core: Make sure npad is initialized

vcpkg: Update fmt to 10.1.1

.codespellrc

@@ -3,4 +3,4 @@
 
 [codespell]
 skip = ./.git,./build,./dist,./Doxyfile,./externals,./LICENSES,./src/android/app/src/main/res
-ignore-words-list = aci,allright,ba,canonicalizations,deques,froms,hda,inout,lod,masia,nam,nax,nce,nd,optin,pullrequests,pullrequest,te,transfered,unstall,uscaled,vas,zink
+ignore-words-list = aci,allright,ba,canonicalizations,deques,fpr,froms,hda,inout,lod,masia,nam,nax,nce,nd,optin,pullrequests,pullrequest,te,transfered,unstall,uscaled,vas,zink

dist/72-yuzu-input.rules

@@ -0,0 +1,19 @@
+# SPDX-FileCopyrightText: 2023 yuzu Emulator Project
+# SPDX-License-Identifier: GPL-2.0-or-later
+
+# Allow systemd-logind to manage user access to hidraw with this file
+# On most systems, this file should be installed to /etc/udev/rules.d/72-yuzu-input.rules
+# Consult your distro if this is not the case
+
+# Switch Pro Controller (USB/Bluetooth)
+KERNEL=="hidraw*", ATTRS{idVendor}=="057e", ATTRS{idProduct}=="2009", MODE="0660", TAG+="uaccess"
+KERNEL=="hidraw*", KERNELS=="*057e:2009*", MODE="0660", TAG+="uaccess"
+
+# Joy-Con L (Bluetooth)
+KERNEL=="hidraw*", KERNELS=="*057e:2006*", MODE="0660", TAG+="uaccess"
+
+# Joy-Con R (Bluetooth)
+KERNEL=="hidraw*", KERNELS=="*057e:2007*", MODE="0660", TAG+="uaccess"
+
+# Joy-Con Charging Grip (USB)
+KERNEL=="hidraw*", ATTRS{idVendor}=="057e", ATTRS{idProduct}=="200e", MODE="0660", TAG+="uaccess"

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

@@ -14,8 +14,10 @@ import org.yuzu.yuzu_emu.R
 import org.yuzu.yuzu_emu.databinding.DialogAddFolderBinding
 import org.yuzu.yuzu_emu.model.GameDir
 import org.yuzu.yuzu_emu.model.GamesViewModel
+import org.yuzu.yuzu_emu.model.HomeViewModel
 
 class AddGameFolderDialogFragment : DialogFragment() {
+    private val homeViewModel: HomeViewModel by activityViewModels()
     private val gamesViewModel: GamesViewModel by activityViewModels()
 
     override fun onCreateDialog(savedInstanceState: Bundle?): Dialog {
@@ -30,6 +32,7 @@ class AddGameFolderDialogFragment : DialogFragment() {
             .setTitle(R.string.add_game_folder)
             .setPositiveButton(android.R.string.ok) { _: DialogInterface, _: Int ->
                 val newGameDir = GameDir(folderUriString!!, binding.deepScanSwitch.isChecked)
+                homeViewModel.setGamesDirSelected(true)
                 gamesViewModel.addFolder(newGameDir)
             }
             .setNegativeButton(android.R.string.cancel, null)

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

@@ -4,6 +4,7 @@
 package org.yuzu.yuzu_emu.fragments
 
 import android.Manifest
+import android.annotation.SuppressLint
 import android.content.Intent
 import android.os.Build
 import android.os.Bundle
@@ -75,6 +76,8 @@ class SetupFragment : Fragment() {
         return binding.root
     }
 
+    // This is using the correct scope, lint is just acting up
+    @SuppressLint("UnsafeRepeatOnLifecycleDetector")
     override fun onViewCreated(view: View, savedInstanceState: Bundle?) {
         mainActivity = requireActivity() as MainActivity
 
@@ -206,12 +209,24 @@ class SetupFragment : Fragment() {
             )
         }
 
-        viewLifecycleOwner.lifecycleScope.launch {
-            repeatOnLifecycle(Lifecycle.State.CREATED) {
-                homeViewModel.shouldPageForward.collect {
-                    if (it) {
-                        pageForward()
-                        homeViewModel.setShouldPageForward(false)
+        viewLifecycleOwner.lifecycleScope.apply {
+            launch {
+                repeatOnLifecycle(Lifecycle.State.CREATED) {
+                    homeViewModel.shouldPageForward.collect {
+                        if (it) {
+                            pageForward()
+                            homeViewModel.setShouldPageForward(false)
+                        }
+                    }
+                }
+            }
+            launch {
+                repeatOnLifecycle(Lifecycle.State.CREATED) {
+                    homeViewModel.gamesDirSelected.collect {
+                        if (it) {
+                            gamesDirCallback.onStepCompleted()
+                            homeViewModel.setGamesDirSelected(false)
+                        }
                     }
                 }
             }
@@ -339,7 +354,6 @@ class SetupFragment : Fragment() {
         registerForActivityResult(ActivityResultContracts.OpenDocumentTree()) { result ->
             if (result != null) {
                 mainActivity.processGamesDir(result)
-                gamesDirCallback.onStepCompleted()
             }
         }
 

src/android/app/src/main/java/org/yuzu/yuzu_emu/model/GamesViewModel.kt

@@ -133,7 +133,7 @@ class GamesViewModel : ViewModel() {
         viewModelScope.launch {
             withContext(Dispatchers.IO) {
                 NativeConfig.addGameDir(gameDir)
-                getGameDirs()
+                getGameDirs(true)
             }
         }
 

src/android/app/src/main/java/org/yuzu/yuzu_emu/model/HomeViewModel.kt

@@ -6,6 +6,7 @@ package org.yuzu.yuzu_emu.model
 import androidx.lifecycle.ViewModel
 import kotlinx.coroutines.flow.MutableStateFlow
 import kotlinx.coroutines.flow.StateFlow
+import kotlinx.coroutines.flow.asStateFlow
 
 class HomeViewModel : ViewModel() {
     val navigationVisible: StateFlow<Pair<Boolean, Boolean>> get() = _navigationVisible
@@ -17,6 +18,9 @@ class HomeViewModel : ViewModel() {
     val shouldPageForward: StateFlow<Boolean> get() = _shouldPageForward
     private val _shouldPageForward = MutableStateFlow(false)
 
+    private val _gamesDirSelected = MutableStateFlow(false)
+    val gamesDirSelected get() = _gamesDirSelected.asStateFlow()
+
     var navigatedToSetup = false
 
     fun setNavigationVisibility(visible: Boolean, animated: Boolean) {
@@ -36,4 +40,8 @@ class HomeViewModel : ViewModel() {
     fun setShouldPageForward(pageForward: Boolean) {
         _shouldPageForward.value = pageForward
     }
+
+    fun setGamesDirSelected(selected: Boolean) {
+        _gamesDirSelected.value = selected
+    }
 }

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

@@ -291,9 +291,6 @@ Core::SystemResultStatus EmulationSession::InitializeEmulation(const std::string
     // Initialize filesystem.
     ConfigureFilesystemProvider(filepath);
 
-    // Initialize account manager
-    m_profile_manager = std::make_unique<Service::Account::ProfileManager>();
-
     // Load the ROM.
     m_load_result = m_system.Load(EmulationSession::GetInstance().Window(), filepath);
     if (m_load_result != Core::SystemResultStatus::Success) {
@@ -736,8 +733,8 @@ void Java_org_yuzu_yuzu_1emu_NativeLibrary_initializeEmptyUserDirectory(JNIEnv*
     auto vfs_nand_dir = EmulationSession::GetInstance().System().GetFilesystem()->OpenDirectory(
         Common::FS::PathToUTF8String(nand_dir), FileSys::Mode::Read);
 
-    Service::Account::ProfileManager manager;
-    const auto user_id = manager.GetUser(static_cast<std::size_t>(0));
+    const auto user_id = EmulationSession::GetInstance().System().GetProfileManager().GetUser(
+        static_cast<std::size_t>(0));
     ASSERT(user_id);
 
     const auto user_save_data_path = FileSys::SaveDataFactory::GetFullPath(

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

@@ -73,7 +73,6 @@ private:
     std::atomic<bool> m_is_running = false;
     std::atomic<bool> m_is_paused = false;
     SoftwareKeyboard::AndroidKeyboard* m_software_keyboard{};
-    std::unique_ptr<Service::Account::ProfileManager> m_profile_manager;
     std::unique_ptr<FileSys::ManualContentProvider> m_manual_provider;
 
     // GPU driver parameters

src/common/settings.cpp

@@ -160,12 +160,16 @@ static bool is_nce_enabled = false;
 
 void SetNceEnabled(bool is_39bit) {
     const bool is_nce_selected = values.cpu_backend.GetValue() == CpuBackend::Nce;
-    is_nce_enabled = IsFastmemEnabled() && is_nce_selected && is_39bit;
-    if (is_nce_selected && !is_nce_enabled) {
+    if (is_nce_selected && !IsFastmemEnabled()) {
+        LOG_WARNING(Common, "Fastmem is required to natively execute code in a performant manner, "
+                            "falling back to Dynarmic");
+    }
+    if (is_nce_selected && !is_39bit) {
         LOG_WARNING(
             Common,
             "Program does not utilize 39-bit address space, unable to natively execute code");
     }
+    is_nce_enabled = IsFastmemEnabled() && is_nce_selected && is_39bit;
 }
 
 bool IsNceEnabled() {

src/common/settings.h

@@ -180,14 +180,20 @@ struct Values {
                                              &use_speed_limit};
 
     // Cpu
-    SwitchableSetting<CpuBackend, true> cpu_backend{
-        linkage,         CpuBackend::Dynarmic, CpuBackend::Dynarmic,
+    SwitchableSetting<CpuBackend, true> cpu_backend{linkage,
 #ifdef HAS_NCE
-        CpuBackend::Nce,
+                                                    CpuBackend::Nce,
 #else
                                                     CpuBackend::Dynarmic,
 #endif
-        "cpu_backend",   Category::Cpu};
+                                                    CpuBackend::Dynarmic,
+#ifdef HAS_NCE
+                                                    CpuBackend::Nce,
+#else
+                                                    CpuBackend::Dynarmic,
+#endif
+                                                    "cpu_backend",
+                                                    Category::Cpu};
     SwitchableSetting<CpuAccuracy, true> cpu_accuracy{linkage,           CpuAccuracy::Auto,
                                                       CpuAccuracy::Auto, CpuAccuracy::Paranoid,
                                                       "cpu_accuracy",    Category::Cpu};

src/core/CMakeLists.txt

@@ -4,6 +4,8 @@
 add_library(core STATIC
     arm/arm_interface.h
     arm/arm_interface.cpp
+    arm/debug.cpp
+    arm/debug.h
     arm/exclusive_monitor.cpp
     arm/exclusive_monitor.h
     arm/symbols.cpp
@@ -249,10 +251,16 @@ add_library(core STATIC
     hle/kernel/k_hardware_timer.h
     hle/kernel/k_interrupt_manager.cpp
     hle/kernel/k_interrupt_manager.h
+    hle/kernel/k_light_client_session.cpp
+    hle/kernel/k_light_client_session.h
     hle/kernel/k_light_condition_variable.cpp
     hle/kernel/k_light_condition_variable.h
     hle/kernel/k_light_lock.cpp
     hle/kernel/k_light_lock.h
+    hle/kernel/k_light_server_session.cpp
+    hle/kernel/k_light_server_session.h
+    hle/kernel/k_light_session.cpp
+    hle/kernel/k_light_session.h
     hle/kernel/k_memory_block.h
     hle/kernel/k_memory_block_manager.cpp
     hle/kernel/k_memory_block_manager.h
@@ -541,6 +549,13 @@ add_library(core STATIC
     hle/service/hid/xcd.cpp
     hle/service/hid/xcd.h
     hle/service/hid/errors.h
+    hle/service/hid/controllers/types/debug_pad_types.h
+    hle/service/hid/controllers/types/keyboard_types.h
+    hle/service/hid/controllers/types/mouse_types.h
+    hle/service/hid/controllers/types/npad_types.h
+    hle/service/hid/controllers/types/touch_types.h
+    hle/service/hid/controllers/applet_resource.cpp
+    hle/service/hid/controllers/applet_resource.h
     hle/service/hid/controllers/console_six_axis.cpp
     hle/service/hid/controllers/console_six_axis.h
     hle/service/hid/controllers/controller_base.cpp
@@ -559,14 +574,15 @@ add_library(core STATIC
     hle/service/hid/controllers/palma.h
     hle/service/hid/controllers/seven_six_axis.cpp
     hle/service/hid/controllers/seven_six_axis.h
+    hle/service/hid/controllers/shared_memory_format.h
+    hle/service/hid/controllers/shared_memory_holder.cpp
+    hle/service/hid/controllers/shared_memory_holder.h
     hle/service/hid/controllers/six_axis.cpp
     hle/service/hid/controllers/six_axis.h
     hle/service/hid/controllers/stubbed.cpp
     hle/service/hid/controllers/stubbed.h
     hle/service/hid/controllers/touchscreen.cpp
     hle/service/hid/controllers/touchscreen.h
-    hle/service/hid/controllers/xpad.cpp
-    hle/service/hid/controllers/xpad.h
     hle/service/hid/hidbus/hidbus_base.cpp
     hle/service/hid/hidbus/hidbus_base.h
     hle/service/hid/hidbus/ringcon.cpp
@@ -762,6 +778,12 @@ add_library(core STATIC
     hle/service/kernel_helpers.h
     hle/service/mutex.cpp
     hle/service/mutex.h
+    hle/service/ro/ro_nro_utils.cpp
+    hle/service/ro/ro_nro_utils.h
+    hle/service/ro/ro_results.h
+    hle/service/ro/ro_types.h
+    hle/service/ro/ro.cpp
+    hle/service/ro/ro.h
     hle/service/server_manager.cpp
     hle/service/server_manager.h
     hle/service/service.cpp

src/core/arm/arm_interface.cpp

@@ -1,231 +1,32 @@
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include <map>
-#include <optional>
-
-#include "common/bit_field.h"
-#include "common/common_types.h"
-#include "common/demangle.h"
 #include "common/logging/log.h"
 #include "core/arm/arm_interface.h"
-#include "core/arm/symbols.h"
+#include "core/arm/debug.h"
 #include "core/core.h"
-#include "core/debugger/debugger.h"
 #include "core/hle/kernel/k_process.h"
-#include "core/hle/kernel/k_thread.h"
-#include "core/hle/kernel/svc.h"
-#include "core/loader/loader.h"
-#include "core/memory.h"
 
 namespace Core {
 
-constexpr u64 SEGMENT_BASE = 0x7100000000ull;
+void ArmInterface::LogBacktrace(const Kernel::KProcess* process) const {
+    Kernel::Svc::ThreadContext ctx;
+    this->GetContext(ctx);
 
-std::vector<ARM_Interface::BacktraceEntry> ARM_Interface::GetBacktraceFromContext(
-    Core::System& system, const ARM_Interface::ThreadContext32& ctx) {
-    std::vector<BacktraceEntry> out;
-    auto& memory = system.ApplicationMemory();
-
-    const auto& reg = ctx.cpu_registers;
-    u32 pc = reg[15], lr = reg[14], fp = reg[11];
-    out.push_back({"", 0, pc, 0, ""});
-
-    // fp (= r11) points to the last frame record.
-    // Frame records are two words long:
-    // fp+0 : pointer to previous frame record
-    // fp+4 : value of lr for frame
-    for (size_t i = 0; i < 256; i++) {
-        out.push_back({"", 0, lr, 0, ""});
-        if (!fp || (fp % 4 != 0) || !memory.IsValidVirtualAddressRange(fp, 8)) {
-            break;
-        }
-        lr = memory.Read32(fp + 4);
-        fp = memory.Read32(fp);
-    }
-
-    SymbolicateBacktrace(system, out);
-
-    return out;
-}
-
-std::vector<ARM_Interface::BacktraceEntry> ARM_Interface::GetBacktraceFromContext(
-    Core::System& system, const ARM_Interface::ThreadContext64& ctx) {
-    std::vector<BacktraceEntry> out;
-    auto& memory = system.ApplicationMemory();
-
-    const auto& reg = ctx.cpu_registers;
-    u64 pc = ctx.pc, lr = reg[30], fp = reg[29];
-
-    out.push_back({"", 0, pc, 0, ""});
-
-    // fp (= x29) points to the previous frame record.
-    // Frame records are two words long:
-    // fp+0 : pointer to previous frame record
-    // fp+8 : value of lr for frame
-    for (size_t i = 0; i < 256; i++) {
-        out.push_back({"", 0, lr, 0, ""});
-        if (!fp || (fp % 4 != 0) || !memory.IsValidVirtualAddressRange(fp, 16)) {
-            break;
-        }
-        lr = memory.Read64(fp + 8);
-        fp = memory.Read64(fp);
-    }
-
-    SymbolicateBacktrace(system, out);
-
-    return out;
-}
-
-void ARM_Interface::SymbolicateBacktrace(Core::System& system, std::vector<BacktraceEntry>& out) {
-    std::map<VAddr, std::string> modules;
-    auto& loader{system.GetAppLoader()};
-    if (loader.ReadNSOModules(modules) != Loader::ResultStatus::Success) {
-        return;
-    }
-
-    std::map<std::string, Symbols::Symbols> symbols;
-    for (const auto& module : modules) {
-        symbols.insert_or_assign(module.second,
-                                 Symbols::GetSymbols(module.first, system.ApplicationMemory(),
-                                                     system.ApplicationProcess()->Is64Bit()));
-    }
-
-    for (auto& entry : out) {
-        VAddr base = 0;
-        for (auto iter = modules.rbegin(); iter != modules.rend(); ++iter) {
-            const auto& module{*iter};
-            if (entry.original_address >= module.first) {
-                entry.module = module.second;
-                base = module.first;
-                break;
-            }
-        }
-
-        entry.offset = entry.original_address - base;
-        entry.address = SEGMENT_BASE + entry.offset;
-
-        if (entry.module.empty()) {
-            entry.module = "unknown";
-        }
-
-        const auto symbol_set = symbols.find(entry.module);
-        if (symbol_set != symbols.end()) {
-            const auto symbol = Symbols::GetSymbolName(symbol_set->second, entry.offset);
-            if (symbol) {
-                entry.name = Common::DemangleSymbol(*symbol);
-            }
-        }
-    }
-}
-
-std::vector<ARM_Interface::BacktraceEntry> ARM_Interface::GetBacktrace() const {
-    if (GetArchitecture() == Architecture::Aarch64) {
-        ThreadContext64 ctx;
-        SaveContext(ctx);
-        return GetBacktraceFromContext(system, ctx);
-    } else {
-        ThreadContext32 ctx;
-        SaveContext(ctx);
-        return GetBacktraceFromContext(system, ctx);
-    }
-}
-
-void ARM_Interface::LogBacktrace() const {
-    const VAddr sp = GetSP();
-    const VAddr pc = GetPC();
-    LOG_ERROR(Core_ARM, "Backtrace, sp={:016X}, pc={:016X}", sp, pc);
+    LOG_ERROR(Core_ARM, "Backtrace, sp={:016X}, pc={:016X}", ctx.sp, ctx.pc);
     LOG_ERROR(Core_ARM, "{:20}{:20}{:20}{:20}{}", "Module Name", "Address", "Original Address",
               "Offset", "Symbol");
     LOG_ERROR(Core_ARM, "");
-    const auto backtrace = GetBacktrace();
+    const auto backtrace = GetBacktraceFromContext(process, ctx);
     for (const auto& entry : backtrace) {
         LOG_ERROR(Core_ARM, "{:20}{:016X}    {:016X}    {:016X}    {}", entry.module, entry.address,
                   entry.original_address, entry.offset, entry.name);
     }
 }
 
-void ARM_Interface::Run() {
-    using Kernel::StepState;
-    using Kernel::SuspendType;
-
-    while (true) {
-        Kernel::KThread* current_thread{Kernel::GetCurrentThreadPointer(system.Kernel())};
-        HaltReason hr{};
-
-        // If the thread is scheduled for termination, exit the thread.
-        if (current_thread->HasDpc()) {
-            if (current_thread->IsTerminationRequested()) {
-                current_thread->Exit();
-                UNREACHABLE();
-            }
-        }
-
-        // Notify the debugger and go to sleep if a step was performed
-        // and this thread has been scheduled again.
-        if (current_thread->GetStepState() == StepState::StepPerformed) {
-            system.GetDebugger().NotifyThreadStopped(current_thread);
-            current_thread->RequestSuspend(SuspendType::Debug);
-            break;
-        }
-
-        // Otherwise, run the thread.
-        system.EnterCPUProfile();
-        if (current_thread->GetStepState() == StepState::StepPending) {
-            hr = StepJit();
-
-            if (True(hr & HaltReason::StepThread)) {
-                current_thread->SetStepState(StepState::StepPerformed);
-            }
-        } else {
-            hr = RunJit();
-        }
-        system.ExitCPUProfile();
-
-        // Notify the debugger and go to sleep if a breakpoint was hit,
-        // or if the thread is unable to continue for any reason.
-        if (True(hr & HaltReason::InstructionBreakpoint) || True(hr & HaltReason::PrefetchAbort)) {
-            if (!True(hr & HaltReason::PrefetchAbort)) {
-                RewindBreakpointInstruction();
-            }
-            if (system.DebuggerEnabled()) {
-                system.GetDebugger().NotifyThreadStopped(current_thread);
-            } else {
-                LogBacktrace();
-            }
-            current_thread->RequestSuspend(SuspendType::Debug);
-            break;
-        }
-
-        // Notify the debugger and go to sleep if a watchpoint was hit.
-        if (True(hr & HaltReason::DataAbort)) {
-            if (system.DebuggerEnabled()) {
-                system.GetDebugger().NotifyThreadWatchpoint(current_thread, *HaltedWatchpoint());
-            } else {
-                LogBacktrace();
-            }
-            current_thread->RequestSuspend(SuspendType::Debug);
-            break;
-        }
-
-        // Handle syscalls and scheduling (this may change the current thread/core)
-        if (True(hr & HaltReason::SupervisorCall)) {
-            Kernel::Svc::Call(system, GetSvcNumber());
-            break;
-        }
-        if (True(hr & HaltReason::BreakLoop) || !uses_wall_clock) {
-            break;
-        }
-    }
-}
-
-void ARM_Interface::LoadWatchpointArray(const WatchpointArray* wp) {
-    watchpoints = wp;
-}
-
-const Kernel::DebugWatchpoint* ARM_Interface::MatchingWatchpoint(
+const Kernel::DebugWatchpoint* ArmInterface::MatchingWatchpoint(
     u64 addr, u64 size, Kernel::DebugWatchpointType access_type) const {
-    if (!watchpoints) {
+    if (!m_watchpoints) {
         return nullptr;
     }
 
@@ -233,7 +34,7 @@ const Kernel::DebugWatchpoint* ARM_Interface::MatchingWatchpoint(
     const u64 end_address{addr + size};
 
     for (size_t i = 0; i < Core::Hardware::NUM_WATCHPOINTS; i++) {
-        const auto& watch{(*watchpoints)[i]};
+        const auto& watch{(*m_watchpoints)[i]};
 
         if (end_address <= GetInteger(watch.start_address)) {
             continue;

src/core/arm/arm_interface.h

@@ -12,20 +12,20 @@
 #include "common/common_types.h"
 #include "core/hardware_properties.h"
 
+#include "core/hle/kernel/svc_types.h"
+
 namespace Common {
 struct PageTable;
 }
 
 namespace Kernel {
-enum class VMAPermission : u8;
 enum class DebugWatchpointType : u8;
 struct DebugWatchpoint;
+class KThread;
+class KProcess;
 } // namespace Kernel
 
 namespace Core {
-class System;
-class CPUInterruptHandler;
-
 using WatchpointArray = std::array<Kernel::DebugWatchpoint, Core::Hardware::NUM_WATCHPOINTS>;
 
 // NOTE: these values match the HaltReason enum in Dynarmic
@@ -40,197 +40,74 @@ enum class HaltReason : u64 {
 DECLARE_ENUM_FLAG_OPERATORS(HaltReason);
 
 enum class Architecture {
-    Aarch32,
-    Aarch64,
+    AArch64,
+    AArch32,
 };
 
 /// Generic ARMv8 CPU interface
-class ARM_Interface {
+class ArmInterface {
 public:
-    YUZU_NON_COPYABLE(ARM_Interface);
-    YUZU_NON_MOVEABLE(ARM_Interface);
+    YUZU_NON_COPYABLE(ArmInterface);
+    YUZU_NON_MOVEABLE(ArmInterface);
 
-    explicit ARM_Interface(System& system_, bool uses_wall_clock_)
-        : system{system_}, uses_wall_clock{uses_wall_clock_} {}
-    virtual ~ARM_Interface() = default;
+    explicit ArmInterface(bool uses_wall_clock) : m_uses_wall_clock{uses_wall_clock} {}
+    virtual ~ArmInterface() = default;
 
-    struct ThreadContext32 {
-        std::array<u32, 16> cpu_registers{};
-        std::array<u32, 64> extension_registers{};
-        u32 cpsr{};
-        u32 fpscr{};
-        u32 fpexc{};
-        u32 tpidr{};
-    };
-    // Internally within the kernel, it expects the AArch32 version of the
-    // thread context to be 344 bytes in size.
-    static_assert(sizeof(ThreadContext32) == 0x150);
-
-    struct ThreadContext64 {
-        std::array<u64, 31> cpu_registers{};
-        u64 sp{};
-        u64 pc{};
-        u32 pstate{};
-        std::array<u8, 4> padding{};
-        std::array<u128, 32> vector_registers{};
-        u32 fpcr{};
-        u32 fpsr{};
-        u64 tpidr{};
-    };
-    // Internally within the kernel, it expects the AArch64 version of the
-    // thread context to be 800 bytes in size.
-    static_assert(sizeof(ThreadContext64) == 0x320);
-
-    /// Perform any backend-specific initialization.
+    // Perform any backend-specific initialization.
     virtual void Initialize() {}
 
-    /// Runs the CPU until an event happens
-    void Run();
+    // Runs the CPU until an event happens.
+    virtual HaltReason RunThread(Kernel::KThread* thread) = 0;
 
-    /// Clear all instruction cache
+    // Runs the CPU for one instruction or until an event happens.
+    virtual HaltReason StepThread(Kernel::KThread* thread) = 0;
+
+    // Admits a backend-specific mechanism to lock the thread context.
+    virtual void LockThread(Kernel::KThread* thread) {}
+    virtual void UnlockThread(Kernel::KThread* thread) {}
+
+    // Clear the entire instruction cache for this CPU.
     virtual void ClearInstructionCache() = 0;
 
-    /**
-     * Clear instruction cache range
-     * @param addr Start address of the cache range to clear
-     * @param size Size of the cache range to clear, starting at addr
-     */
+    // Clear a range of the instruction cache for this CPU.
     virtual void InvalidateCacheRange(u64 addr, std::size_t size) = 0;
 
-    /**
-     * Notifies CPU emulation that the current page table has changed.
-     *  @param new_page_table                 The new page table.
-     *  @param new_address_space_size_in_bits The new usable size of the address space in bits.
-     *                                        This can be either 32, 36, or 39 on official software.
-     */
-    virtual void PageTableChanged(Common::PageTable& new_page_table,
-                                  std::size_t new_address_space_size_in_bits) = 0;
-
-    /**
-     * Set the Program Counter to an address
-     * @param addr Address to set PC to
-     */
-    virtual void SetPC(u64 addr) = 0;
-
-    /*
-     * Get the current Program Counter
-     * @return Returns current PC
-     */
-    virtual u64 GetPC() const = 0;
-
-    /**
-     * Get the current Stack Pointer
-     * @return Returns current SP
-     */
-    virtual u64 GetSP() const = 0;
-
-    /**
-     * Get an ARM register
-     * @param index Register index
-     * @return Returns the value in the register
-     */
-    virtual u64 GetReg(int index) const = 0;
-
-    /**
-     * Set an ARM register
-     * @param index Register index
-     * @param value Value to set register to
-     */
-    virtual void SetReg(int index, u64 value) = 0;
-
-    /**
-     * Gets the value of a specified vector register.
-     *
-     * @param index The index of the vector register.
-     * @return the value within the vector register.
-     */
-    virtual u128 GetVectorReg(int index) const = 0;
-
-    /**
-     * Sets a given value into a vector register.
-     *
-     * @param index The index of the vector register.
-     * @param value The new value to place in the register.
-     */
-    virtual void SetVectorReg(int index, u128 value) = 0;
-
-    /**
-     * Get the current PSTATE register
-     * @return Returns the value of the PSTATE register
-     */
-    virtual u32 GetPSTATE() const = 0;
-
-    /**
-     * Set the current PSTATE register
-     * @param pstate Value to set PSTATE to
-     */
-    virtual void SetPSTATE(u32 pstate) = 0;
-
-    virtual u64 GetTlsAddress() const = 0;
-
-    virtual void SetTlsAddress(u64 address) = 0;
-
-    /**
-     * Gets the value within the TPIDR_EL0 (read/write software thread ID) register.
-     *
-     * @return the value within the register.
-     */
-    virtual u64 GetTPIDR_EL0() const = 0;
-
-    /**
-     * Sets a new value within the TPIDR_EL0 (read/write software thread ID) register.
-     *
-     * @param value The new value to place in the register.
-     */
-    virtual void SetTPIDR_EL0(u64 value) = 0;
-
+    // Get the current architecture.
+    // This returns AArch64 when PSTATE.nRW == 0 and AArch32 when PSTATE.nRW == 1.
     virtual Architecture GetArchitecture() const = 0;
-    virtual void SaveContext(ThreadContext32& ctx) const = 0;
-    virtual void SaveContext(ThreadContext64& ctx) const = 0;
-    virtual void LoadContext(const ThreadContext32& ctx) = 0;
-    virtual void LoadContext(const ThreadContext64& ctx) = 0;
-    void LoadWatchpointArray(const WatchpointArray* wp);
 
-    /// Clears the exclusive monitor's state.
-    virtual void ClearExclusiveState() = 0;
+    // Context accessors.
+    // These should not be called if the CPU is running.
+    virtual void GetContext(Kernel::Svc::ThreadContext& ctx) const = 0;
+    virtual void SetContext(const Kernel::Svc::ThreadContext& ctx) = 0;
+    virtual void SetTpidrroEl0(u64 value) = 0;
 
-    /// Signal an interrupt and ask the core to halt as soon as possible.
-    virtual void SignalInterrupt() = 0;
+    virtual void GetSvcArguments(std::span<uint64_t, 8> args) const = 0;
+    virtual void SetSvcArguments(std::span<const uint64_t, 8> args) = 0;
+    virtual u32 GetSvcNumber() const = 0;
 
-    /// Clear a previous interrupt.
-    virtual void ClearInterrupt() = 0;
+    void SetWatchpointArray(const WatchpointArray* watchpoints) {
+        m_watchpoints = watchpoints;
+    }
 
-    struct BacktraceEntry {
-        std::string module;
-        u64 address;
-        u64 original_address;
-        u64 offset;
-        std::string name;
-    };
+    // Signal an interrupt for execution to halt as soon as possible.
+    // It is safe to call this if the CPU is not running.
+    virtual void SignalInterrupt(Kernel::KThread* thread) = 0;
 
-    static std::vector<BacktraceEntry> GetBacktraceFromContext(System& system,
-                                                               const ThreadContext32& ctx);
-    static std::vector<BacktraceEntry> GetBacktraceFromContext(System& system,
-                                                               const ThreadContext64& ctx);
+    // Stack trace generation.
+    void LogBacktrace(const Kernel::KProcess* process) const;
 
-    std::vector<BacktraceEntry> GetBacktrace() const;
-    void LogBacktrace() const;
+    // Debug functionality.
+    virtual const Kernel::DebugWatchpoint* HaltedWatchpoint() const = 0;
+    virtual void RewindBreakpointInstruction() = 0;
 
 protected:
-    /// System context that this ARM interface is running under.
-    System& system;
-    const WatchpointArray* watchpoints;
-    bool uses_wall_clock;
-
-    static void SymbolicateBacktrace(Core::System& system, std::vector<BacktraceEntry>& out);
     const Kernel::DebugWatchpoint* MatchingWatchpoint(
         u64 addr, u64 size, Kernel::DebugWatchpointType access_type) const;
 
-    virtual HaltReason RunJit() = 0;
-    virtual HaltReason StepJit() = 0;
-    virtual u32 GetSvcNumber() const = 0;
-    virtual const Kernel::DebugWatchpoint* HaltedWatchpoint() const = 0;
-    virtual void RewindBreakpointInstruction() = 0;
+protected:
+    const WatchpointArray* m_watchpoints{};
+    bool m_uses_wall_clock{};
 };
 
 } // namespace Core

src/core/arm/debug.cpp

@@ -0,0 +1,354 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "common/demangle.h"
+#include "core/arm/debug.h"
+#include "core/arm/symbols.h"
+#include "core/hle/kernel/k_process.h"
+#include "core/hle/kernel/k_thread.h"
+#include "core/memory.h"
+
+namespace Core {
+
+namespace {
+
+std::optional<std::string> GetNameFromThreadType64(Core::Memory::Memory& memory,
+                                                   const Kernel::KThread& thread) {
+    // Read thread type from TLS
+    const VAddr tls_thread_type{memory.Read64(thread.GetTlsAddress() + 0x1f8)};
+    const VAddr argument_thread_type{thread.GetArgument()};
+
+    if (argument_thread_type && tls_thread_type != argument_thread_type) {
+        // Probably not created by nnsdk, no name available.
+        return std::nullopt;
+    }
+
+    if (!tls_thread_type) {
+        return std::nullopt;
+    }
+
+    const u16 version{memory.Read16(tls_thread_type + 0x46)};
+    VAddr name_pointer{};
+    if (version == 1) {
+        name_pointer = memory.Read64(tls_thread_type + 0x1a0);
+    } else {
+        name_pointer = memory.Read64(tls_thread_type + 0x1a8);
+    }
+
+    if (!name_pointer) {
+        // No name provided.
+        return std::nullopt;
+    }
+
+    return memory.ReadCString(name_pointer, 256);
+}
+
+std::optional<std::string> GetNameFromThreadType32(Core::Memory::Memory& memory,
+                                                   const Kernel::KThread& thread) {
+    // Read thread type from TLS
+    const VAddr tls_thread_type{memory.Read32(thread.GetTlsAddress() + 0x1fc)};
+    const VAddr argument_thread_type{thread.GetArgument()};
+
+    if (argument_thread_type && tls_thread_type != argument_thread_type) {
+        // Probably not created by nnsdk, no name available.
+        return std::nullopt;
+    }
+
+    if (!tls_thread_type) {
+        return std::nullopt;
+    }
+
+    const u16 version{memory.Read16(tls_thread_type + 0x26)};
+    VAddr name_pointer{};
+    if (version == 1) {
+        name_pointer = memory.Read32(tls_thread_type + 0xe4);
+    } else {
+        name_pointer = memory.Read32(tls_thread_type + 0xe8);
+    }
+
+    if (!name_pointer) {
+        // No name provided.
+        return std::nullopt;
+    }
+
+    return memory.ReadCString(name_pointer, 256);
+}
+
+constexpr std::array<u64, 2> SegmentBases{
+    0x60000000ULL,
+    0x7100000000ULL,
+};
+
+void SymbolicateBacktrace(const Kernel::KProcess* process, std::vector<BacktraceEntry>& out) {
+    auto modules = FindModules(process);
+
+    const bool is_64 = process->Is64Bit();
+
+    std::map<std::string, Symbols::Symbols> symbols;
+    for (const auto& module : modules) {
+        symbols.insert_or_assign(module.second,
+                                 Symbols::GetSymbols(module.first, process->GetMemory(), is_64));
+    }
+
+    for (auto& entry : out) {
+        VAddr base = 0;
+        for (auto iter = modules.rbegin(); iter != modules.rend(); ++iter) {
+            const auto& module{*iter};
+            if (entry.original_address >= module.first) {
+                entry.module = module.second;
+                base = module.first;
+                break;
+            }
+        }
+
+        entry.offset = entry.original_address - base;
+        entry.address = SegmentBases[is_64] + entry.offset;
+
+        if (entry.module.empty()) {
+            entry.module = "unknown";
+        }
+
+        const auto symbol_set = symbols.find(entry.module);
+        if (symbol_set != symbols.end()) {
+            const auto symbol = Symbols::GetSymbolName(symbol_set->second, entry.offset);
+            if (symbol) {
+                entry.name = Common::DemangleSymbol(*symbol);
+            }
+        }
+    }
+}
+
+std::vector<BacktraceEntry> GetAArch64Backtrace(const Kernel::KProcess* process,
+                                                const Kernel::Svc::ThreadContext& ctx) {
+    std::vector<BacktraceEntry> out;
+    auto& memory = process->GetMemory();
+    auto pc = ctx.pc, lr = ctx.lr, fp = ctx.fp;
+
+    out.push_back({"", 0, pc, 0, ""});
+
+    // fp (= x29) points to the previous frame record.
+    // Frame records are two words long:
+    // fp+0 : pointer to previous frame record
+    // fp+8 : value of lr for frame
+    for (size_t i = 0; i < 256; i++) {
+        out.push_back({"", 0, lr, 0, ""});
+        if (!fp || (fp % 4 != 0) || !memory.IsValidVirtualAddressRange(fp, 16)) {
+            break;
+        }
+        lr = memory.Read64(fp + 8);
+        fp = memory.Read64(fp);
+    }
+
+    SymbolicateBacktrace(process, out);
+
+    return out;
+}
+
+std::vector<BacktraceEntry> GetAArch32Backtrace(const Kernel::KProcess* process,
+                                                const Kernel::Svc::ThreadContext& ctx) {
+    std::vector<BacktraceEntry> out;
+    auto& memory = process->GetMemory();
+    auto pc = ctx.pc, lr = ctx.lr, fp = ctx.fp;
+
+    out.push_back({"", 0, pc, 0, ""});
+
+    // fp (= r11) points to the last frame record.
+    // Frame records are two words long:
+    // fp+0 : pointer to previous frame record
+    // fp+4 : value of lr for frame
+    for (size_t i = 0; i < 256; i++) {
+        out.push_back({"", 0, lr, 0, ""});
+        if (!fp || (fp % 4 != 0) || !memory.IsValidVirtualAddressRange(fp, 8)) {
+            break;
+        }
+        lr = memory.Read32(fp + 4);
+        fp = memory.Read32(fp);
+    }
+
+    SymbolicateBacktrace(process, out);
+
+    return out;
+}
+
+} // namespace
+
+std::optional<std::string> GetThreadName(const Kernel::KThread* thread) {
+    const auto* process = thread->GetOwnerProcess();
+    if (process->Is64Bit()) {
+        return GetNameFromThreadType64(process->GetMemory(), *thread);
+    } else {
+        return GetNameFromThreadType32(process->GetMemory(), *thread);
+    }
+}
+
+std::string_view GetThreadWaitReason(const Kernel::KThread* thread) {
+    switch (thread->GetWaitReasonForDebugging()) {
+    case Kernel::ThreadWaitReasonForDebugging::Sleep:
+        return "Sleep";
+    case Kernel::ThreadWaitReasonForDebugging::IPC:
+        return "IPC";
+    case Kernel::ThreadWaitReasonForDebugging::Synchronization:
+        return "Synchronization";
+    case Kernel::ThreadWaitReasonForDebugging::ConditionVar:
+        return "ConditionVar";
+    case Kernel::ThreadWaitReasonForDebugging::Arbitration:
+        return "Arbitration";
+    case Kernel::ThreadWaitReasonForDebugging::Suspended:
+        return "Suspended";
+    default:
+        return "Unknown";
+    }
+}
+
+std::string GetThreadState(const Kernel::KThread* thread) {
+    switch (thread->GetState()) {
+    case Kernel::ThreadState::Initialized:
+        return "Initialized";
+    case Kernel::ThreadState::Waiting:
+        return fmt::format("Waiting ({})", GetThreadWaitReason(thread));
+    case Kernel::ThreadState::Runnable:
+        return "Runnable";
+    case Kernel::ThreadState::Terminated:
+        return "Terminated";
+    default:
+        return "Unknown";
+    }
+}
+
+Kernel::KProcessAddress GetModuleEnd(const Kernel::KProcess* process,
+                                     Kernel::KProcessAddress base) {
+    Kernel::KMemoryInfo mem_info;
+    Kernel::Svc::MemoryInfo svc_mem_info;
+    Kernel::Svc::PageInfo page_info;
+    VAddr cur_addr{GetInteger(base)};
+    auto& page_table = process->GetPageTable();
+
+    // Expect: r-x Code (.text)
+    R_ASSERT(page_table.QueryInfo(std::addressof(mem_info), std::addressof(page_info), cur_addr));
+    svc_mem_info = mem_info.GetSvcMemoryInfo();
+    cur_addr = svc_mem_info.base_address + svc_mem_info.size;
+    if (svc_mem_info.state != Kernel::Svc::MemoryState::Code ||
+        svc_mem_info.permission != Kernel::Svc::MemoryPermission::ReadExecute) {
+        return cur_addr - 1;
+    }
+
+    // Expect: r-- Code (.rodata)
+    R_ASSERT(page_table.QueryInfo(std::addressof(mem_info), std::addressof(page_info), cur_addr));
+    svc_mem_info = mem_info.GetSvcMemoryInfo();
+    cur_addr = svc_mem_info.base_address + svc_mem_info.size;
+    if (svc_mem_info.state != Kernel::Svc::MemoryState::Code ||
+        svc_mem_info.permission != Kernel::Svc::MemoryPermission::Read) {
+        return cur_addr - 1;
+    }
+
+    // Expect: rw- CodeData (.data)
+    R_ASSERT(page_table.QueryInfo(std::addressof(mem_info), std::addressof(page_info), cur_addr));
+    svc_mem_info = mem_info.GetSvcMemoryInfo();
+    cur_addr = svc_mem_info.base_address + svc_mem_info.size;
+    return cur_addr - 1;
+}
+
+Loader::AppLoader::Modules FindModules(const Kernel::KProcess* process) {
+    Loader::AppLoader::Modules modules;
+
+    auto& page_table = process->GetPageTable();
+    auto& memory = process->GetMemory();
+    VAddr cur_addr = 0;
+
+    // Look for executable sections in Code or AliasCode regions.
+    while (true) {
+        Kernel::KMemoryInfo mem_info{};
+        Kernel::Svc::PageInfo page_info{};
+        R_ASSERT(
+            page_table.QueryInfo(std::addressof(mem_info), std::addressof(page_info), cur_addr));
+        auto svc_mem_info = mem_info.GetSvcMemoryInfo();
+
+        if (svc_mem_info.permission == Kernel::Svc::MemoryPermission::ReadExecute &&
+            (svc_mem_info.state == Kernel::Svc::MemoryState::Code ||
+             svc_mem_info.state == Kernel::Svc::MemoryState::AliasCode)) {
+            // Try to read the module name from its path.
+            constexpr s32 PathLengthMax = 0x200;
+            struct {
+                u32 zero;
+                s32 path_length;
+                std::array<char, PathLengthMax> path;
+            } module_path;
+
+            if (memory.ReadBlock(svc_mem_info.base_address + svc_mem_info.size, &module_path,
+                                 sizeof(module_path))) {
+                if (module_path.zero == 0 && module_path.path_length > 0) {
+                    // Truncate module name.
+                    module_path.path[PathLengthMax - 1] = '\0';
+
+                    // Ignore leading directories.
+                    char* path_pointer = module_path.path.data();
+                    char* path_end =
+                        path_pointer + std::min(PathLengthMax, module_path.path_length);
+
+                    for (s32 i = 0; i < std::min(PathLengthMax, module_path.path_length) &&
+                                    module_path.path[i] != '\0';
+                         i++) {
+                        if (module_path.path[i] == '/' || module_path.path[i] == '\\') {
+                            path_pointer = module_path.path.data() + i + 1;
+                        }
+                    }
+
+                    // Insert output.
+                    modules.emplace(svc_mem_info.base_address,
+                                    std::string_view(path_pointer, path_end));
+                }
+            }
+        }
+
+        // Check if we're done.
+        const uintptr_t next_address = svc_mem_info.base_address + svc_mem_info.size;
+        if (next_address <= cur_addr) {
+            break;
+        }
+
+        cur_addr = next_address;
+    }
+
+    return modules;
+}
+
+Kernel::KProcessAddress FindMainModuleEntrypoint(const Kernel::KProcess* process) {
+    // Do we have any loaded executable sections?
+    auto modules = FindModules(process);
+
+    if (modules.size() >= 2) {
+        // If we have two or more, the first one is rtld and the second is main.
+        return std::next(modules.begin())->first;
+    } else if (!modules.empty()) {
+        // If we only have one, this is the main module.
+        return modules.begin()->first;
+    }
+
+    // As a last resort, use the start of the code region.
+    return GetInteger(process->GetPageTable().GetCodeRegionStart());
+}
+
+void InvalidateInstructionCacheRange(const Kernel::KProcess* process, u64 address, u64 size) {
+    for (size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
+        auto* interface = process->GetArmInterface(i);
+        if (interface) {
+            interface->InvalidateCacheRange(address, size);
+        }
+    }
+}
+
+std::vector<BacktraceEntry> GetBacktraceFromContext(const Kernel::KProcess* process,
+                                                    const Kernel::Svc::ThreadContext& ctx) {
+    if (process->Is64Bit()) {
+        return GetAArch64Backtrace(process, ctx);
+    } else {
+        return GetAArch32Backtrace(process, ctx);
+    }
+}
+
+std::vector<BacktraceEntry> GetBacktrace(const Kernel::KThread* thread) {
+    Kernel::Svc::ThreadContext ctx = thread->GetContext();
+    return GetBacktraceFromContext(thread->GetOwnerProcess(), ctx);
+}
+
+} // namespace Core

src/core/arm/debug.h

@@ -0,0 +1,35 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <optional>
+
+#include "core/hle/kernel/k_thread.h"
+#include "core/loader/loader.h"
+
+namespace Core {
+
+std::optional<std::string> GetThreadName(const Kernel::KThread* thread);
+std::string_view GetThreadWaitReason(const Kernel::KThread* thread);
+std::string GetThreadState(const Kernel::KThread* thread);
+
+Loader::AppLoader::Modules FindModules(const Kernel::KProcess* process);
+Kernel::KProcessAddress GetModuleEnd(const Kernel::KProcess* process, Kernel::KProcessAddress base);
+Kernel::KProcessAddress FindMainModuleEntrypoint(const Kernel::KProcess* process);
+
+void InvalidateInstructionCacheRange(const Kernel::KProcess* process, u64 address, u64 size);
+
+struct BacktraceEntry {
+    std::string module;
+    u64 address;
+    u64 original_address;
+    u64 offset;
+    std::string name;
+};
+
+std::vector<BacktraceEntry> GetBacktraceFromContext(const Kernel::KProcess* process,
+                                                    const Kernel::Svc::ThreadContext& ctx);
+std::vector<BacktraceEntry> GetBacktrace(const Kernel::KThread* thread);
+
+} // namespace Core

