early-access version 4028

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 4027.
+This is the source code for early-access 4028.
 
 ## Legal Notice
 

src/core/arm/arm_interface.cpp

@@ -9,7 +9,7 @@
 
 namespace Core {
 
-void ArmInterface::LogBacktrace(const Kernel::KProcess* process) const {
+void ArmInterface::LogBacktrace(Kernel::KProcess* process) const {
     Kernel::Svc::ThreadContext ctx;
     this->GetContext(ctx);
 

src/core/arm/arm_interface.h

@@ -95,7 +95,7 @@ public:
     virtual void SignalInterrupt(Kernel::KThread* thread) = 0;
 
     // Stack trace generation.
-    void LogBacktrace(const Kernel::KProcess* process) const;
+    void LogBacktrace(Kernel::KProcess* process) const;
 
     // Debug functionality.
     virtual const Kernel::DebugWatchpoint* HaltedWatchpoint() const = 0;

src/core/arm/debug.cpp

@@ -79,7 +79,7 @@ constexpr std::array<u64, 2> SegmentBases{
     0x7100000000ULL,
 };
 
-void SymbolicateBacktrace(const Kernel::KProcess* process, std::vector<BacktraceEntry>& out) {
+void SymbolicateBacktrace(Kernel::KProcess* process, std::vector<BacktraceEntry>& out) {
     auto modules = FindModules(process);
 
     const bool is_64 = process->Is64Bit();
@@ -118,7 +118,7 @@ void SymbolicateBacktrace(const Kernel::KProcess* process, std::vector<Backtrace
     }
 }
 
-std::vector<BacktraceEntry> GetAArch64Backtrace(const Kernel::KProcess* process,
+std::vector<BacktraceEntry> GetAArch64Backtrace(Kernel::KProcess* process,
                                                 const Kernel::Svc::ThreadContext& ctx) {
     std::vector<BacktraceEntry> out;
     auto& memory = process->GetMemory();
@@ -144,7 +144,7 @@ std::vector<BacktraceEntry> GetAArch64Backtrace(const Kernel::KProcess* process,
     return out;
 }
 
-std::vector<BacktraceEntry> GetAArch32Backtrace(const Kernel::KProcess* process,
+std::vector<BacktraceEntry> GetAArch32Backtrace(Kernel::KProcess* process,
                                                 const Kernel::Svc::ThreadContext& ctx) {
     std::vector<BacktraceEntry> out;
     auto& memory = process->GetMemory();
@@ -173,7 +173,7 @@ std::vector<BacktraceEntry> GetAArch32Backtrace(const Kernel::KProcess* process,
 } // namespace
 
 std::optional<std::string> GetThreadName(const Kernel::KThread* thread) {
-    const auto* process = thread->GetOwnerProcess();
+    auto* process = thread->GetOwnerProcess();
     if (process->Is64Bit()) {
         return GetNameFromThreadType64(process->GetMemory(), *thread);
     } else {
@@ -248,7 +248,7 @@ Kernel::KProcessAddress GetModuleEnd(const Kernel::KProcess* process,
     return cur_addr - 1;
 }
 
-Loader::AppLoader::Modules FindModules(const Kernel::KProcess* process) {
+Loader::AppLoader::Modules FindModules(Kernel::KProcess* process) {
     Loader::AppLoader::Modules modules;
 
     auto& page_table = process->GetPageTable();
@@ -312,7 +312,7 @@ Loader::AppLoader::Modules FindModules(const Kernel::KProcess* process) {
     return modules;
 }
 
-Kernel::KProcessAddress FindMainModuleEntrypoint(const Kernel::KProcess* process) {
+Kernel::KProcessAddress FindMainModuleEntrypoint(Kernel::KProcess* process) {
     // Do we have any loaded executable sections?
     auto modules = FindModules(process);
 
@@ -337,7 +337,7 @@ void InvalidateInstructionCacheRange(const Kernel::KProcess* process, u64 addres
     }
 }
 
-std::vector<BacktraceEntry> GetBacktraceFromContext(const Kernel::KProcess* process,
+std::vector<BacktraceEntry> GetBacktraceFromContext(Kernel::KProcess* process,
                                                     const Kernel::Svc::ThreadContext& ctx) {
     if (process->Is64Bit()) {
         return GetAArch64Backtrace(process, ctx);

src/core/arm/debug.h

@@ -14,9 +14,9 @@ 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);
+Loader::AppLoader::Modules FindModules(Kernel::KProcess* process);
 Kernel::KProcessAddress GetModuleEnd(const Kernel::KProcess* process, Kernel::KProcessAddress base);
-Kernel::KProcessAddress FindMainModuleEntrypoint(const Kernel::KProcess* process);
+Kernel::KProcessAddress FindMainModuleEntrypoint(Kernel::KProcess* process);
 
 void InvalidateInstructionCacheRange(const Kernel::KProcess* process, u64 address, u64 size);
 
@@ -28,7 +28,7 @@ struct BacktraceEntry {
     std::string name;
 };
 
-std::vector<BacktraceEntry> GetBacktraceFromContext(const Kernel::KProcess* process,
+std::vector<BacktraceEntry> GetBacktraceFromContext(Kernel::KProcess* process,
                                                     const Kernel::Svc::ThreadContext& ctx);
 std::vector<BacktraceEntry> GetBacktrace(const Kernel::KThread* thread);
 

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -15,7 +15,7 @@ using namespace Common::Literals;
 
 class DynarmicCallbacks32 : public Dynarmic::A32::UserCallbacks {
 public:
-    explicit DynarmicCallbacks32(ArmDynarmic32& parent, const Kernel::KProcess* process)
+    explicit DynarmicCallbacks32(ArmDynarmic32& parent, 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 ||
@@ -169,7 +169,7 @@ public:
 
     ArmDynarmic32& m_parent;
     Core::Memory::Memory& m_memory;
-    const Kernel::KProcess* m_process{};
+    Kernel::KProcess* m_process{};
     const bool m_debugger_enabled{};
     const bool m_check_memory_access{};
     static constexpr u64 MinimumRunCycles = 10000U;
@@ -370,7 +370,7 @@ void ArmDynarmic32::RewindBreakpointInstruction() {
     this->SetContext(m_breakpoint_context);
 }
 
-ArmDynarmic32::ArmDynarmic32(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+ArmDynarmic32::ArmDynarmic32(System& system, bool uses_wall_clock, 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)),

src/core/arm/dynarmic/arm_dynarmic_32.h

@@ -20,7 +20,7 @@ class System;
 
 class ArmDynarmic32 final : public ArmInterface {
 public:
-    ArmDynarmic32(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+    ArmDynarmic32(System& system, bool uses_wall_clock, Kernel::KProcess* process,
                   DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index);
     ~ArmDynarmic32() override;
 

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -15,7 +15,7 @@ using namespace Common::Literals;
 
 class DynarmicCallbacks64 : public Dynarmic::A64::UserCallbacks {
 public:
-    explicit DynarmicCallbacks64(ArmDynarmic64& parent, const Kernel::KProcess* process)
+    explicit DynarmicCallbacks64(ArmDynarmic64& parent, 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 ||
@@ -216,7 +216,7 @@ public:
     Core::Memory::Memory& m_memory;
     u64 m_tpidrro_el0{};
     u64 m_tpidr_el0{};
-    const Kernel::KProcess* m_process{};
+    Kernel::KProcess* m_process{};
     const bool m_debugger_enabled{};
     const bool m_check_memory_access{};
     static constexpr u64 MinimumRunCycles = 10000U;
@@ -399,7 +399,7 @@ void ArmDynarmic64::RewindBreakpointInstruction() {
     this->SetContext(m_breakpoint_context);
 }
 
-ArmDynarmic64::ArmDynarmic64(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+ArmDynarmic64::ArmDynarmic64(System& system, bool uses_wall_clock, 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} {

src/core/arm/dynarmic/arm_dynarmic_64.h

@@ -25,7 +25,7 @@ class System;
 
 class ArmDynarmic64 final : public ArmInterface {
 public:
-    ArmDynarmic64(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+    ArmDynarmic64(System& system, bool uses_wall_clock, Kernel::KProcess* process,
                   DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index);
     ~ArmDynarmic64() override;
 

src/core/core.cpp

@@ -28,7 +28,6 @@
 #include "core/file_sys/savedata_factory.h"
 #include "core/file_sys/vfs_concat.h"
 #include "core/file_sys/vfs_real.h"
-#include "core/gpu_dirty_memory_manager.h"
 #include "core/hid/hid_core.h"
 #include "core/hle/kernel/k_memory_manager.h"
 #include "core/hle/kernel/k_process.h"
@@ -130,11 +129,8 @@ 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{},
+        : kernel{system}, fs_controller{system}, hid_core{}, room_network{}, cpu_manager{system},
-          cpu_manager{system}, reporter{system}, applet_manager{system}, profile_manager{},
+          reporter{system}, applet_manager{system}, profile_manager{}, time_manager{system} {}
-          time_manager{system}, gpu_dirty_memory_write_manager{} {
-        memory.SetGPUDirtyManagers(gpu_dirty_memory_write_manager);
-    }
 
     void Initialize(System& system) {
         device_memory = std::make_unique<Core::DeviceMemory>();
@@ -241,17 +237,17 @@ struct System::Impl {
         debugger = std::make_unique<Debugger>(system, port);
     }
 
-    SystemResultStatus SetupForApplicationProcess(System& system, Frontend::EmuWindow& emu_window) {
+    void InitializeKernel(System& system) {
         LOG_DEBUG(Core, "initialized OK");
 
         // Setting changes may require a full system reinitialization (e.g., disabling multicore).
         ReinitializeIfNecessary(system);
 
-        memory.SetGPUDirtyManagers(gpu_dirty_memory_write_manager);
-
         kernel.Initialize();
         cpu_manager.Initialize();
+    }
 
+    SystemResultStatus SetupForApplicationProcess(System& system, Frontend::EmuWindow& emu_window) {
         /// Reset all glue registrations
         arp_manager.ResetAll();
 
@@ -300,17 +296,9 @@ struct System::Impl {
             return SystemResultStatus::ErrorGetLoader;
         }
 
-        SystemResultStatus init_result{SetupForApplicationProcess(system, emu_window)};
+        InitializeKernel(system);
-        if (init_result != SystemResultStatus::Success) {
-            LOG_CRITICAL(Core, "Failed to initialize system (Error {})!",
-                         static_cast<int>(init_result));
-            ShutdownMainProcess();
-            return init_result;
-        }
 
