Android 187

* Fix typos in src/core

* Fix typo correction

* Fix indentation of MemoryStateNames

* Fix indent

README.md

@@ -1,3 +1,12 @@
+| Pull Request | Commit | Title | Author | Merged? |
+|----|----|----|----|----|
+| [12579](https://github.com/yuzu-emu/yuzu-android//pull/12579) | [`66ae60a9e`](https://github.com/yuzu-emu/yuzu-android//pull/12579/files) | Core: Implement Device Mapping & GPU SMMU | [FernandoS27](https://github.com/FernandoS27/) | Yes |
+
+
+End of merge log. You can find the original README.md below the break.
+
+-----
+
 <!--
 SPDX-FileCopyrightText: 2018 yuzu Emulator Project
 SPDX-License-Identifier: GPL-2.0-or-later

src/audio_core/device/device_session.cpp

@@ -8,6 +8,7 @@
 #include "audio_core/sink/sink_stream.h"
 #include "core/core.h"
 #include "core/core_timing.h"
+#include "core/guest_memory.h"
 #include "core/memory.h"
 
 namespace AudioCore {

src/audio_core/renderer/command/data_source/decode.cpp

@@ -9,6 +9,7 @@
 #include "common/fixed_point.h"
 #include "common/logging/log.h"
 #include "common/scratch_buffer.h"
+#include "core/guest_memory.h"
 #include "core/memory.h"
 
 namespace AudioCore::Renderer {

src/common/common_types.h

@@ -45,6 +45,7 @@ using f32 = float;  ///< 32-bit floating point
 using f64 = double; ///< 64-bit floating point
 
 using VAddr = u64;    ///< Represents a pointer in the userspace virtual address space.
+using DAddr = u64;    ///< Represents a pointer in the device specific virtual address space.
 using PAddr = u64;    ///< Represents a pointer in the ARM11 physical address space.
 using GPUVAddr = u64; ///< Represents a pointer in the GPU virtual address space.
 

src/core/CMakeLists.txt

@@ -37,6 +37,8 @@ add_library(core STATIC
     debugger/gdbstub_arch.h
     debugger/gdbstub.cpp
     debugger/gdbstub.h
+    device_memory_manager.h
+    device_memory_manager.inc
     device_memory.cpp
     device_memory.h
     file_sys/fssystem/fs_i_storage.h
@@ -605,6 +607,8 @@ add_library(core STATIC
     hle/service/ns/pdm_qry.h
     hle/service/nvdrv/core/container.cpp
     hle/service/nvdrv/core/container.h
+    hle/service/nvdrv/core/heap_mapper.cpp
+    hle/service/nvdrv/core/heap_mapper.h
     hle/service/nvdrv/core/nvmap.cpp
     hle/service/nvdrv/core/nvmap.h
     hle/service/nvdrv/core/syncpoint_manager.cpp
@@ -708,24 +712,24 @@ add_library(core STATIC
     hle/service/server_manager.h
     hle/service/service.cpp
     hle/service/service.h
-    hle/service/set/set.cpp
-    hle/service/set/set.h
     hle/service/set/appln_settings.cpp
     hle/service/set/appln_settings.h
     hle/service/set/device_settings.cpp
     hle/service/set/device_settings.h
+    hle/service/set/factory_settings_server.cpp
+    hle/service/set/factory_settings_server.h
+    hle/service/set/firmware_debug_settings_server.cpp
+    hle/service/set/firmware_debug_settings_server.h
     hle/service/set/private_settings.cpp
     hle/service/set/private_settings.h
-    hle/service/set/set_cal.cpp
-    hle/service/set/set_cal.h
-    hle/service/set/set_fd.cpp
-    hle/service/set/set_fd.h
-    hle/service/set/set_sys.cpp
-    hle/service/set/set_sys.h
     hle/service/set/settings.cpp
     hle/service/set/settings.h
+    hle/service/set/settings_server.cpp
+    hle/service/set/settings_server.h
     hle/service/set/system_settings.cpp
     hle/service/set/system_settings.h
+    hle/service/set/system_settings_server.cpp
+    hle/service/set/system_settings_server.h
     hle/service/sm/sm.cpp
     hle/service/sm/sm.h
     hle/service/sm/sm_controller.cpp

src/core/core.cpp

@@ -28,6 +28,7 @@
 #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/hle/kernel/k_memory_manager.h"
 #include "core/hle/kernel/k_process.h"
 #include "core/hle/kernel/k_resource_limit.h"
@@ -564,6 +565,9 @@ 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_managers;
+
     std::deque<std::vector<u8>> user_channel;
 };
 
@@ -650,8 +654,14 @@ size_t System::GetCurrentHostThreadID() const {
     return impl->kernel.GetCurrentHostThreadID();
 }
 
-void System::GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback) {
-    return this->ApplicationProcess()->GatherGPUDirtyMemory(callback);
+std::span<GPUDirtyMemoryManager> System::GetGPUDirtyMemoryManager() {
+    return impl->gpu_dirty_memory_managers;
+}
+
+void System::GatherGPUDirtyMemory(std::function<void(PAddr, size_t)>& callback) {
+    for (auto& manager : impl->gpu_dirty_memory_managers) {
+        manager.Gather(callback);
+    }
 }
 
 PerfStatsResults System::GetAndResetPerfStats() {

src/core/core.h

@@ -8,6 +8,7 @@
 #include <functional>
 #include <memory>
 #include <mutex>
+#include <span>
 #include <string>
 #include <vector>
 
@@ -116,6 +117,7 @@ class CpuManager;
 class Debugger;
 class DeviceMemory;
 class ExclusiveMonitor;
+class GPUDirtyMemoryManager;
 class PerfStats;
 class Reporter;
 class SpeedLimiter;
@@ -224,7 +226,9 @@ public:
     /// Prepare the core emulation for a reschedule
     void PrepareReschedule(u32 core_index);
 
-    void GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback);
+    std::span<GPUDirtyMemoryManager> GetGPUDirtyMemoryManager();
+
+    void GatherGPUDirtyMemory(std::function<void(PAddr, size_t)>& callback);
 
     [[nodiscard]] size_t GetCurrentHostThreadID() const;
 

src/core/cpu_manager.h

@@ -64,7 +64,7 @@ public:
         return [this] { ShutdownThreadFunction(); };
     }
 
-    void PreemptSingleCore(bool from_running_enviroment = true);
+    void PreemptSingleCore(bool from_running_environment = true);
 
     std::size_t CurrentCore() const {
         return current_core.load();

src/core/debugger/gdbstub.cpp

@@ -559,28 +559,28 @@ void GDBStub::HandleVCont(std::string_view command, std::vector<DebuggerAction>&
 }
 
 constexpr std::array<std::pair<const char*, Kernel::Svc::MemoryState>, 22> MemoryStateNames{{
-    {"----- Free -----", Kernel::Svc::MemoryState::Free},
-    {"Io              ", Kernel::Svc::MemoryState::Io},
-    {"Static          ", Kernel::Svc::MemoryState::Static},
-    {"Code            ", Kernel::Svc::MemoryState::Code},
-    {"CodeData        ", Kernel::Svc::MemoryState::CodeData},
-    {"Normal          ", Kernel::Svc::MemoryState::Normal},
-    {"Shared          ", Kernel::Svc::MemoryState::Shared},
-    {"AliasCode       ", Kernel::Svc::MemoryState::AliasCode},
-    {"AliasCodeData   ", Kernel::Svc::MemoryState::AliasCodeData},
-    {"Ipc             ", Kernel::Svc::MemoryState::Ipc},
-    {"Stack           ", Kernel::Svc::MemoryState::Stack},
-    {"ThreadLocal     ", Kernel::Svc::MemoryState::ThreadLocal},
-    {"Transfered      ", Kernel::Svc::MemoryState::Transfered},
-    {"SharedTransfered", Kernel::Svc::MemoryState::SharedTransfered},
-    {"SharedCode      ", Kernel::Svc::MemoryState::SharedCode},
-    {"Inaccessible    ", Kernel::Svc::MemoryState::Inaccessible},
-    {"NonSecureIpc    ", Kernel::Svc::MemoryState::NonSecureIpc},
-    {"NonDeviceIpc    ", Kernel::Svc::MemoryState::NonDeviceIpc},
-    {"Kernel          ", Kernel::Svc::MemoryState::Kernel},
-    {"GeneratedCode   ", Kernel::Svc::MemoryState::GeneratedCode},
-    {"CodeOut         ", Kernel::Svc::MemoryState::CodeOut},
-    {"Coverage        ", Kernel::Svc::MemoryState::Coverage},
+    {"----- Free ------", Kernel::Svc::MemoryState::Free},
+    {"Io               ", Kernel::Svc::MemoryState::Io},
+    {"Static           ", Kernel::Svc::MemoryState::Static},
+    {"Code             ", Kernel::Svc::MemoryState::Code},
+    {"CodeData         ", Kernel::Svc::MemoryState::CodeData},
+    {"Normal           ", Kernel::Svc::MemoryState::Normal},
+    {"Shared           ", Kernel::Svc::MemoryState::Shared},
+    {"AliasCode        ", Kernel::Svc::MemoryState::AliasCode},
+    {"AliasCodeData    ", Kernel::Svc::MemoryState::AliasCodeData},
+    {"Ipc              ", Kernel::Svc::MemoryState::Ipc},
+    {"Stack            ", Kernel::Svc::MemoryState::Stack},
+    {"ThreadLocal      ", Kernel::Svc::MemoryState::ThreadLocal},
+    {"Transferred      ", Kernel::Svc::MemoryState::Transferred},
+    {"SharedTransferred", Kernel::Svc::MemoryState::SharedTransferred},
+    {"SharedCode       ", Kernel::Svc::MemoryState::SharedCode},
+    {"Inaccessible     ", Kernel::Svc::MemoryState::Inaccessible},
+    {"NonSecureIpc     ", Kernel::Svc::MemoryState::NonSecureIpc},
+    {"NonDeviceIpc     ", Kernel::Svc::MemoryState::NonDeviceIpc},
+    {"Kernel           ", Kernel::Svc::MemoryState::Kernel},
+    {"GeneratedCode    ", Kernel::Svc::MemoryState::GeneratedCode},
+    {"CodeOut          ", Kernel::Svc::MemoryState::CodeOut},
+    {"Coverage         ", Kernel::Svc::MemoryState::Coverage},
 }};
 
 static constexpr const char* GetMemoryStateName(Kernel::Svc::MemoryState state) {

src/core/device_memory.h

@@ -31,6 +31,12 @@ public:
                DramMemoryMap::Base;
     }
 
+    template <typename T>
+    PAddr GetRawPhysicalAddr(const T* ptr) const {
+        return static_cast<PAddr>(reinterpret_cast<uintptr_t>(ptr) -
+                                  reinterpret_cast<uintptr_t>(buffer.BackingBasePointer()));
+    }
+
     template <typename T>
     T* GetPointer(Common::PhysicalAddress addr) {
         return reinterpret_cast<T*>(buffer.BackingBasePointer() +
@@ -43,6 +49,16 @@ public:
                                     (GetInteger(addr) - DramMemoryMap::Base));
     }
 
+    template <typename T>
+    T* GetPointerFromRaw(PAddr addr) {
+        return reinterpret_cast<T*>(buffer.BackingBasePointer() + addr);
+    }
+
+    template <typename T>
+    const T* GetPointerFromRaw(PAddr addr) const {
+        return reinterpret_cast<T*>(buffer.BackingBasePointer() + addr);
+    }
+
     Common::HostMemory buffer;
 };
 

src/core/device_memory_manager.h

@@ -0,0 +1,208 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <array>
+#include <atomic>
+#include <deque>
+#include <memory>
+#include <mutex>
+
+#include "common/common_types.h"
+#include "common/scratch_buffer.h"
+#include "common/virtual_buffer.h"
+
+namespace Core {
+
+constexpr size_t DEVICE_PAGEBITS = 12ULL;
+constexpr size_t DEVICE_PAGESIZE = 1ULL << DEVICE_PAGEBITS;
+constexpr size_t DEVICE_PAGEMASK = DEVICE_PAGESIZE - 1ULL;
+
+class DeviceMemory;
+
+namespace Memory {
+class Memory;
+}
+
+template <typename DTraits>
+struct DeviceMemoryManagerAllocator;
+
+template <typename Traits>
+class DeviceMemoryManager {
+    using DeviceInterface = typename Traits::DeviceInterface;
+    using DeviceMethods = typename Traits::DeviceMethods;
+
+public:
+    DeviceMemoryManager(const DeviceMemory& device_memory);
+    ~DeviceMemoryManager();
+
+    void BindInterface(DeviceInterface* device_inter);
+
+    DAddr Allocate(size_t size);
+    void AllocateFixed(DAddr start, size_t size);
+    void Free(DAddr start, size_t size);
+
+    void Map(DAddr address, VAddr virtual_address, size_t size, size_t process_id,
+             bool track = false);
+
+    void Unmap(DAddr address, size_t size);
+
+    void TrackContinuityImpl(DAddr address, VAddr virtual_address, size_t size, size_t process_id);
+    void TrackContinuity(DAddr address, VAddr virtual_address, size_t size, size_t process_id) {
+        std::scoped_lock lk(mapping_guard);
+        TrackContinuityImpl(address, virtual_address, size, process_id);
+    }
+
+    // Write / Read
+    template <typename T>
+    T* GetPointer(DAddr address);
+
+    template <typename T>
+    const T* GetPointer(DAddr address) const;
+
+    template <typename Func>
+    void ApplyOpOnPAddr(PAddr address, Common::ScratchBuffer<u32>& buffer, Func&& operation) {
+        DAddr subbits = static_cast<DAddr>(address & page_mask);
+        const u32 base = compressed_device_addr[(address >> page_bits)];
+        if ((base >> MULTI_FLAG_BITS) == 0) [[likely]] {
+            const DAddr d_address = (static_cast<DAddr>(base) << page_bits) + subbits;
+            operation(d_address);
+            return;
+        }
+        InnerGatherDeviceAddresses(buffer, address);
+        for (u32 value : buffer) {
+            operation((static_cast<DAddr>(value) << page_bits) + subbits);
+        }
+    }
+
+    template <typename Func>
+    void ApplyOpOnPointer(const u8* p, Common::ScratchBuffer<u32>& buffer, Func&& operation) {
+        PAddr address = GetRawPhysicalAddr<u8>(p);
+        ApplyOpOnPAddr(address, buffer, operation);
+    }
+
+    PAddr GetPhysicalRawAddressFromDAddr(DAddr address) const {
+        PAddr subbits = static_cast<PAddr>(address & page_mask);
+        auto paddr = compressed_physical_ptr[(address >> page_bits)];
+        if (paddr == 0) {
+            return 0;
+        }
+        return (static_cast<PAddr>(paddr - 1) << page_bits) + subbits;
+    }
+
+    template <typename T>
+    void Write(DAddr address, T value);
+
+    template <typename T>
+    T Read(DAddr address) const;
+
+    u8* GetSpan(const DAddr src_addr, const std::size_t size);
+    const u8* GetSpan(const DAddr src_addr, const std::size_t size) const;
+
+    void ReadBlock(DAddr address, void* dest_pointer, size_t size);
+    void ReadBlockUnsafe(DAddr address, void* dest_pointer, size_t size);
+    void WriteBlock(DAddr address, const void* src_pointer, size_t size);
+    void WriteBlockUnsafe(DAddr address, const void* src_pointer, size_t size);
+
+    size_t RegisterProcess(Memory::Memory* memory);
+    void UnregisterProcess(size_t id);
+
+    void UpdatePagesCachedCount(DAddr addr, size_t size, s32 delta);
+
+    static constexpr size_t AS_BITS = Traits::device_virtual_bits;
+
+private:
+    static constexpr size_t device_virtual_bits = Traits::device_virtual_bits;
+    static constexpr size_t device_as_size = 1ULL << device_virtual_bits;
+    static constexpr size_t physical_min_bits = 32;
+    static constexpr size_t physical_max_bits = 33;
+    static constexpr size_t page_bits = 12;
+    static constexpr size_t page_size = 1ULL << page_bits;
+    static constexpr size_t page_mask = page_size - 1ULL;
+    static constexpr u32 physical_address_base = 1U << page_bits;
+    static constexpr u32 MULTI_FLAG_BITS = 31;
+    static constexpr u32 MULTI_FLAG = 1U << MULTI_FLAG_BITS;
+    static constexpr u32 MULTI_MASK = ~MULTI_FLAG;
+
+    template <typename T>
+    T* GetPointerFromRaw(PAddr addr) {
+        return reinterpret_cast<T*>(physical_base + addr);
+    }
+
+    template <typename T>
+    const T* GetPointerFromRaw(PAddr addr) const {
+        return reinterpret_cast<T*>(physical_base + addr);
+    }
+
+    template <typename T>
+    PAddr GetRawPhysicalAddr(const T* ptr) const {
+        return static_cast<PAddr>(reinterpret_cast<uintptr_t>(ptr) - physical_base);
+    }
+
+    void WalkBlock(const DAddr addr, const std::size_t size, auto on_unmapped, auto on_memory,
+                   auto increment);
+
+    void InnerGatherDeviceAddresses(Common::ScratchBuffer<u32>& buffer, PAddr address);
+
+    std::unique_ptr<DeviceMemoryManagerAllocator<Traits>> impl;
+
+    const uintptr_t physical_base;
+    DeviceInterface* device_inter;
+    Common::VirtualBuffer<u32> compressed_physical_ptr;
+    Common::VirtualBuffer<u32> compressed_device_addr;
+    Common::VirtualBuffer<u32> continuity_tracker;
+
+    // Process memory interfaces
+
+    std::deque<size_t> id_pool;
+    std::deque<Memory::Memory*> registered_processes;
+
+    // Memory protection management
+
+    static constexpr size_t guest_max_as_bits = 39;
+    static constexpr size_t guest_as_size = 1ULL << guest_max_as_bits;
+    static constexpr size_t guest_mask = guest_as_size - 1ULL;
+    static constexpr size_t process_id_start_bit = guest_max_as_bits;
+
+    std::pair<size_t, VAddr> ExtractCPUBacking(size_t page_index) {
+        auto content = cpu_backing_address[page_index];
+        const VAddr address = content & guest_mask;
+        const size_t process_id = static_cast<size_t>(content >> process_id_start_bit);
+        return std::make_pair(process_id, address);
+    }
+
+    void InsertCPUBacking(size_t page_index, VAddr address, size_t process_id) {
+        cpu_backing_address[page_index] = address | (process_id << process_id_start_bit);
+    }
+
+    Common::VirtualBuffer<VAddr> cpu_backing_address;
+    static constexpr size_t subentries = 8 / sizeof(u8);
+    static constexpr size_t subentries_mask = subentries - 1;
+    class CounterEntry final {
+    public:
+        CounterEntry() = default;
+
+        std::atomic_uint8_t& Count(std::size_t page) {
+            return values[page & subentries_mask];
+        }
+
+        const std::atomic_uint8_t& Count(std::size_t page) const {
+            return values[page & subentries_mask];
+        }
+
+    private:
+        std::array<std::atomic_uint8_t, subentries> values{};
+    };
+    static_assert(sizeof(CounterEntry) == subentries * sizeof(u8),
+                  "CounterEntry should be 8 bytes!");
+
+    static constexpr size_t num_counter_entries =
+        (1ULL << (device_virtual_bits - page_bits)) / subentries;
+    using CachedPages = std::array<CounterEntry, num_counter_entries>;
+    std::unique_ptr<CachedPages> cached_pages;
+    std::mutex counter_guard;
+    std::mutex mapping_guard;
+};
+
+} // namespace Core

