early-access version 1255

src/core/hle/kernel/address_arbiter.cpp

@@ -0,0 +1,317 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <vector>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "core/arm/exclusive_monitor.h"
+#include "core/core.h"
+#include "core/hle/kernel/address_arbiter.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/time_manager.h"
+#include "core/hle/result.h"
+#include "core/memory.h"
+
+namespace Kernel {
+
+// Wake up num_to_wake (or all) threads in a vector.
+void AddressArbiter::WakeThreads(const std::vector<std::shared_ptr<Thread>>& waiting_threads,
+                                 s32 num_to_wake) {
+    // Only process up to 'target' threads, unless 'target' is <= 0, in which case process
+    // them all.
+    std::size_t last = waiting_threads.size();
+    if (num_to_wake > 0) {
+        last = std::min(last, static_cast<std::size_t>(num_to_wake));
+    }
+
+    // Signal the waiting threads.
+    for (std::size_t i = 0; i < last; i++) {
+        waiting_threads[i]->SetSynchronizationResults(nullptr, RESULT_SUCCESS);
+        RemoveThread(waiting_threads[i]);
+        waiting_threads[i]->WaitForArbitration(false);
+        waiting_threads[i]->ResumeFromWait();
+    }
+}
+
+AddressArbiter::AddressArbiter(Core::System& system) : system{system} {}
+AddressArbiter::~AddressArbiter() = default;
+
+ResultCode AddressArbiter::SignalToAddress(VAddr address, SignalType type, s32 value,
+                                           s32 num_to_wake) {
+    switch (type) {
+    case SignalType::Signal:
+        return SignalToAddressOnly(address, num_to_wake);
+    case SignalType::IncrementAndSignalIfEqual:
+        return IncrementAndSignalToAddressIfEqual(address, value, num_to_wake);
+    case SignalType::ModifyByWaitingCountAndSignalIfEqual:
+        return ModifyByWaitingCountAndSignalToAddressIfEqual(address, value, num_to_wake);
+    default:
+        return ERR_INVALID_ENUM_VALUE;
+    }
+}
+
+ResultCode AddressArbiter::SignalToAddressOnly(VAddr address, s32 num_to_wake) {
+    KScopedSchedulerLock lock(system.Kernel());
+    const std::vector<std::shared_ptr<Thread>> waiting_threads =
+        GetThreadsWaitingOnAddress(address);
+    WakeThreads(waiting_threads, num_to_wake);
+    return RESULT_SUCCESS;
+}
+
+ResultCode AddressArbiter::IncrementAndSignalToAddressIfEqual(VAddr address, s32 value,
+                                                              s32 num_to_wake) {
+    KScopedSchedulerLock lock(system.Kernel());
+    auto& memory = system.Memory();
+
+    // Ensure that we can write to the address.
+    if (!memory.IsValidVirtualAddress(address)) {
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    const std::size_t current_core = system.CurrentCoreIndex();
+    auto& monitor = system.Monitor();
+    u32 current_value;
+    do {
+        current_value = monitor.ExclusiveRead32(current_core, address);
+
+        if (current_value != static_cast<u32>(value)) {
+            return ERR_INVALID_STATE;
+        }
+        current_value++;
+    } while (!monitor.ExclusiveWrite32(current_core, address, current_value));
+
+    return SignalToAddressOnly(address, num_to_wake);
+}
+
+ResultCode AddressArbiter::ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 value,
+                                                                         s32 num_to_wake) {
+    KScopedSchedulerLock lock(system.Kernel());
+    auto& memory = system.Memory();
+
+    // Ensure that we can write to the address.
+    if (!memory.IsValidVirtualAddress(address)) {
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    // Get threads waiting on the address.
+    const std::vector<std::shared_ptr<Thread>> waiting_threads =
+        GetThreadsWaitingOnAddress(address);
+
+    const std::size_t current_core = system.CurrentCoreIndex();
+    auto& monitor = system.Monitor();
+    s32 updated_value;
+    do {
+        updated_value = monitor.ExclusiveRead32(current_core, address);
+
+        if (updated_value != value) {
+            return ERR_INVALID_STATE;
+        }
+        // Determine the modified value depending on the waiting count.
+        if (num_to_wake <= 0) {
+            if (waiting_threads.empty()) {
+                updated_value = value + 1;
+            } else {
+                updated_value = value - 1;
+            }
+        } else {
+            if (waiting_threads.empty()) {
+                updated_value = value + 1;
+            } else if (waiting_threads.size() <= static_cast<u32>(num_to_wake)) {
+                updated_value = value - 1;
+            } else {
+                updated_value = value;
+            }
+        }
+    } while (!monitor.ExclusiveWrite32(current_core, address, updated_value));
+
+    WakeThreads(waiting_threads, num_to_wake);
+    return RESULT_SUCCESS;
+}
+
+ResultCode AddressArbiter::WaitForAddress(VAddr address, ArbitrationType type, s32 value,
+                                          s64 timeout_ns) {
+    switch (type) {
+    case ArbitrationType::WaitIfLessThan:
+        return WaitForAddressIfLessThan(address, value, timeout_ns, false);
+    case ArbitrationType::DecrementAndWaitIfLessThan:
+        return WaitForAddressIfLessThan(address, value, timeout_ns, true);
+    case ArbitrationType::WaitIfEqual:
+        return WaitForAddressIfEqual(address, value, timeout_ns);
+    default:
+        return ERR_INVALID_ENUM_VALUE;
+    }
+}
+
+ResultCode AddressArbiter::WaitForAddressIfLessThan(VAddr address, s32 value, s64 timeout,
+                                                    bool should_decrement) {
+    auto& memory = system.Memory();
+    auto& kernel = system.Kernel();
+    Thread* current_thread = kernel.CurrentScheduler()->GetCurrentThread();
+
+    Handle event_handle = InvalidHandle;
+    {
+        KScopedSchedulerLockAndSleep lock(kernel, event_handle, current_thread, timeout);
+
+        if (current_thread->IsPendingTermination()) {
+            lock.CancelSleep();
+            return ERR_THREAD_TERMINATING;
+        }
+
+        // Ensure that we can read the address.
+        if (!memory.IsValidVirtualAddress(address)) {
+            lock.CancelSleep();
+            return ERR_INVALID_ADDRESS_STATE;
+        }
+
+        s32 current_value = static_cast<s32>(memory.Read32(address));
+        if (current_value >= value) {
+            lock.CancelSleep();
+            return ERR_INVALID_STATE;
+        }
+
+        current_thread->SetSynchronizationResults(nullptr, RESULT_TIMEOUT);
+
+        s32 decrement_value;
+
+        const std::size_t current_core = system.CurrentCoreIndex();
+        auto& monitor = system.Monitor();
+        do {
+            current_value = static_cast<s32>(monitor.ExclusiveRead32(current_core, address));
+            if (should_decrement) {
+                decrement_value = current_value - 1;
+            } else {
+                decrement_value = current_value;
+            }
+        } while (
+            !monitor.ExclusiveWrite32(current_core, address, static_cast<u32>(decrement_value)));
+
+        // Short-circuit without rescheduling, if timeout is zero.
+        if (timeout == 0) {
+            lock.CancelSleep();
+            return RESULT_TIMEOUT;
+        }
+
+        current_thread->SetArbiterWaitAddress(address);
+        InsertThread(SharedFrom(current_thread));
+        current_thread->SetStatus(ThreadStatus::WaitArb);
+        current_thread->WaitForArbitration(true);
+    }
+
+    if (event_handle != InvalidHandle) {
+        auto& time_manager = kernel.TimeManager();
+        time_manager.UnscheduleTimeEvent(event_handle);
+    }
+
+    {
+        KScopedSchedulerLock lock(kernel);
+        if (current_thread->IsWaitingForArbitration()) {
+            RemoveThread(SharedFrom(current_thread));
+            current_thread->WaitForArbitration(false);
+        }
+    }
+
+    return current_thread->GetSignalingResult();
+}
+
+ResultCode AddressArbiter::WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout) {
+    auto& memory = system.Memory();
+    auto& kernel = system.Kernel();
+    Thread* current_thread = kernel.CurrentScheduler()->GetCurrentThread();
+
+    Handle event_handle = InvalidHandle;
+    {
+        KScopedSchedulerLockAndSleep lock(kernel, event_handle, current_thread, timeout);
+
+        if (current_thread->IsPendingTermination()) {
+            lock.CancelSleep();
+            return ERR_THREAD_TERMINATING;
+        }
+
+        // Ensure that we can read the address.
+        if (!memory.IsValidVirtualAddress(address)) {
+            lock.CancelSleep();
+            return ERR_INVALID_ADDRESS_STATE;
+        }
+
+        s32 current_value = static_cast<s32>(memory.Read32(address));
+        if (current_value != value) {
+            lock.CancelSleep();
+            return ERR_INVALID_STATE;
+        }
+
+        // Short-circuit without rescheduling, if timeout is zero.
+        if (timeout == 0) {
+            lock.CancelSleep();
+            return RESULT_TIMEOUT;
+        }
+
+        current_thread->SetSynchronizationResults(nullptr, RESULT_TIMEOUT);
+        current_thread->SetArbiterWaitAddress(address);
+        InsertThread(SharedFrom(current_thread));
+        current_thread->SetStatus(ThreadStatus::WaitArb);
+        current_thread->WaitForArbitration(true);
+    }
+
+    if (event_handle != InvalidHandle) {
+        auto& time_manager = kernel.TimeManager();
+        time_manager.UnscheduleTimeEvent(event_handle);
+    }
+
+    {
+        KScopedSchedulerLock lock(kernel);
+        if (current_thread->IsWaitingForArbitration()) {
+            RemoveThread(SharedFrom(current_thread));
+            current_thread->WaitForArbitration(false);
+        }
+    }
+
+    return current_thread->GetSignalingResult();
+}
+
+void AddressArbiter::InsertThread(std::shared_ptr<Thread> thread) {
+    const VAddr arb_addr = thread->GetArbiterWaitAddress();
+    std::list<std::shared_ptr<Thread>>& thread_list = arb_threads[arb_addr];
+
+    const auto iter =
+        std::find_if(thread_list.cbegin(), thread_list.cend(), [&thread](const auto& entry) {
+            return entry->GetPriority() >= thread->GetPriority();
+        });
+
+    if (iter == thread_list.cend()) {
+        thread_list.push_back(std::move(thread));
+    } else {
+        thread_list.insert(iter, std::move(thread));
+    }
+}
+
+void AddressArbiter::RemoveThread(std::shared_ptr<Thread> thread) {
+    const VAddr arb_addr = thread->GetArbiterWaitAddress();
+    std::list<std::shared_ptr<Thread>>& thread_list = arb_threads[arb_addr];
+
+    const auto iter = std::find_if(thread_list.cbegin(), thread_list.cend(),
+                                   [&thread](const auto& entry) { return thread == entry; });
+
+    if (iter != thread_list.cend()) {
+        thread_list.erase(iter);
+    }
+}
+
+std::vector<std::shared_ptr<Thread>> AddressArbiter::GetThreadsWaitingOnAddress(
+    VAddr address) const {
+    const auto iter = arb_threads.find(address);
+    if (iter == arb_threads.cend()) {
+        return {};
+    }
+
+    const std::list<std::shared_ptr<Thread>>& thread_list = iter->second;
+    return {thread_list.cbegin(), thread_list.cend()};
+}
+} // namespace Kernel

src/core/hle/kernel/address_arbiter.h

@@ -0,0 +1,91 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <list>
+#include <memory>
+#include <unordered_map>
+#include <vector>
+
+#include "common/common_types.h"
+
+union ResultCode;
+
+namespace Core {
+class System;
+}
+
+namespace Kernel {
+
+class Thread;
+
+class AddressArbiter {
+public:
+    enum class ArbitrationType {
+        WaitIfLessThan = 0,
+        DecrementAndWaitIfLessThan = 1,
+        WaitIfEqual = 2,
+    };
+
+    enum class SignalType {
+        Signal = 0,
+        IncrementAndSignalIfEqual = 1,
+        ModifyByWaitingCountAndSignalIfEqual = 2,
+    };
+
+    explicit AddressArbiter(Core::System& system);
+    ~AddressArbiter();
+
+    AddressArbiter(const AddressArbiter&) = delete;
+    AddressArbiter& operator=(const AddressArbiter&) = delete;
+
+    AddressArbiter(AddressArbiter&&) = default;
+    AddressArbiter& operator=(AddressArbiter&&) = delete;
+
+    /// Signals an address being waited on with a particular signaling type.
+    ResultCode SignalToAddress(VAddr address, SignalType type, s32 value, s32 num_to_wake);
+
+    /// Waits on an address with a particular arbitration type.
+    ResultCode WaitForAddress(VAddr address, ArbitrationType type, s32 value, s64 timeout_ns);
+
+private:
+    /// Signals an address being waited on.
+    ResultCode SignalToAddressOnly(VAddr address, s32 num_to_wake);
+
+    /// Signals an address being waited on and increments its value if equal to the value argument.
+    ResultCode IncrementAndSignalToAddressIfEqual(VAddr address, s32 value, s32 num_to_wake);
+
+    /// Signals an address being waited on and modifies its value based on waiting thread count if
+    /// equal to the value argument.
+    ResultCode ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 value,
+                                                             s32 num_to_wake);
+
+    /// Waits on an address if the value passed is less than the argument value,
+    /// optionally decrementing.
+    ResultCode WaitForAddressIfLessThan(VAddr address, s32 value, s64 timeout,
+                                        bool should_decrement);
+
+    /// Waits on an address if the value passed is equal to the argument value.
+    ResultCode WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout);
+
+    /// Wake up num_to_wake (or all) threads in a vector.
+    void WakeThreads(const std::vector<std::shared_ptr<Thread>>& waiting_threads, s32 num_to_wake);
+
+    /// Insert a thread into the address arbiter container
+    void InsertThread(std::shared_ptr<Thread> thread);
+
+    /// Removes a thread from the address arbiter container
+    void RemoveThread(std::shared_ptr<Thread> thread);
+
+    // Gets the threads waiting on an address.
+    std::vector<std::shared_ptr<Thread>> GetThreadsWaitingOnAddress(VAddr address) const;
+
+    /// List of threads waiting for a address arbiter
+    std::unordered_map<VAddr, std::list<std::shared_ptr<Thread>>> arb_threads;
+
+    Core::System& system;
+};
+
+} // namespace Kernel

src/core/hle/kernel/client_port.cpp

@@ -0,0 +1,50 @@
+// Copyright 2016 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/kernel/client_port.h"
+#include "core/hle/kernel/client_session.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/hle_ipc.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/server_port.h"
+#include "core/hle/kernel/session.h"
+
+namespace Kernel {
+
+ClientPort::ClientPort(KernelCore& kernel) : Object{kernel} {}
+ClientPort::~ClientPort() = default;
+
+std::shared_ptr<ServerPort> ClientPort::GetServerPort() const {
+    return server_port;
+}
+
+ResultVal<std::shared_ptr<ClientSession>> ClientPort::Connect() {
+    if (active_sessions >= max_sessions) {
+        return ERR_MAX_CONNECTIONS_REACHED;
+    }
+    active_sessions++;
+
+    auto [client, server] = Kernel::Session::Create(kernel, name);
+
+    if (server_port->HasHLEHandler()) {
+        server_port->GetHLEHandler()->ClientConnected(std::move(server));
+    } else {
+        server_port->AppendPendingSession(std::move(server));
+    }
+
+    // Wake the threads waiting on the ServerPort
+    server_port->Signal();
+
+    return MakeResult(std::move(client));
+}
+
+void ClientPort::ConnectionClosed() {
+    if (active_sessions == 0) {
+        return;
+    }
+
+    --active_sessions;
+}
+
+} // namespace Kernel

src/core/hle/kernel/client_port.h

@@ -0,0 +1,61 @@
+// Copyright 2016 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <string>
+
+#include "common/common_types.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/result.h"
+
+namespace Kernel {
+
+class ClientSession;
+class KernelCore;
+class ServerPort;
+
+class ClientPort final : public Object {
+public:
+    explicit ClientPort(KernelCore& kernel);
+    ~ClientPort() override;
+
+    friend class ServerPort;
+    std::string GetTypeName() const override {
+        return "ClientPort";
+    }
+    std::string GetName() const override {
+        return name;
+    }
+
+    static constexpr HandleType HANDLE_TYPE = HandleType::ClientPort;
+    HandleType GetHandleType() const override {
+        return HANDLE_TYPE;
+    }
+
+    std::shared_ptr<ServerPort> GetServerPort() const;
+
+    /**
+     * Creates a new Session pair, adds the created ServerSession to the associated ServerPort's
+     * list of pending sessions, and signals the ServerPort, causing any threads
+     * waiting on it to awake.
+     * @returns ClientSession The client endpoint of the created Session pair, or error code.
+     */
+    ResultVal<std::shared_ptr<ClientSession>> Connect();
+
+    /**
+     * Signifies that a previously active connection has been closed,
+     * decreasing the total number of active connections to this port.
+     */
+    void ConnectionClosed();
+
+private:
+    std::shared_ptr<ServerPort> server_port; ///< ServerPort associated with this client port.
+    u32 max_sessions = 0;    ///< Maximum number of simultaneous sessions the port can have
+    u32 active_sessions = 0; ///< Number of currently open sessions to this port
+    std::string name;        ///< Name of client port (optional)
+};
+
+} // namespace Kernel

src/core/hle/kernel/client_session.cpp

@@ -0,0 +1,62 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/kernel/client_session.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/hle_ipc.h"
+#include "core/hle/kernel/server_session.h"
+#include "core/hle/kernel/session.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/result.h"
+
+namespace Kernel {
+
+ClientSession::ClientSession(KernelCore& kernel) : SynchronizationObject{kernel} {}
+
+ClientSession::~ClientSession() {
+    // This destructor will be called automatically when the last ClientSession handle is closed by
+    // the emulated application.
+    if (parent->Server()) {
+        parent->Server()->ClientDisconnected();
+    }
+}
+
+bool ClientSession::ShouldWait(const Thread* thread) const {
+    UNIMPLEMENTED();
+    return {};
+}
+
+void ClientSession::Acquire(Thread* thread) {
+    UNIMPLEMENTED();
+}
+
+bool ClientSession::IsSignaled() const {
+    UNIMPLEMENTED();
+    return true;
+}
+
+ResultVal<std::shared_ptr<ClientSession>> ClientSession::Create(KernelCore& kernel,
+                                                                std::shared_ptr<Session> parent,
+                                                                std::string name) {
+    std::shared_ptr<ClientSession> client_session{std::make_shared<ClientSession>(kernel)};
+
+    client_session->name = std::move(name);
+    client_session->parent = std::move(parent);
+
+    return MakeResult(std::move(client_session));
+}
+
+ResultCode ClientSession::SendSyncRequest(std::shared_ptr<Thread> thread,
+                                          Core::Memory::Memory& memory,
+                                          Core::Timing::CoreTiming& core_timing) {
+    // Keep ServerSession alive until we're done working with it.
+    if (!parent->Server()) {
+        return ERR_SESSION_CLOSED_BY_REMOTE;
+    }
+
+    // Signal the server session that new data is available
+    return parent->Server()->HandleSyncRequest(std::move(thread), memory, core_timing);
+}
+
+} // namespace Kernel

src/core/hle/kernel/client_session.h

@@ -0,0 +1,70 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <string>
+
+#include "core/hle/kernel/synchronization_object.h"
+#include "core/hle/result.h"
+
+union ResultCode;
+
+namespace Core::Memory {
+class Memory;
+}
+
+namespace Core::Timing {
+class CoreTiming;
+}
+
+namespace Kernel {
+
+class KernelCore;
+class Session;
+class Thread;
+
+class ClientSession final : public SynchronizationObject {
+public:
+    explicit ClientSession(KernelCore& kernel);
+    ~ClientSession() override;
+
+    friend class Session;
+
+    std::string GetTypeName() const override {
+        return "ClientSession";
+    }
+
+    std::string GetName() const override {
+        return name;
+    }
+
+    static constexpr HandleType HANDLE_TYPE = HandleType::ClientSession;
+    HandleType GetHandleType() const override {
+        return HANDLE_TYPE;
+    }
+
+    ResultCode SendSyncRequest(std::shared_ptr<Thread> thread, Core::Memory::Memory& memory,
+                               Core::Timing::CoreTiming& core_timing);
+
+    bool ShouldWait(const Thread* thread) const override;
+
+    void Acquire(Thread* thread) override;
+
+    bool IsSignaled() const override;
+
+private:
+    static ResultVal<std::shared_ptr<ClientSession>> Create(KernelCore& kernel,
+                                                            std::shared_ptr<Session> parent,
+                                                            std::string name = "Unknown");
+
+    /// The parent session, which links to the server endpoint.
+    std::shared_ptr<Session> parent;
+
+    /// Name of the client session (optional)
+    std::string name;
+};
+
+} // namespace Kernel

src/core/hle/kernel/code_set.cpp

@@ -0,0 +1,12 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/kernel/code_set.h"
+
+namespace Kernel {
+
+CodeSet::CodeSet() = default;
+CodeSet::~CodeSet() = default;
+
+} // namespace Kernel

src/core/hle/kernel/code_set.h

@@ -0,0 +1,90 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <cstddef>
+#include <vector>
+
+#include "common/common_types.h"
+#include "core/hle/kernel/physical_memory.h"
+
+namespace Kernel {
+
+/**
+ * Represents executable data that may be loaded into a kernel process.
+ *
+ * A code set consists of three basic segments:
+ *   - A code (AKA text) segment,
+ *   - A read-only data segment (rodata)
+ *   - A data segment
+ *
+ * The code segment is the portion of the object file that contains
+ * executable instructions.
+ *
+ * The read-only data segment in the portion of the object file that
+ * contains (as one would expect) read-only data, such as fixed constant
+ * values and data structures.
+ *
+ * The data segment is similar to the read-only data segment -- it contains
+ * variables and data structures that have predefined values, however,
+ * entities within this segment can be modified.
+ */
+struct CodeSet final {
+    /// A single segment within a code set.
+    struct Segment final {
+        /// The byte offset that this segment is located at.
+        std::size_t offset = 0;
+
+        /// The address to map this segment to.
+        VAddr addr = 0;
+
+        /// The size of this segment in bytes.
+        u32 size = 0;
+    };
+
+    explicit CodeSet();
+    ~CodeSet();
+
+    CodeSet(const CodeSet&) = delete;
+    CodeSet& operator=(const CodeSet&) = delete;
+
+    CodeSet(CodeSet&&) = default;
+    CodeSet& operator=(CodeSet&&) = default;
+
+    Segment& CodeSegment() {
+        return segments[0];
+    }
+
+    const Segment& CodeSegment() const {
+        return segments[0];
+    }
+
+    Segment& RODataSegment() {
+        return segments[1];
+    }
+
+    const Segment& RODataSegment() const {
+        return segments[1];
+    }
+
+    Segment& DataSegment() {
+        return segments[2];
+    }
+
+    const Segment& DataSegment() const {
+        return segments[2];
+    }
+
+    /// The overall data that backs this code set.
+    Kernel::PhysicalMemory memory;
+
+    /// The segments that comprise this code set.
+    std::array<Segment, 3> segments;
+
+    /// The entry point address for this code set.
+    VAddr entrypoint = 0;
+};
+
+} // namespace Kernel

src/core/hle/kernel/errors.h

@@ -0,0 +1,40 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/result.h"
+
+namespace Kernel {
+
+// Confirmed Switch kernel error codes
+
+constexpr ResultCode ERR_MAX_CONNECTIONS_REACHED{ErrorModule::Kernel, 7};
+constexpr ResultCode ERR_INVALID_CAPABILITY_DESCRIPTOR{ErrorModule::Kernel, 14};
+constexpr ResultCode ERR_THREAD_TERMINATING{ErrorModule::Kernel, 59};
+constexpr ResultCode ERR_INVALID_SIZE{ErrorModule::Kernel, 101};
+constexpr ResultCode ERR_INVALID_ADDRESS{ErrorModule::Kernel, 102};
+constexpr ResultCode ERR_OUT_OF_RESOURCES{ErrorModule::Kernel, 103};
+constexpr ResultCode ERR_OUT_OF_MEMORY{ErrorModule::Kernel, 104};
+constexpr ResultCode ERR_HANDLE_TABLE_FULL{ErrorModule::Kernel, 105};
+constexpr ResultCode ERR_INVALID_ADDRESS_STATE{ErrorModule::Kernel, 106};
+constexpr ResultCode ERR_INVALID_MEMORY_PERMISSIONS{ErrorModule::Kernel, 108};
+constexpr ResultCode ERR_INVALID_MEMORY_RANGE{ErrorModule::Kernel, 110};
+constexpr ResultCode ERR_INVALID_PROCESSOR_ID{ErrorModule::Kernel, 113};
+constexpr ResultCode ERR_INVALID_THREAD_PRIORITY{ErrorModule::Kernel, 112};
+constexpr ResultCode ERR_INVALID_HANDLE{ErrorModule::Kernel, 114};
+constexpr ResultCode ERR_INVALID_POINTER{ErrorModule::Kernel, 115};
+constexpr ResultCode ERR_INVALID_COMBINATION{ErrorModule::Kernel, 116};
+constexpr ResultCode RESULT_TIMEOUT{ErrorModule::Kernel, 117};
+constexpr ResultCode ERR_SYNCHRONIZATION_CANCELED{ErrorModule::Kernel, 118};
+constexpr ResultCode ERR_OUT_OF_RANGE{ErrorModule::Kernel, 119};
+constexpr ResultCode ERR_INVALID_ENUM_VALUE{ErrorModule::Kernel, 120};
+constexpr ResultCode ERR_NOT_FOUND{ErrorModule::Kernel, 121};
+constexpr ResultCode ERR_BUSY{ErrorModule::Kernel, 122};
+constexpr ResultCode ERR_SESSION_CLOSED_BY_REMOTE{ErrorModule::Kernel, 123};
+constexpr ResultCode ERR_INVALID_STATE{ErrorModule::Kernel, 125};
+constexpr ResultCode ERR_RESERVED_VALUE{ErrorModule::Kernel, 126};
+constexpr ResultCode ERR_RESOURCE_LIMIT_EXCEEDED{ErrorModule::Kernel, 132};
+
+} // namespace Kernel

src/core/hle/kernel/global_scheduler_context.cpp

@@ -0,0 +1,52 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <mutex>
+
+#include "common/assert.h"
+#include "core/core.h"
+#include "core/hle/kernel/global_scheduler_context.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/kernel.h"
+
+namespace Kernel {
+
+GlobalSchedulerContext::GlobalSchedulerContext(KernelCore& kernel)
+    : kernel{kernel}, scheduler_lock{kernel} {}
+
+GlobalSchedulerContext::~GlobalSchedulerContext() = default;
+
+void GlobalSchedulerContext::AddThread(std::shared_ptr<Thread> thread) {
+    std::scoped_lock lock{global_list_guard};
+    thread_list.push_back(std::move(thread));
+}
+
+void GlobalSchedulerContext::RemoveThread(std::shared_ptr<Thread> thread) {
+    std::scoped_lock lock{global_list_guard};
+    thread_list.erase(std::remove(thread_list.begin(), thread_list.end(), thread),
+                      thread_list.end());
+}
+
+void GlobalSchedulerContext::PreemptThreads() {
+    // The priority levels at which the global scheduler preempts threads every 10 ms. They are
+    // ordered from Core 0 to Core 3.
+    static constexpr std::array<u32, Core::Hardware::NUM_CPU_CORES> preemption_priorities{
+        59,
+        59,
+        59,
+        63,
+    };
+
+    ASSERT(IsLocked());
+    for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
+        const u32 priority = preemption_priorities[core_id];
+        kernel.Scheduler(core_id).RotateScheduledQueue(core_id, priority);
+    }
+}
+
+bool GlobalSchedulerContext::IsLocked() const {
+    return scheduler_lock.IsLockedByCurrentThread();
+}
+
+} // namespace Kernel

src/core/hle/kernel/global_scheduler_context.h

@@ -0,0 +1,81 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <atomic>
+#include <vector>
+
+#include "common/common_types.h"
+#include "common/spin_lock.h"
+#include "core/hardware_properties.h"
+#include "core/hle/kernel/k_priority_queue.h"
+#include "core/hle/kernel/k_scheduler_lock.h"
+#include "core/hle/kernel/thread.h"
+
+namespace Kernel {
+
+class KernelCore;
+class SchedulerLock;
+
+using KSchedulerPriorityQueue =
+    KPriorityQueue<Thread, Core::Hardware::NUM_CPU_CORES, THREADPRIO_LOWEST, THREADPRIO_HIGHEST>;
+constexpr s32 HighestCoreMigrationAllowedPriority = 2;
+
+class GlobalSchedulerContext final {
+    friend class KScheduler;
+
+public:
+    using LockType = KAbstractSchedulerLock<KScheduler>;
+
+    explicit GlobalSchedulerContext(KernelCore& kernel);
+    ~GlobalSchedulerContext();
+
+    /// Adds a new thread to the scheduler
+    void AddThread(std::shared_ptr<Thread> thread);
+
+    /// Removes a thread from the scheduler
+    void RemoveThread(std::shared_ptr<Thread> thread);
+
+    /// Returns a list of all threads managed by the scheduler
+    [[nodiscard]] const std::vector<std::shared_ptr<Thread>>& GetThreadList() const {
+        return thread_list;
+    }
+
+    /**
+     * Rotates the scheduling queues of threads at a preemption priority and then does
+     * some core rebalancing. Preemption priorities can be found in the array
+     * 'preemption_priorities'.
+     *
+     * @note This operation happens every 10ms.
+     */
+    void PreemptThreads();
+
+    /// Returns true if the global scheduler lock is acquired
+    bool IsLocked() const;
+
+    [[nodiscard]] LockType& SchedulerLock() {
+        return scheduler_lock;
+    }
+
+    [[nodiscard]] const LockType& SchedulerLock() const {
+        return scheduler_lock;
+    }
+
+private:
+    friend class KScopedSchedulerLock;
+    friend class KScopedSchedulerLockAndSleep;
+
+    KernelCore& kernel;
+
+    std::atomic_bool scheduler_update_needed{};
+    KSchedulerPriorityQueue priority_queue;
+    LockType scheduler_lock;
+
+    /// Lists all thread ids that aren't deleted/etc.
+    std::vector<std::shared_ptr<Thread>> thread_list;
+    Common::SpinLock global_list_guard{};
+};
+
+} // namespace Kernel