-        telemetry_session->AddInitialInfo(*app_loader, fs_controller, *content_provider);
+        // Create the application process.
-
-        // Create the process.
         auto main_process = Kernel::KProcess::Create(system.Kernel());
         Kernel::KProcess::Register(system.Kernel(), main_process);
         kernel.AppendNewProcess(main_process);
@@ -323,7 +311,18 @@ struct System::Impl {
             return static_cast<SystemResultStatus>(
                 static_cast<u32>(SystemResultStatus::ErrorLoader) + static_cast<u32>(load_result));
         }
+
+        // Set up the rest of the system.
+        SystemResultStatus init_result{SetupForApplicationProcess(system, emu_window)};
+        if (init_result != SystemResultStatus::Success) {
+            LOG_CRITICAL(Core, "Failed to initialize system (Error {})!",
+                         static_cast<int>(init_result));
+            ShutdownMainProcess();
+            return init_result;
+        }
+
         AddGlueRegistrationForProcess(*app_loader, *main_process);
+        telemetry_session->AddInitialInfo(*app_loader, fs_controller, *content_provider);
 
         // Initialize cheat engine
         if (cheat_engine) {
@@ -426,7 +425,6 @@ struct System::Impl {
         cpu_manager.Shutdown();
         debugger.reset();
         kernel.Shutdown();
-        memory.Reset();
         Network::RestartSocketOperations();
 
         if (auto room_member = room_network.GetRoomMember().lock()) {
@@ -507,7 +505,6 @@ struct System::Impl {
     std::unique_ptr<Tegra::Host1x::Host1x> host1x_core;
     std::unique_ptr<Core::DeviceMemory> device_memory;
     std::unique_ptr<AudioCore::AudioCore> audio_core;
-    Core::Memory::Memory memory;
     Core::HID::HIDCore hid_core;
     Network::RoomNetwork room_network;
 
@@ -567,9 +564,6 @@ struct System::Impl {
     std::array<u64, Core::Hardware::NUM_CPU_CORES> dynarmic_ticks{};
     std::array<MicroProfileToken, Core::Hardware::NUM_CPU_CORES> microprofile_cpu{};
 
-    std::array<Core::GPUDirtyMemoryManager, Core::Hardware::NUM_CPU_CORES>
-        gpu_dirty_memory_write_manager{};
-
     std::deque<std::vector<u8>> user_channel;
 };
 
@@ -652,29 +646,12 @@ void System::PrepareReschedule(const u32 core_index) {
     impl->kernel.PrepareReschedule(core_index);
 }
 
-Core::GPUDirtyMemoryManager& System::CurrentGPUDirtyMemoryManager() {
-    const std::size_t core = impl->kernel.GetCurrentHostThreadID();
-    return impl->gpu_dirty_memory_write_manager[core < Core::Hardware::NUM_CPU_CORES
-                                                    ? core
-                                                    : Core::Hardware::NUM_CPU_CORES - 1];
-}
-
-/// Provides a constant reference to the current gou dirty memory manager.
-const Core::GPUDirtyMemoryManager& System::CurrentGPUDirtyMemoryManager() const {
-    const std::size_t core = impl->kernel.GetCurrentHostThreadID();
-    return impl->gpu_dirty_memory_write_manager[core < Core::Hardware::NUM_CPU_CORES
-                                                    ? core
-                                                    : Core::Hardware::NUM_CPU_CORES - 1];
-}
-
 size_t System::GetCurrentHostThreadID() const {
     return impl->kernel.GetCurrentHostThreadID();
 }
 
 void System::GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback) {
-    for (auto& manager : impl->gpu_dirty_memory_write_manager) {
+    return this->ApplicationProcess()->GatherGPUDirtyMemory(callback);
-        manager.Gather(callback);
-    }
 }
 
 PerfStatsResults System::GetAndResetPerfStats() {
@@ -723,20 +700,12 @@ const Kernel::KProcess* System::ApplicationProcess() const {
     return impl->kernel.ApplicationProcess();
 }
 
-ExclusiveMonitor& System::Monitor() {
-    return impl->kernel.GetExclusiveMonitor();
-}
-
-const ExclusiveMonitor& System::Monitor() const {
-    return impl->kernel.GetExclusiveMonitor();
-}
-
 Memory::Memory& System::ApplicationMemory() {
-    return impl->memory;
+    return impl->kernel.ApplicationProcess()->GetMemory();
 }
 
 const Core::Memory::Memory& System::ApplicationMemory() const {
-    return impl->memory;
+    return impl->kernel.ApplicationProcess()->GetMemory();
 }
 
 Tegra::GPU& System::GPU() {

src/core/core.h

@@ -116,7 +116,6 @@ class CpuManager;
 class Debugger;
 class DeviceMemory;
 class ExclusiveMonitor;
-class GPUDirtyMemoryManager;
 class PerfStats;
 class Reporter;
 class SpeedLimiter;
@@ -225,12 +224,6 @@ public:
     /// Prepare the core emulation for a reschedule
     void PrepareReschedule(u32 core_index);
 
-    /// Provides a reference to the gou dirty memory manager.
-    [[nodiscard]] Core::GPUDirtyMemoryManager& CurrentGPUDirtyMemoryManager();
-
-    /// Provides a constant reference to the current gou dirty memory manager.
-    [[nodiscard]] const Core::GPUDirtyMemoryManager& CurrentGPUDirtyMemoryManager() const;
-
     void GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback);
 
     [[nodiscard]] size_t GetCurrentHostThreadID() const;
@@ -250,12 +243,6 @@ public:
     /// Gets a const reference to the underlying CPU manager
     [[nodiscard]] const CpuManager& GetCpuManager() const;
 
-    /// Gets a reference to the exclusive monitor
-    [[nodiscard]] ExclusiveMonitor& Monitor();
-
-    /// Gets a constant reference to the exclusive monitor
-    [[nodiscard]] const ExclusiveMonitor& Monitor() const;
-
     /// Gets a mutable reference to the system memory instance.
     [[nodiscard]] Core::Memory::Memory& ApplicationMemory();
 

src/core/file_sys/program_metadata.cpp

@@ -166,6 +166,10 @@ u32 ProgramMetadata::GetSystemResourceSize() const {
     return npdm_header.system_resource_size;
 }
 
+PoolPartition ProgramMetadata::GetPoolPartition() const {
+    return acid_header.pool_partition;
+}
+
 const ProgramMetadata::KernelCapabilityDescriptors& ProgramMetadata::GetKernelCapabilities() const {
     return aci_kernel_capabilities;
 }
@@ -201,7 +205,7 @@ void ProgramMetadata::Print() const {
     // Begin ACID printing (potential perms, signed)
     LOG_DEBUG(Service_FS, "Magic:                  {:.4}", acid_header.magic.data());
     LOG_DEBUG(Service_FS, "Flags:                  0x{:02X}", acid_header.flags);
-    LOG_DEBUG(Service_FS, " > Is Retail:           {}", acid_header.is_retail ? "YES" : "NO");
+    LOG_DEBUG(Service_FS, " > Is Retail:           {}", acid_header.production_flag ? "YES" : "NO");
     LOG_DEBUG(Service_FS, "Title ID Min:           0x{:016X}", acid_header.title_id_min);
     LOG_DEBUG(Service_FS, "Title ID Max:           0x{:016X}", acid_header.title_id_max);
     LOG_DEBUG(Service_FS, "Filesystem Access:      0x{:016X}\n", acid_file_access.permissions);

src/core/file_sys/program_metadata.h

@@ -34,6 +34,13 @@ enum class ProgramFilePermission : u64 {
     Everything = 1ULL << 63,
 };
 
+enum class PoolPartition : u32 {
+    Application = 0,
+    Applet = 1,
+    System = 2,
+    SystemNonSecure = 3,
+};
+
 /**
  * Helper which implements an interface to parse Program Description Metadata (NPDM)
  * Data can either be loaded from a file path or with data and an offset into it.
@@ -72,6 +79,7 @@ public:
     u64 GetTitleID() const;
     u64 GetFilesystemPermissions() const;
     u32 GetSystemResourceSize() const;
+    PoolPartition GetPoolPartition() const;
     const KernelCapabilityDescriptors& GetKernelCapabilities() const;
     const std::array<u8, 0x10>& GetName() const {
         return npdm_header.application_name;
@@ -116,8 +124,9 @@ private:
         union {
             u32 flags;
 
-            BitField<0, 1, u32> is_retail;
+            BitField<0, 1, u32> production_flag;
-            BitField<1, 31, u32> flags_unk;
+            BitField<1, 1, u32> unqualified_approval;
+            BitField<2, 4, PoolPartition> pool_partition;
         };
         u64_le title_id_min;
         u64_le title_id_max;

src/core/hle/kernel/k_address_arbiter.cpp

@@ -4,6 +4,7 @@
 #include "core/arm/exclusive_monitor.h"
 #include "core/core.h"
 #include "core/hle/kernel/k_address_arbiter.h"
+#include "core/hle/kernel/k_process.h"
 #include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
 #include "core/hle/kernel/k_thread.h"
@@ -26,9 +27,9 @@ bool ReadFromUser(KernelCore& kernel, s32* out, KProcessAddress address) {
     return true;
 }
 
-bool DecrementIfLessThan(Core::System& system, s32* out, KProcessAddress address, s32 value) {
+bool DecrementIfLessThan(KernelCore& kernel, s32* out, KProcessAddress address, s32 value) {
-    auto& monitor = system.Monitor();
+    auto& monitor = GetCurrentProcess(kernel).GetExclusiveMonitor();
-    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
+    const auto current_core = kernel.CurrentPhysicalCoreIndex();
 
     // NOTE: If scheduler lock is not held here, interrupt disable is required.
     // KScopedInterruptDisable di;
@@ -66,10 +67,10 @@ bool DecrementIfLessThan(Core::System& system, s32* out, KProcessAddress address
     return true;
 }
 
-bool UpdateIfEqual(Core::System& system, s32* out, KProcessAddress address, s32 value,
+bool UpdateIfEqual(KernelCore& kernel, s32* out, KProcessAddress address, s32 value,
                    s32 new_value) {
-    auto& monitor = system.Monitor();
+    auto& monitor = GetCurrentProcess(kernel).GetExclusiveMonitor();
-    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
+    const auto current_core = kernel.CurrentPhysicalCoreIndex();
 
     // NOTE: If scheduler lock is not held here, interrupt disable is required.
     // KScopedInterruptDisable di;
@@ -159,7 +160,7 @@ Result KAddressArbiter::SignalAndIncrementIfEqual(uint64_t addr, s32 value, s32
 
         // Check the userspace value.
         s32 user_value{};
-        R_UNLESS(UpdateIfEqual(m_system, std::addressof(user_value), addr, value, value + 1),
+        R_UNLESS(UpdateIfEqual(m_kernel, std::addressof(user_value), addr, value, value + 1),
                  ResultInvalidCurrentMemory);
         R_UNLESS(user_value == value, ResultInvalidState);
 