src/core/device_memory_manager.inc

@@ -0,0 +1,588 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <atomic>
+#include <limits>
+#include <memory>
+#include <type_traits>
+
+#include "common/address_space.h"
+#include "common/address_space.inc"
+#include "common/alignment.h"
+#include "common/assert.h"
+#include "common/div_ceil.h"
+#include "common/scope_exit.h"
+#include "common/settings.h"
+#include "core/device_memory.h"
+#include "core/device_memory_manager.h"
+#include "core/memory.h"
+
+namespace Core {
+
+namespace {
+
+class MultiAddressContainer {
+public:
+    MultiAddressContainer() = default;
+    ~MultiAddressContainer() = default;
+
+    void GatherValues(u32 start_entry, Common::ScratchBuffer<u32>& buffer) {
+        buffer.resize(8);
+        buffer.resize(0);
+        size_t index = 0;
+        const auto add_value = [&](u32 value) {
+            buffer[index] = value;
+            index++;
+            buffer.resize(index);
+        };
+
+        u32 iter_entry = start_entry;
+        Entry* current = &storage[iter_entry - 1];
+        add_value(current->value);
+        while (current->next_entry != 0) {
+            iter_entry = current->next_entry;
+            current = &storage[iter_entry - 1];
+            add_value(current->value);
+        }
+    }
+
+    u32 Register(u32 value) {
+        return RegisterImplementation(value);
+    }
+
+    void Register(u32 value, u32 start_entry) {
+        auto entry_id = RegisterImplementation(value);
+        u32 iter_entry = start_entry;
+        Entry* current = &storage[iter_entry - 1];
+        while (current->next_entry != 0) {
+            iter_entry = current->next_entry;
+            current = &storage[iter_entry - 1];
+        }
+        current->next_entry = entry_id;
+    }
+
+    std::pair<bool, u32> Unregister(u32 value, u32 start_entry) {
+        u32 iter_entry = start_entry;
+        Entry* previous{};
+        Entry* current = &storage[iter_entry - 1];
+        Entry* next{};
+        bool more_than_one_remaining = false;
+        u32 result_start{start_entry};
+        size_t count = 0;
+        while (current->value != value) {
+            count++;
+            previous = current;
+            iter_entry = current->next_entry;
+            current = &storage[iter_entry - 1];
+        }
+        // Find next
+        u32 next_entry = current->next_entry;
+        if (next_entry != 0) {
+            next = &storage[next_entry - 1];
+            more_than_one_remaining = next->next_entry != 0 || previous != nullptr;
+        }
+        if (previous) {
+            previous->next_entry = next_entry;
+        } else {
+            result_start = next_entry;
+        }
+        free_entries.emplace_back(iter_entry);
+        return std::make_pair(more_than_one_remaining || count > 1, result_start);
+    }
+
+    u32 ReleaseEntry(u32 start_entry) {
+        Entry* current = &storage[start_entry - 1];
+        free_entries.emplace_back(start_entry);
+        return current->value;
+    }
+
+private:
+    u32 RegisterImplementation(u32 value) {
+        auto entry_id = GetNewEntry();
+        auto& entry = storage[entry_id - 1];
+        entry.next_entry = 0;
+        entry.value = value;
+        return entry_id;
+    }
+    u32 GetNewEntry() {
+        if (!free_entries.empty()) {
+            u32 result = free_entries.front();
+            free_entries.pop_front();
+            return result;
+        }
+        storage.emplace_back();
+        u32 new_entry = static_cast<u32>(storage.size());
+        return new_entry;
+    }
+
+    struct Entry {
+        u32 next_entry{};
+        u32 value{};
+    };
+
+    std::deque<Entry> storage;
+    std::deque<u32> free_entries;
+};
+
+struct EmptyAllocator {
+    EmptyAllocator([[maybe_unused]] DAddr address) {}
+};
+
+} // namespace
+
+template <typename DTraits>
+struct DeviceMemoryManagerAllocator {
+    static constexpr size_t device_virtual_bits = DTraits::device_virtual_bits;
+    static constexpr DAddr first_address = 1ULL << Memory::YUZU_PAGEBITS;
+    static constexpr DAddr max_device_area = 1ULL << device_virtual_bits;
+
+    DeviceMemoryManagerAllocator() : main_allocator(first_address) {}
+
+    Common::FlatAllocator<DAddr, 0, device_virtual_bits> main_allocator;
+    MultiAddressContainer multi_dev_address;
+
+    /// Returns true when vaddr -> vaddr+size is fully contained in the buffer
+    template <bool pin_area>
+    [[nodiscard]] bool IsInBounds(VAddr addr, u64 size) const noexcept {
+        return addr >= 0 && addr + size <= max_device_area;
+    }
+
+    DAddr Allocate(size_t size) {
+        return main_allocator.Allocate(size);
+    }
+
+    void AllocateFixed(DAddr b_address, size_t b_size) {
+        main_allocator.AllocateFixed(b_address, b_size);
+    }
+
+    void Free(DAddr b_address, size_t b_size) {
+        main_allocator.Free(b_address, b_size);
+    }
+};
+
+template <typename Traits>
+DeviceMemoryManager<Traits>::DeviceMemoryManager(const DeviceMemory& device_memory_)
+    : physical_base{reinterpret_cast<const uintptr_t>(device_memory_.buffer.BackingBasePointer())},
+      device_inter{nullptr}, compressed_physical_ptr(device_as_size >> Memory::YUZU_PAGEBITS),
+      compressed_device_addr(1ULL << ((Settings::values.memory_layout_mode.GetValue() ==
+                                               Settings::MemoryLayout::Memory_4Gb
+                                           ? physical_min_bits
+                                           : physical_max_bits) -
+                                      Memory::YUZU_PAGEBITS)),
+      continuity_tracker(device_as_size >> Memory::YUZU_PAGEBITS),
+      cpu_backing_address(device_as_size >> Memory::YUZU_PAGEBITS) {
+    impl = std::make_unique<DeviceMemoryManagerAllocator<Traits>>();
+    cached_pages = std::make_unique<CachedPages>();
+
+    const size_t total_virtual = device_as_size >> Memory::YUZU_PAGEBITS;
+    for (size_t i = 0; i < total_virtual; i++) {
+        compressed_physical_ptr[i] = 0;
+        continuity_tracker[i] = 1;
+        cpu_backing_address[i] = 0;
+    }
+    const size_t total_phys = 1ULL << ((Settings::values.memory_layout_mode.GetValue() ==
+                                                Settings::MemoryLayout::Memory_4Gb
+                                            ? physical_min_bits
+                                            : physical_max_bits) -
+                                       Memory::YUZU_PAGEBITS);
+    for (size_t i = 0; i < total_phys; i++) {
+        compressed_device_addr[i] = 0;
+    }
+}
+
+template <typename Traits>
+DeviceMemoryManager<Traits>::~DeviceMemoryManager() = default;
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::BindInterface(DeviceInterface* device_inter_) {
+    device_inter = device_inter_;
+}
+
+template <typename Traits>
+DAddr DeviceMemoryManager<Traits>::Allocate(size_t size) {
+    return impl->Allocate(size);
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::AllocateFixed(DAddr start, size_t size) {
+    return impl->AllocateFixed(start, size);
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::Free(DAddr start, size_t size) {
+    impl->Free(start, size);
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::Map(DAddr address, VAddr virtual_address, size_t size,
+                                      size_t process_id, bool track) {
+    Core::Memory::Memory* process_memory = registered_processes[process_id];
+    size_t start_page_d = address >> Memory::YUZU_PAGEBITS;
+    size_t num_pages = Common::AlignUp(size, Memory::YUZU_PAGESIZE) >> Memory::YUZU_PAGEBITS;
+    std::scoped_lock lk(mapping_guard);
+    for (size_t i = 0; i < num_pages; i++) {
+        const VAddr new_vaddress = virtual_address + i * Memory::YUZU_PAGESIZE;
+        auto* ptr = process_memory->GetPointerSilent(Common::ProcessAddress(new_vaddress));
+        if (ptr == nullptr) [[unlikely]] {
+            compressed_physical_ptr[start_page_d + i] = 0;
+            continue;
+        }
+        auto phys_addr = static_cast<u32>(GetRawPhysicalAddr(ptr) >> Memory::YUZU_PAGEBITS) + 1U;
+        compressed_physical_ptr[start_page_d + i] = phys_addr;
+        InsertCPUBacking(start_page_d + i, new_vaddress, process_id);
+        const u32 base_dev = compressed_device_addr[phys_addr - 1U];
+        const u32 new_dev = static_cast<u32>(start_page_d + i);
+        if (base_dev == 0) [[likely]] {
+            compressed_device_addr[phys_addr - 1U] = new_dev;
+            continue;
+        }
+        u32 start_id = base_dev & MULTI_MASK;
+        if ((base_dev >> MULTI_FLAG_BITS) == 0) {
+            start_id = impl->multi_dev_address.Register(base_dev);
+            compressed_device_addr[phys_addr - 1U] = MULTI_FLAG | start_id;
+        }
+        impl->multi_dev_address.Register(new_dev, start_id);
+    }
+    if (track) {
+        TrackContinuityImpl(address, virtual_address, size, process_id);
+    }
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::Unmap(DAddr address, size_t size) {
+    size_t start_page_d = address >> Memory::YUZU_PAGEBITS;
+    size_t num_pages = Common::AlignUp(size, Memory::YUZU_PAGESIZE) >> Memory::YUZU_PAGEBITS;
+    device_inter->InvalidateRegion(address, size);
+    std::scoped_lock lk(mapping_guard);
+    for (size_t i = 0; i < num_pages; i++) {
+        auto phys_addr = compressed_physical_ptr[start_page_d + i];
+        compressed_physical_ptr[start_page_d + i] = 0;
+        cpu_backing_address[start_page_d + i] = 0;
+        if (phys_addr != 0) [[likely]] {
+            const u32 base_dev = compressed_device_addr[phys_addr - 1U];
+            if ((base_dev >> MULTI_FLAG_BITS) == 0) [[likely]] {
+                compressed_device_addr[phys_addr - 1] = 0;
+                continue;
+            }
+            const auto [more_entries, new_start] = impl->multi_dev_address.Unregister(
+                static_cast<u32>(start_page_d + i), base_dev & MULTI_MASK);
+            if (!more_entries) {
+                compressed_device_addr[phys_addr - 1] =
+                    impl->multi_dev_address.ReleaseEntry(new_start);
+                continue;
+            }
+            compressed_device_addr[phys_addr - 1] = new_start | MULTI_FLAG;
+        }
+    }
+}
+template <typename Traits>
+void DeviceMemoryManager<Traits>::TrackContinuityImpl(DAddr address, VAddr virtual_address,
+                                                      size_t size, size_t process_id) {
+    Core::Memory::Memory* process_memory = registered_processes[process_id];
+    size_t start_page_d = address >> Memory::YUZU_PAGEBITS;
+    size_t num_pages = Common::AlignUp(size, Memory::YUZU_PAGESIZE) >> Memory::YUZU_PAGEBITS;
+    uintptr_t last_ptr = 0;
+    size_t page_count = 1;
+    for (size_t i = num_pages; i > 0; i--) {
+        size_t index = i - 1;
+        const VAddr new_vaddress = virtual_address + index * Memory::YUZU_PAGESIZE;
+        const uintptr_t new_ptr = reinterpret_cast<uintptr_t>(
+            process_memory->GetPointerSilent(Common::ProcessAddress(new_vaddress)));
+        if (new_ptr + page_size == last_ptr) {
+            page_count++;
+        } else {
+            page_count = 1;
+        }
+        last_ptr = new_ptr;
+        continuity_tracker[start_page_d + index] = static_cast<u32>(page_count);
+    }
+}
+template <typename Traits>
+u8* DeviceMemoryManager<Traits>::GetSpan(const DAddr src_addr, const std::size_t size) {
+    size_t page_index = src_addr >> page_bits;
+    size_t subbits = src_addr & page_mask;
+    if ((static_cast<size_t>(continuity_tracker[page_index]) << page_bits) >= size + subbits) {
+        return GetPointer<u8>(src_addr);
+    }
+    return nullptr;
+}
+
+template <typename Traits>
+const u8* DeviceMemoryManager<Traits>::GetSpan(const DAddr src_addr, const std::size_t size) const {
+    size_t page_index = src_addr >> page_bits;
+    size_t subbits = src_addr & page_mask;
+    if ((static_cast<size_t>(continuity_tracker[page_index]) << page_bits) >= size + subbits) {
+        return GetPointer<u8>(src_addr);
+    }
+    return nullptr;
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::InnerGatherDeviceAddresses(Common::ScratchBuffer<u32>& buffer,
+                                                             PAddr address) {
+    size_t phys_addr = address >> page_bits;
+    std::scoped_lock lk(mapping_guard);
+    u32 backing = compressed_device_addr[phys_addr];
+    if ((backing >> MULTI_FLAG_BITS) != 0) {
+        impl->multi_dev_address.GatherValues(backing & MULTI_MASK, buffer);
+        return;
+    }
+    buffer.resize(1);
+    buffer[0] = backing;
+}
+
+template <typename Traits>
+template <typename T>
+T* DeviceMemoryManager<Traits>::GetPointer(DAddr address) {
+    const size_t index = address >> Memory::YUZU_PAGEBITS;
+    const size_t offset = address & Memory::YUZU_PAGEMASK;
+    auto phys_addr = compressed_physical_ptr[index];
+    if (phys_addr == 0) [[unlikely]] {
+        return nullptr;
+    }
+    return GetPointerFromRaw<T>((static_cast<PAddr>(phys_addr - 1) << Memory::YUZU_PAGEBITS) +
+                                offset);
+}
+
+template <typename Traits>
+template <typename T>
+const T* DeviceMemoryManager<Traits>::GetPointer(DAddr address) const {
+    const size_t index = address >> Memory::YUZU_PAGEBITS;
+    const size_t offset = address & Memory::YUZU_PAGEMASK;
+    auto phys_addr = compressed_physical_ptr[index];
+    if (phys_addr == 0) [[unlikely]] {
+        return nullptr;
+    }
+    return GetPointerFromRaw<T>((static_cast<PAddr>(phys_addr - 1) << Memory::YUZU_PAGEBITS) +
+                                offset);
+}
+
+template <typename Traits>
+template <typename T>
+void DeviceMemoryManager<Traits>::Write(DAddr address, T value) {
+    T* ptr = GetPointer<T>(address);
+    if (!ptr) [[unlikely]] {
+        return;
+    }
+    std::memcpy(ptr, &value, sizeof(T));
+}
+
+template <typename Traits>
+template <typename T>
+T DeviceMemoryManager<Traits>::Read(DAddr address) const {
+    const T* ptr = GetPointer<T>(address);
+    T result{};
+    if (!ptr) [[unlikely]] {
+        return result;
+    }
+    std::memcpy(&result, ptr, sizeof(T));
+    return result;
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::WalkBlock(DAddr addr, std::size_t size, auto on_unmapped,
+                                            auto on_memory, auto increment) {
+    std::size_t remaining_size = size;
+    std::size_t page_index = addr >> Memory::YUZU_PAGEBITS;
+    std::size_t page_offset = addr & Memory::YUZU_PAGEMASK;
+
+    while (remaining_size) {
+        const size_t next_pages = static_cast<std::size_t>(continuity_tracker[page_index]);
+        const std::size_t copy_amount =
+            std::min((next_pages << Memory::YUZU_PAGEBITS) - page_offset, remaining_size);
+        const auto current_vaddr =
+            static_cast<u64>((page_index << Memory::YUZU_PAGEBITS) + page_offset);
+        SCOPE_EXIT({
+            page_index += next_pages;
+            page_offset = 0;
+            increment(copy_amount);
+            remaining_size -= copy_amount;
+        });
+
+        auto phys_addr = compressed_physical_ptr[page_index];
+        if (phys_addr == 0) {
+            on_unmapped(copy_amount, current_vaddr);
+            continue;
+        }
+        auto* mem_ptr = GetPointerFromRaw<u8>(
+            (static_cast<PAddr>(phys_addr - 1) << Memory::YUZU_PAGEBITS) + page_offset);
+        on_memory(copy_amount, mem_ptr);
+    }
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::ReadBlock(DAddr address, void* dest_pointer, size_t size) {
+    device_inter->FlushRegion(address, size);
+    WalkBlock(
+        address, size,
+        [&](size_t copy_amount, DAddr current_vaddr) {
+            LOG_ERROR(
+                HW_Memory,
+                "Unmapped Device ReadBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
+                current_vaddr, address, size);
+            std::memset(dest_pointer, 0, copy_amount);
+        },
+        [&](size_t copy_amount, const u8* const src_ptr) {
+            std::memcpy(dest_pointer, src_ptr, copy_amount);
+        },
+        [&](const std::size_t copy_amount) {
+            dest_pointer = static_cast<u8*>(dest_pointer) + copy_amount;
+        });
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::WriteBlock(DAddr address, const void* src_pointer, size_t size) {
+    WalkBlock(
+        address, size,
+        [&](size_t copy_amount, DAddr current_vaddr) {
+            LOG_ERROR(
+                HW_Memory,
+                "Unmapped Device WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
+                current_vaddr, address, size);
+        },
+        [&](size_t copy_amount, u8* const dst_ptr) {
+            std::memcpy(dst_ptr, src_pointer, copy_amount);
+        },
+        [&](const std::size_t copy_amount) {
+            src_pointer = static_cast<const u8*>(src_pointer) + copy_amount;
+        });
+    device_inter->InvalidateRegion(address, size);
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::ReadBlockUnsafe(DAddr address, void* dest_pointer, size_t size) {
+    WalkBlock(
+        address, size,
+        [&](size_t copy_amount, DAddr current_vaddr) {
+            LOG_ERROR(
+                HW_Memory,
+                "Unmapped Device ReadBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
+                current_vaddr, address, size);
+            std::memset(dest_pointer, 0, copy_amount);
+        },
+        [&](size_t copy_amount, const u8* const src_ptr) {
+            std::memcpy(dest_pointer, src_ptr, copy_amount);
+        },
+        [&](const std::size_t copy_amount) {
+            dest_pointer = static_cast<u8*>(dest_pointer) + copy_amount;
+        });
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::WriteBlockUnsafe(DAddr address, const void* src_pointer,
+                                                   size_t size) {
+    WalkBlock(
+        address, size,
+        [&](size_t copy_amount, DAddr current_vaddr) {
+            LOG_ERROR(
+                HW_Memory,
+                "Unmapped Device WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
+                current_vaddr, address, size);
+        },
+        [&](size_t copy_amount, u8* const dst_ptr) {
+            std::memcpy(dst_ptr, src_pointer, copy_amount);
+        },
+        [&](const std::size_t copy_amount) {
+            src_pointer = static_cast<const u8*>(src_pointer) + copy_amount;
+        });
+}
+
+template <typename Traits>
+size_t DeviceMemoryManager<Traits>::RegisterProcess(Memory::Memory* memory_device_inter) {
+    size_t new_id;
+    if (!id_pool.empty()) {
+        new_id = id_pool.front();
+        id_pool.pop_front();
+        registered_processes[new_id] = memory_device_inter;
+    } else {
+        registered_processes.emplace_back(memory_device_inter);
+        new_id = registered_processes.size() - 1U;
+    }
+    return new_id;
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::UnregisterProcess(size_t id) {
+    registered_processes[id] = nullptr;
+    id_pool.push_front(id);
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::UpdatePagesCachedCount(DAddr addr, size_t size, s32 delta) {
+    bool locked = false;
+    auto lock = [&] {
+        if (!locked) {
+            counter_guard.lock();
+            locked = true;
+        }
+    };
+    SCOPE_EXIT({
+        if (locked) {
+            counter_guard.unlock();
+        }
+    });
+    u64 uncache_begin = 0;
+    u64 cache_begin = 0;
+    u64 uncache_bytes = 0;
+    u64 cache_bytes = 0;
+    const auto MarkRegionCaching = &DeviceMemoryManager<Traits>::DeviceMethods::MarkRegionCaching;
+
+    std::atomic_thread_fence(std::memory_order_acquire);
+    const size_t page_end = Common::DivCeil(addr + size, Memory::YUZU_PAGESIZE);
+    size_t page = addr >> Memory::YUZU_PAGEBITS;
+    auto [process_id, base_vaddress] = ExtractCPUBacking(page);
+    size_t vpage = base_vaddress >> Memory::YUZU_PAGEBITS;
+    auto* memory_device_inter = registered_processes[process_id];
+    for (; page != page_end; ++page) {
+        std::atomic_uint8_t& count = cached_pages->at(page >> 3).Count(page);
+
+        if (delta > 0) {
+            ASSERT_MSG(count.load(std::memory_order::relaxed) < std::numeric_limits<u8>::max(),
+                       "Count may overflow!");
+        } else if (delta < 0) {
+            ASSERT_MSG(count.load(std::memory_order::relaxed) > 0, "Count may underflow!");
+        } else {
+            ASSERT_MSG(false, "Delta must be non-zero!");
+        }
+
+        // Adds or subtracts 1, as count is a unsigned 8-bit value
+        count.fetch_add(static_cast<u8>(delta), std::memory_order_release);
+
+        // Assume delta is either -1 or 1
+        if (count.load(std::memory_order::relaxed) == 0) {
+            if (uncache_bytes == 0) {
+                uncache_begin = vpage;
+            }
+            uncache_bytes += Memory::YUZU_PAGESIZE;
+        } else if (uncache_bytes > 0) {
+            lock();
+            MarkRegionCaching(memory_device_inter, uncache_begin << Memory::YUZU_PAGEBITS,
+                              uncache_bytes, false);
+            uncache_bytes = 0;
+        }
+        if (count.load(std::memory_order::relaxed) == 1 && delta > 0) {
+            if (cache_bytes == 0) {
+                cache_begin = vpage;
+            }
+            cache_bytes += Memory::YUZU_PAGESIZE;
+        } else if (cache_bytes > 0) {
+            lock();
+            MarkRegionCaching(memory_device_inter, cache_begin << Memory::YUZU_PAGEBITS, cache_bytes,
+                              true);
+            cache_bytes = 0;
+        }
+        vpage++;
+    }
+    if (uncache_bytes > 0) {
+        lock();
+        MarkRegionCaching(memory_device_inter, uncache_begin << Memory::YUZU_PAGEBITS, uncache_bytes,
+                          false);
+    }
+    if (cache_bytes > 0) {
+        lock();
+        MarkRegionCaching(memory_device_inter, cache_begin << Memory::YUZU_PAGEBITS, cache_bytes,
+                          true);
+    }
+}
+
+} // namespace Core

src/core/file_sys/patch_manager.cpp

@@ -26,7 +26,7 @@
 #include "core/file_sys/vfs_vector.h"
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/ns/language.h"
-#include "core/hle/service/set/set.h"
+#include "core/hle/service/set/settings_server.h"
 #include "core/loader/loader.h"
 #include "core/loader/nso.h"
 #include "core/memory/cheat_engine.h"

src/core/gpu_dirty_memory_manager.h

@@ -10,7 +10,7 @@
 #include <utility>
 #include <vector>
 
-#include "core/memory.h"
+#include "core/device_memory_manager.h"
 
 namespace Core {
 
@@ -23,7 +23,7 @@ public:
 
     ~GPUDirtyMemoryManager() = default;
 
-    void Collect(VAddr address, size_t size) {
+    void Collect(PAddr address, size_t size) {
         TransformAddress t = BuildTransform(address, size);
         TransformAddress tmp, original;
         do {
@@ -47,7 +47,7 @@ public:
                                                 std::memory_order_relaxed));
     }
 
-    void Gather(std::function<void(VAddr, size_t)>& callback) {
+    void Gather(std::function<void(PAddr, size_t)>& callback) {
         {
             std::scoped_lock lk(guard);
             TransformAddress t = current.exchange(default_transform, std::memory_order_relaxed);
@@ -65,7 +65,7 @@ public:
                 mask = mask >> empty_bits;
 
                 const size_t continuous_bits = std::countr_one(mask);
-                callback((static_cast<VAddr>(transform.address) << page_bits) + offset,
+                callback((static_cast<PAddr>(transform.address) << page_bits) + offset,
                          continuous_bits << align_bits);
                 mask = continuous_bits < align_size ? (mask >> continuous_bits) : 0;
                 offset += continuous_bits << align_bits;
@@ -80,7 +80,7 @@ private:
         u32 mask;
     };
 
-    constexpr static size_t page_bits = Memory::YUZU_PAGEBITS - 1;
+    constexpr static size_t page_bits = DEVICE_PAGEBITS - 1;
     constexpr static size_t page_size = 1ULL << page_bits;
     constexpr static size_t page_mask = page_size - 1;
 
@@ -89,7 +89,7 @@ private:
     constexpr static size_t align_mask = align_size - 1;
     constexpr static TransformAddress default_transform = {.address = ~0U, .mask = 0U};
 
-    bool IsValid(VAddr address) {
+    bool IsValid(PAddr address) {
         return address < (1ULL << 39);
     }
 
@@ -103,7 +103,7 @@ private:
         return mask;
     }
 
-    TransformAddress BuildTransform(VAddr address, size_t size) {
+    TransformAddress BuildTransform(PAddr address, size_t size) {
         const size_t minor_address = address & page_mask;
         const size_t minor_bit = minor_address >> align_bits;
         const size_t top_bit = (minor_address + size + align_mask) >> align_bits;

src/core/guest_memory.h

@@ -0,0 +1,214 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <iterator>
+#include <memory>
+#include <optional>
+#include <span>
+#include <vector>
+
+#include "common/assert.h"
+#include "common/scratch_buffer.h"
+
+namespace Core::Memory {
+
+enum GuestMemoryFlags : u32 {
+    Read = 1 << 0,
+    Write = 1 << 1,
+    Safe = 1 << 2,
+    Cached = 1 << 3,
+
+    SafeRead = Read | Safe,
+    SafeWrite = Write | Safe,
+    SafeReadWrite = SafeRead | SafeWrite,
+    SafeReadCachedWrite = SafeReadWrite | Cached,
+
+    UnsafeRead = Read,
+    UnsafeWrite = Write,
+    UnsafeReadWrite = UnsafeRead | UnsafeWrite,
+    UnsafeReadCachedWrite = UnsafeReadWrite | Cached,
+};
+
+namespace {
+template <typename M, typename T, GuestMemoryFlags FLAGS>
+class GuestMemory {
+    using iterator = T*;
+    using const_iterator = const T*;
+    using value_type = T;
+    using element_type = T;
+    using iterator_category = std::contiguous_iterator_tag;
+
+public:
+    GuestMemory() = delete;
+    explicit GuestMemory(M& memory, u64 addr, std::size_t size,
+                         Common::ScratchBuffer<T>* backup = nullptr)
+        : m_memory{memory}, m_addr{addr}, m_size{size} {
+        static_assert(FLAGS & GuestMemoryFlags::Read || FLAGS & GuestMemoryFlags::Write);
+        if constexpr (FLAGS & GuestMemoryFlags::Read) {
+            Read(addr, size, backup);
+        }
+    }
+
+    ~GuestMemory() = default;
+
+    T* data() noexcept {
+        return m_data_span.data();
+    }
+
+    const T* data() const noexcept {
+        return m_data_span.data();
+    }
+
+    size_t size() const noexcept {
+        return m_size;
+    }
+
+    size_t size_bytes() const noexcept {
+        return this->size() * sizeof(T);
+    }
+
+    [[nodiscard]] T* begin() noexcept {
+        return this->data();
+    }
+
+    [[nodiscard]] const T* begin() const noexcept {
+        return this->data();
+    }
+
+    [[nodiscard]] T* end() noexcept {
+        return this->data() + this->size();
+    }
+
+    [[nodiscard]] const T* end() const noexcept {
+        return this->data() + this->size();
+    }
+
+    T& operator[](size_t index) noexcept {
+        return m_data_span[index];
+    }
+
+    const T& operator[](size_t index) const noexcept {
+        return m_data_span[index];
+    }
+
+    void SetAddressAndSize(u64 addr, std::size_t size) noexcept {
+        m_addr = addr;
+        m_size = size;
+        m_addr_changed = true;
+    }
+
+    std::span<T> Read(u64 addr, std::size_t size,
+                      Common::ScratchBuffer<T>* backup = nullptr) noexcept {
+        m_addr = addr;
+        m_size = size;
+        if (m_size == 0) {
+            m_is_data_copy = true;
+            return {};
+        }
+
+        if (this->TrySetSpan()) {
+            if constexpr (FLAGS & GuestMemoryFlags::Safe) {
+                m_memory.FlushRegion(m_addr, this->size_bytes());
+            }
+        } else {
+            if (backup) {
+                backup->resize_destructive(this->size());
+                m_data_span = *backup;
+            } else {
+                m_data_copy.resize(this->size());
+                m_data_span = std::span(m_data_copy);
+            }
+            m_is_data_copy = true;
+            m_span_valid = true;
+            if constexpr (FLAGS & GuestMemoryFlags::Safe) {
+                m_memory.ReadBlock(m_addr, this->data(), this->size_bytes());
+            } else {
+                m_memory.ReadBlockUnsafe(m_addr, this->data(), this->size_bytes());
+            }
+        }
+        return m_data_span;
+    }
+
+    void Write(std::span<T> write_data) noexcept {
+        if constexpr (FLAGS & GuestMemoryFlags::Cached) {
+            m_memory.WriteBlockCached(m_addr, write_data.data(), this->size_bytes());
+        } else if constexpr (FLAGS & GuestMemoryFlags::Safe) {
+            m_memory.WriteBlock(m_addr, write_data.data(), this->size_bytes());
+        } else {
+            m_memory.WriteBlockUnsafe(m_addr, write_data.data(), this->size_bytes());
+        }
+    }
+
+    bool TrySetSpan() noexcept {
+        if (u8* ptr = m_memory.GetSpan(m_addr, this->size_bytes()); ptr) {
+            m_data_span = {reinterpret_cast<T*>(ptr), this->size()};
+            m_span_valid = true;
+            return true;
+        }
+        return false;
+    }
+
+protected:
+    bool IsDataCopy() const noexcept {
+        return m_is_data_copy;
+    }
+
+    bool AddressChanged() const noexcept {
+        return m_addr_changed;
+    }
+
+    M& m_memory;
+    u64 m_addr{};
+    size_t m_size{};
+    std::span<T> m_data_span{};
+    std::vector<T> m_data_copy{};
+    bool m_span_valid{false};
+    bool m_is_data_copy{false};
+    bool m_addr_changed{false};
+};
+
+template <typename M, typename T, GuestMemoryFlags FLAGS>
+class GuestMemoryScoped : public GuestMemory<M, T, FLAGS> {
+public:
+    GuestMemoryScoped() = delete;
+    explicit GuestMemoryScoped(M& memory, u64 addr, std::size_t size,
+                               Common::ScratchBuffer<T>* backup = nullptr)
+        : GuestMemory<M, T, FLAGS>(memory, addr, size, backup) {
+        if constexpr (!(FLAGS & GuestMemoryFlags::Read)) {
+            if (!this->TrySetSpan()) {
+                if (backup) {
+                    this->m_data_span = *backup;
+                    this->m_span_valid = true;
+                    this->m_is_data_copy = true;
+                }
+            }
+        }
+    }
+
+    ~GuestMemoryScoped() {
+        if constexpr (FLAGS & GuestMemoryFlags::Write) {
+            if (this->size() == 0) [[unlikely]] {
+                return;
+            }
+
+            if (this->AddressChanged() || this->IsDataCopy()) {
+                ASSERT(this->m_span_valid);
+                if constexpr (FLAGS & GuestMemoryFlags::Cached) {
+                    this->m_memory.WriteBlockCached(this->m_addr, this->data(), this->size_bytes());
+                } else if constexpr (FLAGS & GuestMemoryFlags::Safe) {
+                    this->m_memory.WriteBlock(this->m_addr, this->data(), this->size_bytes());
+                } else {
+                    this->m_memory.WriteBlockUnsafe(this->m_addr, this->data(), this->size_bytes());
+                }
+            } else if constexpr ((FLAGS & GuestMemoryFlags::Safe) ||
+                                 (FLAGS & GuestMemoryFlags::Cached)) {
+                this->m_memory.InvalidateRegion(this->m_addr, this->size_bytes());
+            }
+        }
+    }
+};
+} // namespace
+
+} // namespace Core::Memory

src/core/hle/kernel/k_memory_block.h

@@ -81,12 +81,12 @@ enum class KMemoryState : u32 {
 
     ThreadLocal = static_cast<u32>(Svc::MemoryState::ThreadLocal) | FlagLinearMapped,
 