src/core/hle/kernel/handle_table.cpp

@@ -0,0 +1,127 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <utility>
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "core/core.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/kernel/thread.h"
+
+namespace Kernel {
+namespace {
+constexpr u16 GetSlot(Handle handle) {
+    return static_cast<u16>(handle >> 15);
+}
+
+constexpr u16 GetGeneration(Handle handle) {
+    return static_cast<u16>(handle & 0x7FFF);
+}
+} // Anonymous namespace
+
+HandleTable::HandleTable(KernelCore& kernel) : kernel{kernel} {
+    Clear();
+}
+
+HandleTable::~HandleTable() = default;
+
+ResultCode HandleTable::SetSize(s32 handle_table_size) {
+    if (static_cast<u32>(handle_table_size) > MAX_COUNT) {
+        LOG_ERROR(Kernel, "Handle table size {} is greater than {}", handle_table_size, MAX_COUNT);
+        return ERR_OUT_OF_MEMORY;
+    }
+
+    // Values less than or equal to zero indicate to use the maximum allowable
+    // size for the handle table in the actual kernel, so we ignore the given
+    // value in that case, since we assume this by default unless this function
+    // is called.
+    if (handle_table_size > 0) {
+        table_size = static_cast<u16>(handle_table_size);
+    }
+
+    return RESULT_SUCCESS;
+}
+
+ResultVal<Handle> HandleTable::Create(std::shared_ptr<Object> obj) {
+    DEBUG_ASSERT(obj != nullptr);
+
+    const u16 slot = next_free_slot;
+    if (slot >= table_size) {
+        LOG_ERROR(Kernel, "Unable to allocate Handle, too many slots in use.");
+        return ERR_HANDLE_TABLE_FULL;
+    }
+    next_free_slot = generations[slot];
+
+    const u16 generation = next_generation++;
+
+    // Overflow count so it fits in the 15 bits dedicated to the generation in the handle.
+    // Horizon OS uses zero to represent an invalid handle, so skip to 1.
+    if (next_generation >= (1 << 15)) {
+        next_generation = 1;
+    }
+
+    generations[slot] = generation;
+    objects[slot] = std::move(obj);
+
+    Handle handle = generation | (slot << 15);
+    return MakeResult<Handle>(handle);
+}
+
+ResultVal<Handle> HandleTable::Duplicate(Handle handle) {
+    std::shared_ptr<Object> object = GetGeneric(handle);
+    if (object == nullptr) {
+        LOG_ERROR(Kernel, "Tried to duplicate invalid handle: {:08X}", handle);
+        return ERR_INVALID_HANDLE;
+    }
+    return Create(std::move(object));
+}
+
+ResultCode HandleTable::Close(Handle handle) {
+    if (!IsValid(handle)) {
+        LOG_ERROR(Kernel, "Handle is not valid! handle={:08X}", handle);
+        return ERR_INVALID_HANDLE;
+    }
+
+    const u16 slot = GetSlot(handle);
+
+    objects[slot] = nullptr;
+
+    generations[slot] = next_free_slot;
+    next_free_slot = slot;
+    return RESULT_SUCCESS;
+}
+
+bool HandleTable::IsValid(Handle handle) const {
+    const std::size_t slot = GetSlot(handle);
+    const u16 generation = GetGeneration(handle);
+
+    return slot < table_size && objects[slot] != nullptr && generations[slot] == generation;
+}
+
+std::shared_ptr<Object> HandleTable::GetGeneric(Handle handle) const {
+    if (handle == CurrentThread) {
+        return SharedFrom(kernel.CurrentScheduler()->GetCurrentThread());
+    } else if (handle == CurrentProcess) {
+        return SharedFrom(kernel.CurrentProcess());
+    }
+
+    if (!IsValid(handle)) {
+        return nullptr;
+    }
+    return objects[GetSlot(handle)];
+}
+
+void HandleTable::Clear() {
+    for (u16 i = 0; i < table_size; ++i) {
+        generations[i] = static_cast<u16>(i + 1);
+        objects[i] = nullptr;
+    }
+    next_free_slot = 0;
+}
+
+} // namespace Kernel

src/core/hle/kernel/handle_table.h

@@ -0,0 +1,144 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <cstddef>
+#include <memory>
+
+#include "common/common_types.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/result.h"
+
+namespace Kernel {
+
+class KernelCore;
+
+enum KernelHandle : Handle {
+    InvalidHandle = 0,
+    CurrentThread = 0xFFFF8000,
+    CurrentProcess = 0xFFFF8001,
+};
+
+/**
+ * This class allows the creation of Handles, which are references to objects that can be tested
+ * for validity and looked up. Here they are used to pass references to kernel objects to/from the
+ * emulated process. it has been designed so that it follows the same handle format and has
+ * approximately the same restrictions as the handle manager in the CTR-OS.
+ *
+ * Handles contain two sub-fields: a slot index (bits 31:15) and a generation value (bits 14:0).
+ * The slot index is used to index into the arrays in this class to access the data corresponding
+ * to the Handle.
+ *
+ * To prevent accidental use of a freed Handle whose slot has already been reused, a global counter
+ * is kept and incremented every time a Handle is created. This is the Handle's "generation". The
+ * value of the counter is stored into the Handle as well as in the handle table (in the
+ * "generations" array). When looking up a handle, the Handle's generation must match with the
+ * value stored on the class, otherwise the Handle is considered invalid.
+ *
+ * To find free slots when allocating a Handle without needing to scan the entire object array, the
+ * generations field of unallocated slots is re-purposed as a linked list of indices to free slots.
+ * When a Handle is created, an index is popped off the list and used for the new Handle. When it
+ * is destroyed, it is again pushed onto the list to be re-used by the next allocation. It is
+ * likely that this allocation strategy differs from the one used in CTR-OS, but this hasn't been
+ * verified and isn't likely to cause any problems.
+ */
+class HandleTable final : NonCopyable {
+public:
+    /// This is the maximum limit of handles allowed per process in Horizon
+    static constexpr std::size_t MAX_COUNT = 1024;
+
+    explicit HandleTable(KernelCore& kernel);
+    ~HandleTable();
+
+    /**
+     * Sets the number of handles that may be in use at one time
+     * for this handle table.
+     *
+     * @param handle_table_size The desired size to limit the handle table to.
+     *
+     * @returns an error code indicating if initialization was successful.
+     *          If initialization was not successful, then ERR_OUT_OF_MEMORY
+     *          will be returned.
+     *
+     * @pre handle_table_size must be within the range [0, 1024]
+     */
+    ResultCode SetSize(s32 handle_table_size);
+
+    /**
+     * Allocates a handle for the given object.
+     * @return The created Handle or one of the following errors:
+     *           - `ERR_HANDLE_TABLE_FULL`: the maximum number of handles has been exceeded.
+     */
+    ResultVal<Handle> Create(std::shared_ptr<Object> obj);
+
+    /**
+     * Returns a new handle that points to the same object as the passed in handle.
+     * @return The duplicated Handle or one of the following errors:
+     *           - `ERR_INVALID_HANDLE`: an invalid handle was passed in.
+     *           - Any errors returned by `Create()`.
+     */
+    ResultVal<Handle> Duplicate(Handle handle);
+
+    /**
+     * Closes a handle, removing it from the table and decreasing the object's ref-count.
+     * @return `RESULT_SUCCESS` or one of the following errors:
+     *           - `ERR_INVALID_HANDLE`: an invalid handle was passed in.
+     */
+    ResultCode Close(Handle handle);
+
+    /// Checks if a handle is valid and points to an existing object.
+    bool IsValid(Handle handle) const;
+
+    /**
+     * Looks up a handle.
+     * @return Pointer to the looked-up object, or `nullptr` if the handle is not valid.
+     */
+    std::shared_ptr<Object> GetGeneric(Handle handle) const;
+
+    /**
+     * Looks up a handle while verifying its type.
+     * @return Pointer to the looked-up object, or `nullptr` if the handle is not valid or its
+     *         type differs from the requested one.
+     */
+    template <class T>
+    std::shared_ptr<T> Get(Handle handle) const {
+        return DynamicObjectCast<T>(GetGeneric(handle));
+    }
+
+    /// Closes all handles held in this table.
+    void Clear();
+
+private:
+    /// Stores the Object referenced by the handle or null if the slot is empty.
+    std::array<std::shared_ptr<Object>, MAX_COUNT> objects;
+
+    /**
+     * The value of `next_generation` when the handle was created, used to check for validity. For
+     * empty slots, contains the index of the next free slot in the list.
+     */
+    std::array<u16, MAX_COUNT> generations;
+
+    /**
+     * The limited size of the handle table. This can be specified by process
+     * capabilities in order to restrict the overall number of handles that
+     * can be created in a process instance
+     */
+    u16 table_size = static_cast<u16>(MAX_COUNT);
+
+    /**
+     * Global counter of the number of created handles. Stored in `generations` when a handle is
+     * created, and wraps around to 1 when it hits 0x8000.
+     */
+    u16 next_generation = 1;
+
+    /// Head of the free slots linked list.
+    u16 next_free_slot = 0;
+
+    /// Underlying kernel instance that this handle table operates under.
+    KernelCore& kernel;
+};
+
+} // namespace Kernel

src/core/hle/kernel/hle_ipc.cpp

@@ -0,0 +1,392 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <array>
+#include <sstream>
+#include <utility>
+
+#include <boost/range/algorithm_ext/erase.hpp>
+
+#include "common/assert.h"
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+#include "common/logging/log.h"
+#include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/hle_ipc.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/kernel/readable_event.h"
+#include "core/hle/kernel/server_session.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/time_manager.h"
+#include "core/hle/kernel/writable_event.h"
+#include "core/memory.h"
+
+namespace Kernel {
+
+SessionRequestHandler::SessionRequestHandler() = default;
+
+SessionRequestHandler::~SessionRequestHandler() = default;
+
+void SessionRequestHandler::ClientConnected(std::shared_ptr<ServerSession> server_session) {
+    server_session->SetHleHandler(shared_from_this());
+    connected_sessions.push_back(std::move(server_session));
+}
+
+void SessionRequestHandler::ClientDisconnected(
+    const std::shared_ptr<ServerSession>& server_session) {
+    server_session->SetHleHandler(nullptr);
+    boost::range::remove_erase(connected_sessions, server_session);
+}
+
+HLERequestContext::HLERequestContext(KernelCore& kernel, Core::Memory::Memory& memory,
+                                     std::shared_ptr<ServerSession> server_session,
+                                     std::shared_ptr<Thread> thread)
+    : server_session(std::move(server_session)),
+      thread(std::move(thread)), kernel{kernel}, memory{memory} {
+    cmd_buf[0] = 0;
+}
+
+HLERequestContext::~HLERequestContext() = default;
+
+void HLERequestContext::ParseCommandBuffer(const HandleTable& handle_table, u32_le* src_cmdbuf,
+                                           bool incoming) {
+    IPC::RequestParser rp(src_cmdbuf);
+    command_header = rp.PopRaw<IPC::CommandHeader>();
+
+    if (command_header->type == IPC::CommandType::Close) {
+        // Close does not populate the rest of the IPC header
+        return;
+    }
+
+    // If handle descriptor is present, add size of it
+    if (command_header->enable_handle_descriptor) {
+        handle_descriptor_header = rp.PopRaw<IPC::HandleDescriptorHeader>();
+        if (handle_descriptor_header->send_current_pid) {
+            rp.Skip(2, false);
+        }
+        if (incoming) {
+            // Populate the object lists with the data in the IPC request.
+            for (u32 handle = 0; handle < handle_descriptor_header->num_handles_to_copy; ++handle) {
+                copy_objects.push_back(handle_table.GetGeneric(rp.Pop<Handle>()));
+            }
+            for (u32 handle = 0; handle < handle_descriptor_header->num_handles_to_move; ++handle) {
+                move_objects.push_back(handle_table.GetGeneric(rp.Pop<Handle>()));
+            }
+        } else {
+            // For responses we just ignore the handles, they're empty and will be populated when
+            // translating the response.
+            rp.Skip(handle_descriptor_header->num_handles_to_copy, false);
+            rp.Skip(handle_descriptor_header->num_handles_to_move, false);
+        }
+    }
+
+    for (unsigned i = 0; i < command_header->num_buf_x_descriptors; ++i) {
+        buffer_x_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorX>());
+    }
+    for (unsigned i = 0; i < command_header->num_buf_a_descriptors; ++i) {
+        buffer_a_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorABW>());
+    }
+    for (unsigned i = 0; i < command_header->num_buf_b_descriptors; ++i) {
+        buffer_b_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorABW>());
+    }
+    for (unsigned i = 0; i < command_header->num_buf_w_descriptors; ++i) {
+        buffer_w_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorABW>());
+    }
+
+    buffer_c_offset = rp.GetCurrentOffset() + command_header->data_size;
+
+    // Padding to align to 16 bytes
+    rp.AlignWithPadding();
+
+    if (Session()->IsDomain() && ((command_header->type == IPC::CommandType::Request ||
+                                   command_header->type == IPC::CommandType::RequestWithContext) ||
+                                  !incoming)) {
+        // If this is an incoming message, only CommandType "Request" has a domain header
+        // All outgoing domain messages have the domain header, if only incoming has it
+        if (incoming || domain_message_header) {
+            domain_message_header = rp.PopRaw<IPC::DomainMessageHeader>();
+        } else {
+            if (Session()->IsDomain()) {
+                LOG_WARNING(IPC, "Domain request has no DomainMessageHeader!");
+            }
+        }
+    }
+
+    data_payload_header = rp.PopRaw<IPC::DataPayloadHeader>();
+
+    data_payload_offset = rp.GetCurrentOffset();
+
+    if (domain_message_header && domain_message_header->command ==
+                                     IPC::DomainMessageHeader::CommandType::CloseVirtualHandle) {
+        // CloseVirtualHandle command does not have SFC* or any data
+        return;
+    }
+
+    if (incoming) {
+        ASSERT(data_payload_header->magic == Common::MakeMagic('S', 'F', 'C', 'I'));
+    } else {
+        ASSERT(data_payload_header->magic == Common::MakeMagic('S', 'F', 'C', 'O'));
+    }
+
+    rp.SetCurrentOffset(buffer_c_offset);
+
+    // For Inline buffers, the response data is written directly to buffer_c_offset
+    // and in this case we don't have any BufferDescriptorC on the request.
+    if (command_header->buf_c_descriptor_flags >
+        IPC::CommandHeader::BufferDescriptorCFlag::InlineDescriptor) {
+        if (command_header->buf_c_descriptor_flags ==
+            IPC::CommandHeader::BufferDescriptorCFlag::OneDescriptor) {
+            buffer_c_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorC>());
+        } else {
+            unsigned num_buf_c_descriptors =
+                static_cast<unsigned>(command_header->buf_c_descriptor_flags.Value()) - 2;
+
+            // This is used to detect possible underflows, in case something is broken
+            // with the two ifs above and the flags value is == 0 || == 1.
+            ASSERT(num_buf_c_descriptors < 14);
+
+            for (unsigned i = 0; i < num_buf_c_descriptors; ++i) {
+                buffer_c_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorC>());
+            }
+        }
+    }
+
+    rp.SetCurrentOffset(data_payload_offset);
+
+    command = rp.Pop<u32_le>();
+    rp.Skip(1, false); // The command is actually an u64, but we don't use the high part.
+}
+
+ResultCode HLERequestContext::PopulateFromIncomingCommandBuffer(const HandleTable& handle_table,
+                                                                u32_le* src_cmdbuf) {
+    ParseCommandBuffer(handle_table, src_cmdbuf, true);
+    if (command_header->type == IPC::CommandType::Close) {
+        // Close does not populate the rest of the IPC header
+        return RESULT_SUCCESS;
+    }
+
+    // The data_size already includes the payload header, the padding and the domain header.
+    std::size_t size = data_payload_offset + command_header->data_size -
+                       sizeof(IPC::DataPayloadHeader) / sizeof(u32) - 4;
+    if (domain_message_header)
+        size -= sizeof(IPC::DomainMessageHeader) / sizeof(u32);
+    std::copy_n(src_cmdbuf, size, cmd_buf.begin());
+    return RESULT_SUCCESS;
+}
+
+ResultCode HLERequestContext::WriteToOutgoingCommandBuffer(Thread& thread) {
+    auto& owner_process = *thread.GetOwnerProcess();
+    auto& handle_table = owner_process.GetHandleTable();
+
+    std::array<u32, IPC::COMMAND_BUFFER_LENGTH> dst_cmdbuf;
+    memory.ReadBlock(owner_process, thread.GetTLSAddress(), dst_cmdbuf.data(),
+                     dst_cmdbuf.size() * sizeof(u32));
+
+    // The header was already built in the internal command buffer. Attempt to parse it to verify
+    // the integrity and then copy it over to the target command buffer.
+    ParseCommandBuffer(handle_table, cmd_buf.data(), false);
+
+    // The data_size already includes the payload header, the padding and the domain header.
+    std::size_t size = data_payload_offset + command_header->data_size -
+                       sizeof(IPC::DataPayloadHeader) / sizeof(u32) - 4;
+    if (domain_message_header)
+        size -= sizeof(IPC::DomainMessageHeader) / sizeof(u32);
+
+    std::copy_n(cmd_buf.begin(), size, dst_cmdbuf.data());
+
+    if (command_header->enable_handle_descriptor) {
+        ASSERT_MSG(!move_objects.empty() || !copy_objects.empty(),
+                   "Handle descriptor bit set but no handles to translate");
+        // We write the translated handles at a specific offset in the command buffer, this space
+        // was already reserved when writing the header.
+        std::size_t current_offset =
+            (sizeof(IPC::CommandHeader) + sizeof(IPC::HandleDescriptorHeader)) / sizeof(u32);
+        ASSERT_MSG(!handle_descriptor_header->send_current_pid, "Sending PID is not implemented");
+
+        ASSERT(copy_objects.size() == handle_descriptor_header->num_handles_to_copy);
+        ASSERT(move_objects.size() == handle_descriptor_header->num_handles_to_move);
+
+        // We don't make a distinction between copy and move handles when translating since HLE
+        // services don't deal with handles directly. However, the guest applications might check
+        // for specific values in each of these descriptors.
+        for (auto& object : copy_objects) {
+            ASSERT(object != nullptr);
+            dst_cmdbuf[current_offset++] = handle_table.Create(object).Unwrap();
+        }
+
+        for (auto& object : move_objects) {
+            ASSERT(object != nullptr);
+            dst_cmdbuf[current_offset++] = handle_table.Create(object).Unwrap();
+        }
+    }
+
+    // TODO(Subv): Translate the X/A/B/W buffers.
+
+    if (Session()->IsDomain() && domain_message_header) {
+        ASSERT(domain_message_header->num_objects == domain_objects.size());
+        // Write the domain objects to the command buffer, these go after the raw untranslated data.
+        // TODO(Subv): This completely ignores C buffers.
+        std::size_t domain_offset = size - domain_message_header->num_objects;
+
+        for (const auto& object : domain_objects) {
+            server_session->AppendDomainRequestHandler(object);
+            dst_cmdbuf[domain_offset++] =
+                static_cast<u32_le>(server_session->NumDomainRequestHandlers());
+        }
+    }
+
+    // Copy the translated command buffer back into the thread's command buffer area.
+    memory.WriteBlock(owner_process, thread.GetTLSAddress(), dst_cmdbuf.data(),
+                      dst_cmdbuf.size() * sizeof(u32));
+
+    return RESULT_SUCCESS;
+}
+
+std::vector<u8> HLERequestContext::ReadBuffer(std::size_t buffer_index) const {
+    std::vector<u8> buffer{};
+    const bool is_buffer_a{BufferDescriptorA().size() > buffer_index &&
+                           BufferDescriptorA()[buffer_index].Size()};
+
+    if (is_buffer_a) {
+        ASSERT_OR_EXECUTE_MSG(
+            BufferDescriptorA().size() > buffer_index, { return buffer; },
+            "BufferDescriptorA invalid buffer_index {}", buffer_index);
+        buffer.resize(BufferDescriptorA()[buffer_index].Size());
+        memory.ReadBlock(BufferDescriptorA()[buffer_index].Address(), buffer.data(), buffer.size());
+    } else {
+        ASSERT_OR_EXECUTE_MSG(
+            BufferDescriptorX().size() > buffer_index, { return buffer; },
+            "BufferDescriptorX invalid buffer_index {}", buffer_index);
+        buffer.resize(BufferDescriptorX()[buffer_index].Size());
+        memory.ReadBlock(BufferDescriptorX()[buffer_index].Address(), buffer.data(), buffer.size());
+    }
+
+    return buffer;
+}
+
+std::size_t HLERequestContext::WriteBuffer(const void* buffer, std::size_t size,
+                                           std::size_t buffer_index) const {
+    if (size == 0) {
+        LOG_WARNING(Core, "skip empty buffer write");
+        return 0;
+    }
+
+    const bool is_buffer_b{BufferDescriptorB().size() > buffer_index &&
+                           BufferDescriptorB()[buffer_index].Size()};
+    const std::size_t buffer_size{GetWriteBufferSize(buffer_index)};
+    if (size > buffer_size) {
+        LOG_CRITICAL(Core, "size ({:016X}) is greater than buffer_size ({:016X})", size,
+                     buffer_size);
+        size = buffer_size; // TODO(bunnei): This needs to be HW tested
+    }
+
+    if (is_buffer_b) {
+        ASSERT_OR_EXECUTE_MSG(
+            BufferDescriptorB().size() > buffer_index &&
+                BufferDescriptorB()[buffer_index].Size() >= size,
+            { return 0; }, "BufferDescriptorB is invalid, index={}, size={}", buffer_index, size);
+        memory.WriteBlock(BufferDescriptorB()[buffer_index].Address(), buffer, size);
+    } else {
+        ASSERT_OR_EXECUTE_MSG(
+            BufferDescriptorC().size() > buffer_index &&
+                BufferDescriptorC()[buffer_index].Size() >= size,
+            { return 0; }, "BufferDescriptorC is invalid, index={}, size={}", buffer_index, size);
+        memory.WriteBlock(BufferDescriptorC()[buffer_index].Address(), buffer, size);
+    }
+
+    return size;
+}
+
+std::size_t HLERequestContext::GetReadBufferSize(std::size_t buffer_index) const {
+    const bool is_buffer_a{BufferDescriptorA().size() > buffer_index &&
+                           BufferDescriptorA()[buffer_index].Size()};
+    if (is_buffer_a) {
+        ASSERT_OR_EXECUTE_MSG(
+            BufferDescriptorA().size() > buffer_index, { return 0; },
+            "BufferDescriptorA invalid buffer_index {}", buffer_index);
+        return BufferDescriptorA()[buffer_index].Size();
+    } else {
+        ASSERT_OR_EXECUTE_MSG(
+            BufferDescriptorX().size() > buffer_index, { return 0; },
+            "BufferDescriptorX invalid buffer_index {}", buffer_index);
+        return BufferDescriptorX()[buffer_index].Size();
+    }
+}
+
+std::size_t HLERequestContext::GetWriteBufferSize(std::size_t buffer_index) const {
+    const bool is_buffer_b{BufferDescriptorB().size() > buffer_index &&
+                           BufferDescriptorB()[buffer_index].Size()};
+    if (is_buffer_b) {
+        ASSERT_OR_EXECUTE_MSG(
+            BufferDescriptorB().size() > buffer_index, { return 0; },
+            "BufferDescriptorB invalid buffer_index {}", buffer_index);
+        return BufferDescriptorB()[buffer_index].Size();
+    } else {
+        ASSERT_OR_EXECUTE_MSG(
+            BufferDescriptorC().size() > buffer_index, { return 0; },
+            "BufferDescriptorC invalid buffer_index {}", buffer_index);
+        return BufferDescriptorC()[buffer_index].Size();
+    }
+    return 0;
+}
+
+std::string HLERequestContext::Description() const {
+    if (!command_header) {
+        return "No command header available";
+    }
+    std::ostringstream s;
+    s << "IPC::CommandHeader: Type:" << static_cast<u32>(command_header->type.Value());
+    s << ", X(Pointer):" << command_header->num_buf_x_descriptors;
+    if (command_header->num_buf_x_descriptors) {
+        s << '[';
+        for (u64 i = 0; i < command_header->num_buf_x_descriptors; ++i) {
+            s << "0x" << std::hex << BufferDescriptorX()[i].Size();
+            if (i < command_header->num_buf_x_descriptors - 1)
+                s << ", ";
+        }
+        s << ']';
+    }
+    s << ", A(Send):" << command_header->num_buf_a_descriptors;
+    if (command_header->num_buf_a_descriptors) {
+        s << '[';
+        for (u64 i = 0; i < command_header->num_buf_a_descriptors; ++i) {
+            s << "0x" << std::hex << BufferDescriptorA()[i].Size();
+            if (i < command_header->num_buf_a_descriptors - 1)
+                s << ", ";
+        }
+        s << ']';
+    }
+    s << ", B(Receive):" << command_header->num_buf_b_descriptors;
+    if (command_header->num_buf_b_descriptors) {
+        s << '[';
+        for (u64 i = 0; i < command_header->num_buf_b_descriptors; ++i) {
+            s << "0x" << std::hex << BufferDescriptorB()[i].Size();
+            if (i < command_header->num_buf_b_descriptors - 1)
+                s << ", ";
+        }
+        s << ']';
+    }
+    s << ", C(ReceiveList):" << BufferDescriptorC().size();
+    if (!BufferDescriptorC().empty()) {
+        s << '[';
+        for (u64 i = 0; i < BufferDescriptorC().size(); ++i) {
+            s << "0x" << std::hex << BufferDescriptorC()[i].Size();
+            if (i < BufferDescriptorC().size() - 1)
+                s << ", ";
+        }
+        s << ']';
+    }
+    s << ", data_size:" << command_header->data_size.Value();
+
+    return s.str();
+}
+
+} // namespace Kernel