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -1,25 +1,13 @@
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include <cinttypes>
-#include <memory>
-#include <dynarmic/interface/A32/a32.h>
-#include <dynarmic/interface/A32/config.h>
-#include "common/assert.h"
-#include "common/literals.h"
-#include "common/logging/log.h"
-#include "common/page_table.h"
 #include "common/settings.h"
 #include "core/arm/dynarmic/arm_dynarmic.h"
 #include "core/arm/dynarmic/arm_dynarmic_32.h"
 #include "core/arm/dynarmic/dynarmic_cp15.h"
 #include "core/arm/dynarmic/dynarmic_exclusive_monitor.h"
-#include "core/core.h"
 #include "core/core_timing.h"
-#include "core/debugger/debugger.h"
 #include "core/hle/kernel/k_process.h"
-#include "core/hle/kernel/svc.h"
-#include "core/memory.h"
 
 namespace Core {
 
@@ -27,78 +15,78 @@ using namespace Common::Literals;
 
 class DynarmicCallbacks32 : public Dynarmic::A32::UserCallbacks {
 public:
-    explicit DynarmicCallbacks32(ARM_Dynarmic_32& parent_)
-        : parent{parent_}, memory(parent.system.ApplicationMemory()),
-          debugger_enabled{parent.system.DebuggerEnabled()},
-          check_memory_access{debugger_enabled ||
-                              !Settings::values.cpuopt_ignore_memory_aborts.GetValue()} {}
+    explicit DynarmicCallbacks32(ArmDynarmic32& parent, const Kernel::KProcess* process)
+        : m_parent{parent}, m_memory(process->GetMemory()),
+          m_process(process), m_debugger_enabled{parent.m_system.DebuggerEnabled()},
+          m_check_memory_access{m_debugger_enabled ||
+                                !Settings::values.cpuopt_ignore_memory_aborts.GetValue()} {}
 
     u8 MemoryRead8(u32 vaddr) override {
         CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Read);
-        return memory.Read8(vaddr);
+        return m_memory.Read8(vaddr);
     }
     u16 MemoryRead16(u32 vaddr) override {
         CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Read);
-        return memory.Read16(vaddr);
+        return m_memory.Read16(vaddr);
     }
     u32 MemoryRead32(u32 vaddr) override {
         CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Read);
-        return memory.Read32(vaddr);
+        return m_memory.Read32(vaddr);
     }
     u64 MemoryRead64(u32 vaddr) override {
         CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Read);
-        return memory.Read64(vaddr);
+        return m_memory.Read64(vaddr);
     }
     std::optional<u32> MemoryReadCode(u32 vaddr) override {
-        if (!memory.IsValidVirtualAddressRange(vaddr, sizeof(u32))) {
+        if (!m_memory.IsValidVirtualAddressRange(vaddr, sizeof(u32))) {
             return std::nullopt;
         }
-        return memory.Read32(vaddr);
+        return m_memory.Read32(vaddr);
     }
 
     void MemoryWrite8(u32 vaddr, u8 value) override {
         if (CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write)) {
-            memory.Write8(vaddr, value);
+            m_memory.Write8(vaddr, value);
         }
     }
     void MemoryWrite16(u32 vaddr, u16 value) override {
         if (CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Write)) {
-            memory.Write16(vaddr, value);
+            m_memory.Write16(vaddr, value);
         }
     }
     void MemoryWrite32(u32 vaddr, u32 value) override {
         if (CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Write)) {
-            memory.Write32(vaddr, value);
+            m_memory.Write32(vaddr, value);
         }
     }
     void MemoryWrite64(u32 vaddr, u64 value) override {
         if (CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Write)) {
-            memory.Write64(vaddr, value);
+            m_memory.Write64(vaddr, value);
         }
     }
 
     bool MemoryWriteExclusive8(u32 vaddr, u8 value, u8 expected) override {
         return CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write) &&
-               memory.WriteExclusive8(vaddr, value, expected);
+               m_memory.WriteExclusive8(vaddr, value, expected);
     }
     bool MemoryWriteExclusive16(u32 vaddr, u16 value, u16 expected) override {
         return CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Write) &&
-               memory.WriteExclusive16(vaddr, value, expected);
+               m_memory.WriteExclusive16(vaddr, value, expected);
     }
     bool MemoryWriteExclusive32(u32 vaddr, u32 value, u32 expected) override {
         return CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Write) &&
-               memory.WriteExclusive32(vaddr, value, expected);
+               m_memory.WriteExclusive32(vaddr, value, expected);
     }
     bool MemoryWriteExclusive64(u32 vaddr, u64 value, u64 expected) override {
         return CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Write) &&
-               memory.WriteExclusive64(vaddr, value, expected);
+               m_memory.WriteExclusive64(vaddr, value, expected);
     }
 
     void InterpreterFallback(u32 pc, std::size_t num_instructions) override {
-        parent.LogBacktrace();
+        m_parent.LogBacktrace(m_process);
         LOG_ERROR(Core_ARM,
                   "Unimplemented instruction @ 0x{:X} for {} instructions (instr = {:08X})", pc,
-                  num_instructions, memory.Read32(pc));
+                  num_instructions, m_memory.Read32(pc));
     }
 
     void ExceptionRaised(u32 pc, Dynarmic::A32::Exception exception) override {
@@ -108,73 +96,64 @@ public:
             ReturnException(pc, PrefetchAbort);
             return;
         default:
-            if (debugger_enabled) {
+            if (m_debugger_enabled) {
                 ReturnException(pc, InstructionBreakpoint);
                 return;
             }
 
-            parent.LogBacktrace();
+            m_parent.LogBacktrace(m_process);
             LOG_CRITICAL(Core_ARM,
                          "ExceptionRaised(exception = {}, pc = {:08X}, code = {:08X}, thumb = {})",
-                         exception, pc, memory.Read32(pc), parent.IsInThumbMode());
+                         exception, pc, m_memory.Read32(pc), m_parent.IsInThumbMode());
         }
     }
 
     void CallSVC(u32 swi) override {
-        parent.svc_swi = swi;
-        parent.jit.load()->HaltExecution(SupervisorCall);
+        m_parent.m_svc_swi = swi;
+        m_parent.m_jit->HaltExecution(SupervisorCall);
     }
 
     void AddTicks(u64 ticks) override {
-        if (parent.uses_wall_clock) {
-            return;
-        }
+        ASSERT_MSG(!m_parent.m_uses_wall_clock, "Dynarmic ticking disabled");
 
         // Divide the number of ticks by the amount of CPU cores. TODO(Subv): This yields only a
         // rough approximation of the amount of executed ticks in the system, it may be thrown off
         // if not all cores are doing a similar amount of work. Instead of doing this, we should
         // device a way so that timing is consistent across all cores without increasing the ticks 4
         // times.
-        u64 amortized_ticks =
-            (ticks - num_interpreted_instructions) / Core::Hardware::NUM_CPU_CORES;
+        u64 amortized_ticks = ticks / Core::Hardware::NUM_CPU_CORES;
         // Always execute at least one tick.
         amortized_ticks = std::max<u64>(amortized_ticks, 1);
 
-        parent.system.CoreTiming().AddTicks(amortized_ticks);
-        num_interpreted_instructions = 0;
+        m_parent.m_system.CoreTiming().AddTicks(amortized_ticks);
     }
 
     u64 GetTicksRemaining() override {
-        if (parent.uses_wall_clock) {
-            if (!IsInterrupted()) {
-                return minimum_run_cycles;
-            }
-            return 0U;
-        }
+        ASSERT_MSG(!m_parent.m_uses_wall_clock, "Dynarmic ticking disabled");
 
-        return std::max<s64>(parent.system.CoreTiming().GetDowncount(), 0);
+        return std::max<s64>(m_parent.m_system.CoreTiming().GetDowncount(), 0);
     }
 
     bool CheckMemoryAccess(u64 addr, u64 size, Kernel::DebugWatchpointType type) {
-        if (!check_memory_access) {
+        if (!m_check_memory_access) {
             return true;
         }
 
-        if (!memory.IsValidVirtualAddressRange(addr, size)) {
+        if (!m_memory.IsValidVirtualAddressRange(addr, size)) {
             LOG_CRITICAL(Core_ARM, "Stopping execution due to unmapped memory access at {:#x}",
                          addr);
-            parent.jit.load()->HaltExecution(PrefetchAbort);
+            m_parent.m_jit->HaltExecution(PrefetchAbort);
             return false;
         }
 
-        if (!debugger_enabled) {
+        if (!m_debugger_enabled) {
             return true;
         }
 
-        const auto match{parent.MatchingWatchpoint(addr, size, type)};
+        const auto match{m_parent.MatchingWatchpoint(addr, size, type)};
         if (match) {
-            parent.halted_watchpoint = match;
-            parent.jit.load()->HaltExecution(DataAbort);
+            m_parent.m_halted_watchpoint = match;
+            m_parent.m_jit->HaltExecution(DataAbort);
             return false;
         }
 
@@ -182,32 +161,31 @@ public:
     }
 
     void ReturnException(u32 pc, Dynarmic::HaltReason hr) {
-        parent.SaveContext(parent.breakpoint_context);
-        parent.breakpoint_context.cpu_registers[15] = pc;
-        parent.jit.load()->HaltExecution(hr);
+        m_parent.GetContext(m_parent.m_breakpoint_context);
+        m_parent.m_breakpoint_context.pc = pc;
+        m_parent.m_breakpoint_context.r[15] = pc;
+        m_parent.m_jit->HaltExecution(hr);
     }
 
-    bool IsInterrupted() {
-        return parent.system.Kernel().PhysicalCore(parent.core_index).IsInterrupted();
-    }
-
-    ARM_Dynarmic_32& parent;
-    Core::Memory::Memory& memory;
-    std::size_t num_interpreted_instructions{};
-    const bool debugger_enabled{};
-    const bool check_memory_access{};
-    static constexpr u64 minimum_run_cycles = 10000U;
+    ArmDynarmic32& m_parent;
+    Core::Memory::Memory& m_memory;
+    const Kernel::KProcess* m_process{};
+    const bool m_debugger_enabled{};
+    const bool m_check_memory_access{};
+    static constexpr u64 MinimumRunCycles = 10000U;
 };
 
-std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable* page_table) const {
+std::shared_ptr<Dynarmic::A32::Jit> ArmDynarmic32::MakeJit(Common::PageTable* page_table) const {
     Dynarmic::A32::UserConfig config;
-    config.callbacks = cb.get();
-    config.coprocessors[15] = cp15;
+    config.callbacks = m_cb.get();
+    config.coprocessors[15] = m_cp15;
     config.define_unpredictable_behaviour = true;
-    static constexpr std::size_t YUZU_PAGEBITS = 12;
-    static constexpr std::size_t NUM_PAGE_TABLE_ENTRIES = 1 << (32 - YUZU_PAGEBITS);
+
     if (page_table) {
-        config.page_table = reinterpret_cast<std::array<std::uint8_t*, NUM_PAGE_TABLE_ENTRIES>*>(
+        constexpr size_t PageBits = 12;
+        constexpr size_t NumPageTableEntries = 1 << (32 - PageBits);
+
+        config.page_table = reinterpret_cast<std::array<std::uint8_t*, NumPageTableEntries>*>(
             page_table->pointers.data());
         config.absolute_offset_page_table = true;
         config.page_table_pointer_mask_bits = Common::PageTable::ATTRIBUTE_BITS;
@@ -221,12 +199,12 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable*
     }
 
     // Multi-process state
-    config.processor_id = core_index;
-    config.global_monitor = &exclusive_monitor.monitor;
+    config.processor_id = m_core_index;
+    config.global_monitor = &m_exclusive_monitor.monitor;
 
     // Timing
-    config.wall_clock_cntpct = uses_wall_clock;
-    config.enable_cycle_counting = true;
+    config.wall_clock_cntpct = m_uses_wall_clock;
+    config.enable_cycle_counting = !m_uses_wall_clock;
 
     // Code cache size
 #ifdef ARCHITECTURE_arm64
@@ -236,7 +214,7 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable*
 #endif
 
     // Allow memory fault handling to work
-    if (system.DebuggerEnabled()) {
+    if (m_system.DebuggerEnabled()) {
         config.check_halt_on_memory_access = true;
     }
 
@@ -325,137 +303,140 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable*
     return std::make_unique<Dynarmic::A32::Jit>(config);
 }
 
-HaltReason ARM_Dynarmic_32::RunJit() {
-    return TranslateHaltReason(jit.load()->Run());
+static std::pair<u32, u32> FpscrToFpsrFpcr(u32 fpscr) {
+    // FPSCR bits [31:27] are mapped to FPSR[31:27].
+    // FPSCR bit [7] is mapped to FPSR[7].
+    // FPSCR bits [4:0] are mapped to FPSR[4:0].
+    const u32 nzcv = fpscr & 0xf8000000;
+    const u32 idc = fpscr & 0x80;
+    const u32 fiq = fpscr & 0x1f;
+    const u32 fpsr = nzcv | idc | fiq;
+
+    // FPSCR bits [26:15] are mapped to FPCR[26:15].
+    // FPSCR bits [12:8] are mapped to FPCR[12:8].
+    const u32 round = fpscr & 0x7ff8000;
+    const u32 trap = fpscr & 0x1f00;
+    const u32 fpcr = round | trap;
+
+    return {fpsr, fpcr};
 }
 
-HaltReason ARM_Dynarmic_32::StepJit() {
-    return TranslateHaltReason(jit.load()->Step());
+static u32 FpsrFpcrToFpscr(u64 fpsr, u64 fpcr) {
+    auto [s, c] = FpscrToFpsrFpcr(static_cast<u32>(fpsr | fpcr));
+    return s | c;
 }
 
-u32 ARM_Dynarmic_32::GetSvcNumber() const {
-    return svc_swi;
+bool ArmDynarmic32::IsInThumbMode() const {
+    return (m_jit->Cpsr() & 0x20) != 0;
 }
 
-const Kernel::DebugWatchpoint* ARM_Dynarmic_32::HaltedWatchpoint() const {
-    return halted_watchpoint;
+HaltReason ArmDynarmic32::RunThread(Kernel::KThread* thread) {
+    m_jit->ClearExclusiveState();
+    return TranslateHaltReason(m_jit->Run());
 }
 
-void ARM_Dynarmic_32::RewindBreakpointInstruction() {
-    LoadContext(breakpoint_context);
+HaltReason ArmDynarmic32::StepThread(Kernel::KThread* thread) {
+    m_jit->ClearExclusiveState();
+    return TranslateHaltReason(m_jit->Step());
 }
 
-ARM_Dynarmic_32::ARM_Dynarmic_32(System& system_, bool uses_wall_clock_,
-                                 DynarmicExclusiveMonitor& exclusive_monitor_,
-                                 std::size_t core_index_)
-    : ARM_Interface{system_, uses_wall_clock_}, cb(std::make_unique<DynarmicCallbacks32>(*this)),
-      cp15(std::make_shared<DynarmicCP15>(*this)), core_index{core_index_},
-      exclusive_monitor{exclusive_monitor_}, null_jit{MakeJit(nullptr)}, jit{null_jit.get()} {}
-
-ARM_Dynarmic_32::~ARM_Dynarmic_32() = default;
-
-void ARM_Dynarmic_32::SetPC(u64 pc) {
-    jit.load()->Regs()[15] = static_cast<u32>(pc);
+u32 ArmDynarmic32::GetSvcNumber() const {
+    return m_svc_swi;
 }
 
-u64 ARM_Dynarmic_32::GetPC() const {
-    return jit.load()->Regs()[15];
-}
+void ArmDynarmic32::GetSvcArguments(std::span<uint64_t, 8> args) const {
+    Dynarmic::A32::Jit& j = *m_jit;
+    auto& gpr = j.Regs();
 
-u64 ARM_Dynarmic_32::GetSP() const {
-    return jit.load()->Regs()[13];
-}
-
-u64 ARM_Dynarmic_32::GetReg(int index) const {
-    return jit.load()->Regs()[index];
-}
-
-void ARM_Dynarmic_32::SetReg(int index, u64 value) {
-    jit.load()->Regs()[index] = static_cast<u32>(value);
-}
-
-u128 ARM_Dynarmic_32::GetVectorReg(int index) const {
-    return {};
-}
-
-void ARM_Dynarmic_32::SetVectorReg(int index, u128 value) {}
-
-u32 ARM_Dynarmic_32::GetPSTATE() const {
-    return jit.load()->Cpsr();
-}
-
-void ARM_Dynarmic_32::SetPSTATE(u32 cpsr) {
-    jit.load()->SetCpsr(cpsr);
-}
-
-u64 ARM_Dynarmic_32::GetTlsAddress() const {
-    return cp15->uro;
-}
-
-void ARM_Dynarmic_32::SetTlsAddress(u64 address) {
-    cp15->uro = static_cast<u32>(address);
-}
-
-u64 ARM_Dynarmic_32::GetTPIDR_EL0() const {
-    return cp15->uprw;
-}
-
-void ARM_Dynarmic_32::SetTPIDR_EL0(u64 value) {
-    cp15->uprw = static_cast<u32>(value);
-}
-
-void ARM_Dynarmic_32::SaveContext(ThreadContext32& ctx) const {
-    Dynarmic::A32::Jit* j = jit.load();
-    ctx.cpu_registers = j->Regs();
-    ctx.extension_registers = j->ExtRegs();
-    ctx.cpsr = j->Cpsr();
-    ctx.fpscr = j->Fpscr();
-}
-
-void ARM_Dynarmic_32::LoadContext(const ThreadContext32& ctx) {
-    Dynarmic::A32::Jit* j = jit.load();
-    j->Regs() = ctx.cpu_registers;
-    j->ExtRegs() = ctx.extension_registers;
-    j->SetCpsr(ctx.cpsr);
-    j->SetFpscr(ctx.fpscr);
-}
-
-void ARM_Dynarmic_32::SignalInterrupt() {
-    jit.load()->HaltExecution(BreakLoop);
-}
-
-void ARM_Dynarmic_32::ClearInterrupt() {
-    jit.load()->ClearHalt(BreakLoop);
-}
-
-void ARM_Dynarmic_32::ClearInstructionCache() {
-    jit.load()->ClearCache();
-}
-
-void ARM_Dynarmic_32::InvalidateCacheRange(u64 addr, std::size_t size) {
-    jit.load()->InvalidateCacheRange(static_cast<u32>(addr), size);
-}
-
-void ARM_Dynarmic_32::ClearExclusiveState() {
-    jit.load()->ClearExclusiveState();
-}
-
-void ARM_Dynarmic_32::PageTableChanged(Common::PageTable& page_table,
-                                       std::size_t new_address_space_size_in_bits) {
-    ThreadContext32 ctx{};
-    SaveContext(ctx);
-
-    auto key = std::make_pair(&page_table, new_address_space_size_in_bits);
-    auto iter = jit_cache.find(key);
-    if (iter != jit_cache.end()) {
-        jit.store(iter->second.get());
-        LoadContext(ctx);
-        return;
+    for (size_t i = 0; i < 8; i++) {
+        args[i] = gpr[i];
     }
-    std::shared_ptr new_jit = MakeJit(&page_table);
-    jit.store(new_jit.get());
-    LoadContext(ctx);
-    jit_cache.emplace(key, std::move(new_jit));
+}
+
+void ArmDynarmic32::SetSvcArguments(std::span<const uint64_t, 8> args) {
+    Dynarmic::A32::Jit& j = *m_jit;
+    auto& gpr = j.Regs();
+
+    for (size_t i = 0; i < 8; i++) {
+        gpr[i] = static_cast<u32>(args[i]);
+    }
+}
+
+const Kernel::DebugWatchpoint* ArmDynarmic32::HaltedWatchpoint() const {
+    return m_halted_watchpoint;
+}
+
+void ArmDynarmic32::RewindBreakpointInstruction() {
+    this->SetContext(m_breakpoint_context);
+}
+
+ArmDynarmic32::ArmDynarmic32(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+                             DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index)
+    : ArmInterface{uses_wall_clock}, m_system{system}, m_exclusive_monitor{exclusive_monitor},
+      m_cb(std::make_unique<DynarmicCallbacks32>(*this, process)),
+      m_cp15(std::make_shared<DynarmicCP15>(*this)), m_core_index{core_index} {
+    auto& page_table_impl = process->GetPageTable().GetBasePageTable().GetImpl();
+    m_jit = MakeJit(&page_table_impl);
+}
+
+ArmDynarmic32::~ArmDynarmic32() = default;
+
+void ArmDynarmic32::SetTpidrroEl0(u64 value) {
+    m_cp15->uro = static_cast<u32>(value);
+}
+
+void ArmDynarmic32::GetContext(Kernel::Svc::ThreadContext& ctx) const {
+    Dynarmic::A32::Jit& j = *m_jit;
+    auto& gpr = j.Regs();
+    auto& fpr = j.ExtRegs();
+
+    for (size_t i = 0; i < 16; i++) {
+        ctx.r[i] = gpr[i];
+    }
+
+    ctx.fp = gpr[11];
+    ctx.sp = gpr[13];
+    ctx.lr = gpr[14];
+    ctx.pc = gpr[15];
+    ctx.pstate = j.Cpsr();
+
+    static_assert(sizeof(fpr) <= sizeof(ctx.v));
+    std::memcpy(ctx.v.data(), &fpr, sizeof(fpr));
+
+    auto [fpsr, fpcr] = FpscrToFpsrFpcr(j.Fpscr());
+    ctx.fpcr = fpcr;
+    ctx.fpsr = fpsr;
+    ctx.tpidr = m_cp15->uprw;
+}
+
+void ArmDynarmic32::SetContext(const Kernel::Svc::ThreadContext& ctx) {
+    Dynarmic::A32::Jit& j = *m_jit;
+    auto& gpr = j.Regs();
+    auto& fpr = j.ExtRegs();
+
+    for (size_t i = 0; i < 16; i++) {
+        gpr[i] = static_cast<u32>(ctx.r[i]);
+    }
+
+    j.SetCpsr(ctx.pstate);
+
+    static_assert(sizeof(fpr) <= sizeof(ctx.v));
+    std::memcpy(&fpr, ctx.v.data(), sizeof(fpr));
+
+    j.SetFpscr(FpsrFpcrToFpscr(ctx.fpsr, ctx.fpcr));
+    m_cp15->uprw = static_cast<u32>(ctx.tpidr);
+}
+
+void ArmDynarmic32::SignalInterrupt(Kernel::KThread* thread) {
+    m_jit->HaltExecution(BreakLoop);
+}
+
+void ArmDynarmic32::ClearInstructionCache() {
+    m_jit->ClearCache();
+}
+
+void ArmDynarmic32::InvalidateCacheRange(u64 addr, std::size_t size) {
+    m_jit->InvalidateCacheRange(static_cast<u32>(addr), size);
 }
 
 } // namespace Core

src/core/arm/dynarmic/arm_dynarmic_32.h

@@ -3,14 +3,8 @@
 
 #pragma once
 
-#include <atomic>
-#include <memory>
-#include <unordered_map>
-
 #include <dynarmic/interface/A32/a32.h>
-#include <dynarmic/interface/A64/a64.h>
-#include "common/common_types.h"
-#include "common/hash.h"
+
 #include "core/arm/arm_interface.h"
 #include "core/arm/dynarmic/dynarmic_exclusive_monitor.h"
 
@@ -20,89 +14,63 @@ class Memory;
 
 namespace Core {
 
-class CPUInterruptHandler;
 class DynarmicCallbacks32;
 class DynarmicCP15;
-class DynarmicExclusiveMonitor;
 class System;
 
-class ARM_Dynarmic_32 final : public ARM_Interface {
+class ArmDynarmic32 final : public ArmInterface {
 public:
-    ARM_Dynarmic_32(System& system_, bool uses_wall_clock_,
-                    DynarmicExclusiveMonitor& exclusive_monitor_, std::size_t core_index_);
-    ~ARM_Dynarmic_32() override;
-
-    void SetPC(u64 pc) override;
-    u64 GetPC() const override;
-    u64 GetSP() const override;
-    u64 GetReg(int index) const override;
-    void SetReg(int index, u64 value) override;
-    u128 GetVectorReg(int index) const override;
-    void SetVectorReg(int index, u128 value) override;
-    u32 GetPSTATE() const override;
-    void SetPSTATE(u32 pstate) override;
-    u64 GetTlsAddress() const override;
-    void SetTlsAddress(u64 address) override;
-    void SetTPIDR_EL0(u64 value) override;
-    u64 GetTPIDR_EL0() const override;
-
-    bool IsInThumbMode() const {
-        return (GetPSTATE() & 0x20) != 0;
-    }
+    ArmDynarmic32(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+                  DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index);
+    ~ArmDynarmic32() override;
 
     Architecture GetArchitecture() const override {
-        return Architecture::Aarch32;
+        return Architecture::AArch32;
     }
-    void SaveContext(ThreadContext32& ctx) const override;
-    void SaveContext(ThreadContext64& ctx) const override {}
-    void LoadContext(const ThreadContext32& ctx) override;
-    void LoadContext(const ThreadContext64& ctx) override {}
 
-    void SignalInterrupt() override;
-    void ClearInterrupt() override;
-    void ClearExclusiveState() override;
+    bool IsInThumbMode() const;
 
+    HaltReason RunThread(Kernel::KThread* thread) override;
+    HaltReason StepThread(Kernel::KThread* thread) override;
+
+    void GetContext(Kernel::Svc::ThreadContext& ctx) const override;
+    void SetContext(const Kernel::Svc::ThreadContext& ctx) override;
+    void SetTpidrroEl0(u64 value) override;
+
+    void GetSvcArguments(std::span<uint64_t, 8> args) const override;
+    void SetSvcArguments(std::span<const uint64_t, 8> args) override;
+    u32 GetSvcNumber() const override;
+
+    void SignalInterrupt(Kernel::KThread* thread) override;
     void ClearInstructionCache() override;
     void InvalidateCacheRange(u64 addr, std::size_t size) override;
-    void PageTableChanged(Common::PageTable& new_page_table,
-                          std::size_t new_address_space_size_in_bits) override;
 
 protected:
-    HaltReason RunJit() override;
-    HaltReason StepJit() override;
-    u32 GetSvcNumber() const override;
     const Kernel::DebugWatchpoint* HaltedWatchpoint() const override;
     void RewindBreakpointInstruction() override;
 
 private:
-    std::shared_ptr<Dynarmic::A32::Jit> MakeJit(Common::PageTable* page_table) const;
-
-    static std::vector<BacktraceEntry> GetBacktrace(Core::System& system, u64 fp, u64 lr, u64 pc);
-
-    using JitCacheKey = std::pair<Common::PageTable*, std::size_t>;
-    using JitCacheType =
-        std::unordered_map<JitCacheKey, std::shared_ptr<Dynarmic::A32::Jit>, Common::PairHash>;
+    System& m_system;
+    DynarmicExclusiveMonitor& m_exclusive_monitor;
 
+private:
     friend class DynarmicCallbacks32;
     friend class DynarmicCP15;
 
-    std::unique_ptr<DynarmicCallbacks32> cb;
-    JitCacheType jit_cache;
-    std::shared_ptr<DynarmicCP15> cp15;
-    std::size_t core_index;
-    DynarmicExclusiveMonitor& exclusive_monitor;
+    std::shared_ptr<Dynarmic::A32::Jit> MakeJit(Common::PageTable* page_table) const;
 
-    std::shared_ptr<Dynarmic::A32::Jit> null_jit;
+    std::unique_ptr<DynarmicCallbacks32> m_cb{};
+    std::shared_ptr<DynarmicCP15> m_cp15{};
+    std::size_t m_core_index{};
 
-    // A raw pointer here is fine; we never delete Jit instances.
-    std::atomic<Dynarmic::A32::Jit*> jit;
+    std::shared_ptr<Dynarmic::A32::Jit> m_jit{};
 
     // SVC callback
-    u32 svc_swi{};
+    u32 m_svc_swi{};
 
     // Watchpoint info
-    const Kernel::DebugWatchpoint* halted_watchpoint;
-    ThreadContext32 breakpoint_context;
+    const Kernel::DebugWatchpoint* m_halted_watchpoint{};
+    Kernel::Svc::ThreadContext m_breakpoint_context{};
 };
 
 } // namespace Core

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -1,25 +1,12 @@
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include <cinttypes>
-#include <memory>
-#include <dynarmic/interface/A64/a64.h>
-#include <dynarmic/interface/A64/config.h>
-#include "common/assert.h"
-#include "common/literals.h"
-#include "common/logging/log.h"
-#include "common/page_table.h"
 #include "common/settings.h"
 #include "core/arm/dynarmic/arm_dynarmic.h"
 #include "core/arm/dynarmic/arm_dynarmic_64.h"
 #include "core/arm/dynarmic/dynarmic_exclusive_monitor.h"
-#include "core/core.h"
 #include "core/core_timing.h"
-#include "core/debugger/debugger.h"
-#include "core/hardware_properties.h"
 #include "core/hle/kernel/k_process.h"
-#include "core/hle/kernel/svc.h"
-#include "core/memory.h"
 
 namespace Core {
 
@@ -28,92 +15,92 @@ using namespace Common::Literals;
 
 class DynarmicCallbacks64 : public Dynarmic::A64::UserCallbacks {
 public:
-    explicit DynarmicCallbacks64(ARM_Dynarmic_64& parent_)
-        : parent{parent_}, memory(parent.system.ApplicationMemory()),
-          debugger_enabled{parent.system.DebuggerEnabled()},
-          check_memory_access{debugger_enabled ||
-                              !Settings::values.cpuopt_ignore_memory_aborts.GetValue()} {}
+    explicit DynarmicCallbacks64(ArmDynarmic64& parent, const Kernel::KProcess* process)
+        : m_parent{parent}, m_memory(process->GetMemory()),
+          m_process(process), m_debugger_enabled{parent.m_system.DebuggerEnabled()},
+          m_check_memory_access{m_debugger_enabled ||
+                                !Settings::values.cpuopt_ignore_memory_aborts.GetValue()} {}
 
     u8 MemoryRead8(u64 vaddr) override {
         CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Read);
-        return memory.Read8(vaddr);
+        return m_memory.Read8(vaddr);
     }
     u16 MemoryRead16(u64 vaddr) override {
         CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Read);
-        return memory.Read16(vaddr);
+        return m_memory.Read16(vaddr);
     }
     u32 MemoryRead32(u64 vaddr) override {
         CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Read);
-        return memory.Read32(vaddr);
+        return m_memory.Read32(vaddr);
     }
     u64 MemoryRead64(u64 vaddr) override {
         CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Read);
-        return memory.Read64(vaddr);
+        return m_memory.Read64(vaddr);
     }
     Vector MemoryRead128(u64 vaddr) override {
         CheckMemoryAccess(vaddr, 16, Kernel::DebugWatchpointType::Read);
-        return {memory.Read64(vaddr), memory.Read64(vaddr + 8)};
+        return {m_memory.Read64(vaddr), m_memory.Read64(vaddr + 8)};
     }
     std::optional<u32> MemoryReadCode(u64 vaddr) override {
-        if (!memory.IsValidVirtualAddressRange(vaddr, sizeof(u32))) {
+        if (!m_memory.IsValidVirtualAddressRange(vaddr, sizeof(u32))) {
             return std::nullopt;
         }
-        return memory.Read32(vaddr);
+        return m_memory.Read32(vaddr);
     }
 
     void MemoryWrite8(u64 vaddr, u8 value) override {
         if (CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write)) {
-            memory.Write8(vaddr, value);
+            m_memory.Write8(vaddr, value);
         }
     }
     void MemoryWrite16(u64 vaddr, u16 value) override {
         if (CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Write)) {
-            memory.Write16(vaddr, value);
+            m_memory.Write16(vaddr, value);
         }
     }
     void MemoryWrite32(u64 vaddr, u32 value) override {
         if (CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Write)) {
-            memory.Write32(vaddr, value);
+            m_memory.Write32(vaddr, value);
         }
     }
     void MemoryWrite64(u64 vaddr, u64 value) override {
         if (CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Write)) {
-            memory.Write64(vaddr, value);
+            m_memory.Write64(vaddr, value);
         }
     }
     void MemoryWrite128(u64 vaddr, Vector value) override {
         if (CheckMemoryAccess(vaddr, 16, Kernel::DebugWatchpointType::Write)) {
-            memory.Write64(vaddr, value[0]);
-            memory.Write64(vaddr + 8, value[1]);
+            m_memory.Write64(vaddr, value[0]);
+            m_memory.Write64(vaddr + 8, value[1]);
         }
     }
 
     bool MemoryWriteExclusive8(u64 vaddr, std::uint8_t value, std::uint8_t expected) override {
         return CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write) &&
-               memory.WriteExclusive8(vaddr, value, expected);
+               m_memory.WriteExclusive8(vaddr, value, expected);
     }
     bool MemoryWriteExclusive16(u64 vaddr, std::uint16_t value, std::uint16_t expected) override {
         return CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Write) &&
-               memory.WriteExclusive16(vaddr, value, expected);
+               m_memory.WriteExclusive16(vaddr, value, expected);
     }
     bool MemoryWriteExclusive32(u64 vaddr, std::uint32_t value, std::uint32_t expected) override {
         return CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Write) &&
-               memory.WriteExclusive32(vaddr, value, expected);
+               m_memory.WriteExclusive32(vaddr, value, expected);
     }
     bool MemoryWriteExclusive64(u64 vaddr, std::uint64_t value, std::uint64_t expected) override {
         return CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Write) &&
-               memory.WriteExclusive64(vaddr, value, expected);
+               m_memory.WriteExclusive64(vaddr, value, expected);
     }
     bool MemoryWriteExclusive128(u64 vaddr, Vector value, Vector expected) override {
         return CheckMemoryAccess(vaddr, 16, Kernel::DebugWatchpointType::Write) &&
-               memory.WriteExclusive128(vaddr, value, expected);
+               m_memory.WriteExclusive128(vaddr, value, expected);
     }
 
     void InterpreterFallback(u64 pc, std::size_t num_instructions) override {
-        parent.LogBacktrace();
+        m_parent.LogBacktrace(m_process);
         LOG_ERROR(Core_ARM,
                   "Unimplemented instruction @ 0x{:X} for {} instructions (instr = {:08X})", pc,
-                  num_instructions, memory.Read32(pc));
+                  num_instructions, m_memory.Read32(pc));
         ReturnException(pc, PrefetchAbort);
     }
 
@@ -124,11 +111,11 @@ public:
             static constexpr u64 ICACHE_LINE_SIZE = 64;
 
             const u64 cache_line_start = value & ~(ICACHE_LINE_SIZE - 1);
-            parent.system.InvalidateCpuInstructionCacheRange(cache_line_start, ICACHE_LINE_SIZE);
+            m_parent.InvalidateCacheRange(cache_line_start, ICACHE_LINE_SIZE);
             break;
         }
         case Dynarmic::A64::InstructionCacheOperation::InvalidateAllToPoU:
-            parent.system.InvalidateCpuInstructionCaches();
+            m_parent.ClearInstructionCache();
             break;
         case Dynarmic::A64::InstructionCacheOperation::InvalidateAllToPoUInnerSharable:
         default:
@@ -136,7 +123,7 @@ public:
             break;
         }
 
-        parent.jit.load()->HaltExecution(Dynarmic::HaltReason::CacheInvalidation);
+        m_parent.m_jit->HaltExecution(Dynarmic::HaltReason::CacheInvalidation);
     }
 
     void ExceptionRaised(u64 pc, Dynarmic::A64::Exception exception) override {
@@ -152,26 +139,24 @@ public:
             ReturnException(pc, PrefetchAbort);
             return;
         default:
-            if (debugger_enabled) {
+            if (m_debugger_enabled) {
                 ReturnException(pc, InstructionBreakpoint);
                 return;
             }
 
-            parent.LogBacktrace();
+            m_parent.LogBacktrace(m_process);
             LOG_CRITICAL(Core_ARM, "ExceptionRaised(exception = {}, pc = {:08X}, code = {:08X})",
-                         static_cast<std::size_t>(exception), pc, memory.Read32(pc));
+                         static_cast<std::size_t>(exception), pc, m_memory.Read32(pc));
         }
     }
 
-    void CallSVC(u32 swi) override {
-        parent.svc_swi = swi;
-        parent.jit.load()->HaltExecution(SupervisorCall);
+    void CallSVC(u32 svc) override {
+        m_parent.m_svc = svc;
+        m_parent.m_jit->HaltExecution(SupervisorCall);
     }
 
     void AddTicks(u64 ticks) override {
-        if (parent.uses_wall_clock) {
-            return;
-        }
+        ASSERT_MSG(!m_parent.m_uses_wall_clock, "Dynarmic ticking disabled");
 
         // Divide the number of ticks by the amount of CPU cores. TODO(Subv): This yields only a
         // rough approximation of the amount of executed ticks in the system, it may be thrown off
@@ -182,44 +167,39 @@ public:
         // Always execute at least one tick.
         amortized_ticks = std::max<u64>(amortized_ticks, 1);
 
-        parent.system.CoreTiming().AddTicks(amortized_ticks);
+        m_parent.m_system.CoreTiming().AddTicks(amortized_ticks);
     }
 
     u64 GetTicksRemaining() override {
-        if (parent.uses_wall_clock) {
-            if (!IsInterrupted()) {
-                return minimum_run_cycles;
-            }
-            return 0U;
-        }
+        ASSERT_MSG(!m_parent.m_uses_wall_clock, "Dynarmic ticking disabled");
 
-        return std::max<s64>(parent.system.CoreTiming().GetDowncount(), 0);
+        return std::max<s64>(m_parent.m_system.CoreTiming().GetDowncount(), 0);
     }
 
     u64 GetCNTPCT() override {
-        return parent.system.CoreTiming().GetClockTicks();
+        return m_parent.m_system.CoreTiming().GetClockTicks();
     }
 
     bool CheckMemoryAccess(u64 addr, u64 size, Kernel::DebugWatchpointType type) {
-        if (!check_memory_access) {
+        if (!m_check_memory_access) {
             return true;
         }
 
-        if (!memory.IsValidVirtualAddressRange(addr, size)) {
+        if (!m_memory.IsValidVirtualAddressRange(addr, size)) {
             LOG_CRITICAL(Core_ARM, "Stopping execution due to unmapped memory access at {:#x}",
                          addr);
-            parent.jit.load()->HaltExecution(PrefetchAbort);
+            m_parent.m_jit->HaltExecution(PrefetchAbort);
             return false;
         }
 
-        if (!debugger_enabled) {
+        if (!m_debugger_enabled) {
             return true;
         }
 
-        const auto match{parent.MatchingWatchpoint(addr, size, type)};
+        const auto match{m_parent.MatchingWatchpoint(addr, size, type)};
         if (match) {
-            parent.halted_watchpoint = match;
-            parent.jit.load()->HaltExecution(DataAbort);
+            m_parent.m_halted_watchpoint = match;
+            m_parent.m_jit->HaltExecution(DataAbort);
             return false;
         }
 
@@ -227,30 +207,27 @@ public:
     }
 
     void ReturnException(u64 pc, Dynarmic::HaltReason hr) {
-        parent.SaveContext(parent.breakpoint_context);
-        parent.breakpoint_context.pc = pc;
-        parent.jit.load()->HaltExecution(hr);
+        m_parent.GetContext(m_parent.m_breakpoint_context);
+        m_parent.m_breakpoint_context.pc = pc;
+        m_parent.m_jit->HaltExecution(hr);
     }
 
-    bool IsInterrupted() {
-        return parent.system.Kernel().PhysicalCore(parent.core_index).IsInterrupted();
-    }
-
-    ARM_Dynarmic_64& parent;
-    Core::Memory::Memory& memory;
-    u64 tpidrro_el0 = 0;
-    u64 tpidr_el0 = 0;
-    const bool debugger_enabled{};
-    const bool check_memory_access{};
-    static constexpr u64 minimum_run_cycles = 10000U;
+    ArmDynarmic64& m_parent;
+    Core::Memory::Memory& m_memory;
+    u64 m_tpidrro_el0{};
+    u64 m_tpidr_el0{};
+    const Kernel::KProcess* m_process{};
+    const bool m_debugger_enabled{};
+    const bool m_check_memory_access{};
+    static constexpr u64 MinimumRunCycles = 10000U;
 };
 
-std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable* page_table,
-                                                             std::size_t address_space_bits) const {
+std::shared_ptr<Dynarmic::A64::Jit> ArmDynarmic64::MakeJit(Common::PageTable* page_table,
+                                                           std::size_t address_space_bits) const {
     Dynarmic::A64::UserConfig config;
 
     // Callbacks
-    config.callbacks = cb.get();
+    config.callbacks = m_cb.get();
 
     // Memory
     if (page_table) {
@@ -271,12 +248,12 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
     }
 
     // Multi-process state
-    config.processor_id = core_index;
-    config.global_monitor = &exclusive_monitor.monitor;
+    config.processor_id = m_core_index;
+    config.global_monitor = &m_exclusive_monitor.monitor;
 
     // System registers
-    config.tpidrro_el0 = &cb->tpidrro_el0;
-    config.tpidr_el0 = &cb->tpidr_el0;
+    config.tpidrro_el0 = &m_cb->m_tpidrro_el0;
+    config.tpidr_el0 = &m_cb->m_tpidr_el0;
     config.dczid_el0 = 4;
     config.ctr_el0 = 0x8444c004;
     config.cntfrq_el0 = Hardware::CNTFREQ;
@@ -285,8 +262,8 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
     config.define_unpredictable_behaviour = true;
 
     // Timing
-    config.wall_clock_cntpct = uses_wall_clock;
-    config.enable_cycle_counting = true;
+    config.wall_clock_cntpct = m_uses_wall_clock;
+    config.enable_cycle_counting = !m_uses_wall_clock;
 
     // Code cache size
 #ifdef ARCHITECTURE_arm64
@@ -296,7 +273,7 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
 #endif
 
     // Allow memory fault handling to work
-    if (system.DebuggerEnabled()) {
+    if (m_system.DebuggerEnabled()) {
         config.check_halt_on_memory_access = true;
     }
 
@@ -384,147 +361,112 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
     return std::make_shared<Dynarmic::A64::Jit>(config);
 }
 
-HaltReason ARM_Dynarmic_64::RunJit() {
-    return TranslateHaltReason(jit.load()->Run());
+HaltReason ArmDynarmic64::RunThread(Kernel::KThread* thread) {
+    m_jit->ClearExclusiveState();
+    return TranslateHaltReason(m_jit->Run());
 }
 
-HaltReason ARM_Dynarmic_64::StepJit() {
-    return TranslateHaltReason(jit.load()->Step());
+HaltReason ArmDynarmic64::StepThread(Kernel::KThread* thread) {
+    m_jit->ClearExclusiveState();
+    return TranslateHaltReason(m_jit->Step());
 }
 
-u32 ARM_Dynarmic_64::GetSvcNumber() const {
-    return svc_swi;
+u32 ArmDynarmic64::GetSvcNumber() const {
+    return m_svc;
 }
 
-const Kernel::DebugWatchpoint* ARM_Dynarmic_64::HaltedWatchpoint() const {
-    return halted_watchpoint;
-}
+void ArmDynarmic64::GetSvcArguments(std::span<uint64_t, 8> args) const {
+    Dynarmic::A64::Jit& j = *m_jit;
 