@@ -219,7 +220,7 @@ Result KAddressArbiter::SignalAndModifyByWaitingCountIfEqual(uint64_t addr, s32
         s32 user_value{};
         bool succeeded{};
         if (value != new_value) {
-            succeeded = UpdateIfEqual(m_system, std::addressof(user_value), addr, value, new_value);
+            succeeded = UpdateIfEqual(m_kernel, std::addressof(user_value), addr, value, new_value);
         } else {
             succeeded = ReadFromUser(m_kernel, std::addressof(user_value), addr);
         }
@@ -262,7 +263,7 @@ Result KAddressArbiter::WaitIfLessThan(uint64_t addr, s32 value, bool decrement,
         s32 user_value{};
         bool succeeded{};
         if (decrement) {
-            succeeded = DecrementIfLessThan(m_system, std::addressof(user_value), addr, value);
+            succeeded = DecrementIfLessThan(m_kernel, std::addressof(user_value), addr, value);
         } else {
             succeeded = ReadFromUser(m_kernel, std::addressof(user_value), addr);
         }

src/core/hle/kernel/k_client_port.cpp

@@ -58,9 +58,8 @@ Result KClientPort::CreateSession(KClientSession** out) {
     KSession* session{};
 
     // Reserve a new session from the resource limit.
-    //! FIXME: we are reserving this from the wrong resource limit!
+    KScopedResourceReservation session_reservation(GetCurrentProcessPointer(m_kernel),
-    KScopedResourceReservation session_reservation(
+                                                   LimitableResource::SessionCountMax);
-        m_kernel.ApplicationProcess()->GetResourceLimit(), LimitableResource::SessionCountMax);
     R_UNLESS(session_reservation.Succeeded(), ResultLimitReached);
 
     // Allocate a session normally.

src/core/hle/kernel/k_condition_variable.cpp

@@ -28,10 +28,10 @@ bool WriteToUser(KernelCore& kernel, KProcessAddress address, const u32* p) {
     return true;
 }
 
-bool UpdateLockAtomic(Core::System& system, u32* out, KProcessAddress address, u32 if_zero,
+bool UpdateLockAtomic(KernelCore& kernel, u32* out, KProcessAddress address, u32 if_zero,
                       u32 new_orr_mask) {
-    auto& monitor = system.Monitor();
+    auto& monitor = GetCurrentProcess(kernel).GetExclusiveMonitor();
-    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
+    const auto current_core = kernel.CurrentPhysicalCoreIndex();
 
     u32 expected{};
 
@@ -208,7 +208,7 @@ void KConditionVariable::SignalImpl(KThread* thread) {
         // TODO(bunnei): We should call CanAccessAtomic(..) here.
         can_access = true;
         if (can_access) [[likely]] {
-            UpdateLockAtomic(m_system, std::addressof(prev_tag), address, own_tag,
+            UpdateLockAtomic(m_kernel, std::addressof(prev_tag), address, own_tag,
                              Svc::HandleWaitMask);
         }
     }

src/core/hle/kernel/k_handle_table.h

@@ -30,7 +30,7 @@ public:
 public:
     explicit KHandleTable(KernelCore& kernel) : m_kernel(kernel) {}
 
-    Result Initialize(s32 size) {
+    Result Initialize(KProcess* owner, s32 size) {
         // Check that the table size is valid.
         R_UNLESS(size <= static_cast<s32>(MaxTableSize), ResultOutOfMemory);
 
@@ -44,6 +44,7 @@ public:
         m_next_linear_id = MinLinearId;
         m_count = 0;
         m_free_head_index = -1;
+        m_owner = owner;
 
         // Free all entries.
         for (s32 i = 0; i < static_cast<s32>(m_table_size); ++i) {
@@ -90,8 +91,8 @@ public:
         // Handle pseudo-handles.
         if constexpr (std::derived_from<KProcess, T>) {
             if (handle == Svc::PseudoHandle::CurrentProcess) {
-                //! FIXME: this is the wrong process!
+                // TODO: this should be the current process
-                auto* const cur_process = m_kernel.ApplicationProcess();
+                auto* const cur_process = m_owner;
                 ASSERT(cur_process != nullptr);
                 return cur_process;
             }
@@ -301,6 +302,7 @@ private:
 
 private:
     KernelCore& m_kernel;
+    KProcess* m_owner{};
     std::array<EntryInfo, MaxTableSize> m_entry_infos{};
     std::array<KAutoObject*, MaxTableSize> m_objects{};
     mutable KSpinLock m_lock;

src/core/hle/kernel/k_process.cpp

@@ -306,12 +306,16 @@ Result KProcess::Initialize(const Svc::CreateProcessParameter& params, const KPa
             False(params.flags & Svc::CreateProcessFlag::DisableDeviceAddressSpaceMerge);
         R_TRY(m_page_table.Initialize(as_type, enable_aslr, enable_das_merge, !enable_aslr, pool,
                                       params.code_address, params.code_num_pages * PageSize,
-                                      m_system_resource, res_limit, this->GetMemory(), 0));
+                                      m_system_resource, res_limit, m_memory, 0));
     }
     ON_RESULT_FAILURE_2 {
         m_page_table.Finalize();
     };
 
+    // Ensure our memory is initialized.
+    m_memory.SetCurrentPageTable(*this);
+    m_memory.SetGPUDirtyManagers(m_dirty_memory_managers);
+
     // Ensure we can insert the code region.
     R_UNLESS(m_page_table.CanContain(params.code_address, params.code_num_pages * PageSize,
                                      KMemoryState::Code),
@@ -399,12 +403,16 @@ Result KProcess::Initialize(const Svc::CreateProcessParameter& params,
             False(params.flags & Svc::CreateProcessFlag::DisableDeviceAddressSpaceMerge);
         R_TRY(m_page_table.Initialize(as_type, enable_aslr, enable_das_merge, !enable_aslr, pool,
                                       params.code_address, code_size, m_system_resource, res_limit,
-                                      this->GetMemory(), aslr_space_start));
+                                      m_memory, aslr_space_start));
     }
     ON_RESULT_FAILURE_2 {
         m_page_table.Finalize();
     };
 
+    // Ensure our memory is initialized.
+    m_memory.SetCurrentPageTable(*this);
+    m_memory.SetGPUDirtyManagers(m_dirty_memory_managers);
+
     // Ensure we can insert the code region.
     R_UNLESS(m_page_table.CanContain(params.code_address, code_size, KMemoryState::Code),
              ResultInvalidMemoryRegion);
@@ -1094,8 +1102,7 @@ void KProcess::UnpinThread(KThread* thread) {
 
 Result KProcess::GetThreadList(s32* out_num_threads, KProcessAddress out_thread_ids,
                                s32 max_out_count) {
-    // TODO: use current memory reference
+    auto& memory = this->GetMemory();
-    auto& memory = m_kernel.System().ApplicationMemory();
 
     // Lock the list.
     KScopedLightLock lk(m_list_lock);
@@ -1128,14 +1135,15 @@ void KProcess::Switch(KProcess* cur_process, KProcess* next_process) {}
 KProcess::KProcess(KernelCore& kernel)
     : KAutoObjectWithSlabHeapAndContainer(kernel), m_page_table{kernel}, m_state_lock{kernel},
       m_list_lock{kernel}, m_cond_var{kernel.System()}, m_address_arbiter{kernel.System()},
-      m_handle_table{kernel} {}
+      m_handle_table{kernel}, m_dirty_memory_managers{},
+      m_exclusive_monitor{}, m_memory{kernel.System()} {}
 KProcess::~KProcess() = default;
 
 Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std::size_t code_size,
                                   KProcessAddress aslr_space_start, bool is_hbl) {
     // Create a resource limit for the process.
-    const auto physical_memory_size =
+    const auto pool = static_cast<KMemoryManager::Pool>(metadata.GetPoolPartition());
-        m_kernel.MemoryManager().GetSize(Kernel::KMemoryManager::Pool::Application);
+    const auto physical_memory_size = m_kernel.MemoryManager().GetSize(pool);
     auto* res_limit =
         Kernel::CreateResourceLimitForProcess(m_kernel.System(), physical_memory_size);
 
@@ -1146,8 +1154,10 @@ Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std:
     Svc::CreateProcessFlag flag{};
     u64 code_address{};
 
-    // We are an application.
+    // Determine if we are an application.
-    flag |= Svc::CreateProcessFlag::IsApplication;
+    if (pool == KMemoryManager::Pool::Application) {
+        flag |= Svc::CreateProcessFlag::IsApplication;
+    }
 
     // If we are 64-bit, create as such.
     if (metadata.Is64BitProgram()) {
@@ -1196,8 +1206,8 @@ Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std:
     std::memcpy(params.name.data(), name.data(), sizeof(params.name));
 
     // Initialize for application process.
-    R_TRY(this->Initialize(params, metadata.GetKernelCapabilities(), res_limit,
+    R_TRY(this->Initialize(params, metadata.GetKernelCapabilities(), res_limit, pool,
-                           KMemoryManager::Pool::Application, aslr_space_start));
+                           aslr_space_start));
 
     // Assign remaining properties.
     m_is_hbl = is_hbl;
@@ -1223,7 +1233,7 @@ void KProcess::LoadModule(CodeSet code_set, KProcessAddress base_addr) {
     ReprotectSegment(code_set.DataSegment(), Svc::MemoryPermission::ReadWrite);
 
 #ifdef HAS_NCE
-    if (Settings::IsNceEnabled()) {
+    if (this->IsApplication() && Settings::IsNceEnabled()) {
         auto& buffer = m_kernel.System().DeviceMemory().buffer;
         const auto& code = code_set.CodeSegment();
         const auto& patch = code_set.PatchSegment();
@@ -1235,10 +1245,11 @@ void KProcess::LoadModule(CodeSet code_set, KProcessAddress base_addr) {
 }
 
 void KProcess::InitializeInterfaces() {
-    this->GetMemory().SetCurrentPageTable(*this);
+    m_exclusive_monitor =
+        Core::MakeExclusiveMonitor(this->GetMemory(), Core::Hardware::NUM_CPU_CORES);
 
 #ifdef HAS_NCE
-    if (this->Is64Bit() && Settings::IsNceEnabled()) {
+    if (this->IsApplication() && 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);
         }
@@ -1248,13 +1259,13 @@ void KProcess::InitializeInterfaces() {
         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);
+                static_cast<Core::DynarmicExclusiveMonitor&>(*m_exclusive_monitor), 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);
+                static_cast<Core::DynarmicExclusiveMonitor&>(*m_exclusive_monitor), i);
         }
     }
 }
@@ -1305,9 +1316,10 @@ bool KProcess::RemoveWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointT
     return true;
 }
 
