14 @@ NvResult nvmap::IocGetId(IocGetIdParams& params) {
     }
 
     params.id = handle_description->id;
+    std::memcpy(output.data(), &params, sizeof(params));
     return NvResult::Success;
 }
 
-NvResult nvmap::IocFromId(IocFromIdParams& params) {
+NvResult nvmap::IocFromId(std::span<const u8> input, std::span<u8> output) {
+    IocFromIdParams params;
+    std::memcpy(&params, input.data(), sizeof(params));
+
     LOG_DEBUG(Service_NVDRV, "called, id:{}", params.id);
 
     // Handles and IDs are always the same value in nvmap however IDs can be used globally given the
@@ -176,12 +188,16 @@ NvResult nvmap::IocFromId(IocFromIdParams& params) {
         return result;
     }
     params.handle = handle_description->id;
+    std::memcpy(output.data(), &params, sizeof(params));
     return NvResult::Success;
 }
 
-NvResult nvmap::IocParam(IocParamParams& params) {
+NvResult nvmap::IocParam(std::span<const u8> input, std::span<u8> output) {
     enum class ParamTypes { Size = 1, Alignment = 2, Base = 3, Heap = 4, Kind = 5, Compr = 6 };
 
+    IocParamParams params;
+    std::memcpy(&params, input.data(), sizeof(params));
+
     LOG_DEBUG(Service_NVDRV, "called type={}", params.param);
 
     if (!params.handle) {
@@ -221,10 +237,14 @@ NvResult nvmap::IocParam(IocParamParams& params) {
         return NvResult::BadValue;
     }
 
+    std::memcpy(output.data(), &params, sizeof(params));
     return NvResult::Success;
 }
 
-NvResult nvmap::IocFree(IocFreeParams& params) {
+NvResult nvmap::IocFree(std::span<const u8> input, std::span<u8> output) {
+    IocFreeParams params;
+    std::memcpy(&params, input.data(), sizeof(params));
+
     LOG_DEBUG(Service_NVDRV, "called");
 
     if (!params.handle) {
@@ -247,6 +267,7 @@ NvResult nvmap::IocFree(IocFreeParams& params) {
         // This is possible when there's internal dups or other duplicates.
     }
 
+    std::memcpy(output.data(), &params, sizeof(params));
     return NvResult::Success;
 }
 

src/core/hle/service/nvdrv/devices/nvmap.h

@@ -99,12 +99,12 @@ public:
     };
     static_assert(sizeof(IocGetIdParams) == 8, "IocGetIdParams has wrong size");
 
-    NvResult IocCreate(IocCreateParams& params);
-    NvResult IocAlloc(IocAllocParams& params);
-    NvResult IocGetId(IocGetIdParams& params);
-    NvResult IocFromId(IocFromIdParams& params);
-    NvResult IocParam(IocParamParams& params);
-    NvResult IocFree(IocFreeParams& params);
+    NvResult IocCreate(std::span<const u8> input, std::span<u8> output);
+    NvResult IocAlloc(std::span<const u8> input, std::span<u8> output);
+    NvResult IocGetId(std::span<const u8> input, std::span<u8> output);
+    NvResult IocFromId(std::span<const u8> input, std::span<u8> output);
+    NvResult IocParam(std::span<const u8> input, std::span<u8> output);
+    NvResult IocFree(std::span<const u8> input, std::span<u8> output);
 
 private:
     /// Id to use for the next handle that is created.

src/core/hle/service/nvnflinger/buffer_transform_flags.h

@@ -3,7 +3,6 @@
 
 #pragma once
 
-#include "common/common_funcs.h"
 #include "common/common_types.h"
 
 namespace Service::android {
@@ -22,6 +21,5 @@ enum class BufferTransformFlags : u32 {
     /// Rotate source image 270 degrees clockwise
     Rotate270 = 0x07,
 };
-DECLARE_ENUM_FLAG_OPERATORS(BufferTransformFlags);
 
 } // namespace Service::android

src/core/hle/service/nvnflinger/fb_share_buffer_manager.cpp

@@ -71,17 +71,24 @@ Result AllocateIoForProcessAddressSpace(Common::ProcessAddress* out_map_address,
     R_SUCCEED();
 }
 