-void ARM_Dynarmic_64::RewindBreakpointInstruction() {
-    LoadContext(breakpoint_context);
-}
-
-ARM_Dynarmic_64::ARM_Dynarmic_64(System& system_, bool uses_wall_clock_,
-                                 DynarmicExclusiveMonitor& exclusive_monitor_,
-                                 std::size_t core_index_)
-    : ARM_Interface{system_, uses_wall_clock_},
-      cb(std::make_unique<DynarmicCallbacks64>(*this)), core_index{core_index_},
-      exclusive_monitor{exclusive_monitor_}, null_jit{MakeJit(nullptr, 48)}, jit{null_jit.get()} {}
-
-ARM_Dynarmic_64::~ARM_Dynarmic_64() = default;
-
-void ARM_Dynarmic_64::SetPC(u64 pc) {
-    jit.load()->SetPC(pc);
-}
-
-u64 ARM_Dynarmic_64::GetPC() const {
-    return jit.load()->GetPC();
-}
-
-u64 ARM_Dynarmic_64::GetSP() const {
-    return jit.load()->GetSP();
-}
-
-u64 ARM_Dynarmic_64::GetReg(int index) const {
-    return jit.load()->GetRegister(index);
-}
-
-void ARM_Dynarmic_64::SetReg(int index, u64 value) {
-    jit.load()->SetRegister(index, value);
-}
-
-u128 ARM_Dynarmic_64::GetVectorReg(int index) const {
-    return jit.load()->GetVector(index);
-}
-
-void ARM_Dynarmic_64::SetVectorReg(int index, u128 value) {
-    jit.load()->SetVector(index, value);
-}
-
-u32 ARM_Dynarmic_64::GetPSTATE() const {
-    return jit.load()->GetPstate();
-}
-
-void ARM_Dynarmic_64::SetPSTATE(u32 pstate) {
-    jit.load()->SetPstate(pstate);
-}
-
-u64 ARM_Dynarmic_64::GetTlsAddress() const {
-    return cb->tpidrro_el0;
-}
-
-void ARM_Dynarmic_64::SetTlsAddress(u64 address) {
-    cb->tpidrro_el0 = address;
-}
-
-u64 ARM_Dynarmic_64::GetTPIDR_EL0() const {
-    return cb->tpidr_el0;
-}
-
-void ARM_Dynarmic_64::SetTPIDR_EL0(u64 value) {
-    cb->tpidr_el0 = value;
-}
-
-void ARM_Dynarmic_64::SaveContext(ThreadContext64& ctx) const {
-    Dynarmic::A64::Jit* j = jit.load();
-    ctx.cpu_registers = j->GetRegisters();
-    ctx.sp = j->GetSP();
-    ctx.pc = j->GetPC();
-    ctx.pstate = j->GetPstate();
-    ctx.vector_registers = j->GetVectors();
-    ctx.fpcr = j->GetFpcr();
-    ctx.fpsr = j->GetFpsr();
-    ctx.tpidr = cb->tpidr_el0;
-}
-
-void ARM_Dynarmic_64::LoadContext(const ThreadContext64& ctx) {
-    Dynarmic::A64::Jit* j = jit.load();
-    j->SetRegisters(ctx.cpu_registers);
-    j->SetSP(ctx.sp);
-    j->SetPC(ctx.pc);
-    j->SetPstate(ctx.pstate);
-    j->SetVectors(ctx.vector_registers);
-    j->SetFpcr(ctx.fpcr);
-    j->SetFpsr(ctx.fpsr);
-    SetTPIDR_EL0(ctx.tpidr);
-}
-
-void ARM_Dynarmic_64::SignalInterrupt() {
-    jit.load()->HaltExecution(BreakLoop);
-}
-
-void ARM_Dynarmic_64::ClearInterrupt() {
-    jit.load()->ClearHalt(BreakLoop);
-}
-
-void ARM_Dynarmic_64::ClearInstructionCache() {
-    jit.load()->ClearCache();
-}
-
-void ARM_Dynarmic_64::InvalidateCacheRange(u64 addr, std::size_t size) {
-    jit.load()->InvalidateCacheRange(addr, size);
-}
-
-void ARM_Dynarmic_64::ClearExclusiveState() {
-    jit.load()->ClearExclusiveState();
-}
-
-void ARM_Dynarmic_64::PageTableChanged(Common::PageTable& page_table,
-                                       std::size_t new_address_space_size_in_bits) {
-    ThreadContext64 ctx{};
-    SaveContext(ctx);
-
-    auto key = std::make_pair(&page_table, new_address_space_size_in_bits);
-    auto iter = jit_cache.find(key);
-    if (iter != jit_cache.end()) {
-        jit.store(iter->second.get());
-        LoadContext(ctx);
-        return;
+    for (size_t i = 0; i < 8; i++) {
+        args[i] = j.GetRegister(i);
     }
-    std::shared_ptr new_jit = MakeJit(&page_table, new_address_space_size_in_bits);
-    jit.store(new_jit.get());
-    LoadContext(ctx);
-    jit_cache.emplace(key, std::move(new_jit));
+}
+
+void ArmDynarmic64::SetSvcArguments(std::span<const uint64_t, 8> args) {
+    Dynarmic::A64::Jit& j = *m_jit;
+
+    for (size_t i = 0; i < 8; i++) {
+        j.SetRegister(i, args[i]);
+    }
+}
+
+const Kernel::DebugWatchpoint* ArmDynarmic64::HaltedWatchpoint() const {
+    return m_halted_watchpoint;
+}
+
+void ArmDynarmic64::RewindBreakpointInstruction() {
+    this->SetContext(m_breakpoint_context);
+}
+
+ArmDynarmic64::ArmDynarmic64(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+                             DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index)
+    : ArmInterface{uses_wall_clock}, m_system{system}, m_exclusive_monitor{exclusive_monitor},
+      m_cb(std::make_unique<DynarmicCallbacks64>(*this, process)), m_core_index{core_index} {
+    auto& page_table = process->GetPageTable().GetBasePageTable();
+    auto& page_table_impl = page_table.GetImpl();
+    m_jit = MakeJit(&page_table_impl, page_table.GetAddressSpaceWidth());
+}
+
+ArmDynarmic64::~ArmDynarmic64() = default;
+
+void ArmDynarmic64::SetTpidrroEl0(u64 value) {
+    m_cb->m_tpidrro_el0 = value;
+}
+
+void ArmDynarmic64::GetContext(Kernel::Svc::ThreadContext& ctx) const {
+    Dynarmic::A64::Jit& j = *m_jit;
+    auto gpr = j.GetRegisters();
+    auto fpr = j.GetVectors();
+
+    // TODO: this is inconvenient
+    for (size_t i = 0; i < 29; i++) {
+        ctx.r[i] = gpr[i];
+    }
+    ctx.fp = gpr[29];
+    ctx.lr = gpr[30];
+
+    ctx.sp = j.GetSP();
+    ctx.pc = j.GetPC();
+    ctx.pstate = j.GetPstate();
+    ctx.v = fpr;
+    ctx.fpcr = j.GetFpcr();
+    ctx.fpsr = j.GetFpsr();
+    ctx.tpidr = m_cb->m_tpidr_el0;
+}
+
+void ArmDynarmic64::SetContext(const Kernel::Svc::ThreadContext& ctx) {
+    Dynarmic::A64::Jit& j = *m_jit;
+
+    // TODO: this is inconvenient
+    std::array<u64, 31> gpr;
+
+    for (size_t i = 0; i < 29; i++) {
+        gpr[i] = ctx.r[i];
+    }
+    gpr[29] = ctx.fp;
+    gpr[30] = ctx.lr;
+
+    j.SetRegisters(gpr);
+    j.SetSP(ctx.sp);
+    j.SetPC(ctx.pc);
+    j.SetPstate(ctx.pstate);
+    j.SetVectors(ctx.v);
+    j.SetFpcr(ctx.fpcr);
+    j.SetFpsr(ctx.fpsr);
+    m_cb->m_tpidr_el0 = ctx.tpidr;
+}
+
+void ArmDynarmic64::SignalInterrupt(Kernel::KThread* thread) {
+    m_jit->HaltExecution(BreakLoop);
+}
+
+void ArmDynarmic64::ClearInstructionCache() {
+    m_jit->ClearCache();
+}
+
+void ArmDynarmic64::InvalidateCacheRange(u64 addr, std::size_t size) {
+    m_jit->InvalidateCacheRange(addr, size);
 }
 
 } // namespace Core

src/core/arm/dynarmic/arm_dynarmic_64.h

@@ -23,76 +23,55 @@ class DynarmicCallbacks64;
 class DynarmicExclusiveMonitor;
 class System;
 
-class ARM_Dynarmic_64 final : public ARM_Interface {
+class ArmDynarmic64 final : public ArmInterface {
 public:
-    ARM_Dynarmic_64(System& system_, bool uses_wall_clock_,
-                    DynarmicExclusiveMonitor& exclusive_monitor_, std::size_t core_index_);
-    ~ARM_Dynarmic_64() override;
-
-    void SetPC(u64 pc) override;
-    u64 GetPC() const override;
-    u64 GetSP() const override;
-    u64 GetReg(int index) const override;
-    void SetReg(int index, u64 value) override;
-    u128 GetVectorReg(int index) const override;
-    void SetVectorReg(int index, u128 value) override;
-    u32 GetPSTATE() const override;
-    void SetPSTATE(u32 pstate) override;
-    u64 GetTlsAddress() const override;
-    void SetTlsAddress(u64 address) override;
-    void SetTPIDR_EL0(u64 value) override;
-    u64 GetTPIDR_EL0() const override;
+    ArmDynarmic64(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+                  DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index);
+    ~ArmDynarmic64() override;
 
     Architecture GetArchitecture() const override {
-        return Architecture::Aarch64;
+        return Architecture::AArch64;
     }
-    void SaveContext(ThreadContext32& ctx) const override {}
-    void SaveContext(ThreadContext64& ctx) const override;
-    void LoadContext(const ThreadContext32& ctx) override {}
-    void LoadContext(const ThreadContext64& ctx) override;
 
-    void SignalInterrupt() override;
-    void ClearInterrupt() override;
-    void ClearExclusiveState() override;
+    HaltReason RunThread(Kernel::KThread* thread) override;
+    HaltReason StepThread(Kernel::KThread* thread) override;
 
+    void GetContext(Kernel::Svc::ThreadContext& ctx) const override;
+    void SetContext(const Kernel::Svc::ThreadContext& ctx) override;
+    void SetTpidrroEl0(u64 value) override;
+
+    void GetSvcArguments(std::span<uint64_t, 8> args) const override;
+    void SetSvcArguments(std::span<const uint64_t, 8> args) override;
+    u32 GetSvcNumber() const override;
+
+    void SignalInterrupt(Kernel::KThread* thread) override;
     void ClearInstructionCache() override;
     void InvalidateCacheRange(u64 addr, std::size_t size) override;
-    void PageTableChanged(Common::PageTable& new_page_table,
-                          std::size_t new_address_space_size_in_bits) override;
 
 protected:
-    HaltReason RunJit() override;
-    HaltReason StepJit() override;
-    u32 GetSvcNumber() const override;
     const Kernel::DebugWatchpoint* HaltedWatchpoint() const override;
     void RewindBreakpointInstruction() override;
 
 private:
+    System& m_system;
+    DynarmicExclusiveMonitor& m_exclusive_monitor;
+
+private:
+    friend class DynarmicCallbacks64;
+
     std::shared_ptr<Dynarmic::A64::Jit> MakeJit(Common::PageTable* page_table,
                                                 std::size_t address_space_bits) const;
+    std::unique_ptr<DynarmicCallbacks64> m_cb{};
+    std::size_t m_core_index{};
 
-    using JitCacheKey = std::pair<Common::PageTable*, std::size_t>;
-    using JitCacheType =
-        std::unordered_map<JitCacheKey, std::shared_ptr<Dynarmic::A64::Jit>, Common::PairHash>;
-
-    friend class DynarmicCallbacks64;
-    std::unique_ptr<DynarmicCallbacks64> cb;
-    JitCacheType jit_cache;
-
-    std::size_t core_index;
-    DynarmicExclusiveMonitor& exclusive_monitor;
-
-    std::shared_ptr<Dynarmic::A64::Jit> null_jit;
-
-    // A raw pointer here is fine; we never delete Jit instances.
-    std::atomic<Dynarmic::A64::Jit*> jit;
+    std::shared_ptr<Dynarmic::A64::Jit> m_jit{};
 
     // SVC callback
-    u32 svc_swi{};
+    u32 m_svc{};
 
-    // Breakpoint info
-    const Kernel::DebugWatchpoint* halted_watchpoint;
-    ThreadContext64 breakpoint_context;
+    // Watchpoint info
+    const Kernel::DebugWatchpoint* m_halted_watchpoint{};
+    Kernel::Svc::ThreadContext m_breakpoint_context{};
 };
 
 } // namespace Core

src/core/arm/dynarmic/dynarmic_cp15.cpp

@@ -124,8 +124,8 @@ CallbackOrAccessTwoWords DynarmicCP15::CompileGetTwoWords(bool two, unsigned opc
     if (!two && opc == 0 && CRm == CoprocReg::C14) {
         // CNTPCT
         const auto callback = [](void* arg, u32, u32) -> u64 {
-            const auto& parent_arg = *static_cast<ARM_Dynarmic_32*>(arg);
-            return parent_arg.system.CoreTiming().GetClockTicks();
+            const auto& parent_arg = *static_cast<ArmDynarmic32*>(arg);
+            return parent_arg.m_system.CoreTiming().GetClockTicks();
         };
         return Callback{callback, &parent};
     }

src/core/arm/dynarmic/dynarmic_cp15.h

@@ -10,13 +10,13 @@
 
 namespace Core {
 
-class ARM_Dynarmic_32;
+class ArmDynarmic32;
 
 class DynarmicCP15 final : public Dynarmic::A32::Coprocessor {
 public:
     using CoprocReg = Dynarmic::A32::CoprocReg;
 
-    explicit DynarmicCP15(ARM_Dynarmic_32& parent_) : parent{parent_} {}
+    explicit DynarmicCP15(ArmDynarmic32& parent_) : parent{parent_} {}
 
     std::optional<Callback> CompileInternalOperation(bool two, unsigned opc1, CoprocReg CRd,
                                                      CoprocReg CRn, CoprocReg CRm,
@@ -32,11 +32,11 @@ public:
     std::optional<Callback> CompileStoreWords(bool two, bool long_transfer, CoprocReg CRd,
                                               std::optional<u8> option) override;
 
-    ARM_Dynarmic_32& parent;
+    ArmDynarmic32& parent;
     u32 uprw = 0;
     u32 uro = 0;
 
-    friend class ARM_Dynarmic_32;
+    friend class ArmDynarmic32;
 };
 
 } // namespace Core

src/core/arm/dynarmic/dynarmic_exclusive_monitor.h

@@ -14,8 +14,8 @@ class Memory;
 
 namespace Core {
 
-class ARM_Dynarmic_32;
-class ARM_Dynarmic_64;
+class ArmDynarmic32;
+class ArmDynarmic64;
 
 class DynarmicExclusiveMonitor final : public ExclusiveMonitor {
 public:
@@ -36,8 +36,8 @@ public:
     bool ExclusiveWrite128(std::size_t core_index, VAddr vaddr, u128 value) override;
 
 private:
-    friend class ARM_Dynarmic_32;
-    friend class ARM_Dynarmic_64;
+    friend class ArmDynarmic32;
+    friend class ArmDynarmic64;
     Dynarmic::ExclusiveMonitor monitor;
     Core::Memory::Memory& memory;
 };

src/core/arm/nce/arm_nce.cpp

@@ -6,6 +6,7 @@
 
 #include "common/signal_chain.h"
 #include "core/arm/nce/arm_nce.h"
+#include "core/arm/nce/guest_context.h"
 #include "core/arm/nce/patcher.h"
 #include "core/core.h"
 #include "core/memory.h"
@@ -38,7 +39,7 @@ fpsimd_context* GetFloatingPointState(mcontext_t& host_ctx) {
 
 } // namespace
 
-void* ARM_NCE::RestoreGuestContext(void* raw_context) {
+void* ArmNce::RestoreGuestContext(void* raw_context) {
     // Retrieve the host context.
     auto& host_ctx = static_cast<ucontext_t*>(raw_context)->uc_mcontext;
 
@@ -71,7 +72,7 @@ void* ARM_NCE::RestoreGuestContext(void* raw_context) {
     return tpidr;
 }
 
-void ARM_NCE::SaveGuestContext(GuestContext* guest_ctx, void* raw_context) {
+void ArmNce::SaveGuestContext(GuestContext* guest_ctx, void* raw_context) {
     // Retrieve the host context.
     auto& host_ctx = static_cast<ucontext_t*>(raw_context)->uc_mcontext;
 
@@ -103,7 +104,7 @@ void ARM_NCE::SaveGuestContext(GuestContext* guest_ctx, void* raw_context) {
     host_ctx.regs[0] = guest_ctx->esr_el1.exchange(0);
 }
 
-bool ARM_NCE::HandleGuestFault(GuestContext* guest_ctx, void* raw_info, void* raw_context) {
+bool ArmNce::HandleGuestFault(GuestContext* guest_ctx, void* raw_info, void* raw_context) {
     auto& host_ctx = static_cast<ucontext_t*>(raw_context)->uc_mcontext;
     auto* info = static_cast<siginfo_t*>(raw_info);
 
@@ -134,7 +135,7 @@ bool ARM_NCE::HandleGuestFault(GuestContext* guest_ctx, void* raw_info, void* ra
     // - If we lose the race, then SignalInterrupt will send us a signal we are masking,
     //   and it will do nothing when it is unmasked, as we have already left guest code.
     // - If we win the race, then SignalInterrupt will wait for us to unlock first.
-    auto& thread_params = guest_ctx->parent->running_thread->GetNativeExecutionParameters();
+    auto& thread_params = guest_ctx->parent->m_running_thread->GetNativeExecutionParameters();
     thread_params.lock.store(SpinLockLocked);
 
     // Return to host.
@@ -142,97 +143,93 @@ bool ARM_NCE::HandleGuestFault(GuestContext* guest_ctx, void* raw_info, void* ra
     return false;
 }
 
-void ARM_NCE::HandleHostFault(int sig, void* raw_info, void* raw_context) {
+void ArmNce::HandleHostFault(int sig, void* raw_info, void* raw_context) {
     return g_orig_action.sa_sigaction(sig, static_cast<siginfo_t*>(raw_info), raw_context);
 }
 
-HaltReason ARM_NCE::RunJit() {
-    // Get the thread parameters.
-    // TODO: pass the current thread down from ::Run
-    auto* thread = Kernel::GetCurrentThreadPointer(system.Kernel());
+void ArmNce::LockThread(Kernel::KThread* thread) {
     auto* thread_params = &thread->GetNativeExecutionParameters();
+    LockThreadParameters(thread_params);
+}
 
-    {
-        // Lock our core context.
-        std::scoped_lock lk{lock};
+void ArmNce::UnlockThread(Kernel::KThread* thread) {
+    auto* thread_params = &thread->GetNativeExecutionParameters();
+    UnlockThreadParameters(thread_params);
+}
 
-        // We should not be running.
-        ASSERT(running_thread == nullptr);
-
-        // Check if we need to run. If we have already been halted, we are done.
-        u64 halt = guest_ctx.esr_el1.exchange(0);
-        if (halt != 0) {
-            return static_cast<HaltReason>(halt);
-        }
-
-        // Mark that we are running.
-        running_thread = thread;
-
-        // Acquire the lock on the thread parameters.
-        // This allows us to force synchronization with SignalInterrupt.
-        LockThreadParameters(thread_params);
+HaltReason ArmNce::RunThread(Kernel::KThread* thread) {
+    // Check if we're already interrupted.
+    // If we are, we can just return immediately.
+    HaltReason hr = static_cast<HaltReason>(m_guest_ctx.esr_el1.exchange(0));
+    if (True(hr)) {
+        return hr;
     }
 
-    // Assign current members.
-    guest_ctx.parent = this;
-    thread_params->native_context = &guest_ctx;
-    thread_params->tpidr_el0 = guest_ctx.tpidr_el0;
-    thread_params->tpidrro_el0 = guest_ctx.tpidrro_el0;
-    thread_params->is_running = true;
+    // Get the thread context.
+    auto* thread_params = &thread->GetNativeExecutionParameters();
+    auto* process = thread->GetOwnerProcess();
 
-    HaltReason halt{};
+    // Assign current members.
+    m_running_thread = thread;
+    m_guest_ctx.parent = this;
+    thread_params->native_context = &m_guest_ctx;
+    thread_params->tpidr_el0 = m_guest_ctx.tpidr_el0;
+    thread_params->tpidrro_el0 = m_guest_ctx.tpidrro_el0;
+    thread_params->is_running = true;
 
     // TODO: finding and creating the post handler needs to be locked
     // to deal with dynamic loading of NROs.
-    const auto& post_handlers = system.ApplicationProcess()->GetPostHandlers();
-    if (auto it = post_handlers.find(guest_ctx.pc); it != post_handlers.end()) {
-        halt = ReturnToRunCodeByTrampoline(thread_params, &guest_ctx, it->second);
+    const auto& post_handlers = process->GetPostHandlers();
+    if (auto it = post_handlers.find(m_guest_ctx.pc); it != post_handlers.end()) {
+        hr = ReturnToRunCodeByTrampoline(thread_params, &m_guest_ctx, it->second);
     } else {
-        halt = ReturnToRunCodeByExceptionLevelChange(thread_id, thread_params);
+        hr = ReturnToRunCodeByExceptionLevelChange(m_thread_id, thread_params);
     }
 
     // Unload members.
     // The thread does not change, so we can persist the old reference.
-    guest_ctx.tpidr_el0 = thread_params->tpidr_el0;
+    m_running_thread = nullptr;
+    m_guest_ctx.tpidr_el0 = thread_params->tpidr_el0;
     thread_params->native_context = nullptr;
     thread_params->is_running = false;
 
-    // Unlock the thread parameters.
-    UnlockThreadParameters(thread_params);
-
-    {
-        // Lock the core context.
-        std::scoped_lock lk{lock};
-
-        // On exit, we no longer have an active thread.
-        running_thread = nullptr;
-    }
-
     // Return the halt reason.
-    return halt;
+    return hr;
 }
 
-HaltReason ARM_NCE::StepJit() {
+HaltReason ArmNce::StepThread(Kernel::KThread* thread) {
     return HaltReason::StepThread;
 }
 
-u32 ARM_NCE::GetSvcNumber() const {
-    return guest_ctx.svc_swi;
+u32 ArmNce::GetSvcNumber() const {
+    return m_guest_ctx.svc;
 }
 
-ARM_NCE::ARM_NCE(System& system_, bool uses_wall_clock_, std::size_t core_index_)
-    : ARM_Interface{system_, uses_wall_clock_}, core_index{core_index_} {
-    guest_ctx.system = &system_;
+void ArmNce::GetSvcArguments(std::span<uint64_t, 8> args) const {
+    for (size_t i = 0; i < 8; i++) {
+        args[i] = m_guest_ctx.cpu_registers[i];
+    }
 }
 
-ARM_NCE::~ARM_NCE() = default;
+void ArmNce::SetSvcArguments(std::span<const uint64_t, 8> args) {
+    for (size_t i = 0; i < 8; i++) {
+        m_guest_ctx.cpu_registers[i] = args[i];
+    }
+}
 
-void ARM_NCE::Initialize() {
-    thread_id = gettid();
+ArmNce::ArmNce(System& system, bool uses_wall_clock, std::size_t core_index)
+    : ArmInterface{uses_wall_clock}, m_system{system}, m_core_index{core_index} {
+    m_guest_ctx.system = &m_system;
+}
+
+ArmNce::~ArmNce() = default;
+
+void ArmNce::Initialize() {
+    m_thread_id = gettid();
 
     // Setup our signals
-    static std::once_flag flag;
-    std::call_once(flag, [] {
+    static std::once_flag signals;
+    std::call_once(signals, [] {
         using HandlerType = decltype(sigaction::sa_sigaction);
 
         sigset_t signal_mask;
@@ -244,7 +241,7 @@ void ARM_NCE::Initialize() {
         struct sigaction return_to_run_code_action {};
         return_to_run_code_action.sa_flags = SA_SIGINFO | SA_ONSTACK;
         return_to_run_code_action.sa_sigaction = reinterpret_cast<HandlerType>(
-            &ARM_NCE::ReturnToRunCodeByExceptionLevelChangeSignalHandler);
+            &ArmNce::ReturnToRunCodeByExceptionLevelChangeSignalHandler);
         return_to_run_code_action.sa_mask = signal_mask;
         Common::SigAction(ReturnToRunCodeByExceptionLevelChangeSignal, &return_to_run_code_action,
                           nullptr);
@@ -252,14 +249,13 @@ void ARM_NCE::Initialize() {
         struct sigaction break_from_run_code_action {};
         break_from_run_code_action.sa_flags = SA_SIGINFO | SA_ONSTACK;
         break_from_run_code_action.sa_sigaction =
-            reinterpret_cast<HandlerType>(&ARM_NCE::BreakFromRunCodeSignalHandler);
+            reinterpret_cast<HandlerType>(&ArmNce::BreakFromRunCodeSignalHandler);
         break_from_run_code_action.sa_mask = signal_mask;
         Common::SigAction(BreakFromRunCodeSignal, &break_from_run_code_action, nullptr);
 
         struct sigaction fault_action {};
         fault_action.sa_flags = SA_SIGINFO | SA_ONSTACK | SA_RESTART;
-        fault_action.sa_sigaction =
-            reinterpret_cast<HandlerType>(&ARM_NCE::GuestFaultSignalHandler);
+        fault_action.sa_sigaction = reinterpret_cast<HandlerType>(&ArmNce::GuestFaultSignalHandler);
         fault_action.sa_mask = signal_mask;
         Common::SigAction(GuestFaultSignal, &fault_action, &g_orig_action);
 
@@ -272,111 +268,59 @@ void ARM_NCE::Initialize() {
     });
 }
 
-void ARM_NCE::SetPC(u64 pc) {
-    guest_ctx.pc = pc;
+void ArmNce::SetTpidrroEl0(u64 value) {
+    m_guest_ctx.tpidrro_el0 = value;
 }
 
-u64 ARM_NCE::GetPC() const {
-    return guest_ctx.pc;
-}
-
-u64 ARM_NCE::GetSP() const {
-    return guest_ctx.sp;
-}
-
-u64 ARM_NCE::GetReg(int index) const {
-    return guest_ctx.cpu_registers[index];
-}
-
-void ARM_NCE::SetReg(int index, u64 value) {
-    guest_ctx.cpu_registers[index] = value;
-}
-
-u128 ARM_NCE::GetVectorReg(int index) const {
-    return guest_ctx.vector_registers[index];
-}
-
-void ARM_NCE::SetVectorReg(int index, u128 value) {
-    guest_ctx.vector_registers[index] = value;
-}
-
-u32 ARM_NCE::GetPSTATE() const {
-    return guest_ctx.pstate;
-}
-
-void ARM_NCE::SetPSTATE(u32 pstate) {
-    guest_ctx.pstate = pstate;
-}
-
-u64 ARM_NCE::GetTlsAddress() const {
-    return guest_ctx.tpidrro_el0;
-}
-
-void ARM_NCE::SetTlsAddress(u64 address) {
-    guest_ctx.tpidrro_el0 = address;
-}
-
-u64 ARM_NCE::GetTPIDR_EL0() const {
-    return guest_ctx.tpidr_el0;
-}
-
-void ARM_NCE::SetTPIDR_EL0(u64 value) {
-    guest_ctx.tpidr_el0 = value;
-}
-
-void ARM_NCE::SaveContext(ThreadContext64& ctx) const {
-    ctx.cpu_registers = guest_ctx.cpu_registers;
-    ctx.sp = guest_ctx.sp;
-    ctx.pc = guest_ctx.pc;
-    ctx.pstate = guest_ctx.pstate;
-    ctx.vector_registers = guest_ctx.vector_registers;
-    ctx.fpcr = guest_ctx.fpcr;
-    ctx.fpsr = guest_ctx.fpsr;
-    ctx.tpidr = guest_ctx.tpidr_el0;
-}
-
-void ARM_NCE::LoadContext(const ThreadContext64& ctx) {
-    guest_ctx.cpu_registers = ctx.cpu_registers;
-    guest_ctx.sp = ctx.sp;
-    guest_ctx.pc = ctx.pc;
-    guest_ctx.pstate = ctx.pstate;
-    guest_ctx.vector_registers = ctx.vector_registers;
-    guest_ctx.fpcr = ctx.fpcr;
-    guest_ctx.fpsr = ctx.fpsr;
-    guest_ctx.tpidr_el0 = ctx.tpidr;
-}
-
-void ARM_NCE::SignalInterrupt() {
-    // Lock core context.
-    std::scoped_lock lk{lock};
-
-    // Add break loop condition.
-    guest_ctx.esr_el1.fetch_or(static_cast<u64>(HaltReason::BreakLoop));
-
-    // If there is no thread running, we are done.
-    if (running_thread == nullptr) {
-        return;
+void ArmNce::GetContext(Kernel::Svc::ThreadContext& ctx) const {
+    for (size_t i = 0; i < 29; i++) {
+        ctx.r[i] = m_guest_ctx.cpu_registers[i];
     }
+    ctx.fp = m_guest_ctx.cpu_registers[29];
+    ctx.lr = m_guest_ctx.cpu_registers[30];
+    ctx.sp = m_guest_ctx.sp;
+    ctx.pc = m_guest_ctx.pc;
+    ctx.pstate = m_guest_ctx.pstate;
+    ctx.v = m_guest_ctx.vector_registers;
+    ctx.fpcr = m_guest_ctx.fpcr;
+    ctx.fpsr = m_guest_ctx.fpsr;
+    ctx.tpidr = m_guest_ctx.tpidr_el0;
+}
+
+void ArmNce::SetContext(const Kernel::Svc::ThreadContext& ctx) {
+    for (size_t i = 0; i < 29; i++) {
+        m_guest_ctx.cpu_registers[i] = ctx.r[i];
+    }
+    m_guest_ctx.cpu_registers[29] = ctx.fp;
+    m_guest_ctx.cpu_registers[30] = ctx.lr;
+    m_guest_ctx.sp = ctx.sp;
+    m_guest_ctx.pc = ctx.pc;
+    m_guest_ctx.pstate = ctx.pstate;
+    m_guest_ctx.vector_registers = ctx.v;
+    m_guest_ctx.fpcr = ctx.fpcr;
+    m_guest_ctx.fpsr = ctx.fpsr;
+    m_guest_ctx.tpidr_el0 = ctx.tpidr;
+}
+
+void ArmNce::SignalInterrupt(Kernel::KThread* thread) {
+    // Add break loop condition.
+    m_guest_ctx.esr_el1.fetch_or(static_cast<u64>(HaltReason::BreakLoop));
 
     // Lock the thread context.
-    auto* params = &running_thread->GetNativeExecutionParameters();
+    auto* params = &thread->GetNativeExecutionParameters();
     LockThreadParameters(params);
 
     if (params->is_running) {
         // We should signal to the running thread.
         // The running thread will unlock the thread context.
-        syscall(SYS_tkill, thread_id, BreakFromRunCodeSignal);
+        syscall(SYS_tkill, m_thread_id, BreakFromRunCodeSignal);
     } else {
         // If the thread is no longer running, we have nothing to do.
         UnlockThreadParameters(params);
     }
 }
 
-void ARM_NCE::ClearInterrupt() {
-    guest_ctx.esr_el1 = {};
-}
-
-void ARM_NCE::ClearInstructionCache() {
+void ArmNce::ClearInstructionCache() {
     // TODO: This is not possible to implement correctly on Linux because
     // we do not have any access to ic iallu.
 
@@ -384,17 +328,8 @@ void ARM_NCE::ClearInstructionCache() {
     std::atomic_thread_fence(std::memory_order_seq_cst);
 }
 
-void ARM_NCE::InvalidateCacheRange(u64 addr, std::size_t size) {
+void ArmNce::InvalidateCacheRange(u64 addr, std::size_t size) {
     this->ClearInstructionCache();
 }
 
-void ARM_NCE::ClearExclusiveState() {
-    // No-op.
-}
-
-void ARM_NCE::PageTableChanged(Common::PageTable& page_table,
-                               std::size_t new_address_space_size_in_bits) {
-    // No-op. Page table is never used.
-}
-
 } // namespace Core

src/core/arm/nce/arm_nce.h

@@ -3,11 +3,7 @@
 
 #pragma once
 
-#include <atomic>
-#include <memory>
-#include <span>
-#include <unordered_map>
-#include <vector>
+#include <mutex>
 
 #include "core/arm/arm_interface.h"
 #include "core/arm/nce/guest_context.h"
@@ -20,51 +16,36 @@ namespace Core {
 
 class System;
 
-class ARM_NCE final : public ARM_Interface {
+class ArmNce final : public ArmInterface {
 public:
-    ARM_NCE(System& system_, bool uses_wall_clock_, std::size_t core_index_);
-
-    ~ARM_NCE() override;
+    ArmNce(System& system, bool uses_wall_clock, std::size_t core_index);
+    ~ArmNce() override;
 
     void Initialize() override;
-    void SetPC(u64 pc) override;
-    u64 GetPC() const override;
-    u64 GetSP() const override;
-    u64 GetReg(int index) const override;
-    void SetReg(int index, u64 value) override;
-    u128 GetVectorReg(int index) const override;
-    void SetVectorReg(int index, u128 value) override;
-
-    u32 GetPSTATE() const override;
-    void SetPSTATE(u32 pstate) override;
-    u64 GetTlsAddress() const override;
-    void SetTlsAddress(u64 address) override;
-    void SetTPIDR_EL0(u64 value) override;
-    u64 GetTPIDR_EL0() const override;
 
     Architecture GetArchitecture() const override {
-        return Architecture::Aarch64;
+        return Architecture::AArch64;
     }
 
-    void SaveContext(ThreadContext32& ctx) const override {}
-    void SaveContext(ThreadContext64& ctx) const override;
-    void LoadContext(const ThreadContext32& ctx) override {}
-    void LoadContext(const ThreadContext64& ctx) override;
+    HaltReason RunThread(Kernel::KThread* thread) override;
+    HaltReason StepThread(Kernel::KThread* thread) override;
 
-    void SignalInterrupt() override;
-    void ClearInterrupt() override;
-    void ClearExclusiveState() override;
-    void ClearInstructionCache() override;
-    void InvalidateCacheRange(u64 addr, std::size_t size) override;
-    void PageTableChanged(Common::PageTable& new_page_table,
-                          std::size_t new_address_space_size_in_bits) override;
-
-protected:
-    HaltReason RunJit() override;
-    HaltReason StepJit() override;
+    void GetContext(Kernel::Svc::ThreadContext& ctx) const override;
+    void SetContext(const Kernel::Svc::ThreadContext& ctx) override;
+    void SetTpidrroEl0(u64 value) override;
 
+    void GetSvcArguments(std::span<uint64_t, 8> args) const override;
+    void SetSvcArguments(std::span<const uint64_t, 8> args) override;
     u32 GetSvcNumber() const override;
 
+    void SignalInterrupt(Kernel::KThread* thread) override;
+    void ClearInstructionCache() override;
+    void InvalidateCacheRange(u64 addr, std::size_t size) override;
+
+    void LockThread(Kernel::KThread* thread) override;
+    void UnlockThread(Kernel::KThread* thread) override;
+
+protected:
     const Kernel::DebugWatchpoint* HaltedWatchpoint() const override {
         return nullptr;
     }
@@ -93,16 +74,15 @@ private:
     static void HandleHostFault(int sig, void* info, void* raw_context);
 
 public:
+    Core::System& m_system;
+
     // Members set on initialization.
-    std::size_t core_index{};
-    pid_t thread_id{-1};
+    std::size_t m_core_index{};
+    pid_t m_thread_id{-1};
 
     // Core context.
-    GuestContext guest_ctx;
-
-    // Thread and invalidation info.
-    std::mutex lock;
-    Kernel::KThread* running_thread{};
+    GuestContext m_guest_ctx{};
+    Kernel::KThread* m_running_thread{};
 };
 
 } // namespace Core

src/core/arm/nce/arm_nce.s

@@ -8,11 +8,11 @@
     movk    reg, #(((val) >> 0x10) & 0xFFFF), lsl #16
 
 
-/* static HaltReason Core::ARM_NCE::ReturnToRunCodeByTrampoline(void* tpidr, Core::GuestContext* ctx, u64 trampoline_addr) */
-.section    .text._ZN4Core7ARM_NCE27ReturnToRunCodeByTrampolineEPvPNS_12GuestContextEm, "ax", %progbits
-.global     _ZN4Core7ARM_NCE27ReturnToRunCodeByTrampolineEPvPNS_12GuestContextEm
-.type       _ZN4Core7ARM_NCE27ReturnToRunCodeByTrampolineEPvPNS_12GuestContextEm, %function
-_ZN4Core7ARM_NCE27ReturnToRunCodeByTrampolineEPvPNS_12GuestContextEm:
+/* static HaltReason Core::ArmNce::ReturnToRunCodeByTrampoline(void* tpidr, Core::GuestContext* ctx, u64 trampoline_addr) */
+.section    .text._ZN4Core6ArmNce27ReturnToRunCodeByTrampolineEPvPNS_12GuestContextEm, "ax", %progbits
+.global     _ZN4Core6ArmNce27ReturnToRunCodeByTrampolineEPvPNS_12GuestContextEm
+.type       _ZN4Core6ArmNce27ReturnToRunCodeByTrampolineEPvPNS_12GuestContextEm, %function
+_ZN4Core6ArmNce27ReturnToRunCodeByTrampolineEPvPNS_12GuestContextEm:
     /* Back up host sp to x3. */
     /* Back up host tpidr_el0 to x4. */
     mov     x3, sp
@@ -49,11 +49,11 @@ _ZN4Core7ARM_NCE27ReturnToRunCodeByTrampolineEPvPNS_12GuestContextEm:
     br      x2
 
 
-/* static HaltReason Core::ARM_NCE::ReturnToRunCodeByExceptionLevelChange(int tid, void* tpidr) */
-.section    .text._ZN4Core7ARM_NCE37ReturnToRunCodeByExceptionLevelChangeEiPv, "ax", %progbits
-.global     _ZN4Core7ARM_NCE37ReturnToRunCodeByExceptionLevelChangeEiPv
-.type       _ZN4Core7ARM_NCE37ReturnToRunCodeByExceptionLevelChangeEiPv, %function
-_ZN4Core7ARM_NCE37ReturnToRunCodeByExceptionLevelChangeEiPv:
+/* static HaltReason Core::ArmNce::ReturnToRunCodeByExceptionLevelChange(int tid, void* tpidr) */
+.section    .text._ZN4Core6ArmNce37ReturnToRunCodeByExceptionLevelChangeEiPv, "ax", %progbits
+.global     _ZN4Core6ArmNce37ReturnToRunCodeByExceptionLevelChangeEiPv
+.type       _ZN4Core6ArmNce37ReturnToRunCodeByExceptionLevelChangeEiPv, %function
+_ZN4Core6ArmNce37ReturnToRunCodeByExceptionLevelChangeEiPv:
     /* This jumps to the signal handler, which will restore the entire context. */
     /* On entry, x0 = thread id, which is already in the right place. */
 
@@ -71,17 +71,17 @@ _ZN4Core7ARM_NCE37ReturnToRunCodeByExceptionLevelChangeEiPv:
     brk     #1000
 
 
-/* static void Core::ARM_NCE::ReturnToRunCodeByExceptionLevelChangeSignalHandler(int sig, void* info, void* raw_context) */
-.section    .text._ZN4Core7ARM_NCE50ReturnToRunCodeByExceptionLevelChangeSignalHandlerEiPvS1_, "ax", %progbits
-.global     _ZN4Core7ARM_NCE50ReturnToRunCodeByExceptionLevelChangeSignalHandlerEiPvS1_
-.type       _ZN4Core7ARM_NCE50ReturnToRunCodeByExceptionLevelChangeSignalHandlerEiPvS1_, %function
-_ZN4Core7ARM_NCE50ReturnToRunCodeByExceptionLevelChangeSignalHandlerEiPvS1_:
+/* static void Core::ArmNce::ReturnToRunCodeByExceptionLevelChangeSignalHandler(int sig, void* info, void* raw_context) */
+.section    .text._ZN4Core6ArmNce50ReturnToRunCodeByExceptionLevelChangeSignalHandlerEiPvS1_, "ax", %progbits
+.global     _ZN4Core6ArmNce50ReturnToRunCodeByExceptionLevelChangeSignalHandlerEiPvS1_
+.type       _ZN4Core6ArmNce50ReturnToRunCodeByExceptionLevelChangeSignalHandlerEiPvS1_, %function
+_ZN4Core6ArmNce50ReturnToRunCodeByExceptionLevelChangeSignalHandlerEiPvS1_:
     stp     x29, x30, [sp, #-0x10]!
     mov     x29, sp
 
     /* Call the context restorer with the raw context. */
     mov     x0, x2
-    bl      _ZN4Core7ARM_NCE19RestoreGuestContextEPv
+    bl      _ZN4Core6ArmNce19RestoreGuestContextEPv
 
     /* Save the old value of tpidr_el0. */
     mrs     x8, tpidr_el0
@@ -92,18 +92,18 @@ _ZN4Core7ARM_NCE50ReturnToRunCodeByExceptionLevelChangeSignalHandlerEiPvS1_:
     msr     tpidr_el0, x0
 
     /* Unlock the context. */
-    bl      _ZN4Core7ARM_NCE22UnlockThreadParametersEPv
+    bl      _ZN4Core6ArmNce22UnlockThreadParametersEPv
 
     /* Returning from here will enter the guest. */
     ldp     x29, x30, [sp], #0x10
     ret
 
 
-/* static void Core::ARM_NCE::BreakFromRunCodeSignalHandler(int sig, void* info, void* raw_context) */
-.section    .text._ZN4Core7ARM_NCE29BreakFromRunCodeSignalHandlerEiPvS1_, "ax", %progbits
-.global     _ZN4Core7ARM_NCE29BreakFromRunCodeSignalHandlerEiPvS1_
-.type       _ZN4Core7ARM_NCE29BreakFromRunCodeSignalHandlerEiPvS1_, %function
-_ZN4Core7ARM_NCE29BreakFromRunCodeSignalHandlerEiPvS1_:
+/* static void Core::ArmNce::BreakFromRunCodeSignalHandler(int sig, void* info, void* raw_context) */
+.section    .text._ZN4Core6ArmNce29BreakFromRunCodeSignalHandlerEiPvS1_, "ax", %progbits
+.global     _ZN4Core6ArmNce29BreakFromRunCodeSignalHandlerEiPvS1_
+.type       _ZN4Core6ArmNce29BreakFromRunCodeSignalHandlerEiPvS1_, %function
+_ZN4Core6ArmNce29BreakFromRunCodeSignalHandlerEiPvS1_:
     /* Check to see if we have the correct TLS magic. */
     mrs     x8, tpidr_el0
     ldr     w9, [x8, #(TpidrEl0TlsMagic)]
@@ -121,7 +121,7 @@ _ZN4Core7ARM_NCE29BreakFromRunCodeSignalHandlerEiPvS1_:
 
     /* Tail call the restorer. */
     mov     x1, x2
-    b       _ZN4Core7ARM_NCE16SaveGuestContextEPNS_12GuestContextEPv
+    b       _ZN4Core6ArmNce16SaveGuestContextEPNS_12GuestContextEPv
 
     /* Returning from here will enter host code. */
 
@@ -130,11 +130,11 @@ _ZN4Core7ARM_NCE29BreakFromRunCodeSignalHandlerEiPvS1_:
     ret
 
 
-/* static void Core::ARM_NCE::GuestFaultSignalHandler(int sig, void* info, void* raw_context) */
-.section    .text._ZN4Core7ARM_NCE23GuestFaultSignalHandlerEiPvS1_, "ax", %progbits
-.global     _ZN4Core7ARM_NCE23GuestFaultSignalHandlerEiPvS1_
-.type       _ZN4Core7ARM_NCE23GuestFaultSignalHandlerEiPvS1_, %function
-_ZN4Core7ARM_NCE23GuestFaultSignalHandlerEiPvS1_:
+/* static void Core::ArmNce::GuestFaultSignalHandler(int sig, void* info, void* raw_context) */
+.section    .text._ZN4Core6ArmNce23GuestFaultSignalHandlerEiPvS1_, "ax", %progbits
+.global     _ZN4Core6ArmNce23GuestFaultSignalHandlerEiPvS1_
+.type       _ZN4Core6ArmNce23GuestFaultSignalHandlerEiPvS1_, %function
+_ZN4Core6ArmNce23GuestFaultSignalHandlerEiPvS1_:
     /* Check to see if we have the correct TLS magic. */
     mrs     x8, tpidr_el0
     ldr     w9, [x8, #(TpidrEl0TlsMagic)]
@@ -146,7 +146,7 @@ _ZN4Core7ARM_NCE23GuestFaultSignalHandlerEiPvS1_:
 
     /* Incorrect TLS magic, so this is a host fault. */
     /* Tail call the handler. */
-    b       _ZN4Core7ARM_NCE15HandleHostFaultEiPvS1_
+    b       _ZN4Core6ArmNce15HandleHostFaultEiPvS1_
 
 1:
     /* Correct TLS magic, so this is a guest fault. */
@@ -163,7 +163,7 @@ _ZN4Core7ARM_NCE23GuestFaultSignalHandlerEiPvS1_:
     msr     tpidr_el0, x3
 
     /* Call the handler. */
-    bl       _ZN4Core7ARM_NCE16HandleGuestFaultEPNS_12GuestContextEPvS3_
+    bl       _ZN4Core6ArmNce16HandleGuestFaultEPNS_12GuestContextEPvS3_
 
     /* If the handler returned false, we want to preserve the host tpidr_el0. */
     cbz     x0, 2f
@@ -177,11 +177,11 @@ _ZN4Core7ARM_NCE23GuestFaultSignalHandlerEiPvS1_:
     ret
 
 
-/* static void Core::ARM_NCE::LockThreadParameters(void* tpidr) */
-.section    .text._ZN4Core7ARM_NCE20LockThreadParametersEPv, "ax", %progbits
-.global     _ZN4Core7ARM_NCE20LockThreadParametersEPv
-.type       _ZN4Core7ARM_NCE20LockThreadParametersEPv, %function
-_ZN4Core7ARM_NCE20LockThreadParametersEPv:
+/* static void Core::ArmNce::LockThreadParameters(void* tpidr) */
+.section    .text._ZN4Core6ArmNce20LockThreadParametersEPv, "ax", %progbits
+.global     _ZN4Core6ArmNce20LockThreadParametersEPv
+.type       _ZN4Core6ArmNce20LockThreadParametersEPv, %function
+_ZN4Core6ArmNce20LockThreadParametersEPv:
     /* Offset to lock member. */
     add     x0, x0, #(TpidrEl0Lock)
 
@@ -205,11 +205,11 @@ _ZN4Core7ARM_NCE20LockThreadParametersEPv:
     ret
 
 
-/* static void Core::ARM_NCE::UnlockThreadParameters(void* tpidr) */
-.section    .text._ZN4Core7ARM_NCE22UnlockThreadParametersEPv, "ax", %progbits
-.global     _ZN4Core7ARM_NCE22UnlockThreadParametersEPv
-.type       _ZN4Core7ARM_NCE22UnlockThreadParametersEPv, %function
-_ZN4Core7ARM_NCE22UnlockThreadParametersEPv:
+/* static void Core::ArmNce::UnlockThreadParameters(void* tpidr) */
+.section    .text._ZN4Core6ArmNce22UnlockThreadParametersEPv, "ax", %progbits
+.global     _ZN4Core6ArmNce22UnlockThreadParametersEPv
+.type       _ZN4Core6ArmNce22UnlockThreadParametersEPv, %function
+_ZN4Core6ArmNce22UnlockThreadParametersEPv:
     /* Offset to lock member. */
     add     x0, x0, #(TpidrEl0Lock)
 

src/core/arm/nce/guest_context.h

@@ -3,6 +3,8 @@
 
 #pragma once
 
+#include <atomic>
+
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "core/arm/arm_interface.h"
@@ -10,7 +12,7 @@
 
 namespace Core {
 
-class ARM_NCE;
+class ArmNce;
 class System;
 
 struct HostContext {
@@ -33,9 +35,9 @@ struct GuestContext {
     u64 tpidr_el0{};
     std::atomic<u64> esr_el1{};
     u32 nzcv{};
-    u32 svc_swi{};
+    u32 svc{};
     System* system{};
-    ARM_NCE* parent{};
+    ArmNce* parent{};
 };
 
 // Verify assembly offsets.

src/core/arm/nce/patcher.cpp

@@ -28,6 +28,8 @@ Patcher::~Patcher() = default;
 
 void Patcher::PatchText(const Kernel::PhysicalMemory& program_image,
                         const Kernel::CodeSet::Segment& code) {
+    // Branch to the first instruction of the module.
+    this->BranchToModule(0);
 
     // Write save context helper function.
     c.l(m_save_context);
@@ -278,7 +280,7 @@ void Patcher::WriteSvcTrampoline(ModuleDestLabel module_dest, u32 svc_id) {
 
     // Store SVC number to execute when we return
     c.MOV(X2, svc_id);
-    c.STR(W2, X1, offsetof(GuestContext, svc_swi));
+    c.STR(W2, X1, offsetof(GuestContext, svc));
 
     // We are calling a SVC. Clear esr_el1 and return it.
     static_assert(std::is_same_v<std::underlying_type_t<HaltReason>, u64>);

src/core/core.cpp

@@ -36,6 +36,7 @@
 #include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/physical_core.h"
+#include "core/hle/service/acc/profile_manager.h"
 #include "core/hle/service/am/applets/applets.h"
 #include "core/hle/service/apm/apm_controller.h"
 #include "core/hle/service/filesystem/filesystem.h"
@@ -130,8 +131,8 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
 struct System::Impl {
     explicit Impl(System& system)
         : kernel{system}, fs_controller{system}, memory{system}, hid_core{}, room_network{},
-          cpu_manager{system}, reporter{system}, applet_manager{system}, time_manager{system},
-          gpu_dirty_memory_write_manager{} {
+          cpu_manager{system}, reporter{system}, applet_manager{system}, profile_manager{},
+          time_manager{system}, gpu_dirty_memory_write_manager{} {
         memory.SetGPUDirtyManagers(gpu_dirty_memory_write_manager);
     }
 
@@ -323,7 +324,6 @@ struct System::Impl {
                 static_cast<u32>(SystemResultStatus::ErrorLoader) + static_cast<u32>(load_result));
         }
         AddGlueRegistrationForProcess(*app_loader, *main_process);
-        kernel.InitializeCores();
 