-Core::Memory::Memory& KProcess::GetMemory() const {
+void KProcess::GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback) {
-    // TODO: per-process memory
+    for (auto& manager : m_dirty_memory_managers) {
-    return m_kernel.System().ApplicationMemory();
+        manager.Gather(callback);
+    }
 }
 
 } // namespace Kernel

src/core/hle/kernel/k_process.h

@@ -7,6 +7,7 @@
 
 #include "core/arm/arm_interface.h"
 #include "core/file_sys/program_metadata.h"
+#include "core/gpu_dirty_memory_manager.h"
 #include "core/hle/kernel/code_set.h"
 #include "core/hle/kernel/k_address_arbiter.h"
 #include "core/hle/kernel/k_capabilities.h"
@@ -17,6 +18,7 @@
 #include "core/hle/kernel/k_system_resource.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/k_thread_local_page.h"
+#include "core/memory.h"
 
 namespace Kernel {
 
@@ -126,6 +128,9 @@ private:
 #ifdef HAS_NCE
     std::unordered_map<u64, u64> m_post_handlers{};
 #endif
+    std::array<Core::GPUDirtyMemoryManager, Core::Hardware::NUM_CPU_CORES> m_dirty_memory_managers;
+    std::unique_ptr<Core::ExclusiveMonitor> m_exclusive_monitor;
+    Core::Memory::Memory m_memory;
 
 private:
     Result StartTermination();
@@ -502,7 +507,15 @@ public:
 
     void InitializeInterfaces();
 
-    Core::Memory::Memory& GetMemory() const;
+    Core::Memory::Memory& GetMemory() {
+        return m_memory;
+    }
+
+    void GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback);
+
+    Core::ExclusiveMonitor& GetExclusiveMonitor() const {
+        return *m_exclusive_monitor;
+    }
 
 public:
     // Overridden parent functions.
@@ -539,7 +552,7 @@ private:
 
     Result InitializeHandleTable(s32 size) {
         // Try to initialize the handle table.
-        R_TRY(m_handle_table.Initialize(size));
+        R_TRY(m_handle_table.Initialize(this, size));
 
         // We succeeded, so note that we did.
         m_is_handle_table_initialized = true;

src/core/hle/kernel/k_server_session.h

@@ -49,14 +49,21 @@ public:
     bool IsSignaled() const override;
     void OnClientClosed();
 
-    /// TODO: flesh these out to match the real kernel
     Result OnRequest(KSessionRequest* request);
-    Result SendReply(bool is_hle = false);
+    Result SendReply(uintptr_t server_message, uintptr_t server_buffer_size,
-    Result ReceiveRequest(std::shared_ptr<Service::HLERequestContext>* out_context = nullptr,
+                     KPhysicalAddress server_message_paddr, bool is_hle = false);
+    Result ReceiveRequest(uintptr_t server_message, uintptr_t server_buffer_size,
+                          KPhysicalAddress server_message_paddr,
+                          std::shared_ptr<Service::HLERequestContext>* out_context = nullptr,
                           std::weak_ptr<Service::SessionRequestManager> manager = {});
 
     Result SendReplyHLE() {
-        return SendReply(true);
+        R_RETURN(this->SendReply(0, 0, 0, true));
+    }
+
+    Result ReceiveRequestHLE(std::shared_ptr<Service::HLERequestContext>* out_context,
+                             std::weak_ptr<Service::SessionRequestManager> manager) {
+        R_RETURN(this->ReceiveRequest(0, 0, 0, out_context, manager));
     }
 
 private:

src/core/hle/kernel/k_session.cpp

@@ -33,8 +33,7 @@ void KSession::Initialize(KClientPort* client_port, uintptr_t name) {
     m_name = name;
 
     // Set our owner process.
-    //! FIXME: this is the wrong process!
+    m_process = GetCurrentProcessPointer(m_kernel);
-    m_process = m_kernel.ApplicationProcess();
     m_process->Open();
 
     // Set our port.

src/core/hle/kernel/k_thread.cpp

@@ -1422,8 +1422,7 @@ s32 GetCurrentCoreId(KernelCore& kernel) {
 }
 
 Core::Memory::Memory& GetCurrentMemory(KernelCore& kernel) {
-    // TODO: per-process memory
+    return GetCurrentProcess(kernel).GetMemory();
-    return kernel.System().ApplicationMemory();
 }
 
 KScopedDisableDispatch::~KScopedDisableDispatch() {

src/core/hle/kernel/k_thread.h

@@ -314,11 +314,7 @@ public:
         m_current_core_id = core;
     }
 
-    KProcess* GetOwnerProcess() {
+    KProcess* GetOwnerProcess() const {
-        return m_parent;
-    }
-
-    const KProcess* GetOwnerProcess() const {
         return m_parent;
     }
 

src/core/hle/kernel/kernel.cpp

@@ -68,8 +68,6 @@ struct KernelCore::Impl {
 
         global_object_list_container = std::make_unique<KAutoObjectWithListContainer>(kernel);
         global_scheduler_context = std::make_unique<Kernel::GlobalSchedulerContext>(kernel);
-        global_handle_table = std::make_unique<Kernel::KHandleTable>(kernel);
-        global_handle_table->Initialize(KHandleTable::MaxTableSize);
 
         is_phantom_mode_for_singlecore = false;
 
@@ -121,13 +119,8 @@ struct KernelCore::Impl {
         next_user_process_id = KProcess::ProcessIdMin;
         next_thread_id = 1;
 
-        global_handle_table->Finalize();
-        global_handle_table.reset();
-
         preemption_event = nullptr;
 
-        exclusive_monitor.reset();
-
         // Cleanup persistent kernel objects
         auto CleanupObject = [](KAutoObject* obj) {
             if (obj) {
@@ -191,8 +184,6 @@ struct KernelCore::Impl {
     }
 
     void InitializePhysicalCores() {
-        exclusive_monitor =
-            Core::MakeExclusiveMonitor(system.ApplicationMemory(), Core::Hardware::NUM_CPU_CORES);
         for (u32 i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
             const s32 core{static_cast<s32>(i)};
 
@@ -791,10 +782,6 @@ struct KernelCore::Impl {
 
     std::shared_ptr<Core::Timing::EventType> preemption_event;
 
-    // This is the kernel's handle table or supervisor handle table which
-    // stores all the objects in place.
-    std::unique_ptr<KHandleTable> global_handle_table;
-
     std::unique_ptr<KAutoObjectWithListContainer> global_object_list_container;
 
     std::unique_ptr<KObjectNameGlobalData> object_name_global_data;
@@ -805,7 +792,6 @@ struct KernelCore::Impl {
     std::mutex server_lock;
     std::vector<std::unique_ptr<Service::ServerManager>> server_managers;
 
-    std::unique_ptr<Core::ExclusiveMonitor> exclusive_monitor;
     std::array<std::unique_ptr<Kernel::PhysicalCore>, Core::Hardware::NUM_CPU_CORES> cores;
 
     // Next host thead ID to use, 0-3 IDs represent core threads, >3 represent others
@@ -882,10 +868,6 @@ KResourceLimit* KernelCore::GetSystemResourceLimit() {
     return impl->system_resource_limit;
 }
 
-KScopedAutoObject<KThread> KernelCore::RetrieveThreadFromGlobalHandleTable(Handle handle) const {
-    return impl->global_handle_table->GetObject<KThread>(handle);
-}
-
 void KernelCore::AppendNewProcess(KProcess* process) {
     impl->process_list.push_back(process);
 }
@@ -959,14 +941,6 @@ Kernel::KHardwareTimer& KernelCore::HardwareTimer() {
     return *impl->hardware_timer;
 }
 
-Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() {
-    return *impl->exclusive_monitor;
-}
-
-const Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() const {
-    return *impl->exclusive_monitor;
-}
-
 KAutoObjectWithListContainer& KernelCore::ObjectListContainer() {
     return *impl->global_object_list_container;
 }
@@ -1030,14 +1004,6 @@ u64 KernelCore::CreateNewUserProcessID() {
     return impl->next_user_process_id++;
 }
 
-KHandleTable& KernelCore::GlobalHandleTable() {
-    return *impl->global_handle_table;
-}
-
-const KHandleTable& KernelCore::GlobalHandleTable() const {
-    return *impl->global_handle_table;
-}
-
 void KernelCore::RegisterCoreThread(std::size_t core_id) {
     impl->RegisterCoreThread(core_id);
 }

src/core/hle/kernel/kernel.h

@@ -116,9 +116,6 @@ public:
     /// Retrieves a shared pointer to the system resource limit instance.
     KResourceLimit* GetSystemResourceLimit();
 
-    /// Retrieves a shared pointer to a Thread instance within the thread wakeup handle table.
-    KScopedAutoObject<KThread> RetrieveThreadFromGlobalHandleTable(Handle handle) const;
-
     /// Adds the given shared pointer to an internal list of active processes.
     void AppendNewProcess(KProcess* process);
 
@@ -170,10 +167,6 @@ public:
     /// Stops execution of 'id' core, in order to reschedule a new thread.
     void PrepareReschedule(std::size_t id);
 
-    Core::ExclusiveMonitor& GetExclusiveMonitor();
-
-    const Core::ExclusiveMonitor& GetExclusiveMonitor() const;
-
     KAutoObjectWithListContainer& ObjectListContainer();
 
     const KAutoObjectWithListContainer& ObjectListContainer() const;

src/core/hle/kernel/message_buffer.h

@@ -18,13 +18,13 @@ public:
         static constexpr inline u64 NullTag = 0;
 
     public:
-        enum class ReceiveListCountType : u32 {
+        enum ReceiveListCountType : u32 {
-            None = 0,
+            ReceiveListCountType_None = 0,
-            ToMessageBuffer = 1,
+            ReceiveListCountType_ToMessageBuffer = 1,
-            ToSingleBuffer = 2,
+            ReceiveListCountType_ToSingleBuffer = 2,
 
-            CountOffset = 2,
+            ReceiveListCountType_CountOffset = 2,
-            CountMax = 13,
+            ReceiveListCountType_CountMax = 13,
         };
 
     private:
@@ -591,16 +591,16 @@ public:
         // Add the size of the receive list.
         const auto count = hdr.GetReceiveListCount();
         switch (count) {
-        case MessageHeader::ReceiveListCountType::None:
+        case MessageHeader::ReceiveListCountType_None:
             break;
-        case MessageHeader::ReceiveListCountType::ToMessageBuffer:
+        case MessageHeader::ReceiveListCountType_ToMessageBuffer:
             break;
-        case MessageHeader::ReceiveListCountType::ToSingleBuffer:
+        case MessageHeader::ReceiveListCountType_ToSingleBuffer:
             msg_size += ReceiveListEntry::GetDataSize();
             break;
         default:
             msg_size += (static_cast<s32>(count) -
-                         static_cast<s32>(MessageHeader::ReceiveListCountType::CountOffset)) *
+                         static_cast<s32>(MessageHeader::ReceiveListCountType_CountOffset)) *
                         ReceiveListEntry::GetDataSize();
             break;
         }