-    Transfered = static_cast<u32>(Svc::MemoryState::Transfered) | FlagsMisc |
-                 FlagCanAlignedDeviceMap | FlagCanChangeAttribute | FlagCanUseIpc |
-                 FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+    Transferred = static_cast<u32>(Svc::MemoryState::Transferred) | FlagsMisc |
+                  FlagCanAlignedDeviceMap | FlagCanChangeAttribute | FlagCanUseIpc |
+                  FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
 
-    SharedTransfered = static_cast<u32>(Svc::MemoryState::SharedTransfered) | FlagsMisc |
-                       FlagCanAlignedDeviceMap | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+    SharedTransferred = static_cast<u32>(Svc::MemoryState::SharedTransferred) | FlagsMisc |
+                        FlagCanAlignedDeviceMap | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
 
     SharedCode = static_cast<u32>(Svc::MemoryState::SharedCode) | FlagMapped |
                  FlagReferenceCounted | FlagLinearMapped | FlagCanUseNonSecureIpc |
@@ -130,8 +130,8 @@ static_assert(static_cast<u32>(KMemoryState::AliasCodeData) == 0x0FFFBD09);
 static_assert(static_cast<u32>(KMemoryState::Ipc) == 0x045C3C0A);
 static_assert(static_cast<u32>(KMemoryState::Stack) == 0x045C3C0B);
 static_assert(static_cast<u32>(KMemoryState::ThreadLocal) == 0x0400000C);
-static_assert(static_cast<u32>(KMemoryState::Transfered) == 0x055C3C0D);
-static_assert(static_cast<u32>(KMemoryState::SharedTransfered) == 0x045C380E);
+static_assert(static_cast<u32>(KMemoryState::Transferred) == 0x055C3C0D);
+static_assert(static_cast<u32>(KMemoryState::SharedTransferred) == 0x045C380E);
 static_assert(static_cast<u32>(KMemoryState::SharedCode) == 0x0440380F);
 static_assert(static_cast<u32>(KMemoryState::Inaccessible) == 0x00000010);
 static_assert(static_cast<u32>(KMemoryState::NonSecureIpc) == 0x045C3811);

src/core/hle/kernel/k_page_table_base.cpp

@@ -486,8 +486,8 @@ KProcessAddress KPageTableBase::GetRegionAddress(Svc::MemoryState state) const {
     case Svc::MemoryState::Shared:
     case Svc::MemoryState::AliasCode:
     case Svc::MemoryState::AliasCodeData:
-    case Svc::MemoryState::Transfered:
-    case Svc::MemoryState::SharedTransfered:
+    case Svc::MemoryState::Transferred:
+    case Svc::MemoryState::SharedTransferred:
     case Svc::MemoryState::SharedCode:
     case Svc::MemoryState::GeneratedCode:
     case Svc::MemoryState::CodeOut:
@@ -522,8 +522,8 @@ size_t KPageTableBase::GetRegionSize(Svc::MemoryState state) const {
     case Svc::MemoryState::Shared:
     case Svc::MemoryState::AliasCode:
     case Svc::MemoryState::AliasCodeData:
-    case Svc::MemoryState::Transfered:
-    case Svc::MemoryState::SharedTransfered:
+    case Svc::MemoryState::Transferred:
+    case Svc::MemoryState::SharedTransferred:
     case Svc::MemoryState::SharedCode:
     case Svc::MemoryState::GeneratedCode:
     case Svc::MemoryState::CodeOut:
@@ -564,8 +564,8 @@ bool KPageTableBase::CanContain(KProcessAddress addr, size_t size, Svc::MemorySt
     case Svc::MemoryState::AliasCodeData:
     case Svc::MemoryState::Stack:
     case Svc::MemoryState::ThreadLocal:
-    case Svc::MemoryState::Transfered:
-    case Svc::MemoryState::SharedTransfered:
+    case Svc::MemoryState::Transferred:
+    case Svc::MemoryState::SharedTransferred:
     case Svc::MemoryState::SharedCode:
     case Svc::MemoryState::GeneratedCode:
     case Svc::MemoryState::CodeOut:

src/core/hle/kernel/k_process.cpp

@@ -5,6 +5,7 @@
 #include "common/scope_exit.h"
 #include "common/settings.h"
 #include "core/core.h"
+#include "core/gpu_dirty_memory_manager.h"
 #include "core/hle/kernel/k_process.h"
 #include "core/hle/kernel/k_scoped_resource_reservation.h"
 #include "core/hle/kernel/k_shared_memory.h"
@@ -314,7 +315,7 @@ Result KProcess::Initialize(const Svc::CreateProcessParameter& params, const KPa
 
     // Ensure our memory is initialized.
     m_memory.SetCurrentPageTable(*this);
-    m_memory.SetGPUDirtyManagers(m_dirty_memory_managers);
+    m_memory.SetGPUDirtyManagers(m_kernel.System().GetGPUDirtyMemoryManager());
 
     // Ensure we can insert the code region.
     R_UNLESS(m_page_table.CanContain(params.code_address, params.code_num_pages * PageSize,
@@ -411,7 +412,7 @@ Result KProcess::Initialize(const Svc::CreateProcessParameter& params,
 
     // Ensure our memory is initialized.
     m_memory.SetCurrentPageTable(*this);
-    m_memory.SetGPUDirtyManagers(m_dirty_memory_managers);
+    m_memory.SetGPUDirtyManagers(m_kernel.System().GetGPUDirtyMemoryManager());
 
     // Ensure we can insert the code region.
     R_UNLESS(m_page_table.CanContain(params.code_address, code_size, KMemoryState::Code),
@@ -1135,8 +1136,7 @@ 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_dirty_memory_managers{},
-      m_exclusive_monitor{}, m_memory{kernel.System()} {}
+      m_handle_table{kernel}, m_exclusive_monitor{}, m_memory{kernel.System()} {}
 KProcess::~KProcess() = default;
 
 Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std::size_t code_size,
@@ -1318,10 +1318,4 @@ bool KProcess::RemoveWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointT
     return true;
 }
 
-void KProcess::GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback) {
-    for (auto& manager : m_dirty_memory_managers) {
-        manager.Gather(callback);
-    }
-}
-
 } // namespace Kernel

src/core/hle/kernel/k_process.h

@@ -7,7 +7,6 @@
 
 #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"
@@ -128,7 +127,6 @@ 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;
 
@@ -511,8 +509,6 @@ public:
         return m_memory;
     }
 
-    void GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback);
-
     Core::ExclusiveMonitor& GetExclusiveMonitor() const {
         return *m_exclusive_monitor;
     }

src/core/hle/kernel/k_transfer_memory.cpp

@@ -76,8 +76,8 @@ Result KTransferMemory::Map(KProcessAddress address, size_t size, Svc::MemoryPer
 
     // Map the memory.
     const KMemoryState state = (m_owner_perm == Svc::MemoryPermission::None)
-                                   ? KMemoryState::Transfered
-                                   : KMemoryState::SharedTransfered;
+                                   ? KMemoryState::Transferred
+                                   : KMemoryState::SharedTransferred;
     R_TRY(GetCurrentProcess(m_kernel).GetPageTable().MapPageGroup(
         address, *m_page_group, state, KMemoryPermission::UserReadWrite));
 
@@ -96,8 +96,8 @@ Result KTransferMemory::Unmap(KProcessAddress address, size_t size) {
 
     // Unmap the memory.
     const KMemoryState state = (m_owner_perm == Svc::MemoryPermission::None)
-                                   ? KMemoryState::Transfered
-                                   : KMemoryState::SharedTransfered;
+                                   ? KMemoryState::Transferred
+                                   : KMemoryState::SharedTransferred;
     R_TRY(GetCurrentProcess(m_kernel).GetPageTable().UnmapPageGroup(address, *m_page_group, state));
 
     // Mark ourselves as unmapped.

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

@@ -90,7 +90,7 @@ Result MapTransferMemory(Core::System& system, Handle trmem_handle, uint64_t add
     // Verify that the mapping is in range.
     R_UNLESS(GetCurrentProcess(system.Kernel())
                  .GetPageTable()
-                 .CanContain(address, size, KMemoryState::Transfered),
+                 .CanContain(address, size, KMemoryState::Transferred),
              ResultInvalidMemoryRegion);
 
     // Map the transfer memory.
@@ -117,7 +117,7 @@ Result UnmapTransferMemory(Core::System& system, Handle trmem_handle, uint64_t a
     // Verify that the mapping is in range.
     R_UNLESS(GetCurrentProcess(system.Kernel())
                  .GetPageTable()
-                 .CanContain(address, size, KMemoryState::Transfered),
+                 .CanContain(address, size, KMemoryState::Transferred),
              ResultInvalidMemoryRegion);
 
     // Unmap the transfer memory.

src/core/hle/kernel/svc_types.h

@@ -27,8 +27,8 @@ enum class MemoryState : u32 {
     Ipc = 0x0A,
     Stack = 0x0B,
     ThreadLocal = 0x0C,
-    Transfered = 0x0D,
-    SharedTransfered = 0x0E,
+    Transferred = 0x0D,
+    SharedTransferred = 0x0E,
     SharedCode = 0x0F,
     Inaccessible = 0x10,
     NonSecureIpc = 0x11,

src/core/hle/service/am/applets/applet_profile_select.h

@@ -76,7 +76,7 @@ struct UiSettingsDisplayOptions {
     bool is_system_or_launcher;
     bool is_registration_permitted;
     bool show_skip_button;
-    bool aditional_select;
+    bool additional_select;
     bool show_user_selector;
     bool is_unqualified_user_selectable;
 };

src/core/hle/service/caps/caps_manager.cpp

@@ -85,7 +85,7 @@ Result AlbumManager::GetAlbumFileList(std::vector<AlbumEntry>& out_entries, Albu
 }
 
 Result AlbumManager::GetAlbumFileList(std::vector<ApplicationAlbumFileEntry>& out_entries,
-                                      ContentType contex_type, s64 start_posix_time,
+                                      ContentType content_type, s64 start_posix_time,
                                       s64 end_posix_time, u64 aruid) const {
     if (!is_mounted) {
         return ResultIsNotMounted;
@@ -94,7 +94,7 @@ Result AlbumManager::GetAlbumFileList(std::vector<ApplicationAlbumFileEntry>& ou
     std::vector<ApplicationAlbumEntry> album_entries;
     const auto start_date = ConvertToAlbumDateTime(start_posix_time);
     const auto end_date = ConvertToAlbumDateTime(end_posix_time);
-    const auto result = GetAlbumFileList(album_entries, contex_type, start_date, end_date, aruid);
+    const auto result = GetAlbumFileList(album_entries, content_type, start_date, end_date, aruid);
 
     if (result.IsError()) {
         return result;
@@ -113,14 +113,14 @@ Result AlbumManager::GetAlbumFileList(std::vector<ApplicationAlbumFileEntry>& ou
 }
 
 Result AlbumManager::GetAlbumFileList(std::vector<ApplicationAlbumEntry>& out_entries,
-                                      ContentType contex_type, AlbumFileDateTime start_date,
+                                      ContentType content_type, AlbumFileDateTime start_date,
                                       AlbumFileDateTime end_date, u64 aruid) const {
     if (!is_mounted) {
         return ResultIsNotMounted;
     }
 
     for (auto& [file_id, path] : album_files) {
-        if (file_id.type != contex_type) {
+        if (file_id.type != content_type) {
             continue;
         }
         if (file_id.date > start_date) {
@@ -139,7 +139,7 @@ Result AlbumManager::GetAlbumFileList(std::vector<ApplicationAlbumEntry>& out_en
             .hash{},
             .datetime = file_id.date,
             .storage = file_id.storage,
-            .content = contex_type,
+            .content = content_type,
             .unknown = 1,
         };
         out_entries.push_back(entry);

src/core/hle/service/caps/caps_manager.h

@@ -45,10 +45,10 @@ public:
     Result GetAlbumFileList(std::vector<AlbumEntry>& out_entries, AlbumStorage storage,
                             u8 flags) const;
     Result GetAlbumFileList(std::vector<ApplicationAlbumFileEntry>& out_entries,
-                            ContentType contex_type, s64 start_posix_time, s64 end_posix_time,
+                            ContentType content_type, s64 start_posix_time, s64 end_posix_time,
                             u64 aruid) const;
     Result GetAlbumFileList(std::vector<ApplicationAlbumEntry>& out_entries,
-                            ContentType contex_type, AlbumFileDateTime start_date,
+                            ContentType content_type, AlbumFileDateTime start_date,
                             AlbumFileDateTime end_date, u64 aruid) const;
     Result GetAutoSavingStorage(bool& out_is_autosaving) const;
     Result LoadAlbumScreenShotImage(LoadAlbumScreenShotImageOutput& out_image_output,

src/core/hle/service/caps/caps_result.h

@@ -12,7 +12,7 @@ constexpr Result ResultUnknown5(ErrorModule::Capture, 5);
 constexpr Result ResultUnknown6(ErrorModule::Capture, 6);
 constexpr Result ResultUnknown7(ErrorModule::Capture, 7);
 constexpr Result ResultOutOfRange(ErrorModule::Capture, 8);
-constexpr Result ResulInvalidTimestamp(ErrorModule::Capture, 12);
+constexpr Result ResultInvalidTimestamp(ErrorModule::Capture, 12);
 constexpr Result ResultInvalidStorage(ErrorModule::Capture, 13);
 constexpr Result ResultInvalidFileContents(ErrorModule::Capture, 14);
 constexpr Result ResultIsNotMounted(ErrorModule::Capture, 21);

src/core/hle/service/friend/friend.cpp

@@ -131,7 +131,7 @@ private:
         u8 is_favorite;
         u8 same_app;
         u8 same_app_played;
-        u8 arbitary_app_played;
+        u8 arbitrary_app_played;
         u64 group_id;
     };
     static_assert(sizeof(SizedFriendFilter) == 0x10, "SizedFriendFilter is an invalid size");

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

@@ -18,23 +18,23 @@ namespace Service::HID {
 
 void LoopProcess(Core::System& system) {
     auto server_manager = std::make_unique<ServerManager>(system);
-    std::shared_ptr<ResourceManager> resouce_manager = std::make_shared<ResourceManager>(system);
+    std::shared_ptr<ResourceManager> resource_manager = std::make_shared<ResourceManager>(system);
     std::shared_ptr<HidFirmwareSettings> firmware_settings =
         std::make_shared<HidFirmwareSettings>();
 
     // TODO: Remove this hack until this service is emulated properly.
     const auto process_list = system.Kernel().GetProcessList();
     if (!process_list.empty()) {
-        resouce_manager->Initialize();
-        resouce_manager->RegisterAppletResourceUserId(process_list[0]->GetId(), true);
+        resource_manager->Initialize();
+        resource_manager->RegisterAppletResourceUserId(process_list[0]->GetId(), true);
     }
 
     server_manager->RegisterNamedService(
-        "hid", std::make_shared<IHidServer>(system, resouce_manager, firmware_settings));
+        "hid", std::make_shared<IHidServer>(system, resource_manager, firmware_settings));
     server_manager->RegisterNamedService(
-        "hid:dbg", std::make_shared<IHidDebugServer>(system, resouce_manager));
+        "hid:dbg", std::make_shared<IHidDebugServer>(system, resource_manager));
     server_manager->RegisterNamedService(
-        "hid:sys", std::make_shared<IHidSystemServer>(system, resouce_manager));
+        "hid:sys", std::make_shared<IHidSystemServer>(system, resource_manager));
 
     server_manager->RegisterNamedService("hidbus", std::make_shared<HidBus>(system));
 

src/core/hle/service/hid/hid_server.cpp

@@ -1444,8 +1444,8 @@ void IHidServer::SetNpadAnalogStickUseCenterClamp(HLERequestContext& ctx) {
 
     const auto parameters{rp.PopRaw<Parameters>()};
 
-    LOG_WARNING(Service_HID, "(STUBBED) called, use_center_clamp={}, applet_resource_user_id={}",
-                parameters.use_center_clamp, parameters.applet_resource_user_id);
+    LOG_INFO(Service_HID, "called, use_center_clamp={}, applet_resource_user_id={}",
+             parameters.use_center_clamp, parameters.applet_resource_user_id);
 
     GetResourceManager()->GetNpad()->SetNpadAnalogStickUseCenterClamp(
         parameters.applet_resource_user_id, parameters.use_center_clamp);
@@ -1466,23 +1466,27 @@ void IHidServer::SetNpadCaptureButtonAssignment(HLERequestContext& ctx) {
 
     const auto parameters{rp.PopRaw<Parameters>()};
 
-    LOG_WARNING(Service_HID,
-                "(STUBBED) called, npad_styleset={}, applet_resource_user_id={}, button={}",
-                parameters.npad_styleset, parameters.applet_resource_user_id, parameters.button);
+    LOG_INFO(Service_HID, "called, npad_styleset={}, applet_resource_user_id={}, button={}",
+             parameters.npad_styleset, parameters.applet_resource_user_id, parameters.button);
+
+    const auto result = GetResourceManager()->GetNpad()->SetNpadCaptureButtonAssignment(
+        parameters.applet_resource_user_id, parameters.npad_styleset, parameters.button);
 
     IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(ResultSuccess);
+    rb.Push(result);
 }
 
 void IHidServer::ClearNpadCaptureButtonAssignment(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto applet_resource_user_id{rp.Pop<u64>()};
 
-    LOG_WARNING(Service_HID, "(STUBBED) called, applet_resource_user_id={}",
-                applet_resource_user_id);
+    LOG_INFO(Service_HID, "called, applet_resource_user_id={}", applet_resource_user_id);
+
+    const auto result =
+        GetResourceManager()->GetNpad()->ClearNpadCaptureButtonAssignment(applet_resource_user_id);
 
     IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(ResultSuccess);
+    rb.Push(result);
 }
 
 void IHidServer::GetVibrationDeviceInfo(HLERequestContext& ctx) {

src/core/hle/service/hid/hid_system_server.cpp

@@ -46,7 +46,7 @@ IHidSystemServer::IHidSystemServer(Core::System& system_, std::shared_ptr<Resour
         {310, &IHidSystemServer::GetMaskedSupportedNpadStyleSet, "GetMaskedSupportedNpadStyleSet"},
         {311, nullptr, "SetNpadPlayerLedBlinkingDevice"},
         {312, &IHidSystemServer::SetSupportedNpadStyleSetAll, "SetSupportedNpadStyleSetAll"},
-        {313, nullptr, "GetNpadCaptureButtonAssignment"},
+        {313, &IHidSystemServer::GetNpadCaptureButtonAssignment, "GetNpadCaptureButtonAssignment"},
         {314, nullptr, "GetAppletFooterUiType"},
         {315, &IHidSystemServer::GetAppletDetailedUiType, "GetAppletDetailedUiType"},
         {316, &IHidSystemServer::GetNpadInterfaceType, "GetNpadInterfaceType"},
@@ -54,8 +54,8 @@ IHidSystemServer::IHidSystemServer(Core::System& system_, std::shared_ptr<Resour
         {318, &IHidSystemServer::HasBattery, "HasBattery"},
         {319, &IHidSystemServer::HasLeftRightBattery, "HasLeftRightBattery"},
         {321, &IHidSystemServer::GetUniquePadsFromNpad, "GetUniquePadsFromNpad"},
-        {322, &IHidSystemServer::GetIrSensorState, "GetIrSensorState"},
-        {323, nullptr, "GetXcdHandleForNpadWithIrSensor"},
+        {322, &IHidSystemServer::SetNpadSystemExtStateEnabled, "SetNpadSystemExtStateEnabled"},
+        {323, nullptr, "GetLastActiveUniquePad"},
         {324, nullptr, "GetUniquePadButtonSet"},
         {325, nullptr, "GetUniquePadColor"},
         {326, nullptr, "GetUniquePadAppletDetailedUiType"},
@@ -251,25 +251,38 @@ void IHidSystemServer::ApplyNpadSystemCommonPolicy(HLERequestContext& ctx) {
 }
 
 void IHidSystemServer::EnableAssigningSingleOnSlSrPress(HLERequestContext& ctx) {
-    LOG_WARNING(Service_HID, "(STUBBED) called");
+    IPC::RequestParser rp{ctx};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+
+    LOG_INFO(Service_HID, "called, applet_resource_user_id={}", applet_resource_user_id);
+
+    GetResourceManager()->GetNpad()->AssigningSingleOnSlSrPress(applet_resource_user_id, true);
 
     IPC::ResponseBuilder rb{ctx, 2};
     rb.Push(ResultSuccess);
 }
 
 void IHidSystemServer::DisableAssigningSingleOnSlSrPress(HLERequestContext& ctx) {
-    LOG_WARNING(Service_HID, "(STUBBED) called");
+    IPC::RequestParser rp{ctx};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+
+    LOG_INFO(Service_HID, "called, applet_resource_user_id={}", applet_resource_user_id);
+
+    GetResourceManager()->GetNpad()->AssigningSingleOnSlSrPress(applet_resource_user_id, false);
 
     IPC::ResponseBuilder rb{ctx, 2};
     rb.Push(ResultSuccess);
 }
 
 void IHidSystemServer::GetLastActiveNpad(HLERequestContext& ctx) {
-    LOG_DEBUG(Service_HID, "(STUBBED) called"); // Spams a lot when controller applet is running
+    Core::HID::NpadIdType npad_id{};
+    const Result result = GetResourceManager()->GetNpad()->GetLastActiveNpad(npad_id);
+
+    LOG_DEBUG(Service_HID, "called, npad_id={}", npad_id);
 