+template <typename T>
+std::span<u8> SerializeIoc(T& params) {
+    return std::span(reinterpret_cast<u8*>(std::addressof(params)), sizeof(T));
+}
+
 Result CreateNvMapHandle(u32* out_nv_map_handle, Nvidia::Devices::nvmap& nvmap, u32 size) {
     // Create a handle.
-    Nvidia::Devices::nvmap::IocCreateParams create_params{
+    Nvidia::Devices::nvmap::IocCreateParams create_in_params{
         .size = size,
         .handle = 0,
     };
-    R_UNLESS(nvmap.IocCreate(create_params) == Nvidia::NvResult::Success,
+    Nvidia::Devices::nvmap::IocCreateParams create_out_params{};
+    R_UNLESS(nvmap.IocCreate(SerializeIoc(create_in_params), SerializeIoc(create_out_params)) ==
+                 Nvidia::NvResult::Success,
              VI::ResultOperationFailed);
 
     // Assign the output handle.
-    *out_nv_map_handle = create_params.handle;
+    *out_nv_map_handle = create_out_params.handle;
 
     // We succeeded.
     R_SUCCEED();
@@ -89,10 +96,13 @@ Result CreateNvMapHandle(u32* out_nv_map_handle, Nvidia::Devices::nvmap& nvmap,
 
 Result FreeNvMapHandle(Nvidia::Devices::nvmap& nvmap, u32 handle) {
     // Free the handle.
-    Nvidia::Devices::nvmap::IocFreeParams free_params{
+    Nvidia::Devices::nvmap::IocFreeParams free_in_params{
         .handle = handle,
     };
-    R_UNLESS(nvmap.IocFree(free_params) == Nvidia::NvResult::Success, VI::ResultOperationFailed);
+    Nvidia::Devices::nvmap::IocFreeParams free_out_params{};
+    R_UNLESS(nvmap.IocFree(SerializeIoc(free_in_params), SerializeIoc(free_out_params)) ==
+                 Nvidia::NvResult::Success,
+             VI::ResultOperationFailed);
 
     // We succeeded.
     R_SUCCEED();
@@ -101,7 +111,7 @@ Result FreeNvMapHandle(Nvidia::Devices::nvmap& nvmap, u32 handle) {
 Result AllocNvMapHandle(Nvidia::Devices::nvmap& nvmap, u32 handle, Common::ProcessAddress buffer,
                         u32 size) {
     // Assign the allocated memory to the handle.
-    Nvidia::Devices::nvmap::IocAllocParams alloc_params{
+    Nvidia::Devices::nvmap::IocAllocParams alloc_in_params{
         .handle = handle,
         .heap_mask = 0,
         .flags = {},
@@ -109,7 +119,10 @@ Result AllocNvMapHandle(Nvidia::Devices::nvmap& nvmap, u32 handle, Common::Proce
         .kind = 0,
         .address = GetInteger(buffer),
     };
-    R_UNLESS(nvmap.IocAlloc(alloc_params) == Nvidia::NvResult::Success, VI::ResultOperationFailed);
+    Nvidia::Devices::nvmap::IocAllocParams alloc_out_params{};
+    R_UNLESS(nvmap.IocAlloc(SerializeIoc(alloc_in_params), SerializeIoc(alloc_out_params)) ==
+                 Nvidia::NvResult::Success,
+             VI::ResultOperationFailed);
 
     // We succeeded.
     R_SUCCEED();

src/video_core/renderer_vulkan/vk_blit_screen.cpp

@@ -137,56 +137,6 @@ BlitScreen::BlitScreen(Core::Memory::Memory& cpu_memory_, Core::Frontend::EmuWin
 
 BlitScreen::~BlitScreen() = default;
 