src/core/hle/kernel/hle_ipc.h

@@ -0,0 +1,310 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <functional>
+#include <memory>
+#include <optional>
+#include <string>
+#include <type_traits>
+#include <vector>
+#include <boost/container/small_vector.hpp>
+#include "common/common_types.h"
+#include "common/concepts.h"
+#include "common/swap.h"
+#include "core/hle/ipc.h"
+#include "core/hle/kernel/object.h"
+
+union ResultCode;
+
+namespace Core::Memory {
+class Memory;
+}
+
+namespace IPC {
+class ResponseBuilder;
+}
+
+namespace Service {
+class ServiceFrameworkBase;
+}
+
+namespace Kernel {
+
+class Domain;
+class HandleTable;
+class HLERequestContext;
+class KernelCore;
+class Process;
+class ServerSession;
+class Thread;
+class ReadableEvent;
+class WritableEvent;
+
+enum class ThreadWakeupReason;
+
+/**
+ * Interface implemented by HLE Session handlers.
+ * This can be provided to a ServerSession in order to hook into several relevant events
+ * (such as a new connection or a SyncRequest) so they can be implemented in the emulator.
+ */
+class SessionRequestHandler : public std::enable_shared_from_this<SessionRequestHandler> {
+public:
+    SessionRequestHandler();
+    virtual ~SessionRequestHandler();
+
+    /**
+     * Handles a sync request from the emulated application.
+     * @param server_session The ServerSession that was triggered for this sync request,
+     * it should be used to differentiate which client (As in ClientSession) we're answering to.
+     * TODO(Subv): Use a wrapper structure to hold all the information relevant to
+     * this request (ServerSession, Originator thread, Translated command buffer, etc).
+     * @returns ResultCode the result code of the translate operation.
+     */
+    virtual ResultCode HandleSyncRequest(Kernel::HLERequestContext& context) = 0;
+
+    /**
+     * Signals that a client has just connected to this HLE handler and keeps the
+     * associated ServerSession alive for the duration of the connection.
+     * @param server_session Owning pointer to the ServerSession associated with the connection.
+     */
+    void ClientConnected(std::shared_ptr<ServerSession> server_session);
+
+    /**
+     * Signals that a client has just disconnected from this HLE handler and releases the
+     * associated ServerSession.
+     * @param server_session ServerSession associated with the connection.
+     */
+    void ClientDisconnected(const std::shared_ptr<ServerSession>& server_session);
+
+protected:
+    /// List of sessions that are connected to this handler.
+    /// A ServerSession whose server endpoint is an HLE implementation is kept alive by this list
+    /// for the duration of the connection.
+    std::vector<std::shared_ptr<ServerSession>> connected_sessions;
+};
+
+/**
+ * Class containing information about an in-flight IPC request being handled by an HLE service
+ * implementation. Services should avoid using old global APIs (e.g. Kernel::GetCommandBuffer()) and
+ * when possible use the APIs in this class to service the request.
+ *
+ * HLE handle protocol
+ * ===================
+ *
+ * To avoid needing HLE services to keep a separate handle table, or having to directly modify the
+ * requester's table, a tweaked protocol is used to receive and send handles in requests. The kernel
+ * will decode the incoming handles into object pointers and insert a id in the buffer where the
+ * handle would normally be. The service then calls GetIncomingHandle() with that id to get the
+ * pointer to the object. Similarly, instead of inserting a handle into the command buffer, the
+ * service calls AddOutgoingHandle() and stores the returned id where the handle would normally go.
+ *
+ * The end result is similar to just giving services their own real handle tables, but since these
+ * ids are local to a specific context, it avoids requiring services to manage handles for objects
+ * across multiple calls and ensuring that unneeded handles are cleaned up.
+ */
+class HLERequestContext {
+public:
+    explicit HLERequestContext(KernelCore& kernel, Core::Memory::Memory& memory,
+                               std::shared_ptr<ServerSession> session,
+                               std::shared_ptr<Thread> thread);
+    ~HLERequestContext();
+
+    /// Returns a pointer to the IPC command buffer for this request.
+    u32* CommandBuffer() {
+        return cmd_buf.data();
+    }
+
+    /**
+     * Returns the session through which this request was made. This can be used as a map key to
+     * access per-client data on services.
+     */
+    const std::shared_ptr<Kernel::ServerSession>& Session() const {
+        return server_session;
+    }
+
+    using WakeupCallback = std::function<void(
+        std::shared_ptr<Thread> thread, HLERequestContext& context, ThreadWakeupReason reason)>;
+
+    /// Populates this context with data from the requesting process/thread.
+    ResultCode PopulateFromIncomingCommandBuffer(const HandleTable& handle_table,
+                                                 u32_le* src_cmdbuf);
+
+    /// Writes data from this context back to the requesting process/thread.
+    ResultCode WriteToOutgoingCommandBuffer(Thread& thread);
+
+    u32_le GetCommand() const {
+        return command;
+    }
+
+    IPC::CommandType GetCommandType() const {
+        return command_header->type;
+    }
+
+    unsigned GetDataPayloadOffset() const {
+        return data_payload_offset;
+    }
+
+    const std::vector<IPC::BufferDescriptorX>& BufferDescriptorX() const {
+        return buffer_x_desciptors;
+    }
+
+    const std::vector<IPC::BufferDescriptorABW>& BufferDescriptorA() const {
+        return buffer_a_desciptors;
+    }
+
+    const std::vector<IPC::BufferDescriptorABW>& BufferDescriptorB() const {
+        return buffer_b_desciptors;
+    }
+
+    const std::vector<IPC::BufferDescriptorC>& BufferDescriptorC() const {
+        return buffer_c_desciptors;
+    }
+
+    const IPC::DomainMessageHeader& GetDomainMessageHeader() const {
+        return domain_message_header.value();
+    }
+
+    bool HasDomainMessageHeader() const {
+        return domain_message_header.has_value();
+    }
+
+    /// Helper function to read a buffer using the appropriate buffer descriptor
+    std::vector<u8> ReadBuffer(std::size_t buffer_index = 0) const;
+
+    /// Helper function to write a buffer using the appropriate buffer descriptor
+    std::size_t WriteBuffer(const void* buffer, std::size_t size,
+                            std::size_t buffer_index = 0) const;
+
+    /* Helper function to write a buffer using the appropriate buffer descriptor
+     *
+     * @tparam T an arbitrary container that satisfies the
+     *         ContiguousContainer concept in the C++ standard library or a trivially copyable type.
+     *
+     * @param data         The container/data to write into a buffer.
+     * @param buffer_index The buffer in particular to write to.
+     */
+    template <typename T, typename = std::enable_if_t<!std::is_pointer_v<T>>>
+    std::size_t WriteBuffer(const T& data, std::size_t buffer_index = 0) const {
+        if constexpr (Common::IsSTLContainer<T>) {
+            using ContiguousType = typename T::value_type;
+            static_assert(std::is_trivially_copyable_v<ContiguousType>,
+                          "Container to WriteBuffer must contain trivially copyable objects");
+            return WriteBuffer(std::data(data), std::size(data) * sizeof(ContiguousType),
+                               buffer_index);
+        } else {
+            static_assert(std::is_trivially_copyable_v<T>, "T must be trivially copyable");
+            return WriteBuffer(&data, sizeof(T), buffer_index);
+        }
+    }
+
+    /// Helper function to get the size of the input buffer
+    std::size_t GetReadBufferSize(std::size_t buffer_index = 0) const;
+
+    /// Helper function to get the size of the output buffer
+    std::size_t GetWriteBufferSize(std::size_t buffer_index = 0) const;
+
+    template <typename T>
+    std::shared_ptr<T> GetCopyObject(std::size_t index) {
+        return DynamicObjectCast<T>(copy_objects.at(index));
+    }
+
+    template <typename T>
+    std::shared_ptr<T> GetMoveObject(std::size_t index) {
+        return DynamicObjectCast<T>(move_objects.at(index));
+    }
+
+    void AddMoveObject(std::shared_ptr<Object> object) {
+        move_objects.emplace_back(std::move(object));
+    }
+
+    void AddCopyObject(std::shared_ptr<Object> object) {
+        copy_objects.emplace_back(std::move(object));
+    }
+
+    void AddDomainObject(std::shared_ptr<SessionRequestHandler> object) {
+        domain_objects.emplace_back(std::move(object));
+    }
+
+    template <typename T>
+    std::shared_ptr<T> GetDomainRequestHandler(std::size_t index) const {
+        return std::static_pointer_cast<T>(domain_request_handlers.at(index));
+    }
+
+    void SetDomainRequestHandlers(
+        const std::vector<std::shared_ptr<SessionRequestHandler>>& handlers) {
+        domain_request_handlers = handlers;
+    }
+
+    /// Clears the list of objects so that no lingering objects are written accidentally to the
+    /// response buffer.
+    void ClearIncomingObjects() {
+        move_objects.clear();
+        copy_objects.clear();
+        domain_objects.clear();
+    }
+
+    std::size_t NumMoveObjects() const {
+        return move_objects.size();
+    }
+
+    std::size_t NumCopyObjects() const {
+        return copy_objects.size();
+    }
+
+    std::size_t NumDomainObjects() const {
+        return domain_objects.size();
+    }
+
+    std::string Description() const;
+
+    Thread& GetThread() {
+        return *thread;
+    }
+
+    const Thread& GetThread() const {
+        return *thread;
+    }
+
+    bool IsThreadWaiting() const {
+        return is_thread_waiting;
+    }
+
+private:
+    friend class IPC::ResponseBuilder;
+
+    void ParseCommandBuffer(const HandleTable& handle_table, u32_le* src_cmdbuf, bool incoming);
+
+    std::array<u32, IPC::COMMAND_BUFFER_LENGTH> cmd_buf;
+    std::shared_ptr<Kernel::ServerSession> server_session;
+    std::shared_ptr<Thread> thread;
+    // TODO(yuriks): Check common usage of this and optimize size accordingly
+    boost::container::small_vector<std::shared_ptr<Object>, 8> move_objects;
+    boost::container::small_vector<std::shared_ptr<Object>, 8> copy_objects;
+    boost::container::small_vector<std::shared_ptr<SessionRequestHandler>, 8> domain_objects;
+
+    std::optional<IPC::CommandHeader> command_header;
+    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;
+
+    unsigned data_payload_offset{};
+    unsigned buffer_c_offset{};
+    u32_le command{};
+
+    std::vector<std::shared_ptr<SessionRequestHandler>> domain_request_handlers;
+    bool is_thread_waiting{};
+
+    KernelCore& kernel;
+    Core::Memory::Memory& memory;
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_affinity_mask.h

@@ -0,0 +1,58 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "core/hardware_properties.h"
+
+namespace Kernel {
+
+class KAffinityMask {
+public:
+    constexpr KAffinityMask() = default;
+
+    [[nodiscard]] constexpr u64 GetAffinityMask() const {
+        return this->mask;
+    }
+
+    constexpr void SetAffinityMask(u64 new_mask) {
+        ASSERT((new_mask & ~AllowedAffinityMask) == 0);
+        this->mask = new_mask;
+    }
+
+    [[nodiscard]] constexpr bool GetAffinity(s32 core) const {
+        return this->mask & GetCoreBit(core);
+    }
+
+    constexpr void SetAffinity(s32 core, bool set) {
+        ASSERT(0 <= core && core < static_cast<s32>(Core::Hardware::NUM_CPU_CORES));
+
+        if (set) {
+            this->mask |= GetCoreBit(core);
+        } else {
+            this->mask &= ~GetCoreBit(core);
+        }
+    }
+
+    constexpr void SetAll() {
+        this->mask = AllowedAffinityMask;
+    }
+
+private:
+    [[nodiscard]] static constexpr u64 GetCoreBit(s32 core) {
+        ASSERT(0 <= core && core < static_cast<s32>(Core::Hardware::NUM_CPU_CORES));
+        return (1ULL << core);
+    }
+
+    static constexpr u64 AllowedAffinityMask = (1ULL << Core::Hardware::NUM_CPU_CORES) - 1;
+
+    u64 mask{};
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_priority_queue.h

@@ -0,0 +1,449 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include <array>
+
+#include "common/assert.h"
+#include "common/bit_set.h"
+#include "common/bit_util.h"
+#include "common/common_types.h"
+
+namespace Kernel {
+
+class Thread;
+
+template <typename T>
+concept KPriorityQueueAffinityMask = !std::is_reference_v<T> && requires(T & t) {
+    { t.GetAffinityMask() }
+    ->std::convertible_to<u64>;
+    {t.SetAffinityMask(std::declval<u64>())};
+
+    { t.GetAffinity(std::declval<int32_t>()) }
+    ->std::same_as<bool>;
+    {t.SetAffinity(std::declval<int32_t>(), std::declval<bool>())};
+    {t.SetAll()};
+};
+
+template <typename T>
+concept KPriorityQueueMember = !std::is_reference_v<T> && requires(T & t) {
+    {typename T::QueueEntry()};
+    {(typename T::QueueEntry()).Initialize()};
+    {(typename T::QueueEntry()).SetPrev(std::addressof(t))};
+    {(typename T::QueueEntry()).SetNext(std::addressof(t))};
+    { (typename T::QueueEntry()).GetNext() }
+    ->std::same_as<T*>;
+    { (typename T::QueueEntry()).GetPrev() }
+    ->std::same_as<T*>;
+    { t.GetPriorityQueueEntry(std::declval<s32>()) }
+    ->std::same_as<typename T::QueueEntry&>;
+
+    {t.GetAffinityMask()};
+    { typename std::remove_cvref<decltype(t.GetAffinityMask())>::type() }
+    ->KPriorityQueueAffinityMask;
+
+    { t.GetActiveCore() }
+    ->std::convertible_to<s32>;
+    { t.GetPriority() }
+    ->std::convertible_to<s32>;
+};
+
+template <typename Member, size_t _NumCores, int LowestPriority, int HighestPriority>
+requires KPriorityQueueMember<Member> class KPriorityQueue {
+public:
+    using AffinityMaskType = typename std::remove_cv_t<
+        typename std::remove_reference<decltype(std::declval<Member>().GetAffinityMask())>::type>;
+
+    static_assert(LowestPriority >= 0);
+    static_assert(HighestPriority >= 0);
+    static_assert(LowestPriority >= HighestPriority);
+    static constexpr size_t NumPriority = LowestPriority - HighestPriority + 1;
+    static constexpr size_t NumCores = _NumCores;
+
+    static constexpr bool IsValidCore(s32 core) {
+        return 0 <= core && core < static_cast<s32>(NumCores);
+    }
+
+    static constexpr bool IsValidPriority(s32 priority) {
+        return HighestPriority <= priority && priority <= LowestPriority + 1;
+    }
+
+private:
+    using Entry = typename Member::QueueEntry;
+
+public:
+    class KPerCoreQueue {
+    private:
+        std::array<Entry, NumCores> root{};
+
+    public:
+        constexpr KPerCoreQueue() {
+            for (auto& per_core_root : root) {
+                per_core_root.Initialize();
+            }
+        }
+
+        constexpr bool PushBack(s32 core, Member* member) {
+            // Get the entry associated with the member.
+            Entry& member_entry = member->GetPriorityQueueEntry(core);
+
+            // Get the entry associated with the end of the queue.
+            Member* tail = this->root[core].GetPrev();
+            Entry& tail_entry =
+                (tail != nullptr) ? tail->GetPriorityQueueEntry(core) : this->root[core];
+
+            // Link the entries.
+            member_entry.SetPrev(tail);
+            member_entry.SetNext(nullptr);
+            tail_entry.SetNext(member);
+            this->root[core].SetPrev(member);
+
+            return tail == nullptr;
+        }
+
+        constexpr bool PushFront(s32 core, Member* member) {
+            // Get the entry associated with the member.
+            Entry& member_entry = member->GetPriorityQueueEntry(core);
+
+            // Get the entry associated with the front of the queue.
+            Member* head = this->root[core].GetNext();
+            Entry& head_entry =
+                (head != nullptr) ? head->GetPriorityQueueEntry(core) : this->root[core];
+
+            // Link the entries.
+            member_entry.SetPrev(nullptr);
+            member_entry.SetNext(head);
+            head_entry.SetPrev(member);
+            this->root[core].SetNext(member);
+
+            return (head == nullptr);
+        }
+
+        constexpr bool Remove(s32 core, Member* member) {
+            // Get the entry associated with the member.
+            Entry& member_entry = member->GetPriorityQueueEntry(core);
+
+            // Get the entries associated with next and prev.
+            Member* prev = member_entry.GetPrev();
+            Member* next = member_entry.GetNext();
+            Entry& prev_entry =
+                (prev != nullptr) ? prev->GetPriorityQueueEntry(core) : this->root[core];
+            Entry& next_entry =
+                (next != nullptr) ? next->GetPriorityQueueEntry(core) : this->root[core];
+
+            // Unlink.
+            prev_entry.SetNext(next);
+            next_entry.SetPrev(prev);
+
+            return (this->GetFront(core) == nullptr);
+        }
+
+        constexpr Member* GetFront(s32 core) const {
+            return this->root[core].GetNext();
+        }
+    };
+
+    class KPriorityQueueImpl {
+    public:
+        constexpr KPriorityQueueImpl() = default;
+
+        constexpr void PushBack(s32 priority, s32 core, Member* member) {
+            ASSERT(IsValidCore(core));
+            ASSERT(IsValidPriority(priority));
+
+            if (priority > LowestPriority) {
+                return;
+            }
+
+            if (this->queues[priority].PushBack(core, member)) {
+                this->available_priorities[core].SetBit(priority);
+            }
+        }
+
+        constexpr void PushFront(s32 priority, s32 core, Member* member) {
+            ASSERT(IsValidCore(core));
+            ASSERT(IsValidPriority(priority));
+
+            if (priority > LowestPriority) {
+                return;
+            }
+
+            if (this->queues[priority].PushFront(core, member)) {
+                this->available_priorities[core].SetBit(priority);
+            }
+        }
+
+        constexpr void Remove(s32 priority, s32 core, Member* member) {
+            ASSERT(IsValidCore(core));
+            ASSERT(IsValidPriority(priority));
+
+            if (priority > LowestPriority) {
+                return;
+            }
+
+            if (this->queues[priority].Remove(core, member)) {
+                this->available_priorities[core].ClearBit(priority);
+            }
+        }
+
+        constexpr Member* GetFront(s32 core) const {
+            ASSERT(IsValidCore(core));
+
+            const s32 priority =
+                static_cast<s32>(this->available_priorities[core].CountLeadingZero());
+            if (priority <= LowestPriority) {
+                return this->queues[priority].GetFront(core);
+            } else {
+                return nullptr;
+            }
+        }
+
+        constexpr Member* GetFront(s32 priority, s32 core) const {
+            ASSERT(IsValidCore(core));
+            ASSERT(IsValidPriority(priority));
+
+            if (priority <= LowestPriority) {
+                return this->queues[priority].GetFront(core);
+            } else {
+                return nullptr;
+            }
+        }
+
+        constexpr Member* GetNext(s32 core, const Member* member) const {
+            ASSERT(IsValidCore(core));
+
+            Member* next = member->GetPriorityQueueEntry(core).GetNext();
+            if (next == nullptr) {
+                const s32 priority = static_cast<s32>(
+                    this->available_priorities[core].GetNextSet(member->GetPriority()));
+                if (priority <= LowestPriority) {
+                    next = this->queues[priority].GetFront(core);
+                }
+            }
+            return next;
+        }
+
+        constexpr void MoveToFront(s32 priority, s32 core, Member* member) {
+            ASSERT(IsValidCore(core));
+            ASSERT(IsValidPriority(priority));
+
+            if (priority <= LowestPriority) {
+                this->queues[priority].Remove(core, member);
+                this->queues[priority].PushFront(core, member);
+            }
+        }
+
+        constexpr Member* MoveToBack(s32 priority, s32 core, Member* member) {
+            ASSERT(IsValidCore(core));
+            ASSERT(IsValidPriority(priority));
+
+            if (priority <= LowestPriority) {
+                this->queues[priority].Remove(core, member);
+                this->queues[priority].PushBack(core, member);
+                return this->queues[priority].GetFront(core);
+            } else {
+                return nullptr;
+            }
+        }
+
+    private:
+        std::array<KPerCoreQueue, NumPriority> queues{};
+        std::array<Common::BitSet64<NumPriority>, NumCores> available_priorities{};
+    };
+
+private:
+    KPriorityQueueImpl scheduled_queue;
+    KPriorityQueueImpl suggested_queue;
+
+private:
+    constexpr void ClearAffinityBit(u64& affinity, s32 core) {
+        affinity &= ~(u64(1) << core);
+    }
+
+    constexpr s32 GetNextCore(u64& affinity) {
+        const s32 core = Common::CountTrailingZeroes64(affinity);
+        ClearAffinityBit(affinity, core);
+        return core;
+    }
+
+    constexpr void PushBack(s32 priority, Member* member) {
+        ASSERT(IsValidPriority(priority));
+
+        // Push onto the scheduled queue for its core, if we can.
+        u64 affinity = member->GetAffinityMask().GetAffinityMask();
+        if (const s32 core = member->GetActiveCore(); core >= 0) {
+            this->scheduled_queue.PushBack(priority, core, member);
+            ClearAffinityBit(affinity, core);
+        }
+
+        // And suggest the thread for all other cores.
+        while (affinity) {
+            this->suggested_queue.PushBack(priority, GetNextCore(affinity), member);
+        }
+    }
+
+    constexpr void PushFront(s32 priority, Member* member) {
+        ASSERT(IsValidPriority(priority));
+
+        // Push onto the scheduled queue for its core, if we can.
+        u64 affinity = member->GetAffinityMask().GetAffinityMask();
+        if (const s32 core = member->GetActiveCore(); core >= 0) {
+            this->scheduled_queue.PushFront(priority, core, member);
+            ClearAffinityBit(affinity, core);
+        }
+
+        // And suggest the thread for all other cores.
+        // Note: Nintendo pushes onto the back of the suggested queue, not the front.
+        while (affinity) {
+            this->suggested_queue.PushBack(priority, GetNextCore(affinity), member);
+        }
+    }
+
+    constexpr void Remove(s32 priority, Member* member) {
+        ASSERT(IsValidPriority(priority));
+
+        // Remove from the scheduled queue for its core.
+        u64 affinity = member->GetAffinityMask().GetAffinityMask();
+        if (const s32 core = member->GetActiveCore(); core >= 0) {
+            this->scheduled_queue.Remove(priority, core, member);
+            ClearAffinityBit(affinity, core);
+        }
+
+        // Remove from the suggested queue for all other cores.
+        while (affinity) {
+            this->suggested_queue.Remove(priority, GetNextCore(affinity), member);
+        }
+    }
+
+public:
+    constexpr KPriorityQueue() = default;
+
+    // Getters.
+    constexpr Member* GetScheduledFront(s32 core) const {
+        return this->scheduled_queue.GetFront(core);
+    }
+
+    constexpr Member* GetScheduledFront(s32 core, s32 priority) const {
+        return this->scheduled_queue.GetFront(priority, core);
+    }
+
+    constexpr Member* GetSuggestedFront(s32 core) const {
+        return this->suggested_queue.GetFront(core);
+    }
+
+    constexpr Member* GetSuggestedFront(s32 core, s32 priority) const {
+        return this->suggested_queue.GetFront(priority, core);
+    }
+
+    constexpr Member* GetScheduledNext(s32 core, const Member* member) const {
+        return this->scheduled_queue.GetNext(core, member);
+    }
+
+    constexpr Member* GetSuggestedNext(s32 core, const Member* member) const {
+        return this->suggested_queue.GetNext(core, member);
+    }
+
+    constexpr Member* GetSamePriorityNext(s32 core, const Member* member) const {
+        return member->GetPriorityQueueEntry(core).GetNext();
+    }
+
+    // Mutators.
+    constexpr void PushBack(Member* member) {
+        this->PushBack(member->GetPriority(), member);
+    }
+
+    constexpr void Remove(Member* member) {
+        this->Remove(member->GetPriority(), member);
+    }
+
+    constexpr void MoveToScheduledFront(Member* member) {
+        this->scheduled_queue.MoveToFront(member->GetPriority(), member->GetActiveCore(), member);
+    }
+
+    constexpr Thread* MoveToScheduledBack(Member* member) {
+        return this->scheduled_queue.MoveToBack(member->GetPriority(), member->GetActiveCore(),
+                                                member);
+    }
+
+    // First class fancy operations.
+    constexpr void ChangePriority(s32 prev_priority, bool is_running, Member* member) {
+        ASSERT(IsValidPriority(prev_priority));
+
+        // Remove the member from the queues.
+        const s32 new_priority = member->GetPriority();
+        this->Remove(prev_priority, member);
+
+        // And enqueue. If the member is running, we want to keep it running.
+        if (is_running) {
+            this->PushFront(new_priority, member);
+        } else {
+            this->PushBack(new_priority, member);
+        }
+    }
+
+    constexpr void ChangeAffinityMask(s32 prev_core, const AffinityMaskType& prev_affinity,
+                                      Member* member) {
+        // Get the new information.
+        const s32 priority = member->GetPriority();
+        const AffinityMaskType& new_affinity = member->GetAffinityMask();
+        const s32 new_core = member->GetActiveCore();
+
+        // Remove the member from all queues it was in before.
+        for (s32 core = 0; core < static_cast<s32>(NumCores); core++) {
+            if (prev_affinity.GetAffinity(core)) {
+                if (core == prev_core) {
+                    this->scheduled_queue.Remove(priority, core, member);
+                } else {
+                    this->suggested_queue.Remove(priority, core, member);
+                }
+            }
+        }
+
+        // And add the member to all queues it should be in now.
+        for (s32 core = 0; core < static_cast<s32>(NumCores); core++) {
+            if (new_affinity.GetAffinity(core)) {
+                if (core == new_core) {
+                    this->scheduled_queue.PushBack(priority, core, member);
+                } else {
+                    this->suggested_queue.PushBack(priority, core, member);
+                }
+            }
+        }
+    }
+
+    constexpr void ChangeCore(s32 prev_core, Member* member, bool to_front = false) {
+        // Get the new information.
+        const s32 new_core = member->GetActiveCore();
+        const s32 priority = member->GetPriority();
+
+        // We don't need to do anything if the core is the same.
+        if (prev_core != new_core) {
+            // Remove from the scheduled queue for the previous core.
+            if (prev_core >= 0) {
+                this->scheduled_queue.Remove(priority, prev_core, member);
+            }
+
+            // Remove from the suggested queue and add to the scheduled queue for the new core.
+            if (new_core >= 0) {
+                this->suggested_queue.Remove(priority, new_core, member);
+                if (to_front) {
+                    this->scheduled_queue.PushFront(priority, new_core, member);
+                } else {
+                    this->scheduled_queue.PushBack(priority, new_core, member);
+                }
+            }
+
+            // Add to the suggested queue for the previous core.
+            if (prev_core >= 0) {
+                this->suggested_queue.PushBack(priority, prev_core, member);
+            }
+        }
+    }
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_scheduler.cpp