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

@@ -118,7 +118,6 @@ Result GetInfo(Core::System& system, u64* result, InfoType info_id_type, Handle
             R_SUCCEED();
 
         case InfoType::IsApplication:
-            LOG_WARNING(Kernel_SVC, "(STUBBED) Assuming process is application");
             *result = process->IsApplication();
             R_SUCCEED();
 

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

@@ -48,8 +48,7 @@ Result ReplyAndReceiveImpl(KernelCore& kernel, int32_t* out_index, uintptr_t mes
         };
 
         // Send the reply.
-        R_TRY(session->SendReply());
+        R_TRY(session->SendReply(message, buffer_size, message_paddr));
-        // R_TRY(session->SendReply(message, buffer_size, message_paddr));
     }
 
     // Receive a message.
@@ -85,8 +84,7 @@ Result ReplyAndReceiveImpl(KernelCore& kernel, int32_t* out_index, uintptr_t mes
             if (R_SUCCEEDED(result)) {
                 KServerSession* session = objs[index]->DynamicCast<KServerSession*>();
                 if (session != nullptr) {
-                    // result = session->ReceiveRequest(message, buffer_size, message_paddr);
+                    result = session->ReceiveRequest(message, buffer_size, message_paddr);
-                    result = session->ReceiveRequest();
                     if (ResultNotFound == result) {
                         continue;
                     }

src/core/hle/kernel/svc_results.h

@@ -38,7 +38,9 @@ constexpr Result ResultInvalidState{ErrorModule::Kernel, 125};
 constexpr Result ResultReservedUsed{ErrorModule::Kernel, 126};
 constexpr Result ResultPortClosed{ErrorModule::Kernel, 131};
 constexpr Result ResultLimitReached{ErrorModule::Kernel, 132};
+constexpr Result ResultReceiveListBroken{ErrorModule::Kernel, 258};
 constexpr Result ResultOutOfAddressSpace{ErrorModule::Kernel, 259};
+constexpr Result ResultMessageTooLarge{ErrorModule::Kernel, 260};
 constexpr Result ResultInvalidId{ErrorModule::Kernel, 519};
 
 } // namespace Kernel

src/core/hle/service/fatal/fatal.cpp

@@ -89,7 +89,7 @@ static void GenerateErrorReport(Core::System& system, Result error_code, const F
         crash_report += fmt::format("    ESR:                         {:016x}\n", info.esr);
         crash_report += fmt::format("    FAR:                         {:016x}\n", info.far);
         crash_report += "\nBacktrace:\n";
-        for (size_t i = 0; i < info.backtrace_size; i++) {
+        for (u32 i = 0; i < std::min<u32>(info.backtrace_size, 32); i++) {
             crash_report +=
                 fmt::format("    Backtrace[{:02d}]:               {:016x}\n", i, info.backtrace[i]);
         }

src/core/hle/service/ipc_helpers.h

@@ -151,8 +151,8 @@ public:
         if (manager->IsDomain()) {
             context->AddDomainObject(std::move(iface));
         } else {
-            kernel.ApplicationProcess()->GetResourceLimit()->Reserve(
+            ASSERT(Kernel::GetCurrentProcess(kernel).GetResourceLimit()->Reserve(
-                Kernel::LimitableResource::SessionCountMax, 1);
+                Kernel::LimitableResource::SessionCountMax, 1));
 
             auto* session = Kernel::KSession::Create(kernel);
             session->Initialize(nullptr, 0);

src/core/hle/service/server_manager.cpp

@@ -47,7 +47,7 @@ ServerManager::~ServerManager() {
     m_stopped.Wait();
     m_threads.clear();
 
-    // Clean up ports.
+    // Clean up server ports.
     for (const auto& [port, handler] : m_ports) {
         port->Close();
     }
@@ -97,22 +97,15 @@ Result ServerManager::RegisterNamedService(const std::string& service_name,
                                            u32 max_sessions) {
     ASSERT(m_sessions.size() + m_ports.size() < MaximumWaitObjects);
 
-    // Add the new server to sm:.
+    // Add the new server to sm: and get the moved server port.
-    ASSERT(R_SUCCEEDED(
+    Kernel::KServerPort* server_port{};
-        m_system.ServiceManager().RegisterService(service_name, max_sessions, handler_factory)));
+    R_ASSERT(m_system.ServiceManager().RegisterService(std::addressof(server_port), service_name,
-
+                                                       max_sessions, handler_factory));
-    // Get the registered port.
-    Kernel::KPort* port{};
-    ASSERT(
-        R_SUCCEEDED(m_system.ServiceManager().GetServicePort(std::addressof(port), service_name)));
-
-    // Open a new reference to the server port.
-    port->GetServerPort().Open();
 
     // Begin tracking the server port.
     {
         std::scoped_lock ll{m_list_mutex};
-        m_ports.emplace(std::addressof(port->GetServerPort()), std::move(handler_factory));
+        m_ports.emplace(server_port, std::move(handler_factory));
     }
 
     // Signal the wakeup event.
@@ -372,7 +365,7 @@ Result ServerManager::OnSessionEvent(Kernel::KServerSession* session,
 
     // Try to receive a message.
     std::shared_ptr<HLERequestContext> context;
-    rc = session->ReceiveRequest(&context, manager);
+    rc = session->ReceiveRequestHLE(&context, manager);
 
     // If the session has been closed, we're done.
     if (rc == Kernel::ResultSessionClosed) {

src/core/hle/service/set/set_sys.cpp

@@ -507,6 +507,14 @@ void SET_SYS::SetTvSettings(HLERequestContext& ctx) {
     rb.Push(ResultSuccess);
 }
 
+void SET_SYS::GetDebugModeFlag(HLERequestContext& ctx) {
+    LOG_DEBUG(Service_SET, "called");
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(ResultSuccess);
+    rb.Push<u32>(0);
+}
+
 void SET_SYS::GetQuestFlag(HLERequestContext& ctx) {
     LOG_WARNING(Service_SET, "(STUBBED) called");
 
@@ -926,7 +934,7 @@ SET_SYS::SET_SYS(Core::System& system_) : ServiceFramework{system_, "set:sys"},
         {59, &SET_SYS::SetNetworkSystemClockContext, "SetNetworkSystemClockContext"},
         {60, &SET_SYS::IsUserSystemClockAutomaticCorrectionEnabled, "IsUserSystemClockAutomaticCorrectionEnabled"},
         {61, &SET_SYS::SetUserSystemClockAutomaticCorrectionEnabled, "SetUserSystemClockAutomaticCorrectionEnabled"},
-        {62, nullptr, "GetDebugModeFlag"},
+        {62, &SET_SYS::GetDebugModeFlag, "GetDebugModeFlag"},
         {63, &SET_SYS::GetPrimaryAlbumStorage, "GetPrimaryAlbumStorage"},
         {64, nullptr, "SetPrimaryAlbumStorage"},
         {65, nullptr, "GetUsb30EnableFlag"},
@@ -1143,6 +1151,8 @@ void SET_SYS::StoreSettings() {
 }
 
 void SET_SYS::StoreSettingsThreadFunc(std::stop_token stop_token) {
+    Common::SetCurrentThreadName("SettingsStore");
+
     while (Common::StoppableTimedWait(stop_token, std::chrono::minutes(1))) {
         std::scoped_lock l{m_save_needed_mutex};
         if (!std::exchange(m_save_needed, false)) {

src/core/hle/service/set/set_sys.h

@@ -98,6 +98,7 @@ private:
     void GetSettingsItemValue(HLERequestContext& ctx);
     void GetTvSettings(HLERequestContext& ctx);
     void SetTvSettings(HLERequestContext& ctx);
+    void GetDebugModeFlag(HLERequestContext& ctx);
     void GetQuestFlag(HLERequestContext& ctx);
     void GetDeviceTimeZoneLocationName(HLERequestContext& ctx);
     void SetDeviceTimeZoneLocationName(HLERequestContext& ctx);

src/core/hle/service/sm/sm.cpp

@@ -29,8 +29,7 @@ ServiceManager::ServiceManager(Kernel::KernelCore& kernel_) : kernel{kernel_} {
 
 ServiceManager::~ServiceManager() {
     for (auto& [name, port] : service_ports) {
-        port->GetClientPort().Close();
+        port->Close();
-        port->GetServerPort().Close();
     }
 
     if (deferral_event) {
@@ -50,8 +49,8 @@ static Result ValidateServiceName(const std::string& name) {
     return ResultSuccess;
 }
 
-Result ServiceManager::RegisterService(std::string name, u32 max_sessions,
+Result ServiceManager::RegisterService(Kernel::KServerPort** out_server_port, std::string name,
-                                       SessionRequestHandlerFactory handler) {
+                                       u32 max_sessions, SessionRequestHandlerFactory handler) {
     R_TRY(ValidateServiceName(name));
 
     std::scoped_lock lk{lock};
@@ -66,13 +65,17 @@ Result ServiceManager::RegisterService(std::string name, u32 max_sessions,
     // Register the port.
     Kernel::KPort::Register(kernel, port);
 
-    service_ports.emplace(name, port);
+    service_ports.emplace(name, std::addressof(port->GetClientPort()));
     registered_services.emplace(name, handler);
     if (deferral_event) {
         deferral_event->Signal();
     }
 
-    return ResultSuccess;
+    // Set our output.
+    *out_server_port = std::addressof(port->GetServerPort());
+
+    // We succeeded.
+    R_SUCCEED();
 }
 
 Result ServiceManager::UnregisterService(const std::string& name) {
@@ -91,7 +94,8 @@ Result ServiceManager::UnregisterService(const std::string& name) {
     return ResultSuccess;
 }
 
-Result ServiceManager::GetServicePort(Kernel::KPort** out_port, const std::string& name) {
+Result ServiceManager::GetServicePort(Kernel::KClientPort** out_client_port,
+                                      const std::string& name) {
     R_TRY(ValidateServiceName(name));
 
     std::scoped_lock lk{lock};
@@ -101,7 +105,7 @@ Result ServiceManager::GetServicePort(Kernel::KPort** out_port, const std::strin
         return Service::SM::ResultNotRegistered;
     }
 
-    *out_port = it->second;
+    *out_client_port = it->second;
     return ResultSuccess;
 }
 