     IPC::ResponseBuilder rb{ctx, 3};
-    rb.Push(ResultSuccess);
-    rb.Push(0); // Dont forget to fix this
+    rb.Push(result);
+    rb.PushEnum(npad_id);
 }
 
 void IHidSystemServer::ApplyNpadSystemCommonPolicyFull(HLERequestContext& ctx) {
@@ -331,6 +344,27 @@ void IHidSystemServer::SetSupportedNpadStyleSetAll(HLERequestContext& ctx) {
     rb.Push(result);
 }
 
+void IHidSystemServer::GetNpadCaptureButtonAssignment(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+    const auto capture_button_list_size{ctx.GetWriteBufferNumElements<Core::HID::NpadButton>()};
+
+    LOG_DEBUG(Service_HID, "called, applet_resource_user_id={}", applet_resource_user_id);
+
+    std::vector<Core::HID::NpadButton> capture_button_list(capture_button_list_size);
+    const auto& npad = GetResourceManager()->GetNpad();
+    const u64 list_size =
+        npad->GetNpadCaptureButtonAssignment(capture_button_list, applet_resource_user_id);
+
+    if (list_size != 0) {
+        ctx.WriteBuffer(capture_button_list);
+    }
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(ResultSuccess);
+    rb.Push(list_size);
+}
+
 void IHidSystemServer::GetAppletDetailedUiType(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto npad_id_type{rp.PopEnum<Core::HID::NpadIdType>()};
@@ -423,13 +457,25 @@ void IHidSystemServer::GetUniquePadsFromNpad(HLERequestContext& ctx) {
     rb.Push(static_cast<u32>(unique_pads.size()));
 }
 
-void IHidSystemServer::GetIrSensorState(HLERequestContext& ctx) {
+void IHidSystemServer::SetNpadSystemExtStateEnabled(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
+    struct Parameters {
+        bool is_enabled;
+        INSERT_PADDING_BYTES_NOINIT(7);
+        u64 applet_resource_user_id;
+    };
+    static_assert(sizeof(Parameters) == 0x10, "Parameters has incorrect size.");
 
-    LOG_WARNING(Service_HID, "(STUBBED) called");
+    const auto parameters{rp.PopRaw<Parameters>()};
+
+    LOG_INFO(Service_HID, "called, is_enabled={}, applet_resource_user_id={}",
+             parameters.is_enabled, parameters.applet_resource_user_id);
+
+    const auto result = GetResourceManager()->GetNpad()->SetNpadSystemExtStateEnabled(
+        parameters.applet_resource_user_id, parameters.is_enabled);
 
     IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(ResultSuccess);
+    rb.Push(result);
 }
 void IHidSystemServer::RegisterAppletResourceUserId(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};

src/core/hle/service/hid/hid_system_server.h

@@ -31,13 +31,14 @@ private:
     void GetNpadFullKeyGripColor(HLERequestContext& ctx);
     void GetMaskedSupportedNpadStyleSet(HLERequestContext& ctx);
     void SetSupportedNpadStyleSetAll(HLERequestContext& ctx);
+    void GetNpadCaptureButtonAssignment(HLERequestContext& ctx);
     void GetAppletDetailedUiType(HLERequestContext& ctx);
     void GetNpadInterfaceType(HLERequestContext& ctx);
     void GetNpadLeftRightInterfaceType(HLERequestContext& ctx);
     void HasBattery(HLERequestContext& ctx);
     void HasLeftRightBattery(HLERequestContext& ctx);
     void GetUniquePadsFromNpad(HLERequestContext& ctx);
-    void GetIrSensorState(HLERequestContext& ctx);
+    void SetNpadSystemExtStateEnabled(HLERequestContext& ctx);
     void RegisterAppletResourceUserId(HLERequestContext& ctx);
     void UnregisterAppletResourceUserId(HLERequestContext& ctx);
     void EnableAppletToGetInput(HLERequestContext& ctx);

src/core/hle/service/hid/hidbus.cpp

@@ -67,7 +67,7 @@ HidBus::~HidBus() {
 void HidBus::UpdateHidbus(std::chrono::nanoseconds ns_late) {
     if (is_hidbus_enabled) {
         for (std::size_t i = 0; i < devices.size(); ++i) {
-            if (!devices[i].is_device_initializated) {
+            if (!devices[i].is_device_initialized) {
                 continue;
             }
             auto& device = devices[i].device;
@@ -213,7 +213,7 @@ void HidBus::Initialize(HLERequestContext& ctx) {
 
         if (bus_handle_.internal_index == 0 && Settings::values.enable_ring_controller) {
             MakeDevice<RingController>(bus_handle_);
-            devices[device_index.value()].is_device_initializated = true;
+            devices[device_index.value()].is_device_initialized = true;
             devices[device_index.value()].device->ActivateDevice();
             cur_entry.is_in_focus = true;
             cur_entry.is_connected = true;
@@ -222,7 +222,7 @@ void HidBus::Initialize(HLERequestContext& ctx) {
             cur_entry.is_polling_mode = false;
         } else {
             MakeDevice<HidbusStubbed>(bus_handle_);
-            devices[device_index.value()].is_device_initializated = true;
+            devices[device_index.value()].is_device_initialized = true;
             cur_entry.is_in_focus = true;
             cur_entry.is_connected = false;
             cur_entry.is_connected_result = ResultSuccess;
@@ -261,7 +261,7 @@ void HidBus::Finalize(HLERequestContext& ctx) {
         const auto entry_index = devices[device_index.value()].handle.internal_index;
         auto& cur_entry = hidbus_status.entries[entry_index];
         auto& device = devices[device_index.value()].device;
-        devices[device_index.value()].is_device_initializated = false;
+        devices[device_index.value()].is_device_initialized = false;
         device->DeactivateDevice();
 
         cur_entry.is_in_focus = true;

src/core/hle/service/hid/hidbus.h

@@ -89,7 +89,7 @@ private:
     static_assert(sizeof(HidbusStatusManager) <= 0x1000, "HidbusStatusManager is an invalid size");
 
     struct HidbusDevice {
-        bool is_device_initializated{};
+        bool is_device_initialized{};
         BusHandle handle{};
         std::unique_ptr<HidbusBase> device{nullptr};
     };

src/core/hle/service/hle_ipc.cpp

@@ -12,6 +12,7 @@
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "common/scratch_buffer.h"
+#include "core/guest_memory.h"
 #include "core/hle/kernel/k_auto_object.h"
 #include "core/hle/kernel/k_handle_table.h"
 #include "core/hle/kernel/k_process.h"
@@ -181,22 +182,22 @@ void HLERequestContext::ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming) {
         }
     }
 
-    buffer_x_desciptors.reserve(command_header->num_buf_x_descriptors);
-    buffer_a_desciptors.reserve(command_header->num_buf_a_descriptors);
-    buffer_b_desciptors.reserve(command_header->num_buf_b_descriptors);
-    buffer_w_desciptors.reserve(command_header->num_buf_w_descriptors);
+    buffer_x_descriptors.reserve(command_header->num_buf_x_descriptors);
+    buffer_a_descriptors.reserve(command_header->num_buf_a_descriptors);
+    buffer_b_descriptors.reserve(command_header->num_buf_b_descriptors);
+    buffer_w_descriptors.reserve(command_header->num_buf_w_descriptors);
 
     for (u32 i = 0; i < command_header->num_buf_x_descriptors; ++i) {
-        buffer_x_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorX>());
+        buffer_x_descriptors.push_back(rp.PopRaw<IPC::BufferDescriptorX>());
     }
     for (u32 i = 0; i < command_header->num_buf_a_descriptors; ++i) {
-        buffer_a_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorABW>());
+        buffer_a_descriptors.push_back(rp.PopRaw<IPC::BufferDescriptorABW>());
     }
     for (u32 i = 0; i < command_header->num_buf_b_descriptors; ++i) {
-        buffer_b_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorABW>());
+        buffer_b_descriptors.push_back(rp.PopRaw<IPC::BufferDescriptorABW>());
     }
     for (u32 i = 0; i < command_header->num_buf_w_descriptors; ++i) {
-        buffer_w_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorABW>());
+        buffer_w_descriptors.push_back(rp.PopRaw<IPC::BufferDescriptorABW>());
     }
 
     const auto buffer_c_offset = rp.GetCurrentOffset() + command_header->data_size;
@@ -246,7 +247,7 @@ void HLERequestContext::ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming) {
         IPC::CommandHeader::BufferDescriptorCFlag::InlineDescriptor) {
         if (command_header->buf_c_descriptor_flags ==
             IPC::CommandHeader::BufferDescriptorCFlag::OneDescriptor) {
-            buffer_c_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorC>());
+            buffer_c_descriptors.push_back(rp.PopRaw<IPC::BufferDescriptorC>());
         } else {
             u32 num_buf_c_descriptors =
                 static_cast<u32>(command_header->buf_c_descriptor_flags.Value()) - 2;
@@ -256,7 +257,7 @@ void HLERequestContext::ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming) {
             ASSERT(num_buf_c_descriptors < 14);
 
             for (u32 i = 0; i < num_buf_c_descriptors; ++i) {
-                buffer_c_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorC>());
+                buffer_c_descriptors.push_back(rp.PopRaw<IPC::BufferDescriptorC>());
             }
         }
     }
@@ -344,9 +345,9 @@ std::vector<u8> HLERequestContext::ReadBufferCopy(std::size_t buffer_index) cons
 
 std::span<const u8> HLERequestContext::ReadBufferA(std::size_t buffer_index) const {
     static thread_local std::array read_buffer_a{
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead>(memory, 0, 0),
     };
 
     ASSERT_OR_EXECUTE_MSG(
@@ -360,9 +361,9 @@ std::span<const u8> HLERequestContext::ReadBufferA(std::size_t buffer_index) con
 
 std::span<const u8> HLERequestContext::ReadBufferX(std::size_t buffer_index) const {
     static thread_local std::array read_buffer_x{
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead>(memory, 0, 0),
     };
 
     ASSERT_OR_EXECUTE_MSG(
@@ -376,14 +377,14 @@ std::span<const u8> HLERequestContext::ReadBufferX(std::size_t buffer_index) con
 
 std::span<const u8> HLERequestContext::ReadBuffer(std::size_t buffer_index) const {
     static thread_local std::array read_buffer_a{
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead>(memory, 0, 0),
     };
     static thread_local std::array read_buffer_x{
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead>(memory, 0, 0),
     };
 
     const bool is_buffer_a{BufferDescriptorA().size() > buffer_index &&

src/core/hle/service/hle_ipc.h

@@ -41,6 +41,8 @@ class KernelCore;
 class KHandleTable;
 class KProcess;
 class KServerSession;
+template <typename T>
+class KScopedAutoObject;
 class KThread;
 } // namespace Kernel
 
@@ -232,19 +234,19 @@ public:
     }
 
     [[nodiscard]] const std::vector<IPC::BufferDescriptorX>& BufferDescriptorX() const {
-        return buffer_x_desciptors;
+        return buffer_x_descriptors;
     }
 
     [[nodiscard]] const std::vector<IPC::BufferDescriptorABW>& BufferDescriptorA() const {
-        return buffer_a_desciptors;
+        return buffer_a_descriptors;
     }
 
     [[nodiscard]] const std::vector<IPC::BufferDescriptorABW>& BufferDescriptorB() const {
-        return buffer_b_desciptors;
+        return buffer_b_descriptors;
     }
 
     [[nodiscard]] const std::vector<IPC::BufferDescriptorC>& BufferDescriptorC() const {
-        return buffer_c_desciptors;
+        return buffer_c_descriptors;
     }
 
     [[nodiscard]] const IPC::DomainMessageHeader& GetDomainMessageHeader() const {
@@ -406,11 +408,11 @@ private:
     std::optional<IPC::HandleDescriptorHeader> handle_descriptor_header;
     std::optional<IPC::DataPayloadHeader> data_payload_header;
     std::optional<IPC::DomainMessageHeader> domain_message_header;
-    std::vector<IPC::BufferDescriptorX> buffer_x_desciptors;
-    std::vector<IPC::BufferDescriptorABW> buffer_a_desciptors;
-    std::vector<IPC::BufferDescriptorABW> buffer_b_desciptors;
-    std::vector<IPC::BufferDescriptorABW> buffer_w_desciptors;
-    std::vector<IPC::BufferDescriptorC> buffer_c_desciptors;
+    std::vector<IPC::BufferDescriptorX> buffer_x_descriptors;
+    std::vector<IPC::BufferDescriptorABW> buffer_a_descriptors;
+    std::vector<IPC::BufferDescriptorABW> buffer_b_descriptors;
+    std::vector<IPC::BufferDescriptorABW> buffer_w_descriptors;
+    std::vector<IPC::BufferDescriptorC> buffer_c_descriptors;
 
     u32_le command{};
     u64 pid{};

src/core/hle/service/nfc/common/amiibo_crypto.cpp

@@ -19,7 +19,7 @@ namespace Service::NFP::AmiiboCrypto {
 bool IsAmiiboValid(const EncryptedNTAG215File& ntag_file) {
     const auto& amiibo_data = ntag_file.user_memory;
     LOG_DEBUG(Service_NFP, "uuid_lock=0x{0:x}", ntag_file.static_lock);
-    LOG_DEBUG(Service_NFP, "compability_container=0x{0:x}", ntag_file.compability_container);
+    LOG_DEBUG(Service_NFP, "compatibility_container=0x{0:x}", ntag_file.compatibility_container);
     LOG_DEBUG(Service_NFP, "write_count={}", static_cast<u16>(amiibo_data.write_counter));
 
     LOG_DEBUG(Service_NFP, "character_id=0x{0:x}", amiibo_data.model_info.character_id);
@@ -49,7 +49,7 @@ bool IsAmiiboValid(const EncryptedNTAG215File& ntag_file) {
     if (ntag_file.static_lock != 0xE00F) {
         return false;
     }
-    if (ntag_file.compability_container != 0xEEFF10F1U) {
+    if (ntag_file.compatibility_container != 0xEEFF10F1U) {
         return false;
     }
     if (amiibo_data.model_info.tag_type != NFC::PackedTagType::Type2) {
@@ -78,7 +78,7 @@ NTAG215File NfcDataToEncodedData(const EncryptedNTAG215File& nfc_data) {
     encoded_data.uid_crc_check2 = nfc_data.uuid_crc_check2;
     encoded_data.internal_number = nfc_data.internal_number;
     encoded_data.static_lock = nfc_data.static_lock;
-    encoded_data.compability_container = nfc_data.compability_container;
+    encoded_data.compatibility_container = nfc_data.compatibility_container;
     encoded_data.hmac_data = nfc_data.user_memory.hmac_data;
     encoded_data.constant_value = nfc_data.user_memory.constant_value;
     encoded_data.write_counter = nfc_data.user_memory.write_counter;
@@ -112,7 +112,7 @@ EncryptedNTAG215File EncodedDataToNfcData(const NTAG215File& encoded_data) {
     nfc_data.uuid_crc_check2 = encoded_data.uid_crc_check2;
     nfc_data.internal_number = encoded_data.internal_number;
     nfc_data.static_lock = encoded_data.static_lock;
-    nfc_data.compability_container = encoded_data.compability_container;
+    nfc_data.compatibility_container = encoded_data.compatibility_container;
     nfc_data.user_memory.hmac_data = encoded_data.hmac_data;
     nfc_data.user_memory.constant_value = encoded_data.constant_value;
     nfc_data.user_memory.write_counter = encoded_data.write_counter;
@@ -257,7 +257,7 @@ void Cipher(const DerivedKeys& keys, const NTAG215File& in_data, NTAG215File& ou
     out_data.uid_crc_check2 = in_data.uid_crc_check2;
     out_data.internal_number = in_data.internal_number;
     out_data.static_lock = in_data.static_lock;
-    out_data.compability_container = in_data.compability_container;
+    out_data.compatibility_container = in_data.compatibility_container;
 
     out_data.constant_value = in_data.constant_value;
     out_data.write_counter = in_data.write_counter;

src/core/hle/service/nfc/common/device.cpp

@@ -75,7 +75,7 @@ void NfcDevice::NpadUpdate(Core::HID::ControllerTriggerType type) {
         return;
     }
 
-    if (!is_initalized) {
+    if (!is_initialized) {
         return;
     }
 
@@ -207,7 +207,7 @@ void NfcDevice::Initialize() {
         return;
     }
 
-    is_initalized = npad_device->AddNfcHandle();
+    is_initialized = npad_device->AddNfcHandle();
 }
 
 void NfcDevice::Finalize() {
@@ -226,7 +226,7 @@ void NfcDevice::Finalize() {
     }
 
     device_state = DeviceState::Unavailable;
-    is_initalized = false;
+    is_initialized = false;
 }
 
 Result NfcDevice::StartDetection(NfcProtocol allowed_protocol) {

src/core/hle/service/nfc/common/device.h

@@ -126,7 +126,7 @@ private:
     Kernel::KEvent* deactivate_event = nullptr;
     Kernel::KEvent* availability_change_event = nullptr;
 
-    bool is_initalized{};
+    bool is_initialized{};
     NfcProtocol allowed_protocols{};
     DeviceState device_state{DeviceState::Unavailable};
 

src/core/hle/service/nfp/nfp_types.h

@@ -243,12 +243,12 @@ static_assert(sizeof(EncryptedAmiiboFile) == 0x1F8, "AmiiboFile is an invalid si
 struct NTAG215File {
     u8 uid_crc_check2;
     u8 internal_number;
-    u16 static_lock;           // Set defined pages as read only
-    u32 compability_container; // Defines available memory
-    HashData hmac_data;        // Hash
-    u8 constant_value;         // Must be A5
-    u16_be write_counter;      // Number of times the amiibo has been written?
-    u8 amiibo_version;         // Amiibo file version
+    u16 static_lock;             // Set defined pages as read only
+    u32 compatibility_container; // Defines available memory
+    HashData hmac_data;          // Hash
+    u8 constant_value;           // Must be A5
+    u16_be write_counter;        // Number of times the amiibo has been written?
+    u8 amiibo_version;           // Amiibo file version
     AmiiboSettings settings;
     Service::Mii::Ver3StoreData owner_mii; // Mii data
     u64_be application_id;                 // Game id
@@ -278,7 +278,7 @@ struct EncryptedNTAG215File {
     u8 uuid_crc_check2;
     u8 internal_number;
     u16 static_lock;                 // Set defined pages as read only
-    u32 compability_container;       // Defines available memory
+    u32 compatibility_container;     // Defines available memory
     EncryptedAmiiboFile user_memory; // Writable data
     u32 dynamic_lock;                // Dynamic lock
     u32 CFG0;                        // Defines memory protected by password

src/core/hle/service/ns/language.cpp

@@ -2,7 +2,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include "core/hle/service/ns/language.h"
-#include "core/hle/service/set/set.h"
+#include "core/hle/service/set/settings_server.h"
 
 namespace Service::NS {
 

src/core/hle/service/ns/ns.cpp

@@ -16,7 +16,7 @@
 #include "core/hle/service/ns/ns.h"
 #include "core/hle/service/ns/pdm_qry.h"
 #include "core/hle/service/server_manager.h"
-#include "core/hle/service/set/set.h"
+#include "core/hle/service/set/settings_server.h"
 
 namespace Service::NS {
 

src/core/hle/service/nvdrv/core/container.cpp

@@ -2,27 +2,134 @@
 // SPDX-FileCopyrightText: 2022 Skyline Team and Contributors
 // SPDX-License-Identifier: GPL-3.0-or-later
 
+#include <atomic>
+#include <deque>
+#include <mutex>
+
+#include "core/hle/kernel/k_process.h"
 #include "core/hle/service/nvdrv/core/container.h"
+#include "core/hle/service/nvdrv/core/heap_mapper.h"
 #include "core/hle/service/nvdrv/core/nvmap.h"
 #include "core/hle/service/nvdrv/core/syncpoint_manager.h"
+#include "core/memory.h"
 #include "video_core/host1x/host1x.h"
 
 namespace Service::Nvidia::NvCore {
 
+Session::Session(size_t id_, Kernel::KProcess* process_, size_t smmu_id_)
+    : id{id_}, process{process_}, smmu_id{smmu_id_},
+      has_preallocated_area{}, mapper{}, is_active{} {}
+
+Session::~Session() = default;
+
 struct ContainerImpl {
-    explicit ContainerImpl(Tegra::Host1x::Host1x& host1x_)
-        : file{host1x_}, manager{host1x_}, device_file_data{} {}
+    explicit ContainerImpl(Container& core, Tegra::Host1x::Host1x& host1x_)
+        : host1x{host1x_}, file{core, host1x_}, manager{host1x_}, device_file_data{} {}
+    Tegra::Host1x::Host1x& host1x;
     NvMap file;
     SyncpointManager manager;
     Container::Host1xDeviceFileData device_file_data;
+    std::deque<Session> sessions;
+    size_t new_ids{};
+    std::deque<size_t> id_pool;
+    std::mutex session_guard;
 };
 
 Container::Container(Tegra::Host1x::Host1x& host1x_) {
-    impl = std::make_unique<ContainerImpl>(host1x_);
+    impl = std::make_unique<ContainerImpl>(*this, host1x_);
 }
 
 Container::~Container() = default;
 
+size_t Container::OpenSession(Kernel::KProcess* process) {
+    std::scoped_lock lk(impl->session_guard);
+    for (auto& session : impl->sessions) {
+        if (!session.is_active) {
+            continue;
+        }
+        if (session.process == process) {
+            return session.id;
+        }
+    }
+    size_t new_id{};
+    auto* memory_interface = &process->GetMemory();
+    auto& smmu = impl->host1x.MemoryManager();
+    auto smmu_id = smmu.RegisterProcess(memory_interface);
+    if (!impl->id_pool.empty()) {
+        new_id = impl->id_pool.front();
+        impl->id_pool.pop_front();
+        impl->sessions[new_id] = Session{new_id, process, smmu_id};
+    } else {
+        new_id = impl->new_ids++;
+        impl->sessions.emplace_back(new_id, process, smmu_id);
+    }
+    auto& session = impl->sessions[new_id];
+    session.is_active = true;
+    // Optimization
+    if (process->IsApplication()) {
+        auto& page_table = process->GetPageTable().GetBasePageTable();
+        auto heap_start = page_table.GetHeapRegionStart();
+
+        Kernel::KProcessAddress cur_addr = heap_start;
+        size_t region_size = 0;
+        VAddr region_start = 0;
+        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();
+
+            // check if this memory block is heap
+            if (svc_mem_info.state == Kernel::Svc::MemoryState::Normal) {
+                if (svc_mem_info.size > region_size) {
+                    region_size = svc_mem_info.size;
+                    region_start = svc_mem_info.base_address;
+                }
+            }
+
+            // Check if we're done.
+            const uintptr_t next_address = svc_mem_info.base_address + svc_mem_info.size;
+            if (next_address <= GetInteger(cur_addr)) {
+                break;
+            }
+
+            cur_addr = next_address;
+        }
+        session.has_preallocated_area = false;
+        auto start_region = (region_size >> 15) >= 1024 ? smmu.Allocate(region_size) : 0;
+        if (start_region != 0) {
+            session.mapper = std::make_unique<HeapMapper>(region_start, start_region, region_size,
+                                                          smmu_id, impl->host1x);
+            smmu.TrackContinuity(start_region, region_start, region_size, smmu_id);
+            session.has_preallocated_area = true;
+            LOG_CRITICAL(Debug, "Preallocation created!");
+        }
+    }
+    return new_id;
+}
+
+void Container::CloseSession(size_t id) {
+    std::scoped_lock lk(impl->session_guard);
+    auto& session = impl->sessions[id];
+    auto& smmu = impl->host1x.MemoryManager();
+    if (session.has_preallocated_area) {
+        const DAddr region_start = session.mapper->GetRegionStart();
+        const size_t region_size = session.mapper->GetRegionSize();
+        session.mapper.reset();
+        smmu.Free(region_start, region_size);
+        session.has_preallocated_area = false;
+    }
+    session.is_active = false;
+    smmu.UnregisterProcess(impl->sessions[id].smmu_id);
+    impl->id_pool.emplace_front(id);
+}
+
+Session* Container::GetSession(size_t id) {
+    std::atomic_thread_fence(std::memory_order_acquire);
+    return &impl->sessions[id];
+}
+
 NvMap& Container::GetNvMapFile() {
     return impl->file;
 }

src/core/hle/service/nvdrv/core/container.h

@@ -10,22 +10,49 @@
 
 #include "core/hle/service/nvdrv/nvdata.h"
 