         // Initialize cheat engine
         if (cheat_engine) {
@@ -533,6 +533,7 @@ struct System::Impl {
 
     /// Service State
     Service::Glue::ARPManager arp_manager;
+    Service::Account::ProfileManager profile_manager;
     Service::Time::TimeManager time_manager;
 
     /// Service manager
@@ -600,14 +601,6 @@ bool System::IsPaused() const {
     return impl->IsPaused();
 }
 
-void System::InvalidateCpuInstructionCaches() {
-    impl->kernel.InvalidateAllInstructionCaches();
-}
-
-void System::InvalidateCpuInstructionCacheRange(u64 addr, std::size_t size) {
-    impl->kernel.InvalidateCpuInstructionCacheRange(addr, size);
-}
-
 void System::ShutdownMainProcess() {
     impl->ShutdownMainProcess();
 }
@@ -696,14 +689,6 @@ const TelemetrySession& System::TelemetrySession() const {
     return *impl->telemetry_session;
 }
 
-ARM_Interface& System::CurrentArmInterface() {
-    return impl->kernel.CurrentPhysicalCore().ArmInterface();
-}
-
-const ARM_Interface& System::CurrentArmInterface() const {
-    return impl->kernel.CurrentPhysicalCore().ArmInterface();
-}
-
 Kernel::PhysicalCore& System::CurrentPhysicalCore() {
     return impl->kernel.CurrentPhysicalCore();
 }
@@ -738,14 +723,6 @@ const Kernel::KProcess* System::ApplicationProcess() const {
     return impl->kernel.ApplicationProcess();
 }
 
-ARM_Interface& System::ArmInterface(std::size_t core_index) {
-    return impl->kernel.PhysicalCore(core_index).ArmInterface();
-}
-
-const ARM_Interface& System::ArmInterface(std::size_t core_index) const {
-    return impl->kernel.PhysicalCore(core_index).ArmInterface();
-}
-
 ExclusiveMonitor& System::Monitor() {
     return impl->kernel.GetExclusiveMonitor();
 }
@@ -946,6 +923,14 @@ const Service::APM::Controller& System::GetAPMController() const {
     return impl->apm_controller;
 }
 
+Service::Account::ProfileManager& System::GetProfileManager() {
+    return impl->profile_manager;
+}
+
+const Service::Account::ProfileManager& System::GetProfileManager() const {
+    return impl->profile_manager;
+}
+
 Service::Time::TimeManager& System::GetTimeManager() {
     return impl->time_manager;
 }

src/core/core.h

@@ -45,6 +45,10 @@ class Memory;
 
 namespace Service {
 
+namespace Account {
+class ProfileManager;
+} // namespace Account
+
 namespace AM::Applets {
 struct AppletFrontendSet;
 class AppletManager;
@@ -108,7 +112,6 @@ class RenderdocAPI;
 
 namespace Core {
 
-class ARM_Interface;
 class CpuManager;
 class Debugger;
 class DeviceMemory;
@@ -171,15 +174,6 @@ public:
     /// Check if the core is currently paused.
     [[nodiscard]] bool IsPaused() const;
 
-    /**
-     * Invalidate the CPU instruction caches
-     * This function should only be used by GDB Stub to support breakpoints, memory updates and
-     * step/continue commands.
-     */
-    void InvalidateCpuInstructionCaches();
-
-    void InvalidateCpuInstructionCacheRange(u64 addr, std::size_t size);
-
     /// Shutdown the main emulated process.
     void ShutdownMainProcess();
 
@@ -244,24 +238,12 @@ public:
     /// Gets and resets core performance statistics
     [[nodiscard]] PerfStatsResults GetAndResetPerfStats();
 
-    /// Gets an ARM interface to the CPU core that is currently running
-    [[nodiscard]] ARM_Interface& CurrentArmInterface();
-
-    /// Gets an ARM interface to the CPU core that is currently running
-    [[nodiscard]] const ARM_Interface& CurrentArmInterface() const;
-
     /// Gets the physical core for the CPU core that is currently running
     [[nodiscard]] Kernel::PhysicalCore& CurrentPhysicalCore();
 
     /// Gets the physical core for the CPU core that is currently running
     [[nodiscard]] const Kernel::PhysicalCore& CurrentPhysicalCore() const;
 
-    /// Gets a reference to an ARM interface for the CPU core with the specified index
-    [[nodiscard]] ARM_Interface& ArmInterface(std::size_t core_index);
-
-    /// Gets a const reference to an ARM interface from the CPU core with the specified index
-    [[nodiscard]] const ARM_Interface& ArmInterface(std::size_t core_index) const;
-
     /// Gets a reference to the underlying CPU manager.
     [[nodiscard]] CpuManager& GetCpuManager();
 
@@ -405,6 +387,9 @@ public:
     [[nodiscard]] Service::APM::Controller& GetAPMController();
     [[nodiscard]] const Service::APM::Controller& GetAPMController() const;
 
+    [[nodiscard]] Service::Account::ProfileManager& GetProfileManager();
+    [[nodiscard]] const Service::Account::ProfileManager& GetProfileManager() const;
+
     [[nodiscard]] Service::Time::TimeManager& GetTimeManager();
     [[nodiscard]] const Service::Time::TimeManager& GetTimeManager() const;
 

src/core/cpu_manager.cpp

@@ -73,12 +73,13 @@ void CpuManager::HandleInterrupt() {
 void CpuManager::MultiCoreRunGuestThread() {
     // Similar to UserModeThreadStarter in HOS
     auto& kernel = system.Kernel();
+    auto* thread = Kernel::GetCurrentThreadPointer(kernel);
     kernel.CurrentScheduler()->OnThreadStart();
 
     while (true) {
         auto* physical_core = &kernel.CurrentPhysicalCore();
         while (!physical_core->IsInterrupted()) {
-            physical_core->Run();
+            physical_core->RunThread(thread);
             physical_core = &kernel.CurrentPhysicalCore();
         }
 
@@ -110,12 +111,13 @@ void CpuManager::MultiCoreRunIdleThread() {
 
 void CpuManager::SingleCoreRunGuestThread() {
     auto& kernel = system.Kernel();
+    auto* thread = Kernel::GetCurrentThreadPointer(kernel);
     kernel.CurrentScheduler()->OnThreadStart();
 
     while (true) {
         auto* physical_core = &kernel.CurrentPhysicalCore();
         if (!physical_core->IsInterrupted()) {
-            physical_core->Run();
+            physical_core->RunThread(thread);
             physical_core = &kernel.CurrentPhysicalCore();
         }
 
@@ -211,8 +213,6 @@ void CpuManager::RunThread(std::stop_token token, std::size_t core) {
         system.GPU().ObtainContext();
     }
 
-    system.ArmInterface(core).Initialize();
-
     auto& kernel = system.Kernel();
     auto& scheduler = *kernel.CurrentScheduler();
     auto* thread = scheduler.GetSchedulerCurrentThread();

src/core/debugger/gdbstub.cpp

@@ -16,6 +16,7 @@
 #include "common/settings.h"
 #include "common/string_util.h"
 #include "core/arm/arm_interface.h"
+#include "core/arm/debug.h"
 #include "core/core.h"
 #include "core/debugger/gdbstub.h"
 #include "core/debugger/gdbstub_arch.h"
@@ -310,7 +311,7 @@ void GDBStub::ExecuteCommand(std::string_view packet, std::vector<DebuggerAction
         const auto mem{Common::HexStringToVector(mem_substr, false)};
 
         if (system.ApplicationMemory().WriteBlock(addr, mem.data(), size)) {
-            system.InvalidateCpuInstructionCacheRange(addr, size);
+            Core::InvalidateInstructionCacheRange(system.ApplicationProcess(), addr, size);
             SendReply(GDB_STUB_REPLY_OK);
         } else {
             SendReply(GDB_STUB_REPLY_ERR);
@@ -363,7 +364,7 @@ void GDBStub::HandleBreakpointInsert(std::string_view command) {
     case BreakpointType::Software:
         replaced_instructions[addr] = system.ApplicationMemory().Read32(addr);
         system.ApplicationMemory().Write32(addr, arch->BreakpointInstruction());
-        system.InvalidateCpuInstructionCacheRange(addr, sizeof(u32));
+        Core::InvalidateInstructionCacheRange(system.ApplicationProcess(), addr, sizeof(u32));
         success = true;
         break;
     case BreakpointType::WriteWatch:
@@ -411,7 +412,7 @@ void GDBStub::HandleBreakpointRemove(std::string_view command) {
         const auto orig_insn{replaced_instructions.find(addr)};
         if (orig_insn != replaced_instructions.end()) {
             system.ApplicationMemory().Write32(addr, orig_insn->second);
-            system.InvalidateCpuInstructionCacheRange(addr, sizeof(u32));
+            Core::InvalidateInstructionCacheRange(system.ApplicationProcess(), addr, sizeof(u32));
             replaced_instructions.erase(addr);
             success = true;
         }
@@ -442,114 +443,6 @@ void GDBStub::HandleBreakpointRemove(std::string_view command) {
     }
 }
 
-// Structure offsets are from Atmosphere
-// See osdbg_thread_local_region.os.horizon.hpp and osdbg_thread_type.os.horizon.hpp
-
-static std::optional<std::string> GetNameFromThreadType32(Core::Memory::Memory& memory,
-                                                          const Kernel::KThread& thread) {
-    // Read thread type from TLS
-    const VAddr tls_thread_type{memory.Read32(thread.GetTlsAddress() + 0x1fc)};
-    const VAddr argument_thread_type{thread.GetArgument()};
-
-    if (argument_thread_type && tls_thread_type != argument_thread_type) {
-        // Probably not created by nnsdk, no name available.
-        return std::nullopt;
-    }
-
-    if (!tls_thread_type) {
-        return std::nullopt;
-    }
-
-    const u16 version{memory.Read16(tls_thread_type + 0x26)};
-    VAddr name_pointer{};
-    if (version == 1) {
-        name_pointer = memory.Read32(tls_thread_type + 0xe4);
-    } else {
-        name_pointer = memory.Read32(tls_thread_type + 0xe8);
-    }
-
-    if (!name_pointer) {
-        // No name provided.
-        return std::nullopt;
-    }
-
-    return memory.ReadCString(name_pointer, 256);
-}
-
-static std::optional<std::string> GetNameFromThreadType64(Core::Memory::Memory& memory,
-                                                          const Kernel::KThread& thread) {
-    // Read thread type from TLS
-    const VAddr tls_thread_type{memory.Read64(thread.GetTlsAddress() + 0x1f8)};
-    const VAddr argument_thread_type{thread.GetArgument()};
-
-    if (argument_thread_type && tls_thread_type != argument_thread_type) {
-        // Probably not created by nnsdk, no name available.
-        return std::nullopt;
-    }
-
-    if (!tls_thread_type) {
-        return std::nullopt;
-    }
-
-    const u16 version{memory.Read16(tls_thread_type + 0x46)};
-    VAddr name_pointer{};
-    if (version == 1) {
-        name_pointer = memory.Read64(tls_thread_type + 0x1a0);
-    } else {
-        name_pointer = memory.Read64(tls_thread_type + 0x1a8);
-    }
-
-    if (!name_pointer) {
-        // No name provided.
-        return std::nullopt;
-    }
-
-    return memory.ReadCString(name_pointer, 256);
-}
-
-static std::optional<std::string> GetThreadName(Core::System& system,
-                                                const Kernel::KThread& thread) {
-    if (system.ApplicationProcess()->Is64Bit()) {
-        return GetNameFromThreadType64(system.ApplicationMemory(), thread);
-    } else {
-        return GetNameFromThreadType32(system.ApplicationMemory(), thread);
-    }
-}
-
-static std::string_view GetThreadWaitReason(const Kernel::KThread& thread) {
-    switch (thread.GetWaitReasonForDebugging()) {
-    case Kernel::ThreadWaitReasonForDebugging::Sleep:
-        return "Sleep";
-    case Kernel::ThreadWaitReasonForDebugging::IPC:
-        return "IPC";
-    case Kernel::ThreadWaitReasonForDebugging::Synchronization:
-        return "Synchronization";
-    case Kernel::ThreadWaitReasonForDebugging::ConditionVar:
-        return "ConditionVar";
-    case Kernel::ThreadWaitReasonForDebugging::Arbitration:
-        return "Arbitration";
-    case Kernel::ThreadWaitReasonForDebugging::Suspended:
-        return "Suspended";
-    default:
-        return "Unknown";
-    }
-}
-
-static std::string GetThreadState(const Kernel::KThread& thread) {
-    switch (thread.GetState()) {
-    case Kernel::ThreadState::Initialized:
-        return "Initialized";
-    case Kernel::ThreadState::Waiting:
-        return fmt::format("Waiting ({})", GetThreadWaitReason(thread));
-    case Kernel::ThreadState::Runnable:
-        return "Runnable";
-    case Kernel::ThreadState::Terminated:
-        return "Terminated";
-    default:
-        return "Unknown";
-    }
-}
-
 static std::string PaginateBuffer(std::string_view buffer, std::string_view request) {
     const auto amount{request.substr(request.find(',') + 1)};
     const auto offset_val{static_cast<u64>(strtoll(request.data(), nullptr, 16))};
@@ -562,120 +455,6 @@ static std::string PaginateBuffer(std::string_view buffer, std::string_view requ
     }
 }
 
-static VAddr GetModuleEnd(Kernel::KProcessPageTable& page_table, VAddr base) {
-    Kernel::KMemoryInfo mem_info;
-    Kernel::Svc::MemoryInfo svc_mem_info;
-    Kernel::Svc::PageInfo page_info;
-    VAddr cur_addr{base};
-
-    // Expect: r-x Code (.text)
-    R_ASSERT(page_table.QueryInfo(std::addressof(mem_info), std::addressof(page_info), cur_addr));
-    svc_mem_info = mem_info.GetSvcMemoryInfo();
-    cur_addr = svc_mem_info.base_address + svc_mem_info.size;
-    if (svc_mem_info.state != Kernel::Svc::MemoryState::Code ||
-        svc_mem_info.permission != Kernel::Svc::MemoryPermission::ReadExecute) {
-        return cur_addr - 1;
-    }
-
-    // Expect: r-- Code (.rodata)
-    R_ASSERT(page_table.QueryInfo(std::addressof(mem_info), std::addressof(page_info), cur_addr));
-    svc_mem_info = mem_info.GetSvcMemoryInfo();
-    cur_addr = svc_mem_info.base_address + svc_mem_info.size;
-    if (svc_mem_info.state != Kernel::Svc::MemoryState::Code ||
-        svc_mem_info.permission != Kernel::Svc::MemoryPermission::Read) {
-        return cur_addr - 1;
-    }
-
-    // Expect: rw- CodeData (.data)
-    R_ASSERT(page_table.QueryInfo(std::addressof(mem_info), std::addressof(page_info), cur_addr));
-    svc_mem_info = mem_info.GetSvcMemoryInfo();
-    cur_addr = svc_mem_info.base_address + svc_mem_info.size;
-    return cur_addr - 1;
-}
-
-static Loader::AppLoader::Modules FindModules(Core::System& system) {
-    Loader::AppLoader::Modules modules;
-
-    auto& page_table = system.ApplicationProcess()->GetPageTable();
-    auto& memory = system.ApplicationMemory();
-    VAddr cur_addr = 0;
-
-    // Look for executable sections in Code or AliasCode regions.
-    while (true) {
-        Kernel::KMemoryInfo mem_info{};
-        Kernel::Svc::PageInfo page_info{};
-        R_ASSERT(
-            page_table.QueryInfo(std::addressof(mem_info), std::addressof(page_info), cur_addr));
-        auto svc_mem_info = mem_info.GetSvcMemoryInfo();
-
-        if (svc_mem_info.permission == Kernel::Svc::MemoryPermission::ReadExecute &&
-            (svc_mem_info.state == Kernel::Svc::MemoryState::Code ||
-             svc_mem_info.state == Kernel::Svc::MemoryState::AliasCode)) {
-            // Try to read the module name from its path.
-            constexpr s32 PathLengthMax = 0x200;
-            struct {
-                u32 zero;
-                s32 path_length;
-                std::array<char, PathLengthMax> path;
-            } module_path;
-
-            if (memory.ReadBlock(svc_mem_info.base_address + svc_mem_info.size, &module_path,
-                                 sizeof(module_path))) {
-                if (module_path.zero == 0 && module_path.path_length > 0) {
-                    // Truncate module name.
-                    module_path.path[PathLengthMax - 1] = '\0';
-
-                    // Ignore leading directories.
-                    char* path_pointer = module_path.path.data();
-
-                    for (s32 i = 0; i < std::min(PathLengthMax, module_path.path_length) &&
-                                    module_path.path[i] != '\0';
-                         i++) {
-                        if (module_path.path[i] == '/' || module_path.path[i] == '\\') {
-                            path_pointer = module_path.path.data() + i + 1;
-                        }
-                    }
-
-                    // Insert output.
-                    modules.emplace(svc_mem_info.base_address, path_pointer);
-                }
-            }
-        }
-
-        // Check if we're done.
-        const uintptr_t next_address = svc_mem_info.base_address + svc_mem_info.size;
-        if (next_address <= cur_addr) {
-            break;
-        }
-
-        cur_addr = next_address;
-    }
-
-    return modules;
-}
-
-static VAddr FindMainModuleEntrypoint(Core::System& system) {
-    Loader::AppLoader::Modules modules;
-    system.GetAppLoader().ReadNSOModules(modules);
-
-    // Do we have a module named main?
-    const auto main = std::find_if(modules.begin(), modules.end(),
-                                   [](const auto& key) { return key.second == "main"; });
-
-    if (main != modules.end()) {
-        return main->first;
-    }
-
-    // Do we have any loaded executable sections?
-    modules = FindModules(system);
-    if (!modules.empty()) {
-        return modules.begin()->first;
-    }
-
-    // As a last resort, use the start of the code region.
-    return GetInteger(system.ApplicationProcess()->GetPageTable().GetCodeRegionStart());
-}
-
 void GDBStub::HandleQuery(std::string_view command) {
     if (command.starts_with("TStatus")) {
         // no tracepoint support
@@ -687,10 +466,10 @@ void GDBStub::HandleQuery(std::string_view command) {
         const auto target_xml{arch->GetTargetXML()};
         SendReply(PaginateBuffer(target_xml, command.substr(30)));
     } else if (command.starts_with("Offsets")) {
-        const auto main_offset = FindMainModuleEntrypoint(system);
-        SendReply(fmt::format("TextSeg={:x}", main_offset));
+        const auto main_offset = Core::FindMainModuleEntrypoint(system.ApplicationProcess());
+        SendReply(fmt::format("TextSeg={:x}", GetInteger(main_offset)));
     } else if (command.starts_with("Xfer:libraries:read::")) {
-        auto modules = FindModules(system);
+        auto modules = Core::FindModules(system.ApplicationProcess());
 
         std::string buffer;
         buffer += R"(<?xml version="1.0"?>)";
@@ -720,14 +499,14 @@ void GDBStub::HandleQuery(std::string_view command) {
 
         const auto& threads = system.ApplicationProcess()->GetThreadList();
         for (const auto& thread : threads) {
-            auto thread_name{GetThreadName(system, thread)};
+            auto thread_name{Core::GetThreadName(&thread)};
             if (!thread_name) {
                 thread_name = fmt::format("Thread {:d}", thread.GetThreadId());
             }
 
             buffer += fmt::format(R"(<thread id="{:x}" core="{:d}" name="{}">{}</thread>)",
                                   thread.GetThreadId(), thread.GetActiveCore(),
-                                  EscapeXML(*thread_name), GetThreadState(thread));
+                                  EscapeXML(*thread_name), GetThreadState(&thread));
         }
 
         buffer += "</threads>";
@@ -856,7 +635,7 @@ void GDBStub::HandleRcmd(const std::vector<u8>& command) {
             reply = "Fastmem is not enabled.\n";
         }
     } else if (command_str == "get info") {
-        auto modules = FindModules(system);
+        auto modules = Core::FindModules(process);
 
         reply = fmt::format("Process:     {:#x} ({})\n"
                             "Program Id:  {:#018x}\n",
@@ -880,7 +659,7 @@ void GDBStub::HandleRcmd(const std::vector<u8>& command) {
 
         for (const auto& [vaddr, name] : modules) {
             reply += fmt::format("  {:#012x} - {:#012x} {}\n", vaddr,
-                                 GetModuleEnd(page_table, vaddr), name);
+                                 GetInteger(Core::GetModuleEnd(process, vaddr)), name);
         }
     } else if (command_str == "get mappings") {
         reply = "Mappings:\n";

src/core/debugger/gdbstub_arch.cpp

@@ -24,21 +24,6 @@ static std::string ValueToHex(const T value) {
     return Common::HexToString(mem);
 }
 
-template <typename T>
-static T GetSIMDRegister(const std::array<u32, 64>& simd_regs, size_t offset) {
-    static_assert(std::is_trivially_copyable_v<T>);
-    T value{};
-    std::memcpy(&value, reinterpret_cast<const u8*>(simd_regs.data()) + sizeof(T) * offset,
-                sizeof(T));
-    return value;
-}
-
-template <typename T>
-static void PutSIMDRegister(std::array<u32, 64>& simd_regs, size_t offset, const T value) {
-    static_assert(std::is_trivially_copyable_v<T>);
-    std::memcpy(reinterpret_cast<u8*>(simd_regs.data()) + sizeof(T) * offset, &value, sizeof(T));
-}
-
 // For sample XML files see the GDB source /gdb/features
 // This XML defines what the registers are for this specific ARM device
 std::string_view GDBStubA64::GetTargetXML() const {
@@ -184,12 +169,16 @@ std::string GDBStubA64::RegRead(const Kernel::KThread* thread, size_t id) const
         return "";
     }
 
-    const auto& context{thread->GetContext64()};
-    const auto& gprs{context.cpu_registers};
-    const auto& fprs{context.vector_registers};
+    const auto& context{thread->GetContext()};
+    const auto& gprs{context.r};
+    const auto& fprs{context.v};
 
-    if (id < SP_REGISTER) {
+    if (id < FP_REGISTER) {
         return ValueToHex(gprs[id]);
+    } else if (id == FP_REGISTER) {
+        return ValueToHex(context.fp);
+    } else if (id == LR_REGISTER) {
+        return ValueToHex(context.lr);
     } else if (id == SP_REGISTER) {
         return ValueToHex(context.sp);
     } else if (id == PC_REGISTER) {
@@ -212,10 +201,14 @@ void GDBStubA64::RegWrite(Kernel::KThread* thread, size_t id, std::string_view v
         return;
     }
 
-    auto& context{thread->GetContext64()};
+    auto& context{thread->GetContext()};
 
-    if (id < SP_REGISTER) {
-        context.cpu_registers[id] = HexToValue<u64>(value);
+    if (id < FP_REGISTER) {
+        context.r[id] = HexToValue<u64>(value);
+    } else if (id == FP_REGISTER) {
+        context.fp = HexToValue<u64>(value);
+    } else if (id == LR_REGISTER) {
+        context.lr = HexToValue<u64>(value);
     } else if (id == SP_REGISTER) {
         context.sp = HexToValue<u64>(value);
     } else if (id == PC_REGISTER) {
@@ -223,7 +216,7 @@ void GDBStubA64::RegWrite(Kernel::KThread* thread, size_t id, std::string_view v
     } else if (id == PSTATE_REGISTER) {
         context.pstate = HexToValue<u32>(value);
     } else if (id >= Q0_REGISTER && id < FPSR_REGISTER) {
-        context.vector_registers[id - Q0_REGISTER] = HexToValue<u128>(value);
+        context.v[id - Q0_REGISTER] = HexToValue<u128>(value);
     } else if (id == FPSR_REGISTER) {
         context.fpsr = HexToValue<u32>(value);
     } else if (id == FPCR_REGISTER) {
@@ -381,22 +374,20 @@ std::string GDBStubA32::RegRead(const Kernel::KThread* thread, size_t id) const
         return "";
     }
 
-    const auto& context{thread->GetContext32()};
-    const auto& gprs{context.cpu_registers};
-    const auto& fprs{context.extension_registers};
+    const auto& context{thread->GetContext()};
+    const auto& gprs{context.r};
+    const auto& fprs{context.v};
 
     if (id <= PC_REGISTER) {
-        return ValueToHex(gprs[id]);
+        return ValueToHex(static_cast<u32>(gprs[id]));
     } else if (id == CPSR_REGISTER) {
-        return ValueToHex(context.cpsr);
+        return ValueToHex(context.pstate);
     } else if (id >= D0_REGISTER && id < Q0_REGISTER) {
-        const u64 dN{GetSIMDRegister<u64>(fprs, id - D0_REGISTER)};
-        return ValueToHex(dN);
+        return ValueToHex(fprs[id - D0_REGISTER][0]);
     } else if (id >= Q0_REGISTER && id < FPSCR_REGISTER) {
-        const u128 qN{GetSIMDRegister<u128>(fprs, id - Q0_REGISTER)};
-        return ValueToHex(qN);
+        return ValueToHex(fprs[id - Q0_REGISTER]);
     } else if (id == FPSCR_REGISTER) {
-        return ValueToHex(context.fpscr);
+        return ValueToHex(context.fpcr | context.fpsr);
     } else {
         return "";
     }
@@ -407,19 +398,20 @@ void GDBStubA32::RegWrite(Kernel::KThread* thread, size_t id, std::string_view v
         return;
     }
 
-    auto& context{thread->GetContext32()};
-    auto& fprs{context.extension_registers};
+    auto& context{thread->GetContext()};
+    auto& fprs{context.v};
 
     if (id <= PC_REGISTER) {
-        context.cpu_registers[id] = HexToValue<u32>(value);
+        context.r[id] = HexToValue<u32>(value);
     } else if (id == CPSR_REGISTER) {
-        context.cpsr = HexToValue<u32>(value);
+        context.pstate = HexToValue<u32>(value);
     } else if (id >= D0_REGISTER && id < Q0_REGISTER) {
-        PutSIMDRegister(fprs, id - D0_REGISTER, HexToValue<u64>(value));
+        fprs[id - D0_REGISTER] = {HexToValue<u64>(value), 0};
     } else if (id >= Q0_REGISTER && id < FPSCR_REGISTER) {
-        PutSIMDRegister(fprs, id - Q0_REGISTER, HexToValue<u128>(value));
+        fprs[id - Q0_REGISTER] = HexToValue<u128>(value);
     } else if (id == FPSCR_REGISTER) {
-        context.fpscr = HexToValue<u32>(value);
+        context.fpcr = HexToValue<u32>(value);
+        context.fpsr = HexToValue<u32>(value);
     }
 }
 

src/core/debugger/gdbstub_arch.h

@@ -36,6 +36,7 @@ public:
     u32 BreakpointInstruction() const override;
 
 private:
+    static constexpr u32 FP_REGISTER = 29;
     static constexpr u32 LR_REGISTER = 30;
     static constexpr u32 SP_REGISTER = 31;
     static constexpr u32 PC_REGISTER = 32;

src/core/file_sys/fsmitm_romfsbuild.cpp

@@ -2,6 +2,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include <cstring>
+#include <span>
 #include <string_view>
 #include "common/alignment.h"
 #include "common/assert.h"
@@ -134,7 +135,7 @@ void RomFSBuildContext::VisitDirectory(VirtualDir romfs_dir, VirtualDir ext_dir,
 
         child->size = child->source->GetSize();
 
-        AddFile(parent, child);
+        AddFile(parent, std::move(child));
     }
 
     for (auto& child_romfs_dir : romfs_dir->GetSubdirectories()) {
@@ -163,36 +164,24 @@ void RomFSBuildContext::VisitDirectory(VirtualDir romfs_dir, VirtualDir ext_dir,
 
 bool RomFSBuildContext::AddDirectory(std::shared_ptr<RomFSBuildDirectoryContext> parent_dir_ctx,
                                      std::shared_ptr<RomFSBuildDirectoryContext> dir_ctx) {
-    // Check whether it's already in the known directories.
-    const auto [it, is_new] = directories.emplace(dir_ctx->path, nullptr);
-    if (!is_new) {
-        return false;
-    }
-
     // Add a new directory.
     num_dirs++;
     dir_table_size +=
         sizeof(RomFSDirectoryEntry) + Common::AlignUp(dir_ctx->path_len - dir_ctx->cur_path_ofs, 4);
-    dir_ctx->parent = parent_dir_ctx;
-    it->second = dir_ctx;
+    dir_ctx->parent = std::move(parent_dir_ctx);
+    directories.emplace_back(std::move(dir_ctx));
 
     return true;
 }
 
 bool RomFSBuildContext::AddFile(std::shared_ptr<RomFSBuildDirectoryContext> parent_dir_ctx,
                                 std::shared_ptr<RomFSBuildFileContext> file_ctx) {
-    // Check whether it's already in the known files.
-    const auto [it, is_new] = files.emplace(file_ctx->path, nullptr);
-    if (!is_new) {
-        return false;
-    }
-
     // Add a new file.
     num_files++;
     file_table_size +=
         sizeof(RomFSFileEntry) + Common::AlignUp(file_ctx->path_len - file_ctx->cur_path_ofs, 4);
-    file_ctx->parent = parent_dir_ctx;
-    it->second = file_ctx;
+    file_ctx->parent = std::move(parent_dir_ctx);
+    files.emplace_back(std::move(file_ctx));
 
     return true;
 }
@@ -201,7 +190,7 @@ RomFSBuildContext::RomFSBuildContext(VirtualDir base_, VirtualDir ext_)
     : base(std::move(base_)), ext(std::move(ext_)) {
     root = std::make_shared<RomFSBuildDirectoryContext>();
     root->path = "\0";
-    directories.emplace(root->path, root);
+    directories.emplace_back(root);
     num_dirs = 1;
     dir_table_size = 0x18;
 
@@ -210,28 +199,43 @@ RomFSBuildContext::RomFSBuildContext(VirtualDir base_, VirtualDir ext_)
 
 RomFSBuildContext::~RomFSBuildContext() = default;
 
-std::multimap<u64, VirtualFile> RomFSBuildContext::Build() {
+std::vector<std::pair<u64, VirtualFile>> RomFSBuildContext::Build() {
     const u64 dir_hash_table_entry_count = romfs_get_hash_table_count(num_dirs);
     const u64 file_hash_table_entry_count = romfs_get_hash_table_count(num_files);
     dir_hash_table_size = 4 * dir_hash_table_entry_count;
     file_hash_table_size = 4 * file_hash_table_entry_count;
 
-    // Assign metadata pointers
+    // Assign metadata pointers.
     RomFSHeader header{};
 
-    std::vector<u32> dir_hash_table(dir_hash_table_entry_count, ROMFS_ENTRY_EMPTY);
-    std::vector<u32> file_hash_table(file_hash_table_entry_count, ROMFS_ENTRY_EMPTY);
+    std::vector<u8> metadata(file_hash_table_size + file_table_size + dir_hash_table_size +
+                             dir_table_size);
+    u32* const dir_hash_table_pointer = reinterpret_cast<u32*>(metadata.data());
+    u8* const dir_table_pointer = metadata.data() + dir_hash_table_size;
+    u32* const file_hash_table_pointer =
+        reinterpret_cast<u32*>(metadata.data() + dir_hash_table_size + dir_table_size);
+    u8* const file_table_pointer =
+        metadata.data() + dir_hash_table_size + dir_table_size + file_hash_table_size;
 
-    std::vector<u8> dir_table(dir_table_size);
-    std::vector<u8> file_table(file_table_size);
+    std::span<u32> dir_hash_table(dir_hash_table_pointer, dir_hash_table_entry_count);
+    std::span<u32> file_hash_table(file_hash_table_pointer, file_hash_table_entry_count);
+    std::span<u8> dir_table(dir_table_pointer, dir_table_size);
+    std::span<u8> file_table(file_table_pointer, file_table_size);
 
-    std::shared_ptr<RomFSBuildFileContext> cur_file;
+    // Initialize hash tables.
+    std::memset(dir_hash_table.data(), 0xFF, dir_hash_table.size_bytes());
+    std::memset(file_hash_table.data(), 0xFF, file_hash_table.size_bytes());
+
+    // Sort tables by name.
+    std::sort(files.begin(), files.end(),
+              [](const auto& a, const auto& b) { return a->path < b->path; });
+    std::sort(directories.begin(), directories.end(),
+              [](const auto& a, const auto& b) { return a->path < b->path; });
 
     // Determine file offsets.
     u32 entry_offset = 0;
     std::shared_ptr<RomFSBuildFileContext> prev_file = nullptr;
-    for (const auto& it : files) {
-        cur_file = it.second;
+    for (const auto& cur_file : files) {
         file_partition_size = Common::AlignUp(file_partition_size, 16);
         cur_file->offset = file_partition_size;
         file_partition_size += cur_file->size;
@@ -243,34 +247,48 @@ std::multimap<u64, VirtualFile> RomFSBuildContext::Build() {
     }
     // Assign deferred parent/sibling ownership.
     for (auto it = files.rbegin(); it != files.rend(); ++it) {
-        cur_file = it->second;
+        auto& cur_file = *it;
         cur_file->sibling = cur_file->parent->file;
         cur_file->parent->file = cur_file;
     }
 
-    std::shared_ptr<RomFSBuildDirectoryContext> cur_dir;
-
     // Determine directory offsets.
     entry_offset = 0;
-    for (const auto& it : directories) {
-        cur_dir = it.second;
+    for (const auto& cur_dir : directories) {
         cur_dir->entry_offset = entry_offset;
         entry_offset +=
             static_cast<u32>(sizeof(RomFSDirectoryEntry) +
                              Common::AlignUp(cur_dir->path_len - cur_dir->cur_path_ofs, 4));
     }
     // Assign deferred parent/sibling ownership.
-    for (auto it = directories.rbegin(); it->second != root; ++it) {
-        cur_dir = it->second;
+    for (auto it = directories.rbegin(); (*it) != root; ++it) {
+        auto& cur_dir = *it;
         cur_dir->sibling = cur_dir->parent->child;
         cur_dir->parent->child = cur_dir;
     }
 
-    std::multimap<u64, VirtualFile> out;
+    // Create output map.
+    std::vector<std::pair<u64, VirtualFile>> out;
+    out.reserve(num_files + 2);
+
+    // Set header fields.
+    header.header_size = sizeof(RomFSHeader);
+    header.file_hash_table_size = file_hash_table_size;
+    header.file_table_size = file_table_size;
+    header.dir_hash_table_size = dir_hash_table_size;
+    header.dir_table_size = dir_table_size;
+    header.file_partition_ofs = ROMFS_FILEPARTITION_OFS;
+    header.dir_hash_table_ofs = Common::AlignUp(header.file_partition_ofs + file_partition_size, 4);
+    header.dir_table_ofs = header.dir_hash_table_ofs + header.dir_hash_table_size;
+    header.file_hash_table_ofs = header.dir_table_ofs + header.dir_table_size;
+    header.file_table_ofs = header.file_hash_table_ofs + header.file_hash_table_size;
+
+    std::vector<u8> header_data(sizeof(RomFSHeader));
+    std::memcpy(header_data.data(), &header, header_data.size());
+    out.emplace_back(0, std::make_shared<VectorVfsFile>(std::move(header_data)));
 
     // Populate file tables.
-    for (const auto& it : files) {
-        cur_file = it.second;
+    for (const auto& cur_file : files) {
         RomFSFileEntry cur_entry{};
 
         cur_entry.parent = cur_file->parent->entry_offset;
@@ -287,7 +305,7 @@ std::multimap<u64, VirtualFile> RomFSBuildContext::Build() {
 
         cur_entry.name_size = name_size;
 
-        out.emplace(cur_file->offset + ROMFS_FILEPARTITION_OFS, std::move(cur_file->source));
+        out.emplace_back(cur_file->offset + ROMFS_FILEPARTITION_OFS, std::move(cur_file->source));
         std::memcpy(file_table.data() + cur_file->entry_offset, &cur_entry, sizeof(RomFSFileEntry));
         std::memset(file_table.data() + cur_file->entry_offset + sizeof(RomFSFileEntry), 0,
                     Common::AlignUp(cur_entry.name_size, 4));
@@ -296,8 +314,7 @@ std::multimap<u64, VirtualFile> RomFSBuildContext::Build() {
     }
 
     // Populate dir tables.
-    for (const auto& it : directories) {
-        cur_dir = it.second;
+    for (const auto& cur_dir : directories) {
         RomFSDirectoryEntry cur_entry{};
 
         cur_entry.parent = cur_dir == root ? 0 : cur_dir->parent->entry_offset;
@@ -323,34 +340,13 @@ std::multimap<u64, VirtualFile> RomFSBuildContext::Build() {
                     cur_dir->path.data() + cur_dir->cur_path_ofs, name_size);
     }
 
-    // Set header fields.
-    header.header_size = sizeof(RomFSHeader);
-    header.file_hash_table_size = file_hash_table_size;
-    header.file_table_size = file_table_size;
-    header.dir_hash_table_size = dir_hash_table_size;
-    header.dir_table_size = dir_table_size;
-    header.file_partition_ofs = ROMFS_FILEPARTITION_OFS;
-    header.dir_hash_table_ofs = Common::AlignUp(header.file_partition_ofs + file_partition_size, 4);
-    header.dir_table_ofs = header.dir_hash_table_ofs + header.dir_hash_table_size;
-    header.file_hash_table_ofs = header.dir_table_ofs + header.dir_table_size;
-    header.file_table_ofs = header.file_hash_table_ofs + header.file_hash_table_size;
+    // Write metadata.
+    out.emplace_back(header.dir_hash_table_ofs,
+                     std::make_shared<VectorVfsFile>(std::move(metadata)));
 
-    std::vector<u8> header_data(sizeof(RomFSHeader));
-    std::memcpy(header_data.data(), &header, header_data.size());
-    out.emplace(0, std::make_shared<VectorVfsFile>(std::move(header_data)));
-
-    std::vector<u8> metadata(file_hash_table_size + file_table_size + dir_hash_table_size +
-                             dir_table_size);
-    std::size_t index = 0;
-    std::memcpy(metadata.data(), dir_hash_table.data(), dir_hash_table.size() * sizeof(u32));
-    index += dir_hash_table.size() * sizeof(u32);
-    std::memcpy(metadata.data() + index, dir_table.data(), dir_table.size());
-    index += dir_table.size();
-    std::memcpy(metadata.data() + index, file_hash_table.data(),
-                file_hash_table.size() * sizeof(u32));
-    index += file_hash_table.size() * sizeof(u32);
-    std::memcpy(metadata.data() + index, file_table.data(), file_table.size());
-    out.emplace(header.dir_hash_table_ofs, std::make_shared<VectorVfsFile>(std::move(metadata)));
+    // Sort the output.
+    std::sort(out.begin(), out.end(),
+              [](const auto& a, const auto& b) { return a.first < b.first; });
 
     return out;
 }

src/core/file_sys/fsmitm_romfsbuild.h

@@ -22,14 +22,14 @@ public:
     ~RomFSBuildContext();
 
     // This finalizes the context.
-    std::multimap<u64, VirtualFile> Build();
+    std::vector<std::pair<u64, VirtualFile>> Build();
 
 private:
     VirtualDir base;
     VirtualDir ext;
     std::shared_ptr<RomFSBuildDirectoryContext> root;
-    std::map<std::string, std::shared_ptr<RomFSBuildDirectoryContext>, std::less<>> directories;
-    std::map<std::string, std::shared_ptr<RomFSBuildFileContext>, std::less<>> files;
+    std::vector<std::shared_ptr<RomFSBuildDirectoryContext>> directories;
+    std::vector<std::shared_ptr<RomFSBuildFileContext>> files;
     u64 num_dirs = 0;
     u64 num_files = 0;
     u64 dir_table_size = 0;

src/core/file_sys/romfs.cpp

@@ -55,44 +55,68 @@ struct FileEntry {
 };
 static_assert(sizeof(FileEntry) == 0x20, "FileEntry has incorrect size.");
 
-template <typename Entry>
-std::pair<Entry, std::string> GetEntry(const VirtualFile& file, std::size_t offset) {
-    Entry entry{};
-    if (file->ReadObject(&entry, offset) != sizeof(Entry))
+struct RomFSTraversalContext {
+    RomFSHeader header;
+    VirtualFile file;
+    std::vector<u8> directory_meta;
+    std::vector<u8> file_meta;
+};
+
+template <typename EntryType, auto Member>
+std::pair<EntryType, std::string> GetEntry(const RomFSTraversalContext& ctx, size_t offset) {
+    const size_t entry_end = offset + sizeof(EntryType);
+    const std::vector<u8>& vec = ctx.*Member;
+    const size_t size = vec.size();
+    const u8* data = vec.data();
+    EntryType entry{};
+
+    if (entry_end > size) {
         return {};
-    std::string string(entry.name_length, '\0');
-    if (file->ReadArray(&string[0], string.size(), offset + sizeof(Entry)) != string.size())
-        return {};
-    return {entry, string};
+    }
+    std::memcpy(&entry, data + offset, sizeof(EntryType));
+
+    const size_t name_length = std::min(entry_end + entry.name_length, size) - entry_end;
+    std::string name(reinterpret_cast<const char*>(data + entry_end), name_length);
+
+    return {entry, std::move(name)};
 }
 
-void ProcessFile(const VirtualFile& file, std::size_t file_offset, std::size_t data_offset,
-                 u32 this_file_offset, std::shared_ptr<VectorVfsDirectory>& parent) {
-    while (this_file_offset != ROMFS_ENTRY_EMPTY) {
-        auto entry = GetEntry<FileEntry>(file, file_offset + this_file_offset);
+std::pair<DirectoryEntry, std::string> GetDirectoryEntry(const RomFSTraversalContext& ctx,
+                                                         size_t directory_offset) {
+    return GetEntry<DirectoryEntry, &RomFSTraversalContext::directory_meta>(ctx, directory_offset);
+}
 
-        parent->AddFile(std::make_shared<OffsetVfsFile>(
-            file, entry.first.size, entry.first.offset + data_offset, entry.second));
+std::pair<FileEntry, std::string> GetFileEntry(const RomFSTraversalContext& ctx,
+                                               size_t file_offset) {
+    return GetEntry<FileEntry, &RomFSTraversalContext::file_meta>(ctx, file_offset);
+}
+
+void ProcessFile(const RomFSTraversalContext& ctx, u32 this_file_offset,
+                 std::shared_ptr<VectorVfsDirectory>& parent) {
+    while (this_file_offset != ROMFS_ENTRY_EMPTY) {
+        auto entry = GetFileEntry(ctx, this_file_offset);
+
+        parent->AddFile(std::make_shared<OffsetVfsFile>(ctx.file, entry.first.size,
+                                                        entry.first.offset + ctx.header.data_offset,
+                                                        std::move(entry.second)));
 
         this_file_offset = entry.first.sibling;
     }
 }
 
-void ProcessDirectory(const VirtualFile& file, std::size_t dir_offset, std::size_t file_offset,
-                      std::size_t data_offset, u32 this_dir_offset,
+void ProcessDirectory(const RomFSTraversalContext& ctx, u32 this_dir_offset,
                       std::shared_ptr<VectorVfsDirectory>& parent) {
     while (this_dir_offset != ROMFS_ENTRY_EMPTY) {
-        auto entry = GetEntry<DirectoryEntry>(file, dir_offset + this_dir_offset);
+        auto entry = GetDirectoryEntry(ctx, this_dir_offset);
         auto current = std::make_shared<VectorVfsDirectory>(
             std::vector<VirtualFile>{}, std::vector<VirtualDir>{}, entry.second);
 
         if (entry.first.child_file != ROMFS_ENTRY_EMPTY) {
-            ProcessFile(file, file_offset, data_offset, entry.first.child_file, current);
+            ProcessFile(ctx, entry.first.child_file, current);
         }
 
         if (entry.first.child_dir != ROMFS_ENTRY_EMPTY) {
-            ProcessDirectory(file, dir_offset, file_offset, data_offset, entry.first.child_dir,
-                             current);
+            ProcessDirectory(ctx, entry.first.child_dir, current);
         }
 
         parent->AddDirectory(current);
@@ -107,22 +131,25 @@ VirtualDir ExtractRomFS(VirtualFile file) {
         return root_container;
     }
 
-    RomFSHeader header{};
-    if (file->ReadObject(&header) != sizeof(RomFSHeader)) {
-        return root_container;
+    RomFSTraversalContext ctx{};
+
+    if (file->ReadObject(&ctx.header) != sizeof(RomFSHeader)) {
+        return nullptr;
     }
 
-    if (header.header_size != sizeof(RomFSHeader)) {
-        return root_container;
+    if (ctx.header.header_size != sizeof(RomFSHeader)) {
+        return nullptr;
     }
 
-    const u64 file_offset = header.file_meta.offset;
-    const u64 dir_offset = header.directory_meta.offset;
+    ctx.file = file;
+    ctx.directory_meta =
+        file->ReadBytes(ctx.header.directory_meta.size, ctx.header.directory_meta.offset);
+    ctx.file_meta = file->ReadBytes(ctx.header.file_meta.size, ctx.header.file_meta.offset);
 
-    ProcessDirectory(file, dir_offset, file_offset, header.data_offset, 0, root_container);
+    ProcessDirectory(ctx, 0, root_container);
 
     if (auto root = root_container->GetSubdirectory(""); root) {
-        return std::make_shared<CachedVfsDirectory>(std::move(root));
+        return root;
     }
 
     ASSERT(false);

src/core/file_sys/savedata_factory.cpp

@@ -12,8 +12,6 @@
 
 namespace FileSys {
 
-constexpr char SAVE_DATA_SIZE_FILENAME[] = ".yuzu_save_size";
-
 namespace {
 
 void PrintSaveDataAttributeWarnings(SaveDataAttribute meta) {
@@ -197,7 +195,7 @@ SaveDataSize SaveDataFactory::ReadSaveDataSize(SaveDataType type, u64 title_id,
         GetFullPath(system, dir, SaveDataSpaceId::NandUser, type, title_id, user_id, 0);
     const auto relative_dir = GetOrCreateDirectoryRelative(dir, path);
 
-    const auto size_file = relative_dir->GetFile(SAVE_DATA_SIZE_FILENAME);
+    const auto size_file = relative_dir->GetFile(GetSaveDataSizeFileName());
     if (size_file == nullptr || size_file->GetSize() < sizeof(SaveDataSize)) {
         return {0, 0};
     }
@@ -216,7 +214,7 @@ void SaveDataFactory::WriteSaveDataSize(SaveDataType type, u64 title_id, u128 us
         GetFullPath(system, dir, SaveDataSpaceId::NandUser, type, title_id, user_id, 0);
     const auto relative_dir = GetOrCreateDirectoryRelative(dir, path);
 
-    const auto size_file = relative_dir->CreateFile(SAVE_DATA_SIZE_FILENAME);
+    const auto size_file = relative_dir->CreateFile(GetSaveDataSizeFileName());
     if (size_file == nullptr) {
         return;
     }

src/core/file_sys/savedata_factory.h

@@ -83,6 +83,10 @@ struct SaveDataSize {
     u64 journal;
 };
 