@@ -172,8 +176,8 @@ Result SM::GetServiceImpl(Kernel::KClientSession** out_client_session, HLEReques
     std::string name(PopServiceName(rp));
 
     // Find the named port.
-    Kernel::KPort* port{};
+    Kernel::KClientPort* client_port{};
-    auto port_result = service_manager.GetServicePort(&port, name);
+    auto port_result = service_manager.GetServicePort(&client_port, name);
     if (port_result == Service::SM::ResultInvalidServiceName) {
         LOG_ERROR(Service_SM, "Invalid service name '{}'", name);
         return Service::SM::ResultInvalidServiceName;
@@ -187,7 +191,7 @@ Result SM::GetServiceImpl(Kernel::KClientSession** out_client_session, HLEReques
 
     // Create a new session.
     Kernel::KClientSession* session{};
-    if (const auto result = port->GetClientPort().CreateSession(&session); result.IsError()) {
+    if (const auto result = client_port->CreateSession(&session); result.IsError()) {
         LOG_ERROR(Service_SM, "called service={} -> error 0x{:08X}", name, result.raw);
         return result;
     }
@@ -221,7 +225,9 @@ void SM::RegisterServiceImpl(HLERequestContext& ctx, std::string name, u32 max_s
     LOG_DEBUG(Service_SM, "called with name={}, max_session_count={}, is_light={}", name,
               max_session_count, is_light);
 
-    if (const auto result = service_manager.RegisterService(name, max_session_count, nullptr);
+    Kernel::KServerPort* server_port{};
+    if (const auto result = service_manager.RegisterService(std::addressof(server_port), name,
+                                                            max_session_count, nullptr);
         result.IsError()) {
         LOG_ERROR(Service_SM, "failed to register service with error_code={:08X}", result.raw);
         IPC::ResponseBuilder rb{ctx, 2};
@@ -229,13 +235,9 @@ void SM::RegisterServiceImpl(HLERequestContext& ctx, std::string name, u32 max_s
         return;
     }
 
-    auto* port = Kernel::KPort::Create(kernel);
-    port->Initialize(ServerSessionCountMax, is_light, 0);
-    SCOPE_EXIT({ port->GetClientPort().Close(); });
-
     IPC::ResponseBuilder rb{ctx, 2, 0, 1, IPC::ResponseBuilder::Flags::AlwaysMoveHandles};
     rb.Push(ResultSuccess);
-    rb.PushMoveObjects(port->GetServerPort());
+    rb.PushMoveObjects(server_port);
 }
 
 void SM::UnregisterService(HLERequestContext& ctx) {

src/core/hle/service/sm/sm.h

@@ -56,10 +56,10 @@ public:
     explicit ServiceManager(Kernel::KernelCore& kernel_);
     ~ServiceManager();
 
-    Result RegisterService(std::string name, u32 max_sessions,
+    Result RegisterService(Kernel::KServerPort** out_server_port, std::string name,
-                           SessionRequestHandlerFactory handler_factory);
+                           u32 max_sessions, SessionRequestHandlerFactory handler_factory);
     Result UnregisterService(const std::string& name);
-    Result GetServicePort(Kernel::KPort** out_port, const std::string& name);
+    Result GetServicePort(Kernel::KClientPort** out_client_port, const std::string& name);
 
     template <Common::DerivedFrom<SessionRequestHandler> T>
     std::shared_ptr<T> GetService(const std::string& service_name) const {
@@ -84,7 +84,7 @@ private:
     /// Map of registered services, retrieved using GetServicePort.
     std::mutex lock;
     std::unordered_map<std::string, SessionRequestHandlerFactory> registered_services;
-    std::unordered_map<std::string, Kernel::KPort*> service_ports;
+    std::unordered_map<std::string, Kernel::KClientPort*> service_ports;
 
     /// Kernel context
     Kernel::KernelCore& kernel;

src/core/hle/service/sm/sm_controller.cpp

@@ -28,7 +28,6 @@ void Controller::ConvertCurrentObjectToDomain(HLERequestContext& ctx) {
 void Controller::CloneCurrentObject(HLERequestContext& ctx) {
     LOG_DEBUG(Service, "called");
 
-    auto& process = *ctx.GetThread().GetOwnerProcess();
     auto session_manager = ctx.GetManager();
 
     // FIXME: this is duplicated from the SVC, it should just call it instead
@@ -36,11 +35,11 @@ void Controller::CloneCurrentObject(HLERequestContext& ctx) {
 
     // Reserve a new session from the process resource limit.
     Kernel::KScopedResourceReservation session_reservation(
-        &process, Kernel::LimitableResource::SessionCountMax);
+        Kernel::GetCurrentProcessPointer(kernel), Kernel::LimitableResource::SessionCountMax);
     ASSERT(session_reservation.Succeeded());
 
     // Create the session.
-    Kernel::KSession* session = Kernel::KSession::Create(system.Kernel());
+    Kernel::KSession* session = Kernel::KSession::Create(kernel);
     ASSERT(session != nullptr);
 
     // Initialize the session.
@@ -50,7 +49,7 @@ void Controller::CloneCurrentObject(HLERequestContext& ctx) {
     session_reservation.Commit();
 
     // Register the session.
-    Kernel::KSession::Register(system.Kernel(), session);
+    Kernel::KSession::Register(kernel, session);
 
     // Register with server manager.
     session_manager->GetServerManager().RegisterSession(&session->GetServerSession(),

src/core/loader/deconstructed_rom_directory.cpp

@@ -129,9 +129,10 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
     }
     metadata.Print();
 
-    // Enable NCE only for programs with 39-bit address space.
+    // Enable NCE only for applications with 39-bit address space.
     const bool is_39bit =
         metadata.GetAddressSpaceType() == FileSys::ProgramAddressSpaceType::Is39Bit;
+    const bool is_application = metadata.GetPoolPartition() == FileSys::PoolPartition::Application;
     Settings::SetNceEnabled(is_39bit);
 
     const std::array static_modules = {"rtld",    "main",    "subsdk0", "subsdk1", "subsdk2",
@@ -147,7 +148,7 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
 
     const auto GetPatcher = [&](size_t i) -> Core::NCE::Patcher* {
 #ifdef HAS_NCE
-        if (Settings::IsNceEnabled()) {
+        if (is_application && Settings::IsNceEnabled()) {
             return &module_patchers[i];
         }
 #endif
@@ -175,7 +176,7 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
 
     // Enable direct memory mapping in case of NCE.
     const u64 fastmem_base = [&]() -> size_t {
-        if (Settings::IsNceEnabled()) {
+        if (is_application && Settings::IsNceEnabled()) {
             auto& buffer = system.DeviceMemory().buffer;
             buffer.EnableDirectMappedAddress();
             return reinterpret_cast<u64>(buffer.VirtualBasePointer());

src/core/memory.cpp

@@ -45,7 +45,13 @@ struct Memory::Impl {
 
     void SetCurrentPageTable(Kernel::KProcess& process) {
         current_page_table = &process.GetPageTable().GetImpl();
-        current_page_table->fastmem_arena = system.DeviceMemory().buffer.VirtualBasePointer();
+
+        if (std::addressof(process) == system.ApplicationProcess() &&
+            Settings::IsFastmemEnabled()) {
+            current_page_table->fastmem_arena = system.DeviceMemory().buffer.VirtualBasePointer();
+        } else {
+            current_page_table->fastmem_arena = nullptr;
+        }
     }
 
     void MapMemoryRegion(Common::PageTable& page_table, Common::ProcessAddress base, u64 size,
@@ -57,7 +63,7 @@ struct Memory::Impl {
         MapPages(page_table, base / YUZU_PAGESIZE, size / YUZU_PAGESIZE, target,
                  Common::PageType::Memory);
 
-        if (Settings::IsFastmemEnabled()) {
+        if (current_page_table->fastmem_arena) {
             system.DeviceMemory().buffer.Map(GetInteger(base),
                                              GetInteger(target) - DramMemoryMap::Base, size, perms);
         }
@@ -69,7 +75,7 @@ struct Memory::Impl {
         MapPages(page_table, base / YUZU_PAGESIZE, size / YUZU_PAGESIZE, 0,
                  Common::PageType::Unmapped);
 
-        if (Settings::IsFastmemEnabled()) {
+        if (current_page_table->fastmem_arena) {
             system.DeviceMemory().buffer.Unmap(GetInteger(base), size);
         }
     }
@@ -79,7 +85,7 @@ struct Memory::Impl {
         ASSERT_MSG((size & YUZU_PAGEMASK) == 0, "non-page aligned size: {:016X}", size);
         ASSERT_MSG((vaddr & YUZU_PAGEMASK) == 0, "non-page aligned base: {:016X}", vaddr);
 
-        if (!Settings::IsFastmemEnabled()) {
+        if (!current_page_table->fastmem_arena) {
             return;
         }
 
@@ -88,11 +94,6 @@ struct Memory::Impl {
         const bool is_x =
             True(perms & Common::MemoryPermission::Execute) && Settings::IsNceEnabled();
 
-        if (!current_page_table) {
-            system.DeviceMemory().buffer.Protect(vaddr, size, is_r, is_w, is_x);
-            return;
-        }
-
         u64 protect_bytes{};
         u64 protect_begin{};
         for (u64 addr = vaddr; addr < vaddr + size; addr += YUZU_PAGESIZE) {
@@ -239,7 +240,7 @@ struct Memory::Impl {
 
     bool WalkBlock(const Common::ProcessAddress addr, const std::size_t size, auto on_unmapped,
                    auto on_memory, auto on_rasterizer, auto increment) {
-        const auto& page_table = system.ApplicationProcess()->GetPageTable().GetImpl();
+        const auto& page_table = *current_page_table;
         std::size_t remaining_size = size;
         std::size_t page_index = addr >> YUZU_PAGEBITS;
         std::size_t page_offset = addr & YUZU_PAGEMASK;
@@ -484,7 +485,7 @@ struct Memory::Impl {
             return;
         }
 
-        if (Settings::IsFastmemEnabled()) {
+        if (current_page_table->fastmem_arena) {
             system.DeviceMemory().buffer.Protect(vaddr, size, !debug, !debug);
         }
 