@@ -0,0 +1,784 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#include "common/assert.h"
+#include "common/bit_util.h"
+#include "common/fiber.h"
+#include "common/logging/log.h"
+#include "core/arm/arm_interface.h"
+#include "core/core.h"
+#include "core/core_timing.h"
+#include "core/cpu_manager.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/physical_core.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/time_manager.h"
+
+namespace Kernel {
+
+static void IncrementScheduledCount(Kernel::Thread* thread) {
+    if (auto process = thread->GetOwnerProcess(); process) {
+        process->IncrementScheduledCount();
+    }
+}
+
+void KScheduler::RescheduleCores(KernelCore& kernel, u64 cores_pending_reschedule,
+                                 Core::EmuThreadHandle global_thread) {
+    u32 current_core = global_thread.host_handle;
+    bool must_context_switch = global_thread.guest_handle != InvalidHandle &&
+                               (current_core < Core::Hardware::NUM_CPU_CORES);
+
+    while (cores_pending_reschedule != 0) {
+        u32 core = Common::CountTrailingZeroes64(cores_pending_reschedule);
+        ASSERT(core < Core::Hardware::NUM_CPU_CORES);
+        if (!must_context_switch || core != current_core) {
+            auto& phys_core = kernel.PhysicalCore(core);
+            phys_core.Interrupt();
+        } else {
+            must_context_switch = true;
+        }
+        cores_pending_reschedule &= ~(1ULL << core);
+    }
+    if (must_context_switch) {
+        auto core_scheduler = kernel.CurrentScheduler();
+        kernel.ExitSVCProfile();
+        core_scheduler->RescheduleCurrentCore();
+        kernel.EnterSVCProfile();
+    }
+}
+
+u64 KScheduler::UpdateHighestPriorityThread(Thread* highest_thread) {
+    std::scoped_lock lock{guard};
+    if (Thread* prev_highest_thread = this->state.highest_priority_thread;
+        prev_highest_thread != highest_thread) {
+        if (prev_highest_thread != nullptr) {
+            IncrementScheduledCount(prev_highest_thread);
+            prev_highest_thread->SetLastScheduledTick(system.CoreTiming().GetCPUTicks());
+        }
+        if (this->state.should_count_idle) {
+            if (highest_thread != nullptr) {
+                // if (Process* process = highest_thread->GetOwnerProcess(); process != nullptr) {
+                //    process->SetRunningThread(this->core_id, highest_thread,
+                //                              this->state.idle_count);
+                //}
+            } else {
+                this->state.idle_count++;
+            }
+        }
+
+        this->state.highest_priority_thread = highest_thread;
+        this->state.needs_scheduling = true;
+        return (1ULL << this->core_id);
+    } else {
+        return 0;
+    }
+}
+
+u64 KScheduler::UpdateHighestPriorityThreadsImpl(KernelCore& kernel) {
+    ASSERT(kernel.GlobalSchedulerContext().IsLocked());
+
+    // Clear that we need to update.
+    ClearSchedulerUpdateNeeded(kernel);
+
+    u64 cores_needing_scheduling = 0, idle_cores = 0;
+    Thread* top_threads[Core::Hardware::NUM_CPU_CORES];
+    auto& priority_queue = GetPriorityQueue(kernel);
+
+    /// We want to go over all cores, finding the highest priority thread and determining if
+    /// scheduling is needed for that core.
+    for (size_t core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
+        Thread* top_thread = priority_queue.GetScheduledFront(static_cast<s32>(core_id));
+        if (top_thread != nullptr) {
+            // If the thread has no waiters, we need to check if the process has a thread pinned.
+            // TODO(bunnei): Implement thread pinning
+        } else {
+            idle_cores |= (1ULL << core_id);
+        }
+
+        top_threads[core_id] = top_thread;
+        cores_needing_scheduling |=
+            kernel.Scheduler(core_id).UpdateHighestPriorityThread(top_threads[core_id]);
+    }
+
+    // Idle cores are bad. We're going to try to migrate threads to each idle core in turn.
+    while (idle_cores != 0) {
+        u32 core_id = Common::CountTrailingZeroes64(idle_cores);
+        if (Thread* suggested = priority_queue.GetSuggestedFront(core_id); suggested != nullptr) {
+            s32 migration_candidates[Core::Hardware::NUM_CPU_CORES];
+            size_t num_candidates = 0;
+
+            // While we have a suggested thread, try to migrate it!
+            while (suggested != nullptr) {
+                // Check if the suggested thread is the top thread on its core.
+                const s32 suggested_core = suggested->GetActiveCore();
+                if (Thread* top_thread =
+                        (suggested_core >= 0) ? top_threads[suggested_core] : nullptr;
+                    top_thread != suggested) {
+                    // Make sure we're not dealing with threads too high priority for migration.
+                    if (top_thread != nullptr &&
+                        top_thread->GetPriority() < HighestCoreMigrationAllowedPriority) {
+                        break;
+                    }
+
+                    // The suggested thread isn't bound to its core, so we can migrate it!
+                    suggested->SetActiveCore(core_id);
+                    priority_queue.ChangeCore(suggested_core, suggested);
+
+                    top_threads[core_id] = suggested;
+                    cores_needing_scheduling |=
+                        kernel.Scheduler(core_id).UpdateHighestPriorityThread(top_threads[core_id]);
+                    break;
+                }
+
+                // Note this core as a candidate for migration.
+                ASSERT(num_candidates < Core::Hardware::NUM_CPU_CORES);
+                migration_candidates[num_candidates++] = suggested_core;
+                suggested = priority_queue.GetSuggestedNext(core_id, suggested);
+            }
+
+            // If suggested is nullptr, we failed to migrate a specific thread. So let's try all our
+            // candidate cores' top threads.
+            if (suggested == nullptr) {
+                for (size_t i = 0; i < num_candidates; i++) {
+                    // Check if there's some other thread that can run on the candidate core.
+                    const s32 candidate_core = migration_candidates[i];
+                    suggested = top_threads[candidate_core];
+                    if (Thread* next_on_candidate_core =
+                            priority_queue.GetScheduledNext(candidate_core, suggested);
+                        next_on_candidate_core != nullptr) {
+                        // The candidate core can run some other thread! We'll migrate its current
+                        // top thread to us.
+                        top_threads[candidate_core] = next_on_candidate_core;
+                        cores_needing_scheduling |=
+                            kernel.Scheduler(candidate_core)
+                                .UpdateHighestPriorityThread(top_threads[candidate_core]);
+
+                        // Perform the migration.
+                        suggested->SetActiveCore(core_id);
+                        priority_queue.ChangeCore(candidate_core, suggested);
+
+                        top_threads[core_id] = suggested;
+                        cores_needing_scheduling |=
+                            kernel.Scheduler(core_id).UpdateHighestPriorityThread(
+                                top_threads[core_id]);
+                        break;
+                    }
+                }
+            }
+        }
+
+        idle_cores &= ~(1ULL << core_id);
+    }
+
+    return cores_needing_scheduling;
+}
+
+void KScheduler::OnThreadStateChanged(KernelCore& kernel, Thread* thread, u32 old_state) {
+    ASSERT(kernel.GlobalSchedulerContext().IsLocked());
+
+    // Check if the state has changed, because if it hasn't there's nothing to do.
+    const auto cur_state = thread->scheduling_state;
+    if (cur_state == old_state) {
+        return;
+    }
+
+    // Update the priority queues.
+    if (old_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+        // If we were previously runnable, then we're not runnable now, and we should remove.
+        GetPriorityQueue(kernel).Remove(thread);
+        IncrementScheduledCount(thread);
+        SetSchedulerUpdateNeeded(kernel);
+    } else if (cur_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+        // If we're now runnable, then we weren't previously, and we should add.
+        GetPriorityQueue(kernel).PushBack(thread);
+        IncrementScheduledCount(thread);
+        SetSchedulerUpdateNeeded(kernel);
+    }
+}
+
+void KScheduler::OnThreadPriorityChanged(KernelCore& kernel, Thread* thread, Thread* current_thread,
+                                         u32 old_priority) {
+
+    ASSERT(kernel.GlobalSchedulerContext().IsLocked());
+
+    // If the thread is runnable, we want to change its priority in the queue.
+    if (thread->scheduling_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+        GetPriorityQueue(kernel).ChangePriority(
+            old_priority, thread == kernel.CurrentScheduler()->GetCurrentThread(), thread);
+        IncrementScheduledCount(thread);
+        SetSchedulerUpdateNeeded(kernel);
+    }
+}
+
+void KScheduler::OnThreadAffinityMaskChanged(KernelCore& kernel, Thread* thread,
+                                             const KAffinityMask& old_affinity, s32 old_core) {
+    ASSERT(kernel.GlobalSchedulerContext().IsLocked());
+
+    // If the thread is runnable, we want to change its affinity in the queue.
+    if (thread->scheduling_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+        GetPriorityQueue(kernel).ChangeAffinityMask(old_core, old_affinity, thread);
+        IncrementScheduledCount(thread);
+        SetSchedulerUpdateNeeded(kernel);
+    }
+}
+
+void KScheduler::RotateScheduledQueue(s32 core_id, s32 priority) {
+    ASSERT(system.GlobalSchedulerContext().IsLocked());
+
+    // Get a reference to the priority queue.
+    auto& kernel = system.Kernel();
+    auto& priority_queue = GetPriorityQueue(kernel);
+
+    // Rotate the front of the queue to the end.
+    Thread* top_thread = priority_queue.GetScheduledFront(core_id, priority);
+    Thread* next_thread = nullptr;
+    if (top_thread != nullptr) {
+        next_thread = priority_queue.MoveToScheduledBack(top_thread);
+        if (next_thread != top_thread) {
+            IncrementScheduledCount(top_thread);
+            IncrementScheduledCount(next_thread);
+        }
+    }
+
+    // While we have a suggested thread, try to migrate it!
+    {
+        Thread* suggested = priority_queue.GetSuggestedFront(core_id, priority);
+        while (suggested != nullptr) {
+            // Check if the suggested thread is the top thread on its core.
+            const s32 suggested_core = suggested->GetActiveCore();
+            if (Thread* top_on_suggested_core =
+                    (suggested_core >= 0) ? priority_queue.GetScheduledFront(suggested_core)
+                                          : nullptr;
+                top_on_suggested_core != suggested) {
+                // If the next thread is a new thread that has been waiting longer than our
+                // suggestion, we prefer it to our suggestion.
+                if (top_thread != next_thread && next_thread != nullptr &&
+                    next_thread->GetLastScheduledTick() < suggested->GetLastScheduledTick()) {
+                    suggested = nullptr;
+                    break;
+                }
+
+                // If we're allowed to do a migration, do one.
+                // NOTE: Unlike migrations in UpdateHighestPriorityThread, this moves the suggestion
+                // to the front of the queue.
+                if (top_on_suggested_core == nullptr ||
+                    top_on_suggested_core->GetPriority() >= HighestCoreMigrationAllowedPriority) {
+                    suggested->SetActiveCore(core_id);
+                    priority_queue.ChangeCore(suggested_core, suggested, true);
+                    IncrementScheduledCount(suggested);
+                    break;
+                }
+            }
+
+            // Get the next suggestion.
+            suggested = priority_queue.GetSamePriorityNext(core_id, suggested);
+        }
+    }
+
+    // Now that we might have migrated a thread with the same priority, check if we can do better.
+
+    {
+        Thread* best_thread = priority_queue.GetScheduledFront(core_id);
+        if (best_thread == GetCurrentThread()) {
+            best_thread = priority_queue.GetScheduledNext(core_id, best_thread);
+        }
+
+        // If the best thread we can choose has a priority the same or worse than ours, try to
+        // migrate a higher priority thread.
+        if (best_thread != nullptr && best_thread->GetPriority() >= static_cast<u32>(priority)) {
+            Thread* suggested = priority_queue.GetSuggestedFront(core_id);
+            while (suggested != nullptr) {
+                // If the suggestion's priority is the same as ours, don't bother.
+                if (suggested->GetPriority() >= best_thread->GetPriority()) {
+                    break;
+                }
+
+                // Check if the suggested thread is the top thread on its core.
+                const s32 suggested_core = suggested->GetActiveCore();
+                if (Thread* top_on_suggested_core =
+                        (suggested_core >= 0) ? priority_queue.GetScheduledFront(suggested_core)
+                                              : nullptr;
+                    top_on_suggested_core != suggested) {
+                    // If we're allowed to do a migration, do one.
+                    // NOTE: Unlike migrations in UpdateHighestPriorityThread, this moves the
+                    // suggestion to the front of the queue.
+                    if (top_on_suggested_core == nullptr ||
+                        top_on_suggested_core->GetPriority() >=
+                            HighestCoreMigrationAllowedPriority) {
+                        suggested->SetActiveCore(core_id);
+                        priority_queue.ChangeCore(suggested_core, suggested, true);
+                        IncrementScheduledCount(suggested);
+                        break;
+                    }
+                }
+
+                // Get the next suggestion.
+                suggested = priority_queue.GetSuggestedNext(core_id, suggested);
+            }
+        }
+    }
+
+    // After a rotation, we need a scheduler update.
+    SetSchedulerUpdateNeeded(kernel);
+}
+
+bool KScheduler::CanSchedule(KernelCore& kernel) {
+    return kernel.CurrentScheduler()->GetCurrentThread()->GetDisableDispatchCount() <= 1;
+}
+
+bool KScheduler::IsSchedulerUpdateNeeded(const KernelCore& kernel) {
+    return kernel.GlobalSchedulerContext().scheduler_update_needed.load(std::memory_order_acquire);
+}
+
+void KScheduler::SetSchedulerUpdateNeeded(KernelCore& kernel) {
+    kernel.GlobalSchedulerContext().scheduler_update_needed.store(true, std::memory_order_release);
+}
+
+void KScheduler::ClearSchedulerUpdateNeeded(KernelCore& kernel) {
+    kernel.GlobalSchedulerContext().scheduler_update_needed.store(false, std::memory_order_release);
+}
+
+void KScheduler::DisableScheduling(KernelCore& kernel) {
+    if (auto* scheduler = kernel.CurrentScheduler(); scheduler) {
+        ASSERT(scheduler->GetCurrentThread()->GetDisableDispatchCount() >= 0);
+        scheduler->GetCurrentThread()->DisableDispatch();
+    }
+}
+
+void KScheduler::EnableScheduling(KernelCore& kernel, u64 cores_needing_scheduling,
+                                  Core::EmuThreadHandle global_thread) {
+    if (auto* scheduler = kernel.CurrentScheduler(); scheduler) {
+        scheduler->GetCurrentThread()->EnableDispatch();
+    }
+    RescheduleCores(kernel, cores_needing_scheduling, global_thread);
+}
+
+u64 KScheduler::UpdateHighestPriorityThreads(KernelCore& kernel) {
+    if (IsSchedulerUpdateNeeded(kernel)) {
+        return UpdateHighestPriorityThreadsImpl(kernel);
+    } else {
+        return 0;
+    }
+}
+
+KSchedulerPriorityQueue& KScheduler::GetPriorityQueue(KernelCore& kernel) {
+    return kernel.GlobalSchedulerContext().priority_queue;
+}
+
+void KScheduler::YieldWithoutCoreMigration() {
+    auto& kernel = system.Kernel();
+
+    // Validate preconditions.
+    ASSERT(CanSchedule(kernel));
+    ASSERT(kernel.CurrentProcess() != nullptr);
+
+    // Get the current thread and process.
+    Thread& cur_thread = *GetCurrentThread();
+    Process& cur_process = *kernel.CurrentProcess();
+
+    // If the thread's yield count matches, there's nothing for us to do.
+    if (cur_thread.GetYieldScheduleCount() == cur_process.GetScheduledCount()) {
+        return;
+    }
+
+    // Get a reference to the priority queue.
+    auto& priority_queue = GetPriorityQueue(kernel);
+
+    // Perform the yield.
+    {
+        KScopedSchedulerLock lock(kernel);
+
+        const auto cur_state = cur_thread.scheduling_state;
+        if (cur_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+            // Put the current thread at the back of the queue.
+            Thread* next_thread = priority_queue.MoveToScheduledBack(std::addressof(cur_thread));
+            IncrementScheduledCount(std::addressof(cur_thread));
+
+            // If the next thread is different, we have an update to perform.
+            if (next_thread != std::addressof(cur_thread)) {
+                SetSchedulerUpdateNeeded(kernel);
+            } else {
+                // Otherwise, set the thread's yield count so that we won't waste work until the
+                // process is scheduled again.
+                cur_thread.SetYieldScheduleCount(cur_process.GetScheduledCount());
+            }
+        }
+    }
+}
+
+void KScheduler::YieldWithCoreMigration() {
+    auto& kernel = system.Kernel();
+
+    // Validate preconditions.
+    ASSERT(CanSchedule(kernel));
+    ASSERT(kernel.CurrentProcess() != nullptr);
+
+    // Get the current thread and process.
+    Thread& cur_thread = *GetCurrentThread();
+    Process& cur_process = *kernel.CurrentProcess();
+
+    // If the thread's yield count matches, there's nothing for us to do.
+    if (cur_thread.GetYieldScheduleCount() == cur_process.GetScheduledCount()) {
+        return;
+    }
+
+    // Get a reference to the priority queue.
+    auto& priority_queue = GetPriorityQueue(kernel);
+
+    // Perform the yield.
+    {
+        KScopedSchedulerLock lock(kernel);
+
+        const auto cur_state = cur_thread.scheduling_state;
+        if (cur_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+            // Get the current active core.
+            const s32 core_id = cur_thread.GetActiveCore();
+
+            // Put the current thread at the back of the queue.
+            Thread* next_thread = priority_queue.MoveToScheduledBack(std::addressof(cur_thread));
+            IncrementScheduledCount(std::addressof(cur_thread));
+
+            // While we have a suggested thread, try to migrate it!
+            bool recheck = false;
+            Thread* suggested = priority_queue.GetSuggestedFront(core_id);
+            while (suggested != nullptr) {
+                // Check if the suggested thread is the thread running on its core.
+                const s32 suggested_core = suggested->GetActiveCore();
+
+                if (Thread* running_on_suggested_core =
+                        (suggested_core >= 0)
+                            ? kernel.Scheduler(suggested_core).state.highest_priority_thread
+                            : nullptr;
+                    running_on_suggested_core != suggested) {
+                    // If the current thread's priority is higher than our suggestion's we prefer
+                    // the next thread to the suggestion. We also prefer the next thread when the
+                    // current thread's priority is equal to the suggestions, but the next thread
+                    // has been waiting longer.
+                    if ((suggested->GetPriority() > cur_thread.GetPriority()) ||
+                        (suggested->GetPriority() == cur_thread.GetPriority() &&
+                         next_thread != std::addressof(cur_thread) &&
+                         next_thread->GetLastScheduledTick() < suggested->GetLastScheduledTick())) {
+                        suggested = nullptr;
+                        break;
+                    }
+
+                    // If we're allowed to do a migration, do one.
+                    // NOTE: Unlike migrations in UpdateHighestPriorityThread, this moves the
+                    // suggestion to the front of the queue.
+                    if (running_on_suggested_core == nullptr ||
+                        running_on_suggested_core->GetPriority() >=
+                            HighestCoreMigrationAllowedPriority) {
+                        suggested->SetActiveCore(core_id);
+                        priority_queue.ChangeCore(suggested_core, suggested, true);
+                        IncrementScheduledCount(suggested);
+                        break;
+                    } else {
+                        // We couldn't perform a migration, but we should check again on a future
+                        // yield.
+                        recheck = true;
+                    }
+                }
+
+                // Get the next suggestion.
+                suggested = priority_queue.GetSuggestedNext(core_id, suggested);
+            }
+
+            // If we still have a suggestion or the next thread is different, we have an update to
+            // perform.
+            if (suggested != nullptr || next_thread != std::addressof(cur_thread)) {
+                SetSchedulerUpdateNeeded(kernel);
+            } else if (!recheck) {
+                // Otherwise if we don't need to re-check, set the thread's yield count so that we
+                // won't waste work until the process is scheduled again.
+                cur_thread.SetYieldScheduleCount(cur_process.GetScheduledCount());
+            }
+        }
+    }
+}
+
+void KScheduler::YieldToAnyThread() {
+    auto& kernel = system.Kernel();
+
+    // Validate preconditions.
+    ASSERT(CanSchedule(kernel));
+    ASSERT(kernel.CurrentProcess() != nullptr);
+
+    // Get the current thread and process.
+    Thread& cur_thread = *GetCurrentThread();
+    Process& cur_process = *kernel.CurrentProcess();
+
+    // If the thread's yield count matches, there's nothing for us to do.
+    if (cur_thread.GetYieldScheduleCount() == cur_process.GetScheduledCount()) {
+        return;
+    }
+
+    // Get a reference to the priority queue.
+    auto& priority_queue = GetPriorityQueue(kernel);
+
+    // Perform the yield.
+    {
+        KScopedSchedulerLock lock(kernel);
+
+        const auto cur_state = cur_thread.scheduling_state;
+        if (cur_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+            // Get the current active core.
+            const s32 core_id = cur_thread.GetActiveCore();
+
+            // Migrate the current thread to core -1.
+            cur_thread.SetActiveCore(-1);
+            priority_queue.ChangeCore(core_id, std::addressof(cur_thread));
+            IncrementScheduledCount(std::addressof(cur_thread));
+
+            // If there's nothing scheduled, we can try to perform a migration.
+            if (priority_queue.GetScheduledFront(core_id) == nullptr) {
+                // While we have a suggested thread, try to migrate it!
+                Thread* suggested = priority_queue.GetSuggestedFront(core_id);
+                while (suggested != nullptr) {
+                    // Check if the suggested thread is the top thread on its core.
+                    const s32 suggested_core = suggested->GetActiveCore();
+                    if (Thread* top_on_suggested_core =
+                            (suggested_core >= 0) ? priority_queue.GetScheduledFront(suggested_core)
+                                                  : nullptr;
+                        top_on_suggested_core != suggested) {
+                        // If we're allowed to do a migration, do one.
+                        if (top_on_suggested_core == nullptr ||
+                            top_on_suggested_core->GetPriority() >=
+                                HighestCoreMigrationAllowedPriority) {
+                            suggested->SetActiveCore(core_id);
+                            priority_queue.ChangeCore(suggested_core, suggested);
+                            IncrementScheduledCount(suggested);
+                        }
+
+                        // Regardless of whether we migrated, we had a candidate, so we're done.
+                        break;
+                    }
+
+                    // Get the next suggestion.
+                    suggested = priority_queue.GetSuggestedNext(core_id, suggested);
+                }
+
+                // If the suggestion is different from the current thread, we need to perform an
+                // update.
+                if (suggested != std::addressof(cur_thread)) {
+                    SetSchedulerUpdateNeeded(kernel);
+                } else {
+                    // Otherwise, set the thread's yield count so that we won't waste work until the
+                    // process is scheduled again.
+                    cur_thread.SetYieldScheduleCount(cur_process.GetScheduledCount());
+                }
+            } else {
+                // Otherwise, we have an update to perform.
+                SetSchedulerUpdateNeeded(kernel);
+            }
+        }
+    }
+}
+
+KScheduler::KScheduler(Core::System& system, std::size_t core_id)
+    : system(system), core_id(core_id) {
+    switch_fiber = std::make_shared<Common::Fiber>(OnSwitch, this);
+    this->state.needs_scheduling = true;
+    this->state.interrupt_task_thread_runnable = false;
+    this->state.should_count_idle = false;
+    this->state.idle_count = 0;
+    this->state.idle_thread_stack = nullptr;
+    this->state.highest_priority_thread = nullptr;
+}
+
+KScheduler::~KScheduler() = default;
+
+Thread* KScheduler::GetCurrentThread() const {
+    if (current_thread) {
+        return current_thread;
+    }
+    return idle_thread;
+}
+
+u64 KScheduler::GetLastContextSwitchTicks() const {
+    return last_context_switch_time;
+}
+
+void KScheduler::RescheduleCurrentCore() {
+    ASSERT(GetCurrentThread()->GetDisableDispatchCount() == 1);
+
+    auto& phys_core = system.Kernel().PhysicalCore(core_id);
+    if (phys_core.IsInterrupted()) {
+        phys_core.ClearInterrupt();
+    }
+    guard.lock();
+    if (this->state.needs_scheduling) {
+        Schedule();
+    } else {
+        guard.unlock();
+    }
+}
+
+void KScheduler::OnThreadStart() {
+    SwitchContextStep2();
+}
+
+void KScheduler::Unload(Thread* thread) {
+    if (thread) {
+        thread->SetIsRunning(false);
+        if (thread->IsContinuousOnSVC() && !thread->IsHLEThread()) {
+            system.ArmInterface(core_id).ExceptionalExit();
+            thread->SetContinuousOnSVC(false);
+        }
+        if (!thread->IsHLEThread() && !thread->HasExited()) {
+            Core::ARM_Interface& cpu_core = system.ArmInterface(core_id);
+            cpu_core.SaveContext(thread->GetContext32());
+            cpu_core.SaveContext(thread->GetContext64());
+            // Save the TPIDR_EL0 system register in case it was modified.
+            thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
+            cpu_core.ClearExclusiveState();
+        }
+        thread->context_guard.unlock();
+    }
+}
+
+void KScheduler::Reload(Thread* thread) {
+    if (thread) {
+        ASSERT_MSG(thread->GetSchedulingStatus() == ThreadSchedStatus::Runnable,
+                   "Thread must be runnable.");
+
+        // Cancel any outstanding wakeup events for this thread
+        thread->SetIsRunning(true);
+        thread->SetWasRunning(false);
+
+        auto* const thread_owner_process = thread->GetOwnerProcess();
+        if (thread_owner_process != nullptr) {
+            system.Kernel().MakeCurrentProcess(thread_owner_process);
+        }
+        if (!thread->IsHLEThread()) {
+            Core::ARM_Interface& cpu_core = system.ArmInterface(core_id);
+            cpu_core.LoadContext(thread->GetContext32());
+            cpu_core.LoadContext(thread->GetContext64());
+            cpu_core.SetTlsAddress(thread->GetTLSAddress());
+            cpu_core.SetTPIDR_EL0(thread->GetTPIDR_EL0());
+            cpu_core.ClearExclusiveState();
+        }
+    }
+}
+
+void KScheduler::SwitchContextStep2() {
+    // Load context of new thread
+    Reload(current_thread);
+
+    RescheduleCurrentCore();
+}
+
+void KScheduler::ScheduleImpl() {
+    Thread* previous_thread = current_thread;
+    current_thread = state.highest_priority_thread;
+
+    this->state.needs_scheduling = false;
+
+    if (current_thread == previous_thread) {
+        guard.unlock();
+        return;
+    }
+
+    Process* const previous_process = system.Kernel().CurrentProcess();
+
+    UpdateLastContextSwitchTime(previous_thread, previous_process);
+
+    // Save context for previous thread
+    Unload(previous_thread);
+
+    std::shared_ptr<Common::Fiber>* old_context;
+    if (previous_thread != nullptr) {
+        old_context = &previous_thread->GetHostContext();
+    } else {
+        old_context = &idle_thread->GetHostContext();
+    }
+    guard.unlock();
+
+    Common::Fiber::YieldTo(*old_context, switch_fiber);
+    /// When a thread wakes up, the scheduler may have changed to other in another core.
+    auto& next_scheduler = *system.Kernel().CurrentScheduler();
+    next_scheduler.SwitchContextStep2();
+}
+
+void KScheduler::OnSwitch(void* this_scheduler) {
+    KScheduler* sched = static_cast<KScheduler*>(this_scheduler);
+    sched->SwitchToCurrent();
+}
+
+void KScheduler::SwitchToCurrent() {
+    while (true) {
+        {
+            std::scoped_lock lock{guard};
+            current_thread = state.highest_priority_thread;
+            this->state.needs_scheduling = false;
+        }
+        const auto is_switch_pending = [this] {
+            std::scoped_lock lock{guard};
+            return state.needs_scheduling.load(std::memory_order_relaxed);
+        };
+        do {
+            if (current_thread != nullptr && !current_thread->IsHLEThread()) {
+                current_thread->context_guard.lock();
+                if (!current_thread->IsRunnable()) {
+                    current_thread->context_guard.unlock();
+                    break;
+                }
+                if (static_cast<u32>(current_thread->GetProcessorID()) != core_id) {
+                    current_thread->context_guard.unlock();
+                    break;
+                }
+            }
+            std::shared_ptr<Common::Fiber>* next_context;
+            if (current_thread != nullptr) {
+                next_context = &current_thread->GetHostContext();
+            } else {
+                next_context = &idle_thread->GetHostContext();
+            }
+            Common::Fiber::YieldTo(switch_fiber, *next_context);
+        } while (!is_switch_pending());
+    }
+}
+
+void KScheduler::UpdateLastContextSwitchTime(Thread* thread, Process* process) {
+    const u64 prev_switch_ticks = last_context_switch_time;
+    const u64 most_recent_switch_ticks = system.CoreTiming().GetCPUTicks();
+    const u64 update_ticks = most_recent_switch_ticks - prev_switch_ticks;
+
+    if (thread != nullptr) {
+        thread->UpdateCPUTimeTicks(update_ticks);
+    }
+
+    if (process != nullptr) {
+        process->UpdateCPUTimeTicks(update_ticks);
+    }
+
+    last_context_switch_time = most_recent_switch_ticks;
+}
+
+void KScheduler::Initialize() {
+    std::string name = "Idle Thread Id:" + std::to_string(core_id);
+    std::function<void(void*)> init_func = Core::CpuManager::GetIdleThreadStartFunc();
+    void* init_func_parameter = system.GetCpuManager().GetStartFuncParamater();
+    ThreadType type = static_cast<ThreadType>(THREADTYPE_KERNEL | THREADTYPE_HLE | THREADTYPE_IDLE);
+    auto thread_res = Thread::Create(system, type, name, 0, 64, 0, static_cast<u32>(core_id), 0,
+                                     nullptr, std::move(init_func), init_func_parameter);
+    idle_thread = thread_res.Unwrap().get();
+
+    {
+        KScopedSchedulerLock lock{system.Kernel()};
+        idle_thread->SetStatus(ThreadStatus::Ready);
+    }
+}
+
+KScopedSchedulerLock::KScopedSchedulerLock(KernelCore& kernel)
+    : KScopedLock(kernel.GlobalSchedulerContext().SchedulerLock()) {}
+
+KScopedSchedulerLock::~KScopedSchedulerLock() = default;
+
+} // namespace Kernel

src/core/hle/kernel/k_scheduler.h

@@ -0,0 +1,201 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include <atomic>
+
+#include "common/common_types.h"
+#include "common/spin_lock.h"
+#include "core/hle/kernel/global_scheduler_context.h"
+#include "core/hle/kernel/k_priority_queue.h"
+#include "core/hle/kernel/k_scheduler_lock.h"
+#include "core/hle/kernel/k_scoped_lock.h"
+
+namespace Common {
+class Fiber;
+}
+
+namespace Core {
+class System;
+}
+
+namespace Kernel {
+
+class KernelCore;
+class Process;
+class SchedulerLock;
+class Thread;
+
+class KScheduler final {
+public:
+    explicit KScheduler(Core::System& system, std::size_t core_id);
+    ~KScheduler();
+
+    /// Reschedules to the next available thread (call after current thread is suspended)
+    void RescheduleCurrentCore();
+
+    /// Reschedules cores pending reschedule, to be called on EnableScheduling.
+    static void RescheduleCores(KernelCore& kernel, u64 cores_pending_reschedule,
+                                Core::EmuThreadHandle global_thread);
+
+    /// The next two are for SingleCore Only.
+    /// Unload current thread before preempting core.
+    void Unload(Thread* thread);
+
+    /// Reload current thread after core preemption.
+    void Reload(Thread* thread);
+
+    /// Gets the current running thread
+    [[nodiscard]] Thread* GetCurrentThread() const;
+
+    /// Gets the timestamp for the last context switch in ticks.
+    [[nodiscard]] u64 GetLastContextSwitchTicks() const;
+
+    [[nodiscard]] bool ContextSwitchPending() const {
+        return state.needs_scheduling.load(std::memory_order_relaxed);
+    }
+
+    void Initialize();
+
+    void OnThreadStart();
+
+    [[nodiscard]] std::shared_ptr<Common::Fiber>& ControlContext() {
+        return switch_fiber;
+    }
+
+    [[nodiscard]] const std::shared_ptr<Common::Fiber>& ControlContext() const {
+        return switch_fiber;
+    }
+
+    [[nodiscard]] u64 UpdateHighestPriorityThread(Thread* highest_thread);
+
+    /**
+     * Takes a thread and moves it to the back of the it's priority list.
+     *
+     * @note This operation can be redundant and no scheduling is changed if marked as so.
+     */
+    void YieldWithoutCoreMigration();
+
+    /**
+     * Takes a thread and moves it to the back of the it's priority list.
+     * Afterwards, tries to pick a suggested thread from the suggested queue that has worse time or
+     * a better priority than the next thread in the core.
+     *
+     * @note This operation can be redundant and no scheduling is changed if marked as so.
+     */
+    void YieldWithCoreMigration();
+
+    /**
+     * Takes a thread and moves it out of the scheduling queue.
+     * and into the suggested queue. If no thread can be scheduled afterwards in that core,
+     * a suggested thread is obtained instead.
+     *
+     * @note This operation can be redundant and no scheduling is changed if marked as so.
+     */
+    void YieldToAnyThread();
+
+    /// Notify the scheduler a thread's status has changed.
+    static void OnThreadStateChanged(KernelCore& kernel, Thread* thread, u32 old_state);
+
+    /// Notify the scheduler a thread's priority has changed.
+    static void OnThreadPriorityChanged(KernelCore& kernel, Thread* thread, Thread* current_thread,
+                                        u32 old_priority);
+
+    /// Notify the scheduler a thread's core and/or affinity mask has changed.
+    static void OnThreadAffinityMaskChanged(KernelCore& kernel, Thread* thread,
+                                            const KAffinityMask& old_affinity, s32 old_core);
+
+    static bool CanSchedule(KernelCore& kernel);
+    static bool IsSchedulerUpdateNeeded(const KernelCore& kernel);
+    static void SetSchedulerUpdateNeeded(KernelCore& kernel);
+    static void ClearSchedulerUpdateNeeded(KernelCore& kernel);
+    static void DisableScheduling(KernelCore& kernel);
+    static void EnableScheduling(KernelCore& kernel, u64 cores_needing_scheduling,
+                                 Core::EmuThreadHandle global_thread);
+    [[nodiscard]] static u64 UpdateHighestPriorityThreads(KernelCore& kernel);
+
+private:
+    friend class GlobalSchedulerContext;
+
+    /**
+     * Takes care of selecting the new scheduled threads in three steps:
+     *
+     * 1. First a thread is selected from the top of the priority queue. If no thread
+     *    is obtained then we move to step two, else we are done.
+     *
+     * 2. Second we try to get a suggested thread that's not assigned to any core or
+     *    that is not the top thread in that core.
+     *
+     * 3. Third is no suggested thread is found, we do a second pass and pick a running
+     *    thread in another core and swap it with its current thread.
+     *
+     * returns the cores needing scheduling.
+     */
+    [[nodiscard]] static u64 UpdateHighestPriorityThreadsImpl(KernelCore& kernel);
+
+    [[nodiscard]] static KSchedulerPriorityQueue& GetPriorityQueue(KernelCore& kernel);
+
+    void RotateScheduledQueue(s32 core_id, s32 priority);
+
+    void Schedule() {
+        ASSERT(GetCurrentThread()->GetDisableDispatchCount() == 1);
+        this->ScheduleImpl();
+    }
+
+    /// Switches the CPU's active thread context to that of the specified thread
+    void ScheduleImpl();
+
+    /// When a thread wakes up, it must run this through it's new scheduler
+    void SwitchContextStep2();
+
+    /**
+     * Called on every context switch to update the internal timestamp
+     * This also updates the running time ticks for the given thread and
+     * process using the following difference:
+     *
+     * ticks += most_recent_ticks - last_context_switch_ticks
+     *
+     * The internal tick timestamp for the scheduler is simply the
+     * most recent tick count retrieved. No special arithmetic is
+     * applied to it.
+     */
+    void UpdateLastContextSwitchTime(Thread* thread, Process* process);
+
+    static void OnSwitch(void* this_scheduler);
+    void SwitchToCurrent();
+
+    Thread* current_thread{};
+    Thread* idle_thread{};
+
+    std::shared_ptr<Common::Fiber> switch_fiber{};
+
+    struct SchedulingState {
+        std::atomic<bool> needs_scheduling;
+        bool interrupt_task_thread_runnable{};
+        bool should_count_idle{};
+        u64 idle_count{};
+        Thread* highest_priority_thread{};
+        void* idle_thread_stack{};
+    };
+
+    SchedulingState state;
+
+    Core::System& system;
+    u64 last_context_switch_time{};
+    const std::size_t core_id;
+
+    Common::SpinLock guard{};
+};
+
+class KScopedSchedulerLock : KScopedLock<GlobalSchedulerContext::LockType> {
+public:
+    explicit KScopedSchedulerLock(KernelCore& kernel);
+    ~KScopedSchedulerLock();
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_scheduler_lock.h