+constexpr const char* GetSaveDataSizeFileName() {
+    return ".yuzu_save_size";
+}
+
 /// File system interface to the SaveData archive
 class SaveDataFactory {
 public:

src/core/file_sys/vfs_concat.cpp

@@ -59,8 +59,8 @@ VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(std::string&& name,
     return VirtualFile(new ConcatenatedVfsFile(std::move(name), std::move(concatenation_map)));
 }
 
-VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(u8 filler_byte, std::string&& name,
-                                                      std::multimap<u64, VirtualFile>&& files) {
+VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(
+    u8 filler_byte, std::string&& name, std::vector<std::pair<u64, VirtualFile>>&& files) {
     // Fold trivial cases.
     if (files.empty()) {
         return nullptr;

src/core/file_sys/vfs_concat.h

@@ -37,7 +37,7 @@ public:
     /// Convenience function that turns a map of offsets to files into a concatenated file, filling
     /// gaps with a given filler byte.
     static VirtualFile MakeConcatenatedFile(u8 filler_byte, std::string&& name,
-                                            std::multimap<u64, VirtualFile>&& files);
+                                            std::vector<std::pair<u64, VirtualFile>>&& files);
 
     std::string GetName() const override;
     std::size_t GetSize() const override;

src/core/file_sys/vfs_layered.cpp

@@ -3,6 +3,7 @@
 
 #include <algorithm>
 #include <set>
+#include <unordered_set>
 #include <utility>
 #include "core/file_sys/vfs_layered.h"
 
@@ -59,13 +60,12 @@ std::string LayeredVfsDirectory::GetFullPath() const {
 
 std::vector<VirtualFile> LayeredVfsDirectory::GetFiles() const {
     std::vector<VirtualFile> out;
-    std::set<std::string, std::less<>> out_names;
+    std::unordered_set<std::string> out_names;
 
     for (const auto& layer : dirs) {
         for (auto& file : layer->GetFiles()) {
-            auto file_name = file->GetName();
-            if (!out_names.contains(file_name)) {
-                out_names.emplace(std::move(file_name));
+            const auto [it, is_new] = out_names.emplace(file->GetName());
+            if (is_new) {
                 out.emplace_back(std::move(file));
             }
         }
@@ -75,18 +75,19 @@ std::vector<VirtualFile> LayeredVfsDirectory::GetFiles() const {
 }
 
 std::vector<VirtualDir> LayeredVfsDirectory::GetSubdirectories() const {
-    std::vector<std::string> names;
+    std::vector<VirtualDir> out;
+    std::unordered_set<std::string> out_names;
+
     for (const auto& layer : dirs) {
         for (const auto& sd : layer->GetSubdirectories()) {
-            if (std::find(names.begin(), names.end(), sd->GetName()) == names.end())
-                names.push_back(sd->GetName());
+            out_names.emplace(sd->GetName());
         }
     }
 
-    std::vector<VirtualDir> out;
-    out.reserve(names.size());
-    for (const auto& subdir : names)
+    out.reserve(out_names.size());
+    for (const auto& subdir : out_names) {
         out.emplace_back(GetSubdirectory(subdir));
+    }
 
     return out;
 }

src/core/hid/input_interpreter.cpp

@@ -20,6 +20,9 @@ InputInterpreter::InputInterpreter(Core::System& system)
 InputInterpreter::~InputInterpreter() = default;
 
 void InputInterpreter::PollInput() {
+    if (npad == nullptr) {
+        return;
+    }
     const auto button_state = npad->GetAndResetPressState();
 
     previous_index = current_index;

src/core/hle/kernel/k_client_port.cpp

@@ -3,6 +3,7 @@
 
 #include "common/scope_exit.h"
 #include "core/hle/kernel/k_client_port.h"
+#include "core/hle/kernel/k_light_session.h"
 #include "core/hle/kernel/k_port.h"
 #include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/k_scoped_resource_reservation.h"
@@ -63,6 +64,7 @@ Result KClientPort::CreateSession(KClientSession** out) {
     R_UNLESS(session_reservation.Succeeded(), ResultLimitReached);
 
     // Allocate a session normally.
+    // TODO: Dynamic resource limits
     session = KSession::Create(m_kernel);
 
     // Check that we successfully created a session.
@@ -119,4 +121,71 @@ Result KClientPort::CreateSession(KClientSession** out) {
     R_SUCCEED();
 }
 
+Result KClientPort::CreateLightSession(KLightClientSession** out) {
+    // Declare the session we're going to allocate.
+    KLightSession* session{};
+
+    // Reserve a new session from the resource limit.
+    KScopedResourceReservation session_reservation(GetCurrentProcessPointer(m_kernel),
+                                                   Svc::LimitableResource::SessionCountMax);
+    R_UNLESS(session_reservation.Succeeded(), ResultLimitReached);
+
+    // Allocate a session normally.
+    // TODO: Dynamic resource limits
+    session = KLightSession::Create(m_kernel);
+
+    // Check that we successfully created a session.
+    R_UNLESS(session != nullptr, ResultOutOfResource);
+
+    // Update the session counts.
+    {
+        ON_RESULT_FAILURE {
+            session->Close();
+        };
+
+        // Atomically increment the number of sessions.
+        s32 new_sessions;
+        {
+            const auto max = m_max_sessions;
+            auto cur_sessions = m_num_sessions.load(std::memory_order_acquire);
+            do {
+                R_UNLESS(cur_sessions < max, ResultOutOfSessions);
+                new_sessions = cur_sessions + 1;
+            } while (!m_num_sessions.compare_exchange_weak(cur_sessions, new_sessions,
+                                                           std::memory_order_relaxed));
+        }
+
+        // Atomically update the peak session tracking.
+        {
+            auto peak = m_peak_sessions.load(std::memory_order_acquire);
+            do {
+                if (peak >= new_sessions) {
+                    break;
+                }
+            } while (!m_peak_sessions.compare_exchange_weak(peak, new_sessions,
+                                                            std::memory_order_relaxed));
+        }
+    }
+
+    // Initialize the session.
+    session->Initialize(this, m_parent->GetName());
+
+    // Commit the session reservation.
+    session_reservation.Commit();
+
+    // Register the session.
+    KLightSession::Register(m_kernel, session);
+    ON_RESULT_FAILURE {
+        session->GetClientSession().Close();
+        session->GetServerSession().Close();
+    };
+
+    // Enqueue the session with our parent.
+    R_TRY(m_parent->EnqueueSession(std::addressof(session->GetServerSession())));
+
+    // We succeeded, so set the output.
+    *out = std::addressof(session->GetClientSession());
+    R_SUCCEED();
+}
+
 } // namespace Kernel

src/core/hle/kernel/k_client_port.h

@@ -11,6 +11,7 @@
 
 namespace Kernel {
 
+class KLightClientSession;
 class KClientSession;
 class KernelCore;
 class KPort;
@@ -51,6 +52,7 @@ public:
     bool IsSignaled() const override;
 
     Result CreateSession(KClientSession** out);
+    Result CreateLightSession(KLightClientSession** out);
 
 private:
     std::atomic<s32> m_num_sessions{};

src/core/hle/kernel/k_client_session.cpp

@@ -10,9 +10,7 @@
 
 namespace Kernel {
 
-static constexpr u32 MessageBufferSize = 0x100;
-
-KClientSession::KClientSession(KernelCore& kernel) : KAutoObjectWithSlabHeapAndContainer{kernel} {}
+KClientSession::KClientSession(KernelCore& kernel) : KAutoObject{kernel} {}
 KClientSession::~KClientSession() = default;
 
 void KClientSession::Destroy() {
@@ -22,18 +20,30 @@ void KClientSession::Destroy() {
 
 void KClientSession::OnServerClosed() {}
 
-Result KClientSession::SendSyncRequest() {
+Result KClientSession::SendSyncRequest(uintptr_t address, size_t size) {
     // Create a session request.
     KSessionRequest* request = KSessionRequest::Create(m_kernel);
     R_UNLESS(request != nullptr, ResultOutOfResource);
     SCOPE_EXIT({ request->Close(); });
 
     // Initialize the request.
-    request->Initialize(nullptr, GetInteger(GetCurrentThread(m_kernel).GetTlsAddress()),
-                        MessageBufferSize);
+    request->Initialize(nullptr, address, size);
 
     // Send the request.
-    R_RETURN(m_parent->GetServerSession().OnRequest(request));
+    R_RETURN(m_parent->OnRequest(request));
+}
+
+Result KClientSession::SendAsyncRequest(KEvent* event, uintptr_t address, size_t size) {
+    // Create a session request.
+    KSessionRequest* request = KSessionRequest::Create(m_kernel);
+    R_UNLESS(request != nullptr, ResultOutOfResource);
+    SCOPE_EXIT({ request->Close(); });
+
+    // Initialize the request.
+    request->Initialize(event, address, size);
+
+    // Send the request.
+    R_RETURN(m_parent->OnRequest(request));
 }
 
 } // namespace Kernel

src/core/hle/kernel/k_client_session.h

@@ -9,24 +9,12 @@
 #include "core/hle/kernel/slab_helpers.h"
 #include "core/hle/result.h"
 
-union Result;
-
-namespace Core::Memory {
-class Memory;
-}
-
-namespace Core::Timing {
-class CoreTiming;
-}
-
 namespace Kernel {
 
 class KernelCore;
 class KSession;
-class KThread;
 
-class KClientSession final
-    : public KAutoObjectWithSlabHeapAndContainer<KClientSession, KAutoObjectWithList> {
+class KClientSession final : public KAutoObject {
     KERNEL_AUTOOBJECT_TRAITS(KClientSession, KAutoObject);
 
 public:
@@ -39,13 +27,13 @@ public:
     }
 
     void Destroy() override;
-    static void PostDestroy(uintptr_t arg) {}
 
     KSession* GetParent() const {
         return m_parent;
     }
 
-    Result SendSyncRequest();
+    Result SendSyncRequest(uintptr_t address, size_t size);
+    Result SendAsyncRequest(KEvent* event, uintptr_t address, size_t size);
 
     void OnServerClosed();
 

src/core/hle/kernel/k_light_client_session.cpp

@@ -0,0 +1,31 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "core/hle/kernel/k_light_client_session.h"
+#include "core/hle/kernel/k_light_session.h"
+#include "core/hle/kernel/k_thread.h"
+
+namespace Kernel {
+
+KLightClientSession::KLightClientSession(KernelCore& kernel) : KAutoObject(kernel) {}
+
+KLightClientSession::~KLightClientSession() = default;
+
+void KLightClientSession::Destroy() {
+    m_parent->OnClientClosed();
+}
+
+void KLightClientSession::OnServerClosed() {}
+
+Result KLightClientSession::SendSyncRequest(u32* data) {
+    // Get the request thread.
+    KThread* cur_thread = GetCurrentThreadPointer(m_kernel);
+
+    // Set the light data.
+    cur_thread->SetLightSessionData(data);
+
+    // Send the request.
+    R_RETURN(m_parent->OnRequest(cur_thread));
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_light_client_session.h

@@ -0,0 +1,39 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "core/hle/kernel/k_auto_object.h"
+#include "core/hle/result.h"
+
+namespace Kernel {
+
+class KLightSession;
+
+class KLightClientSession final : public KAutoObject {
+    KERNEL_AUTOOBJECT_TRAITS(KLightClientSession, KAutoObject);
+
+public:
+    explicit KLightClientSession(KernelCore& kernel);
+    ~KLightClientSession();
+
+    void Initialize(KLightSession* parent) {
+        // Set member variables.
+        m_parent = parent;
+    }
+
+    virtual void Destroy() override;
+
+    const KLightSession* GetParent() const {
+        return m_parent;
+    }
+
+    Result SendSyncRequest(u32* data);
+
+    void OnServerClosed();
+
+private:
+    KLightSession* m_parent;
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_light_server_session.cpp

@@ -0,0 +1,247 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "core/hle/kernel/k_light_server_session.h"
+#include "core/hle/kernel/k_light_session.h"
+#include "core/hle/kernel/k_thread.h"
+#include "core/hle/kernel/k_thread_queue.h"
+#include "core/hle/kernel/svc_results.h"
+
+namespace Kernel {
+
+namespace {
+
+constexpr u64 InvalidThreadId = std::numeric_limits<u64>::max();
+
+class ThreadQueueImplForKLightServerSessionRequest final : public KThreadQueue {
+private:
+    KThread::WaiterList* m_wait_list;
+
+public:
+    ThreadQueueImplForKLightServerSessionRequest(KernelCore& kernel, KThread::WaiterList* wl)
+        : KThreadQueue(kernel), m_wait_list(wl) {}
+
+    virtual void EndWait(KThread* waiting_thread, Result wait_result) override {
+        // Remove the thread from our wait list.
+        m_wait_list->erase(m_wait_list->iterator_to(*waiting_thread));
+
+        // Invoke the base end wait handler.
+        KThreadQueue::EndWait(waiting_thread, wait_result);
+    }
+
+    virtual void CancelWait(KThread* waiting_thread, Result wait_result,
+                            bool cancel_timer_task) override {
+        // Remove the thread from our wait list.
+        m_wait_list->erase(m_wait_list->iterator_to(*waiting_thread));
+
+        // Invoke the base cancel wait handler.
+        KThreadQueue::CancelWait(waiting_thread, wait_result, cancel_timer_task);
+    }
+};
+
+class ThreadQueueImplForKLightServerSessionReceive final : public KThreadQueue {
+private:
+    KThread** m_server_thread;
+
+public:
+    ThreadQueueImplForKLightServerSessionReceive(KernelCore& kernel, KThread** st)
+        : KThreadQueue(kernel), m_server_thread(st) {}
+
+    virtual void EndWait(KThread* waiting_thread, Result wait_result) override {
+        // Clear the server thread.
+        *m_server_thread = nullptr;
+
+        // Set the waiting thread as not cancelable.
+        waiting_thread->ClearCancellable();
+
+        // Invoke the base end wait handler.
+        KThreadQueue::EndWait(waiting_thread, wait_result);
+    }
+
+    virtual void CancelWait(KThread* waiting_thread, Result wait_result,
+                            bool cancel_timer_task) override {
+        // Clear the server thread.
+        *m_server_thread = nullptr;
+
+        // Set the waiting thread as not cancelable.
+        waiting_thread->ClearCancellable();
+
+        // Invoke the base cancel wait handler.
+        KThreadQueue::CancelWait(waiting_thread, wait_result, cancel_timer_task);
+    }
+};
+
+} // namespace
+
+KLightServerSession::KLightServerSession(KernelCore& kernel) : KAutoObject(kernel) {}
+KLightServerSession::~KLightServerSession() = default;
+
+void KLightServerSession::Destroy() {
+    this->CleanupRequests();
+
+    m_parent->OnServerClosed();
+}
+
+void KLightServerSession::OnClientClosed() {
+    this->CleanupRequests();
+}
+
+Result KLightServerSession::OnRequest(KThread* request_thread) {
+    ThreadQueueImplForKLightServerSessionRequest wait_queue(m_kernel,
+                                                            std::addressof(m_request_list));
+
+    // Send the request.
+    {
+        // Lock the scheduler.
+        KScopedSchedulerLock sl(m_kernel);
+
+        // Check that the server isn't closed.
+        R_UNLESS(!m_parent->IsServerClosed(), ResultSessionClosed);
+
+        // Check that the request thread isn't terminating.
+        R_UNLESS(!request_thread->IsTerminationRequested(), ResultTerminationRequested);
+
+        // Add the request thread to our list.
+        m_request_list.push_back(*request_thread);
+
+        // Begin waiting on the request.
+        request_thread->SetWaitReasonForDebugging(ThreadWaitReasonForDebugging::IPC);
+        request_thread->BeginWait(std::addressof(wait_queue));
+
+        // If we have a server thread, end its wait.
+        if (m_server_thread != nullptr) {
+            m_server_thread->EndWait(ResultSuccess);
+        }
+    }
+
+    // NOTE: Nintendo returns GetCurrentThread().GetWaitResult() here.
+    // This is technically incorrect, although it doesn't cause problems in practice
+    // because this is only ever called with request_thread = GetCurrentThreadPointer().
+    R_RETURN(request_thread->GetWaitResult());
+}
+
+Result KLightServerSession::ReplyAndReceive(u32* data) {
+    // Set the server context.
+    GetCurrentThread(m_kernel).SetLightSessionData(data);
+
+    // Reply, if we need to.
+    if (data[0] & KLightSession::ReplyFlag) {
+        KScopedSchedulerLock sl(m_kernel);
+
+        // Check that we're open.
+        R_UNLESS(!m_parent->IsClientClosed(), ResultSessionClosed);
+        R_UNLESS(!m_parent->IsServerClosed(), ResultSessionClosed);
+
+        // Check that we have a request to reply to.
+        R_UNLESS(m_current_request != nullptr, ResultInvalidState);
+
+        // Check that the server thread id is correct.
+        R_UNLESS(m_server_thread_id == GetCurrentThread(m_kernel).GetId(), ResultInvalidState);
+
+        // If we can reply, do so.
+        if (!m_current_request->IsTerminationRequested()) {
+            std::memcpy(m_current_request->GetLightSessionData(),
+                        GetCurrentThread(m_kernel).GetLightSessionData(), KLightSession::DataSize);
+            m_current_request->EndWait(ResultSuccess);
+        }
+
+        // Close our current request.
+        m_current_request->Close();
+
+        // Clear our current request.
+        m_current_request = nullptr;
+        m_server_thread_id = InvalidThreadId;
+    }
+
+    // Create the wait queue for our receive.
+    ThreadQueueImplForKLightServerSessionReceive wait_queue(m_kernel,
+                                                            std::addressof(m_server_thread));
+
+    // Receive.
+    while (true) {
+        // Try to receive a request.
+        {
+            KScopedSchedulerLock sl(m_kernel);
+
+            // Check that we aren't already receiving.
+            R_UNLESS(m_server_thread == nullptr, ResultInvalidState);
+            R_UNLESS(m_server_thread_id == InvalidThreadId, ResultInvalidState);
+
+            // Check that we're open.
+            R_UNLESS(!m_parent->IsClientClosed(), ResultSessionClosed);
+            R_UNLESS(!m_parent->IsServerClosed(), ResultSessionClosed);
+
+            // Check that we're not terminating.
+            R_UNLESS(!GetCurrentThread(m_kernel).IsTerminationRequested(),
+                     ResultTerminationRequested);
+
+            // If we have a request available, use it.
+            if (auto head = m_request_list.begin(); head != m_request_list.end()) {
+                // Set our current request.
+                m_current_request = std::addressof(*head);
+                m_current_request->Open();
+
+                // Set our server thread id.
+                m_server_thread_id = GetCurrentThread(m_kernel).GetId();
+
+                // Copy the client request data.
+                std::memcpy(GetCurrentThread(m_kernel).GetLightSessionData(),
+                            m_current_request->GetLightSessionData(), KLightSession::DataSize);
+
+                // We successfully received.
+                R_SUCCEED();
+            }
+
+            // We need to wait for a request to come in.
+
+            // Check if we were cancelled.
+            if (GetCurrentThread(m_kernel).IsWaitCancelled()) {
+                GetCurrentThread(m_kernel).ClearWaitCancelled();
+                R_THROW(ResultCancelled);
+            }
+
+            // Mark ourselves as cancellable.
+            GetCurrentThread(m_kernel).SetCancellable();
+
+            // Wait for a request to come in.
+            m_server_thread = GetCurrentThreadPointer(m_kernel);
+            GetCurrentThread(m_kernel).SetWaitReasonForDebugging(ThreadWaitReasonForDebugging::IPC);
+            GetCurrentThread(m_kernel).BeginWait(std::addressof(wait_queue));
+        }
+
+        // We waited to receive a request; if our wait failed, return the failing result.
+        R_TRY(GetCurrentThread(m_kernel).GetWaitResult());
+    }
+}
+
+void KLightServerSession::CleanupRequests() {
+    // Cleanup all pending requests.
+    {
+        KScopedSchedulerLock sl(m_kernel);
+
+        // Handle the current request.
+        if (m_current_request != nullptr) {
+            // Reply to the current request.
+            if (!m_current_request->IsTerminationRequested()) {
+                m_current_request->EndWait(ResultSessionClosed);
+            }
+
+            // Clear our current request.
+            m_current_request->Close();
+            m_current_request = nullptr;
+            m_server_thread_id = InvalidThreadId;
+        }
+
+        // Reply to all other requests.
+        for (auto& thread : m_request_list) {
+            thread.EndWait(ResultSessionClosed);
+        }
+
+        // Wait up our server thread, if we have one.
+        if (m_server_thread != nullptr) {
+            m_server_thread->EndWait(ResultSessionClosed);
+        }
+    }
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_light_server_session.h

@@ -0,0 +1,49 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "core/hle/kernel/k_auto_object.h"
+#include "core/hle/kernel/k_thread.h"
+#include "core/hle/result.h"
+
+namespace Kernel {
+
+class KLightSession;
+
+class KLightServerSession final : public KAutoObject,
+                                  public Common::IntrusiveListBaseNode<KLightServerSession> {
+    KERNEL_AUTOOBJECT_TRAITS(KLightServerSession, KAutoObject);
+
+private:
+    KLightSession* m_parent{};
+    KThread::WaiterList m_request_list{};
+    KThread* m_current_request{};
+    u64 m_server_thread_id{std::numeric_limits<u64>::max()};
+    KThread* m_server_thread{};
+
+public:
+    explicit KLightServerSession(KernelCore& kernel);
+    ~KLightServerSession();
+
+    void Initialize(KLightSession* parent) {
+        // Set member variables. */
+        m_parent = parent;
+    }
+
+    virtual void Destroy() override;
+
+    constexpr const KLightSession* GetParent() const {
+        return m_parent;
+    }
+
+    Result OnRequest(KThread* request_thread);
+    Result ReplyAndReceive(u32* data);
+
+    void OnClientClosed();
+
+private:
+    void CleanupRequests();
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_light_session.cpp

@@ -0,0 +1,81 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "core/hle/kernel/k_client_port.h"
+#include "core/hle/kernel/k_light_client_session.h"
+#include "core/hle/kernel/k_light_server_session.h"
+#include "core/hle/kernel/k_light_session.h"
+#include "core/hle/kernel/k_process.h"
+
+namespace Kernel {
+
+KLightSession::KLightSession(KernelCore& kernel)
+    : KAutoObjectWithSlabHeapAndContainer(kernel), m_server(kernel), m_client(kernel) {}
+KLightSession::~KLightSession() = default;
+
+void KLightSession::Initialize(KClientPort* client_port, uintptr_t name) {
+    // Increment reference count.
+    // Because reference count is one on creation, this will result
+    // in a reference count of two. Thus, when both server and client are closed
+    // this object will be destroyed.
+    this->Open();
+
+    // Create our sub sessions.
+    KAutoObject::Create(std::addressof(m_server));
+    KAutoObject::Create(std::addressof(m_client));
+
+    // Initialize our sub sessions.
+    m_server.Initialize(this);
+    m_client.Initialize(this);
+
+    // Set state and name.
+    m_state = State::Normal;
+    m_name = name;
+
+    // Set our owner process.
+    m_process = GetCurrentProcessPointer(m_kernel);
+    m_process->Open();
+
+    // Set our port.
+    m_port = client_port;
+    if (m_port != nullptr) {
+        m_port->Open();
+    }
+
+    // Mark initialized.
+    m_initialized = true;
+}
+
+void KLightSession::Finalize() {
+    if (m_port != nullptr) {
+        m_port->OnSessionFinalized();
+        m_port->Close();
+    }
+}
+
+void KLightSession::OnServerClosed() {
+    if (m_state == State::Normal) {
+        m_state = State::ServerClosed;
+        m_client.OnServerClosed();
+    }
+
+    this->Close();
+}
+
+void KLightSession::OnClientClosed() {
+    if (m_state == State::Normal) {
+        m_state = State::ClientClosed;
+        m_server.OnClientClosed();
+    }
+
+    this->Close();
+}
+
+void KLightSession::PostDestroy(uintptr_t arg) {
+    // Release the session count resource the owner process holds.
+    KProcess* owner = reinterpret_cast<KProcess*>(arg);
+    owner->ReleaseResource(Svc::LimitableResource::SessionCountMax, 1);
+    owner->Close();
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_light_session.h

@@ -0,0 +1,86 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "core/hle/kernel/k_light_client_session.h"
+#include "core/hle/kernel/k_light_server_session.h"
+#include "core/hle/kernel/slab_helpers.h"
+#include "core/hle/result.h"
+
+namespace Kernel {
+
+class KClientPort;
+class KProcess;
+
+// TODO: SupportDynamicExpansion for SlabHeap
+class KLightSession final
+    : public KAutoObjectWithSlabHeapAndContainer<KLightSession, KAutoObjectWithList> {
+    KERNEL_AUTOOBJECT_TRAITS(KLightSession, KAutoObject);
+
+private:
+    enum class State : u8 {
+        Invalid = 0,
+        Normal = 1,
+        ClientClosed = 2,
+        ServerClosed = 3,
+    };
+
+public:
+    static constexpr size_t DataSize = sizeof(u32) * 7;
+    static constexpr u32 ReplyFlag = (1U << 31);
+
+private:
+    KLightServerSession m_server;
+    KLightClientSession m_client;
+    State m_state{State::Invalid};
+    KClientPort* m_port{};
+    uintptr_t m_name{};
+    KProcess* m_process{};
+    bool m_initialized{};
+
+public:
+    explicit KLightSession(KernelCore& kernel);
+    ~KLightSession();
+
+    void Initialize(KClientPort* client_port, uintptr_t name);
+    void Finalize() override;
+
+    bool IsInitialized() const override {
+        return m_initialized;
+    }
+    uintptr_t GetPostDestroyArgument() const override {
+        return reinterpret_cast<uintptr_t>(m_process);
+    }
+
+    static void PostDestroy(uintptr_t arg);
+
+    void OnServerClosed();
+    void OnClientClosed();
+
+    bool IsServerClosed() const {
+        return m_state != State::Normal;
+    }
+    bool IsClientClosed() const {
+        return m_state != State::Normal;
+    }
+
+    Result OnRequest(KThread* request_thread) {
+        R_RETURN(m_server.OnRequest(request_thread));
+    }
+
+    KLightClientSession& GetClientSession() {
+        return m_client;
+    }
+    KLightServerSession& GetServerSession() {
+        return m_server;
+    }
+    const KLightClientSession& GetClientSession() const {
+        return m_client;
+    }
+    const KLightServerSession& GetServerSession() const {
+        return m_server;
+    }
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_page_table_base.cpp

@@ -69,8 +69,16 @@ public:
 };
 
 template <typename AddressType>
-void InvalidateInstructionCache(Core::System& system, AddressType addr, u64 size) {
-    system.InvalidateCpuInstructionCacheRange(GetInteger(addr), size);
+void InvalidateInstructionCache(KernelCore& kernel, AddressType addr, u64 size) {
+    // TODO: lock the process list
+    for (auto& process : kernel.GetProcessList()) {
+        for (size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
+            auto* interface = process->GetArmInterface(i);
+            if (interface) {
+                interface->InvalidateCacheRange(GetInteger(addr), size);
+            }
+        }
+    }
 }
 
 template <typename AddressType>
@@ -1261,7 +1269,7 @@ Result KPageTableBase::UnmapCodeMemory(KProcessAddress dst_address, KProcessAddr
     bool reprotected_pages = false;
     SCOPE_EXIT({
         if (reprotected_pages && any_code_pages) {
-            InvalidateInstructionCache(m_system, dst_address, size);
+            InvalidateInstructionCache(m_kernel, dst_address, size);
         }
     });
 
@@ -1997,7 +2005,7 @@ Result KPageTableBase::SetProcessMemoryPermission(KProcessAddress addr, size_t s
         for (const auto& block : pg) {
             StoreDataCache(GetHeapVirtualPointer(m_kernel, block.GetAddress()), block.GetSize());
         }
-        InvalidateInstructionCache(m_system, addr, size);
+        InvalidateInstructionCache(m_kernel, addr, size);
     }
 
     R_SUCCEED();
@@ -3239,7 +3247,7 @@ Result KPageTableBase::WriteDebugMemory(KProcessAddress dst_address, KProcessAdd
     R_TRY(PerformCopy());
 
     // Invalidate the instruction cache, as this svc allows modifying executable pages.
-    InvalidateInstructionCache(m_system, dst_address, size);
+    InvalidateInstructionCache(m_kernel, dst_address, size);
 
     R_SUCCEED();
 }

src/core/hle/kernel/k_port.cpp

@@ -58,4 +58,13 @@ Result KPort::EnqueueSession(KServerSession* session) {
     R_SUCCEED();
 }
 
+Result KPort::EnqueueSession(KLightServerSession* session) {
+    KScopedSchedulerLock sl{m_kernel};
+
+    R_UNLESS(m_state == State::Normal, ResultPortClosed);
+
+    m_server.EnqueueSession(session);
+    R_SUCCEED();
+}
+
 } // namespace Kernel

src/core/hle/kernel/k_port.h

@@ -13,6 +13,7 @@
 
 namespace Kernel {
 
+class KLightServerSession;
 class KServerSession;
 
 class KPort final : public KAutoObjectWithSlabHeapAndContainer<KPort, KAutoObjectWithList> {
@@ -38,6 +39,7 @@ public:
     bool IsServerClosed() const;
 
     Result EnqueueSession(KServerSession* session);
+    Result EnqueueSession(KLightServerSession* session);
 
     KClientPort& GetClientPort() {
         return m_client;

src/core/hle/kernel/k_process.cpp

@@ -13,6 +13,12 @@
 #include "core/hle/kernel/k_thread_queue.h"
 #include "core/hle/kernel/k_worker_task_manager.h"
 
+#include "core/arm/dynarmic/arm_dynarmic_32.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
+#ifdef HAS_NCE
+#include "core/arm/nce/arm_nce.h"
+#endif
+
 namespace Kernel {
 
 namespace {
@@ -957,10 +963,8 @@ Result KProcess::Run(s32 priority, size_t stack_size) {
     R_TRY(m_handle_table.Add(std::addressof(thread_handle), main_thread));
 
     // Set the thread arguments.
-    main_thread->GetContext32().cpu_registers[0] = 0;
-    main_thread->GetContext64().cpu_registers[0] = 0;
-    main_thread->GetContext32().cpu_registers[1] = thread_handle;
-    main_thread->GetContext64().cpu_registers[1] = thread_handle;
+    main_thread->GetContext().r[0] = 0;
+    main_thread->GetContext().r[1] = thread_handle;
 
     // Update our state.
     this->ChangeState((state == State::Created) ? State::Running : State::RunningAttached);
@@ -1199,6 +1203,9 @@ Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std:
     m_is_hbl = is_hbl;
     m_ideal_core_id = metadata.GetMainThreadCore();
 
+    // Set up emulation context.
+    this->InitializeInterfaces();
+
     // We succeeded.
     R_SUCCEED();
 }
@@ -1227,6 +1234,31 @@ void KProcess::LoadModule(CodeSet code_set, KProcessAddress base_addr) {
 #endif
 }
 
+void KProcess::InitializeInterfaces() {
+    this->GetMemory().SetCurrentPageTable(*this);
+
+#ifdef HAS_NCE
+    if (this->Is64Bit() && Settings::IsNceEnabled()) {
+        for (size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
+            m_arm_interfaces[i] = std::make_unique<Core::ArmNce>(m_kernel.System(), true, i);
+        }
+    } else
+#endif
+        if (this->Is64Bit()) {
+        for (size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
+            m_arm_interfaces[i] = std::make_unique<Core::ArmDynarmic64>(
+                m_kernel.System(), m_kernel.IsMulticore(), this,
+                static_cast<Core::DynarmicExclusiveMonitor&>(m_kernel.GetExclusiveMonitor()), i);
+        }
+    } else {
+        for (size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
+            m_arm_interfaces[i] = std::make_unique<Core::ArmDynarmic32>(
+                m_kernel.System(), m_kernel.IsMulticore(), this,
+                static_cast<Core::DynarmicExclusiveMonitor&>(m_kernel.GetExclusiveMonitor()), i);
+        }
+    }
+}
+
 bool KProcess::InsertWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointType type) {
     const auto watch{std::find_if(m_watchpoints.begin(), m_watchpoints.end(), [&](const auto& wp) {
         return wp.type == DebugWatchpointType::None;

src/core/hle/kernel/k_process.h

@@ -5,6 +5,7 @@
 
 #include <map>
 
+#include "core/arm/arm_interface.h"
 #include "core/file_sys/program_metadata.h"
 #include "core/hle/kernel/code_set.h"
 #include "core/hle/kernel/k_address_arbiter.h"
@@ -106,6 +107,8 @@ private:
     bool m_is_suspended{};
     bool m_is_immortal{};
     bool m_is_handle_table_initialized{};
+    std::array<std::unique_ptr<Core::ArmInterface>, Core::Hardware::NUM_CPU_CORES>
+        m_arm_interfaces{};
     std::array<KThread*, Core::Hardware::NUM_CPU_CORES> m_running_threads{};
     std::array<u64, Core::Hardware::NUM_CPU_CORES> m_running_thread_idle_counts{};
     std::array<u64, Core::Hardware::NUM_CPU_CORES> m_running_thread_switch_counts{};
@@ -476,6 +479,10 @@ public:
     }
 #endif
 
+    Core::ArmInterface* GetArmInterface(size_t core_index) const {
+        return m_arm_interfaces[core_index].get();
+    }
+
 public:
     // Attempts to insert a watchpoint into a free slot. Returns false if none are available.
     bool InsertWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointType type);
@@ -493,6 +500,8 @@ public:
 
     void LoadModule(CodeSet code_set, KProcessAddress base_addr);
 
+    void InitializeInterfaces();
+
     Core::Memory::Memory& GetMemory() const;
 
 public:

src/core/hle/kernel/k_process_page_table.h

@@ -7,10 +7,6 @@
 #include "core/hle/kernel/k_scoped_lock.h"
 #include "core/hle/kernel/svc_types.h"
 
-namespace Core {
-class ARM_Interface;
-}
-
 namespace Kernel {
 
 class KProcessPageTable {

src/core/hle/kernel/k_scheduler.cpp

@@ -494,12 +494,7 @@ void KScheduler::ScheduleImplFiber() {
 }
 
 void KScheduler::Unload(KThread* thread) {
-    auto& cpu_core = m_kernel.System().ArmInterface(m_core_id);
-    cpu_core.SaveContext(thread->GetContext32());
-    cpu_core.SaveContext(thread->GetContext64());
-    // Save the TPIDR_EL0 system register in case it was modified.
-    thread->SetTpidrEl0(cpu_core.GetTPIDR_EL0());
-    cpu_core.ClearExclusiveState();
+    m_kernel.PhysicalCore(m_core_id).SaveContext(thread);
 
     // Check if the thread is terminated by checking the DPC flags.
     if ((thread->GetStackParameters().dpc_flags & static_cast<u32>(DpcFlag::Terminated)) == 0) {
@@ -509,14 +504,7 @@ void KScheduler::Unload(KThread* thread) {
 }
 
 void KScheduler::Reload(KThread* thread) {
-    auto& cpu_core = m_kernel.System().ArmInterface(m_core_id);
-    auto* process = thread->GetOwnerProcess();
-    cpu_core.LoadContext(thread->GetContext32());
-    cpu_core.LoadContext(thread->GetContext64());
-    cpu_core.SetTlsAddress(GetInteger(thread->GetTlsAddress()));
-    cpu_core.SetTPIDR_EL0(thread->GetTpidrEl0());
-    cpu_core.LoadWatchpointArray(process ? &process->GetWatchpoints() : nullptr);
-    cpu_core.ClearExclusiveState();
+    m_kernel.PhysicalCore(m_core_id).LoadContext(thread);
 }
 
 void KScheduler::ClearPreviousThread(KernelCore& kernel, KThread* thread) {

src/core/hle/kernel/k_server_port.cpp

@@ -27,12 +27,14 @@ bool KServerPort::IsLight() const {
 void KServerPort::CleanupSessions() {
     // Ensure our preconditions are met.
     if (this->IsLight()) {
-        UNIMPLEMENTED();
+        ASSERT(m_session_list.empty());
+    } else {
+        ASSERT(m_light_session_list.empty());
     }
 
     // Cleanup the session list.
     while (true) {
-        // Get the last session in the list
+        // Get the last session in the list.
         KServerSession* session = nullptr;
         {
             KScopedSchedulerLock sl{m_kernel};
@@ -49,6 +51,26 @@ void KServerPort::CleanupSessions() {
             break;
         }
     }
+
+    // Cleanup the light session list.
+    while (true) {
+        // Get the last session in the list.
+        KLightServerSession* session = nullptr;
+        {
+            KScopedSchedulerLock sl{m_kernel};
+            if (!m_light_session_list.empty()) {
+                session = std::addressof(m_light_session_list.front());
+                m_light_session_list.pop_front();
+            }
+        }
+
+        // Close the session.
+        if (session != nullptr) {
+            session->Close();
+        } else {
+            break;
+        }
+    }
 }
 
 void KServerPort::Destroy() {
@@ -64,8 +86,7 @@ void KServerPort::Destroy() {
 
 bool KServerPort::IsSignaled() const {
     if (this->IsLight()) {
-        UNIMPLEMENTED();
-        return false;
+        return !m_light_session_list.empty();
     } else {
         return !m_session_list.empty();
     }
@@ -83,6 +104,18 @@ void KServerPort::EnqueueSession(KServerSession* session) {
     }
 }
 
+void KServerPort::EnqueueSession(KLightServerSession* session) {
+    ASSERT(this->IsLight());
+
+    KScopedSchedulerLock sl{m_kernel};
+
+    // Add the session to our queue.
+    m_light_session_list.push_back(*session);
+    if (m_light_session_list.size() == 1) {
+        this->NotifyAvailable();
+    }
+}
+
 KServerSession* KServerPort::AcceptSession() {
     ASSERT(!this->IsLight());
 
@@ -98,4 +131,19 @@ KServerSession* KServerPort::AcceptSession() {
     return session;
 }
 
+KLightServerSession* KServerPort::AcceptLightSession() {
+    ASSERT(this->IsLight());
+
+    KScopedSchedulerLock sl{m_kernel};
+
+    // Return the first session in the list.
+    if (m_light_session_list.empty()) {
+        return nullptr;
+    }
+
+    KLightServerSession* session = std::addressof(m_light_session_list.front());
+    m_light_session_list.pop_front();
+    return session;
+}
+
 } // namespace Kernel

src/core/hle/kernel/k_server_port.h

@@ -9,6 +9,7 @@
 
 #include "common/intrusive_list.h"
 
+#include "core/hle/kernel/k_light_server_session.h"
 #include "core/hle/kernel/k_server_session.h"
 #include "core/hle/kernel/k_synchronization_object.h"
 
@@ -28,8 +29,10 @@ public:
     void Initialize(KPort* parent);
 
     void EnqueueSession(KServerSession* session);
+    void EnqueueSession(KLightServerSession* session);
 
     KServerSession* AcceptSession();
+    KLightServerSession* AcceptLightSession();
 
     const KPort* GetParent() const {
         return m_parent;
@@ -43,10 +46,12 @@ public:
 
 private:
     using SessionList = Common::IntrusiveListBaseTraits<KServerSession>::ListType;
+    using LightSessionList = Common::IntrusiveListBaseTraits<KLightServerSession>::ListType;
 
     void CleanupSessions();
 
     SessionList m_session_list{};
+    LightSessionList m_light_session_list{};
     KPort* m_parent{};
 };
 

src/core/hle/kernel/k_server_session.cpp

@@ -453,6 +453,11 @@ Result KServerSession::ReceiveRequest(std::shared_ptr<Service::HLERequestContext
     size_t client_buffer_size = request->GetSize();
     // bool recv_list_broken = false;
 
+    if (!client_message) {
+        client_message = GetInteger(client_thread->GetTlsAddress());
+        client_buffer_size = MessageBufferSize;
+    }
+
     // Receive the message.
     Core::Memory::Memory& memory{client_thread->GetOwnerProcess()->GetMemory()};
     if (out_context != nullptr) {
@@ -462,8 +467,7 @@ Result KServerSession::ReceiveRequest(std::shared_ptr<Service::HLERequestContext
             std::make_shared<Service::HLERequestContext>(m_kernel, memory, this, client_thread);
         (*out_context)->SetSessionRequestManager(manager);
         (*out_context)
-            ->PopulateFromIncomingCommandBuffer(client_thread->GetOwnerProcess()->GetHandleTable(),
-                                                cmd_buf);
+            ->PopulateFromIncomingCommandBuffer(*client_thread->GetOwnerProcess(), cmd_buf);
     } else {
         KThread* server_thread = GetCurrentThreadPointer(m_kernel);
         KProcess& src_process = *client_thread->GetOwnerProcess();

src/core/hle/kernel/k_session.h

@@ -46,6 +46,10 @@ public:
         return this->GetState() != State::Normal;
     }
 
+    Result OnRequest(KSessionRequest* request) {
+        R_RETURN(m_server.OnRequest(request));
+    }
+
     KClientSession& GetClientSession() {
         return m_client;
     }

src/core/hle/kernel/k_thread.cpp

@@ -41,24 +41,25 @@ namespace {
 
 constexpr inline s32 TerminatingThreadPriority = Kernel::Svc::SystemThreadPriorityHighest - 1;
 
-static void ResetThreadContext32(Kernel::KThread::ThreadContext32& context, u32 stack_top,
-                                 u32 entry_point, u32 arg) {
-    context = {};
-    context.cpu_registers[0] = arg;
-    context.cpu_registers[15] = entry_point;
-    context.cpu_registers[13] = stack_top;
-    context.fpscr = 0;
+static void ResetThreadContext32(Kernel::Svc::ThreadContext& ctx, u64 stack_top, u64 entry_point,
+                                 u64 arg) {
+    ctx = {};
+    ctx.r[0] = arg;
+    ctx.r[15] = entry_point;
+    ctx.r[13] = stack_top;
+    ctx.fpcr = 0;
+    ctx.fpsr = 0;
 }
 
-static void ResetThreadContext64(Kernel::KThread::ThreadContext64& context, u64 stack_top,
-                                 u64 entry_point, u64 arg) {
-    context = {};
-    context.cpu_registers[0] = arg;
-    context.cpu_registers[18] = Kernel::KSystemControl::GenerateRandomU64() | 1;
-    context.pc = entry_point;
-    context.sp = stack_top;
-    context.fpcr = 0;
-    context.fpsr = 0;
+static void ResetThreadContext64(Kernel::Svc::ThreadContext& ctx, u64 stack_top, u64 entry_point,
+                                 u64 arg) {
+    ctx = {};
+    ctx.r[0] = arg;
+    ctx.r[18] = Kernel::KSystemControl::GenerateRandomU64() | 1;
+    ctx.pc = entry_point;
+    ctx.sp = stack_top;
+    ctx.fpcr = 0;
+    ctx.fpsr = 0;
 }
 } // namespace
 
@@ -223,9 +224,11 @@ Result KThread::Initialize(KThreadFunction func, uintptr_t arg, KProcessAddress
     }
 
     // Initialize thread context.
-    ResetThreadContext64(m_thread_context_64, GetInteger(user_stack_top), GetInteger(func), arg);
-    ResetThreadContext32(m_thread_context_32, static_cast<u32>(GetInteger(user_stack_top)),
-                         static_cast<u32>(GetInteger(func)), static_cast<u32>(arg));
+    if (m_parent != nullptr && !m_parent->Is64Bit()) {
+        ResetThreadContext32(m_thread_context, GetInteger(user_stack_top), GetInteger(func), arg);
+    } else {
+        ResetThreadContext64(m_thread_context, GetInteger(user_stack_top), GetInteger(func), arg);
+    }
 
     // Setup the stack parameters.
     StackParameters& sp = this->GetStackParameters();
@@ -823,20 +826,7 @@ void KThread::CloneFpuStatus() {
     ASSERT(this->GetOwnerProcess() != nullptr);
     ASSERT(this->GetOwnerProcess() == GetCurrentProcessPointer(m_kernel));
 
-    if (this->GetOwnerProcess()->Is64Bit()) {
-        // Clone FPSR and FPCR.
-        ThreadContext64 cur_ctx{};
-        m_kernel.System().CurrentArmInterface().SaveContext(cur_ctx);
-
-        this->GetContext64().fpcr = cur_ctx.fpcr;
-        this->GetContext64().fpsr = cur_ctx.fpsr;
-    } else {
-        // Clone FPSCR.
-        ThreadContext32 cur_ctx{};
-        m_kernel.System().CurrentArmInterface().SaveContext(cur_ctx);
-
-        this->GetContext32().fpscr = cur_ctx.fpscr;
-    }
+    m_kernel.CurrentPhysicalCore().CloneFpuStatus(this);
 }
 
 Result KThread::SetActivity(Svc::ThreadActivity activity) {
@@ -912,7 +902,7 @@ Result KThread::SetActivity(Svc::ThreadActivity activity) {
     R_SUCCEED();
 }
 
-Result KThread::GetThreadContext3(Common::ScratchBuffer<u8>& out) {
+Result KThread::GetThreadContext3(Svc::ThreadContext* out) {
     // Lock ourselves.
     KScopedLightLock lk{m_activity_pause_lock};
 
@@ -926,18 +916,16 @@ Result KThread::GetThreadContext3(Common::ScratchBuffer<u8>& out) {
 
         // If we're not terminating, get the thread's user context.
         if (!this->IsTerminationRequested()) {
+            *out = m_thread_context;
+
+            // Mask away mode bits, interrupt bits, IL bit, and other reserved bits.
+            constexpr u32 El0Aarch64PsrMask = 0xF0000000;
+            constexpr u32 El0Aarch32PsrMask = 0xFE0FFE20;
+
             if (m_parent->Is64Bit()) {
-                // Mask away mode bits, interrupt bits, IL bit, and other reserved bits.
-                auto context = GetContext64();
-                context.pstate &= 0xFF0FFE20;
-                out.resize_destructive(sizeof(context));
-                std::memcpy(out.data(), std::addressof(context), sizeof(context));
+                out->pstate &= El0Aarch64PsrMask;
             } else {
-                // Mask away mode bits, interrupt bits, IL bit, and other reserved bits.
-                auto context = GetContext32();
-                context.cpsr &= 0xFF0FFE20;
-                out.resize_destructive(sizeof(context));
-                std::memcpy(out.data(), std::addressof(context), sizeof(context));
+                out->pstate &= El0Aarch32PsrMask;
             }
         }
     }

src/core/hle/kernel/k_thread.h

@@ -38,7 +38,6 @@ namespace Core {
 namespace Memory {
 class Memory;
 }
-class ARM_Interface;
 class System;
 } // namespace Core
 
@@ -137,8 +136,6 @@ public:
     ~KThread() override;
 
 public:
-    using ThreadContext32 = Core::ARM_Interface::ThreadContext32;
-    using ThreadContext64 = Core::ARM_Interface::ThreadContext64;
     using WaiterList = Common::IntrusiveListBaseTraits<KThread>::ListType;
 