@@ -541,7 +542,7 @@ struct Memory::Impl {
             return;
         }
 
-        if (Settings::IsFastmemEnabled()) {
+        if (current_page_table->fastmem_arena) {
             const bool is_read_enable =
                 !Settings::values.use_reactive_flushing.GetValue() || !cached;
             system.DeviceMemory().buffer.Protect(vaddr, size, is_read_enable, !cached);
@@ -886,8 +887,7 @@ void Memory::ProtectRegion(Common::PageTable& page_table, Common::ProcessAddress
 }
 
 bool Memory::IsValidVirtualAddress(const Common::ProcessAddress vaddr) const {
-    const Kernel::KProcess& process = *system.ApplicationProcess();
+    const auto& page_table = *impl->current_page_table;
-    const auto& page_table = process.GetPageTable().GetImpl();
     const size_t page = vaddr >> YUZU_PAGEBITS;
     if (page >= page_table.pointers.size()) {
         return false;

src/video_core/engines/maxwell_3d.cpp

@@ -586,14 +586,22 @@ void Maxwell3D::ProcessQueryCondition() {
 }
 
 void Maxwell3D::ProcessCounterReset() {
-    switch (regs.clear_report_value) {
+    const auto query_type = [clear_report = regs.clear_report_value]() {
-    case Regs::ClearReport::ZPassPixelCount:
+        switch (clear_report) {
-        rasterizer->ResetCounter(VideoCommon::QueryType::ZPassPixelCount64);
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::ZPassPixelCount:
-        break;
+            return VideoCommon::QueryType::ZPassPixelCount64;
-    default:
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::StreamingPrimitivesSucceeded:
-        LOG_DEBUG(Render_OpenGL, "Unimplemented counter reset={}", regs.clear_report_value);
+            return VideoCommon::QueryType::StreamingPrimitivesSucceeded;
-        break;
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::PrimitivesGenerated:
-    }
+            return VideoCommon::QueryType::PrimitivesGenerated;
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::VtgPrimitivesOut:
+            return VideoCommon::QueryType::VtgPrimitivesOut;
+        default:
+            LOG_DEBUG(HW_GPU, "Unimplemented counter reset={}", clear_report);
+            return VideoCommon::QueryType::Payload;
+        }
+    }();
+    rasterizer->ResetCounter(query_type);
 }
 
 void Maxwell3D::ProcessSyncPoint() {

src/video_core/query_cache.h

@@ -28,8 +28,11 @@
 namespace VideoCore {
 enum class QueryType {
     SamplesPassed,
+    PrimitivesGenerated,
+    TfbPrimitivesWritten,
+    Count,
 };
-constexpr std::size_t NumQueryTypes = 1;
+constexpr std::size_t NumQueryTypes = static_cast<size_t>(QueryType::Count);
 } // namespace VideoCore
 
 namespace VideoCommon {
@@ -44,15 +47,6 @@ public:
     explicit CounterStreamBase(QueryCache& cache_, VideoCore::QueryType type_)
         : cache{cache_}, type{type_} {}
 
-    /// Updates the state of the stream, enabling or disabling as needed.
-    void Update(bool enabled) {
-        if (enabled) {
-            Enable();
-        } else {
-            Disable();
-        }
-    }
-
     /// Resets the stream to zero. It doesn't disable the query after resetting.
     void Reset() {
         if (current) {
@@ -80,7 +74,6 @@ public:
         return current != nullptr;
     }
 
-private:
     /// Enables the stream.
     void Enable() {
         if (current) {
@@ -97,6 +90,7 @@ private:
         last = std::exchange(current, nullptr);
     }
 
+private:
     QueryCache& cache;
     const VideoCore::QueryType type;
 
@@ -112,8 +106,14 @@ public:
         : rasterizer{rasterizer_},
           // Use reinterpret_cast instead of static_cast as workaround for
           // UBSan bug (https://github.com/llvm/llvm-project/issues/59060)
-          cpu_memory{cpu_memory_}, streams{{CounterStream{reinterpret_cast<QueryCache&>(*this),
+          cpu_memory{cpu_memory_}, streams{{
-                                                          VideoCore::QueryType::SamplesPassed}}} {
+                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::SamplesPassed}},
+                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::PrimitivesGenerated}},
+                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::TfbPrimitivesWritten}},
+                                   }} {
         (void)slot_async_jobs.insert(); // Null value
     }
 
@@ -157,12 +157,11 @@ public:
         AsyncFlushQuery(query, timestamp, lock);
     }
 
-    /// Updates counters from GPU state. Expected to be called once per draw, clear or dispatch.
+    /// Enables all available GPU counters
-    void UpdateCounters() {
+    void EnableCounters() {
         std::unique_lock lock{mutex};
-        if (maxwell3d) {
+        for (auto& stream : streams) {
-            const auto& regs = maxwell3d->regs;
+            stream.Enable();
-            Stream(VideoCore::QueryType::SamplesPassed).Update(regs.zpass_pixel_count_enable);
         }
     }
 
@@ -176,7 +175,7 @@ public:
     void DisableStreams() {
         std::unique_lock lock{mutex};
         for (auto& stream : streams) {
-            stream.Update(false);
+            stream.Disable();
         }
     }
 
@@ -353,7 +352,7 @@ private:
 
     std::shared_ptr<std::vector<AsyncJobId>> uncommitted_flushes{};
     std::list<std::shared_ptr<std::vector<AsyncJobId>>> committed_flushes;
-};
+}; // namespace VideoCommon
 
 template <class QueryCache, class HostCounter>
 class HostCounterBase {

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -58,6 +58,9 @@ Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rast
         glObjectLabel(GL_BUFFER, buffer.handle, static_cast<GLsizei>(name.size()), name.data());
     }
     glNamedBufferData(buffer.handle, SizeBytes(), nullptr, GL_DYNAMIC_DRAW);
+    if (runtime.has_unified_vertex_buffers) {
+        glGetNamedBufferParameterui64vNV(buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, &address);
+    }
 }
 
 void Buffer::ImmediateUpload(size_t offset, std::span<const u8> data) noexcept {
@@ -109,6 +112,7 @@ BufferCacheRuntime::BufferCacheRuntime(const Device& device_,
     : device{device_}, staging_buffer_pool{staging_buffer_pool_},
       has_fast_buffer_sub_data{device.HasFastBufferSubData()},
       use_assembly_shaders{device.UseAssemblyShaders()},
+      has_unified_vertex_buffers{device.HasVertexBufferUnifiedMemory()},
       stream_buffer{has_fast_buffer_sub_data ? std::nullopt : std::make_optional<StreamBuffer>()} {
     GLint gl_max_attributes;
     glGetIntegerv(GL_MAX_VERTEX_ATTRIBS, &gl_max_attributes);
@@ -210,8 +214,14 @@ void BufferCacheRuntime::ClearBuffer(Buffer& dest_buffer, u32 offset, size_t siz
 }
 
 void BufferCacheRuntime::BindIndexBuffer(Buffer& buffer, u32 offset, u32 size) {
-    glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, buffer.Handle());
+    if (has_unified_vertex_buffers) {
-    index_buffer_offset = offset;
+        buffer.MakeResident(GL_READ_ONLY);
+        glBufferAddressRangeNV(GL_ELEMENT_ARRAY_ADDRESS_NV, 0, buffer.HostGpuAddr() + offset,
+                               static_cast<GLsizeiptr>(Common::AlignUp(size, 4)));
+    } else {
+        glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, buffer.Handle());
+        index_buffer_offset = offset;
+    }
 }
 
 void BufferCacheRuntime::BindVertexBuffer(u32 index, Buffer& buffer, u32 offset, u32 size,
@@ -219,8 +229,15 @@ void BufferCacheRuntime::BindVertexBuffer(u32 index, Buffer& buffer, u32 offset,
     if (index >= max_attributes) {
         return;
     }
-    glBindVertexBuffer(index, buffer.Handle(), static_cast<GLintptr>(offset),
+    if (has_unified_vertex_buffers) {
-                       static_cast<GLsizei>(stride));
+        buffer.MakeResident(GL_READ_ONLY);
+        glBindVertexBuffer(index, 0, 0, static_cast<GLsizei>(stride));
+        glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, index,
+                               buffer.HostGpuAddr() + offset, static_cast<GLsizeiptr>(size));
+    } else {
+        glBindVertexBuffer(index, buffer.Handle(), static_cast<GLintptr>(offset),
+                           static_cast<GLsizei>(stride));
+    }
 }
 
 void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bindings) {
@@ -233,9 +250,23 @@ void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bi
                            [](u64 stride) { return static_cast<GLsizei>(stride); });
     const u32 count =
         std::min(static_cast<u32>(bindings.buffers.size()), max_attributes - bindings.min_index);
-    glBindVertexBuffers(bindings.min_index, static_cast<GLsizei>(count), buffer_handles.data(),
+    if (has_unified_vertex_buffers) {
-                        reinterpret_cast<const GLintptr*>(bindings.offsets.data()),
+        for (u32 index = 0; index < count; ++index) {
-                        buffer_strides.data());
+            Buffer& buffer = *bindings.buffers[index];
+            buffer.MakeResident(GL_READ_ONLY);
+            glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, bindings.min_index + index,
+                                   buffer.HostGpuAddr() + bindings.offsets[index],
+                                   static_cast<GLsizeiptr>(bindings.sizes[index]));
+        }
+        static constexpr std::array<size_t, 32> ZEROS{};
+        glBindVertexBuffers(bindings.min_index, static_cast<GLsizei>(count),
+                            reinterpret_cast<const GLuint*>(ZEROS.data()),
+                            reinterpret_cast<const GLintptr*>(ZEROS.data()), buffer_strides.data());
+    } else {
+        glBindVertexBuffers(bindings.min_index, static_cast<GLsizei>(count), buffer_handles.data(),
+                            reinterpret_cast<const GLintptr*>(bindings.offsets.data()),
+                            buffer_strides.data());
+    }
 }
 
 void BufferCacheRuntime::BindUniformBuffer(size_t stage, u32 binding_index, Buffer& buffer,