@@ -0,0 +1,74 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include "common/assert.h"
+#include "common/spin_lock.h"
+#include "core/hardware_properties.h"
+
+namespace Kernel {
+
+class KernelCore;
+
+template <typename SchedulerType>
+class KAbstractSchedulerLock {
+public:
+    explicit KAbstractSchedulerLock(KernelCore& kernel) : kernel{kernel} {}
+
+    bool IsLockedByCurrentThread() const {
+        return this->owner_thread == kernel.GetCurrentEmuThreadID();
+    }
+
+    void Lock() {
+        if (this->IsLockedByCurrentThread()) {
+            // If we already own the lock, we can just increment the count.
+            ASSERT(this->lock_count > 0);
+            this->lock_count++;
+        } else {
+            // Otherwise, we want to disable scheduling and acquire the spinlock.
+            SchedulerType::DisableScheduling(kernel);
+            this->spin_lock.lock();
+
+            // For debug, ensure that our state is valid.
+            ASSERT(this->lock_count == 0);
+            ASSERT(this->owner_thread == Core::EmuThreadHandle::InvalidHandle());
+
+            // Increment count, take ownership.
+            this->lock_count = 1;
+            this->owner_thread = kernel.GetCurrentEmuThreadID();
+        }
+    }
+
+    void Unlock() {
+        ASSERT(this->IsLockedByCurrentThread());
+        ASSERT(this->lock_count > 0);
+
+        // Release an instance of the lock.
+        if ((--this->lock_count) == 0) {
+            // We're no longer going to hold the lock. Take note of what cores need scheduling.
+            const u64 cores_needing_scheduling =
+                SchedulerType::UpdateHighestPriorityThreads(kernel);
+            Core::EmuThreadHandle leaving_thread = owner_thread;
+
+            // Note that we no longer hold the lock, and unlock the spinlock.
+            this->owner_thread = Core::EmuThreadHandle::InvalidHandle();
+            this->spin_lock.unlock();
+
+            // Enable scheduling, and perform a rescheduling operation.
+            SchedulerType::EnableScheduling(kernel, cores_needing_scheduling, leaving_thread);
+        }
+    }
+
+private:
+    KernelCore& kernel;
+    Common::SpinLock spin_lock{};
+    s32 lock_count{};
+    Core::EmuThreadHandle owner_thread{Core::EmuThreadHandle::InvalidHandle()};
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_scoped_lock.h

@@ -0,0 +1,41 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Kernel {
+
+template <typename T>
+concept KLockable = !std::is_reference_v<T> && requires(T & t) {
+    { t.Lock() }
+    ->std::same_as<void>;
+    { t.Unlock() }
+    ->std::same_as<void>;
+};
+
+template <typename T>
+requires KLockable<T> class KScopedLock {
+public:
+    explicit KScopedLock(T* l) : lock_ptr(l) {
+        this->lock_ptr->Lock();
+    }
+    explicit KScopedLock(T& l) : KScopedLock(std::addressof(l)) { /* ... */
+    }
+    ~KScopedLock() {
+        this->lock_ptr->Unlock();
+    }
+
+    KScopedLock(const KScopedLock&) = delete;
+    KScopedLock(KScopedLock&&) = delete;
+
+private:
+    T* lock_ptr;
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h

@@ -0,0 +1,50 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include "common/common_types.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/time_manager.h"
+
+namespace Kernel {
+
+class KScopedSchedulerLockAndSleep {
+public:
+    explicit KScopedSchedulerLockAndSleep(KernelCore& kernel, Handle& event_handle, Thread* t,
+                                          s64 timeout)
+        : kernel(kernel), event_handle(event_handle), thread(t), timeout_tick(timeout) {
+        event_handle = InvalidHandle;
+
+        // Lock the scheduler.
+        kernel.GlobalSchedulerContext().scheduler_lock.Lock();
+    }
+
+    ~KScopedSchedulerLockAndSleep() {
+        // Register the sleep.
+        if (this->timeout_tick > 0) {
+            kernel.TimeManager().ScheduleTimeEvent(event_handle, this->thread, this->timeout_tick);
+        }
+
+        // Unlock the scheduler.
+        kernel.GlobalSchedulerContext().scheduler_lock.Unlock();
+    }
+
+    void CancelSleep() {
+        this->timeout_tick = 0;
+    }
+
+private:
+    KernelCore& kernel;
+    Handle& event_handle;
+    Thread* thread{};
+    s64 timeout_tick{};
+};
+
+} // namespace Kernel

src/core/hle/kernel/kernel.cpp

@@ -0,0 +1,661 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <array>
+#include <atomic>
+#include <bitset>
+#include <functional>
+#include <memory>
+#include <thread>
+#include <unordered_set>
+#include <utility>
+
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "common/microprofile.h"
+#include "common/thread.h"
+#include "core/arm/arm_interface.h"
+#include "core/arm/cpu_interrupt_handler.h"
+#include "core/arm/exclusive_monitor.h"
+#include "core/core.h"
+#include "core/core_timing.h"
+#include "core/core_timing_util.h"
+#include "core/cpu_manager.h"
+#include "core/device_memory.h"
+#include "core/hardware_properties.h"
+#include "core/hle/kernel/client_port.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/memory/memory_layout.h"
+#include "core/hle/kernel/memory/memory_manager.h"
+#include "core/hle/kernel/memory/slab_heap.h"
+#include "core/hle/kernel/physical_core.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/kernel/resource_limit.h"
+#include "core/hle/kernel/service_thread.h"
+#include "core/hle/kernel/shared_memory.h"
+#include "core/hle/kernel/synchronization.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/time_manager.h"
+#include "core/hle/lock.h"
+#include "core/hle/result.h"
+#include "core/memory.h"
+
+MICROPROFILE_DEFINE(Kernel_SVC, "Kernel", "SVC", MP_RGB(70, 200, 70));
+
+namespace Kernel {
+
+struct KernelCore::Impl {
+    explicit Impl(Core::System& system, KernelCore& kernel)
+        : synchronization{system}, time_manager{system}, global_handle_table{kernel}, system{
+                                                                                          system} {}
+
+    void SetMulticore(bool is_multicore) {
+        this->is_multicore = is_multicore;
+    }
+
+    void Initialize(KernelCore& kernel) {
+        process_list.clear();
+
+        RegisterHostThread();
+
+        global_scheduler_context = std::make_unique<Kernel::GlobalSchedulerContext>(kernel);
+
+        InitializePhysicalCores();
+        InitializeSystemResourceLimit(kernel);
+        InitializeMemoryLayout();
+        InitializePreemption(kernel);
+        InitializeSchedulers();
+        InitializeSuspendThreads();
+    }
+
+    void InitializeCores() {
+        for (auto& core : cores) {
+            core.Initialize(current_process->Is64BitProcess());
+        }
+    }
+
+    void Shutdown() {
+        next_object_id = 0;
+        next_kernel_process_id = Process::InitialKIPIDMin;
+        next_user_process_id = Process::ProcessIDMin;
+        next_thread_id = 1;
+
+        for (std::size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
+            if (suspend_threads[i]) {
+                suspend_threads[i].reset();
+            }
+        }
+
+        cores.clear();
+
+        current_process = nullptr;
+
+        system_resource_limit = nullptr;
+
+        global_handle_table.Clear();
+
+        preemption_event = nullptr;
+
+        named_ports.clear();
+
+        exclusive_monitor.reset();
+
+        num_host_threads = 0;
+        std::fill(register_host_thread_keys.begin(), register_host_thread_keys.end(),
+                  std::thread::id{});
+        std::fill(register_host_thread_values.begin(), register_host_thread_values.end(), 0);
+
+        // Ensures all service threads gracefully shutdown
+        service_threads.clear();
+    }
+
+    void InitializePhysicalCores() {
+        exclusive_monitor =
+            Core::MakeExclusiveMonitor(system.Memory(), Core::Hardware::NUM_CPU_CORES);
+        for (std::size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
+            schedulers[i] = std::make_unique<Kernel::KScheduler>(system, i);
+            cores.emplace_back(i, system, *schedulers[i], interrupts);
+        }
+    }
+
+    void InitializeSchedulers() {
+        for (std::size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
+            cores[i].Scheduler().Initialize();
+        }
+    }
+
+    // Creates the default system resource limit
+    void InitializeSystemResourceLimit(KernelCore& kernel) {
+        system_resource_limit = ResourceLimit::Create(kernel);
+
+        // If setting the default system values fails, then something seriously wrong has occurred.
+        ASSERT(system_resource_limit->SetLimitValue(ResourceType::PhysicalMemory, 0x100000000)
+                   .IsSuccess());
+        ASSERT(system_resource_limit->SetLimitValue(ResourceType::Threads, 800).IsSuccess());
+        ASSERT(system_resource_limit->SetLimitValue(ResourceType::Events, 700).IsSuccess());
+        ASSERT(system_resource_limit->SetLimitValue(ResourceType::TransferMemory, 200).IsSuccess());
+        ASSERT(system_resource_limit->SetLimitValue(ResourceType::Sessions, 900).IsSuccess());
+
+        if (!system_resource_limit->Reserve(ResourceType::PhysicalMemory, 0) ||
+            !system_resource_limit->Reserve(ResourceType::PhysicalMemory, 0x60000)) {
+            UNREACHABLE();
+        }
+    }
+
+    void InitializePreemption(KernelCore& kernel) {
+        preemption_event = Core::Timing::CreateEvent(
+            "PreemptionCallback", [this, &kernel](std::uintptr_t, std::chrono::nanoseconds) {
+                {
+                    KScopedSchedulerLock lock(kernel);
+                    global_scheduler_context->PreemptThreads();
+                }
+                const auto time_interval = std::chrono::nanoseconds{
+                    Core::Timing::msToCycles(std::chrono::milliseconds(10))};
+                system.CoreTiming().ScheduleEvent(time_interval, preemption_event);
+            });
+
+        const auto time_interval =
+            std::chrono::nanoseconds{Core::Timing::msToCycles(std::chrono::milliseconds(10))};
+        system.CoreTiming().ScheduleEvent(time_interval, preemption_event);
+    }
+
+    void InitializeSuspendThreads() {
+        for (std::size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
+            std::string name = "Suspend Thread Id:" + std::to_string(i);
+            std::function<void(void*)> init_func = Core::CpuManager::GetSuspendThreadStartFunc();
+            void* init_func_parameter = system.GetCpuManager().GetStartFuncParamater();
+            const auto type =
+                static_cast<ThreadType>(THREADTYPE_KERNEL | THREADTYPE_HLE | THREADTYPE_SUSPEND);
+            auto thread_res =
+                Thread::Create(system, type, std::move(name), 0, 0, 0, static_cast<u32>(i), 0,
+                               nullptr, std::move(init_func), init_func_parameter);
+
+            suspend_threads[i] = std::move(thread_res).Unwrap();
+        }
+    }
+
+    void MakeCurrentProcess(Process* process) {
+        current_process = process;
+        if (process == nullptr) {
+            return;
+        }
+
+        const u32 core_id = GetCurrentHostThreadID();
+        if (core_id < Core::Hardware::NUM_CPU_CORES) {
+            system.Memory().SetCurrentPageTable(*process, core_id);
+        }
+    }
+
+    void RegisterCoreThread(std::size_t core_id) {
+        const std::thread::id this_id = std::this_thread::get_id();
+        if (!is_multicore) {
+            single_core_thread_id = this_id;
+        }
+        const auto end =
+            register_host_thread_keys.begin() + static_cast<ptrdiff_t>(num_host_threads);
+        const auto it = std::find(register_host_thread_keys.begin(), end, this_id);
+        ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
+        ASSERT(it == end);
+        InsertHostThread(static_cast<u32>(core_id));
+    }
+
+    void RegisterHostThread() {
+        const std::thread::id this_id = std::this_thread::get_id();
+        const auto end =
+            register_host_thread_keys.begin() + static_cast<ptrdiff_t>(num_host_threads);
+        const auto it = std::find(register_host_thread_keys.begin(), end, this_id);
+        if (it == end) {
+            InsertHostThread(registered_thread_ids++);
+        }
+    }
+
+    void InsertHostThread(u32 value) {
+        const size_t index = num_host_threads++;
+        ASSERT_MSG(index < NUM_REGISTRABLE_HOST_THREADS, "Too many host threads");
+        register_host_thread_values[index] = value;
+        register_host_thread_keys[index] = std::this_thread::get_id();
+    }
+
+    [[nodiscard]] u32 GetCurrentHostThreadID() const {
+        const std::thread::id this_id = std::this_thread::get_id();
+        if (!is_multicore && single_core_thread_id == this_id) {
+            return static_cast<u32>(system.GetCpuManager().CurrentCore());
+        }
+        const auto end =
+            register_host_thread_keys.begin() + static_cast<ptrdiff_t>(num_host_threads);
+        const auto it = std::find(register_host_thread_keys.begin(), end, this_id);
+        if (it == end) {
+            return Core::INVALID_HOST_THREAD_ID;
+        }
+        return register_host_thread_values[static_cast<size_t>(
+            std::distance(register_host_thread_keys.begin(), it))];
+    }
+
+    Core::EmuThreadHandle GetCurrentEmuThreadID() const {
+        Core::EmuThreadHandle result = Core::EmuThreadHandle::InvalidHandle();
+        result.host_handle = GetCurrentHostThreadID();
+        if (result.host_handle >= Core::Hardware::NUM_CPU_CORES) {
+            return result;
+        }
+        const Kernel::KScheduler& sched = cores[result.host_handle].Scheduler();
+        const Kernel::Thread* current = sched.GetCurrentThread();
+        if (current != nullptr && !current->IsPhantomMode()) {
+            result.guest_handle = current->GetGlobalHandle();
+        } else {
+            result.guest_handle = InvalidHandle;
+        }
+        return result;
+    }
+
+    void InitializeMemoryLayout() {
+        // Initialize memory layout
+        constexpr Memory::MemoryLayout layout{Memory::MemoryLayout::GetDefaultLayout()};
+        constexpr std::size_t hid_size{0x40000};
+        constexpr std::size_t font_size{0x1100000};
+        constexpr std::size_t irs_size{0x8000};
+        constexpr std::size_t time_size{0x1000};
+        constexpr PAddr hid_addr{layout.System().StartAddress()};
+        constexpr PAddr font_pa{layout.System().StartAddress() + hid_size};
+        constexpr PAddr irs_addr{layout.System().StartAddress() + hid_size + font_size};
+        constexpr PAddr time_addr{layout.System().StartAddress() + hid_size + font_size + irs_size};
+
+        // Initialize memory manager
+        memory_manager = std::make_unique<Memory::MemoryManager>();
+        memory_manager->InitializeManager(Memory::MemoryManager::Pool::Application,
+                                          layout.Application().StartAddress(),
+                                          layout.Application().EndAddress());
+        memory_manager->InitializeManager(Memory::MemoryManager::Pool::Applet,
+                                          layout.Applet().StartAddress(),
+                                          layout.Applet().EndAddress());
+        memory_manager->InitializeManager(Memory::MemoryManager::Pool::System,
+                                          layout.System().StartAddress(),
+                                          layout.System().EndAddress());
+
+        hid_shared_mem = Kernel::SharedMemory::Create(
+            system.Kernel(), system.DeviceMemory(), nullptr,
+            {hid_addr, hid_size / Memory::PageSize}, Memory::MemoryPermission::None,
+            Memory::MemoryPermission::Read, hid_addr, hid_size, "HID:SharedMemory");
+        font_shared_mem = Kernel::SharedMemory::Create(
+            system.Kernel(), system.DeviceMemory(), nullptr,
+            {font_pa, font_size / Memory::PageSize}, Memory::MemoryPermission::None,
+            Memory::MemoryPermission::Read, font_pa, font_size, "Font:SharedMemory");
+        irs_shared_mem = Kernel::SharedMemory::Create(
+            system.Kernel(), system.DeviceMemory(), nullptr,
+            {irs_addr, irs_size / Memory::PageSize}, Memory::MemoryPermission::None,
+            Memory::MemoryPermission::Read, irs_addr, irs_size, "IRS:SharedMemory");
+        time_shared_mem = Kernel::SharedMemory::Create(
+            system.Kernel(), system.DeviceMemory(), nullptr,
+            {time_addr, time_size / Memory::PageSize}, Memory::MemoryPermission::None,
+            Memory::MemoryPermission::Read, time_addr, time_size, "Time:SharedMemory");
+
+        // Allocate slab heaps
+        user_slab_heap_pages = std::make_unique<Memory::SlabHeap<Memory::Page>>();
+
+        // Initialize slab heaps
+        constexpr u64 user_slab_heap_size{0x3de000};
+        user_slab_heap_pages->Initialize(
+            system.DeviceMemory().GetPointer(Core::DramMemoryMap::SlabHeapBase),
+            user_slab_heap_size);
+    }
+
+    std::atomic<u32> next_object_id{0};
+    std::atomic<u64> next_kernel_process_id{Process::InitialKIPIDMin};
+    std::atomic<u64> next_user_process_id{Process::ProcessIDMin};
+    std::atomic<u64> next_thread_id{1};
+
+    // Lists all processes that exist in the current session.
+    std::vector<std::shared_ptr<Process>> process_list;
+    Process* current_process = nullptr;
+    std::unique_ptr<Kernel::GlobalSchedulerContext> global_scheduler_context;
+    Kernel::Synchronization synchronization;
+    Kernel::TimeManager time_manager;
+
+    std::shared_ptr<ResourceLimit> system_resource_limit;
+
+    std::shared_ptr<Core::Timing::EventType> preemption_event;
+
+    // This is the kernel's handle table or supervisor handle table which
+    // stores all the objects in place.
+    HandleTable global_handle_table;
+
+    /// Map of named ports managed by the kernel, which can be retrieved using
+    /// the ConnectToPort SVC.
+    NamedPortTable named_ports;
+
+    std::unique_ptr<Core::ExclusiveMonitor> exclusive_monitor;
+    std::vector<Kernel::PhysicalCore> cores;
+
+    // 0-3 IDs represent core threads, >3 represent others
+    std::atomic<u32> registered_thread_ids{Core::Hardware::NUM_CPU_CORES};
+
+    // Number of host threads is a relatively high number to avoid overflowing
+    static constexpr size_t NUM_REGISTRABLE_HOST_THREADS = 1024;
+    std::atomic<size_t> num_host_threads{0};
+    std::array<std::atomic<std::thread::id>, NUM_REGISTRABLE_HOST_THREADS>
+        register_host_thread_keys{};
+    std::array<std::atomic<u32>, NUM_REGISTRABLE_HOST_THREADS> register_host_thread_values{};
+
+    // Kernel memory management
+    std::unique_ptr<Memory::MemoryManager> memory_manager;
+    std::unique_ptr<Memory::SlabHeap<Memory::Page>> user_slab_heap_pages;
+
+    // Shared memory for services
+    std::shared_ptr<Kernel::SharedMemory> hid_shared_mem;
+    std::shared_ptr<Kernel::SharedMemory> font_shared_mem;
+    std::shared_ptr<Kernel::SharedMemory> irs_shared_mem;
+    std::shared_ptr<Kernel::SharedMemory> time_shared_mem;
+
+    // Threads used for services
+    std::unordered_set<std::shared_ptr<Kernel::ServiceThread>> service_threads;
+
+    std::array<std::shared_ptr<Thread>, Core::Hardware::NUM_CPU_CORES> suspend_threads{};
+    std::array<Core::CPUInterruptHandler, Core::Hardware::NUM_CPU_CORES> interrupts{};
+    std::array<std::unique_ptr<Kernel::KScheduler>, Core::Hardware::NUM_CPU_CORES> schedulers{};
+
+    bool is_multicore{};
+    std::thread::id single_core_thread_id{};
+
+    std::array<u64, Core::Hardware::NUM_CPU_CORES> svc_ticks{};
+
+    // System context
+    Core::System& system;
+};
+
+KernelCore::KernelCore(Core::System& system) : impl{std::make_unique<Impl>(system, *this)} {}
+KernelCore::~KernelCore() {
+    Shutdown();
+}
+
+void KernelCore::SetMulticore(bool is_multicore) {
+    impl->SetMulticore(is_multicore);
+}
+
+void KernelCore::Initialize() {
+    impl->Initialize(*this);
+}
+
+void KernelCore::InitializeCores() {
+    impl->InitializeCores();
+}
+
+void KernelCore::Shutdown() {
+    impl->Shutdown();
+}
+
+std::shared_ptr<ResourceLimit> KernelCore::GetSystemResourceLimit() const {
+    return impl->system_resource_limit;
+}
+
+std::shared_ptr<Thread> KernelCore::RetrieveThreadFromGlobalHandleTable(Handle handle) const {
+    return impl->global_handle_table.Get<Thread>(handle);
+}
+
+void KernelCore::AppendNewProcess(std::shared_ptr<Process> process) {
+    impl->process_list.push_back(std::move(process));
+}
+
+void KernelCore::MakeCurrentProcess(Process* process) {
+    impl->MakeCurrentProcess(process);
+}
+
+Process* KernelCore::CurrentProcess() {
+    return impl->current_process;
+}
+
+const Process* KernelCore::CurrentProcess() const {
+    return impl->current_process;
+}
+
+const std::vector<std::shared_ptr<Process>>& KernelCore::GetProcessList() const {
+    return impl->process_list;
+}
+
+Kernel::GlobalSchedulerContext& KernelCore::GlobalSchedulerContext() {
+    return *impl->global_scheduler_context;
+}
+
+const Kernel::GlobalSchedulerContext& KernelCore::GlobalSchedulerContext() const {
+    return *impl->global_scheduler_context;
+}
+
+Kernel::KScheduler& KernelCore::Scheduler(std::size_t id) {
+    return *impl->schedulers[id];
+}
+
+const Kernel::KScheduler& KernelCore::Scheduler(std::size_t id) const {
+    return *impl->schedulers[id];
+}
+
+Kernel::PhysicalCore& KernelCore::PhysicalCore(std::size_t id) {
+    return impl->cores[id];
+}
+
+const Kernel::PhysicalCore& KernelCore::PhysicalCore(std::size_t id) const {
+    return impl->cores[id];
+}
+
+Kernel::PhysicalCore& KernelCore::CurrentPhysicalCore() {
+    u32 core_id = impl->GetCurrentHostThreadID();
+    ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
+    return impl->cores[core_id];
+}
+
+const Kernel::PhysicalCore& KernelCore::CurrentPhysicalCore() const {
+    u32 core_id = impl->GetCurrentHostThreadID();
+    ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
+    return impl->cores[core_id];
+}
+
+Kernel::KScheduler* KernelCore::CurrentScheduler() {
+    u32 core_id = impl->GetCurrentHostThreadID();
+    if (core_id >= Core::Hardware::NUM_CPU_CORES) {
+        // This is expected when called from not a guest thread
+        return {};
+    }
+    return impl->schedulers[core_id].get();
+}
+
+std::array<Core::CPUInterruptHandler, Core::Hardware::NUM_CPU_CORES>& KernelCore::Interrupts() {
+    return impl->interrupts;
+}
+
+const std::array<Core::CPUInterruptHandler, Core::Hardware::NUM_CPU_CORES>& KernelCore::Interrupts()
+    const {
+    return impl->interrupts;
+}
+
+Kernel::Synchronization& KernelCore::Synchronization() {
+    return impl->synchronization;
+}
+
+const Kernel::Synchronization& KernelCore::Synchronization() const {
+    return impl->synchronization;
+}
+
+Kernel::TimeManager& KernelCore::TimeManager() {
+    return impl->time_manager;
+}
+
+const Kernel::TimeManager& KernelCore::TimeManager() const {
+    return impl->time_manager;
+}
+
+Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() {
+    return *impl->exclusive_monitor;
+}
+
+const Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() const {
+    return *impl->exclusive_monitor;
+}
+
+void KernelCore::InvalidateAllInstructionCaches() {
+    for (auto& physical_core : impl->cores) {
+        physical_core.ArmInterface().ClearInstructionCache();
+    }
+}
+
+void KernelCore::InvalidateCpuInstructionCacheRange(VAddr addr, std::size_t size) {
+    for (auto& physical_core : impl->cores) {
+        if (!physical_core.IsInitialized()) {
+            continue;
+        }
+        physical_core.ArmInterface().InvalidateCacheRange(addr, size);
+    }
+}
+
+void KernelCore::PrepareReschedule(std::size_t id) {
+    // TODO: Reimplement, this
+}
+
+void KernelCore::AddNamedPort(std::string name, std::shared_ptr<ClientPort> port) {
+    impl->named_ports.emplace(std::move(name), std::move(port));
+}
+
+KernelCore::NamedPortTable::iterator KernelCore::FindNamedPort(const std::string& name) {
+    return impl->named_ports.find(name);
+}
+
+KernelCore::NamedPortTable::const_iterator KernelCore::FindNamedPort(
+    const std::string& name) const {
+    return impl->named_ports.find(name);
+}
+
+bool KernelCore::IsValidNamedPort(NamedPortTable::const_iterator port) const {
+    return port != impl->named_ports.cend();
+}
+
+u32 KernelCore::CreateNewObjectID() {
+    return impl->next_object_id++;
+}
+
+u64 KernelCore::CreateNewThreadID() {
+    return impl->next_thread_id++;
+}
+
+u64 KernelCore::CreateNewKernelProcessID() {
+    return impl->next_kernel_process_id++;
+}
+
+u64 KernelCore::CreateNewUserProcessID() {
+    return impl->next_user_process_id++;
+}
+
+Kernel::HandleTable& KernelCore::GlobalHandleTable() {
+    return impl->global_handle_table;
+}
+
+const Kernel::HandleTable& KernelCore::GlobalHandleTable() const {
+    return impl->global_handle_table;
+}
+
+void KernelCore::RegisterCoreThread(std::size_t core_id) {
+    impl->RegisterCoreThread(core_id);
+}
+
+void KernelCore::RegisterHostThread() {
+    impl->RegisterHostThread();
+}
+
+u32 KernelCore::GetCurrentHostThreadID() const {
+    return impl->GetCurrentHostThreadID();
+}
+
+Core::EmuThreadHandle KernelCore::GetCurrentEmuThreadID() const {
+    return impl->GetCurrentEmuThreadID();
+}
+
+Memory::MemoryManager& KernelCore::MemoryManager() {
+    return *impl->memory_manager;
+}
+
+const Memory::MemoryManager& KernelCore::MemoryManager() const {
+    return *impl->memory_manager;
+}
+
+Memory::SlabHeap<Memory::Page>& KernelCore::GetUserSlabHeapPages() {
+    return *impl->user_slab_heap_pages;
+}
+
+const Memory::SlabHeap<Memory::Page>& KernelCore::GetUserSlabHeapPages() const {
+    return *impl->user_slab_heap_pages;
+}
+
+Kernel::SharedMemory& KernelCore::GetHidSharedMem() {
+    return *impl->hid_shared_mem;
+}
+
+const Kernel::SharedMemory& KernelCore::GetHidSharedMem() const {
+    return *impl->hid_shared_mem;
+}
+
+Kernel::SharedMemory& KernelCore::GetFontSharedMem() {
+    return *impl->font_shared_mem;
+}
+
+const Kernel::SharedMemory& KernelCore::GetFontSharedMem() const {
+    return *impl->font_shared_mem;
+}
+
+Kernel::SharedMemory& KernelCore::GetIrsSharedMem() {
+    return *impl->irs_shared_mem;
+}
+
+const Kernel::SharedMemory& KernelCore::GetIrsSharedMem() const {
+    return *impl->irs_shared_mem;
+}
+
+Kernel::SharedMemory& KernelCore::GetTimeSharedMem() {
+    return *impl->time_shared_mem;
+}
+
+const Kernel::SharedMemory& KernelCore::GetTimeSharedMem() const {
+    return *impl->time_shared_mem;
+}
+
+void KernelCore::Suspend(bool in_suspention) {
+    const bool should_suspend = exception_exited || in_suspention;
+    {
+        KScopedSchedulerLock lock(*this);
+        ThreadStatus status = should_suspend ? ThreadStatus::Ready : ThreadStatus::WaitSleep;
+        for (std::size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
+            impl->suspend_threads[i]->SetStatus(status);
+        }
+    }
+}
+
+bool KernelCore::IsMulticore() const {
+    return impl->is_multicore;
+}
+
+void KernelCore::ExceptionalExit() {
+    exception_exited = true;
+    Suspend(true);
+}
+
+void KernelCore::EnterSVCProfile() {
+    std::size_t core = impl->GetCurrentHostThreadID();
+    impl->svc_ticks[core] = MicroProfileEnter(MICROPROFILE_TOKEN(Kernel_SVC));
+}
+
+void KernelCore::ExitSVCProfile() {
+    std::size_t core = impl->GetCurrentHostThreadID();
+    MicroProfileLeave(MICROPROFILE_TOKEN(Kernel_SVC), impl->svc_ticks[core]);
+}
+
+std::weak_ptr<Kernel::ServiceThread> KernelCore::CreateServiceThread(const std::string& name) {
+    auto service_thread = std::make_shared<Kernel::ServiceThread>(*this, 1, name);
+    impl->service_threads.emplace(service_thread);
+    return service_thread;
+}
+
+void KernelCore::ReleaseServiceThread(std::weak_ptr<Kernel::ServiceThread> service_thread) {
+    if (auto strong_ptr = service_thread.lock()) {
+        impl->service_threads.erase(strong_ptr);
+    }
+}
+
+} // namespace Kernel