     /**
@@ -246,31 +243,22 @@ public:
      * @returns The value of the TPIDR_EL0 register.
      */
     u64 GetTpidrEl0() const {
-        return m_thread_context_64.tpidr;
+        return m_thread_context.tpidr;
     }
 
     /// Sets the value of the TPIDR_EL0 Read/Write system register for this thread.
     void SetTpidrEl0(u64 value) {
-        m_thread_context_64.tpidr = value;
-        m_thread_context_32.tpidr = static_cast<u32>(value);
+        m_thread_context.tpidr = value;
     }
 
     void CloneFpuStatus();
 
-    ThreadContext32& GetContext32() {
-        return m_thread_context_32;
+    Svc::ThreadContext& GetContext() {
+        return m_thread_context;
     }
 
-    const ThreadContext32& GetContext32() const {
-        return m_thread_context_32;
-    }
-
-    ThreadContext64& GetContext64() {
-        return m_thread_context_64;
-    }
-
-    const ThreadContext64& GetContext64() const {
-        return m_thread_context_64;
+    const Svc::ThreadContext& GetContext() const {
+        return m_thread_context;
     }
 
     std::shared_ptr<Common::Fiber>& GetHostContext();
@@ -397,6 +385,13 @@ public:
         m_cancellable = false;
     }
 
+    u32* GetLightSessionData() const {
+        return m_light_ipc_data;
+    }
+    void SetLightSessionData(u32* data) {
+        m_light_ipc_data = data;
+    }
+
     bool IsTerminationRequested() const {
         return m_termination_requested || GetRawState() == ThreadState::Terminated;
     }
@@ -577,7 +572,7 @@ public:
 
     void RemoveWaiter(KThread* thread);
 
-    Result GetThreadContext3(Common::ScratchBuffer<u8>& out);
+    Result GetThreadContext3(Svc::ThreadContext* out);
 
     KThread* RemoveUserWaiterByKey(bool* out_has_waiters, KProcessAddress key) {
         return this->RemoveWaiterByKey(out_has_waiters, key, false);
@@ -734,8 +729,7 @@ private:
                                    std::function<void()>&& init_func);
 
     // For core KThread implementation
-    ThreadContext32 m_thread_context_32{};
-    ThreadContext64 m_thread_context_64{};
+    Svc::ThreadContext m_thread_context{};
     Common::IntrusiveListNode m_process_list_node;
     Common::IntrusiveRedBlackTreeNode m_condvar_arbiter_tree_node{};
     s32 m_priority{};

src/core/hle/kernel/kernel.cpp

@@ -99,13 +99,6 @@ struct KernelCore::Impl {
         RegisterHostThread(nullptr);
     }
 
-    void InitializeCores() {
-        for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
-            cores[core_id]->Initialize((*application_process).Is64Bit());
-            system.ApplicationMemory().SetCurrentPageTable(*application_process, core_id);
-        }
-    }
-
     void TerminateApplicationProcess() {
         application_process.load()->Terminate();
     }
@@ -142,7 +135,6 @@ struct KernelCore::Impl {
                 obj = nullptr;
             }
         };
-        CleanupObject(hid_shared_mem);
         CleanupObject(font_shared_mem);
         CleanupObject(irs_shared_mem);
         CleanupObject(time_shared_mem);
@@ -205,7 +197,7 @@ struct KernelCore::Impl {
             const s32 core{static_cast<s32>(i)};
 
             schedulers[i] = std::make_unique<Kernel::KScheduler>(system.Kernel());
-            cores[i] = std::make_unique<Kernel::PhysicalCore>(i, system, *schedulers[i]);
+            cores[i] = std::make_unique<Kernel::PhysicalCore>(system.Kernel(), i);
 
             auto* main_thread{Kernel::KThread::Create(system.Kernel())};
             main_thread->SetCurrentCore(core);
@@ -751,22 +743,16 @@ struct KernelCore::Impl {
     void InitializeHackSharedMemory(KernelCore& kernel) {
         // Setup memory regions for emulated processes
         // TODO(bunnei): These should not be hardcoded regions initialized within the kernel
-        constexpr std::size_t hid_size{0x40000};
         constexpr std::size_t font_size{0x1100000};
         constexpr std::size_t irs_size{0x8000};
         constexpr std::size_t time_size{0x1000};
         constexpr std::size_t hidbus_size{0x1000};
 
-        hid_shared_mem = KSharedMemory::Create(system.Kernel());
         font_shared_mem = KSharedMemory::Create(system.Kernel());
         irs_shared_mem = KSharedMemory::Create(system.Kernel());
         time_shared_mem = KSharedMemory::Create(system.Kernel());
         hidbus_shared_mem = KSharedMemory::Create(system.Kernel());
 
-        hid_shared_mem->Initialize(system.DeviceMemory(), nullptr, Svc::MemoryPermission::None,
-                                   Svc::MemoryPermission::Read, hid_size);
-        KSharedMemory::Register(kernel, hid_shared_mem);
-
         font_shared_mem->Initialize(system.DeviceMemory(), nullptr, Svc::MemoryPermission::None,
                                     Svc::MemoryPermission::Read, font_size);
         KSharedMemory::Register(kernel, font_shared_mem);
@@ -880,10 +866,6 @@ void KernelCore::Initialize() {
     impl->Initialize(*this);
 }
 
-void KernelCore::InitializeCores() {
-    impl->InitializeCores();
-}
-
 void KernelCore::Shutdown() {
     impl->Shutdown();
 }
@@ -993,21 +975,6 @@ const KAutoObjectWithListContainer& KernelCore::ObjectListContainer() const {
     return *impl->global_object_list_container;
 }
 
-void KernelCore::InvalidateAllInstructionCaches() {
-    for (auto& physical_core : impl->cores) {
-        physical_core->ArmInterface().ClearInstructionCache();
-    }
-}
-
-void KernelCore::InvalidateCpuInstructionCacheRange(KProcessAddress addr, std::size_t size) {
-    for (auto& physical_core : impl->cores) {
-        if (!physical_core->IsInitialized()) {
-            continue;
-        }
-        physical_core->ArmInterface().InvalidateCacheRange(GetInteger(addr), size);
-    }
-}
-
 void KernelCore::PrepareReschedule(std::size_t id) {
     // TODO: Reimplement, this
 }
@@ -1216,14 +1183,6 @@ const KSystemResource& KernelCore::GetSystemSystemResource() const {
     return *impl->sys_system_resource;
 }
 
-Kernel::KSharedMemory& KernelCore::GetHidSharedMem() {
-    return *impl->hid_shared_mem;
-}
-
-const Kernel::KSharedMemory& KernelCore::GetHidSharedMem() const {
-    return *impl->hid_shared_mem;
-}
-
 Kernel::KSharedMemory& KernelCore::GetFontSharedMem() {
     return *impl->font_shared_mem;
 }
@@ -1366,6 +1325,7 @@ struct KernelCore::SlabHeapContainer {
     KSlabHeap<KProcess> process;
     KSlabHeap<KResourceLimit> resource_limit;
     KSlabHeap<KSession> session;
+    KSlabHeap<KLightSession> light_session;
     KSlabHeap<KSharedMemory> shared_memory;
     KSlabHeap<KSharedMemoryInfo> shared_memory_info;
     KSlabHeap<KThread> thread;
@@ -1396,6 +1356,8 @@ KSlabHeap<T>& KernelCore::SlabHeap() {
         return slab_heap_container->resource_limit;
     } else if constexpr (std::is_same_v<T, KSession>) {
         return slab_heap_container->session;
+    } else if constexpr (std::is_same_v<T, KLightSession>) {
+        return slab_heap_container->light_session;
     } else if constexpr (std::is_same_v<T, KSharedMemory>) {
         return slab_heap_container->shared_memory;
     } else if constexpr (std::is_same_v<T, KSharedMemoryInfo>) {
@@ -1433,6 +1395,7 @@ template KSlabHeap<KPort>& KernelCore::SlabHeap();
 template KSlabHeap<KProcess>& KernelCore::SlabHeap();
 template KSlabHeap<KResourceLimit>& KernelCore::SlabHeap();
 template KSlabHeap<KSession>& KernelCore::SlabHeap();
+template KSlabHeap<KLightSession>& KernelCore::SlabHeap();
 template KSlabHeap<KSharedMemory>& KernelCore::SlabHeap();
 template KSlabHeap<KSharedMemoryInfo>& KernelCore::SlabHeap();
 template KSlabHeap<KThread>& KernelCore::SlabHeap();

src/core/hle/kernel/kernel.h

@@ -104,9 +104,6 @@ public:
     /// Resets the kernel to a clean slate for use.
     void Initialize();
 
-    /// Initializes the CPU cores.
-    void InitializeCores();
-
     /// Clears all resources in use by the kernel instance.
     void Shutdown();
 
@@ -181,10 +178,6 @@ public:
 
     const KAutoObjectWithListContainer& ObjectListContainer() const;
 
-    void InvalidateAllInstructionCaches();
-
-    void InvalidateCpuInstructionCacheRange(KProcessAddress addr, std::size_t size);
-
     /// Registers all kernel objects with the global emulation state, this is purely for tracking
     /// leaks after emulation has been shutdown.
     void RegisterKernelObject(KAutoObject* object);
@@ -246,12 +239,6 @@ public:
     /// Gets the system resource manager.
     const KSystemResource& GetSystemSystemResource() const;
 
-    /// Gets the shared memory object for HID services.
-    Kernel::KSharedMemory& GetHidSharedMem();
-
-    /// Gets the shared memory object for HID services.
-    const Kernel::KSharedMemory& GetHidSharedMem() const;
-
     /// Gets the shared memory object for font services.
     Kernel::KSharedMemory& GetFontSharedMem();
 

src/core/hle/kernel/physical_core.cpp

@@ -1,62 +1,206 @@
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include "common/scope_exit.h"
 #include "common/settings.h"
-#include "core/arm/dynarmic/arm_dynarmic_32.h"
-#include "core/arm/dynarmic/arm_dynarmic_64.h"
-#ifdef HAS_NCE
-#include "core/arm/nce/arm_nce.h"
-#endif
 #include "core/core.h"
-#include "core/hle/kernel/k_scheduler.h"
+#include "core/debugger/debugger.h"
+#include "core/hle/kernel/k_process.h"
+#include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/physical_core.h"
+#include "core/hle/kernel/svc.h"
 
 namespace Kernel {
 
-PhysicalCore::PhysicalCore(std::size_t core_index, Core::System& system, KScheduler& scheduler)
-    : m_core_index{core_index}, m_system{system}, m_scheduler{scheduler} {
-#if defined(ARCHITECTURE_x86_64) || defined(ARCHITECTURE_arm64)
-    // TODO(bunnei): Initialization relies on a core being available. We may later replace this with
-    // an NCE interface or a 32-bit instance of Dynarmic. This should be abstracted out to a CPU
-    // manager.
-    auto& kernel = system.Kernel();
-    m_arm_interface = std::make_unique<Core::ARM_Dynarmic_64>(
-        system, kernel.IsMulticore(),
-        reinterpret_cast<Core::DynarmicExclusiveMonitor&>(kernel.GetExclusiveMonitor()),
-        m_core_index);
-#else
-#error Platform not supported yet.
-#endif
+PhysicalCore::PhysicalCore(KernelCore& kernel, std::size_t core_index)
+    : m_kernel{kernel}, m_core_index{core_index} {
+    m_is_single_core = !kernel.IsMulticore();
 }
-
 PhysicalCore::~PhysicalCore() = default;
 
-void PhysicalCore::Initialize(bool is_64_bit) {
-#if defined(HAS_NCE)
-    if (Settings::IsNceEnabled()) {
-        m_arm_interface = std::make_unique<Core::ARM_NCE>(m_system, m_system.Kernel().IsMulticore(),
-                                                          m_core_index);
-        return;
+void PhysicalCore::RunThread(Kernel::KThread* thread) {
+    auto* process = thread->GetOwnerProcess();
+    auto& system = m_kernel.System();
+    auto* interface = process->GetArmInterface(m_core_index);
+
+    interface->Initialize();
+
+    const auto EnterContext = [&]() {
+        system.EnterCPUProfile();
+
+        // Lock the core context.
+        std::scoped_lock lk{m_guard};
+
+        // Check if we are already interrupted. If we are, we can just stop immediately.
+        if (m_is_interrupted) {
+            return false;
+        }
+
+        // Mark that we are running.
+        m_arm_interface = interface;
+        m_current_thread = thread;
+
+        // Acquire the lock on the thread parameters.
+        // This allows us to force synchronization with Interrupt.
+        interface->LockThread(thread);
+
+        return true;
+    };
+
+    const auto ExitContext = [&]() {
+        // Unlock the thread.
+        interface->UnlockThread(thread);
+
+        // Lock the core context.
+        std::scoped_lock lk{m_guard};
+
+        // On exit, we no longer are running.
+        m_arm_interface = nullptr;
+        m_current_thread = nullptr;
+
+        system.ExitCPUProfile();
+    };
+
+    while (true) {
+        // If the thread is scheduled for termination, exit.
+        if (thread->HasDpc() && thread->IsTerminationRequested()) {
+            thread->Exit();
+        }
+
+        // Notify the debugger and go to sleep if a step was performed
+        // and this thread has been scheduled again.
+        if (thread->GetStepState() == StepState::StepPerformed) {
+            system.GetDebugger().NotifyThreadStopped(thread);
+            thread->RequestSuspend(SuspendType::Debug);
+            return;
+        }
+
+        // Otherwise, run the thread.
+        Core::HaltReason hr{};
+        {
+            // If we were interrupted, exit immediately.
+            if (!EnterContext()) {
+                return;
+            }
+
+            if (thread->GetStepState() == StepState::StepPending) {
+                hr = interface->StepThread(thread);
+
+                if (True(hr & Core::HaltReason::StepThread)) {
+                    thread->SetStepState(StepState::StepPerformed);
+                }
+            } else {
+                hr = interface->RunThread(thread);
+            }
+
+            ExitContext();
+        }
+
+        // Determine why we stopped.
+        const bool supervisor_call = True(hr & Core::HaltReason::SupervisorCall);
+        const bool prefetch_abort = True(hr & Core::HaltReason::PrefetchAbort);
+        const bool breakpoint = True(hr & Core::HaltReason::InstructionBreakpoint);
+        const bool data_abort = True(hr & Core::HaltReason::DataAbort);
+        const bool interrupt = True(hr & Core::HaltReason::BreakLoop);
+
+        // Since scheduling may occur here, we cannot use any cached
+        // state after returning from calls we make.
+
+        // Notify the debugger and go to sleep if a breakpoint was hit,
+        // or if the thread is unable to continue for any reason.
+        if (breakpoint || prefetch_abort) {
+            if (breakpoint) {
+                interface->RewindBreakpointInstruction();
+            }
+            if (system.DebuggerEnabled()) {
+                system.GetDebugger().NotifyThreadStopped(thread);
+            } else {
+                interface->LogBacktrace(process);
+            }
+            thread->RequestSuspend(SuspendType::Debug);
+            return;
+        }
+
+        // Notify the debugger and go to sleep on data abort.
+        if (data_abort) {
+            if (system.DebuggerEnabled()) {
+                system.GetDebugger().NotifyThreadWatchpoint(thread, *interface->HaltedWatchpoint());
+            }
+            thread->RequestSuspend(SuspendType::Debug);
+            return;
+        }
+
+        // Handle system calls.
+        if (supervisor_call) {
+            // Perform call.
+            Svc::Call(system, interface->GetSvcNumber());
+            return;
+        }
+
+        // Handle external interrupt sources.
+        if (interrupt || m_is_single_core) {
+            return;
+        }
     }
-#endif
-#if defined(ARCHITECTURE_x86_64) || defined(ARCHITECTURE_arm64)
-    auto& kernel = m_system.Kernel();
-    if (!is_64_bit) {
-        // We already initialized a 64-bit core, replace with a 32-bit one.
-        m_arm_interface = std::make_unique<Core::ARM_Dynarmic_32>(
-            m_system, kernel.IsMulticore(),
-            reinterpret_cast<Core::DynarmicExclusiveMonitor&>(kernel.GetExclusiveMonitor()),
-            m_core_index);
-    }
-#else
-#error Platform not supported yet.
-#endif
 }
 
-void PhysicalCore::Run() {
-    m_arm_interface->Run();
-    m_arm_interface->ClearExclusiveState();
+void PhysicalCore::LoadContext(const KThread* thread) {
+    auto* const process = thread->GetOwnerProcess();
+    if (!process) {
+        // Kernel threads do not run on emulated CPU cores.
+        return;
+    }
+
+    auto* interface = process->GetArmInterface(m_core_index);
+    if (interface) {
+        interface->SetContext(thread->GetContext());
+        interface->SetTpidrroEl0(GetInteger(thread->GetTlsAddress()));
+        interface->SetWatchpointArray(&process->GetWatchpoints());
+    }
+}
+
+void PhysicalCore::LoadSvcArguments(const KProcess& process, std::span<const uint64_t, 8> args) {
+    process.GetArmInterface(m_core_index)->SetSvcArguments(args);
+}
+
+void PhysicalCore::SaveContext(KThread* thread) const {
+    auto* const process = thread->GetOwnerProcess();
+    if (!process) {
+        // Kernel threads do not run on emulated CPU cores.
+        return;
+    }
+
+    auto* interface = process->GetArmInterface(m_core_index);
+    if (interface) {
+        interface->GetContext(thread->GetContext());
+    }
+}
+
+void PhysicalCore::SaveSvcArguments(KProcess& process, std::span<uint64_t, 8> args) const {
+    process.GetArmInterface(m_core_index)->GetSvcArguments(args);
+}
+
+void PhysicalCore::CloneFpuStatus(KThread* dst) const {
+    auto* process = dst->GetOwnerProcess();
+
+    Svc::ThreadContext ctx{};
+    process->GetArmInterface(m_core_index)->GetContext(ctx);
+
+    dst->GetContext().fpcr = ctx.fpcr;
+    dst->GetContext().fpsr = ctx.fpsr;
+}
+
+void PhysicalCore::LogBacktrace() {
+    auto* process = GetCurrentProcessPointer(m_kernel);
+    if (!process) {
+        return;
+    }
+
+    auto* interface = process->GetArmInterface(m_core_index);
+    if (interface) {
+        interface->LogBacktrace(process);
+    }
 }
 
 void PhysicalCore::Idle() {
@@ -69,16 +213,31 @@ bool PhysicalCore::IsInterrupted() const {
 }
 
 void PhysicalCore::Interrupt() {
-    std::unique_lock lk{m_guard};
+    // Lock core context.
+    std::scoped_lock lk{m_guard};
+
+    // Load members.
+    auto* arm_interface = m_arm_interface;
+    auto* thread = m_current_thread;
+
+    // Add interrupt flag.
     m_is_interrupted = true;
-    m_arm_interface->SignalInterrupt();
-    m_on_interrupt.notify_all();
+
+    // Interrupt ourselves.
+    m_on_interrupt.notify_one();
+
+    // If there is no thread running, we are done.
+    if (arm_interface == nullptr) {
+        return;
+    }
+
+    // Interrupt the CPU.
+    arm_interface->SignalInterrupt(thread);
 }
 
 void PhysicalCore::ClearInterrupt() {
-    std::unique_lock lk{m_guard};
+    std::scoped_lock lk{m_guard};
     m_is_interrupted = false;
-    m_arm_interface->ClearInterrupt();
 }
 
 } // namespace Kernel

src/core/hle/kernel/physical_core.h

@@ -11,7 +11,7 @@
 #include "core/arm/arm_interface.h"
 
 namespace Kernel {
-class KScheduler;
+class KernelCore;
 } // namespace Kernel
 
 namespace Core {
@@ -23,62 +23,55 @@ namespace Kernel {
 
 class PhysicalCore {
 public:
-    PhysicalCore(std::size_t core_index_, Core::System& system_, KScheduler& scheduler_);
+    PhysicalCore(KernelCore& kernel, std::size_t core_index);
     ~PhysicalCore();
 
     YUZU_NON_COPYABLE(PhysicalCore);
     YUZU_NON_MOVEABLE(PhysicalCore);
 
-    /// Initialize the core for the specified parameters.
-    void Initialize(bool is_64_bit);
+    // Execute guest code running on the given thread.
+    void RunThread(KThread* thread);
 
-    /// Execute current jit state
-    void Run();
+    // Copy context from thread to current core.
+    void LoadContext(const KThread* thread);
+    void LoadSvcArguments(const KProcess& process, std::span<const uint64_t, 8> args);
 
+    // Copy context from current core to thread.
+    void SaveContext(KThread* thread) const;
+    void SaveSvcArguments(KProcess& process, std::span<uint64_t, 8> args) const;
+
+    // Copy floating point status registers to the target thread.
+    void CloneFpuStatus(KThread* dst) const;
+
+    // Log backtrace of current processor state.
+    void LogBacktrace();
+
+    // Wait for an interrupt.
     void Idle();
 
-    /// Interrupt this physical core.
+    // Interrupt this core.
     void Interrupt();
 
-    /// Clear this core's interrupt
+    // Clear this core's interrupt.
     void ClearInterrupt();
 
-    /// Check if this core is interrupted
+    // Check if this core is interrupted.
     bool IsInterrupted() const;
 
-    bool IsInitialized() const {
-        return m_arm_interface != nullptr;
-    }
-
-    Core::ARM_Interface& ArmInterface() {
-        return *m_arm_interface;
-    }
-
-    const Core::ARM_Interface& ArmInterface() const {
-        return *m_arm_interface;
-    }
-
     std::size_t CoreIndex() const {
         return m_core_index;
     }
 
-    Kernel::KScheduler& Scheduler() {
-        return m_scheduler;
-    }
-
-    const Kernel::KScheduler& Scheduler() const {
-        return m_scheduler;
-    }
-
 private:
+    KernelCore& m_kernel;
     const std::size_t m_core_index;
-    Core::System& m_system;
-    Kernel::KScheduler& m_scheduler;
 
     std::mutex m_guard;
     std::condition_variable m_on_interrupt;
-    std::unique_ptr<Core::ARM_Interface> m_arm_interface;
+    Core::ArmInterface* m_arm_interface{};
+    KThread* m_current_thread{};
     bool m_is_interrupted{};
+    bool m_is_single_core{};
 };
 
 } // namespace Kernel

src/core/hle/kernel/svc.h

@@ -9,6 +9,8 @@ namespace Core {
 class System;
 }
 
+#include <span>
+
 #include "common/common_types.h"
 #include "core/hle/kernel/svc_types.h"
 #include "core/hle/result.h"
@@ -520,15 +522,15 @@ void CallSecureMonitor64From32(Core::System& system, ilp32::SecureMonitorArgumen
 void CallSecureMonitor64(Core::System& system, lp64::SecureMonitorArguments* args);
 
 // Defined in svc_light_ipc.cpp.
-void SvcWrap_ReplyAndReceiveLight64From32(Core::System& system);
-void SvcWrap_ReplyAndReceiveLight64(Core::System& system);
+void SvcWrap_ReplyAndReceiveLight64From32(Core::System& system, std::span<uint64_t, 8> args);
+void SvcWrap_ReplyAndReceiveLight64(Core::System& system, std::span<uint64_t, 8> args);
 
-void SvcWrap_SendSyncRequestLight64From32(Core::System& system);
-void SvcWrap_SendSyncRequestLight64(Core::System& system);
+void SvcWrap_SendSyncRequestLight64From32(Core::System& system, std::span<uint64_t, 8> args);
+void SvcWrap_SendSyncRequestLight64(Core::System& system, std::span<uint64_t, 8> args);
 
 // Defined in svc_secure_monitor_call.cpp.
-void SvcWrap_CallSecureMonitor64From32(Core::System& system);
-void SvcWrap_CallSecureMonitor64(Core::System& system);
+void SvcWrap_CallSecureMonitor64From32(Core::System& system, std::span<uint64_t, 8> args);
+void SvcWrap_CallSecureMonitor64(Core::System& system, std::span<uint64_t, 8> args);
 
 // Perform a supervisor call by index.
 void Call(Core::System& system, u32 imm);

src/core/hle/kernel/svc/svc_exception.cpp

@@ -103,9 +103,7 @@ void Break(Core::System& system, BreakReason reason, u64 info1, u64 info2) {
 
         handle_debug_buffer(info1, info2);
 
-        auto* const current_thread = GetCurrentThreadPointer(system.Kernel());
-        const auto thread_processor_id = current_thread->GetActiveCore();
-        system.ArmInterface(static_cast<std::size_t>(thread_processor_id)).LogBacktrace();
+        system.CurrentPhysicalCore().LogBacktrace();
     }
 
     const bool is_hbl = GetCurrentProcess(system.Kernel()).IsHbl();

src/core/hle/kernel/svc/svc_ipc.cpp

@@ -7,71 +7,39 @@
 #include "core/hle/kernel/k_client_session.h"
 #include "core/hle/kernel/k_hardware_timer.h"
 #include "core/hle/kernel/k_process.h"
+#include "core/hle/kernel/k_scoped_resource_reservation.h"
 #include "core/hle/kernel/k_server_session.h"
+#include "core/hle/kernel/k_session.h"
 #include "core/hle/kernel/svc.h"
 #include "core/hle/kernel/svc_results.h"
 
 namespace Kernel::Svc {
 
-/// Makes a blocking IPC call to a service.
-Result SendSyncRequest(Core::System& system, Handle handle) {
-    // Get the client session from its handle.
+namespace {
+
+Result SendSyncRequestImpl(KernelCore& kernel, uintptr_t message, size_t buffer_size,
+                           Handle session_handle) {
+    // Get the client session.
     KScopedAutoObject session =
-        GetCurrentProcess(system.Kernel()).GetHandleTable().GetObject<KClientSession>(handle);
+        GetCurrentProcess(kernel).GetHandleTable().GetObject<KClientSession>(session_handle);
     R_UNLESS(session.IsNotNull(), ResultInvalidHandle);
 
-    LOG_TRACE(Kernel_SVC, "called handle=0x{:08X}", handle);
+    // Get the parent, and persist a reference to it until we're done.
+    KScopedAutoObject parent = session->GetParent();
+    ASSERT(parent.IsNotNull());
 
-    R_RETURN(session->SendSyncRequest());
+    // Send the request.
+    R_RETURN(session->SendSyncRequest(message, buffer_size));
 }
 
-Result SendSyncRequestWithUserBuffer(Core::System& system, uint64_t message_buffer,
-                                     uint64_t message_buffer_size, Handle session_handle) {
-    UNIMPLEMENTED();
-    R_THROW(ResultNotImplemented);
-}
-
-Result SendAsyncRequestWithUserBuffer(Core::System& system, Handle* out_event_handle,
-                                      uint64_t message_buffer, uint64_t message_buffer_size,
-                                      Handle session_handle) {
-    UNIMPLEMENTED();
-    R_THROW(ResultNotImplemented);
-}
-
-Result ReplyAndReceive(Core::System& system, s32* out_index, uint64_t handles_addr, s32 num_handles,
-                       Handle reply_target, s64 timeout_ns) {
-    // Ensure number of handles is valid.
-    R_UNLESS(0 <= num_handles && num_handles <= ArgumentHandleCountMax, ResultOutOfRange);
-
-    // Get the synchronization context.
-    auto& kernel = system.Kernel();
-    auto& handle_table = GetCurrentProcess(kernel).GetHandleTable();
-    auto objs = GetCurrentThread(kernel).GetSynchronizationObjectBuffer();
-    auto handles = GetCurrentThread(kernel).GetHandleBuffer();
-
-    // Copy user handles.
-    if (num_handles > 0) {
-        // Get the handles.
-        R_UNLESS(GetCurrentMemory(kernel).ReadBlock(handles_addr, handles.data(),
-                                                    sizeof(Handle) * num_handles),
-                 ResultInvalidPointer);
-
-        // Convert the handles to objects.
-        R_UNLESS(handle_table.GetMultipleObjects<KSynchronizationObject>(
-                     objs.data(), handles.data(), num_handles),
-                 ResultInvalidHandle);
-    }
-
-    // Ensure handles are closed when we're done.
-    SCOPE_EXIT({
-        for (auto i = 0; i < num_handles; ++i) {
-            objs[i]->Close();
-        }
-    });
-
+Result ReplyAndReceiveImpl(KernelCore& kernel, int32_t* out_index, uintptr_t message,
+                           size_t buffer_size, KPhysicalAddress message_paddr,
+                           KSynchronizationObject** objs, int32_t num_objects, Handle reply_target,
+                           int64_t timeout_ns) {
     // Reply to the target, if one is specified.
     if (reply_target != InvalidHandle) {
-        KScopedAutoObject session = handle_table.GetObject<KServerSession>(reply_target);
+        KScopedAutoObject session =
+            GetCurrentProcess(kernel).GetHandleTable().GetObject<KServerSession>(reply_target);
         R_UNLESS(session.IsNotNull(), ResultInvalidHandle);
 
         // If we fail to reply, we want to set the output index to -1.
@@ -81,57 +49,223 @@ Result ReplyAndReceive(Core::System& system, s32* out_index, uint64_t handles_ad
 
         // Send the reply.
         R_TRY(session->SendReply());
+        // R_TRY(session->SendReply(message, buffer_size, message_paddr));
     }
 
-    // Convert the timeout from nanoseconds to ticks.
-    // NOTE: Nintendo does not use this conversion logic in WaitSynchronization...
-    s64 timeout;
-    if (timeout_ns > 0) {
-        const s64 offset_tick(timeout_ns);
-        if (offset_tick > 0) {
-            timeout = kernel.HardwareTimer().GetTick() + offset_tick + 2;
-            if (timeout <= 0) {
+    // Receive a message.
+    {
+        // Convert the timeout from nanoseconds to ticks.
+        // NOTE: Nintendo does not use this conversion logic in WaitSynchronization...
+        s64 timeout;
+        if (timeout_ns > 0) {
+            const s64 offset_tick(timeout_ns);
+            if (offset_tick > 0) {
+                timeout = kernel.HardwareTimer().GetTick() + offset_tick + 2;
+                if (timeout <= 0) {
+                    timeout = std::numeric_limits<s64>::max();
+                }
+            } else {
                 timeout = std::numeric_limits<s64>::max();
             }
         } else {
-            timeout = std::numeric_limits<s64>::max();
-        }
-    } else {
-        timeout = timeout_ns;
-    }
-
-    // Wait for a message.
-    while (true) {
-        // Wait for an object.
-        s32 index;
-        Result result = KSynchronizationObject::Wait(kernel, std::addressof(index), objs.data(),
-                                                     num_handles, timeout);
-        if (result == ResultTimedOut) {
-            R_RETURN(result);
+            timeout = timeout_ns;
         }
 
-        // Receive the request.
-        if (R_SUCCEEDED(result)) {
-            KServerSession* session = objs[index]->DynamicCast<KServerSession*>();
-            if (session != nullptr) {
-                result = session->ReceiveRequest();
-                if (result == ResultNotFound) {
-                    continue;
+        // Wait for a message.
+        while (true) {
+            // Wait for an object.
+            s32 index;
+            Result result = KSynchronizationObject::Wait(kernel, std::addressof(index), objs,
+                                                         num_objects, timeout);
+            if (ResultTimedOut == result) {
+                R_THROW(result);
+            }
+
+            // Receive the request.
+            if (R_SUCCEEDED(result)) {
+                KServerSession* session = objs[index]->DynamicCast<KServerSession*>();
+                if (session != nullptr) {
+                    // result = session->ReceiveRequest(message, buffer_size, message_paddr);
+                    result = session->ReceiveRequest();
+                    if (ResultNotFound == result) {
+                        continue;
+                    }
                 }
             }
-        }
 
-        *out_index = index;
-        R_RETURN(result);
+            *out_index = index;
+            R_RETURN(result);
+        }
     }
 }
 
-Result ReplyAndReceiveWithUserBuffer(Core::System& system, int32_t* out_index,
-                                     uint64_t message_buffer, uint64_t message_buffer_size,
-                                     uint64_t handles, int32_t num_handles, Handle reply_target,
-                                     int64_t timeout_ns) {
-    UNIMPLEMENTED();
-    R_THROW(ResultNotImplemented);
+Result ReplyAndReceiveImpl(KernelCore& kernel, int32_t* out_index, uintptr_t message,
+                           size_t buffer_size, KPhysicalAddress message_paddr,
+                           KProcessAddress user_handles, int32_t num_handles, Handle reply_target,
+                           int64_t timeout_ns) {
+    // Ensure number of handles is valid.
+    R_UNLESS(0 <= num_handles && num_handles <= Svc::ArgumentHandleCountMax, ResultOutOfRange);
+
+    // Get the synchronization context.
+    auto& process = GetCurrentProcess(kernel);
+    auto& thread = GetCurrentThread(kernel);
+    auto& handle_table = process.GetHandleTable();
+    KSynchronizationObject** objs = thread.GetSynchronizationObjectBuffer().data();
+    Handle* handles = thread.GetHandleBuffer().data();
+
+    // Copy user handles.
+    if (num_handles > 0) {
+        // Ensure that we can try to get the handles.
+        R_UNLESS(process.GetPageTable().Contains(user_handles, num_handles * sizeof(Handle)),
+                 ResultInvalidPointer);
+
+        // Get the handles
+        R_UNLESS(
+            GetCurrentMemory(kernel).ReadBlock(user_handles, handles, sizeof(Handle) * num_handles),
+            ResultInvalidPointer);
+
+        // Convert the handles to objects.
+        R_UNLESS(
+            handle_table.GetMultipleObjects<KSynchronizationObject>(objs, handles, num_handles),
+            ResultInvalidHandle);
+    }
+
+    // Ensure handles are closed when we're done.
+    SCOPE_EXIT({
+        for (auto i = 0; i < num_handles; ++i) {
+            objs[i]->Close();
+        }
+    });
+
+    R_RETURN(ReplyAndReceiveImpl(kernel, out_index, message, buffer_size, message_paddr, objs,
+                                 num_handles, reply_target, timeout_ns));
+}
+
+} // namespace
+
+/// Makes a blocking IPC call to a service.
+Result SendSyncRequest(Core::System& system, Handle session_handle) {
+    R_RETURN(SendSyncRequestImpl(system.Kernel(), 0, 0, session_handle));
+}
+
+Result SendSyncRequestWithUserBuffer(Core::System& system, uint64_t message, uint64_t buffer_size,
+                                     Handle session_handle) {
+    auto& kernel = system.Kernel();
+
+    // Validate that the message buffer is page aligned and does not overflow.
+    R_UNLESS(Common::IsAligned(message, PageSize), ResultInvalidAddress);
+    R_UNLESS(buffer_size > 0, ResultInvalidSize);
+    R_UNLESS(Common::IsAligned(buffer_size, PageSize), ResultInvalidSize);
+    R_UNLESS(message < message + buffer_size, ResultInvalidCurrentMemory);
+
+    // Get the process page table.
+    auto& page_table = GetCurrentProcess(kernel).GetPageTable();
+
+    // Lock the message buffer.
+    R_TRY(page_table.LockForIpcUserBuffer(nullptr, message, buffer_size));
+
+    {
+        // If we fail to send the message, unlock the message buffer.
+        ON_RESULT_FAILURE {
+            page_table.UnlockForIpcUserBuffer(message, buffer_size);
+        };
+
+        // Send the request.
+        ASSERT(message != 0);
+        R_TRY(SendSyncRequestImpl(kernel, message, buffer_size, session_handle));
+    }
+
+    // We successfully processed, so try to unlock the message buffer.
+    R_RETURN(page_table.UnlockForIpcUserBuffer(message, buffer_size));
+}
+
+Result SendAsyncRequestWithUserBuffer(Core::System& system, Handle* out_event_handle,
+                                      uint64_t message, uint64_t buffer_size,
+                                      Handle session_handle) {
+    // Get the process and handle table.
+    auto& process = GetCurrentProcess(system.Kernel());
+    auto& handle_table = process.GetHandleTable();
+
+    // Reserve a new event from the process resource limit.
+    KScopedResourceReservation event_reservation(std::addressof(process),
+                                                 Svc::LimitableResource::EventCountMax);
+    R_UNLESS(event_reservation.Succeeded(), ResultLimitReached);
+
+    // Get the client session.
+    KScopedAutoObject session = process.GetHandleTable().GetObject<KClientSession>(session_handle);
+    R_UNLESS(session.IsNotNull(), ResultInvalidHandle);
+
+    // Get the parent, and persist a reference to it until we're done.
+    KScopedAutoObject parent = session->GetParent();
+    ASSERT(parent.IsNotNull());
+
+    // Create a new event.
+    KEvent* event = KEvent::Create(system.Kernel());
+    R_UNLESS(event != nullptr, ResultOutOfResource);
+
+    // Initialize the event.
+    event->Initialize(std::addressof(process));
+
+    // Commit our reservation.
+    event_reservation.Commit();
+
+    // At end of scope, kill the standing references to the sub events.
+    SCOPE_EXIT({
+        event->GetReadableEvent().Close();
+        event->Close();
+    });
+
+    // Register the event.
+    KEvent::Register(system.Kernel(), event);
+
+    // Add the readable event to the handle table.
+    R_TRY(handle_table.Add(out_event_handle, std::addressof(event->GetReadableEvent())));
+
+    // Ensure that if we fail to send the request, we close the readable handle.
+    ON_RESULT_FAILURE {
+        handle_table.Remove(*out_event_handle);
+    };
+
+    // Send the async request.
+    R_RETURN(session->SendAsyncRequest(event, message, buffer_size));
+}
+
+Result ReplyAndReceive(Core::System& system, s32* out_index, uint64_t handles, s32 num_handles,
+                       Handle reply_target, s64 timeout_ns) {
+    R_RETURN(ReplyAndReceiveImpl(system.Kernel(), out_index, 0, 0, 0, handles, num_handles,
+                                 reply_target, timeout_ns));
+}
+
+Result ReplyAndReceiveWithUserBuffer(Core::System& system, int32_t* out_index, uint64_t message,
+                                     uint64_t buffer_size, uint64_t handles, int32_t num_handles,
+                                     Handle reply_target, int64_t timeout_ns) {
+    // Validate that the message buffer is page aligned and does not overflow.
+    R_UNLESS(Common::IsAligned(message, PageSize), ResultInvalidAddress);
+    R_UNLESS(buffer_size > 0, ResultInvalidSize);
+    R_UNLESS(Common::IsAligned(buffer_size, PageSize), ResultInvalidSize);
+    R_UNLESS(message < message + buffer_size, ResultInvalidCurrentMemory);
+
+    // Get the process page table.
+    auto& page_table = GetCurrentProcess(system.Kernel()).GetPageTable();
+
+    // Lock the message buffer, getting its physical address.
+    KPhysicalAddress message_paddr;
+    R_TRY(page_table.LockForIpcUserBuffer(std::addressof(message_paddr), message, buffer_size));
+
+    {
+        // If we fail to send the message, unlock the message buffer.
+        ON_RESULT_FAILURE {
+            page_table.UnlockForIpcUserBuffer(message, buffer_size);
+        };
+
+        // Reply/Receive the request.
+        ASSERT(message != 0);
+        R_TRY(ReplyAndReceiveImpl(system.Kernel(), out_index, message, buffer_size, message_paddr,
+                                  handles, num_handles, reply_target, timeout_ns));
+    }
+
+    // We successfully processed, so try to unlock the message buffer.
+    R_RETURN(page_table.UnlockForIpcUserBuffer(message, buffer_size));
 }
 
 Result SendSyncRequest64(Core::System& system, Handle session_handle) {

src/core/hle/kernel/svc/svc_light_ipc.cpp

@@ -1,21 +1,40 @@
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include "core/arm/arm_interface.h"
 #include "core/core.h"
+#include "core/hle/kernel/k_light_client_session.h"
+#include "core/hle/kernel/k_light_server_session.h"
+#include "core/hle/kernel/k_process.h"
+#include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/svc.h"
 #include "core/hle/kernel/svc_results.h"
 
 namespace Kernel::Svc {
 
 Result SendSyncRequestLight(Core::System& system, Handle session_handle, u32* args) {
-    UNIMPLEMENTED();
-    R_THROW(ResultNotImplemented);
+    // Get the light client session from its handle.
+    KScopedAutoObject session = GetCurrentProcess(system.Kernel())
+                                    .GetHandleTable()
+                                    .GetObject<KLightClientSession>(session_handle);
+    R_UNLESS(session.IsNotNull(), ResultInvalidHandle);
+
+    // Send the request.
+    R_TRY(session->SendSyncRequest(args));
+
+    R_SUCCEED();
 }
 
 Result ReplyAndReceiveLight(Core::System& system, Handle session_handle, u32* args) {
-    UNIMPLEMENTED();
-    R_THROW(ResultNotImplemented);
+    // Get the light server session from its handle.
+    KScopedAutoObject session = GetCurrentProcess(system.Kernel())
+                                    .GetHandleTable()
+                                    .GetObject<KLightServerSession>(session_handle);
+    R_UNLESS(session.IsNotNull(), ResultInvalidHandle);
+
+    // Handle the request.
+    R_TRY(session->ReplyAndReceive(args));
+
+    R_SUCCEED();
 }
 
 Result SendSyncRequestLight64(Core::System& system, Handle session_handle, u32* args) {
@@ -37,37 +56,36 @@ Result ReplyAndReceiveLight64From32(Core::System& system, Handle session_handle,
 // Custom ABI implementation for light IPC.
 
 template <typename F>
-static void SvcWrap_LightIpc(Core::System& system, F&& cb) {
-    auto& core = system.CurrentArmInterface();
-    std::array<u32, 7> arguments{};
+static void SvcWrap_LightIpc(Core::System& system, std::span<uint64_t, 8> args, F&& cb) {
+    std::array<u32, 7> ipc_args{};
 
-    Handle session_handle = static_cast<Handle>(core.GetReg(0));
+    Handle session_handle = static_cast<Handle>(args[0]);
     for (int i = 0; i < 7; i++) {
-        arguments[i] = static_cast<u32>(core.GetReg(i + 1));
+        ipc_args[i] = static_cast<u32>(args[i + 1]);
     }
 
-    Result ret = cb(system, session_handle, arguments.data());
+    Result ret = cb(system, session_handle, ipc_args.data());
 
-    core.SetReg(0, ret.raw);
+    args[0] = ret.raw;
     for (int i = 0; i < 7; i++) {
-        core.SetReg(i + 1, arguments[i]);
+        args[i + 1] = ipc_args[i];
     }
 }
 
-void SvcWrap_SendSyncRequestLight64(Core::System& system) {
-    SvcWrap_LightIpc(system, SendSyncRequestLight64);
+void SvcWrap_SendSyncRequestLight64(Core::System& system, std::span<uint64_t, 8> args) {
+    SvcWrap_LightIpc(system, args, SendSyncRequestLight64);
 }
 
-void SvcWrap_ReplyAndReceiveLight64(Core::System& system) {
-    SvcWrap_LightIpc(system, ReplyAndReceiveLight64);
+void SvcWrap_ReplyAndReceiveLight64(Core::System& system, std::span<uint64_t, 8> args) {
+    SvcWrap_LightIpc(system, args, ReplyAndReceiveLight64);
 }
 
-void SvcWrap_SendSyncRequestLight64From32(Core::System& system) {
-    SvcWrap_LightIpc(system, SendSyncRequestLight64From32);
+void SvcWrap_SendSyncRequestLight64From32(Core::System& system, std::span<uint64_t, 8> args) {
+    SvcWrap_LightIpc(system, args, SendSyncRequestLight64From32);
 }
 
-void SvcWrap_ReplyAndReceiveLight64From32(Core::System& system) {
-    SvcWrap_LightIpc(system, ReplyAndReceiveLight64From32);
+void SvcWrap_ReplyAndReceiveLight64From32(Core::System& system, std::span<uint64_t, 8> args) {
+    SvcWrap_LightIpc(system, args, ReplyAndReceiveLight64From32);
 }
 
 } // namespace Kernel::Svc

src/core/hle/kernel/svc/svc_port.cpp

@@ -5,6 +5,7 @@
 #include "core/core.h"
 #include "core/hle/kernel/k_client_port.h"
 #include "core/hle/kernel/k_client_session.h"
+#include "core/hle/kernel/k_light_client_session.h"
 #include "core/hle/kernel/k_object_name.h"
 #include "core/hle/kernel/k_port.h"
 #include "core/hle/kernel/k_process.h"
@@ -51,13 +52,73 @@ Result ConnectToNamedPort(Core::System& system, Handle* out, u64 user_name) {
 
 Result CreatePort(Core::System& system, Handle* out_server, Handle* out_client,
                   int32_t max_sessions, bool is_light, uint64_t name) {
-    UNIMPLEMENTED();
-    R_THROW(ResultNotImplemented);
+    auto& kernel = system.Kernel();
+
+    // Ensure max sessions is valid.
+    R_UNLESS(max_sessions > 0, ResultOutOfRange);
+
+    // Get the current handle table.
+    auto& handle_table = GetCurrentProcess(kernel).GetHandleTable();
+
+    // Create a new port.
+    KPort* port = KPort::Create(kernel);
+    R_UNLESS(port != nullptr, ResultOutOfResource);
+
+    // Initialize the port.
+    port->Initialize(max_sessions, is_light, name);
+
+    // Ensure that we clean up the port (and its only references are handle table) on function end.
+    SCOPE_EXIT({
+        port->GetServerPort().Close();
+        port->GetClientPort().Close();
+    });
+
+    // Register the port.
+    KPort::Register(kernel, port);
+
+    // Add the client to the handle table.
+    R_TRY(handle_table.Add(out_client, std::addressof(port->GetClientPort())));
+
+    // Ensure that we maintain a clean handle state on exit.
+    ON_RESULT_FAILURE {
+        handle_table.Remove(*out_client);
+    };
+
+    // Add the server to the handle table.
+    R_RETURN(handle_table.Add(out_server, std::addressof(port->GetServerPort())));
 }
 
-Result ConnectToPort(Core::System& system, Handle* out_handle, Handle port) {
-    UNIMPLEMENTED();
-    R_THROW(ResultNotImplemented);
+Result ConnectToPort(Core::System& system, Handle* out, Handle port) {
+    // Get the current handle table.
+    auto& handle_table = GetCurrentProcess(system.Kernel()).GetHandleTable();
+
+    // Get the client port.
+    KScopedAutoObject client_port = handle_table.GetObject<KClientPort>(port);
+    R_UNLESS(client_port.IsNotNull(), ResultInvalidHandle);
+
+    // Reserve a handle for the port.
+    // NOTE: Nintendo really does write directly to the output handle here.
+    R_TRY(handle_table.Reserve(out));
+    ON_RESULT_FAILURE {
+        handle_table.Unreserve(*out);
+    };
+
+    // Create the session.
+    KAutoObject* session;
+    if (client_port->IsLight()) {
+        R_TRY(client_port->CreateLightSession(
+            reinterpret_cast<KLightClientSession**>(std::addressof(session))));
+    } else {
+        R_TRY(client_port->CreateSession(
+            reinterpret_cast<KClientSession**>(std::addressof(session))));
+    }
+
+    // Register the session.
+    handle_table.Register(*out, session);
+    session->Close();
+
+    // We succeeded.
+    R_SUCCEED();
 }
 
 Result ManageNamedPort(Core::System& system, Handle* out_server_handle, uint64_t user_name,

src/core/hle/kernel/svc/svc_secure_monitor_call.cpp

@@ -22,31 +22,29 @@ void CallSecureMonitor64From32(Core::System& system, ilp32::SecureMonitorArgumen
 
 // Custom ABI for CallSecureMonitor.
 