+namespace Kernel {
+class KProcess;
+}
+
 namespace Tegra::Host1x {
 class Host1x;
 } // namespace Tegra::Host1x
 
 namespace Service::Nvidia::NvCore {
 
+class HeapMapper;
 class NvMap;
 class SyncpointManager;
 
 struct ContainerImpl;
 
+struct Session {
+    Session(size_t id_, Kernel::KProcess* process_, size_t smmu_id_);
+    ~Session();
+
+    Session(const Session&) = delete;
+    Session& operator=(const Session&) = delete;
+    Session(Session&&) = default;
+    Session& operator=(Session&&) = default;
+
+    size_t id;
+    Kernel::KProcess* process;
+    size_t smmu_id;
+    bool has_preallocated_area{};
+    std::unique_ptr<HeapMapper> mapper{};
+    bool is_active{};
+};
+
 class Container {
 public:
     explicit Container(Tegra::Host1x::Host1x& host1x);
     ~Container();
 
+    size_t OpenSession(Kernel::KProcess* process);
+    void CloseSession(size_t id);
+
+    Session* GetSession(size_t id);
+
     NvMap& GetNvMapFile();
 
     const NvMap& GetNvMapFile() const;

src/core/hle/service/nvdrv/core/heap_mapper.cpp

@@ -0,0 +1,175 @@
+// SPDX-FileCopyrightText: 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#include <mutex>
+
+#include <boost/container/small_vector.hpp>
+#define BOOST_NO_MT
+#include <boost/pool/detail/mutex.hpp>
+#undef BOOST_NO_MT
+#include <boost/icl/interval.hpp>
+#include <boost/icl/interval_base_set.hpp>
+#include <boost/icl/interval_set.hpp>
+#include <boost/icl/split_interval_map.hpp>
+#include <boost/pool/pool.hpp>
+#include <boost/pool/pool_alloc.hpp>
+#include <boost/pool/poolfwd.hpp>
+
+#include "core/hle/service/nvdrv/core/heap_mapper.h"
+#include "video_core/host1x/host1x.h"
+
+namespace boost {
+template <typename T>
+class fast_pool_allocator<T, default_user_allocator_new_delete, details::pool::null_mutex, 4096, 0>;
+}
+
+namespace Service::Nvidia::NvCore {
+
+using IntervalCompare = std::less<DAddr>;
+using IntervalInstance = boost::icl::interval_type_default<DAddr, std::less>;
+using IntervalAllocator = boost::fast_pool_allocator<DAddr>;
+using IntervalSet = boost::icl::interval_set<DAddr>;
+using IntervalType = typename IntervalSet::interval_type;
+
+template <typename Type>
+struct counter_add_functor : public boost::icl::identity_based_inplace_combine<Type> {
+    // types
+    typedef counter_add_functor<Type> type;
+    typedef boost::icl::identity_based_inplace_combine<Type> base_type;
+
+    // public member functions
+    void operator()(Type& current, const Type& added) const {
+        current += added;
+        if (current < base_type::identity_element()) {
+            current = base_type::identity_element();
+        }
+    }
+
+    // public static functions
+    static void version(Type&){};
+};
+
+using OverlapCombine = counter_add_functor<int>;
+using OverlapSection = boost::icl::inter_section<int>;
+using OverlapCounter = boost::icl::split_interval_map<DAddr, int>;
+
+struct HeapMapper::HeapMapperInternal {
+    HeapMapperInternal(Tegra::Host1x::Host1x& host1x) : device_memory{host1x.MemoryManager()} {}
+    ~HeapMapperInternal() = default;
+
+    template <typename Func>
+    void ForEachInOverlapCounter(OverlapCounter& current_range, VAddr cpu_addr, u64 size,
+                                 Func&& func) {
+        const DAddr start_address = cpu_addr;
+        const DAddr end_address = start_address + size;
+        const IntervalType search_interval{start_address, end_address};
+        auto it = current_range.lower_bound(search_interval);
+        if (it == current_range.end()) {
+            return;
+        }
+        auto end_it = current_range.upper_bound(search_interval);
+        for (; it != end_it; it++) {
+            auto& inter = it->first;
+            DAddr inter_addr_end = inter.upper();
+            DAddr inter_addr = inter.lower();
+            if (inter_addr_end > end_address) {
+                inter_addr_end = end_address;
+            }
+            if (inter_addr < start_address) {
+                inter_addr = start_address;
+            }
+            func(inter_addr, inter_addr_end, it->second);
+        }
+    }
+
+    void RemoveEachInOverlapCounter(OverlapCounter& current_range,
+                                    const IntervalType search_interval, int subtract_value) {
+        bool any_removals = false;
+        current_range.add(std::make_pair(search_interval, subtract_value));
+        do {
+            any_removals = false;
+            auto it = current_range.lower_bound(search_interval);
+            if (it == current_range.end()) {
+                return;
+            }
+            auto end_it = current_range.upper_bound(search_interval);
+            for (; it != end_it; it++) {
+                if (it->second <= 0) {
+                    any_removals = true;
+                    current_range.erase(it);
+                    break;
+                }
+            }
+        } while (any_removals);
+    }
+
+    IntervalSet base_set;
+    OverlapCounter mapping_overlaps;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
+    std::mutex guard;
+};
+
+HeapMapper::HeapMapper(VAddr start_vaddress, DAddr start_daddress, size_t size, size_t smmu_id,
+                       Tegra::Host1x::Host1x& host1x)
+    : m_vaddress{start_vaddress}, m_daddress{start_daddress}, m_size{size}, m_smmu_id{smmu_id} {
+    m_internal = std::make_unique<HeapMapperInternal>(host1x);
+}
+
+HeapMapper::~HeapMapper() {
+    m_internal->device_memory.Unmap(m_daddress, m_size);
+}
+
+DAddr HeapMapper::Map(VAddr start, size_t size) {
+    std::scoped_lock lk(m_internal->guard);
+    m_internal->base_set.clear();
+    const IntervalType interval{start, start + size};
+    m_internal->base_set.insert(interval);
+    m_internal->ForEachInOverlapCounter(m_internal->mapping_overlaps, start, size,
+                                        [this](VAddr start_addr, VAddr end_addr, int) {
+                                            const IntervalType other{start_addr, end_addr};
+                                            m_internal->base_set.subtract(other);
+                                        });
+    if (!m_internal->base_set.empty()) {
+        auto it = m_internal->base_set.begin();
+        auto end_it = m_internal->base_set.end();
+        for (; it != end_it; it++) {
+            const VAddr inter_addr_end = it->upper();
+            const VAddr inter_addr = it->lower();
+            const size_t offset = inter_addr - m_vaddress;
+            const size_t sub_size = inter_addr_end - inter_addr;
+            m_internal->device_memory.Map(m_daddress + offset, m_vaddress + offset, sub_size,
+                                          m_smmu_id);
+        }
+    }
+    m_internal->mapping_overlaps += std::make_pair(interval, 1);
+    m_internal->base_set.clear();
+    return m_daddress + (start - m_vaddress);
+}
+
+void HeapMapper::Unmap(VAddr start, size_t size) {
+    std::scoped_lock lk(m_internal->guard);
+    m_internal->base_set.clear();
+    m_internal->ForEachInOverlapCounter(m_internal->mapping_overlaps, start, size,
+                                        [this](VAddr start_addr, VAddr end_addr, int value) {
+                                            if (value <= 1) {
+                                                const IntervalType other{start_addr, end_addr};
+                                                m_internal->base_set.insert(other);
+                                            }
+                                        });
+    if (!m_internal->base_set.empty()) {
+        auto it = m_internal->base_set.begin();
+        auto end_it = m_internal->base_set.end();
+        for (; it != end_it; it++) {
+            const VAddr inter_addr_end = it->upper();
+            const VAddr inter_addr = it->lower();
+            const size_t offset = inter_addr - m_vaddress;
+            const size_t sub_size = inter_addr_end - inter_addr;
+            m_internal->device_memory.Unmap(m_daddress + offset, sub_size);
+        }
+    }
+    const IntervalType to_remove{start, start + size};
+    m_internal->RemoveEachInOverlapCounter(m_internal->mapping_overlaps, to_remove, -1);
+    m_internal->base_set.clear();
+}
+
+} // namespace Service::Nvidia::NvCore

src/core/hle/service/nvdrv/core/heap_mapper.h

@@ -0,0 +1,48 @@
+// SPDX-FileCopyrightText: 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+#include <memory>
+
+#include "common/common_types.h"
+
+namespace Tegra::Host1x {
+class Host1x;
+} // namespace Tegra::Host1x
+
+namespace Service::Nvidia::NvCore {
+
+class HeapMapper {
+public:
+    HeapMapper(VAddr start_vaddress, DAddr start_daddress, size_t size, size_t smmu_id,
+               Tegra::Host1x::Host1x& host1x);
+    ~HeapMapper();
+
+    bool IsInBounds(VAddr start, size_t size) const {
+        VAddr end = start + size;
+        return start >= m_vaddress && end <= (m_vaddress + m_size);
+    }
+
+    DAddr Map(VAddr start, size_t size);
+
+    void Unmap(VAddr start, size_t size);
+
+    DAddr GetRegionStart() const {
+        return m_daddress;
+    }
+
+    size_t GetRegionSize() const {
+        return m_size;
+    }
+
+private:
+    struct HeapMapperInternal;
+    VAddr m_vaddress;
+    DAddr m_daddress;
+    size_t m_size;
+    size_t m_smmu_id;
+    std::unique_ptr<HeapMapperInternal> m_internal;
+};
+
+} // namespace Service::Nvidia::NvCore

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

@@ -2,14 +2,19 @@
 // SPDX-FileCopyrightText: 2022 Skyline Team and Contributors
 // SPDX-License-Identifier: GPL-3.0-or-later
 
+#include <functional>
+
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "core/hle/service/nvdrv/core/container.h"
+#include "core/hle/service/nvdrv/core/heap_mapper.h"
 #include "core/hle/service/nvdrv/core/nvmap.h"
 #include "core/memory.h"
 #include "video_core/host1x/host1x.h"
 
 using Core::Memory::YUZU_PAGESIZE;
+constexpr size_t BIG_PAGE_SIZE = YUZU_PAGESIZE * 16;
 
 namespace Service::Nvidia::NvCore {
 NvMap::Handle::Handle(u64 size_, Id id_)
@@ -17,9 +22,8 @@ NvMap::Handle::Handle(u64 size_, Id id_)
     flags.raw = 0;
 }
 
-NvResult NvMap::Handle::Alloc(Flags pFlags, u32 pAlign, u8 pKind, u64 pAddress) {
+NvResult NvMap::Handle::Alloc(Flags pFlags, u32 pAlign, u8 pKind, u64 pAddress, size_t pSessionId) {
     std::scoped_lock lock(mutex);
-
     // Handles cannot be allocated twice
     if (allocated) {
         return NvResult::AccessDenied;
@@ -28,6 +32,7 @@ NvResult NvMap::Handle::Alloc(Flags pFlags, u32 pAlign, u8 pKind, u64 pAddress)
     flags = pFlags;
     kind = pKind;
     align = pAlign < YUZU_PAGESIZE ? YUZU_PAGESIZE : pAlign;
+    session_id = pSessionId;
 
     // This flag is only applicable for handles with an address passed
     if (pAddress) {
@@ -63,7 +68,7 @@ NvResult NvMap::Handle::Duplicate(bool internal_session) {
     return NvResult::Success;
 }
 
-NvMap::NvMap(Tegra::Host1x::Host1x& host1x_) : host1x{host1x_} {}
+NvMap::NvMap(Container& core_, Tegra::Host1x::Host1x& host1x_) : host1x{host1x_}, core{core_} {}
 
 void NvMap::AddHandle(std::shared_ptr<Handle> handle_description) {
     std::scoped_lock lock(handles_lock);
@@ -78,12 +83,30 @@ void NvMap::UnmapHandle(Handle& handle_description) {
         handle_description.unmap_queue_entry.reset();
     }
 
+    // Free and unmap the handle from Host1x GMMU
+    if (handle_description.pin_virt_address) {
+        host1x.GMMU().Unmap(static_cast<GPUVAddr>(handle_description.pin_virt_address),
+                            handle_description.aligned_size);
+        host1x.Allocator().Free(handle_description.pin_virt_address,
+                                static_cast<u32>(handle_description.aligned_size));
+        handle_description.pin_virt_address = 0;
+    }
+
     // Free and unmap the handle from the SMMU
-    host1x.MemoryManager().Unmap(static_cast<GPUVAddr>(handle_description.pin_virt_address),
-                                 handle_description.aligned_size);
-    host1x.Allocator().Free(handle_description.pin_virt_address,
-                            static_cast<u32>(handle_description.aligned_size));
-    handle_description.pin_virt_address = 0;
+    const size_t map_size = handle_description.aligned_size;
+    if (!handle_description.in_heap) {
+        auto& smmu = host1x.MemoryManager();
+        size_t aligned_up = Common::AlignUp(map_size, BIG_PAGE_SIZE);
+        smmu.Unmap(handle_description.d_address, map_size);
+        smmu.Free(handle_description.d_address, static_cast<size_t>(aligned_up));
+        handle_description.d_address = 0;
+        return;
+    }
+    const VAddr vaddress = handle_description.address;
+    auto* session = core.GetSession(handle_description.session_id);
+    session->mapper->Unmap(vaddress, map_size);
+    handle_description.d_address = 0;
+    handle_description.in_heap = false;
 }
 
 bool NvMap::TryRemoveHandle(const Handle& handle_description) {
@@ -124,22 +147,33 @@ std::shared_ptr<NvMap::Handle> NvMap::GetHandle(Handle::Id handle) {
     }
 }
 
-VAddr NvMap::GetHandleAddress(Handle::Id handle) {
+DAddr NvMap::GetHandleAddress(Handle::Id handle) {
     std::scoped_lock lock(handles_lock);
     try {
-        return handles.at(handle)->address;
+        return handles.at(handle)->d_address;
     } catch (std::out_of_range&) {
         return 0;
     }
 }
 
-u32 NvMap::PinHandle(NvMap::Handle::Id handle) {
+DAddr NvMap::PinHandle(NvMap::Handle::Id handle, bool low_area_pin) {
     auto handle_description{GetHandle(handle)};
     if (!handle_description) [[unlikely]] {
         return 0;
     }
 
     std::scoped_lock lock(handle_description->mutex);
+    const auto map_low_area = [&] {
+        if (handle_description->pin_virt_address == 0) {
+            auto& gmmu_allocator = host1x.Allocator();
+            auto& gmmu = host1x.GMMU();
+            u32 address =
+                gmmu_allocator.Allocate(static_cast<u32>(handle_description->aligned_size));
+            gmmu.Map(static_cast<GPUVAddr>(address), handle_description->d_address,
+                     handle_description->aligned_size);
+            handle_description->pin_virt_address = address;
+        }
+    };
     if (!handle_description->pins) {
         // If we're in the unmap queue we can just remove ourselves and return since we're already
         // mapped
@@ -151,37 +185,58 @@ u32 NvMap::PinHandle(NvMap::Handle::Id handle) {
                 unmap_queue.erase(*handle_description->unmap_queue_entry);
                 handle_description->unmap_queue_entry.reset();
 
+                if (low_area_pin) {
+                    map_low_area();
+                    handle_description->pins++;
+                    return static_cast<DAddr>(handle_description->pin_virt_address);
+                }
+
                 handle_description->pins++;
-                return handle_description->pin_virt_address;
+                return handle_description->d_address;
             }
         }
 
+        using namespace std::placeholders;
         // If not then allocate some space and map it
-        u32 address{};
-        auto& smmu_allocator = host1x.Allocator();
-        auto& smmu_memory_manager = host1x.MemoryManager();
-        while ((address = smmu_allocator.Allocate(
-                    static_cast<u32>(handle_description->aligned_size))) == 0) {
-            // Free handles until the allocation succeeds
-            std::scoped_lock queueLock(unmap_queue_lock);
-            if (auto freeHandleDesc{unmap_queue.front()}) {
-                // Handles in the unmap queue are guaranteed not to be pinned so don't bother
-                // checking if they are before unmapping
-                std::scoped_lock freeLock(freeHandleDesc->mutex);
-                if (handle_description->pin_virt_address)
-                    UnmapHandle(*freeHandleDesc);
-            } else {
-                LOG_CRITICAL(Service_NVDRV, "Ran out of SMMU address space!");
+        DAddr address{};
+        auto& smmu = host1x.MemoryManager();
+        auto* session = core.GetSession(handle_description->session_id);
+        const VAddr vaddress = handle_description->address;
+        const size_t map_size = handle_description->aligned_size;
+        if (session->has_preallocated_area && session->mapper->IsInBounds(vaddress, map_size)) {
+            handle_description->d_address = session->mapper->Map(vaddress, map_size);
+            handle_description->in_heap = true;
+        } else {
+            size_t aligned_up = Common::AlignUp(map_size, BIG_PAGE_SIZE);
+            while ((address = smmu.Allocate(aligned_up)) == 0) {
+                // Free handles until the allocation succeeds
+                std::scoped_lock queueLock(unmap_queue_lock);
+                if (auto freeHandleDesc{unmap_queue.front()}) {
+                    // Handles in the unmap queue are guaranteed not to be pinned so don't bother
+                    // checking if they are before unmapping
+                    std::scoped_lock freeLock(freeHandleDesc->mutex);
+                    if (handle_description->d_address)
+                        UnmapHandle(*freeHandleDesc);
+                } else {
+                    LOG_CRITICAL(Service_NVDRV, "Ran out of SMMU address space!");
+                }
             }
-        }
 
-        smmu_memory_manager.Map(static_cast<GPUVAddr>(address), handle_description->address,
-                                handle_description->aligned_size);
-        handle_description->pin_virt_address = address;
+            handle_description->d_address = address;
+            smmu.Map(address, vaddress, map_size, session->smmu_id, true);
+            handle_description->in_heap = false;
+        }
+    }
+
+    if (low_area_pin) {
+        map_low_area();
     }
 
     handle_description->pins++;
-    return handle_description->pin_virt_address;
+    if (low_area_pin) {
+        return static_cast<DAddr>(handle_description->pin_virt_address);
+    }
+    return handle_description->d_address;
 }
 
 void NvMap::UnpinHandle(Handle::Id handle) {
@@ -232,7 +287,7 @@ std::optional<NvMap::FreeInfo> NvMap::FreeHandle(Handle::Id handle, bool interna
                 LOG_WARNING(Service_NVDRV, "User duplicate count imbalance detected!");
             } else if (handle_description->dupes == 0) {
                 // Force unmap the handle
-                if (handle_description->pin_virt_address) {
+                if (handle_description->d_address) {
                     std::scoped_lock queueLock(unmap_queue_lock);
                     UnmapHandle(*handle_description);
                 }

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

@@ -25,6 +25,8 @@ class Host1x;
 } // namespace Tegra
 
 namespace Service::Nvidia::NvCore {
+
+class Container;
 /**
  * @brief The nvmap core class holds the global state for nvmap and provides methods to manage
  * handles
@@ -48,7 +50,7 @@ public:
         using Id = u32;
         Id id; //!< A globally unique identifier for this handle
 
-        s32 pins{};
+        s64 pins{};
         u32 pin_virt_address{};
         std::optional<typename std::list<std::shared_ptr<Handle>>::iterator> unmap_queue_entry{};
 
@@ -61,15 +63,18 @@ public:
         } flags{};
         static_assert(sizeof(Flags) == sizeof(u32));
 
-        u64 address{}; //!< The memory location in the guest's AS that this handle corresponds to,
-                       //!< this can also be in the nvdrv tmem
+        VAddr address{}; //!< The memory location in the guest's AS that this handle corresponds to,
+                         //!< this can also be in the nvdrv tmem
         bool is_shared_mem_mapped{}; //!< If this nvmap has been mapped with the MapSharedMem IPC
                                      //!< call
 
         u8 kind{};        //!< Used for memory compression
         bool allocated{}; //!< If the handle has been allocated with `Alloc`
+        bool in_heap{};
+        size_t session_id{};
 
-        u64 dma_map_addr{}; //! remove me after implementing pinning.
+        DAddr d_address{}; //!< The memory location in the device's AS that this handle corresponds
+                           //!< to, this can also be in the nvdrv tmem
 
         Handle(u64 size, Id id);
 
@@ -77,7 +82,8 @@ public:
          * @brief Sets up the handle with the given memory config, can allocate memory from the tmem
          * if a 0 address is passed
          */
-        [[nodiscard]] NvResult Alloc(Flags pFlags, u32 pAlign, u8 pKind, u64 pAddress);
+        [[nodiscard]] NvResult Alloc(Flags pFlags, u32 pAlign, u8 pKind, u64 pAddress,
+                                     size_t pSessionId);
 
         /**
          * @brief Increases the dupe counter of the handle for the given session
@@ -108,7 +114,7 @@ public:
         bool can_unlock;   //!< If the address region is ready to be unlocked
     };
 
-    explicit NvMap(Tegra::Host1x::Host1x& host1x);
+    explicit NvMap(Container& core, Tegra::Host1x::Host1x& host1x);
 
     /**
      * @brief Creates an unallocated handle of the given size
@@ -117,7 +123,7 @@ public:
 
     std::shared_ptr<Handle> GetHandle(Handle::Id handle);
 
-    VAddr GetHandleAddress(Handle::Id handle);
+    DAddr GetHandleAddress(Handle::Id handle);
 
     /**
      * @brief Maps a handle into the SMMU address space
@@ -125,7 +131,7 @@ public:
      * number of calls to `UnpinHandle`
      * @return The SMMU virtual address that the handle has been mapped to
      */
-    u32 PinHandle(Handle::Id handle);
+    DAddr PinHandle(Handle::Id handle, bool low_area_pin);
 
     /**
      * @brief When this has been called an equal number of times to `PinHandle` for the supplied
@@ -172,5 +178,7 @@ private:
      * @return If the handle was removed from the map
      */
     bool TryRemoveHandle(const Handle& handle_description);
+
+    Container& core;
 };
 } // namespace Service::Nvidia::NvCore

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

@@ -62,7 +62,7 @@ public:
      * Called once a device is opened
      * @param fd The device fd
      */
-    virtual void OnOpen(DeviceFD fd) = 0;
+    virtual void OnOpen(size_t session_id, DeviceFD fd) = 0;
 
     /**
      * Called once a device is closed

src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp

@@ -35,14 +35,14 @@ NvResult nvdisp_disp0::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> in
     return NvResult::NotImplemented;
 }
 
-void nvdisp_disp0::OnOpen(DeviceFD fd) {}
+void nvdisp_disp0::OnOpen(size_t session_id, DeviceFD fd) {}
 void nvdisp_disp0::OnClose(DeviceFD fd) {}
 
 void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, android::PixelFormat format, u32 width,
                         u32 height, u32 stride, android::BufferTransformFlags transform,
                         const Common::Rectangle<int>& crop_rect,
                         std::array<Service::Nvidia::NvFence, 4>& fences, u32 num_fences) {
-    const VAddr addr = nvmap.GetHandleAddress(buffer_handle);
+    const DAddr addr = nvmap.GetHandleAddress(buffer_handle);
     LOG_TRACE(Service,
               "Drawing from address {:X} offset {:08X} Width {} Height {} Stride {} Format {}",
               addr, offset, width, height, stride, format);

src/core/hle/service/nvdrv/devices/nvdisp_disp0.h

@@ -32,7 +32,7 @@ public:
     NvResult Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, std::span<u8> output,
                     std::span<u8> inline_output) override;
 
-    void OnOpen(DeviceFD fd) override;
+    void OnOpen(size_t session_id, DeviceFD fd) override;
     void OnClose(DeviceFD fd) override;
 
     /// Performs a screen flip, drawing the buffer pointed to by the handle.