src/core/hle/kernel/kernel.h

@@ -0,0 +1,275 @@
+// Copyright 2014 Citra Emulator Project / PPSSPP Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <memory>
+#include <string>
+#include <unordered_map>
+#include <vector>
+#include "core/arm/cpu_interrupt_handler.h"
+#include "core/hardware_properties.h"
+#include "core/hle/kernel/memory/memory_types.h"
+#include "core/hle/kernel/object.h"
+
+namespace Core {
+class CPUInterruptHandler;
+class ExclusiveMonitor;
+class System;
+} // namespace Core
+
+namespace Core::Timing {
+class CoreTiming;
+struct EventType;
+} // namespace Core::Timing
+
+namespace Kernel {
+
+namespace Memory {
+class MemoryManager;
+template <typename T>
+class SlabHeap;
+} // namespace Memory
+
+class AddressArbiter;
+class ClientPort;
+class GlobalSchedulerContext;
+class HandleTable;
+class PhysicalCore;
+class Process;
+class ResourceLimit;
+class KScheduler;
+class SharedMemory;
+class ServiceThread;
+class Synchronization;
+class Thread;
+class TimeManager;
+
+/// Represents a single instance of the kernel.
+class KernelCore {
+private:
+    using NamedPortTable = std::unordered_map<std::string, std::shared_ptr<ClientPort>>;
+
+public:
+    /// Constructs an instance of the kernel using the given System
+    /// instance as a context for any necessary system-related state,
+    /// such as threads, CPU core state, etc.
+    ///
+    /// @post After execution of the constructor, the provided System
+    ///       object *must* outlive the kernel instance itself.
+    ///
+    explicit KernelCore(Core::System& system);
+    ~KernelCore();
+
+    KernelCore(const KernelCore&) = delete;
+    KernelCore& operator=(const KernelCore&) = delete;
+
+    KernelCore(KernelCore&&) = delete;
+    KernelCore& operator=(KernelCore&&) = delete;
+
+    /// Sets if emulation is multicore or single core, must be set before Initialize
+    void SetMulticore(bool is_multicore);
+
+    /// Resets the kernel to a clean slate for use.
+    void Initialize();
+
+    /// Initializes the CPU cores.
+    void InitializeCores();
+
+    /// Clears all resources in use by the kernel instance.
+    void Shutdown();
+
+    /// Retrieves a shared pointer to the system resource limit instance.
+    std::shared_ptr<ResourceLimit> GetSystemResourceLimit() const;
+
+    /// Retrieves a shared pointer to a Thread instance within the thread wakeup handle table.
+    std::shared_ptr<Thread> RetrieveThreadFromGlobalHandleTable(Handle handle) const;
+
+    /// Adds the given shared pointer to an internal list of active processes.
+    void AppendNewProcess(std::shared_ptr<Process> process);
+
+    /// Makes the given process the new current process.
+    void MakeCurrentProcess(Process* process);
+
+    /// Retrieves a pointer to the current process.
+    Process* CurrentProcess();
+
+    /// Retrieves a const pointer to the current process.
+    const Process* CurrentProcess() const;
+
+    /// Retrieves the list of processes.
+    const std::vector<std::shared_ptr<Process>>& GetProcessList() const;
+
+    /// Gets the sole instance of the global scheduler
+    Kernel::GlobalSchedulerContext& GlobalSchedulerContext();
+
+    /// Gets the sole instance of the global scheduler
+    const Kernel::GlobalSchedulerContext& GlobalSchedulerContext() const;
+
+    /// Gets the sole instance of the Scheduler assoviated with cpu core 'id'
+    Kernel::KScheduler& Scheduler(std::size_t id);
+
+    /// Gets the sole instance of the Scheduler assoviated with cpu core 'id'
+    const Kernel::KScheduler& Scheduler(std::size_t id) const;
+
+    /// Gets the an instance of the respective physical CPU core.
+    Kernel::PhysicalCore& PhysicalCore(std::size_t id);
+
+    /// Gets the an instance of the respective physical CPU core.
+    const Kernel::PhysicalCore& PhysicalCore(std::size_t id) const;
+
+    /// Gets the sole instance of the Scheduler at the current running core.
+    Kernel::KScheduler* CurrentScheduler();
+
+    /// Gets the an instance of the current physical CPU core.
+    Kernel::PhysicalCore& CurrentPhysicalCore();
+
+    /// Gets the an instance of the current physical CPU core.
+    const Kernel::PhysicalCore& CurrentPhysicalCore() const;
+
+    /// Gets the an instance of the Synchronization Interface.
+    Kernel::Synchronization& Synchronization();
+
+    /// Gets the an instance of the Synchronization Interface.
+    const Kernel::Synchronization& Synchronization() const;
+
+    /// Gets the an instance of the TimeManager Interface.
+    Kernel::TimeManager& TimeManager();
+
+    /// Gets the an instance of the TimeManager Interface.
+    const Kernel::TimeManager& TimeManager() const;
+
+    /// Stops execution of 'id' core, in order to reschedule a new thread.
+    void PrepareReschedule(std::size_t id);
+
+    Core::ExclusiveMonitor& GetExclusiveMonitor();
+
+    const Core::ExclusiveMonitor& GetExclusiveMonitor() const;
+
+    std::array<Core::CPUInterruptHandler, Core::Hardware::NUM_CPU_CORES>& Interrupts();
+
+    const std::array<Core::CPUInterruptHandler, Core::Hardware::NUM_CPU_CORES>& Interrupts() const;
+
+    void InvalidateAllInstructionCaches();
+
+    void InvalidateCpuInstructionCacheRange(VAddr addr, std::size_t size);
+
+    /// Adds a port to the named port table
+    void AddNamedPort(std::string name, std::shared_ptr<ClientPort> port);
+
+    /// Finds a port within the named port table with the given name.
+    NamedPortTable::iterator FindNamedPort(const std::string& name);
+
+    /// Finds a port within the named port table with the given name.
+    NamedPortTable::const_iterator FindNamedPort(const std::string& name) const;
+
+    /// Determines whether or not the given port is a valid named port.
+    bool IsValidNamedPort(NamedPortTable::const_iterator port) const;
+
+    /// Gets the current host_thread/guest_thread handle.
+    Core::EmuThreadHandle GetCurrentEmuThreadID() const;
+
+    /// Gets the current host_thread handle.
+    u32 GetCurrentHostThreadID() const;
+
+    /// Register the current thread as a CPU Core Thread.
+    void RegisterCoreThread(std::size_t core_id);
+
+    /// Register the current thread as a non CPU core thread.
+    void RegisterHostThread();
+
+    /// Gets the virtual memory manager for the kernel.
+    Memory::MemoryManager& MemoryManager();
+
+    /// Gets the virtual memory manager for the kernel.
+    const Memory::MemoryManager& MemoryManager() const;
+
+    /// Gets the slab heap allocated for user space pages.
+    Memory::SlabHeap<Memory::Page>& GetUserSlabHeapPages();
+
+    /// Gets the slab heap allocated for user space pages.
+    const Memory::SlabHeap<Memory::Page>& GetUserSlabHeapPages() const;
+
+    /// Gets the shared memory object for HID services.
+    Kernel::SharedMemory& GetHidSharedMem();
+
+    /// Gets the shared memory object for HID services.
+    const Kernel::SharedMemory& GetHidSharedMem() const;
+
+    /// Gets the shared memory object for font services.
+    Kernel::SharedMemory& GetFontSharedMem();
+
+    /// Gets the shared memory object for font services.
+    const Kernel::SharedMemory& GetFontSharedMem() const;
+
+    /// Gets the shared memory object for IRS services.
+    Kernel::SharedMemory& GetIrsSharedMem();
+
+    /// Gets the shared memory object for IRS services.
+    const Kernel::SharedMemory& GetIrsSharedMem() const;
+
+    /// Gets the shared memory object for Time services.
+    Kernel::SharedMemory& GetTimeSharedMem();
+
+    /// Gets the shared memory object for Time services.
+    const Kernel::SharedMemory& GetTimeSharedMem() const;
+
+    /// Suspend/unsuspend the OS.
+    void Suspend(bool in_suspention);
+
+    /// Exceptional exit the OS.
+    void ExceptionalExit();
+
+    bool IsMulticore() const;
+
+    void EnterSVCProfile();
+
+    void ExitSVCProfile();
+
+    /**
+     * Creates an HLE service thread, which are used to execute service routines asynchronously.
+     * While these are allocated per ServerSession, these need to be owned and managed outside of
+     * ServerSession to avoid a circular dependency.
+     * @param name String name for the ServerSession creating this thread, used for debug purposes.
+     * @returns The a weak pointer newly created service thread.
+     */
+    std::weak_ptr<Kernel::ServiceThread> CreateServiceThread(const std::string& name);
+
+    /**
+     * Releases a HLE service thread, instructing KernelCore to free it. This should be called when
+     * the ServerSession associated with the thread is destroyed.
+     * @param service_thread Service thread to release.
+     */
+    void ReleaseServiceThread(std::weak_ptr<Kernel::ServiceThread> service_thread);
+
+private:
+    friend class Object;
+    friend class Process;
+    friend class Thread;
+
+    /// Creates a new object ID, incrementing the internal object ID counter.
+    u32 CreateNewObjectID();
+
+    /// Creates a new process ID, incrementing the internal process ID counter;
+    u64 CreateNewKernelProcessID();
+
+    /// Creates a new process ID, incrementing the internal process ID counter;
+    u64 CreateNewUserProcessID();
+
+    /// Creates a new thread ID, incrementing the internal thread ID counter.
+    u64 CreateNewThreadID();
+
+    /// Provides a reference to the global handle table.
+    Kernel::HandleTable& GlobalHandleTable();
+
+    /// Provides a const reference to the global handle table.
+    const Kernel::HandleTable& GlobalHandleTable() const;
+
+    struct Impl;
+    std::unique_ptr<Impl> impl;
+    bool exception_exited{};
+};
+
+} // namespace Kernel

src/core/hle/kernel/memory/address_space_info.cpp

@@ -0,0 +1,117 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#include <array>
+
+#include "common/assert.h"
+#include "core/hle/kernel/memory/address_space_info.h"
+
+namespace Kernel::Memory {
+
+namespace {
+
+enum : u64 {
+    Size_1_MB = 0x100000,
+    Size_2_MB = 2 * Size_1_MB,
+    Size_128_MB = 128 * Size_1_MB,
+    Size_1_GB = 0x40000000,
+    Size_2_GB = 2 * Size_1_GB,
+    Size_4_GB = 4 * Size_1_GB,
+    Size_6_GB = 6 * Size_1_GB,
+    Size_64_GB = 64 * Size_1_GB,
+    Size_512_GB = 512 * Size_1_GB,
+    Invalid = std::numeric_limits<u64>::max(),
+};
+
+// clang-format off
+constexpr std::array<AddressSpaceInfo, 13> AddressSpaceInfos{{
+   { .bit_width = 32, .address = Size_2_MB  , .size = Size_1_GB   - Size_2_MB  , .type = AddressSpaceInfo::Type::Is32Bit,    },
+   { .bit_width = 32, .address = Size_1_GB  , .size = Size_4_GB   - Size_1_GB  , .type = AddressSpaceInfo::Type::Small64Bit, },
+   { .bit_width = 32, .address = Invalid    , .size = Size_1_GB                , .type = AddressSpaceInfo::Type::Heap,       },
+   { .bit_width = 32, .address = Invalid    , .size = Size_1_GB                , .type = AddressSpaceInfo::Type::Alias,      },
+   { .bit_width = 36, .address = Size_128_MB, .size = Size_2_GB   - Size_128_MB, .type = AddressSpaceInfo::Type::Is32Bit,    },
+   { .bit_width = 36, .address = Size_2_GB  , .size = Size_64_GB  - Size_2_GB  , .type = AddressSpaceInfo::Type::Small64Bit, },
+   { .bit_width = 36, .address = Invalid    , .size = Size_6_GB                , .type = AddressSpaceInfo::Type::Heap,       },
+   { .bit_width = 36, .address = Invalid    , .size = Size_6_GB                , .type = AddressSpaceInfo::Type::Alias,      },
+   { .bit_width = 39, .address = Size_128_MB, .size = Size_512_GB - Size_128_MB, .type = AddressSpaceInfo::Type::Large64Bit, },
+   { .bit_width = 39, .address = Invalid    , .size = Size_64_GB               , .type = AddressSpaceInfo::Type::Is32Bit     },
+   { .bit_width = 39, .address = Invalid    , .size = Size_6_GB                , .type = AddressSpaceInfo::Type::Heap,       },
+   { .bit_width = 39, .address = Invalid    , .size = Size_64_GB               , .type = AddressSpaceInfo::Type::Alias,      },
+   { .bit_width = 39, .address = Invalid    , .size = Size_2_GB                , .type = AddressSpaceInfo::Type::Stack,      },
+}};
+// clang-format on
+
+constexpr bool IsAllowedIndexForAddress(std::size_t index) {
+    return index < AddressSpaceInfos.size() && AddressSpaceInfos[index].address != Invalid;
+}
+
+using IndexArray = std::array<std::size_t, static_cast<std::size_t>(AddressSpaceInfo::Type::Count)>;
+
+constexpr IndexArray AddressSpaceIndices32Bit{
+    0, 1, 0, 2, 0, 3,
+};
+
+constexpr IndexArray AddressSpaceIndices36Bit{
+    4, 5, 4, 6, 4, 7,
+};
+
+constexpr IndexArray AddressSpaceIndices39Bit{
+    9, 8, 8, 10, 12, 11,
+};
+
+constexpr bool IsAllowed32BitType(AddressSpaceInfo::Type type) {
+    return type < AddressSpaceInfo::Type::Count && type != AddressSpaceInfo::Type::Large64Bit &&
+           type != AddressSpaceInfo::Type::Stack;
+}
+
+constexpr bool IsAllowed36BitType(AddressSpaceInfo::Type type) {
+    return type < AddressSpaceInfo::Type::Count && type != AddressSpaceInfo::Type::Large64Bit &&
+           type != AddressSpaceInfo::Type::Stack;
+}
+
+constexpr bool IsAllowed39BitType(AddressSpaceInfo::Type type) {
+    return type < AddressSpaceInfo::Type::Count && type != AddressSpaceInfo::Type::Small64Bit;
+}
+
+} // namespace
+
+u64 AddressSpaceInfo::GetAddressSpaceStart(std::size_t width, Type type) {
+    const std::size_t index{static_cast<std::size_t>(type)};
+    switch (width) {
+    case 32:
+        ASSERT(IsAllowed32BitType(type));
+        ASSERT(IsAllowedIndexForAddress(AddressSpaceIndices32Bit[index]));
+        return AddressSpaceInfos[AddressSpaceIndices32Bit[index]].address;
+    case 36:
+        ASSERT(IsAllowed36BitType(type));
+        ASSERT(IsAllowedIndexForAddress(AddressSpaceIndices36Bit[index]));
+        return AddressSpaceInfos[AddressSpaceIndices36Bit[index]].address;
+    case 39:
+        ASSERT(IsAllowed39BitType(type));
+        ASSERT(IsAllowedIndexForAddress(AddressSpaceIndices39Bit[index]));
+        return AddressSpaceInfos[AddressSpaceIndices39Bit[index]].address;
+    }
+    UNREACHABLE();
+}
+
+std::size_t AddressSpaceInfo::GetAddressSpaceSize(std::size_t width, Type type) {
+    const std::size_t index{static_cast<std::size_t>(type)};
+    switch (width) {
+    case 32:
+        ASSERT(IsAllowed32BitType(type));
+        return AddressSpaceInfos[AddressSpaceIndices32Bit[index]].size;
+    case 36:
+        ASSERT(IsAllowed36BitType(type));
+        return AddressSpaceInfos[AddressSpaceIndices36Bit[index]].size;
+    case 39:
+        ASSERT(IsAllowed39BitType(type));
+        return AddressSpaceInfos[AddressSpaceIndices39Bit[index]].size;
+    }
+    UNREACHABLE();
+}
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/address_space_info.h

@@ -0,0 +1,34 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Kernel::Memory {
+
+struct AddressSpaceInfo final {
+    enum class Type : u32 {
+        Is32Bit = 0,
+        Small64Bit = 1,
+        Large64Bit = 2,
+        Heap = 3,
+        Stack = 4,
+        Alias = 5,
+        Count,
+    };
+
+    static u64 GetAddressSpaceStart(std::size_t width, Type type);
+    static std::size_t GetAddressSpaceSize(std::size_t width, Type type);
+
+    const std::size_t bit_width{};
+    const std::size_t address{};
+    const std::size_t size{};
+    const Type type{};
+};
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/memory_block.h

@@ -0,0 +1,335 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include "common/alignment.h"
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "core/hle/kernel/memory/memory_types.h"
+#include "core/hle/kernel/svc_types.h"
+
+namespace Kernel::Memory {
+
+enum class MemoryState : u32 {
+    None = 0,
+    Mask = 0xFF,
+    All = ~None,
+
+    FlagCanReprotect = (1 << 8),
+    FlagCanDebug = (1 << 9),
+    FlagCanUseIpc = (1 << 10),
+    FlagCanUseNonDeviceIpc = (1 << 11),
+    FlagCanUseNonSecureIpc = (1 << 12),
+    FlagMapped = (1 << 13),
+    FlagCode = (1 << 14),
+    FlagCanAlias = (1 << 15),
+    FlagCanCodeAlias = (1 << 16),
+    FlagCanTransfer = (1 << 17),
+    FlagCanQueryPhysical = (1 << 18),
+    FlagCanDeviceMap = (1 << 19),
+    FlagCanAlignedDeviceMap = (1 << 20),
+    FlagCanIpcUserBuffer = (1 << 21),
+    FlagReferenceCounted = (1 << 22),
+    FlagCanMapProcess = (1 << 23),
+    FlagCanChangeAttribute = (1 << 24),
+    FlagCanCodeMemory = (1 << 25),
+
+    FlagsData = FlagCanReprotect | FlagCanUseIpc | FlagCanUseNonDeviceIpc | FlagCanUseNonSecureIpc |
+                FlagMapped | FlagCanAlias | FlagCanTransfer | FlagCanQueryPhysical |
+                FlagCanDeviceMap | FlagCanAlignedDeviceMap | FlagCanIpcUserBuffer |
+                FlagReferenceCounted | FlagCanChangeAttribute,
+
+    FlagsCode = FlagCanDebug | FlagCanUseIpc | FlagCanUseNonDeviceIpc | FlagCanUseNonSecureIpc |
+                FlagMapped | FlagCode | FlagCanQueryPhysical | FlagCanDeviceMap |
+                FlagCanAlignedDeviceMap | FlagReferenceCounted,
+
+    FlagsMisc = FlagMapped | FlagReferenceCounted | FlagCanQueryPhysical | FlagCanDeviceMap,
+
+    Free = static_cast<u32>(Svc::MemoryState::Free),
+    Io = static_cast<u32>(Svc::MemoryState::Io) | FlagMapped,
+    Static = static_cast<u32>(Svc::MemoryState::Static) | FlagMapped | FlagCanQueryPhysical,
+    Code = static_cast<u32>(Svc::MemoryState::Code) | FlagsCode | FlagCanMapProcess,
+    CodeData = static_cast<u32>(Svc::MemoryState::CodeData) | FlagsData | FlagCanMapProcess |
+               FlagCanCodeMemory,
+    Shared = static_cast<u32>(Svc::MemoryState::Shared) | FlagMapped | FlagReferenceCounted,
+    Normal = static_cast<u32>(Svc::MemoryState::Normal) | FlagsData | FlagCanCodeMemory,
+
+    AliasCode = static_cast<u32>(Svc::MemoryState::AliasCode) | FlagsCode | FlagCanMapProcess |
+                FlagCanCodeAlias,
+    AliasCodeData = static_cast<u32>(Svc::MemoryState::AliasCodeData) | FlagsData |
+                    FlagCanMapProcess | FlagCanCodeAlias | FlagCanCodeMemory,
+
+    Ipc = static_cast<u32>(Svc::MemoryState::Ipc) | FlagsMisc | FlagCanAlignedDeviceMap |
+          FlagCanUseIpc | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+
+    Stack = static_cast<u32>(Svc::MemoryState::Stack) | FlagsMisc | FlagCanAlignedDeviceMap |
+            FlagCanUseIpc | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+
+    ThreadLocal =
+        static_cast<u32>(Svc::MemoryState::ThreadLocal) | FlagMapped | FlagReferenceCounted,
+
+    Transfered = static_cast<u32>(Svc::MemoryState::Transfered) | FlagsMisc |
+                 FlagCanAlignedDeviceMap | FlagCanChangeAttribute | FlagCanUseIpc |
+                 FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+
+    SharedTransfered = static_cast<u32>(Svc::MemoryState::SharedTransfered) | FlagsMisc |
+                       FlagCanAlignedDeviceMap | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+
+    SharedCode = static_cast<u32>(Svc::MemoryState::SharedCode) | FlagMapped |
+                 FlagReferenceCounted | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+
+    Inaccessible = static_cast<u32>(Svc::MemoryState::Inaccessible),
+
+    NonSecureIpc = static_cast<u32>(Svc::MemoryState::NonSecureIpc) | FlagsMisc |
+                   FlagCanAlignedDeviceMap | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+
+    NonDeviceIpc =
+        static_cast<u32>(Svc::MemoryState::NonDeviceIpc) | FlagsMisc | FlagCanUseNonDeviceIpc,
+
+    Kernel = static_cast<u32>(Svc::MemoryState::Kernel) | FlagMapped,
+
+    GeneratedCode = static_cast<u32>(Svc::MemoryState::GeneratedCode) | FlagMapped |
+                    FlagReferenceCounted | FlagCanDebug,
+    CodeOut = static_cast<u32>(Svc::MemoryState::CodeOut) | FlagMapped | FlagReferenceCounted,
+};
+DECLARE_ENUM_FLAG_OPERATORS(MemoryState);
+
+static_assert(static_cast<u32>(MemoryState::Free) == 0x00000000);
+static_assert(static_cast<u32>(MemoryState::Io) == 0x00002001);
+static_assert(static_cast<u32>(MemoryState::Static) == 0x00042002);
+static_assert(static_cast<u32>(MemoryState::Code) == 0x00DC7E03);
+static_assert(static_cast<u32>(MemoryState::CodeData) == 0x03FEBD04);
+static_assert(static_cast<u32>(MemoryState::Normal) == 0x037EBD05);
+static_assert(static_cast<u32>(MemoryState::Shared) == 0x00402006);
+static_assert(static_cast<u32>(MemoryState::AliasCode) == 0x00DD7E08);
+static_assert(static_cast<u32>(MemoryState::AliasCodeData) == 0x03FFBD09);
+static_assert(static_cast<u32>(MemoryState::Ipc) == 0x005C3C0A);
+static_assert(static_cast<u32>(MemoryState::Stack) == 0x005C3C0B);
+static_assert(static_cast<u32>(MemoryState::ThreadLocal) == 0x0040200C);
+static_assert(static_cast<u32>(MemoryState::Transfered) == 0x015C3C0D);
+static_assert(static_cast<u32>(MemoryState::SharedTransfered) == 0x005C380E);
+static_assert(static_cast<u32>(MemoryState::SharedCode) == 0x0040380F);
+static_assert(static_cast<u32>(MemoryState::Inaccessible) == 0x00000010);
+static_assert(static_cast<u32>(MemoryState::NonSecureIpc) == 0x005C3811);
+static_assert(static_cast<u32>(MemoryState::NonDeviceIpc) == 0x004C2812);
+static_assert(static_cast<u32>(MemoryState::Kernel) == 0x00002013);
+static_assert(static_cast<u32>(MemoryState::GeneratedCode) == 0x00402214);
+static_assert(static_cast<u32>(MemoryState::CodeOut) == 0x00402015);
+
+enum class MemoryPermission : u8 {
+    None = 0,
+    Mask = static_cast<u8>(~None),
+
+    Read = 1 << 0,
+    Write = 1 << 1,
+    Execute = 1 << 2,
+
+    ReadAndWrite = Read | Write,
+    ReadAndExecute = Read | Execute,
+
+    UserMask = static_cast<u8>(Svc::MemoryPermission::Read | Svc::MemoryPermission::Write |
+                               Svc::MemoryPermission::Execute),
+};
+DECLARE_ENUM_FLAG_OPERATORS(MemoryPermission);
+
+enum class MemoryAttribute : u8 {
+    None = 0x00,
+    Mask = 0x7F,
+    All = Mask,
+    DontCareMask = 0x80,
+
+    Locked = static_cast<u8>(Svc::MemoryAttribute::Locked),
+    IpcLocked = static_cast<u8>(Svc::MemoryAttribute::IpcLocked),
+    DeviceShared = static_cast<u8>(Svc::MemoryAttribute::DeviceShared),
+    Uncached = static_cast<u8>(Svc::MemoryAttribute::Uncached),
+
+    IpcAndDeviceMapped = IpcLocked | DeviceShared,
+    LockedAndIpcLocked = Locked | IpcLocked,
+    DeviceSharedAndUncached = DeviceShared | Uncached
+};
+DECLARE_ENUM_FLAG_OPERATORS(MemoryAttribute);
+
+static_assert((static_cast<u8>(MemoryAttribute::Mask) &
+               static_cast<u8>(MemoryAttribute::DontCareMask)) == 0);
+
+struct MemoryInfo {
+    VAddr addr{};
+    std::size_t size{};
+    MemoryState state{};
+    MemoryPermission perm{};
+    MemoryAttribute attribute{};
+    MemoryPermission original_perm{};
+    u16 ipc_lock_count{};
+    u16 device_use_count{};
+
+    constexpr Svc::MemoryInfo GetSvcMemoryInfo() const {
+        return {
+            addr,
+            size,
+            static_cast<Svc::MemoryState>(state & MemoryState::Mask),
+            static_cast<Svc::MemoryAttribute>(attribute & MemoryAttribute::Mask),
+            static_cast<Svc::MemoryPermission>(perm & MemoryPermission::UserMask),
+            ipc_lock_count,
+            device_use_count,
+        };
+    }
+
+    constexpr VAddr GetAddress() const {
+        return addr;
+    }
+    constexpr std::size_t GetSize() const {
+        return size;
+    }
+    constexpr std::size_t GetNumPages() const {
+        return GetSize() / PageSize;
+    }
+    constexpr VAddr GetEndAddress() const {
+        return GetAddress() + GetSize();
+    }
+    constexpr VAddr GetLastAddress() const {
+        return GetEndAddress() - 1;
+    }
+};
+
+class MemoryBlock final {
+    friend class MemoryBlockManager;
+
+private:
+    VAddr addr{};
+    std::size_t num_pages{};
+    MemoryState state{MemoryState::None};
+    u16 ipc_lock_count{};
+    u16 device_use_count{};
+    MemoryPermission perm{MemoryPermission::None};
+    MemoryPermission original_perm{MemoryPermission::None};
+    MemoryAttribute attribute{MemoryAttribute::None};
+
+public:
+    static constexpr int Compare(const MemoryBlock& lhs, const MemoryBlock& rhs) {
+        if (lhs.GetAddress() < rhs.GetAddress()) {
+            return -1;
+        } else if (lhs.GetAddress() <= rhs.GetLastAddress()) {
+            return 0;
+        } else {
+            return 1;
+        }
+    }
+
+public:
+    constexpr MemoryBlock() = default;
+    constexpr MemoryBlock(VAddr addr_, std::size_t num_pages_, MemoryState state_,
+                          MemoryPermission perm_, MemoryAttribute attribute_)
+        : addr{addr_}, num_pages(num_pages_), state{state_}, perm{perm_}, attribute{attribute_} {}
+
+    constexpr VAddr GetAddress() const {
+        return addr;
+    }
+
+    constexpr std::size_t GetNumPages() const {
+        return num_pages;
+    }
+
+    constexpr std::size_t GetSize() const {
+        return GetNumPages() * PageSize;
+    }
+
+    constexpr VAddr GetEndAddress() const {
+        return GetAddress() + GetSize();
+    }
+
+    constexpr VAddr GetLastAddress() const {
+        return GetEndAddress() - 1;
+    }
+
+    constexpr MemoryInfo GetMemoryInfo() const {
+        return {
+            GetAddress(), GetSize(),     state,          perm,
+            attribute,    original_perm, ipc_lock_count, device_use_count,
+        };
+    }
+
+    void ShareToDevice(MemoryPermission /*new_perm*/) {
+        ASSERT((attribute & MemoryAttribute::DeviceShared) == MemoryAttribute::DeviceShared ||
+               device_use_count == 0);
+        attribute |= MemoryAttribute::DeviceShared;
+        const u16 new_use_count{++device_use_count};
+        ASSERT(new_use_count > 0);
+    }
+
+    void UnshareToDevice(MemoryPermission /*new_perm*/) {
+        ASSERT((attribute & MemoryAttribute::DeviceShared) == MemoryAttribute::DeviceShared);
+        const u16 prev_use_count{device_use_count--};
+        ASSERT(prev_use_count > 0);
+        if (prev_use_count == 1) {
+            attribute &= ~MemoryAttribute::DeviceShared;
+        }
+    }
+
+private:
+    constexpr bool HasProperties(MemoryState s, MemoryPermission p, MemoryAttribute a) const {
+        constexpr MemoryAttribute AttributeIgnoreMask{MemoryAttribute::DontCareMask |
+                                                      MemoryAttribute::IpcLocked |
+                                                      MemoryAttribute::DeviceShared};
+        return state == s && perm == p &&
+               (attribute | AttributeIgnoreMask) == (a | AttributeIgnoreMask);
+    }
+
+    constexpr bool HasSameProperties(const MemoryBlock& rhs) const {
+        return state == rhs.state && perm == rhs.perm && original_perm == rhs.original_perm &&
+               attribute == rhs.attribute && ipc_lock_count == rhs.ipc_lock_count &&
+               device_use_count == rhs.device_use_count;
+    }
+
+    constexpr bool Contains(VAddr start) const {
+        return GetAddress() <= start && start <= GetEndAddress();
+    }
+
+    constexpr void Add(std::size_t count) {
+        ASSERT(count > 0);
+        ASSERT(GetAddress() + count * PageSize - 1 < GetEndAddress() + count * PageSize - 1);
+
+        num_pages += count;
+    }
+
+    constexpr void Update(MemoryState new_state, MemoryPermission new_perm,
+                          MemoryAttribute new_attribute) {
+        ASSERT(original_perm == MemoryPermission::None);
+        ASSERT((attribute & MemoryAttribute::IpcLocked) == MemoryAttribute::None);
+
+        state = new_state;
+        perm = new_perm;
+
+        attribute = static_cast<MemoryAttribute>(
+            new_attribute |
+            (attribute & (MemoryAttribute::IpcLocked | MemoryAttribute::DeviceShared)));
+    }
+
+    constexpr MemoryBlock Split(VAddr split_addr) {
+        ASSERT(GetAddress() < split_addr);
+        ASSERT(Contains(split_addr));
+        ASSERT(Common::IsAligned(split_addr, PageSize));
+
+        MemoryBlock block;
+        block.addr = addr;
+        block.num_pages = (split_addr - GetAddress()) / PageSize;
+        block.state = state;
+        block.ipc_lock_count = ipc_lock_count;
+        block.device_use_count = device_use_count;
+        block.perm = perm;
+        block.original_perm = original_perm;
+        block.attribute = attribute;
+
+        addr = split_addr;
+        num_pages -= block.num_pages;
+
+        return block;
+    }
+};
+static_assert(std::is_trivially_destructible<MemoryBlock>::value);
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/memory_block_manager.cpp