-static Common::Rectangle<f32> NormalizeCrop(const Tegra::FramebufferConfig& framebuffer,
-                                            const ScreenInfo& screen_info) {
-    f32 left, top, right, bottom;
-
-    if (!framebuffer.crop_rect.IsEmpty()) {
-        // If crop rectangle is not empty, apply properties from rectangle.
-        left = static_cast<f32>(framebuffer.crop_rect.left);
-        top = static_cast<f32>(framebuffer.crop_rect.top);
-        right = static_cast<f32>(framebuffer.crop_rect.right);
-        bottom = static_cast<f32>(framebuffer.crop_rect.bottom);
-    } else {
-        // Otherwise, fall back to framebuffer dimensions.
-        left = 0;
-        top = 0;
-        right = static_cast<f32>(framebuffer.width);
-        bottom = static_cast<f32>(framebuffer.height);
-    }
-
-    // Apply transformation flags.
-    auto framebuffer_transform_flags = framebuffer.transform_flags;
-
-    if (True(framebuffer_transform_flags & Service::android::BufferTransformFlags::FlipH)) {
-        // Switch left and right.
-        std::swap(left, right);
-    }
-    if (True(framebuffer_transform_flags & Service::android::BufferTransformFlags::FlipV)) {
-        // Switch top and bottom.
-        std::swap(top, bottom);
-    }
-
-    framebuffer_transform_flags &= ~Service::android::BufferTransformFlags::FlipH;
-    framebuffer_transform_flags &= ~Service::android::BufferTransformFlags::FlipV;
-    if (True(framebuffer_transform_flags)) {
-        UNIMPLEMENTED_MSG("Unsupported framebuffer_transform_flags={}",
-                          static_cast<u32>(framebuffer_transform_flags));
-    }
-
-    // Get the screen properties.
-    const f32 screen_width = static_cast<f32>(screen_info.width);
-    const f32 screen_height = static_cast<f32>(screen_info.height);
-
-    // Normalize coordinate space.
-    left /= screen_width;
-    top /= screen_height;
-    right /= screen_width;
-    bottom /= screen_height;
-
-    return Common::Rectangle<f32>(left, top, right, bottom);
-}
-
 void BlitScreen::Recreate() {
     present_manager.WaitPresent();
     scheduler.Finish();
@@ -404,10 +354,17 @@ void BlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
         source_image_view = smaa->Draw(scheduler, image_index, source_image, source_image_view);
     }
     if (fsr) {
-        const auto crop_rect = NormalizeCrop(framebuffer, screen_info);
-        const VkExtent2D fsr_input_size{
-            .width = Settings::values.resolution_info.ScaleUp(screen_info.width),
-            .height = Settings::values.resolution_info.ScaleUp(screen_info.height),
+        auto crop_rect = framebuffer.crop_rect;
+        if (crop_rect.GetWidth() == 0) {
+            crop_rect.right = framebuffer.width;
+        }
+        if (crop_rect.GetHeight() == 0) {
+            crop_rect.bottom = framebuffer.height;
+        }
+        crop_rect = crop_rect.Scale(Settings::values.resolution_info.up_factor);
+        VkExtent2D fsr_input_size{
+            .width = Settings::values.resolution_info.ScaleUp(framebuffer.width),
+            .height = Settings::values.resolution_info.ScaleUp(framebuffer.height),
         };
         VkImageView fsr_image_view =
             fsr->Draw(scheduler, image_index, source_image_view, fsr_input_size, crop_rect);
@@ -1440,37 +1397,61 @@ void BlitScreen::SetUniformData(BufferData& data, const Layout::FramebufferLayou
 
 void BlitScreen::SetVertexData(BufferData& data, const Tegra::FramebufferConfig& framebuffer,
                                const Layout::FramebufferLayout layout) const {
-    f32 left, top, right, bottom;
+    const auto& framebuffer_transform_flags = framebuffer.transform_flags;
+    const auto& framebuffer_crop_rect = framebuffer.crop_rect;
 
-    if (fsr) {
-        // FSR has already applied the crop, so we just want to render the image
-        // it has produced.
-        left = 0;
-        top = 0;
-        right = 1;
-        bottom = 1;
-    } else {
-        // Get the normalized crop rectangle.
-        const auto crop = NormalizeCrop(framebuffer, screen_info);
+    static constexpr Common::Rectangle<f32> texcoords{0.f, 0.f, 1.f, 1.f};
+    auto left = texcoords.left;
+    auto right = texcoords.right;
 
-        // Apply the crop.
-        left = crop.left;
-        top = crop.top;
-        right = crop.right;
-        bottom = crop.bottom;
+    switch (framebuffer_transform_flags) {
+    case Service::android::BufferTransformFlags::Unset:
+        break;
+    case Service::android::BufferTransformFlags::FlipV:
+        // Flip the framebuffer vertically
+        left = texcoords.right;
+        right = texcoords.left;
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Unsupported framebuffer_transform_flags={}",
+                          static_cast<u32>(framebuffer_transform_flags));
+        break;
+    }
+
+    UNIMPLEMENTED_IF(framebuffer_crop_rect.left != 0);
+
+    f32 left_start{};
+    if (framebuffer_crop_rect.Top() > 0) {
+        left_start = static_cast<f32>(framebuffer_crop_rect.Top()) /
+                     static_cast<f32>(framebuffer_crop_rect.Bottom());
+    }
+    f32 scale_u = static_cast<f32>(framebuffer.width) / static_cast<f32>(screen_info.width);
+    f32 scale_v = static_cast<f32>(framebuffer.height) / static_cast<f32>(screen_info.height);
+    // Scale the output by the crop width/height. This is commonly used with 1280x720 rendering
+    // (e.g. handheld mode) on a 1920x1080 framebuffer.
+    if (!fsr) {
+        if (framebuffer_crop_rect.GetWidth() > 0) {
+            scale_u = static_cast<f32>(framebuffer_crop_rect.GetWidth()) /
+                      static_cast<f32>(screen_info.width);
+        }
+        if (framebuffer_crop_rect.GetHeight() > 0) {
+            scale_v = static_cast<f32>(framebuffer_crop_rect.GetHeight()) /
+                      static_cast<f32>(screen_info.height);
+        }
     }
 