-void SvcWrap_CallSecureMonitor64(Core::System& system) {
-    auto& core = system.CurrentPhysicalCore().ArmInterface();
-    lp64::SecureMonitorArguments args{};
+void SvcWrap_CallSecureMonitor64(Core::System& system, std::span<uint64_t, 8> args) {
+    lp64::SecureMonitorArguments smc_args{};
     for (int i = 0; i < 8; i++) {
-        args.r[i] = core.GetReg(i);
+        smc_args.r[i] = args[i];
     }
 
-    CallSecureMonitor64(system, std::addressof(args));
+    CallSecureMonitor64(system, std::addressof(smc_args));
 
     for (int i = 0; i < 8; i++) {
-        core.SetReg(i, args.r[i]);
+        args[i] = smc_args.r[i];
     }
 }
 
-void SvcWrap_CallSecureMonitor64From32(Core::System& system) {
-    auto& core = system.CurrentPhysicalCore().ArmInterface();
-    ilp32::SecureMonitorArguments args{};
+void SvcWrap_CallSecureMonitor64From32(Core::System& system, std::span<uint64_t, 8> args) {
+    ilp32::SecureMonitorArguments smc_args{};
     for (int i = 0; i < 8; i++) {
-        args.r[i] = static_cast<u32>(core.GetReg(i));
+        smc_args.r[i] = static_cast<u32>(args[i]);
     }
 
-    CallSecureMonitor64From32(system, std::addressof(args));
+    CallSecureMonitor64From32(system, std::addressof(smc_args));
 
     for (int i = 0; i < 8; i++) {
-        core.SetReg(i, args.r[i]);
+        args[i] = smc_args.r[i];
     }
 }
 

src/core/hle/kernel/svc/svc_session.cpp

@@ -3,8 +3,10 @@
 
 #include "common/scope_exit.h"
 #include "core/core.h"
+#include "core/hle/kernel/k_light_session.h"
 #include "core/hle/kernel/k_process.h"
 #include "core/hle/kernel/k_scoped_resource_reservation.h"
+#include "core/hle/kernel/k_server_port.h"
 #include "core/hle/kernel/k_session.h"
 #include "core/hle/kernel/svc.h"
 
@@ -20,7 +22,7 @@ Result CreateSession(Core::System& system, Handle* out_server, Handle* out_clien
     T* session;
 
     // Reserve a new session from the process resource limit.
-    // FIXME: LimitableResource_SessionCountMax
+    // TODO: Dynamic resource limits
     KScopedResourceReservation session_reservation(std::addressof(process),
                                                    LimitableResource::SessionCountMax);
     if (session_reservation.Succeeded()) {
@@ -92,16 +94,42 @@ Result CreateSession(Core::System& system, Handle* out_server, Handle* out_clien
 Result CreateSession(Core::System& system, Handle* out_server, Handle* out_client, bool is_light,
                      u64 name) {
     if (is_light) {
-        // return CreateSession<KLightSession>(system, out_server, out_client, name);
-        R_THROW(ResultNotImplemented);
+        R_RETURN(CreateSession<KLightSession>(system, out_server, out_client, name));
     } else {
         R_RETURN(CreateSession<KSession>(system, out_server, out_client, name));
     }
 }
 
-Result AcceptSession(Core::System& system, Handle* out_handle, Handle port_handle) {
-    UNIMPLEMENTED();
-    R_THROW(ResultNotImplemented);
+Result AcceptSession(Core::System& system, Handle* out, Handle port_handle) {
+    // Get the current handle table.
+    auto& handle_table = GetCurrentProcess(system.Kernel()).GetHandleTable();
+
+    // Get the server port.
+    KScopedAutoObject port = handle_table.GetObject<KServerPort>(port_handle);
+    R_UNLESS(port.IsNotNull(), ResultInvalidHandle);
+
+    // Reserve an entry for the new session.
+    R_TRY(handle_table.Reserve(out));
+    ON_RESULT_FAILURE {
+        handle_table.Unreserve(*out);
+    };
+
+    // Accept the session.
+    KAutoObject* session;
+    if (port->IsLight()) {
+        session = port->AcceptLightSession();
+    } else {
+        session = port->AcceptSession();
+    }
+
+    // Ensure we accepted successfully.
+    R_UNLESS(session != nullptr, ResultNotFound);
+
+    // Register the session.
+    handle_table.Register(*out, session);
+    session->Close();
+
+    R_SUCCEED();
 }
 
 Result CreateSession64(Core::System& system, Handle* out_server_session_handle,

src/core/hle/kernel/svc/svc_thread.cpp

@@ -90,8 +90,6 @@ Result StartThread(Core::System& system, Handle thread_handle) {
 
 /// Called when a thread exits
 void ExitThread(Core::System& system) {
-    LOG_DEBUG(Kernel_SVC, "called, pc=0x{:08X}", system.CurrentArmInterface().GetPC());
-
     auto* const current_thread = GetCurrentThreadPointer(system.Kernel());
     system.GlobalSchedulerContext().RemoveThread(current_thread);
     current_thread->Exit();
@@ -147,47 +145,19 @@ Result GetThreadContext3(Core::System& system, u64 out_context, Handle thread_ha
     R_UNLESS(thread.IsNotNull(), ResultInvalidHandle);
 
     // Require the handle be to a non-current thread in the current process.
-    const auto* current_process = GetCurrentProcessPointer(kernel);
-    R_UNLESS(current_process == thread->GetOwnerProcess(), ResultInvalidId);
+    R_UNLESS(thread->GetOwnerProcess() == GetCurrentProcessPointer(kernel), ResultInvalidHandle);
+    R_UNLESS(thread.GetPointerUnsafe() != GetCurrentThreadPointer(kernel), ResultBusy);
 
-    // Verify that the thread isn't terminated.
-    R_UNLESS(thread->GetState() != ThreadState::Terminated, ResultTerminationRequested);
+    // Get the thread context.
+    Svc::ThreadContext context{};
+    R_TRY(thread->GetThreadContext3(std::addressof(context)));
 
-    /// Check that the thread is not the current one.
-    /// NOTE: Nintendo does not check this, and thus the following loop will deadlock.
-    R_UNLESS(thread.GetPointerUnsafe() != GetCurrentThreadPointer(kernel), ResultInvalidId);
+    // Copy the thread context to user space.
+    R_UNLESS(
+        GetCurrentMemory(kernel).WriteBlock(out_context, std::addressof(context), sizeof(context)),
+        ResultInvalidPointer);
 
-    // Try to get the thread context until the thread isn't current on any core.
-    while (true) {
-        KScopedSchedulerLock sl{kernel};
-
-        // TODO(bunnei): Enforce that thread is suspended for debug here.
-
-        // If the thread's raw state isn't runnable, check if it's current on some core.
-        if (thread->GetRawState() != ThreadState::Runnable) {
-            bool current = false;
-            for (auto i = 0; i < static_cast<s32>(Core::Hardware::NUM_CPU_CORES); ++i) {
-                if (thread.GetPointerUnsafe() == kernel.Scheduler(i).GetSchedulerCurrentThread()) {
-                    current = true;
-                    break;
-                }
-            }
-
-            // If the thread is current, retry until it isn't.
-            if (current) {
-                continue;
-            }
-        }
-
-        // Get the thread context.
-        static thread_local Common::ScratchBuffer<u8> context;
-        R_TRY(thread->GetThreadContext3(context));
-
-        // Copy the thread context to user space.
-        GetCurrentMemory(kernel).WriteBlock(out_context, context.data(), context.size());
-
-        R_SUCCEED();
-    }
+    R_SUCCEED();
 }
 
 /// Gets the priority for the specified thread

src/core/hle/kernel/svc_generator.py

@@ -374,11 +374,11 @@ def get_registers(parse_result, bitness):
 
 # Collects possibly multiple source registers into the named C++ value.
 def emit_gather(sources, name, type_name, reg_size):
-    get_fn = f"GetReg{reg_size*8}"
+    get_fn = f"GetArg{reg_size*8}"
 
     if len(sources) == 1:
         s, = sources
-        line = f"{name} = Convert<{type_name}>({get_fn}(system, {s}));"
+        line = f"{name} = Convert<{type_name}>({get_fn}(args, {s}));"
         return [line]
 
     var_type = f"std::array<uint{reg_size*8}_t, {len(sources)}>"
@@ -387,7 +387,7 @@ def emit_gather(sources, name, type_name, reg_size):
     ]
     for i in range(0, len(sources)):
         lines.append(
-            f"{name}_gather[{i}] = {get_fn}(system, {sources[i]});")
+            f"{name}_gather[{i}] = {get_fn}(args, {sources[i]});")
 
     lines.append(f"{name} = Convert<{type_name}>({name}_gather);")
     return lines
@@ -396,12 +396,12 @@ def emit_gather(sources, name, type_name, reg_size):
 # Produces one or more statements which assign the named C++ value
 # into possibly multiple registers.
 def emit_scatter(destinations, name, reg_size):
-    set_fn = f"SetReg{reg_size*8}"
+    set_fn = f"SetArg{reg_size*8}"
     reg_type = f"uint{reg_size*8}_t"
 
     if len(destinations) == 1:
         d, = destinations
-        line = f"{set_fn}(system, {d}, Convert<{reg_type}>({name}));"
+        line = f"{set_fn}(args, {d}, Convert<{reg_type}>({name}));"
         return [line]
 
     var_type = f"std::array<{reg_type}, {len(destinations)}>"
@@ -411,7 +411,7 @@ def emit_scatter(destinations, name, reg_size):
 
     for i in range(0, len(destinations)):
         lines.append(
-            f"{set_fn}(system, {destinations[i]}, {name}_scatter[{i}]);")
+            f"{set_fn}(args, {destinations[i]}, {name}_scatter[{i}]);")
 
     return lines
 
@@ -433,7 +433,7 @@ def emit_lines(lines, indent='    '):
 def emit_wrapper(wrapped_fn, suffix, register_info, arguments, byte_size):
     return_write, output_writes, input_reads = register_info
     lines = [
-        f"static void SvcWrap_{wrapped_fn}{suffix}(Core::System& system) {{"
+        f"static void SvcWrap_{wrapped_fn}{suffix}(Core::System& system, std::span<uint64_t, 8> args) {{"
     ]
 
     # Get everything ready.
@@ -498,6 +498,8 @@ namespace Core {
 class System;
 }
 
+#include <span>
+
 #include "common/common_types.h"
 #include "core/hle/kernel/svc_types.h"
 #include "core/hle/result.h"
@@ -524,15 +526,15 @@ void CallSecureMonitor64From32(Core::System& system, ilp32::SecureMonitorArgumen
 void CallSecureMonitor64(Core::System& system, lp64::SecureMonitorArguments* args);
 
 // Defined in svc_light_ipc.cpp.
-void SvcWrap_ReplyAndReceiveLight64From32(Core::System& system);
-void SvcWrap_ReplyAndReceiveLight64(Core::System& system);
+void SvcWrap_ReplyAndReceiveLight64From32(Core::System& system, std::span<uint64_t, 8> args);
+void SvcWrap_ReplyAndReceiveLight64(Core::System& system, std::span<uint64_t, 8> args);
 
-void SvcWrap_SendSyncRequestLight64From32(Core::System& system);
-void SvcWrap_SendSyncRequestLight64(Core::System& system);
+void SvcWrap_SendSyncRequestLight64From32(Core::System& system, std::span<uint64_t, 8> args);
+void SvcWrap_SendSyncRequestLight64(Core::System& system, std::span<uint64_t, 8> args);
 
 // Defined in svc_secure_monitor_call.cpp.
-void SvcWrap_CallSecureMonitor64From32(Core::System& system);
-void SvcWrap_CallSecureMonitor64(Core::System& system);
+void SvcWrap_CallSecureMonitor64From32(Core::System& system, std::span<uint64_t, 8> args);
+void SvcWrap_CallSecureMonitor64(Core::System& system, std::span<uint64_t, 8> args);
 
 // Perform a supervisor call by index.
 void Call(Core::System& system, u32 imm);
@@ -550,20 +552,20 @@ PROLOGUE_CPP = """
 
 namespace Kernel::Svc {
 
-static uint32_t GetReg32(Core::System& system, int n) {
-    return static_cast<uint32_t>(system.CurrentArmInterface().GetReg(n));
+static uint32_t GetArg32(std::span<uint64_t, 8> args, int n) {
+    return static_cast<uint32_t>(args[n]);
 }
 
-static void SetReg32(Core::System& system, int n, uint32_t result) {
-    system.CurrentArmInterface().SetReg(n, static_cast<uint64_t>(result));
+static void SetArg32(std::span<uint64_t, 8> args, int n, uint32_t result) {
+    args[n] = result;
 }
 
-static uint64_t GetReg64(Core::System& system, int n) {
-    return system.CurrentArmInterface().GetReg(n);
+static uint64_t GetArg64(std::span<uint64_t, 8> args, int n) {
+    return args[n];
 }
 
-static void SetReg64(Core::System& system, int n, uint64_t result) {
-    system.CurrentArmInterface().SetReg(n, result);
+static void SetArg64(std::span<uint64_t, 8> args, int n, uint64_t result) {
+    args[n] = result;
 }
 
 // Like bit_cast, but handles the case when the source and dest
@@ -590,15 +592,20 @@ EPILOGUE_CPP = """
 
 void Call(Core::System& system, u32 imm) {
     auto& kernel = system.Kernel();
+    auto& process = GetCurrentProcess(kernel);
+
+    std::array<uint64_t, 8> args;
+    kernel.CurrentPhysicalCore().SaveSvcArguments(process, args);
     kernel.EnterSVCProfile();
 
-    if (GetCurrentProcess(system.Kernel()).Is64Bit()) {
-        Call64(system, imm);
+    if (process.Is64Bit()) {
+        Call64(system, imm, args);
     } else {
-        Call32(system, imm);
+        Call32(system, imm, args);
     }
 
     kernel.ExitSVCProfile();
+    kernel.CurrentPhysicalCore().LoadSvcArguments(process, args);
 }
 
 } // namespace Kernel::Svc
@@ -609,13 +616,13 @@ def emit_call(bitness, names, suffix):
     bit_size = REG_SIZES[bitness]*8
     indent = "    "
     lines = [
-        f"static void Call{bit_size}(Core::System& system, u32 imm) {{",
+        f"static void Call{bit_size}(Core::System& system, u32 imm, std::span<uint64_t, 8> args) {{",
         f"{indent}switch (static_cast<SvcId>(imm)) {{"
     ]
 
     for _, name in names:
         lines.append(f"{indent}case SvcId::{name}:")
-        lines.append(f"{indent*2}return SvcWrap_{name}{suffix}(system);")
+        lines.append(f"{indent*2}return SvcWrap_{name}{suffix}(system, args);")
 
     lines.append(f"{indent}default:")
     lines.append(

src/core/hle/service/filesystem/fsp_srv.cpp

@@ -246,7 +246,13 @@ static void BuildEntryIndex(std::vector<FileSys::Entry>& entries, const std::vec
     entries.reserve(entries.size() + new_data.size());
 
     for (const auto& new_entry : new_data) {
-        entries.emplace_back(new_entry->GetName(), type,
+        auto name = new_entry->GetName();
+
+        if (type == FileSys::EntryType::File && name == FileSys::GetSaveDataSizeFileName()) {
+            continue;
+        }
+
+        entries.emplace_back(name, type,
                              type == FileSys::EntryType::Directory ? 0 : new_entry->GetSize());
     }
 }

src/core/hle/service/hid/controllers/applet_resource.cpp

@@ -0,0 +1,313 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#include "core/core.h"
+#include "core/hle/kernel/k_shared_memory.h"
+#include "core/hle/service/hid/controllers/applet_resource.h"
+#include "core/hle/service/hid/controllers/shared_memory_format.h"
+#include "core/hle/service/hid/errors.h"
+
+namespace Service::HID {
+
+AppletResource::AppletResource(Core::System& system_) : system{system_} {}
+
+AppletResource::~AppletResource() = default;
+
+Result AppletResource::CreateAppletResource(u64 aruid) {
+    const u64 index = GetIndexFromAruid(aruid);
+
+    if (index >= AruidIndexMax) {
+        return ResultAruidNotRegistered;
+    }
+
+    if (data[index].flag.is_assigned) {
+        return ResultAruidAlreadyRegistered;
+    }
+
+    auto& shared_memory = shared_memory_holder[index];
+    if (!shared_memory.IsMapped()) {
+        const Result result = shared_memory.Initialize(system);
+        if (result.IsError()) {
+            return result;
+        }
+        if (shared_memory.GetAddress() == nullptr) {
+            shared_memory.Finalize();
+            return ResultSharedMemoryNotInitialized;
+        }
+    }
+
+    auto* shared_memory_format = shared_memory.GetAddress();
+    if (shared_memory_format != nullptr) {
+        shared_memory_format->Initialize();
+    }
+
+    data[index].shared_memory_format = shared_memory_format;
+    data[index].flag.is_assigned.Assign(true);
+    // TODO: InitializeSixAxisControllerConfig(false);
+    active_aruid = aruid;
+    return ResultSuccess;
+}
+
+Result AppletResource::RegisterAppletResourceUserId(u64 aruid, bool enable_input) {
+    const u64 index = GetIndexFromAruid(aruid);
+
+    if (index < AruidIndexMax) {
+        return ResultAruidAlreadyRegistered;
+    }
+
+    std::size_t data_index = AruidIndexMax;
+    for (std::size_t i = 0; i < AruidIndexMax; i++) {
+        if (!data[i].flag.is_initialized) {
+            data_index = i;
+            break;
+        }
+    }
+
+    if (data_index == AruidIndexMax) {
+        return ResultAruidNoAvailableEntries;
+    }
+
+    AruidData& aruid_data = data[data_index];
+
+    aruid_data.aruid = aruid;
+    aruid_data.flag.is_initialized.Assign(true);
+    if (enable_input) {
+        aruid_data.flag.enable_pad_input.Assign(true);
+        aruid_data.flag.enable_six_axis_sensor.Assign(true);
+        aruid_data.flag.bit_18.Assign(true);
+        aruid_data.flag.enable_touchscreen.Assign(true);
+    }
+
+    data_index = AruidIndexMax;
+    for (std::size_t i = 0; i < AruidIndexMax; i++) {
+        if (registration_list.flag[i] == RegistrationStatus::Initialized) {
+            if (registration_list.aruid[i] != aruid) {
+                continue;
+            }
+            data_index = i;
+            break;
+        }
+        if (registration_list.flag[i] == RegistrationStatus::None) {
+            data_index = i;
+            break;
+        }
+    }
+
+    if (data_index == AruidIndexMax) {
+        return ResultSuccess;
+    }
+
+    registration_list.flag[data_index] = RegistrationStatus::Initialized;
+    registration_list.aruid[data_index] = aruid;
+
+    return ResultSuccess;
+}
+
+void AppletResource::UnregisterAppletResourceUserId(u64 aruid) {
+    u64 index = GetIndexFromAruid(aruid);
+
+    if (index < AruidIndexMax) {
+        if (data[index].flag.is_assigned) {
+            data[index].shared_memory_format = nullptr;
+            data[index].flag.is_assigned.Assign(false);
+        }
+    }
+
+    index = GetIndexFromAruid(aruid);
+    if (index < AruidIndexMax) {
+        DestroySevenSixAxisTransferMemory();
+        data[index].flag.raw = 0;
+        data[index].aruid = 0;
+
+        index = GetIndexFromAruid(aruid);
+        if (index < AruidIndexMax) {
+            registration_list.flag[index] = RegistrationStatus::PendingDelete;
+        }
+    }
+}
+
+void AppletResource::FreeAppletResourceId(u64 aruid) {
+    u64 index = GetIndexFromAruid(aruid);
+    if (index >= AruidIndexMax) {
+        return;
+    }
+
+    auto& aruid_data = data[index];
+    if (aruid_data.flag.is_assigned) {
+        aruid_data.shared_memory_format = nullptr;
+        aruid_data.flag.is_assigned.Assign(false);
+    }
+}
+
+u64 AppletResource::GetActiveAruid() {
+    return active_aruid;
+}
+
+Result AppletResource::GetSharedMemoryHandle(Kernel::KSharedMemory** out_handle, u64 aruid) {
+    u64 index = GetIndexFromAruid(aruid);
+    if (index >= AruidIndexMax) {
+        return ResultAruidNotRegistered;
+    }
+
+    *out_handle = shared_memory_holder[index].GetHandle();
+    return ResultSuccess;
+}
+
+Result AppletResource::GetSharedMemoryFormat(SharedMemoryFormat** out_shared_memory_format,
+                                             u64 aruid) {
+    u64 index = GetIndexFromAruid(aruid);
+    if (index >= AruidIndexMax) {
+        return ResultAruidNotRegistered;
+    }
+
+    *out_shared_memory_format = data[index].shared_memory_format;
+    return ResultSuccess;
+}
+
+u64 AppletResource::GetIndexFromAruid(u64 aruid) {
+    for (std::size_t i = 0; i < AruidIndexMax; i++) {
+        if (registration_list.flag[i] == RegistrationStatus::Initialized &&
+            registration_list.aruid[i] == aruid) {
+            return i;
+        }
+    }
+    return AruidIndexMax;
+}
+
+Result AppletResource::DestroySevenSixAxisTransferMemory() {
+    // TODO
+    return ResultSuccess;
+}
+
+void AppletResource::EnableInput(u64 aruid, bool is_enabled) {
+    const u64 index = GetIndexFromAruid(aruid);
+    if (index >= AruidIndexMax) {
+        return;
+    }
+
+    data[index].flag.enable_pad_input.Assign(is_enabled);
+    data[index].flag.enable_touchscreen.Assign(is_enabled);
+}
+
+void AppletResource::EnableSixAxisSensor(u64 aruid, bool is_enabled) {
+    const u64 index = GetIndexFromAruid(aruid);
+    if (index >= AruidIndexMax) {
+        return;
+    }
+
+    data[index].flag.enable_six_axis_sensor.Assign(is_enabled);
+}
+
+void AppletResource::EnablePadInput(u64 aruid, bool is_enabled) {
+    const u64 index = GetIndexFromAruid(aruid);
+    if (index >= AruidIndexMax) {
+        return;
+    }
+
+    data[index].flag.enable_pad_input.Assign(is_enabled);
+}
+
+void AppletResource::EnableTouchScreen(u64 aruid, bool is_enabled) {
+    const u64 index = GetIndexFromAruid(aruid);
+    if (index >= AruidIndexMax) {
+        return;
+    }
+
+    data[index].flag.enable_touchscreen.Assign(is_enabled);
+}
+
+void AppletResource::SetIsPalmaConnectable(u64 aruid, bool is_connectable) {
+    const u64 index = GetIndexFromAruid(aruid);
+    if (index >= AruidIndexMax) {
+        return;
+    }
+
+    data[index].flag.is_palma_connectable.Assign(is_connectable);
+}
+
+void AppletResource::EnablePalmaBoostMode(u64 aruid, bool is_enabled) {
+    const u64 index = GetIndexFromAruid(aruid);
+    if (index >= AruidIndexMax) {
+        return;
+    }
+
+    data[index].flag.enable_palma_boost_mode.Assign(is_enabled);
+}
+
+Result AppletResource::RegisterCoreAppletResource() {
+    if (ref_counter == std::numeric_limits<s32>::max() - 1) {
+        return ResultAppletResourceOverflow;
+    }
+    if (ref_counter == 0) {
+        const u64 index = GetIndexFromAruid(0);
+        if (index < AruidIndexMax) {
+            return ResultAruidAlreadyRegistered;
+        }
+
+        std::size_t data_index = AruidIndexMax;
+        for (std::size_t i = 0; i < AruidIndexMax; i++) {
+            if (!data[i].flag.is_initialized) {
+                data_index = i;
+                break;
+            }
+        }
+
+        if (data_index == AruidIndexMax) {
+            return ResultAruidNoAvailableEntries;
+        }
+
+        AruidData& aruid_data = data[data_index];
+
+        aruid_data.aruid = 0;
+        aruid_data.flag.is_initialized.Assign(true);
+        aruid_data.flag.enable_pad_input.Assign(true);
+        aruid_data.flag.enable_six_axis_sensor.Assign(true);
+        aruid_data.flag.bit_18.Assign(true);
+        aruid_data.flag.enable_touchscreen.Assign(true);
+
+        data_index = AruidIndexMax;
+        for (std::size_t i = 0; i < AruidIndexMax; i++) {
+            if (registration_list.flag[i] == RegistrationStatus::Initialized) {
+                if (registration_list.aruid[i] != 0) {
+                    continue;
+                }
+                data_index = i;
+                break;
+            }
+            if (registration_list.flag[i] == RegistrationStatus::None) {
+                data_index = i;
+                break;
+            }
+        }
+
+        Result result = ResultSuccess;
+
+        if (data_index == AruidIndexMax) {
+            result = CreateAppletResource(0);
+        } else {
+            registration_list.flag[data_index] = RegistrationStatus::Initialized;
+            registration_list.aruid[data_index] = 0;
+        }
+
+        if (result.IsError()) {
+            UnregisterAppletResourceUserId(0);
+            return result;
+        }
+    }
+    ref_counter++;
+    return ResultSuccess;
+}
+
+Result AppletResource::UnregisterCoreAppletResource() {
+    if (ref_counter == 0) {
+        return ResultAppletResourceNotInitialized;
+    }
+
+    if (--ref_counter == 0) {
+        UnregisterAppletResourceUserId(0);
+    }
+
+    return ResultSuccess;
+}
+
+} // namespace Service::HID

src/core/hle/service/hid/controllers/applet_resource.h

@@ -0,0 +1,98 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+#include <array>
+
+#include "common/bit_field.h"
+#include "common/common_types.h"
+#include "core/hle/result.h"
+#include "core/hle/service/hid/controllers/shared_memory_holder.h"
+
+namespace Core {
+class System;
+}
+
+namespace Kernel {
+class KSharedMemory;
+}
+
+namespace Service::HID {
+struct SharedMemoryFormat;
+
+class AppletResource {
+public:
+    explicit AppletResource(Core::System& system_);
+    ~AppletResource();
+
+    Result CreateAppletResource(u64 aruid);
+
+    Result RegisterAppletResourceUserId(u64 aruid, bool enable_input);
+    void UnregisterAppletResourceUserId(u64 aruid);
+
+    void FreeAppletResourceId(u64 aruid);
+
+    u64 GetActiveAruid();
+    Result GetSharedMemoryHandle(Kernel::KSharedMemory** out_handle, u64 aruid);
+    Result GetSharedMemoryFormat(SharedMemoryFormat** out_shared_memory_format, u64 aruid);
+
+    u64 GetIndexFromAruid(u64 aruid);
+
+    Result DestroySevenSixAxisTransferMemory();
+
+    void EnableInput(u64 aruid, bool is_enabled);
+    void EnableSixAxisSensor(u64 aruid, bool is_enabled);
+    void EnablePadInput(u64 aruid, bool is_enabled);
+    void EnableTouchScreen(u64 aruid, bool is_enabled);
+    void SetIsPalmaConnectable(u64 aruid, bool is_connectable);
+    void EnablePalmaBoostMode(u64 aruid, bool is_enabled);
+
+    Result RegisterCoreAppletResource();
+    Result UnregisterCoreAppletResource();
+
+private:
+    static constexpr std::size_t AruidIndexMax = 0x20;
+
+    enum RegistrationStatus : u32 {
+        None,
+        Initialized,
+        PendingDelete,
+    };
+
+    struct DataStatusFlag {
+        union {
+            u32 raw{};
+
+            BitField<0, 1, u32> is_initialized;
+            BitField<1, 1, u32> is_assigned;
+            BitField<16, 1, u32> enable_pad_input;
+            BitField<17, 1, u32> enable_six_axis_sensor;
+            BitField<18, 1, u32> bit_18;
+            BitField<19, 1, u32> is_palma_connectable;
+            BitField<20, 1, u32> enable_palma_boost_mode;
+            BitField<21, 1, u32> enable_touchscreen;
+        };
+    };
+
+    struct AruidRegisterList {
+        std::array<RegistrationStatus, AruidIndexMax> flag{};
+        std::array<u64, AruidIndexMax> aruid{};
+    };
+    static_assert(sizeof(AruidRegisterList) == 0x180, "AruidRegisterList is an invalid size");
+
+    struct AruidData {
+        DataStatusFlag flag{};
+        u64 aruid{};
+        SharedMemoryFormat* shared_memory_format{nullptr};
+    };
+
+    u64 active_aruid{};
+    AruidRegisterList registration_list{};
+    std::array<AruidData, AruidIndexMax> data{};
+    std::array<SharedMemoryHolder, AruidIndexMax> shared_memory_holder{};
+    s32 ref_counter{};
+
+    Core::System& system;
+};
+} // namespace Service::HID

src/core/hle/service/hid/controllers/console_six_axis.cpp

@@ -1,23 +1,18 @@
 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include "core/core.h"
 #include "core/core_timing.h"
 #include "core/hid/emulated_console.h"
 #include "core/hid/hid_core.h"
 #include "core/hle/service/hid/controllers/console_six_axis.h"
-#include "core/memory.h"
+#include "core/hle/service/hid/controllers/shared_memory_format.h"
 
 namespace Service::HID {
-constexpr std::size_t SHARED_MEMORY_OFFSET = 0x3C200;
 
-ConsoleSixAxis::ConsoleSixAxis(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_)
-    : ControllerBase{hid_core_} {
+ConsoleSixAxis::ConsoleSixAxis(Core::HID::HIDCore& hid_core_,
+                               ConsoleSixAxisSensorSharedMemoryFormat& console_shared_memory)
+    : ControllerBase{hid_core_}, shared_memory{console_shared_memory} {
     console = hid_core.GetEmulatedConsole();
-    static_assert(SHARED_MEMORY_OFFSET + sizeof(ConsoleSharedMemory) < shared_memory_size,
-                  "ConsoleSharedMemory is bigger than the shared memory");
-    shared_memory = std::construct_at(
-        reinterpret_cast<ConsoleSharedMemory*>(raw_shared_memory_ + SHARED_MEMORY_OFFSET));
 }
 
 ConsoleSixAxis::~ConsoleSixAxis() = default;
@@ -33,10 +28,10 @@ void ConsoleSixAxis::OnUpdate(const Core::Timing::CoreTiming& core_timing) {
 
     const auto motion_status = console->GetMotion();
 
-    shared_memory->sampling_number++;
-    shared_memory->is_seven_six_axis_sensor_at_rest = motion_status.is_at_rest;
-    shared_memory->verticalization_error = motion_status.verticalization_error;
-    shared_memory->gyro_bias = motion_status.gyro_bias;
+    shared_memory.sampling_number++;
+    shared_memory.is_seven_six_axis_sensor_at_rest = motion_status.is_at_rest;
+    shared_memory.verticalization_error = motion_status.verticalization_error;
+    shared_memory.gyro_bias = motion_status.gyro_bias;
 }
 
 } // namespace Service::HID

src/core/hle/service/hid/controllers/console_six_axis.h

@@ -3,7 +3,6 @@
 
 #pragma once
 
-#include "common/vector_math.h"
 #include "core/hle/service/hid/controllers/controller_base.h"
 
 namespace Core::HID {
@@ -11,9 +10,12 @@ class EmulatedConsole;
 } // namespace Core::HID
 
 namespace Service::HID {
+struct ConsoleSixAxisSensorSharedMemoryFormat;
+
 class ConsoleSixAxis final : public ControllerBase {
 public:
-    explicit ConsoleSixAxis(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_);
+    explicit ConsoleSixAxis(Core::HID::HIDCore& hid_core_,
+                            ConsoleSixAxisSensorSharedMemoryFormat& console_shared_memory);
     ~ConsoleSixAxis() override;
 
     // Called when the controller is initialized
@@ -26,18 +28,7 @@ public:
     void OnUpdate(const Core::Timing::CoreTiming& core_timing) override;
 
 private:
-    // This is nn::hid::detail::ConsoleSixAxisSensorSharedMemoryFormat
-    struct ConsoleSharedMemory {
-        u64 sampling_number{};
-        bool is_seven_six_axis_sensor_at_rest{};
-        INSERT_PADDING_BYTES(3); // padding
-        f32 verticalization_error{};
-        Common::Vec3f gyro_bias{};
-        INSERT_PADDING_BYTES(4); // padding
-    };
-    static_assert(sizeof(ConsoleSharedMemory) == 0x20, "ConsoleSharedMemory is an invalid size");
-
-    ConsoleSharedMemory* shared_memory = nullptr;
+    ConsoleSixAxisSensorSharedMemoryFormat& shared_memory;
     Core::HID::EmulatedConsole* console = nullptr;
 };
 } // namespace Service::HID

src/core/hle/service/hid/controllers/controller_base.h

@@ -39,9 +39,6 @@ public:
 
     bool IsControllerActivated() const;
 
-    static const std::size_t hid_entry_count = 17;
-    static const std::size_t shared_memory_size = 0x40000;
-
 protected:
     bool is_activated{false};
 

src/core/hle/service/hid/controllers/debug_pad.cpp

@@ -1,24 +1,19 @@
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include <cstring>
-#include "common/common_types.h"
 #include "common/settings.h"
 #include "core/core_timing.h"
 #include "core/hid/emulated_controller.h"
 #include "core/hid/hid_core.h"
 #include "core/hid/hid_types.h"
 #include "core/hle/service/hid/controllers/debug_pad.h"
+#include "core/hle/service/hid/controllers/shared_memory_format.h"
 
 namespace Service::HID {
-constexpr std::size_t SHARED_MEMORY_OFFSET = 0x00000;
 
-DebugPad::DebugPad(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_)
-    : ControllerBase{hid_core_} {
-    static_assert(SHARED_MEMORY_OFFSET + sizeof(DebugPadSharedMemory) < shared_memory_size,
-                  "DebugPadSharedMemory is bigger than the shared memory");
-    shared_memory = std::construct_at(
-        reinterpret_cast<DebugPadSharedMemory*>(raw_shared_memory_ + SHARED_MEMORY_OFFSET));
+DebugPad::DebugPad(Core::HID::HIDCore& hid_core_,
+                   DebugPadSharedMemoryFormat& debug_pad_shared_memory)
+    : ControllerBase{hid_core_}, shared_memory{debug_pad_shared_memory} {
     controller = hid_core.GetEmulatedController(Core::HID::NpadIdType::Other);
 }
 
@@ -30,12 +25,12 @@ void DebugPad::OnRelease() {}
 
 void DebugPad::OnUpdate(const Core::Timing::CoreTiming& core_timing) {
     if (!IsControllerActivated()) {
-        shared_memory->debug_pad_lifo.buffer_count = 0;
-        shared_memory->debug_pad_lifo.buffer_tail = 0;
+        shared_memory.debug_pad_lifo.buffer_count = 0;
+        shared_memory.debug_pad_lifo.buffer_tail = 0;
         return;
     }
 
-    const auto& last_entry = shared_memory->debug_pad_lifo.ReadCurrentEntry().state;
+    const auto& last_entry = shared_memory.debug_pad_lifo.ReadCurrentEntry().state;
     next_state.sampling_number = last_entry.sampling_number + 1;
 
     if (Settings::values.debug_pad_enabled) {
@@ -49,7 +44,7 @@ void DebugPad::OnUpdate(const Core::Timing::CoreTiming& core_timing) {
         next_state.r_stick = stick_state.right;
     }
 
-    shared_memory->debug_pad_lifo.WriteNextEntry(next_state);
+    shared_memory.debug_pad_lifo.WriteNextEntry(next_state);
 }
 
 } // namespace Service::HID

src/core/hle/service/hid/controllers/debug_pad.h

@@ -3,21 +3,24 @@
 
 #pragma once
 
-#include "common/bit_field.h"
-#include "common/common_types.h"
 #include "core/hle/service/hid/controllers/controller_base.h"
-#include "core/hle/service/hid/ring_lifo.h"
+#include "core/hle/service/hid/controllers/types/debug_pad_types.h"
 
 namespace Core::HID {
-class EmulatedController;
-struct DebugPadButton;
-struct AnalogStickState;
-} // namespace Core::HID
+class HIDCore;
+}
+
+namespace Core::Timing {
+class CoreTiming;
+}
 
 namespace Service::HID {
+struct DebugPadSharedMemoryFormat;
+
 class DebugPad final : public ControllerBase {
 public:
-    explicit DebugPad(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_);
+    explicit DebugPad(Core::HID::HIDCore& hid_core_,
+                      DebugPadSharedMemoryFormat& debug_pad_shared_memory);
     ~DebugPad() override;
 
     // Called when the controller is initialized
@@ -30,35 +33,8 @@ public:
     void OnUpdate(const Core::Timing::CoreTiming& core_timing) override;
 
 private:
-    // This is nn::hid::DebugPadAttribute
-    struct DebugPadAttribute {
-        union {
-            u32 raw{};
-            BitField<0, 1, u32> connected;
-        };
-    };
-    static_assert(sizeof(DebugPadAttribute) == 0x4, "DebugPadAttribute is an invalid size");
-
-    // This is nn::hid::DebugPadState
-    struct DebugPadState {
-        s64 sampling_number{};
-        DebugPadAttribute attribute{};
-        Core::HID::DebugPadButton pad_state{};
-        Core::HID::AnalogStickState r_stick{};
-        Core::HID::AnalogStickState l_stick{};
-    };
-    static_assert(sizeof(DebugPadState) == 0x20, "DebugPadState is an invalid state");
-
-    struct DebugPadSharedMemory {
-        // This is nn::hid::detail::DebugPadLifo
-        Lifo<DebugPadState, hid_entry_count> debug_pad_lifo{};
-        static_assert(sizeof(debug_pad_lifo) == 0x2C8, "debug_pad_lifo is an invalid size");
-        INSERT_PADDING_WORDS(0x4E);
-    };
-    static_assert(sizeof(DebugPadSharedMemory) == 0x400, "DebugPadSharedMemory is an invalid size");
-
     DebugPadState next_state{};
-    DebugPadSharedMemory* shared_memory = nullptr;
+    DebugPadSharedMemoryFormat& shared_memory;
     Core::HID::EmulatedController* controller = nullptr;
 };
 } // namespace Service::HID

src/core/hle/service/hid/controllers/gesture.cpp

@@ -1,17 +1,15 @@
 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include "common/logging/log.h"
 #include "common/math_util.h"
 #include "common/settings.h"
-#include "core/core_timing.h"
 #include "core/frontend/emu_window.h"
+#include "core/hid/emulated_console.h"
 #include "core/hid/hid_core.h"
 #include "core/hle/service/hid/controllers/gesture.h"
+#include "core/hle/service/hid/controllers/shared_memory_format.h"
 
 namespace Service::HID {
-constexpr std::size_t SHARED_MEMORY_OFFSET = 0x3BA00;
-
 // HW is around 700, value is set to 400 to make it easier to trigger with mouse
 constexpr f32 swipe_threshold = 400.0f; // Threshold in pixels/s
 constexpr f32 angle_threshold = 0.015f; // Threshold in radians
@@ -23,19 +21,15 @@ constexpr f32 Square(s32 num) {
     return static_cast<f32>(num * num);
 }
 
-Gesture::Gesture(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_)
-    : ControllerBase(hid_core_) {
-    static_assert(SHARED_MEMORY_OFFSET + sizeof(GestureSharedMemory) < shared_memory_size,
-                  "GestureSharedMemory is bigger than the shared memory");
-    shared_memory = std::construct_at(
-        reinterpret_cast<GestureSharedMemory*>(raw_shared_memory_ + SHARED_MEMORY_OFFSET));
+Gesture::Gesture(Core::HID::HIDCore& hid_core_, GestureSharedMemoryFormat& gesture_shared_memory)
+    : ControllerBase(hid_core_), shared_memory{gesture_shared_memory} {
     console = hid_core.GetEmulatedConsole();
 }
 Gesture::~Gesture() = default;
 
 void Gesture::OnInit() {
-    shared_memory->gesture_lifo.buffer_count = 0;
-    shared_memory->gesture_lifo.buffer_tail = 0;
+    shared_memory.gesture_lifo.buffer_count = 0;
+    shared_memory.gesture_lifo.buffer_tail = 0;
     force_update = true;
 }
 
@@ -43,8 +37,8 @@ void Gesture::OnRelease() {}
 
 void Gesture::OnUpdate(const Core::Timing::CoreTiming& core_timing) {
     if (!IsControllerActivated()) {
-        shared_memory->gesture_lifo.buffer_count = 0;
-        shared_memory->gesture_lifo.buffer_tail = 0;
+        shared_memory.gesture_lifo.buffer_count = 0;
+        shared_memory.gesture_lifo.buffer_tail = 0;
         return;
     }
 
@@ -52,7 +46,7 @@ void Gesture::OnUpdate(const Core::Timing::CoreTiming& core_timing) {
 
     GestureProperties gesture = GetGestureProperties();
     f32 time_difference =
-        static_cast<f32>(shared_memory->gesture_lifo.timestamp - last_update_timestamp) /
+        static_cast<f32>(shared_memory.gesture_lifo.timestamp - last_update_timestamp) /
         (1000 * 1000 * 1000);
 
     // Only update if necessary
@@ -60,7 +54,7 @@ void Gesture::OnUpdate(const Core::Timing::CoreTiming& core_timing) {
         return;
     }
 
-    last_update_timestamp = shared_memory->gesture_lifo.timestamp;
+    last_update_timestamp = shared_memory.gesture_lifo.timestamp;
     UpdateGestureSharedMemory(gesture, time_difference);
 }
 
@@ -103,7 +97,7 @@ void Gesture::UpdateGestureSharedMemory(GestureProperties& gesture, f32 time_dif
     GestureType type = GestureType::Idle;
     GestureAttribute attributes{};
 
-    const auto& last_entry = shared_memory->gesture_lifo.ReadCurrentEntry().state;
+    const auto& last_entry = shared_memory.gesture_lifo.ReadCurrentEntry().state;
 
     // Reset next state to default
     next_state.sampling_number = last_entry.sampling_number + 1;
@@ -133,7 +127,7 @@ void Gesture::UpdateGestureSharedMemory(GestureProperties& gesture, f32 time_dif
     next_state.points = gesture.points;
     last_gesture = gesture;
 
-    shared_memory->gesture_lifo.WriteNextEntry(next_state);
+    shared_memory.gesture_lifo.WriteNextEntry(next_state);
 }
 
 void Gesture::NewGesture(GestureProperties& gesture, GestureType& type,
@@ -305,11 +299,11 @@ void Gesture::SetSwipeEvent(GestureProperties& gesture, GestureProperties& last_
     next_state.direction = GestureDirection::Up;
 }
 
-const Gesture::GestureState& Gesture::GetLastGestureEntry() const {
-    return shared_memory->gesture_lifo.ReadCurrentEntry().state;
+const GestureState& Gesture::GetLastGestureEntry() const {
+    return shared_memory.gesture_lifo.ReadCurrentEntry().state;
 }
 
-Gesture::GestureProperties Gesture::GetGestureProperties() {
+GestureProperties Gesture::GetGestureProperties() {
     GestureProperties gesture;
     std::array<Core::HID::TouchFinger, MAX_POINTS> active_fingers;
     const auto end_iter = std::copy_if(fingers.begin(), fingers.end(), active_fingers.begin(),

src/core/hle/service/hid/controllers/gesture.h

@@ -4,17 +4,22 @@
 #pragma once
 
 #include <array>
-#include "common/bit_field.h"
+
 #include "common/common_types.h"
-#include "common/point.h"
-#include "core/hid/emulated_console.h"
 #include "core/hle/service/hid/controllers/controller_base.h"
-#include "core/hle/service/hid/ring_lifo.h"
+#include "core/hle/service/hid/controllers/types/touch_types.h"
+
+namespace Core::HID {
+class EmulatedConsole;
+}
 
 namespace Service::HID {
+struct GestureSharedMemoryFormat;
+
 class Gesture final : public ControllerBase {
 public:
-    explicit Gesture(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_);
+    explicit Gesture(Core::HID::HIDCore& hid_core_,
+                     GestureSharedMemoryFormat& gesture_shared_memory);
     ~Gesture() override;
 
     // Called when the controller is initialized
@@ -27,79 +32,6 @@ public:
     void OnUpdate(const Core::Timing::CoreTiming& core_timing) override;
 
 private:
-    static constexpr size_t MAX_FINGERS = 16;
-    static constexpr size_t MAX_POINTS = 4;
-
-    // This is nn::hid::GestureType
-    enum class GestureType : u32 {
-        Idle,     // Nothing touching the screen
-        Complete, // Set at the end of a touch event
-        Cancel,   // Set when the number of fingers change
-        Touch,    // A finger just touched the screen
-        Press,    // Set if last type is touch and the finger hasn't moved
-        Tap,      // Fast press then release
-        Pan,      // All points moving together across the screen
-        Swipe,    // Fast press movement and release of a single point
-        Pinch,    // All points moving away/closer to the midpoint
-        Rotate,   // All points rotating from the midpoint
-    };
-
-    // This is nn::hid::GestureDirection
-    enum class GestureDirection : u32 {
-        None,
-        Left,
-        Up,
-        Right,
-        Down,
-    };
-
-    // This is nn::hid::GestureAttribute
-    struct GestureAttribute {
-        union {
-            u32 raw{};
-
-            BitField<4, 1, u32> is_new_touch;
-            BitField<8, 1, u32> is_double_tap;
-        };
-    };
-    static_assert(sizeof(GestureAttribute) == 4, "GestureAttribute is an invalid size");
-
-    // This is nn::hid::GestureState
-    struct GestureState {
-        s64 sampling_number{};
-        s64 detection_count{};
-        GestureType type{GestureType::Idle};
-        GestureDirection direction{GestureDirection::None};
-        Common::Point<s32> pos{};
-        Common::Point<s32> delta{};
-        f32 vel_x{};
-        f32 vel_y{};
-        GestureAttribute attributes{};
-        f32 scale{};
-        f32 rotation_angle{};
-        s32 point_count{};
-        std::array<Common::Point<s32>, 4> points{};
-    };
-    static_assert(sizeof(GestureState) == 0x60, "GestureState is an invalid size");
-
-    struct GestureProperties {
-        std::array<Common::Point<s32>, MAX_POINTS> points{};
-        std::size_t active_points{};
-        Common::Point<s32> mid_point{};
-        s64 detection_count{};
-        u64 delta_time{};
-        f32 average_distance{};
-        f32 angle{};
-    };
-
-    struct GestureSharedMemory {
-        // This is nn::hid::detail::GestureLifo
-        Lifo<GestureState, hid_entry_count> gesture_lifo{};
-        static_assert(sizeof(gesture_lifo) == 0x708, "gesture_lifo is an invalid size");
-        INSERT_PADDING_WORDS(0x3E);
-    };
-    static_assert(sizeof(GestureSharedMemory) == 0x800, "GestureSharedMemory is an invalid size");
-
     // Reads input from all available input engines
     void ReadTouchInput();
 