@@ -0,0 +1,223 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/kernel/memory/memory_block_manager.h"
+#include "core/hle/kernel/memory/memory_types.h"
+
+namespace Kernel::Memory {
+
+MemoryBlockManager::MemoryBlockManager(VAddr start_addr, VAddr end_addr)
+    : start_addr{start_addr}, end_addr{end_addr} {
+    const u64 num_pages{(end_addr - start_addr) / PageSize};
+    memory_block_tree.emplace_back(start_addr, num_pages, MemoryState::Free, MemoryPermission::None,
+                                   MemoryAttribute::None);
+}
+
+MemoryBlockManager::iterator MemoryBlockManager::FindIterator(VAddr addr) {
+    auto node{memory_block_tree.begin()};
+    while (node != end()) {
+        const VAddr end_addr{node->GetNumPages() * PageSize + node->GetAddress()};
+        if (node->GetAddress() <= addr && end_addr - 1 >= addr) {
+            return node;
+        }
+        node = std::next(node);
+    }
+    return end();
+}
+
+VAddr MemoryBlockManager::FindFreeArea(VAddr region_start, std::size_t region_num_pages,
+                                       std::size_t num_pages, std::size_t align, std::size_t offset,
+                                       std::size_t guard_pages) {
+    if (num_pages == 0) {
+        return {};
+    }
+
+    const VAddr region_end{region_start + region_num_pages * PageSize};
+    const VAddr region_last{region_end - 1};
+    for (auto it{FindIterator(region_start)}; it != memory_block_tree.cend(); it++) {
+        const auto info{it->GetMemoryInfo()};
+        if (region_last < info.GetAddress()) {
+            break;
+        }
+
+        if (info.state != MemoryState::Free) {
+            continue;
+        }
+
+        VAddr area{(info.GetAddress() <= region_start) ? region_start : info.GetAddress()};
+        area += guard_pages * PageSize;
+
+        const VAddr offset_area{Common::AlignDown(area, align) + offset};
+        area = (area <= offset_area) ? offset_area : offset_area + align;
+
+        const VAddr area_end{area + num_pages * PageSize + guard_pages * PageSize};
+        const VAddr area_last{area_end - 1};
+
+        if (info.GetAddress() <= area && area < area_last && area_last <= region_last &&
+            area_last <= info.GetLastAddress()) {
+            return area;
+        }
+    }
+
+    return {};
+}
+
+void MemoryBlockManager::Update(VAddr addr, std::size_t num_pages, MemoryState prev_state,
+                                MemoryPermission prev_perm, MemoryAttribute prev_attribute,
+                                MemoryState state, MemoryPermission perm,
+                                MemoryAttribute attribute) {
+    const VAddr end_addr{addr + num_pages * PageSize};
+    iterator node{memory_block_tree.begin()};
+
+    prev_attribute |= MemoryAttribute::IpcAndDeviceMapped;
+
+    while (node != memory_block_tree.end()) {
+        MemoryBlock* block{&(*node)};
+        iterator next_node{std::next(node)};
+        const VAddr cur_addr{block->GetAddress()};
+        const VAddr cur_end_addr{block->GetNumPages() * PageSize + cur_addr};
+
+        if (addr < cur_end_addr && cur_addr < end_addr) {
+            if (!block->HasProperties(prev_state, prev_perm, prev_attribute)) {
+                node = next_node;
+                continue;
+            }
+
+            iterator new_node{node};
+            if (addr > cur_addr) {
+                memory_block_tree.insert(node, block->Split(addr));
+            }
+
+            if (end_addr < cur_end_addr) {
+                new_node = memory_block_tree.insert(node, block->Split(end_addr));
+            }
+
+            new_node->Update(state, perm, attribute);
+
+            MergeAdjacent(new_node, next_node);
+        }
+
+        if (cur_end_addr - 1 >= end_addr - 1) {
+            break;
+        }
+
+        node = next_node;
+    }
+}
+
+void MemoryBlockManager::Update(VAddr addr, std::size_t num_pages, MemoryState state,
+                                MemoryPermission perm, MemoryAttribute attribute) {
+    const VAddr end_addr{addr + num_pages * PageSize};
+    iterator node{memory_block_tree.begin()};
+
+    while (node != memory_block_tree.end()) {
+        MemoryBlock* block{&(*node)};
+        iterator next_node{std::next(node)};
+        const VAddr cur_addr{block->GetAddress()};
+        const VAddr cur_end_addr{block->GetNumPages() * PageSize + cur_addr};
+
+        if (addr < cur_end_addr && cur_addr < end_addr) {
+            iterator new_node{node};
+
+            if (addr > cur_addr) {
+                memory_block_tree.insert(node, block->Split(addr));
+            }
+
+            if (end_addr < cur_end_addr) {
+                new_node = memory_block_tree.insert(node, block->Split(end_addr));
+            }
+
+            new_node->Update(state, perm, attribute);
+
+            MergeAdjacent(new_node, next_node);
+        }
+
+        if (cur_end_addr - 1 >= end_addr - 1) {
+            break;
+        }
+
+        node = next_node;
+    }
+}
+
+void MemoryBlockManager::UpdateLock(VAddr addr, std::size_t num_pages, LockFunc&& lock_func,
+                                    MemoryPermission perm) {
+    const VAddr end_addr{addr + num_pages * PageSize};
+    iterator node{memory_block_tree.begin()};
+
+    while (node != memory_block_tree.end()) {
+        MemoryBlock* block{&(*node)};
+        iterator next_node{std::next(node)};
+        const VAddr cur_addr{block->GetAddress()};
+        const VAddr cur_end_addr{block->GetNumPages() * PageSize + cur_addr};
+
+        if (addr < cur_end_addr && cur_addr < end_addr) {
+            iterator new_node{node};
+
+            if (addr > cur_addr) {
+                memory_block_tree.insert(node, block->Split(addr));
+            }
+
+            if (end_addr < cur_end_addr) {
+                new_node = memory_block_tree.insert(node, block->Split(end_addr));
+            }
+
+            lock_func(new_node, perm);
+
+            MergeAdjacent(new_node, next_node);
+        }
+
+        if (cur_end_addr - 1 >= end_addr - 1) {
+            break;
+        }
+
+        node = next_node;
+    }
+}
+
+void MemoryBlockManager::IterateForRange(VAddr start, VAddr end, IterateFunc&& func) {
+    const_iterator it{FindIterator(start)};
+    MemoryInfo info{};
+    do {
+        info = it->GetMemoryInfo();
+        func(info);
+        it = std::next(it);
+    } while (info.addr + info.size - 1 < end - 1 && it != cend());
+}
+
+void MemoryBlockManager::MergeAdjacent(iterator it, iterator& next_it) {
+    MemoryBlock* block{&(*it)};
+
+    auto EraseIt = [&](const iterator it_to_erase) {
+        if (next_it == it_to_erase) {
+            next_it = std::next(next_it);
+        }
+        memory_block_tree.erase(it_to_erase);
+    };
+
+    if (it != memory_block_tree.begin()) {
+        MemoryBlock* prev{&(*std::prev(it))};
+
+        if (block->HasSameProperties(*prev)) {
+            const iterator prev_it{std::prev(it)};
+
+            prev->Add(block->GetNumPages());
+            EraseIt(it);
+
+            it = prev_it;
+            block = prev;
+        }
+    }
+
+    if (it != cend()) {
+        const MemoryBlock* const next{&(*std::next(it))};
+
+        if (block->HasSameProperties(*next)) {
+            block->Add(next->GetNumPages());
+            EraseIt(std::next(it));
+        }
+    }
+}
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/memory_block_manager.h

@@ -0,0 +1,66 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <functional>
+#include <list>
+
+#include "common/common_types.h"
+#include "core/hle/kernel/memory/memory_block.h"
+
+namespace Kernel::Memory {
+
+class MemoryBlockManager final {
+public:
+    using MemoryBlockTree = std::list<MemoryBlock>;
+    using iterator = MemoryBlockTree::iterator;
+    using const_iterator = MemoryBlockTree::const_iterator;
+
+public:
+    MemoryBlockManager(VAddr start_addr, VAddr end_addr);
+
+    iterator end() {
+        return memory_block_tree.end();
+    }
+    const_iterator end() const {
+        return memory_block_tree.end();
+    }
+    const_iterator cend() const {
+        return memory_block_tree.cend();
+    }
+
+    iterator FindIterator(VAddr addr);
+
+    VAddr FindFreeArea(VAddr region_start, std::size_t region_num_pages, std::size_t num_pages,
+                       std::size_t align, std::size_t offset, std::size_t guard_pages);
+
+    void Update(VAddr addr, std::size_t num_pages, MemoryState prev_state,
+                MemoryPermission prev_perm, MemoryAttribute prev_attribute, MemoryState state,
+                MemoryPermission perm, MemoryAttribute attribute);
+
+    void Update(VAddr addr, std::size_t num_pages, MemoryState state,
+                MemoryPermission perm = MemoryPermission::None,
+                MemoryAttribute attribute = MemoryAttribute::None);
+
+    using LockFunc = std::function<void(iterator, MemoryPermission)>;
+    void UpdateLock(VAddr addr, std::size_t num_pages, LockFunc&& lock_func, MemoryPermission perm);
+
+    using IterateFunc = std::function<void(const MemoryInfo&)>;
+    void IterateForRange(VAddr start, VAddr end, IterateFunc&& func);
+
+    MemoryBlock& FindBlock(VAddr addr) {
+        return *FindIterator(addr);
+    }
+
+private:
+    void MergeAdjacent(iterator it, iterator& next_it);
+
+    [[maybe_unused]] const VAddr start_addr;
+    [[maybe_unused]] const VAddr end_addr;
+
+    MemoryBlockTree memory_block_tree;
+};
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/memory_layout.h

@@ -0,0 +1,71 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Kernel::Memory {
+
+class MemoryRegion final {
+    friend class MemoryLayout;
+
+public:
+    constexpr PAddr StartAddress() const {
+        return start_address;
+    }
+
+    constexpr PAddr EndAddress() const {
+        return end_address;
+    }
+
+private:
+    constexpr MemoryRegion() = default;
+    constexpr MemoryRegion(PAddr start_address, PAddr end_address)
+        : start_address{start_address}, end_address{end_address} {}
+
+    const PAddr start_address{};
+    const PAddr end_address{};
+};
+
+class MemoryLayout final {
+public:
+    constexpr const MemoryRegion& Application() const {
+        return application;
+    }
+
+    constexpr const MemoryRegion& Applet() const {
+        return applet;
+    }
+
+    constexpr const MemoryRegion& System() const {
+        return system;
+    }
+
+    static constexpr MemoryLayout GetDefaultLayout() {
+        constexpr std::size_t application_size{0xcd500000};
+        constexpr std::size_t applet_size{0x1fb00000};
+        constexpr PAddr application_start_address{Core::DramMemoryMap::End - application_size};
+        constexpr PAddr application_end_address{Core::DramMemoryMap::End};
+        constexpr PAddr applet_start_address{application_start_address - applet_size};
+        constexpr PAddr applet_end_address{applet_start_address + applet_size};
+        constexpr PAddr system_start_address{Core::DramMemoryMap::SlabHeapEnd};
+        constexpr PAddr system_end_address{applet_start_address};
+        return {application_start_address, application_end_address, applet_start_address,
+                applet_end_address,        system_start_address,    system_end_address};
+    }
+
+private:
+    constexpr MemoryLayout(PAddr application_start_address, std::size_t application_size,
+                           PAddr applet_start_address, std::size_t applet_size,
+                           PAddr system_start_address, std::size_t system_size)
+        : application{application_start_address, application_size},
+          applet{applet_start_address, applet_size}, system{system_start_address, system_size} {}
+
+    const MemoryRegion application;
+    const MemoryRegion applet;
+    const MemoryRegion system;
+};
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/memory_manager.cpp

@@ -0,0 +1,175 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+
+#include "common/alignment.h"
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "common/scope_exit.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/memory/memory_manager.h"
+#include "core/hle/kernel/memory/page_linked_list.h"
+
+namespace Kernel::Memory {
+
+std::size_t MemoryManager::Impl::Initialize(Pool new_pool, u64 start_address, u64 end_address) {
+    const auto size{end_address - start_address};
+
+    // Calculate metadata sizes
+    const auto ref_count_size{(size / PageSize) * sizeof(u16)};
+    const auto optimize_map_size{(Common::AlignUp((size / PageSize), 64) / 64) * sizeof(u64)};
+    const auto manager_size{Common::AlignUp(optimize_map_size + ref_count_size, PageSize)};
+    const auto page_heap_size{PageHeap::CalculateMetadataOverheadSize(size)};
+    const auto total_metadata_size{manager_size + page_heap_size};
+    ASSERT(manager_size <= total_metadata_size);
+    ASSERT(Common::IsAligned(total_metadata_size, PageSize));
+
+    // Setup region
+    pool = new_pool;
+
+    // Initialize the manager's KPageHeap
+    heap.Initialize(start_address, size, page_heap_size);
+
+    // Free the memory to the heap
+    heap.Free(start_address, size / PageSize);
+
+    // Update the heap's used size
+    heap.UpdateUsedSize();
+
+    return total_metadata_size;
+}
+
+void MemoryManager::InitializeManager(Pool pool, u64 start_address, u64 end_address) {
+    ASSERT(pool < Pool::Count);
+    managers[static_cast<std::size_t>(pool)].Initialize(pool, start_address, end_address);
+}
+
+VAddr MemoryManager::AllocateContinuous(std::size_t num_pages, std::size_t align_pages, Pool pool,
+                                        Direction dir) {
+    // Early return if we're allocating no pages
+    if (num_pages == 0) {
+        return {};
+    }
+
+    // Lock the pool that we're allocating from
+    const auto pool_index{static_cast<std::size_t>(pool)};
+    std::lock_guard lock{pool_locks[pool_index]};
+
+    // Choose a heap based on our page size request
+    const s32 heap_index{PageHeap::GetAlignedBlockIndex(num_pages, align_pages)};
+
+    // Loop, trying to iterate from each block
+    // TODO (bunnei): Support multiple managers
+    Impl& chosen_manager{managers[pool_index]};
+    VAddr allocated_block{chosen_manager.AllocateBlock(heap_index)};
+
+    // If we failed to allocate, quit now
+    if (!allocated_block) {
+        return {};
+    }
+
+    // If we allocated more than we need, free some
+    const auto allocated_pages{PageHeap::GetBlockNumPages(heap_index)};
+    if (allocated_pages > num_pages) {
+        chosen_manager.Free(allocated_block + num_pages * PageSize, allocated_pages - num_pages);
+    }
+
+    return allocated_block;
+}
+
+ResultCode MemoryManager::Allocate(PageLinkedList& page_list, std::size_t num_pages, Pool pool,
+                                   Direction dir) {
+    ASSERT(page_list.GetNumPages() == 0);
+
+    // Early return if we're allocating no pages
+    if (num_pages == 0) {
+        return RESULT_SUCCESS;
+    }
+
+    // Lock the pool that we're allocating from
+    const auto pool_index{static_cast<std::size_t>(pool)};
+    std::lock_guard lock{pool_locks[pool_index]};
+
+    // Choose a heap based on our page size request
+    const s32 heap_index{PageHeap::GetBlockIndex(num_pages)};
+    if (heap_index < 0) {
+        return ERR_OUT_OF_MEMORY;
+    }
+
+    // TODO (bunnei): Support multiple managers
+    Impl& chosen_manager{managers[pool_index]};
+
+    // Ensure that we don't leave anything un-freed
+    auto group_guard = detail::ScopeExit([&] {
+        for (const auto& it : page_list.Nodes()) {
+            const auto min_num_pages{std::min<size_t>(
+                it.GetNumPages(), (chosen_manager.GetEndAddress() - it.GetAddress()) / PageSize)};
+            chosen_manager.Free(it.GetAddress(), min_num_pages);
+        }
+    });
+
+    // Keep allocating until we've allocated all our pages
+    for (s32 index{heap_index}; index >= 0 && num_pages > 0; index--) {
+        const auto pages_per_alloc{PageHeap::GetBlockNumPages(index)};
+
+        while (num_pages >= pages_per_alloc) {
+            // Allocate a block
+            VAddr allocated_block{chosen_manager.AllocateBlock(index)};
+            if (!allocated_block) {
+                break;
+            }
+
+            // Safely add it to our group
+            {
+                auto block_guard = detail::ScopeExit(
+                    [&] { chosen_manager.Free(allocated_block, pages_per_alloc); });
+
+                if (const ResultCode result{page_list.AddBlock(allocated_block, pages_per_alloc)};
+                    result.IsError()) {
+                    return result;
+                }
+
+                block_guard.Cancel();
+            }
+
+            num_pages -= pages_per_alloc;
+        }
+    }
+
+    // Only succeed if we allocated as many pages as we wanted
+    if (num_pages) {
+        return ERR_OUT_OF_MEMORY;
+    }
+
+    // We succeeded!
+    group_guard.Cancel();
+    return RESULT_SUCCESS;
+}
+
+ResultCode MemoryManager::Free(PageLinkedList& page_list, std::size_t num_pages, Pool pool,
+                               Direction dir) {
+    // Early return if we're freeing no pages
+    if (!num_pages) {
+        return RESULT_SUCCESS;
+    }
+
+    // Lock the pool that we're freeing from
+    const auto pool_index{static_cast<std::size_t>(pool)};
+    std::lock_guard lock{pool_locks[pool_index]};
+
+    // TODO (bunnei): Support multiple managers
+    Impl& chosen_manager{managers[pool_index]};
+
+    // Free all of the pages
+    for (const auto& it : page_list.Nodes()) {
+        const auto min_num_pages{std::min<size_t>(
+            it.GetNumPages(), (chosen_manager.GetEndAddress() - it.GetAddress()) / PageSize)};
+        chosen_manager.Free(it.GetAddress(), min_num_pages);
+    }
+
+    return RESULT_SUCCESS;
+}
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/memory_manager.h

@@ -0,0 +1,96 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <mutex>
+
+#include "common/common_types.h"
+#include "core/hle/kernel/memory/page_heap.h"
+#include "core/hle/result.h"
+
+namespace Kernel::Memory {
+
+class PageLinkedList;
+
+class MemoryManager final : NonCopyable {
+public:
+    enum class Pool : u32 {
+        Application = 0,
+        Applet = 1,
+        System = 2,
+        SystemNonSecure = 3,
+
+        Count,
+
+        Shift = 4,
+        Mask = (0xF << Shift),
+    };
+
+    enum class Direction : u32 {
+        FromFront = 0,
+        FromBack = 1,
+
+        Shift = 0,
+        Mask = (0xF << Shift),
+    };
+
+    MemoryManager() = default;
+
+    constexpr std::size_t GetSize(Pool pool) const {
+        return managers[static_cast<std::size_t>(pool)].GetSize();
+    }
+
+    void InitializeManager(Pool pool, u64 start_address, u64 end_address);
+    VAddr AllocateContinuous(std::size_t num_pages, std::size_t align_pages, Pool pool,
+                             Direction dir = Direction::FromFront);
+    ResultCode Allocate(PageLinkedList& page_list, std::size_t num_pages, Pool pool,
+                        Direction dir = Direction::FromFront);
+    ResultCode Free(PageLinkedList& page_list, std::size_t num_pages, Pool pool,
+                    Direction dir = Direction::FromFront);
+
+    static constexpr std::size_t MaxManagerCount = 10;
+
+private:
+    class Impl final : NonCopyable {
+    private:
+        using RefCount = u16;
+
+    private:
+        PageHeap heap;
+        Pool pool{};
+
+    public:
+        Impl() = default;
+
+        std::size_t Initialize(Pool new_pool, u64 start_address, u64 end_address);
+
+        VAddr AllocateBlock(s32 index) {
+            return heap.AllocateBlock(index);
+        }
+
+        void Free(VAddr addr, std::size_t num_pages) {
+            heap.Free(addr, num_pages);
+        }
+
+        constexpr std::size_t GetSize() const {
+            return heap.GetSize();
+        }
+
+        constexpr VAddr GetAddress() const {
+            return heap.GetAddress();
+        }
+
+        constexpr VAddr GetEndAddress() const {
+            return heap.GetEndAddress();
+        }
+    };
+
+private:
+    std::array<std::mutex, static_cast<std::size_t>(Pool::Count)> pool_locks;
+    std::array<Impl, MaxManagerCount> managers;
+};
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/memory_types.h

@@ -0,0 +1,18 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+
+#include "common/common_types.h"
+
+namespace Kernel::Memory {
+
+constexpr std::size_t PageBits{12};
+constexpr std::size_t PageSize{1 << PageBits};
+
+using Page = std::array<u8, PageSize>;
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/page_heap.cpp

@@ -0,0 +1,119 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#include "core/core.h"
+#include "core/hle/kernel/memory/page_heap.h"
+#include "core/memory.h"
+
+namespace Kernel::Memory {
+
+void PageHeap::Initialize(VAddr address, std::size_t size, std::size_t metadata_size) {
+    // Check our assumptions
+    ASSERT(Common::IsAligned((address), PageSize));
+    ASSERT(Common::IsAligned(size, PageSize));
+
+    // Set our members
+    heap_address = address;
+    heap_size = size;
+
+    // Setup bitmaps
+    metadata.resize(metadata_size / sizeof(u64));
+    u64* cur_bitmap_storage{metadata.data()};
+    for (std::size_t i = 0; i < MemoryBlockPageShifts.size(); i++) {
+        const std::size_t cur_block_shift{MemoryBlockPageShifts[i]};
+        const std::size_t next_block_shift{
+            (i != MemoryBlockPageShifts.size() - 1) ? MemoryBlockPageShifts[i + 1] : 0};
+        cur_bitmap_storage = blocks[i].Initialize(heap_address, heap_size, cur_block_shift,
+                                                  next_block_shift, cur_bitmap_storage);
+    }
+}
+
+VAddr PageHeap::AllocateBlock(s32 index) {
+    const std::size_t needed_size{blocks[index].GetSize()};
+
+    for (s32 i{index}; i < static_cast<s32>(MemoryBlockPageShifts.size()); i++) {
+        if (const VAddr addr{blocks[i].PopBlock()}; addr) {
+            if (const std::size_t allocated_size{blocks[i].GetSize()};
+                allocated_size > needed_size) {
+                Free(addr + needed_size, (allocated_size - needed_size) / PageSize);
+            }
+            return addr;
+        }
+    }
+
+    return 0;
+}
+
+void PageHeap::FreeBlock(VAddr block, s32 index) {
+    do {
+        block = blocks[index++].PushBlock(block);
+    } while (block != 0);
+}
+
+void PageHeap::Free(VAddr addr, std::size_t num_pages) {
+    // Freeing no pages is a no-op
+    if (num_pages == 0) {
+        return;
+    }
+
+    // Find the largest block size that we can free, and free as many as possible
+    s32 big_index{static_cast<s32>(MemoryBlockPageShifts.size()) - 1};
+    const VAddr start{addr};
+    const VAddr end{(num_pages * PageSize) + addr};
+    VAddr before_start{start};
+    VAddr before_end{start};
+    VAddr after_start{end};
+    VAddr after_end{end};
+    while (big_index >= 0) {
+        const std::size_t block_size{blocks[big_index].GetSize()};
+        const VAddr big_start{Common::AlignUp((start), block_size)};
+        const VAddr big_end{Common::AlignDown((end), block_size)};
+        if (big_start < big_end) {
+            // Free as many big blocks as we can
+            for (auto block{big_start}; block < big_end; block += block_size) {
+                FreeBlock(block, big_index);
+            }
+            before_end = big_start;
+            after_start = big_end;
+            break;
+        }
+        big_index--;
+    }
+    ASSERT(big_index >= 0);
+
+    // Free space before the big blocks
+    for (s32 i{big_index - 1}; i >= 0; i--) {
+        const std::size_t block_size{blocks[i].GetSize()};
+        while (before_start + block_size <= before_end) {
+            before_end -= block_size;
+            FreeBlock(before_end, i);
+        }
+    }
+
+    // Free space after the big blocks
+    for (s32 i{big_index - 1}; i >= 0; i--) {
+        const std::size_t block_size{blocks[i].GetSize()};
+        while (after_start + block_size <= after_end) {
+            FreeBlock(after_start, i);
+            after_start += block_size;
+        }
+    }
+}
+
+std::size_t PageHeap::CalculateMetadataOverheadSize(std::size_t region_size) {
+    std::size_t overhead_size = 0;
+    for (std::size_t i = 0; i < MemoryBlockPageShifts.size(); i++) {
+        const std::size_t cur_block_shift{MemoryBlockPageShifts[i]};
+        const std::size_t next_block_shift{
+            (i != MemoryBlockPageShifts.size() - 1) ? MemoryBlockPageShifts[i + 1] : 0};
+        overhead_size += PageHeap::Block::CalculateMetadataOverheadSize(
+            region_size, cur_block_shift, next_block_shift);
+    }
+    return Common::AlignUp(overhead_size, PageSize);
+}
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/page_heap.h