-    // Map the coordinates to the screen.
     const auto& screen = layout.screen;
     const auto x = static_cast<f32>(screen.left);
     const auto y = static_cast<f32>(screen.top);
     const auto w = static_cast<f32>(screen.GetWidth());
     const auto h = static_cast<f32>(screen.GetHeight());
-
-    data.vertices[0] = ScreenRectVertex(x, y, left, top);
-    data.vertices[1] = ScreenRectVertex(x + w, y, right, top);
-    data.vertices[2] = ScreenRectVertex(x, y + h, left, bottom);
-    data.vertices[3] = ScreenRectVertex(x + w, y + h, right, bottom);
+    data.vertices[0] = ScreenRectVertex(x, y, texcoords.top * scale_u, left_start + left * scale_v);
+    data.vertices[1] =
+        ScreenRectVertex(x + w, y, texcoords.bottom * scale_u, left_start + left * scale_v);
+    data.vertices[2] =
+        ScreenRectVertex(x, y + h, texcoords.top * scale_u, left_start + right * scale_v);
+    data.vertices[3] =
+        ScreenRectVertex(x + w, y + h, texcoords.bottom * scale_u, left_start + right * scale_v);
 }
 
 void BlitScreen::CreateSMAA(VkExtent2D smaa_size) {

src/video_core/renderer_vulkan/vk_fsr.cpp

@@ -34,7 +34,7 @@ FSR::FSR(const Device& device_, MemoryAllocator& memory_allocator_, size_t image
 }
 
 VkImageView FSR::Draw(Scheduler& scheduler, size_t image_index, VkImageView image_view,
-                      VkExtent2D input_image_extent, const Common::Rectangle<f32>& crop_rect) {
+                      VkExtent2D input_image_extent, const Common::Rectangle<int>& crop_rect) {
 
     UpdateDescriptorSet(image_index, image_view);
 
@@ -61,21 +61,15 @@ VkImageView FSR::Draw(Scheduler& scheduler, size_t image_index, VkImageView imag
 
         cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *easu_pipeline);
 
-        const f32 input_image_width = static_cast<f32>(input_image_extent.width);
-        const f32 input_image_height = static_cast<f32>(input_image_extent.height);
-        const f32 output_image_width = static_cast<f32>(output_size.width);
-        const f32 output_image_height = static_cast<f32>(output_size.height);
-        const f32 viewport_width = (crop_rect.right - crop_rect.left) * input_image_width;
-        const f32 viewport_x = crop_rect.left * input_image_width;
-        const f32 viewport_height = (crop_rect.bottom - crop_rect.top) * input_image_height;
-        const f32 viewport_y = crop_rect.top * input_image_height;
-
         std::array<u32, 4 * 4> push_constants;
-        FsrEasuConOffset(push_constants.data() + 0, push_constants.data() + 4,
-                         push_constants.data() + 8, push_constants.data() + 12,
+        FsrEasuConOffset(
+            push_constants.data() + 0, push_constants.data() + 4, push_constants.data() + 8,
+            push_constants.data() + 12,
 
-                         viewport_width, viewport_height, input_image_width, input_image_height,
-                         output_image_width, output_image_height, viewport_x, viewport_y);
+            static_cast<f32>(crop_rect.GetWidth()), static_cast<f32>(crop_rect.GetHeight()),
+            static_cast<f32>(input_image_extent.width), static_cast<f32>(input_image_extent.height),
+            static_cast<f32>(output_size.width), static_cast<f32>(output_size.height),
+            static_cast<f32>(crop_rect.left), static_cast<f32>(crop_rect.top));
         cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_COMPUTE_BIT, push_constants);
 
         {

src/video_core/renderer_vulkan/vk_fsr.h

@@ -17,7 +17,7 @@ public:
     explicit FSR(const Device& device, MemoryAllocator& memory_allocator, size_t image_count,
                  VkExtent2D output_size);
     VkImageView Draw(Scheduler& scheduler, size_t image_index, VkImageView image_view,
-                     VkExtent2D input_image_extent, const Common::Rectangle<f32>& crop_rect);
+                     VkExtent2D input_image_extent, const Common::Rectangle<int>& crop_rect);
 
 private:
     void CreateDescriptorPool();