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

@@ -86,7 +86,7 @@ NvResult nvhost_as_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> i
     return NvResult::NotImplemented;
 }
 
-void nvhost_as_gpu::OnOpen(DeviceFD fd) {}
+void nvhost_as_gpu::OnOpen(size_t session_id, DeviceFD fd) {}
 void nvhost_as_gpu::OnClose(DeviceFD fd) {}
 
 NvResult nvhost_as_gpu::AllocAsEx(IoctlAllocAsEx& params) {
@@ -206,6 +206,8 @@ void nvhost_as_gpu::FreeMappingLocked(u64 offset) {
                        static_cast<u32>(aligned_size >> page_size_bits));
     }
 
+    nvmap.UnpinHandle(mapping->handle);
+
     // Sparse mappings shouldn't be fully unmapped, just returned to their sparse state
     // Only FreeSpace can unmap them fully
     if (mapping->sparse_alloc) {
@@ -293,12 +295,12 @@ NvResult nvhost_as_gpu::Remap(std::span<IoctlRemapEntry> entries) {
                 return NvResult::BadValue;
             }
 
-            VAddr cpu_address{static_cast<VAddr>(
-                handle->address +
-                (static_cast<u64>(entry.handle_offset_big_pages) << vm.big_page_size_bits))};
+            DAddr base = nvmap.PinHandle(entry.handle, false);
+            DAddr device_address{static_cast<DAddr>(
+                base + (static_cast<u64>(entry.handle_offset_big_pages) << vm.big_page_size_bits))};
 
-            gmmu->Map(virtual_address, cpu_address, size, static_cast<Tegra::PTEKind>(entry.kind),
-                      use_big_pages);
+            gmmu->Map(virtual_address, device_address, size,
+                      static_cast<Tegra::PTEKind>(entry.kind), use_big_pages);
         }
     }
 
@@ -331,9 +333,9 @@ NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
             }
 
             u64 gpu_address{static_cast<u64>(params.offset + params.buffer_offset)};
-            VAddr cpu_address{mapping->ptr + params.buffer_offset};
+            VAddr device_address{mapping->ptr + params.buffer_offset};
 
-            gmmu->Map(gpu_address, cpu_address, params.mapping_size,
+            gmmu->Map(gpu_address, device_address, params.mapping_size,
                       static_cast<Tegra::PTEKind>(params.kind), mapping->big_page);
 
             return NvResult::Success;
@@ -349,7 +351,8 @@ NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
         return NvResult::BadValue;
     }
 
-    VAddr cpu_address{static_cast<VAddr>(handle->address + params.buffer_offset)};
+    DAddr device_address{
+        static_cast<DAddr>(nvmap.PinHandle(params.handle, false) + params.buffer_offset)};
     u64 size{params.mapping_size ? params.mapping_size : handle->orig_size};
 
     bool big_page{[&]() {
@@ -373,15 +376,14 @@ NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
         }
 
         const bool use_big_pages = alloc->second.big_pages && big_page;
-        gmmu->Map(params.offset, cpu_address, size, static_cast<Tegra::PTEKind>(params.kind),
+        gmmu->Map(params.offset, device_address, size, static_cast<Tegra::PTEKind>(params.kind),
                   use_big_pages);
 
-        auto mapping{std::make_shared<Mapping>(cpu_address, params.offset, size, true,
-                                               use_big_pages, alloc->second.sparse)};
+        auto mapping{std::make_shared<Mapping>(params.handle, device_address, params.offset, size,
+                                               true, use_big_pages, alloc->second.sparse)};
         alloc->second.mappings.push_back(mapping);
         mapping_map[params.offset] = mapping;
     } else {
-
         auto& allocator{big_page ? *vm.big_page_allocator : *vm.small_page_allocator};
         u32 page_size{big_page ? vm.big_page_size : VM::YUZU_PAGESIZE};
         u32 page_size_bits{big_page ? vm.big_page_size_bits : VM::PAGE_SIZE_BITS};
@@ -394,11 +396,11 @@ NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
             return NvResult::InsufficientMemory;
         }
 
-        gmmu->Map(params.offset, cpu_address, Common::AlignUp(size, page_size),
+        gmmu->Map(params.offset, device_address, Common::AlignUp(size, page_size),
                   static_cast<Tegra::PTEKind>(params.kind), big_page);
 
-        auto mapping{
-            std::make_shared<Mapping>(cpu_address, params.offset, size, false, big_page, false)};
+        auto mapping{std::make_shared<Mapping>(params.handle, device_address, params.offset, size,
+                                               false, big_page, false)};
         mapping_map[params.offset] = mapping;
     }
 
@@ -433,6 +435,8 @@ NvResult nvhost_as_gpu::UnmapBuffer(IoctlUnmapBuffer& params) {
             gmmu->Unmap(params.offset, mapping->size);
         }
 
+        nvmap.UnpinHandle(mapping->handle);
+
         mapping_map.erase(params.offset);
     } catch (const std::out_of_range&) {
         LOG_WARNING(Service_NVDRV, "Couldn't find region to unmap at 0x{:X}", params.offset);

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

@@ -55,7 +55,7 @@ public:
     NvResult Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, std::span<u8> output,
                     std::span<u8> inline_output) override;
 
-    void OnOpen(DeviceFD fd) override;
+    void OnOpen(size_t session_id, DeviceFD fd) override;
     void OnClose(DeviceFD fd) override;
 
     Kernel::KEvent* QueryEvent(u32 event_id) override;
@@ -90,7 +90,7 @@ private:
             u64_le align;
         };
     };
-    static_assert(sizeof(IoctlAllocSpace) == 24, "IoctlInitalizeEx is incorrect size");
+    static_assert(sizeof(IoctlAllocSpace) == 24, "IoctlInitializeEx is incorrect size");
 
     struct IoctlFreeSpace {
         u64_le offset{};
@@ -159,16 +159,18 @@ private:
     NvCore::NvMap& nvmap;
 
     struct Mapping {
-        VAddr ptr;
+        NvCore::NvMap::Handle::Id handle;
+        DAddr ptr;
         u64 offset;
         u64 size;
         bool fixed;
         bool big_page; // Only valid if fixed == false
         bool sparse_alloc;
 
-        Mapping(VAddr ptr_, u64 offset_, u64 size_, bool fixed_, bool big_page_, bool sparse_alloc_)
-            : ptr(ptr_), offset(offset_), size(size_), fixed(fixed_), big_page(big_page_),
-              sparse_alloc(sparse_alloc_) {}
+        Mapping(NvCore::NvMap::Handle::Id handle_, DAddr ptr_, u64 offset_, u64 size_, bool fixed_,
+                bool big_page_, bool sparse_alloc_)
+            : handle(handle_), ptr(ptr_), offset(offset_), size(size_), fixed(fixed_),
+              big_page(big_page_), sparse_alloc(sparse_alloc_) {}
     };
 
     struct Allocation {
@@ -212,9 +214,6 @@ private:
         bool initialised{};
     } vm;
     std::shared_ptr<Tegra::MemoryManager> gmmu;
-
-    // s32 channel{};
-    // u32 big_page_size{VM::DEFAULT_BIG_PAGE_SIZE};
 };
 
 } // namespace Service::Nvidia::Devices

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

@@ -76,7 +76,7 @@ NvResult nvhost_ctrl::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> inp
     return NvResult::NotImplemented;
 }
 
-void nvhost_ctrl::OnOpen(DeviceFD fd) {}
+void nvhost_ctrl::OnOpen(size_t session_id, DeviceFD fd) {}
 
 void nvhost_ctrl::OnClose(DeviceFD fd) {}
 

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

@@ -32,7 +32,7 @@ public:
     NvResult Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, std::span<u8> output,
                     std::span<u8> inline_output) override;
 
-    void OnOpen(DeviceFD fd) override;
+    void OnOpen(size_t session_id, DeviceFD fd) override;
     void OnClose(DeviceFD fd) override;
 
     Kernel::KEvent* QueryEvent(u32 event_id) override;

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

@@ -15,7 +15,7 @@ namespace Service::Nvidia::Devices {
 nvhost_ctrl_gpu::nvhost_ctrl_gpu(Core::System& system_, EventInterface& events_interface_)
     : nvdevice{system_}, events_interface{events_interface_} {
     error_notifier_event = events_interface.CreateEvent("CtrlGpuErrorNotifier");
-    unknown_event = events_interface.CreateEvent("CtrlGpuUknownEvent");
+    unknown_event = events_interface.CreateEvent("CtrlGpuUnknownEvent");
 }
 nvhost_ctrl_gpu::~nvhost_ctrl_gpu() {
     events_interface.FreeEvent(error_notifier_event);
@@ -82,7 +82,7 @@ NvResult nvhost_ctrl_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8>
     return NvResult::NotImplemented;
 }
 
-void nvhost_ctrl_gpu::OnOpen(DeviceFD fd) {}
+void nvhost_ctrl_gpu::OnOpen(size_t session_id, DeviceFD fd) {}
 void nvhost_ctrl_gpu::OnClose(DeviceFD fd) {}
 
 NvResult nvhost_ctrl_gpu::GetCharacteristics1(IoctlCharacteristics& params) {

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

@@ -28,7 +28,7 @@ public:
     NvResult Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, std::span<u8> output,
                     std::span<u8> inline_output) override;
 
-    void OnOpen(DeviceFD fd) override;
+    void OnOpen(size_t session_id, DeviceFD fd) override;
     void OnClose(DeviceFD fd) override;
 
     Kernel::KEvent* QueryEvent(u32 event_id) override;

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

@@ -120,7 +120,7 @@ NvResult nvhost_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     return NvResult::NotImplemented;
 }
 
-void nvhost_gpu::OnOpen(DeviceFD fd) {}
+void nvhost_gpu::OnOpen(size_t session_id, DeviceFD fd) {}
 void nvhost_gpu::OnClose(DeviceFD fd) {}
 
 NvResult nvhost_gpu::SetNVMAPfd(IoctlSetNvmapFD& params) {

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

@@ -47,7 +47,7 @@ public:
     NvResult Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, std::span<u8> output,
                     std::span<u8> inline_output) override;
 
-    void OnOpen(DeviceFD fd) override;
+    void OnOpen(size_t session_id, DeviceFD fd) override;
     void OnClose(DeviceFD fd) override;
 
     Kernel::KEvent* QueryEvent(u32 event_id) override;

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

@@ -35,7 +35,7 @@ NvResult nvhost_nvdec::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> in
         case 0x7:
             return WrapFixed(this, &nvhost_nvdec::SetSubmitTimeout, input, output);
         case 0x9:
-            return WrapFixedVariable(this, &nvhost_nvdec::MapBuffer, input, output);
+            return WrapFixedVariable(this, &nvhost_nvdec::MapBuffer, input, output, fd);
         case 0xa:
             return WrapFixedVariable(this, &nvhost_nvdec::UnmapBuffer, input, output);
         default:
@@ -68,9 +68,10 @@ NvResult nvhost_nvdec::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> in
     return NvResult::NotImplemented;
 }
 
-void nvhost_nvdec::OnOpen(DeviceFD fd) {
+void nvhost_nvdec::OnOpen(size_t session_id, DeviceFD fd) {
     LOG_INFO(Service_NVDRV, "NVDEC video stream started");
     system.SetNVDECActive(true);
+    sessions[fd] = session_id;
 }
 
 void nvhost_nvdec::OnClose(DeviceFD fd) {
@@ -81,6 +82,10 @@ void nvhost_nvdec::OnClose(DeviceFD fd) {
         system.GPU().ClearCdmaInstance(iter->second);
     }
     system.SetNVDECActive(false);
+    auto it = sessions.find(fd);
+    if (it != sessions.end()) {
+        sessions.erase(it);
+    }
 }
 
 } // namespace Service::Nvidia::Devices

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

@@ -20,7 +20,7 @@ public:
     NvResult Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, std::span<u8> output,
                     std::span<u8> inline_output) override;
 
-    void OnOpen(DeviceFD fd) override;
+    void OnOpen(size_t session_id, DeviceFD fd) override;
     void OnClose(DeviceFD fd) override;
 };
 

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

@@ -8,6 +8,7 @@
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "core/core.h"
+#include "core/hle/kernel/k_process.h"
 #include "core/hle/service/nvdrv/core/container.h"
 #include "core/hle/service/nvdrv/core/nvmap.h"
 #include "core/hle/service/nvdrv/core/syncpoint_manager.h"
@@ -95,6 +96,8 @@ NvResult nvhost_nvdec_common::Submit(IoctlSubmit& params, std::span<u8> data, De
     offset += SliceVectors(data, fence_thresholds, params.fence_count, offset);
 
     auto& gpu = system.GPU();
+    auto* session = core.GetSession(sessions[fd]);
+
     if (gpu.UseNvdec()) {
         for (std::size_t i = 0; i < syncpt_increments.size(); i++) {
             const SyncptIncr& syncpt_incr = syncpt_increments[i];
@@ -106,8 +109,8 @@ NvResult nvhost_nvdec_common::Submit(IoctlSubmit& params, std::span<u8> data, De
         const auto object = nvmap.GetHandle(cmd_buffer.memory_id);
         ASSERT_OR_EXECUTE(object, return NvResult::InvalidState;);
         Tegra::ChCommandHeaderList cmdlist(cmd_buffer.word_count);
-        system.ApplicationMemory().ReadBlock(object->address + cmd_buffer.offset, cmdlist.data(),
-                                             cmdlist.size() * sizeof(u32));
+        session->process->GetMemory().ReadBlock(object->address + cmd_buffer.offset, cmdlist.data(),
+                                                cmdlist.size() * sizeof(u32));
         gpu.PushCommandBuffer(core.Host1xDeviceFile().fd_to_id[fd], cmdlist);
     }
     // Some games expect command_buffers to be written back
@@ -133,10 +136,12 @@ NvResult nvhost_nvdec_common::GetWaitbase(IoctlGetWaitbase& params) {
     return NvResult::Success;
 }
 
-NvResult nvhost_nvdec_common::MapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries) {
+NvResult nvhost_nvdec_common::MapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries,
+                                        DeviceFD fd) {
     const size_t num_entries = std::min(params.num_entries, static_cast<u32>(entries.size()));
     for (size_t i = 0; i < num_entries; i++) {
-        entries[i].map_address = nvmap.PinHandle(entries[i].map_handle);
+        DAddr pin_address = nvmap.PinHandle(entries[i].map_handle, true);
+        entries[i].map_address = static_cast<u32>(pin_address);
     }
 
     return NvResult::Success;

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

@@ -4,7 +4,9 @@
 #pragma once
 
 #include <deque>
+#include <unordered_map>
 #include <vector>
+
 #include "common/common_types.h"
 #include "common/swap.h"
 #include "core/hle/service/nvdrv/core/syncpoint_manager.h"
@@ -111,7 +113,7 @@ protected:
     NvResult Submit(IoctlSubmit& params, std::span<u8> input, DeviceFD fd);
     NvResult GetSyncpoint(IoctlGetSyncpoint& params);
     NvResult GetWaitbase(IoctlGetWaitbase& params);
-    NvResult MapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries);
+    NvResult MapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries, DeviceFD fd);
     NvResult UnmapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries);
     NvResult SetSubmitTimeout(u32 timeout);
 
@@ -125,6 +127,7 @@ protected:
     NvCore::NvMap& nvmap;
     NvCore::ChannelType channel_type;
     std::array<u32, MaxSyncPoints> device_syncpoints{};
+    std::unordered_map<DeviceFD, size_t> sessions;
 };
 }; // namespace Devices
 } // namespace Service::Nvidia

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

@@ -44,7 +44,7 @@ NvResult nvhost_nvjpg::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> in
     return NvResult::NotImplemented;
 }
 
-void nvhost_nvjpg::OnOpen(DeviceFD fd) {}
+void nvhost_nvjpg::OnOpen(size_t session_id, DeviceFD fd) {}
 void nvhost_nvjpg::OnClose(DeviceFD fd) {}
 
 NvResult nvhost_nvjpg::SetNVMAPfd(IoctlSetNvmapFD& params) {

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

@@ -22,7 +22,7 @@ public:
     NvResult Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, std::span<u8> output,
                     std::span<u8> inline_output) override;
 
-    void OnOpen(DeviceFD fd) override;
+    void OnOpen(size_t session_id, DeviceFD fd) override;
     void OnClose(DeviceFD fd) override;
 
 private:

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

@@ -33,7 +33,7 @@ NvResult nvhost_vic::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpu
         case 0x3:
             return WrapFixed(this, &nvhost_vic::GetWaitbase, input, output);
         case 0x9:
-            return WrapFixedVariable(this, &nvhost_vic::MapBuffer, input, output);
+            return WrapFixedVariable(this, &nvhost_vic::MapBuffer, input, output, fd);
         case 0xa:
             return WrapFixedVariable(this, &nvhost_vic::UnmapBuffer, input, output);
         default:
@@ -68,7 +68,9 @@ NvResult nvhost_vic::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     return NvResult::NotImplemented;
 }
 
-void nvhost_vic::OnOpen(DeviceFD fd) {}
+void nvhost_vic::OnOpen(size_t session_id, DeviceFD fd) {
+    sessions[fd] = session_id;
+}
 
 void nvhost_vic::OnClose(DeviceFD fd) {
     auto& host1x_file = core.Host1xDeviceFile();
@@ -76,6 +78,10 @@ void nvhost_vic::OnClose(DeviceFD fd) {
     if (iter != host1x_file.fd_to_id.end()) {
         system.GPU().ClearCdmaInstance(iter->second);
     }
+    auto it = sessions.find(fd);
+    if (it != sessions.end()) {
+        sessions.erase(it);
+    }
 }
 
 } // namespace Service::Nvidia::Devices

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

@@ -19,7 +19,7 @@ public:
     NvResult Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, std::span<u8> output,
                     std::span<u8> inline_output) override;
 
-    void OnOpen(DeviceFD fd) override;
+    void OnOpen(size_t session_id, DeviceFD fd) override;
     void OnClose(DeviceFD fd) override;
 };
 } // namespace Service::Nvidia::Devices

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

@@ -36,9 +36,9 @@ NvResult nvmap::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> input,
         case 0x3:
             return WrapFixed(this, &nvmap::IocFromId, input, output);
         case 0x4:
-            return WrapFixed(this, &nvmap::IocAlloc, input, output);
+            return WrapFixed(this, &nvmap::IocAlloc, input, output, fd);
         case 0x5:
-            return WrapFixed(this, &nvmap::IocFree, input, output);
+            return WrapFixed(this, &nvmap::IocFree, input, output, fd);
         case 0x9:
             return WrapFixed(this, &nvmap::IocParam, input, output);
         case 0xe:
@@ -67,8 +67,15 @@ NvResult nvmap::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, st
     return NvResult::NotImplemented;
 }
 
-void nvmap::OnOpen(DeviceFD fd) {}
-void nvmap::OnClose(DeviceFD fd) {}
+void nvmap::OnOpen(size_t session_id, DeviceFD fd) {
+    sessions[fd] = session_id;
+}
+void nvmap::OnClose(DeviceFD fd) {
+    auto it = sessions.find(fd);
+    if (it != sessions.end()) {
+        sessions.erase(it);
+    }
+}
 
 NvResult nvmap::IocCreate(IocCreateParams& params) {
     LOG_DEBUG(Service_NVDRV, "called, size=0x{:08X}", params.size);
@@ -87,7 +94,7 @@ NvResult nvmap::IocCreate(IocCreateParams& params) {
     return NvResult::Success;
 }
 
-NvResult nvmap::IocAlloc(IocAllocParams& params) {
+NvResult nvmap::IocAlloc(IocAllocParams& params, DeviceFD fd) {
     LOG_DEBUG(Service_NVDRV, "called, addr={:X}", params.address);
 
     if (!params.handle) {
@@ -116,15 +123,15 @@ NvResult nvmap::IocAlloc(IocAllocParams& params) {
         return NvResult::InsufficientMemory;
     }
 
-    const auto result =
-        handle_description->Alloc(params.flags, params.align, params.kind, params.address);
+    const auto result = handle_description->Alloc(params.flags, params.align, params.kind,
+                                                  params.address, sessions[fd]);
     if (result != NvResult::Success) {
         LOG_CRITICAL(Service_NVDRV, "Object failed to allocate, handle={:08X}", params.handle);
         return result;
     }
     bool is_out_io{};
-    ASSERT(system.ApplicationProcess()
-               ->GetPageTable()
+    auto process = container.GetSession(sessions[fd])->process;
+    ASSERT(process->GetPageTable()
                .LockForMapDeviceAddressSpace(&is_out_io, handle_description->address,
                                              handle_description->size,
                                              Kernel::KMemoryPermission::None, true, false)
@@ -224,7 +231,7 @@ NvResult nvmap::IocParam(IocParamParams& params) {
     return NvResult::Success;
 }
 
-NvResult nvmap::IocFree(IocFreeParams& params) {
+NvResult nvmap::IocFree(IocFreeParams& params, DeviceFD fd) {
     LOG_DEBUG(Service_NVDRV, "called");
 
     if (!params.handle) {
@@ -233,9 +240,9 @@ NvResult nvmap::IocFree(IocFreeParams& params) {
     }
 
     if (auto freeInfo{file.FreeHandle(params.handle, false)}) {
+        auto process = container.GetSession(sessions[fd])->process;
         if (freeInfo->can_unlock) {
-            ASSERT(system.ApplicationProcess()
-                       ->GetPageTable()
+            ASSERT(process->GetPageTable()
                        .UnlockForDeviceAddressSpace(freeInfo->address, freeInfo->size)
                        .IsSuccess());
         }

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

@@ -33,7 +33,7 @@ public:
     NvResult Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, std::span<u8> output,
                     std::span<u8> inline_output) override;
 
-    void OnOpen(DeviceFD fd) override;
+    void OnOpen(size_t session_id, DeviceFD fd) override;
     void OnClose(DeviceFD fd) override;
 
     enum class HandleParameterType : u32_le {
@@ -100,11 +100,11 @@ public:
     static_assert(sizeof(IocGetIdParams) == 8, "IocGetIdParams has wrong size");
 
     NvResult IocCreate(IocCreateParams& params);
-    NvResult IocAlloc(IocAllocParams& params);
+    NvResult IocAlloc(IocAllocParams& params, DeviceFD fd);
     NvResult IocGetId(IocGetIdParams& params);
     NvResult IocFromId(IocFromIdParams& params);
     NvResult IocParam(IocParamParams& params);
-    NvResult IocFree(IocFreeParams& params);
+    NvResult IocFree(IocFreeParams& params, DeviceFD fd);
 
 private:
     /// Id to use for the next handle that is created.
@@ -115,6 +115,7 @@ private:
 
     NvCore::Container& container;
     NvCore::NvMap& file;
+    std::unordered_map<DeviceFD, size_t> sessions;
 };
 
 } // namespace Service::Nvidia::Devices

src/core/hle/service/nvdrv/nvdata.h

@@ -51,7 +51,7 @@ enum class NvResult : u32 {
     DispNoDisplaysAttached = 0x20003,
     DispModeNotSupported = 0x20004,
     DispNotFound = 0x20005,
-    DispAttachDissallowed = 0x20006,
+    DispAttachDisallowed = 0x20006,
     DispTypeNotSupported = 0x20007,
     DispAuthenticationFailed = 0x20008,
     DispNotAttached = 0x20009,

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

@@ -45,13 +45,22 @@ void EventInterface::FreeEvent(Kernel::KEvent* event) {
 void LoopProcess(Nvnflinger::Nvnflinger& nvnflinger, Core::System& system) {
     auto server_manager = std::make_unique<ServerManager>(system);
     auto module = std::make_shared<Module>(system);
-    server_manager->RegisterNamedService("nvdrv", std::make_shared<NVDRV>(system, module, "nvdrv"));
-    server_manager->RegisterNamedService("nvdrv:a",
-                                         std::make_shared<NVDRV>(system, module, "nvdrv:a"));
-    server_manager->RegisterNamedService("nvdrv:s",
-                                         std::make_shared<NVDRV>(system, module, "nvdrv:s"));
-    server_manager->RegisterNamedService("nvdrv:t",
-                                         std::make_shared<NVDRV>(system, module, "nvdrv:t"));
+    const auto NvdrvInterfaceFactoryForApplication = [&, module] {
+        return std::make_shared<NVDRV>(system, module, "nvdrv");
+    };
+    const auto NvdrvInterfaceFactoryForApplets = [&, module] {
+        return std::make_shared<NVDRV>(system, module, "nvdrv:a");
+    };
+    const auto NvdrvInterfaceFactoryForSysmodules = [&, module] {
+        return std::make_shared<NVDRV>(system, module, "nvdrv:a");
+    };
+    const auto NvdrvInterfaceFactory = [&, module] {
+        return std::make_shared<NVDRV>(system, module, "nvdrv:t");
+    };
+    server_manager->RegisterNamedService("nvdrv", NvdrvInterfaceFactoryForApplication);
+    server_manager->RegisterNamedService("nvdrv:a", NvdrvInterfaceFactoryForApplets);
+    server_manager->RegisterNamedService("nvdrv:s", NvdrvInterfaceFactoryForSysmodules);
+    server_manager->RegisterNamedService("nvdrv:t", NvdrvInterfaceFactory);
     server_manager->RegisterNamedService("nvmemp", std::make_shared<NVMEMP>(system));
     nvnflinger.SetNVDrvInstance(module);
     ServerManager::RunServer(std::move(server_manager));
@@ -113,7 +122,7 @@ NvResult Module::VerifyFD(DeviceFD fd) const {
     return NvResult::Success;
 }
 