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -209,6 +209,7 @@ private:
 
     bool has_fast_buffer_sub_data = false;
     bool use_assembly_shaders = false;
+    bool has_unified_vertex_buffers = false;
 
     bool use_storage_buffers = false;
 

src/video_core/renderer_opengl/gl_device.cpp

@@ -200,6 +200,7 @@ Device::Device(Core::Frontend::EmuWindow& emu_window) {
     has_broken_texture_view_formats = is_amd || (!is_linux && is_intel);
     has_nv_viewport_array2 = GLAD_GL_NV_viewport_array2;
     has_derivative_control = GLAD_GL_ARB_derivative_control;
+    has_vertex_buffer_unified_memory = GLAD_GL_NV_vertex_buffer_unified_memory;
     has_debugging_tool_attached = IsDebugToolAttached(extensions);
     has_depth_buffer_float = HasExtension(extensions, "GL_NV_depth_buffer_float");
     has_geometry_shader_passthrough = GLAD_GL_NV_geometry_shader_passthrough;

src/video_core/renderer_opengl/gl_device.h

@@ -72,6 +72,10 @@ public:
         return has_texture_shadow_lod;
     }
 
+    bool HasVertexBufferUnifiedMemory() const {
+        return has_vertex_buffer_unified_memory;
+    }
+
     bool HasASTC() const {
         return has_astc;
     }
@@ -211,6 +215,7 @@ private:
     bool has_vertex_viewport_layer{};
     bool has_image_load_formatted{};
     bool has_texture_shadow_lod{};
+    bool has_vertex_buffer_unified_memory{};
     bool has_astc{};
     bool has_variable_aoffi{};
     bool has_component_indexing_bug{};

src/video_core/renderer_opengl/gl_query_cache.cpp

@@ -18,16 +18,27 @@ namespace OpenGL {
 
 namespace {
 
-constexpr std::array<GLenum, VideoCore::NumQueryTypes> QueryTargets = {GL_SAMPLES_PASSED};
-
 constexpr GLenum GetTarget(VideoCore::QueryType type) {
-    return QueryTargets[static_cast<std::size_t>(type)];
+    switch (type) {
+    case VideoCore::QueryType::SamplesPassed:
+        return GL_SAMPLES_PASSED;
+    case VideoCore::QueryType::PrimitivesGenerated:
+        return GL_PRIMITIVES_GENERATED;
+    case VideoCore::QueryType::TfbPrimitivesWritten:
+        return GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN;
+    default:
+        break;
+    }
+    UNIMPLEMENTED_MSG("Query type {}", type);
+    return 0;
 }
 
 } // Anonymous namespace
 
 QueryCache::QueryCache(RasterizerOpenGL& rasterizer_, Core::Memory::Memory& cpu_memory_)
-    : QueryCacheLegacy(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {}
+    : QueryCacheLegacy(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {
+    EnableCounters();
+}
 
 QueryCache::~QueryCache() = default;
 
@@ -103,13 +114,13 @@ u64 CachedQuery::Flush([[maybe_unused]] bool async) {
     auto& stream = cache->Stream(type);
     const bool slice_counter = WaitPending() && stream.IsEnabled();
     if (slice_counter) {
-        stream.Update(false);
+        stream.Disable();
     }
 
     auto result = VideoCommon::CachedQueryBase<HostCounter>::Flush();
 
     if (slice_counter) {
-        stream.Update(true);
+        stream.Enable();
     }
 
     return result;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -51,6 +51,22 @@ constexpr size_t NUM_SUPPORTED_VERTEX_ATTRIBUTES = 16;
 void oglEnable(GLenum cap, bool state) {
     (state ? glEnable : glDisable)(cap);
 }
+
+std::optional<VideoCore::QueryType> MaxwellToVideoCoreQuery(VideoCommon::QueryType type) {
+    switch (type) {
+    case VideoCommon::QueryType::PrimitivesGenerated:
+    case VideoCommon::QueryType::VtgPrimitivesOut:
+        return VideoCore::QueryType::PrimitivesGenerated;
+    case VideoCommon::QueryType::ZPassPixelCount64:
+        return VideoCore::QueryType::SamplesPassed;
+    case VideoCommon::QueryType::StreamingPrimitivesSucceeded:
+        // case VideoCommon::QueryType::StreamingByteCount:
+        // TODO: StreamingByteCount = StreamingPrimitivesSucceeded * num_verts * vert_stride
+        return VideoCore::QueryType::TfbPrimitivesWritten;
+    default:
+        return std::nullopt;
+    }
+}
 } // Anonymous namespace
 
 RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_,
@@ -162,14 +178,18 @@ void RasterizerOpenGL::Clear(u32 layer_count) {
         SyncFramebufferSRGB();
     }
     if (regs.clear_surface.Z) {
-        ASSERT_MSG(regs.zeta_enable != 0, "Tried to clear Z but buffer is not enabled!");
+        if (regs.zeta_enable != 0) {
+            LOG_DEBUG(Render_OpenGL, "Tried to clear Z but buffer is not enabled!");
+        }
         use_depth = true;
 
         state_tracker.NotifyDepthMask();
         glDepthMask(GL_TRUE);
     }
     if (regs.clear_surface.S) {
-        ASSERT_MSG(regs.zeta_enable, "Tried to clear stencil but buffer is not enabled!");
+        if (regs.zeta_enable) {
+            LOG_DEBUG(Render_OpenGL, "Tried to clear stencil but buffer is not enabled!");
+        }
         use_stencil = true;
     }
 
@@ -212,7 +232,6 @@ void RasterizerOpenGL::PrepareDraw(bool is_indexed, Func&& draw_func) {
 
     SCOPE_EXIT({ gpu.TickWork(); });
     gpu_memory->FlushCaching();
-    query_cache.UpdateCounters();
 
     GraphicsPipeline* const pipeline{shader_cache.CurrentGraphicsPipeline()};
     if (!pipeline) {
@@ -330,7 +349,6 @@ void RasterizerOpenGL::DrawTexture() {
     MICROPROFILE_SCOPE(OpenGL_Drawing);
 
     SCOPE_EXIT({ gpu.TickWork(); });
-    query_cache.UpdateCounters();
 
     texture_cache.SynchronizeGraphicsDescriptors();
     texture_cache.UpdateRenderTargets(false);
@@ -397,21 +415,28 @@ void RasterizerOpenGL::DispatchCompute() {
 }
 
 void RasterizerOpenGL::ResetCounter(VideoCommon::QueryType type) {
-    if (type == VideoCommon::QueryType::ZPassPixelCount64) {
+    const auto query_cache_type = MaxwellToVideoCoreQuery(type);
-        query_cache.ResetCounter(VideoCore::QueryType::SamplesPassed);
+    if (!query_cache_type.has_value()) {
+        UNIMPLEMENTED_MSG("Reset query type: {}", type);
+        return;
     }
+    query_cache.ResetCounter(*query_cache_type);
 }
 
 void RasterizerOpenGL::Query(GPUVAddr gpu_addr, VideoCommon::QueryType type,
                              VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport) {
-    if (type == VideoCommon::QueryType::ZPassPixelCount64) {
+    const auto query_cache_type = MaxwellToVideoCoreQuery(type);
-        if (True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout)) {
+    if (!query_cache_type.has_value()) {
-            query_cache.Query(gpu_addr, VideoCore::QueryType::SamplesPassed, {gpu.GetTicks()});
+        return QueryFallback(gpu_addr, type, flags, payload, subreport);
-        } else {
-            query_cache.Query(gpu_addr, VideoCore::QueryType::SamplesPassed, std::nullopt);
-        }
-        return;
     }
+    const bool has_timeout = True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout);
+    const auto timestamp = has_timeout ? std::optional<u64>{gpu.GetTicks()} : std::nullopt;
+    query_cache.Query(gpu_addr, *query_cache_type, timestamp);
+}
+
+void RasterizerOpenGL::QueryFallback(GPUVAddr gpu_addr, VideoCommon::QueryType type,
+                                     VideoCommon::QueryPropertiesFlags flags, u32 payload,
+                                     u32 subreport) {
     if (type != VideoCommon::QueryType::Payload) {
         payload = 1u;
     }
@@ -1294,15 +1319,13 @@ void RasterizerOpenGL::BeginTransformFeedback(GraphicsPipeline* program, GLenum
     program->ConfigureTransformFeedback();
 
     UNIMPLEMENTED_IF(regs.IsShaderConfigEnabled(Maxwell::ShaderType::TessellationInit) ||
-                     regs.IsShaderConfigEnabled(Maxwell::ShaderType::Tessellation) ||
+                     regs.IsShaderConfigEnabled(Maxwell::ShaderType::Tessellation));
-                     regs.IsShaderConfigEnabled(Maxwell::ShaderType::Geometry));
-    UNIMPLEMENTED_IF(primitive_mode != GL_POINTS);
 
     // We may have to call BeginTransformFeedbackNV here since they seem to call different
     // implementations on Nvidia's driver (the pointer is different) but we are using
     // ARB_transform_feedback3 features with NV_transform_feedback interactions and the ARB
     // extension doesn't define BeginTransformFeedback (without NV) interactions. It just works.
-    glBeginTransformFeedback(GL_POINTS);
+    glBeginTransformFeedback(primitive_mode);
 }
 
 void RasterizerOpenGL::EndTransformFeedback() {

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -225,6 +225,9 @@ private:
     /// End a transform feedback
     void EndTransformFeedback();
 
+    void QueryFallback(GPUVAddr gpu_addr, VideoCommon::QueryType type,
+                       VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport);
+
     Tegra::GPU& gpu;
 
     const Device& device;

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -168,6 +168,14 @@ RendererOpenGL::RendererOpenGL(Core::TelemetrySession& telemetry_session_,
     if (!GLAD_GL_ARB_seamless_cubemap_per_texture && !GLAD_GL_AMD_seamless_cubemap_per_texture) {
         glEnable(GL_TEXTURE_CUBE_MAP_SEAMLESS);
     }
+    // Enable unified vertex attributes and query vertex buffer address when the driver supports it
+    if (device.HasVertexBufferUnifiedMemory()) {
+        glEnableClientState(GL_VERTEX_ATTRIB_ARRAY_UNIFIED_NV);
+        glEnableClientState(GL_ELEMENT_ARRAY_UNIFIED_NV);
+        glMakeNamedBufferResidentNV(vertex_buffer.handle, GL_READ_ONLY);
+        glGetNamedBufferParameterui64vNV(vertex_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV,
+                                         &vertex_buffer_address);
+    }
 }
 
 RendererOpenGL::~RendererOpenGL() = default;
@@ -667,7 +675,13 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
                          offsetof(ScreenRectVertex, tex_coord));
     glVertexAttribBinding(PositionLocation, 0);
     glVertexAttribBinding(TexCoordLocation, 0);
-    glBindVertexBuffer(0, vertex_buffer.handle, 0, sizeof(ScreenRectVertex));
+    if (device.HasVertexBufferUnifiedMemory()) {
+        glBindVertexBuffer(0, 0, 0, sizeof(ScreenRectVertex));
+        glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, 0, vertex_buffer_address,
+                               sizeof(vertices));
+    } else {
+        glBindVertexBuffer(0, vertex_buffer.handle, 0, sizeof(ScreenRectVertex));
+    }
 
     if (Settings::values.scaling_filter.GetValue() != Settings::ScalingFilter::NearestNeighbor) {
         glBindSampler(0, present_sampler.handle);

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -485,6 +485,10 @@ void RasterizerVulkan::DispatchCompute() {
 }
 
 void RasterizerVulkan::ResetCounter(VideoCommon::QueryType type) {
+    if (type != VideoCommon::QueryType::ZPassPixelCount64) {
+        LOG_DEBUG(Render_Vulkan, "Unimplemented counter reset={}", type);
+        return;
+    }
     query_cache.CounterReset(type);
 }