@@ -142,7 +74,7 @@ private:
     GestureProperties GetGestureProperties();
 
     GestureState next_state{};
-    GestureSharedMemory* shared_memory = nullptr;
+    GestureSharedMemoryFormat& shared_memory;
     Core::HID::EmulatedConsole* console = nullptr;
 
     std::array<Core::HID::TouchFinger, MAX_POINTS> fingers{};

src/core/hle/service/hid/controllers/keyboard.cpp

@@ -1,23 +1,18 @@
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include <cstring>
-#include "common/common_types.h"
 #include "common/settings.h"
 #include "core/core_timing.h"
 #include "core/hid/emulated_devices.h"
 #include "core/hid/hid_core.h"
 #include "core/hle/service/hid/controllers/keyboard.h"
+#include "core/hle/service/hid/controllers/shared_memory_format.h"
 
 namespace Service::HID {
-constexpr std::size_t SHARED_MEMORY_OFFSET = 0x3800;
 
-Keyboard::Keyboard(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_)
-    : ControllerBase{hid_core_} {
-    static_assert(SHARED_MEMORY_OFFSET + sizeof(KeyboardSharedMemory) < shared_memory_size,
-                  "KeyboardSharedMemory is bigger than the shared memory");
-    shared_memory = std::construct_at(
-        reinterpret_cast<KeyboardSharedMemory*>(raw_shared_memory_ + SHARED_MEMORY_OFFSET));
+Keyboard::Keyboard(Core::HID::HIDCore& hid_core_,
+                   KeyboardSharedMemoryFormat& keyboard_shared_memory)
+    : ControllerBase{hid_core_}, shared_memory{keyboard_shared_memory} {
     emulated_devices = hid_core.GetEmulatedDevices();
 }
 
@@ -29,12 +24,12 @@ void Keyboard::OnRelease() {}
 
 void Keyboard::OnUpdate(const Core::Timing::CoreTiming& core_timing) {
     if (!IsControllerActivated()) {
-        shared_memory->keyboard_lifo.buffer_count = 0;
-        shared_memory->keyboard_lifo.buffer_tail = 0;
+        shared_memory.keyboard_lifo.buffer_count = 0;
+        shared_memory.keyboard_lifo.buffer_tail = 0;
         return;
     }
 
-    const auto& last_entry = shared_memory->keyboard_lifo.ReadCurrentEntry().state;
+    const auto& last_entry = shared_memory.keyboard_lifo.ReadCurrentEntry().state;
     next_state.sampling_number = last_entry.sampling_number + 1;
 
     if (Settings::values.keyboard_enabled) {
@@ -46,7 +41,7 @@ void Keyboard::OnUpdate(const Core::Timing::CoreTiming& core_timing) {
         next_state.attribute.is_connected.Assign(1);
     }
 
-    shared_memory->keyboard_lifo.WriteNextEntry(next_state);
+    shared_memory.keyboard_lifo.WriteNextEntry(next_state);
 }
 
 } // namespace Service::HID

src/core/hle/service/hid/controllers/keyboard.h

@@ -3,20 +3,16 @@
 
 #pragma once
 
-#include "common/common_types.h"
 #include "core/hle/service/hid/controllers/controller_base.h"
-#include "core/hle/service/hid/ring_lifo.h"
-
-namespace Core::HID {
-class EmulatedDevices;
-struct KeyboardModifier;
-struct KeyboardKey;
-} // namespace Core::HID
+#include "core/hle/service/hid/controllers/types/keyboard_types.h"
 
 namespace Service::HID {
+struct KeyboardSharedMemoryFormat;
+
 class Keyboard final : public ControllerBase {
 public:
-    explicit Keyboard(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_);
+    explicit Keyboard(Core::HID::HIDCore& hid_core_,
+                      KeyboardSharedMemoryFormat& keyboard_shared_memory);
     ~Keyboard() override;
 
     // Called when the controller is initialized
@@ -29,25 +25,8 @@ public:
     void OnUpdate(const Core::Timing::CoreTiming& core_timing) override;
 
 private:
-    // This is nn::hid::detail::KeyboardState
-    struct KeyboardState {
-        s64 sampling_number{};
-        Core::HID::KeyboardModifier modifier{};
-        Core::HID::KeyboardAttribute attribute{};
-        Core::HID::KeyboardKey key{};
-    };
-    static_assert(sizeof(KeyboardState) == 0x30, "KeyboardState is an invalid size");
-
-    struct KeyboardSharedMemory {
-        // This is nn::hid::detail::KeyboardLifo
-        Lifo<KeyboardState, hid_entry_count> keyboard_lifo{};
-        static_assert(sizeof(keyboard_lifo) == 0x3D8, "keyboard_lifo is an invalid size");
-        INSERT_PADDING_WORDS(0xA);
-    };
-    static_assert(sizeof(KeyboardSharedMemory) == 0x400, "KeyboardSharedMemory is an invalid size");
-
     KeyboardState next_state{};
-    KeyboardSharedMemory* shared_memory = nullptr;
+    KeyboardSharedMemoryFormat& shared_memory;
     Core::HID::EmulatedDevices* emulated_devices = nullptr;
 };
 } // namespace Service::HID

src/core/hle/service/hid/controllers/mouse.cpp

@@ -1,22 +1,17 @@
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include <cstring>
-#include "common/common_types.h"
 #include "core/core_timing.h"
 #include "core/frontend/emu_window.h"
 #include "core/hid/emulated_devices.h"
 #include "core/hid/hid_core.h"
 #include "core/hle/service/hid/controllers/mouse.h"
+#include "core/hle/service/hid/controllers/shared_memory_format.h"
 
 namespace Service::HID {
-constexpr std::size_t SHARED_MEMORY_OFFSET = 0x3400;
 
-Mouse::Mouse(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_) : ControllerBase{hid_core_} {
-    static_assert(SHARED_MEMORY_OFFSET + sizeof(MouseSharedMemory) < shared_memory_size,
-                  "MouseSharedMemory is bigger than the shared memory");
-    shared_memory = std::construct_at(
-        reinterpret_cast<MouseSharedMemory*>(raw_shared_memory_ + SHARED_MEMORY_OFFSET));
+Mouse::Mouse(Core::HID::HIDCore& hid_core_, MouseSharedMemoryFormat& mouse_shared_memory)
+    : ControllerBase{hid_core_}, shared_memory{mouse_shared_memory} {
     emulated_devices = hid_core.GetEmulatedDevices();
 }
 
@@ -27,14 +22,14 @@ void Mouse::OnRelease() {}
 
 void Mouse::OnUpdate(const Core::Timing::CoreTiming& core_timing) {
     if (!IsControllerActivated()) {
-        shared_memory->mouse_lifo.buffer_count = 0;
-        shared_memory->mouse_lifo.buffer_tail = 0;
+        shared_memory.mouse_lifo.buffer_count = 0;
+        shared_memory.mouse_lifo.buffer_tail = 0;
         return;
     }
 
     next_state = {};
 
-    const auto& last_entry = shared_memory->mouse_lifo.ReadCurrentEntry().state;
+    const auto& last_entry = shared_memory.mouse_lifo.ReadCurrentEntry().state;
     next_state.sampling_number = last_entry.sampling_number + 1;
 
     if (Settings::values.mouse_enabled) {
@@ -53,7 +48,7 @@ void Mouse::OnUpdate(const Core::Timing::CoreTiming& core_timing) {
         next_state.button = mouse_button_state;
     }
 
-    shared_memory->mouse_lifo.WriteNextEntry(next_state);
+    shared_memory.mouse_lifo.WriteNextEntry(next_state);
 }
 
 } // namespace Service::HID

src/core/hle/service/hid/controllers/mouse.h

@@ -3,9 +3,7 @@
 
 #pragma once
 
-#include "common/common_types.h"
 #include "core/hle/service/hid/controllers/controller_base.h"
-#include "core/hle/service/hid/ring_lifo.h"
 
 namespace Core::HID {
 class EmulatedDevices;
@@ -14,9 +12,11 @@ struct AnalogStickState;
 } // namespace Core::HID
 
 namespace Service::HID {
+struct MouseSharedMemoryFormat;
+
 class Mouse final : public ControllerBase {
 public:
-    explicit Mouse(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_);
+    explicit Mouse(Core::HID::HIDCore& hid_core_, MouseSharedMemoryFormat& mouse_shared_memory);
     ~Mouse() override;
 
     // Called when the controller is initialized
@@ -29,17 +29,9 @@ public:
     void OnUpdate(const Core::Timing::CoreTiming& core_timing) override;
 
 private:
-    struct MouseSharedMemory {
-        // This is nn::hid::detail::MouseLifo
-        Lifo<Core::HID::MouseState, hid_entry_count> mouse_lifo{};
-        static_assert(sizeof(mouse_lifo) == 0x350, "mouse_lifo is an invalid size");
-        INSERT_PADDING_WORDS(0x2C);
-    };
-    static_assert(sizeof(MouseSharedMemory) == 0x400, "MouseSharedMemory is an invalid size");
-
     Core::HID::MouseState next_state{};
     Core::HID::AnalogStickState last_mouse_wheel_state{};
-    MouseSharedMemory* shared_memory = nullptr;
+    MouseSharedMemoryFormat& shared_memory;
     Core::HID::EmulatedDevices* emulated_devices = nullptr;
 };
 } // namespace Service::HID

src/core/hle/service/hid/controllers/npad.cpp

@@ -17,12 +17,12 @@
 #include "core/hle/kernel/k_event.h"
 #include "core/hle/kernel/k_readable_event.h"
 #include "core/hle/service/hid/controllers/npad.h"
+#include "core/hle/service/hid/controllers/shared_memory_format.h"
 #include "core/hle/service/hid/errors.h"
 #include "core/hle/service/hid/hid_util.h"
 #include "core/hle/service/kernel_helpers.h"
 
 namespace Service::HID {
-constexpr std::size_t NPAD_OFFSET = 0x9A00;
 constexpr std::array<Core::HID::NpadIdType, 10> npad_id_list{
     Core::HID::NpadIdType::Player1,  Core::HID::NpadIdType::Player2, Core::HID::NpadIdType::Player3,
     Core::HID::NpadIdType::Player4,  Core::HID::NpadIdType::Player5, Core::HID::NpadIdType::Player6,
@@ -30,14 +30,12 @@ constexpr std::array<Core::HID::NpadIdType, 10> npad_id_list{
     Core::HID::NpadIdType::Handheld,
 };
 
-NPad::NPad(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_,
+NPad::NPad(Core::HID::HIDCore& hid_core_, NpadSharedMemoryFormat& npad_shared_memory_format,
            KernelHelpers::ServiceContext& service_context_)
     : ControllerBase{hid_core_}, service_context{service_context_} {
-    static_assert(NPAD_OFFSET + (NPAD_COUNT * sizeof(NpadInternalState)) < shared_memory_size);
     for (std::size_t i = 0; i < controller_data.size(); ++i) {
         auto& controller = controller_data[i];
-        controller.shared_memory = std::construct_at(reinterpret_cast<NpadInternalState*>(
-            raw_shared_memory_ + NPAD_OFFSET + (i * sizeof(NpadInternalState))));
+        controller.shared_memory = &npad_shared_memory_format.npad_entry[i].internal_state;
         controller.device = hid_core.GetEmulatedControllerByIndex(i);
         controller.vibration[Core::HID::EmulatedDeviceIndex::LeftIndex].latest_vibration_value =
             Core::HID::DEFAULT_VIBRATION_VALUE;
@@ -617,7 +615,7 @@ void NPad::SetHoldType(NpadJoyHoldType joy_hold_type) {
     hold_type = joy_hold_type;
 }
 
-NPad::NpadJoyHoldType NPad::GetHoldType() const {
+NpadJoyHoldType NPad::GetHoldType() const {
     return hold_type;
 }
 
@@ -630,7 +628,7 @@ void NPad::SetNpadHandheldActivationMode(NpadHandheldActivationMode activation_m
     handheld_activation_mode = activation_mode;
 }
 
-NPad::NpadHandheldActivationMode NPad::GetNpadHandheldActivationMode() const {
+NpadHandheldActivationMode NPad::GetNpadHandheldActivationMode() const {
     return handheld_activation_mode;
 }
 
@@ -638,7 +636,7 @@ void NPad::SetNpadCommunicationMode(NpadCommunicationMode communication_mode_) {
     communication_mode = communication_mode_;
 }
 
-NPad::NpadCommunicationMode NPad::GetNpadCommunicationMode() const {
+NpadCommunicationMode NPad::GetNpadCommunicationMode() const {
     return communication_mode;
 }
 
@@ -978,27 +976,27 @@ Result NPad::ResetIsSixAxisSensorDeviceNewlyAssigned(
     return ResultSuccess;
 }
 
-NPad::SixAxisLifo& NPad::GetSixAxisFullkeyLifo(Core::HID::NpadIdType npad_id) {
+NpadSixAxisSensorLifo& NPad::GetSixAxisFullkeyLifo(Core::HID::NpadIdType npad_id) {
     return GetControllerFromNpadIdType(npad_id).shared_memory->sixaxis_fullkey_lifo;
 }
 
-NPad::SixAxisLifo& NPad::GetSixAxisHandheldLifo(Core::HID::NpadIdType npad_id) {
+NpadSixAxisSensorLifo& NPad::GetSixAxisHandheldLifo(Core::HID::NpadIdType npad_id) {
     return GetControllerFromNpadIdType(npad_id).shared_memory->sixaxis_handheld_lifo;
 }
 
-NPad::SixAxisLifo& NPad::GetSixAxisDualLeftLifo(Core::HID::NpadIdType npad_id) {
+NpadSixAxisSensorLifo& NPad::GetSixAxisDualLeftLifo(Core::HID::NpadIdType npad_id) {
     return GetControllerFromNpadIdType(npad_id).shared_memory->sixaxis_dual_left_lifo;
 }
 
-NPad::SixAxisLifo& NPad::GetSixAxisDualRightLifo(Core::HID::NpadIdType npad_id) {
+NpadSixAxisSensorLifo& NPad::GetSixAxisDualRightLifo(Core::HID::NpadIdType npad_id) {
     return GetControllerFromNpadIdType(npad_id).shared_memory->sixaxis_dual_right_lifo;
 }
 
-NPad::SixAxisLifo& NPad::GetSixAxisLeftLifo(Core::HID::NpadIdType npad_id) {
+NpadSixAxisSensorLifo& NPad::GetSixAxisLeftLifo(Core::HID::NpadIdType npad_id) {
     return GetControllerFromNpadIdType(npad_id).shared_memory->sixaxis_left_lifo;
 }
 
-NPad::SixAxisLifo& NPad::GetSixAxisRightLifo(Core::HID::NpadIdType npad_id) {
+NpadSixAxisSensorLifo& NPad::GetSixAxisRightLifo(Core::HID::NpadIdType npad_id) {
     return GetControllerFromNpadIdType(npad_id).shared_memory->sixaxis_right_lifo;
 }
 
@@ -1343,7 +1341,7 @@ const Core::HID::SixAxisSensorProperties& NPad::GetSixaxisProperties(
     }
 }
 
-NPad::AppletDetailedUiType NPad::GetAppletDetailedUiType(Core::HID::NpadIdType npad_id) {
+AppletDetailedUiType NPad::GetAppletDetailedUiType(Core::HID::NpadIdType npad_id) {
     const auto& shared_memory = GetControllerFromNpadIdType(npad_id).shared_memory;
 
     return {

src/core/hle/service/hid/controllers/npad.h

@@ -8,12 +8,10 @@
 #include <mutex>
 #include <span>
 
-#include "common/bit_field.h"
 #include "common/common_types.h"
-
 #include "core/hid/hid_types.h"
 #include "core/hle/service/hid/controllers/controller_base.h"
-#include "core/hle/service/hid/ring_lifo.h"
+#include "core/hle/service/hid/controllers/types/npad_types.h"
 
 namespace Core::HID {
 class EmulatedController;
@@ -32,10 +30,13 @@ class ServiceContext;
 union Result;
 
 namespace Service::HID {
+struct NpadInternalState;
+struct NpadSixAxisSensorLifo;
+struct NpadSharedMemoryFormat;
 
 class NPad final : public ControllerBase {
 public:
-    explicit NPad(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_,
+    explicit NPad(Core::HID::HIDCore& hid_core_, NpadSharedMemoryFormat& npad_shared_memory_format,
                   KernelHelpers::ServiceContext& service_context_);
     ~NPad() override;
 
@@ -48,89 +49,6 @@ public:
     // When the controller is requesting an update for the shared memory
     void OnUpdate(const Core::Timing::CoreTiming& core_timing) override;
 
-    // This is nn::hid::NpadJoyHoldType
-    enum class NpadJoyHoldType : u64 {
-        Vertical = 0,
-        Horizontal = 1,
-    };
-
-    // This is nn::hid::NpadJoyAssignmentMode
-    enum class NpadJoyAssignmentMode : u32 {
-        Dual = 0,
-        Single = 1,
-    };
-
-    // This is nn::hid::NpadJoyDeviceType
-    enum class NpadJoyDeviceType : s64 {
-        Left = 0,
-        Right = 1,
-    };
-
-    // This is nn::hid::NpadHandheldActivationMode
-    enum class NpadHandheldActivationMode : u64 {
-        Dual = 0,
-        Single = 1,
-        None = 2,
-        MaxActivationMode = 3,
-    };
-
-    // This is nn::hid::system::AppletFooterUiAttributesSet
-    struct AppletFooterUiAttributes {
-        INSERT_PADDING_BYTES(0x4);
-    };
-
-    // This is nn::hid::system::AppletFooterUiType
-    enum class AppletFooterUiType : u8 {
-        None = 0,
-        HandheldNone = 1,
-        HandheldJoyConLeftOnly = 2,
-        HandheldJoyConRightOnly = 3,
-        HandheldJoyConLeftJoyConRight = 4,
-        JoyDual = 5,
-        JoyDualLeftOnly = 6,
-        JoyDualRightOnly = 7,
-        JoyLeftHorizontal = 8,
-        JoyLeftVertical = 9,
-        JoyRightHorizontal = 10,
-        JoyRightVertical = 11,
-        SwitchProController = 12,
-        CompatibleProController = 13,
-        CompatibleJoyCon = 14,
-        LarkHvc1 = 15,
-        LarkHvc2 = 16,
-        LarkNesLeft = 17,
-        LarkNesRight = 18,
-        Lucia = 19,
-        Verification = 20,
-        Lagon = 21,
-    };
-
-    using AppletFooterUiVariant = u8;
-
-    // This is "nn::hid::system::AppletDetailedUiType".
-    struct AppletDetailedUiType {
-        AppletFooterUiVariant ui_variant;
-        INSERT_PADDING_BYTES(0x2);
-        AppletFooterUiType footer;
-    };
-    static_assert(sizeof(AppletDetailedUiType) == 0x4, "AppletDetailedUiType is an invalid size");
-    // This is nn::hid::NpadCommunicationMode
-    enum class NpadCommunicationMode : u64 {
-        Mode_5ms = 0,
-        Mode_10ms = 1,
-        Mode_15ms = 2,
-        Default = 3,
-    };
-
-    enum class NpadRevision : u32 {
-        Revision0 = 0,
-        Revision1 = 1,
-        Revision2 = 2,
-        Revision3 = 3,
-    };
-
-    using SixAxisLifo = Lifo<Core::HID::SixAxisSensorState, hid_entry_count>;
-
     void SetSupportedStyleSet(Core::HID::NpadStyleTag style_set);
     Core::HID::NpadStyleTag GetSupportedStyleSet() const;
 
@@ -188,12 +106,12 @@ public:
     Result ResetIsSixAxisSensorDeviceNewlyAssigned(
         const Core::HID::SixAxisSensorHandle& sixaxis_handle);
 
-    SixAxisLifo& GetSixAxisFullkeyLifo(Core::HID::NpadIdType npad_id);
-    SixAxisLifo& GetSixAxisHandheldLifo(Core::HID::NpadIdType npad_id);
-    SixAxisLifo& GetSixAxisDualLeftLifo(Core::HID::NpadIdType npad_id);
-    SixAxisLifo& GetSixAxisDualRightLifo(Core::HID::NpadIdType npad_id);
-    SixAxisLifo& GetSixAxisLeftLifo(Core::HID::NpadIdType npad_id);
-    SixAxisLifo& GetSixAxisRightLifo(Core::HID::NpadIdType npad_id);
+    NpadSixAxisSensorLifo& GetSixAxisFullkeyLifo(Core::HID::NpadIdType npad_id);
+    NpadSixAxisSensorLifo& GetSixAxisHandheldLifo(Core::HID::NpadIdType npad_id);
+    NpadSixAxisSensorLifo& GetSixAxisDualLeftLifo(Core::HID::NpadIdType npad_id);
+    NpadSixAxisSensorLifo& GetSixAxisDualRightLifo(Core::HID::NpadIdType npad_id);
+    NpadSixAxisSensorLifo& GetSixAxisLeftLifo(Core::HID::NpadIdType npad_id);
+    NpadSixAxisSensorLifo& GetSixAxisRightLifo(Core::HID::NpadIdType npad_id);
 
     Result GetLedPattern(Core::HID::NpadIdType npad_id, Core::HID::LedPattern& pattern) const;
     Result IsUnintendedHomeButtonInputProtectionEnabled(Core::HID::NpadIdType npad_id,
@@ -221,214 +139,6 @@ public:
     AppletDetailedUiType GetAppletDetailedUiType(Core::HID::NpadIdType npad_id);
 
 private:
-    static constexpr std::size_t NPAD_COUNT = 10;
-
-    // This is nn::hid::detail::ColorAttribute
-    enum class ColorAttribute : u32 {
-        Ok = 0,
-        ReadError = 1,
-        NoController = 2,
-    };
-    static_assert(sizeof(ColorAttribute) == 4, "ColorAttribute is an invalid size");
-
-    // This is nn::hid::detail::NpadFullKeyColorState
-    struct NpadFullKeyColorState {
-        ColorAttribute attribute{ColorAttribute::NoController};
-        Core::HID::NpadControllerColor fullkey{};
-    };
-    static_assert(sizeof(NpadFullKeyColorState) == 0xC, "NpadFullKeyColorState is an invalid size");
-
-    // This is nn::hid::detail::NpadJoyColorState
-    struct NpadJoyColorState {
-        ColorAttribute attribute{ColorAttribute::NoController};
-        Core::HID::NpadControllerColor left{};
-        Core::HID::NpadControllerColor right{};
-    };
-    static_assert(sizeof(NpadJoyColorState) == 0x14, "NpadJoyColorState is an invalid size");
-
-    // This is nn::hid::NpadAttribute
-    struct NpadAttribute {
-        union {
-            u32 raw{};
-            BitField<0, 1, u32> is_connected;
-            BitField<1, 1, u32> is_wired;
-            BitField<2, 1, u32> is_left_connected;
-            BitField<3, 1, u32> is_left_wired;
-            BitField<4, 1, u32> is_right_connected;
-            BitField<5, 1, u32> is_right_wired;
-        };
-    };
-    static_assert(sizeof(NpadAttribute) == 4, "NpadAttribute is an invalid size");
-
-    // This is nn::hid::NpadFullKeyState
-    // This is nn::hid::NpadHandheldState
-    // This is nn::hid::NpadJoyDualState
-    // This is nn::hid::NpadJoyLeftState
-    // This is nn::hid::NpadJoyRightState
-    // This is nn::hid::NpadPalmaState
-    // This is nn::hid::NpadSystemExtState
-    struct NPadGenericState {
-        s64_le sampling_number{};
-        Core::HID::NpadButtonState npad_buttons{};
-        Core::HID::AnalogStickState l_stick{};
-        Core::HID::AnalogStickState r_stick{};
-        NpadAttribute connection_status{};
-        INSERT_PADDING_BYTES(4); // Reserved
-    };
-    static_assert(sizeof(NPadGenericState) == 0x28, "NPadGenericState is an invalid size");
-
-    // This is nn::hid::server::NpadGcTriggerState
-    struct NpadGcTriggerState {
-        s64 sampling_number{};
-        s32 l_analog{};
-        s32 r_analog{};
-    };
-    static_assert(sizeof(NpadGcTriggerState) == 0x10, "NpadGcTriggerState is an invalid size");
-
-    // This is nn::hid::NpadSystemProperties
-    struct NPadSystemProperties {
-        union {
-            s64 raw{};
-            BitField<0, 1, s64> is_charging_joy_dual;
-            BitField<1, 1, s64> is_charging_joy_left;
-            BitField<2, 1, s64> is_charging_joy_right;
-            BitField<3, 1, s64> is_powered_joy_dual;
-            BitField<4, 1, s64> is_powered_joy_left;
-            BitField<5, 1, s64> is_powered_joy_right;
-            BitField<9, 1, s64> is_system_unsupported_button;
-            BitField<10, 1, s64> is_system_ext_unsupported_button;
-            BitField<11, 1, s64> is_vertical;
-            BitField<12, 1, s64> is_horizontal;
-            BitField<13, 1, s64> use_plus;
-            BitField<14, 1, s64> use_minus;
-            BitField<15, 1, s64> use_directional_buttons;
-        };
-    };
-    static_assert(sizeof(NPadSystemProperties) == 0x8, "NPadSystemProperties is an invalid size");
-
-    // This is nn::hid::NpadSystemButtonProperties
-    struct NpadSystemButtonProperties {
-        union {
-            s32 raw{};
-            BitField<0, 1, s32> is_home_button_protection_enabled;
-        };
-    };
-    static_assert(sizeof(NpadSystemButtonProperties) == 0x4,
-                  "NPadButtonProperties is an invalid size");
-
-    // This is nn::hid::system::DeviceType
-    struct DeviceType {
-        union {
-            u32 raw{};
-            BitField<0, 1, s32> fullkey;
-            BitField<1, 1, s32> debug_pad;
-            BitField<2, 1, s32> handheld_left;
-            BitField<3, 1, s32> handheld_right;
-            BitField<4, 1, s32> joycon_left;
-            BitField<5, 1, s32> joycon_right;
-            BitField<6, 1, s32> palma;
-            BitField<7, 1, s32> lark_hvc_left;
-            BitField<8, 1, s32> lark_hvc_right;
-            BitField<9, 1, s32> lark_nes_left;
-            BitField<10, 1, s32> lark_nes_right;
-            BitField<11, 1, s32> handheld_lark_hvc_left;
-            BitField<12, 1, s32> handheld_lark_hvc_right;
-            BitField<13, 1, s32> handheld_lark_nes_left;
-            BitField<14, 1, s32> handheld_lark_nes_right;
-            BitField<15, 1, s32> lucia;
-            BitField<16, 1, s32> lagon;
-            BitField<17, 1, s32> lager;
-            BitField<31, 1, s32> system;
-        };
-    };
-
-    // This is nn::hid::detail::NfcXcdDeviceHandleStateImpl
-    struct NfcXcdDeviceHandleStateImpl {
-        u64 handle{};
-        bool is_available{};
-        bool is_activated{};
-        INSERT_PADDING_BYTES(0x6); // Reserved
-        u64 sampling_number{};
-    };
-    static_assert(sizeof(NfcXcdDeviceHandleStateImpl) == 0x18,
-                  "NfcXcdDeviceHandleStateImpl is an invalid size");
-
-    // This is nn::hid::NpadLarkType
-    enum class NpadLarkType : u32 {
-        Invalid,
-        H1,
-        H2,
-        NL,
-        NR,
-    };
-
-    // This is nn::hid::NpadLuciaType
-    enum class NpadLuciaType : u32 {
-        Invalid,
-        J,
-        E,
-        U,
-    };
-
-    // This is nn::hid::NpadLagonType
-    enum class NpadLagonType : u32 {
-        Invalid,
-    };
-
-    // This is nn::hid::NpadLagerType
-    enum class NpadLagerType : u32 {
-        Invalid,
-        J,
-        E,
-        U,
-    };
-
-    // This is nn::hid::detail::NpadInternalState
-    struct NpadInternalState {
-        Core::HID::NpadStyleTag style_tag{Core::HID::NpadStyleSet::None};
-        NpadJoyAssignmentMode assignment_mode{NpadJoyAssignmentMode::Dual};
-        NpadFullKeyColorState fullkey_color{};
-        NpadJoyColorState joycon_color{};
-        Lifo<NPadGenericState, hid_entry_count> fullkey_lifo{};
-        Lifo<NPadGenericState, hid_entry_count> handheld_lifo{};
-        Lifo<NPadGenericState, hid_entry_count> joy_dual_lifo{};
-        Lifo<NPadGenericState, hid_entry_count> joy_left_lifo{};
-        Lifo<NPadGenericState, hid_entry_count> joy_right_lifo{};
-        Lifo<NPadGenericState, hid_entry_count> palma_lifo{};
-        Lifo<NPadGenericState, hid_entry_count> system_ext_lifo{};
-        Lifo<Core::HID::SixAxisSensorState, hid_entry_count> sixaxis_fullkey_lifo{};
-        Lifo<Core::HID::SixAxisSensorState, hid_entry_count> sixaxis_handheld_lifo{};
-        Lifo<Core::HID::SixAxisSensorState, hid_entry_count> sixaxis_dual_left_lifo{};
-        Lifo<Core::HID::SixAxisSensorState, hid_entry_count> sixaxis_dual_right_lifo{};
-        Lifo<Core::HID::SixAxisSensorState, hid_entry_count> sixaxis_left_lifo{};
-        Lifo<Core::HID::SixAxisSensorState, hid_entry_count> sixaxis_right_lifo{};
-        DeviceType device_type{};
-        INSERT_PADDING_BYTES(0x4); // Reserved
-        NPadSystemProperties system_properties{};
-        NpadSystemButtonProperties button_properties{};
-        Core::HID::NpadBatteryLevel battery_level_dual{};
-        Core::HID::NpadBatteryLevel battery_level_left{};
-        Core::HID::NpadBatteryLevel battery_level_right{};
-        AppletFooterUiAttributes applet_footer_attributes{};
-        AppletFooterUiType applet_footer_type{AppletFooterUiType::None};
-        INSERT_PADDING_BYTES(0x5B); // Reserved
-        INSERT_PADDING_BYTES(0x20); // Unknown
-        Lifo<NpadGcTriggerState, hid_entry_count> gc_trigger_lifo{};
-        NpadLarkType lark_type_l_and_main{};
-        NpadLarkType lark_type_r{};
-        NpadLuciaType lucia_type{};
-        NpadLagonType lagon_type{};
-        NpadLagerType lager_type{};
-        Core::HID::SixAxisSensorProperties sixaxis_fullkey_properties;
-        Core::HID::SixAxisSensorProperties sixaxis_handheld_properties;
-        Core::HID::SixAxisSensorProperties sixaxis_dual_left_properties;
-        Core::HID::SixAxisSensorProperties sixaxis_dual_right_properties;
-        Core::HID::SixAxisSensorProperties sixaxis_left_properties;
-        Core::HID::SixAxisSensorProperties sixaxis_right_properties;
-        INSERT_PADDING_BYTES(0xc06); // Unknown
-    };
-    static_assert(sizeof(NpadInternalState) == 0x5000, "NpadInternalState is an invalid size");
-
     struct VibrationData {
         bool device_mounted{};
         Core::HID::VibrationValue latest_vibration_value{};
@@ -479,7 +189,7 @@ private:
 
     std::atomic<u64> press_state{};
 
-    std::array<NpadControllerData, NPAD_COUNT> controller_data{};
+    std::array<NpadControllerData, NpadCount> controller_data{};
     KernelHelpers::ServiceContext& service_context;
     std::mutex mutex;
     std::vector<Core::HID::NpadIdType> supported_npad_id_types{};

src/core/hle/service/hid/controllers/palma.cpp

@@ -12,8 +12,7 @@
 
 namespace Service::HID {
 
-Palma::Palma(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_,
-             KernelHelpers::ServiceContext& service_context_)
+Palma::Palma(Core::HID::HIDCore& hid_core_, KernelHelpers::ServiceContext& service_context_)
     : ControllerBase{hid_core_}, service_context{service_context_} {
     controller = hid_core.GetEmulatedController(Core::HID::NpadIdType::Other);
     operation_complete_event = service_context.CreateEvent("hid:PalmaOperationCompleteEvent");

src/core/hle/service/hid/controllers/palma.h

@@ -97,8 +97,7 @@ public:
     static_assert(sizeof(PalmaConnectionHandle) == 0x8,
                   "PalmaConnectionHandle has incorrect size.");
 
-    explicit Palma(Core::HID::HIDCore& hid_core_, u8* raw_shared_memory_,
-                   KernelHelpers::ServiceContext& service_context_);
+    explicit Palma(Core::HID::HIDCore& hid_core_, KernelHelpers::ServiceContext& service_context_);
     ~Palma() override;
 
     // Called when the controller is initialized

src/core/hle/service/hid/controllers/shared_memory_format.h

@@ -0,0 +1,240 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+#include "common/vector_math.h"
+#include "core/hid/hid_types.h"
+#include "core/hle/service/hid//controllers/types/debug_pad_types.h"
+#include "core/hle/service/hid//controllers/types/keyboard_types.h"
+#include "core/hle/service/hid//controllers/types/mouse_types.h"
+#include "core/hle/service/hid//controllers/types/npad_types.h"
+#include "core/hle/service/hid//controllers/types/touch_types.h"
+#include "core/hle/service/hid/ring_lifo.h"
+
+namespace Service::HID {
+static const std::size_t HidEntryCount = 17;
+
+struct CommonHeader {
+    s64 timestamp{};
+    s64 total_entry_count{};
+    s64 last_entry_index{};
+    s64 entry_count{};
+};
+static_assert(sizeof(CommonHeader) == 0x20, "CommonHeader is an invalid size");
+
+// This is nn::hid::detail::DebugPadSharedMemoryFormat
+struct DebugPadSharedMemoryFormat {
+    // This is nn::hid::detail::DebugPadLifo
+    Lifo<DebugPadState, HidEntryCount> debug_pad_lifo{};
+    static_assert(sizeof(debug_pad_lifo) == 0x2C8, "debug_pad_lifo is an invalid size");
+    INSERT_PADDING_WORDS(0x4E);
+};
+static_assert(sizeof(DebugPadSharedMemoryFormat) == 0x400,
+              "DebugPadSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::TouchScreenSharedMemoryFormat
+struct TouchScreenSharedMemoryFormat {
+    // This is nn::hid::detail::TouchScreenLifo
+    Lifo<TouchScreenState, HidEntryCount> touch_screen_lifo{};
+    static_assert(sizeof(touch_screen_lifo) == 0x2C38, "touch_screen_lifo is an invalid size");
+    INSERT_PADDING_WORDS(0xF2);
+};
+static_assert(sizeof(TouchScreenSharedMemoryFormat) == 0x3000,
+              "TouchScreenSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::MouseSharedMemoryFormat
+struct MouseSharedMemoryFormat {
+    // This is nn::hid::detail::MouseLifo
+    Lifo<Core::HID::MouseState, HidEntryCount> mouse_lifo{};
+    static_assert(sizeof(mouse_lifo) == 0x350, "mouse_lifo is an invalid size");
+    INSERT_PADDING_WORDS(0x2C);
+};
+static_assert(sizeof(MouseSharedMemoryFormat) == 0x400,
+              "MouseSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::KeyboardSharedMemoryFormat
+struct KeyboardSharedMemoryFormat {
+    // This is nn::hid::detail::KeyboardLifo
+    Lifo<KeyboardState, HidEntryCount> keyboard_lifo{};
+    static_assert(sizeof(keyboard_lifo) == 0x3D8, "keyboard_lifo is an invalid size");
+    INSERT_PADDING_WORDS(0xA);
+};
+static_assert(sizeof(KeyboardSharedMemoryFormat) == 0x400,
+              "KeyboardSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::DigitizerSharedMemoryFormat
+struct DigitizerSharedMemoryFormat {
+    CommonHeader header;
+    INSERT_PADDING_BYTES(0xFE0);
+};
+static_assert(sizeof(DigitizerSharedMemoryFormat) == 0x1000,
+              "DigitizerSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::HomeButtonSharedMemoryFormat
+struct HomeButtonSharedMemoryFormat {
+    CommonHeader header;
+    INSERT_PADDING_BYTES(0x1E0);
+};
+static_assert(sizeof(HomeButtonSharedMemoryFormat) == 0x200,
+              "HomeButtonSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::SleepButtonSharedMemoryFormat
+struct SleepButtonSharedMemoryFormat {
+    CommonHeader header;
+    INSERT_PADDING_BYTES(0x1E0);
+};
+static_assert(sizeof(SleepButtonSharedMemoryFormat) == 0x200,
+              "SleepButtonSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::CaptureButtonSharedMemoryFormat
+struct CaptureButtonSharedMemoryFormat {
+    CommonHeader header;
+    INSERT_PADDING_BYTES(0x1E0);
+};
+static_assert(sizeof(CaptureButtonSharedMemoryFormat) == 0x200,
+              "CaptureButtonSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::InputDetectorSharedMemoryFormat
+struct InputDetectorSharedMemoryFormat {
+    CommonHeader header;
+    INSERT_PADDING_BYTES(0x7E0);
+};
+static_assert(sizeof(InputDetectorSharedMemoryFormat) == 0x800,
+              "InputDetectorSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::UniquePadSharedMemoryFormat
+struct UniquePadSharedMemoryFormat {
+    CommonHeader header;
+    INSERT_PADDING_BYTES(0x3FE0);
+};
+static_assert(sizeof(UniquePadSharedMemoryFormat) == 0x4000,
+              "UniquePadSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::NpadSixAxisSensorLifo
+struct NpadSixAxisSensorLifo {
+    Lifo<Core::HID::SixAxisSensorState, HidEntryCount> lifo;
+};
+
+// This is nn::hid::detail::NpadInternalState
+struct NpadInternalState {
+    Core::HID::NpadStyleTag style_tag{Core::HID::NpadStyleSet::None};
+    NpadJoyAssignmentMode assignment_mode{NpadJoyAssignmentMode::Dual};
+    NpadFullKeyColorState fullkey_color{};
+    NpadJoyColorState joycon_color{};
+    Lifo<NPadGenericState, HidEntryCount> fullkey_lifo{};
+    Lifo<NPadGenericState, HidEntryCount> handheld_lifo{};
+    Lifo<NPadGenericState, HidEntryCount> joy_dual_lifo{};
+    Lifo<NPadGenericState, HidEntryCount> joy_left_lifo{};
+    Lifo<NPadGenericState, HidEntryCount> joy_right_lifo{};
+    Lifo<NPadGenericState, HidEntryCount> palma_lifo{};
+    Lifo<NPadGenericState, HidEntryCount> system_ext_lifo{};
+    NpadSixAxisSensorLifo sixaxis_fullkey_lifo{};
+    NpadSixAxisSensorLifo sixaxis_handheld_lifo{};
+    NpadSixAxisSensorLifo sixaxis_dual_left_lifo{};
+    NpadSixAxisSensorLifo sixaxis_dual_right_lifo{};
+    NpadSixAxisSensorLifo sixaxis_left_lifo{};
+    NpadSixAxisSensorLifo sixaxis_right_lifo{};
+    DeviceType device_type{};
+    INSERT_PADDING_BYTES(0x4); // Reserved
+    NPadSystemProperties system_properties{};
+    NpadSystemButtonProperties button_properties{};
+    Core::HID::NpadBatteryLevel battery_level_dual{};
+    Core::HID::NpadBatteryLevel battery_level_left{};
+    Core::HID::NpadBatteryLevel battery_level_right{};
+    AppletFooterUiAttributes applet_footer_attributes{};
+    AppletFooterUiType applet_footer_type{AppletFooterUiType::None};
+    INSERT_PADDING_BYTES(0x5B); // Reserved
+    INSERT_PADDING_BYTES(0x20); // Unknown
+    Lifo<NpadGcTriggerState, HidEntryCount> gc_trigger_lifo{};
+    NpadLarkType lark_type_l_and_main{};
+    NpadLarkType lark_type_r{};
+    NpadLuciaType lucia_type{};
+    NpadLagerType lager_type{};
+    Core::HID::SixAxisSensorProperties sixaxis_fullkey_properties;
+    Core::HID::SixAxisSensorProperties sixaxis_handheld_properties;
+    Core::HID::SixAxisSensorProperties sixaxis_dual_left_properties;
+    Core::HID::SixAxisSensorProperties sixaxis_dual_right_properties;
+    Core::HID::SixAxisSensorProperties sixaxis_left_properties;
+    Core::HID::SixAxisSensorProperties sixaxis_right_properties;
+};
+static_assert(sizeof(NpadInternalState) == 0x43F8, "NpadInternalState is an invalid size");
+
+// This is nn::hid::detail::NpadSharedMemoryEntry
+struct NpadSharedMemoryEntry {
+    NpadInternalState internal_state;
+    INSERT_PADDING_BYTES(0xC08);
+};
+static_assert(sizeof(NpadSharedMemoryEntry) == 0x5000, "NpadSharedMemoryEntry is an invalid size");
+
+// This is nn::hid::detail::NpadSharedMemoryFormat
+struct NpadSharedMemoryFormat {
+    std::array<NpadSharedMemoryEntry, NpadCount> npad_entry;
+};
+static_assert(sizeof(NpadSharedMemoryFormat) == 0x32000,
+              "NpadSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::GestureSharedMemoryFormat
+struct GestureSharedMemoryFormat {
+    // This is nn::hid::detail::GestureLifo
+    Lifo<GestureState, HidEntryCount> gesture_lifo{};
+    static_assert(sizeof(gesture_lifo) == 0x708, "gesture_lifo is an invalid size");
+    INSERT_PADDING_WORDS(0x3E);
+};
+static_assert(sizeof(GestureSharedMemoryFormat) == 0x800,
+              "GestureSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::ConsoleSixAxisSensorSharedMemoryFormat
+struct ConsoleSixAxisSensorSharedMemoryFormat {
+    u64 sampling_number{};
+    bool is_seven_six_axis_sensor_at_rest{};
+    INSERT_PADDING_BYTES(3); // padding
+    f32 verticalization_error{};
+    Common::Vec3f gyro_bias{};
+    INSERT_PADDING_BYTES(4); // padding
+};
+static_assert(sizeof(ConsoleSixAxisSensorSharedMemoryFormat) == 0x20,
+              "ConsoleSixAxisSensorSharedMemoryFormat is an invalid size");
+
+// This is nn::hid::detail::SharedMemoryFormat
+struct SharedMemoryFormat {
+    void Initialize() {}
+
+    DebugPadSharedMemoryFormat debug_pad;
+    TouchScreenSharedMemoryFormat touch_screen;
+    MouseSharedMemoryFormat mouse;
+    KeyboardSharedMemoryFormat keyboard;
+    DigitizerSharedMemoryFormat digitizer;
+    HomeButtonSharedMemoryFormat home_button;
+    SleepButtonSharedMemoryFormat sleep_button;
+    CaptureButtonSharedMemoryFormat capture_button;
+    InputDetectorSharedMemoryFormat input_detector;
+    UniquePadSharedMemoryFormat unique_pad;
+    NpadSharedMemoryFormat npad;
+    GestureSharedMemoryFormat gesture;
+    ConsoleSixAxisSensorSharedMemoryFormat console;
+    INSERT_PADDING_BYTES(0x19E0);
+    MouseSharedMemoryFormat debug_mouse;
+    INSERT_PADDING_BYTES(0x2000);
+};
+static_assert(offsetof(SharedMemoryFormat, debug_pad) == 0x0, "debug_pad has wrong offset");
+static_assert(offsetof(SharedMemoryFormat, touch_screen) == 0x400, "touch_screen has wrong offset");
+static_assert(offsetof(SharedMemoryFormat, mouse) == 0x3400, "mouse has wrong offset");
+static_assert(offsetof(SharedMemoryFormat, keyboard) == 0x3800, "keyboard has wrong offset");
+static_assert(offsetof(SharedMemoryFormat, digitizer) == 0x3C00, "digitizer has wrong offset");
+static_assert(offsetof(SharedMemoryFormat, home_button) == 0x4C00, "home_button has wrong offset");
+static_assert(offsetof(SharedMemoryFormat, sleep_button) == 0x4E00,
+              "sleep_button has wrong offset");
+static_assert(offsetof(SharedMemoryFormat, capture_button) == 0x5000,
+              "capture_button has wrong offset");
+static_assert(offsetof(SharedMemoryFormat, input_detector) == 0x5200,
+              "input_detector has wrong offset");
+static_assert(offsetof(SharedMemoryFormat, npad) == 0x9A00, "npad has wrong offset");
+static_assert(offsetof(SharedMemoryFormat, gesture) == 0x3BA00, "gesture has wrong offset");
+static_assert(offsetof(SharedMemoryFormat, console) == 0x3C200, "console has wrong offset");
+static_assert(offsetof(SharedMemoryFormat, debug_mouse) == 0x3DC00, "debug_mouse has wrong offset");
+static_assert(sizeof(SharedMemoryFormat) == 0x40000, "SharedMemoryFormat is an invalid size");
+
+} // namespace Service::HID

src/core/hle/service/hid/controllers/shared_memory_holder.cpp

@@ -0,0 +1,53 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#include "core/core.h"
+#include "core/hle/kernel/k_shared_memory.h"
+#include "core/hle/service/hid/controllers/shared_memory_format.h"
+#include "core/hle/service/hid/controllers/shared_memory_holder.h"
+#include "core/hle/service/hid/errors.h"
+
+namespace Service::HID {
+SharedMemoryHolder::SharedMemoryHolder() {}
+
+SharedMemoryHolder::~SharedMemoryHolder() {
+    Finalize();
+}
+
+Result SharedMemoryHolder::Initialize(Core::System& system) {
+    shared_memory = Kernel::KSharedMemory::Create(system.Kernel());
+    const Result result = shared_memory->Initialize(
+        system.DeviceMemory(), nullptr, Kernel::Svc::MemoryPermission::None,
+        Kernel::Svc::MemoryPermission::Read, sizeof(SharedMemoryFormat));
+    if (result.IsError()) {
+        return result;
+    }
+    Kernel::KSharedMemory::Register(system.Kernel(), shared_memory);
+
+    is_created = true;
+    is_mapped = true;
+    address = std::construct_at(reinterpret_cast<SharedMemoryFormat*>(shared_memory->GetPointer()));
+    return ResultSuccess;
+}
+
+void SharedMemoryHolder::Finalize() {
+    if (address != nullptr) {
+        shared_memory->Close();
+    }
+    is_created = false;
+    is_mapped = false;
+    address = nullptr;
+}
+
+bool SharedMemoryHolder::IsMapped() {
+    return is_mapped;
+}
+
+SharedMemoryFormat* SharedMemoryHolder::GetAddress() {
+    return address;
+}
+
+Kernel::KSharedMemory* SharedMemoryHolder::GetHandle() {
+    return shared_memory;
+}
+} // namespace Service::HID