-DeviceFD Module::Open(const std::string& device_name) {
+DeviceFD Module::Open(const std::string& device_name, size_t session_id) {
     auto it = builders.find(device_name);
     if (it == builders.end()) {
         LOG_ERROR(Service_NVDRV, "Trying to open unknown device {}", device_name);
@@ -124,7 +133,7 @@ DeviceFD Module::Open(const std::string& device_name) {
     auto& builder = it->second;
     auto device = builder(fd)->second;
 
-    device->OnOpen(fd);
+    device->OnOpen(session_id, fd);
 
     return fd;
 }

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

@@ -77,7 +77,7 @@ public:
     NvResult VerifyFD(DeviceFD fd) const;
 
     /// Opens a device node and returns a file descriptor to it.
-    DeviceFD Open(const std::string& device_name);
+    DeviceFD Open(const std::string& device_name, size_t session_id);
 
     /// Sends an ioctl command to the specified file descriptor.
     NvResult Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> input, std::span<u8> output);
@@ -93,6 +93,10 @@ public:
 
     NvResult QueryEvent(DeviceFD fd, u32 event_id, Kernel::KEvent*& event);
 
+    NvCore::Container& GetContainer() {
+        return container;
+    }
+
 private:
     friend class EventInterface;
     friend class Service::Nvnflinger::Nvnflinger;

src/core/hle/service/nvdrv/nvdrv_interface.cpp

@@ -3,8 +3,10 @@
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 #include "common/logging/log.h"
+#include "common/scope_exit.h"
 #include "core/core.h"
 #include "core/hle/kernel/k_event.h"
+#include "core/hle/kernel/k_process.h"
 #include "core/hle/kernel/k_readable_event.h"
 #include "core/hle/service/ipc_helpers.h"
 #include "core/hle/service/nvdrv/nvdata.h"
@@ -37,7 +39,7 @@ void NVDRV::Open(HLERequestContext& ctx) {
         return;
     }
 
-    DeviceFD fd = nvdrv->Open(device_name);
+    DeviceFD fd = nvdrv->Open(device_name, session_id);
 
     rb.Push<DeviceFD>(fd);
     rb.PushEnum(fd != INVALID_NVDRV_FD ? NvResult::Success : NvResult::FileOperationFailed);
@@ -150,12 +152,29 @@ void NVDRV::Close(HLERequestContext& ctx) {
 
 void NVDRV::Initialize(HLERequestContext& ctx) {
     LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    IPC::ResponseBuilder rb{ctx, 3};
+    SCOPE_EXIT({
+        rb.Push(ResultSuccess);
+        rb.PushEnum(NvResult::Success);
+    });
+
+    if (is_initialized) {
+        // No need to initialize again
+        return;
+    }
+
+    IPC::RequestParser rp{ctx};
+    const auto process_handle{ctx.GetCopyHandle(0)};
+    // The transfer memory is lent to nvdrv as a work buffer since nvdrv is
+    // unable to allocate as much memory on its own. For HLE it's unnecessary to handle it
+    [[maybe_unused]] const auto transfer_memory_handle{ctx.GetCopyHandle(1)};
+    [[maybe_unused]] const auto transfer_memory_size = rp.Pop<u32>();
+
+    auto& container = nvdrv->GetContainer();
+    auto process = ctx.GetObjectFromHandle<Kernel::KProcess>(process_handle);
+    session_id = container.OpenSession(process.GetPointerUnsafe());
 
     is_initialized = true;
-
-    IPC::ResponseBuilder rb{ctx, 3};
-    rb.Push(ResultSuccess);
-    rb.PushEnum(NvResult::Success);
 }
 
 void NVDRV::QueryEvent(HLERequestContext& ctx) {
@@ -242,6 +261,9 @@ NVDRV::NVDRV(Core::System& system_, std::shared_ptr<Module> nvdrv_, const char*
     RegisterHandlers(functions);
 }
 
-NVDRV::~NVDRV() = default;
+NVDRV::~NVDRV() {
+    auto& container = nvdrv->GetContainer();
+    container.CloseSession(session_id);
+}
 
 } // namespace Service::Nvidia

src/core/hle/service/nvdrv/nvdrv_interface.h

@@ -35,6 +35,7 @@ private:
 
     u64 pid{};
     bool is_initialized{};
+    size_t session_id{};
     Common::ScratchBuffer<u8> output_buffer;
     Common::ScratchBuffer<u8> inline_output_buffer;
 };

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

@@ -87,19 +87,20 @@ Result CreateNvMapHandle(u32* out_nv_map_handle, Nvidia::Devices::nvmap& nvmap,
     R_SUCCEED();
 }
 
-Result FreeNvMapHandle(Nvidia::Devices::nvmap& nvmap, u32 handle) {
+Result FreeNvMapHandle(Nvidia::Devices::nvmap& nvmap, u32 handle, Nvidia::DeviceFD nvmap_fd) {
     // Free the handle.
     Nvidia::Devices::nvmap::IocFreeParams free_params{
         .handle = handle,
     };
-    R_UNLESS(nvmap.IocFree(free_params) == Nvidia::NvResult::Success, VI::ResultOperationFailed);
+    R_UNLESS(nvmap.IocFree(free_params, nvmap_fd) == Nvidia::NvResult::Success,
+             VI::ResultOperationFailed);
 
     // We succeeded.
     R_SUCCEED();
 }
 
 Result AllocNvMapHandle(Nvidia::Devices::nvmap& nvmap, u32 handle, Common::ProcessAddress buffer,
-                        u32 size) {
+                        u32 size, Nvidia::DeviceFD nvmap_fd) {
     // Assign the allocated memory to the handle.
     Nvidia::Devices::nvmap::IocAllocParams alloc_params{
         .handle = handle,
@@ -109,16 +110,16 @@ Result AllocNvMapHandle(Nvidia::Devices::nvmap& nvmap, u32 handle, Common::Proce
         .kind = 0,
         .address = GetInteger(buffer),
     };
-    R_UNLESS(nvmap.IocAlloc(alloc_params) == Nvidia::NvResult::Success, VI::ResultOperationFailed);
+    R_UNLESS(nvmap.IocAlloc(alloc_params, nvmap_fd) == Nvidia::NvResult::Success,
+             VI::ResultOperationFailed);
 
     // We succeeded.
     R_SUCCEED();
 }
 
-Result AllocateHandleForBuffer(u32* out_handle, Nvidia::Module& nvdrv,
+Result AllocateHandleForBuffer(u32* out_handle, Nvidia::Module& nvdrv, Nvidia::DeviceFD nvmap_fd,
                                Common::ProcessAddress buffer, u32 size) {
     // Get the nvmap device.
-    auto nvmap_fd = nvdrv.Open("/dev/nvmap");
     auto nvmap = nvdrv.GetDevice<Nvidia::Devices::nvmap>(nvmap_fd);
     ASSERT(nvmap != nullptr);
 
@@ -127,11 +128,11 @@ Result AllocateHandleForBuffer(u32* out_handle, Nvidia::Module& nvdrv,
 
     // Ensure we maintain a clean state on failure.
     ON_RESULT_FAILURE {
-        ASSERT(R_SUCCEEDED(FreeNvMapHandle(*nvmap, *out_handle)));
+        ASSERT(R_SUCCEEDED(FreeNvMapHandle(*nvmap, *out_handle, nvmap_fd)));
     };
 
     // Assign the allocated memory to the handle.
-    R_RETURN(AllocNvMapHandle(*nvmap, *out_handle, buffer, size));
+    R_RETURN(AllocNvMapHandle(*nvmap, *out_handle, buffer, size, nvmap_fd));
 }
 
 constexpr auto SharedBufferBlockLinearFormat = android::PixelFormat::Rgba8888;
@@ -197,9 +198,13 @@ Result FbShareBufferManager::Initialize(u64* out_buffer_id, u64* out_layer_id, u
                                            std::addressof(m_buffer_page_group), m_system,
                                            SharedBufferSize));
 
+    auto& container = m_nvdrv->GetContainer();
+    m_session_id = container.OpenSession(m_system.ApplicationProcess());
+    m_nvmap_fd = m_nvdrv->Open("/dev/nvmap", m_session_id);
+
     // Create an nvmap handle for the buffer and assign the memory to it.
-    R_TRY(AllocateHandleForBuffer(std::addressof(m_buffer_nvmap_handle), *m_nvdrv, map_address,
-                                  SharedBufferSize));
+    R_TRY(AllocateHandleForBuffer(std::addressof(m_buffer_nvmap_handle), *m_nvdrv, m_nvmap_fd,
+                                  map_address, SharedBufferSize));
 
     // Record the display id.
     m_display_id = display_id;

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

@@ -4,6 +4,7 @@
 #pragma once
 
 #include "common/math_util.h"
+#include "core/hle/service/nvdrv/nvdata.h"
 #include "core/hle/service/nvnflinger/nvnflinger.h"
 #include "core/hle/service/nvnflinger/ui/fence.h"
 
@@ -53,7 +54,8 @@ private:
     u64 m_layer_id = 0;
     u32 m_buffer_nvmap_handle = 0;
     SharedMemoryPoolLayout m_pool_layout = {};
-
+    Nvidia::DeviceFD m_nvmap_fd = {};
+    size_t m_session_id = {};
     std::unique_ptr<Kernel::KPageGroup> m_buffer_page_group;
 
     std::mutex m_guard;

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

@@ -126,7 +126,7 @@ void Nvnflinger::ShutdownLayers() {
 
 void Nvnflinger::SetNVDrvInstance(std::shared_ptr<Nvidia::Module> instance) {
     nvdrv = std::move(instance);
-    disp_fd = nvdrv->Open("/dev/nvdisp_disp0");
+    disp_fd = nvdrv->Open("/dev/nvdisp_disp0", 0);
 }
 
 std::optional<u64> Nvnflinger::OpenDisplay(std::string_view name) {

src/core/hle/service/nvnflinger/ui/graphic_buffer.cpp

@@ -22,11 +22,13 @@ GraphicBuffer::GraphicBuffer(Service::Nvidia::NvCore::NvMap& nvmap,
     : NvGraphicBuffer(GetBuffer(buffer)), m_nvmap(std::addressof(nvmap)) {
     if (this->BufferId() > 0) {
         m_nvmap->DuplicateHandle(this->BufferId(), true);
+        m_nvmap->PinHandle(this->BufferId(), false);
     }
 }
 
 GraphicBuffer::~GraphicBuffer() {
     if (m_nvmap != nullptr && this->BufferId() > 0) {
+        m_nvmap->UnpinHandle(this->BufferId());
         m_nvmap->FreeHandle(this->BufferId(), true);
     }
 }

src/core/hle/service/pcv/pcv.cpp

@@ -54,8 +54,8 @@ public:
 
 class IClkrstSession final : public ServiceFramework<IClkrstSession> {
 public:
-    explicit IClkrstSession(Core::System& system_, DeviceCode deivce_code_)
-        : ServiceFramework{system_, "IClkrstSession"}, deivce_code(deivce_code_) {
+    explicit IClkrstSession(Core::System& system_, DeviceCode device_code_)
+        : ServiceFramework{system_, "IClkrstSession"}, device_code(device_code_) {
         // clang-format off
         static const FunctionInfo functions[] = {
             {0, nullptr, "SetClockEnabled"},
@@ -93,7 +93,7 @@ private:
         rb.Push<u32>(clock_rate);
     }
 
-    DeviceCode deivce_code;
+    DeviceCode device_code;
     u32 clock_rate{};
 };
 
@@ -118,9 +118,9 @@ private:
     void OpenSession(HLERequestContext& ctx) {
         IPC::RequestParser rp{ctx};
         const auto device_code = static_cast<DeviceCode>(rp.Pop<u32>());
-        const auto unkonwn_input = rp.Pop<u32>();
+        const auto unknown_input = rp.Pop<u32>();
 
-        LOG_DEBUG(Service_PCV, "called, device_code={}, input={}", device_code, unkonwn_input);
+        LOG_DEBUG(Service_PCV, "called, device_code={}, input={}", device_code, unknown_input);
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(ResultSuccess);

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

@@ -1,11 +1,12 @@
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include "core/hle/service/set/set_cal.h"
+#include "core/hle/service/set/factory_settings_server.h"
 
 namespace Service::Set {
 
-SET_CAL::SET_CAL(Core::System& system_) : ServiceFramework{system_, "set:cal"} {
+IFactorySettingsServer::IFactorySettingsServer(Core::System& system_)
+    : ServiceFramework{system_, "set:cal"} {
     // clang-format off
     static const FunctionInfo functions[] = {
         {0, nullptr, "GetBluetoothBdAddress"},
@@ -57,6 +58,6 @@ SET_CAL::SET_CAL(Core::System& system_) : ServiceFramework{system_, "set:cal"} {
     RegisterHandlers(functions);
 }
 
-SET_CAL::~SET_CAL() = default;
+IFactorySettingsServer::~IFactorySettingsServer() = default;
 
 } // namespace Service::Set

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

@@ -11,10 +11,10 @@ class System;
 
 namespace Service::Set {
 
-class SET_FD final : public ServiceFramework<SET_FD> {
+class IFactorySettingsServer final : public ServiceFramework<IFactorySettingsServer> {
 public:
-    explicit SET_FD(Core::System& system_);
-    ~SET_FD() override;
+    explicit IFactorySettingsServer(Core::System& system_);
+    ~IFactorySettingsServer() override;
 };
 
 } // namespace Service::Set

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

@@ -1,11 +1,12 @@
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include "core/hle/service/set/set_fd.h"
+#include "core/hle/service/set/firmware_debug_settings_server.h"
 
 namespace Service::Set {
 
-SET_FD::SET_FD(Core::System& system_) : ServiceFramework{system_, "set:fd"} {
+IFirmwareDebugSettingsServer::IFirmwareDebugSettingsServer(Core::System& system_)
+    : ServiceFramework{system_, "set:fd"} {
     // clang-format off
     static const FunctionInfo functions[] = {
         {2, nullptr, "SetSettingsItemValue"},
@@ -23,6 +24,6 @@ SET_FD::SET_FD(Core::System& system_) : ServiceFramework{system_, "set:fd"} {
     RegisterHandlers(functions);
 }
 
-SET_FD::~SET_FD() = default;
+IFirmwareDebugSettingsServer::~IFirmwareDebugSettingsServer() = default;
 
 } // namespace Service::Set

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

@@ -11,10 +11,10 @@ class System;
 
 namespace Service::Set {
 
-class SET_CAL final : public ServiceFramework<SET_CAL> {
+class IFirmwareDebugSettingsServer final : public ServiceFramework<IFirmwareDebugSettingsServer> {
 public:
-    explicit SET_CAL(Core::System& system_);
-    ~SET_CAL() override;
+    explicit IFirmwareDebugSettingsServer(Core::System& system_);
+    ~IFirmwareDebugSettingsServer() override;
 };
 
 } // namespace Service::Set

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

@@ -2,21 +2,24 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include "core/hle/service/server_manager.h"
-#include "core/hle/service/set/set.h"
-#include "core/hle/service/set/set_cal.h"
-#include "core/hle/service/set/set_fd.h"
-#include "core/hle/service/set/set_sys.h"
+#include "core/hle/service/set/factory_settings_server.h"
+#include "core/hle/service/set/firmware_debug_settings_server.h"
 #include "core/hle/service/set/settings.h"
+#include "core/hle/service/set/settings_server.h"
+#include "core/hle/service/set/system_settings_server.h"
 
 namespace Service::Set {
 
 void LoopProcess(Core::System& system) {
     auto server_manager = std::make_unique<ServerManager>(system);
 
-    server_manager->RegisterNamedService("set", std::make_shared<SET>(system));
-    server_manager->RegisterNamedService("set:cal", std::make_shared<SET_CAL>(system));
-    server_manager->RegisterNamedService("set:fd", std::make_shared<SET_FD>(system));
-    server_manager->RegisterNamedService("set:sys", std::make_shared<SET_SYS>(system));
+    server_manager->RegisterNamedService("set", std::make_shared<ISettingsServer>(system));
+    server_manager->RegisterNamedService("set:cal",
+                                         std::make_shared<IFactorySettingsServer>(system));
+    server_manager->RegisterNamedService("set:fd",
+                                         std::make_shared<IFirmwareDebugSettingsServer>(system));
+    server_manager->RegisterNamedService("set:sys",
+                                         std::make_shared<ISystemSettingsServer>(system));
     ServerManager::RunServer(std::move(server_manager));
 }
 

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

@@ -7,7 +7,7 @@
 #include "common/logging/log.h"
 #include "common/settings.h"
 #include "core/hle/service/ipc_helpers.h"
-#include "core/hle/service/set/set.h"
+#include "core/hle/service/set/settings_server.h"
 
 namespace Service::Set {
 namespace {
@@ -58,13 +58,36 @@ LanguageCode GetLanguageCodeFromIndex(std::size_t index) {
     return available_language_codes.at(index);
 }
 
-void SET::GetAvailableLanguageCodes(HLERequestContext& ctx) {
+ISettingsServer::ISettingsServer(Core::System& system_) : ServiceFramework{system_, "set"} {
+    // clang-format off
+    static const FunctionInfo functions[] = {
+        {0, &ISettingsServer::GetLanguageCode, "GetLanguageCode"},
+        {1, &ISettingsServer::GetAvailableLanguageCodes, "GetAvailableLanguageCodes"},
+        {2, &ISettingsServer::MakeLanguageCode, "MakeLanguageCode"},
+        {3, &ISettingsServer::GetAvailableLanguageCodeCount, "GetAvailableLanguageCodeCount"},
+        {4, &ISettingsServer::GetRegionCode, "GetRegionCode"},
+        {5, &ISettingsServer::GetAvailableLanguageCodes2, "GetAvailableLanguageCodes2"},
+        {6, &ISettingsServer::GetAvailableLanguageCodeCount2, "GetAvailableLanguageCodeCount2"},
+        {7, &ISettingsServer::GetKeyCodeMap, "GetKeyCodeMap"},
+        {8, &ISettingsServer::GetQuestFlag, "GetQuestFlag"},
+        {9, &ISettingsServer::GetKeyCodeMap2, "GetKeyCodeMap2"},
+        {10, nullptr, "GetFirmwareVersionForDebug"},
+        {11, &ISettingsServer::GetDeviceNickName, "GetDeviceNickName"},
+    };
+    // clang-format on
+
+    RegisterHandlers(functions);
+}
+
+ISettingsServer::~ISettingsServer() = default;
+
+void ISettingsServer::GetAvailableLanguageCodes(HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
 
     GetAvailableLanguageCodesImpl(ctx, PRE_4_0_0_MAX_ENTRIES);
 }
 
-void SET::MakeLanguageCode(HLERequestContext& ctx) {
+void ISettingsServer::MakeLanguageCode(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto index = rp.Pop<u32>();
 
@@ -80,25 +103,25 @@ void SET::MakeLanguageCode(HLERequestContext& ctx) {
     rb.PushEnum(available_language_codes[index]);
 }
 
-void SET::GetAvailableLanguageCodes2(HLERequestContext& ctx) {
+void ISettingsServer::GetAvailableLanguageCodes2(HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
 
     GetAvailableLanguageCodesImpl(ctx, POST_4_0_0_MAX_ENTRIES);
 }
 
-void SET::GetAvailableLanguageCodeCount(HLERequestContext& ctx) {
+void ISettingsServer::GetAvailableLanguageCodeCount(HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
 
     PushResponseLanguageCode(ctx, PRE_4_0_0_MAX_ENTRIES);
 }
 
-void SET::GetAvailableLanguageCodeCount2(HLERequestContext& ctx) {
+void ISettingsServer::GetAvailableLanguageCodeCount2(HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
 
     PushResponseLanguageCode(ctx, POST_4_0_0_MAX_ENTRIES);
 }
 
-void SET::GetQuestFlag(HLERequestContext& ctx) {
+void ISettingsServer::GetQuestFlag(HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
 
     IPC::ResponseBuilder rb{ctx, 3};
@@ -106,7 +129,7 @@ void SET::GetQuestFlag(HLERequestContext& ctx) {
     rb.Push(static_cast<s32>(Settings::values.quest_flag.GetValue()));
 }
 
-void SET::GetLanguageCode(HLERequestContext& ctx) {
+void ISettingsServer::GetLanguageCode(HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called {}", Settings::values.language_index.GetValue());
 
     IPC::ResponseBuilder rb{ctx, 4};
@@ -115,7 +138,7 @@ void SET::GetLanguageCode(HLERequestContext& ctx) {
         available_language_codes[static_cast<s32>(Settings::values.language_index.GetValue())]);
 }
 
-void SET::GetRegionCode(HLERequestContext& ctx) {
+void ISettingsServer::GetRegionCode(HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
 
     IPC::ResponseBuilder rb{ctx, 3};
@@ -123,44 +146,21 @@ void SET::GetRegionCode(HLERequestContext& ctx) {
     rb.Push(static_cast<u32>(Settings::values.region_index.GetValue()));
 }
 
-void SET::GetKeyCodeMap(HLERequestContext& ctx) {
+void ISettingsServer::GetKeyCodeMap(HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "Called {}", ctx.Description());
     GetKeyCodeMapImpl(ctx);
 }
 
-void SET::GetKeyCodeMap2(HLERequestContext& ctx) {
+void ISettingsServer::GetKeyCodeMap2(HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "Called {}", ctx.Description());
     GetKeyCodeMapImpl(ctx);
 }
 
-void SET::GetDeviceNickName(HLERequestContext& ctx) {
+void ISettingsServer::GetDeviceNickName(HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
     IPC::ResponseBuilder rb{ctx, 2};
     rb.Push(ResultSuccess);
     ctx.WriteBuffer(Settings::values.device_name.GetValue());
 }
 
-SET::SET(Core::System& system_) : ServiceFramework{system_, "set"} {
-    // clang-format off
-    static const FunctionInfo functions[] = {
-        {0, &SET::GetLanguageCode, "GetLanguageCode"},
-        {1, &SET::GetAvailableLanguageCodes, "GetAvailableLanguageCodes"},
-        {2, &SET::MakeLanguageCode, "MakeLanguageCode"},
-        {3, &SET::GetAvailableLanguageCodeCount, "GetAvailableLanguageCodeCount"},
-        {4, &SET::GetRegionCode, "GetRegionCode"},
-        {5, &SET::GetAvailableLanguageCodes2, "GetAvailableLanguageCodes2"},
-        {6, &SET::GetAvailableLanguageCodeCount2, "GetAvailableLanguageCodeCount2"},
-        {7, &SET::GetKeyCodeMap, "GetKeyCodeMap"},
-        {8, &SET::GetQuestFlag, "GetQuestFlag"},
-        {9, &SET::GetKeyCodeMap2, "GetKeyCodeMap2"},
-        {10, nullptr, "GetFirmwareVersionForDebug"},
-        {11, &SET::GetDeviceNickName, "GetDeviceNickName"},
-    };
-    // clang-format on
-
-    RegisterHandlers(functions);
-}
-
-SET::~SET() = default;
-
 } // namespace Service::Set

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

@@ -73,10 +73,10 @@ static constexpr std::array<std::pair<LanguageCode, KeyboardLayout>, 18> languag
 
 LanguageCode GetLanguageCodeFromIndex(std::size_t idx);
 
-class SET final : public ServiceFramework<SET> {
+class ISettingsServer final : public ServiceFramework<ISettingsServer> {
 public:
-    explicit SET(Core::System& system_);
-    ~SET() override;
+    explicit ISettingsServer(Core::System& system_);
+    ~ISettingsServer() override;
 
 private:
     void GetLanguageCode(HLERequestContext& ctx);

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

@@ -28,7 +28,7 @@ SystemSettings DefaultSystemSettings() {
         .cmu_mode = CmuMode::None,
         .tv_underscan = {},
         .tv_gama = 1.0f,
-        .constrast_ratio = 0.5f,
+        .contrast_ratio = 0.5f,
     };
 
     settings.initial_launch_settings_packed = {

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

@@ -208,7 +208,7 @@ struct TvSettings {
     CmuMode cmu_mode;
     u32 tv_underscan;
     f32 tv_gama;
-    f32 constrast_ratio;
+    f32 contrast_ratio;
 };
 static_assert(sizeof(TvSettings) == 0x20, "TvSettings is an invalid size");
 
@@ -341,7 +341,7 @@ struct SystemSettings {
     std::array<u8, 0x3C> reserved_09934;
 
     // nn::settings::system::ErrorReportSharePermission
-    ErrorReportSharePermission error_report_share_permssion;
+    ErrorReportSharePermission error_report_share_permission;
 
     std::array<u8, 0x3C> reserved_09974;
 

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

@@ -47,10 +47,10 @@ static_assert(sizeof(FirmwareVersionFormat) == 0x100, "FirmwareVersionFormat is
 Result GetFirmwareVersionImpl(FirmwareVersionFormat& out_firmware, Core::System& system,
                               GetFirmwareVersionType type);
 
-class SET_SYS final : public ServiceFramework<SET_SYS> {
+class ISystemSettingsServer final : public ServiceFramework<ISystemSettingsServer> {
 public:
-    explicit SET_SYS(Core::System& system_);
-    ~SET_SYS() override;
+    explicit ISystemSettingsServer(Core::System& system_);
+    ~ISystemSettingsServer() override;
 
     Result GetSettingsItemValue(std::vector<u8>& out_value, const std::string& category,
                                 const std::string& name);

src/core/memory.cpp

@@ -24,6 +24,8 @@
 #include "core/hle/kernel/k_process.h"
 #include "core/memory.h"
 #include "video_core/gpu.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
+#include "video_core/host1x/host1x.h"
 #include "video_core/rasterizer_download_area.h"
 
 namespace Core::Memory {
@@ -637,17 +639,6 @@ struct Memory::Impl {
         LOG_DEBUG(HW_Memory, "Mapping {:016X} onto {:016X}-{:016X}", GetInteger(target),
                   base * YUZU_PAGESIZE, (base + size) * YUZU_PAGESIZE);
 
-        // During boot, current_page_table might not be set yet, in which case we need not flush
-        if (system.IsPoweredOn()) {
-            auto& gpu = system.GPU();
-            for (u64 i = 0; i < size; i++) {
-                const auto page = base + i;
-                if (page_table.pointers[page].Type() == Common::PageType::RasterizerCachedMemory) {
-                    gpu.FlushAndInvalidateRegion(page << YUZU_PAGEBITS, YUZU_PAGESIZE);
-                }
-            }
-        }
-
         const auto end = base + size;
         ASSERT_MSG(end <= page_table.pointers.size(), "out of range mapping at {:016X}",
                    base + page_table.pointers.size());
@@ -811,21 +802,33 @@ struct Memory::Impl {
         return true;
     }
 
-    void HandleRasterizerDownload(VAddr address, size_t size) {
+    void HandleRasterizerDownload(VAddr v_address, size_t size) {
+        const auto* p = GetPointerImpl(
+            v_address, []() {}, []() {});
+        if (!gpu_device_memory) [[unlikely]] {
+            gpu_device_memory = &system.Host1x().MemoryManager();
+        }
         const size_t core = system.GetCurrentHostThreadID();
         auto& current_area = rasterizer_read_areas[core];
-        const VAddr end_address = address + size;
-        if (current_area.start_address <= address && end_address <= current_area.end_address)
-            [[likely]] {
-            return;
-        }
-        current_area = system.GPU().OnCPURead(address, size);
+        gpu_device_memory->ApplyOpOnPointer(p, scratch_buffers[core], [&](DAddr address) {
+            const DAddr end_address = address + size;
+            if (current_area.start_address <= address && end_address <= current_area.end_address)
+                [[likely]] {
+                return;
+            }
+            current_area = system.GPU().OnCPURead(address, size);
+        });
     }
 
-    void HandleRasterizerWrite(VAddr address, size_t size) {
+    void HandleRasterizerWrite(VAddr v_address, size_t size) {
+        const auto* p = GetPointerImpl(
+            v_address, []() {}, []() {});
         constexpr size_t sys_core = Core::Hardware::NUM_CPU_CORES - 1;
         const size_t core = std::min(system.GetCurrentHostThreadID(),
                                      sys_core); // any other calls threads go to syscore.
+        if (!gpu_device_memory) [[unlikely]] {
+            gpu_device_memory = &system.Host1x().MemoryManager();
+        }
         // Guard on sys_core;
         if (core == sys_core) [[unlikely]] {
             sys_core_guard.lock();
@@ -835,36 +838,53 @@ struct Memory::Impl {
                 sys_core_guard.unlock();
             }
         });
-        auto& current_area = rasterizer_write_areas[core];
-        VAddr subaddress = address >> YUZU_PAGEBITS;
-        bool do_collection = current_area.last_address == subaddress;
-        if (!do_collection) [[unlikely]] {
-            do_collection = system.GPU().OnCPUWrite(address, size);
-            if (!do_collection) {
-                return;
+        gpu_device_memory->ApplyOpOnPointer(p, scratch_buffers[core], [&](DAddr address) {
+            auto& current_area = rasterizer_write_areas[core];
+            PAddr subaddress = address >> YUZU_PAGEBITS;
+            bool do_collection = current_area.last_address == subaddress;
+            if (!do_collection) [[unlikely]] {
+                do_collection = system.GPU().OnCPUWrite(address, size);
+                if (!do_collection) {
+                    return;
+                }
+                current_area.last_address = subaddress;
             }
-            current_area.last_address = subaddress;
-        }
-        gpu_dirty_managers[core].Collect(address, size);
+            gpu_dirty_managers[core].Collect(address, size);
+        });
     }
 
     struct GPUDirtyState {
-        VAddr last_address;
+        PAddr last_address;
     };
 
-    void InvalidateRegion(Common::ProcessAddress dest_addr, size_t size) {
-        system.GPU().InvalidateRegion(GetInteger(dest_addr), size);
-    }
-
-    void FlushRegion(Common::ProcessAddress dest_addr, size_t size) {
-        system.GPU().FlushRegion(GetInteger(dest_addr), size);
+    void InvalidateGPUMemory(u8* p, size_t size) {
+        constexpr size_t sys_core = Core::Hardware::NUM_CPU_CORES - 1;
+        const size_t core = std::min(system.GetCurrentHostThreadID(),
+                                     sys_core); // any other calls threads go to syscore.
+        if (!gpu_device_memory) [[unlikely]] {
+            gpu_device_memory = &system.Host1x().MemoryManager();
+        }
+        // Guard on sys_core;
+        if (core == sys_core) [[unlikely]] {
+            sys_core_guard.lock();
+        }
+        SCOPE_EXIT({
+            if (core == sys_core) [[unlikely]] {
+                sys_core_guard.unlock();
+            }
+        });
+        auto& gpu = system.GPU();
+        gpu_device_memory->ApplyOpOnPointer(
+            p, scratch_buffers[core], [&](DAddr address) { gpu.InvalidateRegion(address, size); });
     }
 
     Core::System& system;
+    Tegra::MaxwellDeviceMemoryManager* gpu_device_memory{};
     Common::PageTable* current_page_table = nullptr;
     std::array<VideoCore::RasterizerDownloadArea, Core::Hardware::NUM_CPU_CORES>
         rasterizer_read_areas{};
     std::array<GPUDirtyState, Core::Hardware::NUM_CPU_CORES> rasterizer_write_areas{};
+    std::array<Common::ScratchBuffer<u32>, Core::Hardware::NUM_CPU_CORES> scratch_buffers{};
     std::span<Core::GPUDirtyMemoryManager> gpu_dirty_managers;
     std::mutex sys_core_guard;
 
@@ -1059,14 +1079,6 @@ void Memory::MarkRegionDebug(Common::ProcessAddress vaddr, u64 size, bool debug)
     impl->MarkRegionDebug(GetInteger(vaddr), size, debug);
 }
 
-void Memory::InvalidateRegion(Common::ProcessAddress dest_addr, size_t size) {
-    impl->InvalidateRegion(dest_addr, size);
-}
-
-void Memory::FlushRegion(Common::ProcessAddress dest_addr, size_t size) {
-    impl->FlushRegion(dest_addr, size);
-}
-
 bool Memory::InvalidateNCE(Common::ProcessAddress vaddr, size_t size) {
     [[maybe_unused]] bool mapped = true;
     [[maybe_unused]] bool rasterizer = false;
@@ -1078,10 +1090,10 @@ bool Memory::InvalidateNCE(Common::ProcessAddress vaddr, size_t size) {
                       GetInteger(vaddr));
             mapped = false;
         },
-        [&] {
-            impl->system.GPU().InvalidateRegion(GetInteger(vaddr), size);
-            rasterizer = true;
-        });
+        [&] { rasterizer = true; });
+    if (rasterizer) {
+        impl->InvalidateGPUMemory(ptr, size);
+    }
 
 #ifdef __linux__
     if (!rasterizer && mapped) {

src/core/memory.h

@@ -12,6 +12,7 @@
 
 #include "common/scratch_buffer.h"
 #include "common/typed_address.h"
+#include "core/guest_memory.h"
 #include "core/hle/result.h"
 
 namespace Common {
@@ -486,10 +487,10 @@ public:
     void MarkRegionDebug(Common::ProcessAddress vaddr, u64 size, bool debug);
 
     void SetGPUDirtyManagers(std::span<Core::GPUDirtyMemoryManager> managers);
-    void InvalidateRegion(Common::ProcessAddress dest_addr, size_t size);
+
     bool InvalidateNCE(Common::ProcessAddress vaddr, size_t size);
+
     bool InvalidateSeparateHeap(void* fault_address);
-    void FlushRegion(Common::ProcessAddress dest_addr, size_t size);
 
 private:
     Core::System& system;
@@ -498,209 +499,9 @@ private:
     std::unique_ptr<Impl> impl;
 };
 
-enum GuestMemoryFlags : u32 {
-    Read = 1 << 0,
-    Write = 1 << 1,
-    Safe = 1 << 2,
-    Cached = 1 << 3,
-
-    SafeRead = Read | Safe,
-    SafeWrite = Write | Safe,
-    SafeReadWrite = SafeRead | SafeWrite,
-    SafeReadCachedWrite = SafeReadWrite | Cached,
-
-    UnsafeRead = Read,
-    UnsafeWrite = Write,
-    UnsafeReadWrite = UnsafeRead | UnsafeWrite,
-    UnsafeReadCachedWrite = UnsafeReadWrite | Cached,
-};
-
-namespace {
-template <typename M, typename T, GuestMemoryFlags FLAGS>
-class GuestMemory {
-    using iterator = T*;
-    using const_iterator = const T*;
-    using value_type = T;
-    using element_type = T;
-    using iterator_category = std::contiguous_iterator_tag;
-
-public:
-    GuestMemory() = delete;
-    explicit GuestMemory(M& memory, u64 addr, std::size_t size,
-                         Common::ScratchBuffer<T>* backup = nullptr)
-        : m_memory{memory}, m_addr{addr}, m_size{size} {
-        static_assert(FLAGS & GuestMemoryFlags::Read || FLAGS & GuestMemoryFlags::Write);
-        if constexpr (FLAGS & GuestMemoryFlags::Read) {
-            Read(addr, size, backup);
-        }
-    }
-
-    ~GuestMemory() = default;
-
-    T* data() noexcept {
-        return m_data_span.data();
-    }
-
-    const T* data() const noexcept {
-        return m_data_span.data();
-    }
-
-    size_t size() const noexcept {
-        return m_size;
-    }
-
-    size_t size_bytes() const noexcept {
-        return this->size() * sizeof(T);
-    }
-
-    [[nodiscard]] T* begin() noexcept {
-        return this->data();
-    }
-
-    [[nodiscard]] const T* begin() const noexcept {
-        return this->data();
-    }
-
-    [[nodiscard]] T* end() noexcept {
-        return this->data() + this->size();
-    }
-
-    [[nodiscard]] const T* end() const noexcept {
-        return this->data() + this->size();
-    }
-
-    T& operator[](size_t index) noexcept {
-        return m_data_span[index];
-    }
-
-    const T& operator[](size_t index) const noexcept {
-        return m_data_span[index];
-    }
-
-    void SetAddressAndSize(u64 addr, std::size_t size) noexcept {
-        m_addr = addr;
-        m_size = size;
-        m_addr_changed = true;
-    }
-
-    std::span<T> Read(u64 addr, std::size_t size,
-                      Common::ScratchBuffer<T>* backup = nullptr) noexcept {
-        m_addr = addr;
-        m_size = size;
-        if (m_size == 0) {
-            m_is_data_copy = true;
-            return {};
-        }
-
-        if (this->TrySetSpan()) {
-            if constexpr (FLAGS & GuestMemoryFlags::Safe) {
-                m_memory.FlushRegion(m_addr, this->size_bytes());
-            }
-        } else {
-            if (backup) {
-                backup->resize_destructive(this->size());
-                m_data_span = *backup;
-            } else {
-                m_data_copy.resize(this->size());
-                m_data_span = std::span(m_data_copy);
-            }
-            m_is_data_copy = true;
-            m_span_valid = true;
-            if constexpr (FLAGS & GuestMemoryFlags::Safe) {
-                m_memory.ReadBlock(m_addr, this->data(), this->size_bytes());
-            } else {
-                m_memory.ReadBlockUnsafe(m_addr, this->data(), this->size_bytes());
-            }
-        }
-        return m_data_span;
-    }
-
-    void Write(std::span<T> write_data) noexcept {
-        if constexpr (FLAGS & GuestMemoryFlags::Cached) {
-            m_memory.WriteBlockCached(m_addr, write_data.data(), this->size_bytes());
-        } else if constexpr (FLAGS & GuestMemoryFlags::Safe) {
-            m_memory.WriteBlock(m_addr, write_data.data(), this->size_bytes());
-        } else {
-            m_memory.WriteBlockUnsafe(m_addr, write_data.data(), this->size_bytes());
-        }
-    }
-
-    bool TrySetSpan() noexcept {
-        if (u8* ptr = m_memory.GetSpan(m_addr, this->size_bytes()); ptr) {
-            m_data_span = {reinterpret_cast<T*>(ptr), this->size()};
-            m_span_valid = true;
-            return true;
-        }
-        return false;
-    }
-
-protected:
-    bool IsDataCopy() const noexcept {
-        return m_is_data_copy;
-    }
-
-    bool AddressChanged() const noexcept {
-        return m_addr_changed;
-    }
-
-    M& m_memory;
-    u64 m_addr{};
-    size_t m_size{};
-    std::span<T> m_data_span{};
-    std::vector<T> m_data_copy{};
-    bool m_span_valid{false};
-    bool m_is_data_copy{false};
-    bool m_addr_changed{false};
-};
-
-template <typename M, typename T, GuestMemoryFlags FLAGS>
-class GuestMemoryScoped : public GuestMemory<M, T, FLAGS> {
-public:
-    GuestMemoryScoped() = delete;
-    explicit GuestMemoryScoped(M& memory, u64 addr, std::size_t size,
-                               Common::ScratchBuffer<T>* backup = nullptr)
-        : GuestMemory<M, T, FLAGS>(memory, addr, size, backup) {
-        if constexpr (!(FLAGS & GuestMemoryFlags::Read)) {
-            if (!this->TrySetSpan()) {
-                if (backup) {
-                    this->m_data_span = *backup;
-                    this->m_span_valid = true;
-                    this->m_is_data_copy = true;
-                }
-            }
-        }
-    }
-
-    ~GuestMemoryScoped() {
-        if constexpr (FLAGS & GuestMemoryFlags::Write) {
-            if (this->size() == 0) [[unlikely]] {
-                return;
-            }
-
-            if (this->AddressChanged() || this->IsDataCopy()) {
-                ASSERT(this->m_span_valid);
-                if constexpr (FLAGS & GuestMemoryFlags::Cached) {
-                    this->m_memory.WriteBlockCached(this->m_addr, this->data(), this->size_bytes());
-                } else if constexpr (FLAGS & GuestMemoryFlags::Safe) {
-                    this->m_memory.WriteBlock(this->m_addr, this->data(), this->size_bytes());
-                } else {
-                    this->m_memory.WriteBlockUnsafe(this->m_addr, this->data(), this->size_bytes());
-                }
-            } else if constexpr ((FLAGS & GuestMemoryFlags::Safe) ||
-                                 (FLAGS & GuestMemoryFlags::Cached)) {
-                this->m_memory.InvalidateRegion(this->m_addr, this->size_bytes());
-            }
-        }
-    }
-};
-} // namespace
-
 template <typename T, GuestMemoryFlags FLAGS>
-using CpuGuestMemory = GuestMemory<Memory, T, FLAGS>;
+using CpuGuestMemory = GuestMemory<Core::Memory::Memory, T, FLAGS>;
 template <typename T, GuestMemoryFlags FLAGS>
-using CpuGuestMemoryScoped = GuestMemoryScoped<Memory, T, FLAGS>;
-template <typename T, GuestMemoryFlags FLAGS>
-using GpuGuestMemory = GuestMemory<Tegra::MemoryManager, T, FLAGS>;
-template <typename T, GuestMemoryFlags FLAGS>
-using GpuGuestMemoryScoped = GuestMemoryScoped<Tegra::MemoryManager, T, FLAGS>;
+using CpuGuestMemoryScoped = GuestMemoryScoped<Core::Memory::Memory, T, FLAGS>;
+
 } // namespace Core::Memory

src/hid_core/resources/applet_resource.cpp

@@ -130,6 +130,7 @@ void AppletResource::FreeAppletResourceId(u64 aruid) {
     if (aruid_data.flag.is_assigned) {
         aruid_data.shared_memory_format = nullptr;
         aruid_data.flag.is_assigned.Assign(false);
+        shared_memory_holder[index].Finalize();
     }
 }
 

src/hid_core/resources/npad/npad.cpp

@@ -1344,4 +1344,49 @@ AppletDetailedUiType NPad::GetAppletDetailedUiType(Core::HID::NpadIdType npad_id
     };
 }
 
+Result NPad::SetNpadCaptureButtonAssignment(u64 aruid, Core::HID::NpadStyleSet npad_style_set,
+                                            Core::HID::NpadButton button_assignment) {
+    std::scoped_lock lock{mutex};
+    return npad_resource.SetNpadCaptureButtonAssignment(aruid, npad_style_set, button_assignment);
+}
+
+Result NPad::ClearNpadCaptureButtonAssignment(u64 aruid) {
+    std::scoped_lock lock{mutex};
+    return npad_resource.ClearNpadCaptureButtonAssignment(aruid);
+}
+
+std::size_t NPad::GetNpadCaptureButtonAssignment(std::span<Core::HID::NpadButton> out_list,
+                                                 u64 aruid) const {
+    std::scoped_lock lock{mutex};
+    return npad_resource.GetNpadCaptureButtonAssignment(out_list, aruid);
+}
+
+Result NPad::SetNpadSystemExtStateEnabled(u64 aruid, bool is_enabled) {
+    std::scoped_lock lock{mutex};
+    const auto result = npad_resource.SetNpadSystemExtStateEnabled(aruid, is_enabled);
+
+    if (result.IsSuccess()) {
+        std::scoped_lock shared_lock{*applet_resource_holder.shared_mutex};
+        // TODO: abstracted_pad->EnableAppletToGetInput(aruid);
+    }
+
+    return result;
+}
+
+Result NPad::AssigningSingleOnSlSrPress(u64 aruid, bool is_enabled) {
+    std::scoped_lock lock{mutex};
+    bool is_currently_enabled{};
+    Result result = npad_resource.IsAssigningSingleOnSlSrPressEnabled(is_currently_enabled, aruid);
+    if (result.IsSuccess() && is_enabled != is_currently_enabled) {
+        result = npad_resource.SetAssigningSingleOnSlSrPress(aruid, is_enabled);
+    }
+    return result;
+}
+
+Result NPad::GetLastActiveNpad(Core::HID::NpadIdType& out_npad_id) const {
+    std::scoped_lock lock{mutex};
+    out_npad_id = hid_core.GetLastActiveController();
+    return ResultSuccess;
+}
+
 } // namespace Service::HID

src/hid_core/resources/npad/npad.h

@@ -149,6 +149,18 @@ public:
 
     AppletDetailedUiType GetAppletDetailedUiType(Core::HID::NpadIdType npad_id);
 
+    Result SetNpadCaptureButtonAssignment(u64 aruid, Core::HID::NpadStyleSet npad_style_set,
+                                          Core::HID::NpadButton button_assignment);
+    Result ClearNpadCaptureButtonAssignment(u64 aruid);
+    std::size_t GetNpadCaptureButtonAssignment(std::span<Core::HID::NpadButton> out_list,
+                                               u64 aruid) const;
+
+    Result SetNpadSystemExtStateEnabled(u64 aruid, bool is_enabled);
+
+    Result AssigningSingleOnSlSrPress(u64 aruid, bool is_enabled);
+
+    Result GetLastActiveNpad(Core::HID::NpadIdType& out_npad_id) const;
+
 private:
     struct VibrationData {
         bool device_mounted{};

src/shader_recompiler/environment.h

@@ -59,8 +59,8 @@ public:
         return start_address;
     }
 
-    [[nodiscard]] bool IsPropietaryDriver() const noexcept {
-        return is_propietary_driver;
+    [[nodiscard]] bool IsProprietaryDriver() const noexcept {
+        return is_proprietary_driver;
     }
 
 protected:
@@ -68,7 +68,7 @@ protected:
     std::array<u32, 8> gp_passthrough_mask{};
     Stage stage{};
     u32 start_address{};
-    bool is_propietary_driver{};
+    bool is_proprietary_driver{};
 };
 
 } // namespace Shader

src/shader_recompiler/ir_opt/constant_propagation_pass.cpp

@@ -1084,7 +1084,7 @@ void ConstantPropagation(Environment& env, IR::Block& block, IR::Inst& inst) {
         if (env.HasHLEMacroState()) {
             FoldConstBuffer(env, block, inst);
         }
-        if (env.IsPropietaryDriver()) {
+        if (env.IsProprietaryDriver()) {
             FoldDriverConstBuffer(env, block, inst, 1);
         }
         break;

src/tests/video_core/memory_tracker.cpp

@@ -24,9 +24,8 @@ constexpr VAddr c = 16 * HIGH_PAGE_SIZE;
 class RasterizerInterface {
 public:
     void UpdatePagesCachedCount(VAddr addr, u64 size, int delta) {
-        const u64 page_start{addr >> Core::Memory::YUZU_PAGEBITS};
-        const u64 page_end{(addr + size + Core::Memory::YUZU_PAGESIZE - 1) >>
-                           Core::Memory::YUZU_PAGEBITS};
+        const u64 page_start{addr >> Core::DEVICE_PAGEBITS};
+        const u64 page_end{(addr + size + Core::DEVICE_PAGESIZE - 1) >> Core::DEVICE_PAGEBITS};
         for (u64 page = page_start; page < page_end; ++page) {
             int& value = page_table[page];
             value += delta;
@@ -40,7 +39,7 @@ public:
     }
 
     [[nodiscard]] int Count(VAddr addr) const noexcept {
-        const auto it = page_table.find(addr >> Core::Memory::YUZU_PAGEBITS);
+        const auto it = page_table.find(addr >> Core::DEVICE_PAGEBITS);
         return it == page_table.end() ? 0 : it->second;
     }
 

src/video_core/CMakeLists.txt

@@ -71,6 +71,8 @@ add_library(video_core STATIC
     host1x/ffmpeg/ffmpeg.h
     host1x/control.cpp
     host1x/control.h
+    host1x/gpu_device_memory_manager.cpp
+    host1x/gpu_device_memory_manager.h
     host1x/host1x.cpp
     host1x/host1x.h
     host1x/nvdec.cpp
@@ -93,6 +95,7 @@ add_library(video_core STATIC
     gpu.h
     gpu_thread.cpp
     gpu_thread.h
+    guest_memory.h
     invalidation_accumulator.h
     memory_manager.cpp
     memory_manager.h
@@ -105,8 +108,6 @@ add_library(video_core STATIC
     query_cache/query_stream.h
     query_cache/types.h
     query_cache.h
-    rasterizer_accelerated.cpp
-    rasterizer_accelerated.h
     rasterizer_interface.h
     renderer_base.cpp
     renderer_base.h

src/video_core/buffer_cache/buffer_base.h

@@ -33,13 +33,12 @@ struct NullBufferParams {};
  *
  * The buffer size and address is forcefully aligned to CPU page boundaries.
  */
-template <class RasterizerInterface>
 class BufferBase {
 public:
     static constexpr u64 BASE_PAGE_BITS = 16;
     static constexpr u64 BASE_PAGE_SIZE = 1ULL << BASE_PAGE_BITS;
 
-    explicit BufferBase(RasterizerInterface& rasterizer_, VAddr cpu_addr_, u64 size_bytes_)
+    explicit BufferBase(VAddr cpu_addr_, u64 size_bytes_)
         : cpu_addr{cpu_addr_}, size_bytes{size_bytes_} {}
 
     explicit BufferBase(NullBufferParams) {}