@@ -0,0 +1,370 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include <array>
+#include <vector>
+
+#include "common/alignment.h"
+#include "common/assert.h"
+#include "common/bit_util.h"
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+#include "core/hle/kernel/memory/memory_types.h"
+
+namespace Kernel::Memory {
+
+class PageHeap final : NonCopyable {
+public:
+    static constexpr s32 GetAlignedBlockIndex(std::size_t num_pages, std::size_t align_pages) {
+        const auto target_pages{std::max(num_pages, align_pages)};
+        for (std::size_t i = 0; i < NumMemoryBlockPageShifts; i++) {
+            if (target_pages <=
+                (static_cast<std::size_t>(1) << MemoryBlockPageShifts[i]) / PageSize) {
+                return static_cast<s32>(i);
+            }
+        }
+        return -1;
+    }
+
+    static constexpr s32 GetBlockIndex(std::size_t num_pages) {
+        for (s32 i{static_cast<s32>(NumMemoryBlockPageShifts) - 1}; i >= 0; i--) {
+            if (num_pages >= (static_cast<std::size_t>(1) << MemoryBlockPageShifts[i]) / PageSize) {
+                return i;
+            }
+        }
+        return -1;
+    }
+
+    static constexpr std::size_t GetBlockSize(std::size_t index) {
+        return static_cast<std::size_t>(1) << MemoryBlockPageShifts[index];
+    }
+
+    static constexpr std::size_t GetBlockNumPages(std::size_t index) {
+        return GetBlockSize(index) / PageSize;
+    }
+
+private:
+    static constexpr std::size_t NumMemoryBlockPageShifts{7};
+    static constexpr std::array<std::size_t, NumMemoryBlockPageShifts> MemoryBlockPageShifts{
+        0xC, 0x10, 0x15, 0x16, 0x19, 0x1D, 0x1E,
+    };
+
+    class Block final : NonCopyable {
+    private:
+        class Bitmap final : NonCopyable {
+        public:
+            static constexpr std::size_t MaxDepth{4};
+
+        private:
+            std::array<u64*, MaxDepth> bit_storages{};
+            std::size_t num_bits{};
+            std::size_t used_depths{};
+
+        public:
+            constexpr Bitmap() = default;
+
+            constexpr std::size_t GetNumBits() const {
+                return num_bits;
+            }
+            constexpr s32 GetHighestDepthIndex() const {
+                return static_cast<s32>(used_depths) - 1;
+            }
+
+            constexpr u64* Initialize(u64* storage, std::size_t size) {
+                //* Initially, everything is un-set
+                num_bits = 0;
+
+                // Calculate the needed bitmap depth
+                used_depths = static_cast<std::size_t>(GetRequiredDepth(size));
+                ASSERT(used_depths <= MaxDepth);
+
+                // Set the bitmap pointers
+                for (s32 depth{GetHighestDepthIndex()}; depth >= 0; depth--) {
+                    bit_storages[depth] = storage;
+                    size = Common::AlignUp(size, 64) / 64;
+                    storage += size;
+                }
+
+                return storage;
+            }
+
+            s64 FindFreeBlock() const {
+                uintptr_t offset{};
+                s32 depth{};
+
+                do {
+                    const u64 v{bit_storages[depth][offset]};
+                    if (v == 0) {
+                        // Non-zero depth indicates that a previous level had a free block
+                        ASSERT(depth == 0);
+                        return -1;
+                    }
+                    offset = offset * 64 + Common::CountTrailingZeroes64(v);
+                    ++depth;
+                } while (depth < static_cast<s32>(used_depths));
+
+                return static_cast<s64>(offset);
+            }
+
+            constexpr void SetBit(std::size_t offset) {
+                SetBit(GetHighestDepthIndex(), offset);
+                num_bits++;
+            }
+
+            constexpr void ClearBit(std::size_t offset) {
+                ClearBit(GetHighestDepthIndex(), offset);
+                num_bits--;
+            }
+
+            constexpr bool ClearRange(std::size_t offset, std::size_t count) {
+                const s32 depth{GetHighestDepthIndex()};
+                const auto bit_ind{offset / 64};
+                u64* bits{bit_storages[depth]};
+                if (count < 64) {
+                    const auto shift{offset % 64};
+                    ASSERT(shift + count <= 64);
+                    // Check that all the bits are set
+                    const u64 mask{((1ULL << count) - 1) << shift};
+                    u64 v{bits[bit_ind]};
+                    if ((v & mask) != mask) {
+                        return false;
+                    }
+
+                    // Clear the bits
+                    v &= ~mask;
+                    bits[bit_ind] = v;
+                    if (v == 0) {
+                        ClearBit(depth - 1, bit_ind);
+                    }
+                } else {
+                    ASSERT(offset % 64 == 0);
+                    ASSERT(count % 64 == 0);
+                    // Check that all the bits are set
+                    std::size_t remaining{count};
+                    std::size_t i = 0;
+                    do {
+                        if (bits[bit_ind + i++] != ~u64(0)) {
+                            return false;
+                        }
+                        remaining -= 64;
+                    } while (remaining > 0);
+
+                    // Clear the bits
+                    remaining = count;
+                    i = 0;
+                    do {
+                        bits[bit_ind + i] = 0;
+                        ClearBit(depth - 1, bit_ind + i);
+                        i++;
+                        remaining -= 64;
+                    } while (remaining > 0);
+                }
+
+                num_bits -= count;
+                return true;
+            }
+
+        private:
+            constexpr void SetBit(s32 depth, std::size_t offset) {
+                while (depth >= 0) {
+                    const auto ind{offset / 64};
+                    const auto which{offset % 64};
+                    const u64 mask{1ULL << which};
+
+                    u64* bit{std::addressof(bit_storages[depth][ind])};
+                    const u64 v{*bit};
+                    ASSERT((v & mask) == 0);
+                    *bit = v | mask;
+                    if (v) {
+                        break;
+                    }
+                    offset = ind;
+                    depth--;
+                }
+            }
+
+            constexpr void ClearBit(s32 depth, std::size_t offset) {
+                while (depth >= 0) {
+                    const auto ind{offset / 64};
+                    const auto which{offset % 64};
+                    const u64 mask{1ULL << which};
+
+                    u64* bit{std::addressof(bit_storages[depth][ind])};
+                    u64 v{*bit};
+                    ASSERT((v & mask) != 0);
+                    v &= ~mask;
+                    *bit = v;
+                    if (v) {
+                        break;
+                    }
+                    offset = ind;
+                    depth--;
+                }
+            }
+
+        private:
+            static constexpr s32 GetRequiredDepth(std::size_t region_size) {
+                s32 depth = 0;
+                while (true) {
+                    region_size /= 64;
+                    depth++;
+                    if (region_size == 0) {
+                        return depth;
+                    }
+                }
+            }
+
+        public:
+            static constexpr std::size_t CalculateMetadataOverheadSize(std::size_t region_size) {
+                std::size_t overhead_bits = 0;
+                for (s32 depth{GetRequiredDepth(region_size) - 1}; depth >= 0; depth--) {
+                    region_size = Common::AlignUp(region_size, 64) / 64;
+                    overhead_bits += region_size;
+                }
+                return overhead_bits * sizeof(u64);
+            }
+        };
+
+    private:
+        Bitmap bitmap;
+        VAddr heap_address{};
+        uintptr_t end_offset{};
+        std::size_t block_shift{};
+        std::size_t next_block_shift{};
+
+    public:
+        constexpr Block() = default;
+
+        constexpr std::size_t GetShift() const {
+            return block_shift;
+        }
+        constexpr std::size_t GetNextShift() const {
+            return next_block_shift;
+        }
+        constexpr std::size_t GetSize() const {
+            return static_cast<std::size_t>(1) << GetShift();
+        }
+        constexpr std::size_t GetNumPages() const {
+            return GetSize() / PageSize;
+        }
+        constexpr std::size_t GetNumFreeBlocks() const {
+            return bitmap.GetNumBits();
+        }
+        constexpr std::size_t GetNumFreePages() const {
+            return GetNumFreeBlocks() * GetNumPages();
+        }
+
+        constexpr u64* Initialize(VAddr addr, std::size_t size, std::size_t bs, std::size_t nbs,
+                                  u64* bit_storage) {
+            // Set shifts
+            block_shift = bs;
+            next_block_shift = nbs;
+
+            // Align up the address
+            VAddr end{addr + size};
+            const auto align{(next_block_shift != 0) ? (1ULL << next_block_shift)
+                                                     : (1ULL << block_shift)};
+            addr = Common::AlignDown((addr), align);
+            end = Common::AlignUp((end), align);
+
+            heap_address = addr;
+            end_offset = (end - addr) / (1ULL << block_shift);
+            return bitmap.Initialize(bit_storage, end_offset);
+        }
+
+        constexpr VAddr PushBlock(VAddr address) {
+            // Set the bit for the free block
+            std::size_t offset{(address - heap_address) >> GetShift()};
+            bitmap.SetBit(offset);
+
+            // If we have a next shift, try to clear the blocks below and return the address
+            if (GetNextShift()) {
+                const auto diff{1ULL << (GetNextShift() - GetShift())};
+                offset = Common::AlignDown(offset, diff);
+                if (bitmap.ClearRange(offset, diff)) {
+                    return heap_address + (offset << GetShift());
+                }
+            }
+
+            // We couldn't coalesce, or we're already as big as possible
+            return 0;
+        }
+
+        VAddr PopBlock() {
+            // Find a free block
+            const s64 soffset{bitmap.FindFreeBlock()};
+            if (soffset < 0) {
+                return 0;
+            }
+            const auto offset{static_cast<std::size_t>(soffset)};
+
+            // Update our tracking and return it
+            bitmap.ClearBit(offset);
+            return heap_address + (offset << GetShift());
+        }
+
+    public:
+        static constexpr std::size_t CalculateMetadataOverheadSize(std::size_t region_size,
+                                                                   std::size_t cur_block_shift,
+                                                                   std::size_t next_block_shift) {
+            const auto cur_block_size{(1ULL << cur_block_shift)};
+            const auto next_block_size{(1ULL << next_block_shift)};
+            const auto align{(next_block_shift != 0) ? next_block_size : cur_block_size};
+            return Bitmap::CalculateMetadataOverheadSize(
+                (align * 2 + Common::AlignUp(region_size, align)) / cur_block_size);
+        }
+    };
+
+public:
+    PageHeap() = default;
+
+    constexpr VAddr GetAddress() const {
+        return heap_address;
+    }
+    constexpr std::size_t GetSize() const {
+        return heap_size;
+    }
+    constexpr VAddr GetEndAddress() const {
+        return GetAddress() + GetSize();
+    }
+    constexpr std::size_t GetPageOffset(VAddr block) const {
+        return (block - GetAddress()) / PageSize;
+    }
+
+    void Initialize(VAddr heap_address, std::size_t heap_size, std::size_t metadata_size);
+    VAddr AllocateBlock(s32 index);
+    void Free(VAddr addr, std::size_t num_pages);
+
+    void UpdateUsedSize() {
+        used_size = heap_size - (GetNumFreePages() * PageSize);
+    }
+
+    static std::size_t CalculateMetadataOverheadSize(std::size_t region_size);
+
+private:
+    constexpr std::size_t GetNumFreePages() const {
+        std::size_t num_free{};
+
+        for (const auto& block : blocks) {
+            num_free += block.GetNumFreePages();
+        }
+
+        return num_free;
+    }
+
+    void FreeBlock(VAddr block, s32 index);
+
+    VAddr heap_address{};
+    std::size_t heap_size{};
+    std::size_t used_size{};
+    std::array<Block, NumMemoryBlockPageShifts> blocks{};
+    std::vector<u64> metadata;
+};
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/page_linked_list.h

@@ -0,0 +1,92 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <list>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "core/hle/kernel/memory/memory_types.h"
+#include "core/hle/result.h"
+
+namespace Kernel::Memory {
+
+class PageLinkedList final {
+public:
+    class Node final {
+    public:
+        constexpr Node(u64 addr, std::size_t num_pages) : addr{addr}, num_pages{num_pages} {}
+
+        constexpr u64 GetAddress() const {
+            return addr;
+        }
+
+        constexpr std::size_t GetNumPages() const {
+            return num_pages;
+        }
+
+    private:
+        u64 addr{};
+        std::size_t num_pages{};
+    };
+
+public:
+    PageLinkedList() = default;
+    PageLinkedList(u64 address, u64 num_pages) {
+        ASSERT(AddBlock(address, num_pages).IsSuccess());
+    }
+
+    constexpr std::list<Node>& Nodes() {
+        return nodes;
+    }
+
+    constexpr const std::list<Node>& Nodes() const {
+        return nodes;
+    }
+
+    std::size_t GetNumPages() const {
+        std::size_t num_pages = 0;
+        for (const Node& node : nodes) {
+            num_pages += node.GetNumPages();
+        }
+        return num_pages;
+    }
+
+    bool IsEqual(PageLinkedList& other) const {
+        auto this_node = nodes.begin();
+        auto other_node = other.nodes.begin();
+        while (this_node != nodes.end() && other_node != other.nodes.end()) {
+            if (this_node->GetAddress() != other_node->GetAddress() ||
+                this_node->GetNumPages() != other_node->GetNumPages()) {
+                return false;
+            }
+            this_node = std::next(this_node);
+            other_node = std::next(other_node);
+        }
+
+        return this_node == nodes.end() && other_node == other.nodes.end();
+    }
+
+    ResultCode AddBlock(u64 address, u64 num_pages) {
+        if (!num_pages) {
+            return RESULT_SUCCESS;
+        }
+        if (!nodes.empty()) {
+            const auto node = nodes.back();
+            if (node.GetAddress() + node.GetNumPages() * PageSize == address) {
+                address = node.GetAddress();
+                num_pages += node.GetNumPages();
+                nodes.pop_back();
+            }
+        }
+        nodes.push_back({address, num_pages});
+        return RESULT_SUCCESS;
+    }
+
+private:
+    std::list<Node> nodes;
+};
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/page_table.h

@@ -0,0 +1,277 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <mutex>
+
+#include "common/common_types.h"
+#include "common/page_table.h"
+#include "core/file_sys/program_metadata.h"
+#include "core/hle/kernel/memory/memory_block.h"
+#include "core/hle/kernel/memory/memory_manager.h"
+#include "core/hle/result.h"
+
+namespace Core {
+class System;
+}
+
+namespace Kernel::Memory {
+
+class MemoryBlockManager;
+
+class PageTable final : NonCopyable {
+public:
+    explicit PageTable(Core::System& system);
+
+    ResultCode InitializeForProcess(FileSys::ProgramAddressSpaceType as_type, bool enable_aslr,
+                                    VAddr code_addr, std::size_t code_size,
+                                    Memory::MemoryManager::Pool pool);
+    ResultCode MapProcessCode(VAddr addr, std::size_t pages_count, MemoryState state,
+                              MemoryPermission perm);
+    ResultCode MapProcessCodeMemory(VAddr dst_addr, VAddr src_addr, std::size_t size);
+    ResultCode UnmapProcessCodeMemory(VAddr dst_addr, VAddr src_addr, std::size_t size);
+    ResultCode MapPhysicalMemory(VAddr addr, std::size_t size);
+    ResultCode UnmapPhysicalMemory(VAddr addr, std::size_t size);
+    ResultCode UnmapMemory(VAddr addr, std::size_t size);
+    ResultCode Map(VAddr dst_addr, VAddr src_addr, std::size_t size);
+    ResultCode Unmap(VAddr dst_addr, VAddr src_addr, std::size_t size);
+    ResultCode MapPages(VAddr addr, PageLinkedList& page_linked_list, MemoryState state,
+                        MemoryPermission perm);
+    ResultCode SetCodeMemoryPermission(VAddr addr, std::size_t size, MemoryPermission perm);
+    MemoryInfo QueryInfo(VAddr addr);
+    ResultCode ReserveTransferMemory(VAddr addr, std::size_t size, MemoryPermission perm);
+    ResultCode ResetTransferMemory(VAddr addr, std::size_t size);
+    ResultCode SetMemoryAttribute(VAddr addr, std::size_t size, MemoryAttribute mask,
+                                  MemoryAttribute value);
+    ResultCode SetHeapCapacity(std::size_t new_heap_capacity);
+    ResultVal<VAddr> SetHeapSize(std::size_t size);
+    ResultVal<VAddr> AllocateAndMapMemory(std::size_t needed_num_pages, std::size_t align,
+                                          bool is_map_only, VAddr region_start,
+                                          std::size_t region_num_pages, MemoryState state,
+                                          MemoryPermission perm, PAddr map_addr = 0);
+    ResultCode LockForDeviceAddressSpace(VAddr addr, std::size_t size);
+    ResultCode UnlockForDeviceAddressSpace(VAddr addr, std::size_t size);
+
+    Common::PageTable& PageTableImpl() {
+        return page_table_impl;
+    }
+
+    const Common::PageTable& PageTableImpl() const {
+        return page_table_impl;
+    }
+
+private:
+    enum class OperationType : u32 {
+        Map,
+        MapGroup,
+        Unmap,
+        ChangePermissions,
+        ChangePermissionsAndRefresh,
+    };
+
+    static constexpr MemoryAttribute DefaultMemoryIgnoreAttr =
+        MemoryAttribute::DontCareMask | MemoryAttribute::IpcLocked | MemoryAttribute::DeviceShared;
+
+    ResultCode InitializeMemoryLayout(VAddr start, VAddr end);
+    ResultCode MapPages(VAddr addr, const PageLinkedList& page_linked_list, MemoryPermission perm);
+    void MapPhysicalMemory(PageLinkedList& page_linked_list, VAddr start, VAddr end);
+    bool IsRegionMapped(VAddr address, u64 size);
+    bool IsRegionContiguous(VAddr addr, u64 size) const;
+    void AddRegionToPages(VAddr start, std::size_t num_pages, PageLinkedList& page_linked_list);
+    MemoryInfo QueryInfoImpl(VAddr addr);
+    VAddr AllocateVirtualMemory(VAddr start, std::size_t region_num_pages, u64 needed_num_pages,
+                                std::size_t align);
+    ResultCode Operate(VAddr addr, std::size_t num_pages, const PageLinkedList& page_group,
+                       OperationType operation);
+    ResultCode Operate(VAddr addr, std::size_t num_pages, MemoryPermission perm,
+                       OperationType operation, PAddr map_addr = 0);
+    constexpr VAddr GetRegionAddress(MemoryState state) const;
+    constexpr std::size_t GetRegionSize(MemoryState state) const;
+    constexpr bool CanContain(VAddr addr, std::size_t size, MemoryState state) const;
+
+    constexpr ResultCode CheckMemoryState(const MemoryInfo& info, MemoryState state_mask,
+                                          MemoryState state, MemoryPermission perm_mask,
+                                          MemoryPermission perm, MemoryAttribute attr_mask,
+                                          MemoryAttribute attr) const;
+    ResultCode CheckMemoryState(MemoryState* out_state, MemoryPermission* out_perm,
+                                MemoryAttribute* out_attr, VAddr addr, std::size_t size,
+                                MemoryState state_mask, MemoryState state,
+                                MemoryPermission perm_mask, MemoryPermission perm,
+                                MemoryAttribute attr_mask, MemoryAttribute attr,
+                                MemoryAttribute ignore_attr = DefaultMemoryIgnoreAttr);
+    ResultCode CheckMemoryState(VAddr addr, std::size_t size, MemoryState state_mask,
+                                MemoryState state, MemoryPermission perm_mask,
+                                MemoryPermission perm, MemoryAttribute attr_mask,
+                                MemoryAttribute attr,
+                                MemoryAttribute ignore_attr = DefaultMemoryIgnoreAttr) {
+        return CheckMemoryState(nullptr, nullptr, nullptr, addr, size, state_mask, state, perm_mask,
+                                perm, attr_mask, attr, ignore_attr);
+    }
+
+    std::recursive_mutex page_table_lock;
+    std::unique_ptr<MemoryBlockManager> block_manager;
+
+public:
+    constexpr VAddr GetAddressSpaceStart() const {
+        return address_space_start;
+    }
+    constexpr VAddr GetAddressSpaceEnd() const {
+        return address_space_end;
+    }
+    constexpr std::size_t GetAddressSpaceSize() const {
+        return address_space_end - address_space_start;
+    }
+    constexpr VAddr GetHeapRegionStart() const {
+        return heap_region_start;
+    }
+    constexpr VAddr GetHeapRegionEnd() const {
+        return heap_region_end;
+    }
+    constexpr std::size_t GetHeapRegionSize() const {
+        return heap_region_end - heap_region_start;
+    }
+    constexpr VAddr GetAliasRegionStart() const {
+        return alias_region_start;
+    }
+    constexpr VAddr GetAliasRegionEnd() const {
+        return alias_region_end;
+    }
+    constexpr std::size_t GetAliasRegionSize() const {
+        return alias_region_end - alias_region_start;
+    }
+    constexpr VAddr GetStackRegionStart() const {
+        return stack_region_start;
+    }
+    constexpr VAddr GetStackRegionEnd() const {
+        return stack_region_end;
+    }
+    constexpr std::size_t GetStackRegionSize() const {
+        return stack_region_end - stack_region_start;
+    }
+    constexpr VAddr GetKernelMapRegionStart() const {
+        return kernel_map_region_start;
+    }
+    constexpr VAddr GetKernelMapRegionEnd() const {
+        return kernel_map_region_end;
+    }
+    constexpr VAddr GetCodeRegionStart() const {
+        return code_region_start;
+    }
+    constexpr VAddr GetCodeRegionEnd() const {
+        return code_region_end;
+    }
+    constexpr VAddr GetAliasCodeRegionStart() const {
+        return alias_code_region_start;
+    }
+    constexpr VAddr GetAliasCodeRegionSize() const {
+        return alias_code_region_end - alias_code_region_start;
+    }
+    constexpr std::size_t GetAddressSpaceWidth() const {
+        return address_space_width;
+    }
+    constexpr std::size_t GetHeapSize() {
+        return current_heap_addr - heap_region_start;
+    }
+    constexpr std::size_t GetTotalHeapSize() {
+        return GetHeapSize() + physical_memory_usage;
+    }
+    constexpr bool IsInsideAddressSpace(VAddr address, std::size_t size) const {
+        return address_space_start <= address && address + size - 1 <= address_space_end - 1;
+    }
+    constexpr bool IsOutsideAliasRegion(VAddr address, std::size_t size) const {
+        return alias_region_start > address || address + size - 1 > alias_region_end - 1;
+    }
+    constexpr bool IsOutsideStackRegion(VAddr address, std::size_t size) const {
+        return stack_region_start > address || address + size - 1 > stack_region_end - 1;
+    }
+    constexpr bool IsInvalidRegion(VAddr address, std::size_t size) const {
+        return address + size - 1 > GetAliasCodeRegionStart() + GetAliasCodeRegionSize() - 1;
+    }
+    constexpr bool IsInsideHeapRegion(VAddr address, std::size_t size) const {
+        return address + size > heap_region_start && heap_region_end > address;
+    }
+    constexpr bool IsInsideAliasRegion(VAddr address, std::size_t size) const {
+        return address + size > alias_region_start && alias_region_end > address;
+    }
+    constexpr bool IsOutsideASLRRegion(VAddr address, std::size_t size) const {
+        if (IsInvalidRegion(address, size)) {
+            return true;
+        }
+        if (IsInsideHeapRegion(address, size)) {
+            return true;
+        }
+        if (IsInsideAliasRegion(address, size)) {
+            return true;
+        }
+        return {};
+    }
+    constexpr bool IsInsideASLRRegion(VAddr address, std::size_t size) const {
+        return !IsOutsideASLRRegion(address, size);
+    }
+    constexpr PAddr GetPhysicalAddr(VAddr addr) {
+        return page_table_impl.backing_addr[addr >> Memory::PageBits] + addr;
+    }
+
+private:
+    constexpr bool Contains(VAddr addr) const {
+        return address_space_start <= addr && addr <= address_space_end - 1;
+    }
+    constexpr bool Contains(VAddr addr, std::size_t size) const {
+        return address_space_start <= addr && addr < addr + size &&
+               addr + size - 1 <= address_space_end - 1;
+    }
+    constexpr bool IsKernel() const {
+        return is_kernel;
+    }
+    constexpr bool IsAslrEnabled() const {
+        return is_aslr_enabled;
+    }
+
+    constexpr std::size_t GetNumGuardPages() const {
+        return IsKernel() ? 1 : 4;
+    }
+
+    constexpr bool ContainsPages(VAddr addr, std::size_t num_pages) const {
+        return (address_space_start <= addr) &&
+               (num_pages <= (address_space_end - address_space_start) / PageSize) &&
+               (addr + num_pages * PageSize - 1 <= address_space_end - 1);
+    }
+
+private:
+    VAddr address_space_start{};
+    VAddr address_space_end{};
+    VAddr heap_region_start{};
+    VAddr heap_region_end{};
+    VAddr current_heap_end{};
+    VAddr alias_region_start{};
+    VAddr alias_region_end{};
+    VAddr stack_region_start{};
+    VAddr stack_region_end{};
+    VAddr kernel_map_region_start{};
+    VAddr kernel_map_region_end{};
+    VAddr code_region_start{};
+    VAddr code_region_end{};
+    VAddr alias_code_region_start{};
+    VAddr alias_code_region_end{};
+    VAddr current_heap_addr{};
+
+    std::size_t heap_capacity{};
+    std::size_t physical_memory_usage{};
+    std::size_t max_heap_size{};
+    std::size_t max_physical_memory_size{};
+    std::size_t address_space_width{};
+
+    bool is_kernel{};
+    bool is_aslr_enabled{};
+
+    MemoryManager::Pool memory_pool{MemoryManager::Pool::Application};
+
+    Common::PageTable page_table_impl;
+
+    Core::System& system;
+};
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/slab_heap.h

@@ -0,0 +1,163 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include <atomic>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+
+namespace Kernel::Memory {
+
+namespace impl {
+
+class SlabHeapImpl final : NonCopyable {
+public:
+    struct Node {
+        Node* next{};
+    };
+
+    constexpr SlabHeapImpl() = default;
+
+    void Initialize(std::size_t size) {
+        ASSERT(head == nullptr);
+        obj_size = size;
+    }
+
+    constexpr std::size_t GetObjectSize() const {
+        return obj_size;
+    }
+
+    Node* GetHead() const {
+        return head;
+    }
+
+    void* Allocate() {
+        Node* ret = head.load();
+
+        do {
+            if (ret == nullptr) {
+                break;
+            }
+        } while (!head.compare_exchange_weak(ret, ret->next));
+
+        return ret;
+    }
+
+    void Free(void* obj) {
+        Node* node = static_cast<Node*>(obj);
+
+        Node* cur_head = head.load();
+        do {
+            node->next = cur_head;
+        } while (!head.compare_exchange_weak(cur_head, node));
+    }
+
+private:
+    std::atomic<Node*> head{};
+    std::size_t obj_size{};
+};
+
+} // namespace impl
+
+class SlabHeapBase : NonCopyable {
+public:
+    constexpr SlabHeapBase() = default;
+
+    constexpr bool Contains(uintptr_t addr) const {
+        return start <= addr && addr < end;
+    }
+
+    constexpr std::size_t GetSlabHeapSize() const {
+        return (end - start) / GetObjectSize();
+    }
+
+    constexpr std::size_t GetObjectSize() const {
+        return impl.GetObjectSize();
+    }
+
+    constexpr uintptr_t GetSlabHeapAddress() const {
+        return start;
+    }
+
+    std::size_t GetObjectIndexImpl(const void* obj) const {
+        return (reinterpret_cast<uintptr_t>(obj) - start) / GetObjectSize();
+    }
+
+    std::size_t GetPeakIndex() const {
+        return GetObjectIndexImpl(reinterpret_cast<const void*>(peak));
+    }
+
+    void* AllocateImpl() {
+        return impl.Allocate();
+    }
+
+    void FreeImpl(void* obj) {
+        // Don't allow freeing an object that wasn't allocated from this heap
+        ASSERT(Contains(reinterpret_cast<uintptr_t>(obj)));
+        impl.Free(obj);
+    }
+
+    void InitializeImpl(std::size_t obj_size, void* memory, std::size_t memory_size) {
+        // Ensure we don't initialize a slab using null memory
+        ASSERT(memory != nullptr);
+
+        // Initialize the base allocator
+        impl.Initialize(obj_size);
+
+        // Set our tracking variables
+        const std::size_t num_obj = (memory_size / obj_size);
+        start = reinterpret_cast<uintptr_t>(memory);
+        end = start + num_obj * obj_size;
+        peak = start;
+
+        // Free the objects
+        u8* cur = reinterpret_cast<u8*>(end);
+
+        for (std::size_t i{}; i < num_obj; i++) {
+            cur -= obj_size;
+            impl.Free(cur);
+        }
+    }
+
+private:
+    using Impl = impl::SlabHeapImpl;
+
+    Impl impl;
+    uintptr_t peak{};
+    uintptr_t start{};
+    uintptr_t end{};
+};
+
+template <typename T>
+class SlabHeap final : public SlabHeapBase {
+public:
+    constexpr SlabHeap() : SlabHeapBase() {}
+
+    void Initialize(void* memory, std::size_t memory_size) {
+        InitializeImpl(sizeof(T), memory, memory_size);
+    }
+
+    T* Allocate() {
+        T* obj = static_cast<T*>(AllocateImpl());
+        if (obj != nullptr) {
+            new (obj) T();
+        }
+        return obj;
+    }
+
+    void Free(T* obj) {
+        FreeImpl(obj);
+    }
+
+    constexpr std::size_t GetObjectIndex(const T* obj) const {
+        return GetObjectIndexImpl(obj);
+    }
+};
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/system_control.cpp

@@ -0,0 +1,40 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <random>
+
+#include "core/hle/kernel/memory/system_control.h"
+
+namespace Kernel::Memory::SystemControl {
+namespace {
+template <typename F>
+u64 GenerateUniformRange(u64 min, u64 max, F f) {
+    // Handle the case where the difference is too large to represent.
+    if (max == std::numeric_limits<u64>::max() && min == std::numeric_limits<u64>::min()) {
+        return f();
+    }
+
+    // Iterate until we get a value in range.
+    const u64 range_size = ((max + 1) - min);
+    const u64 effective_max = (std::numeric_limits<u64>::max() / range_size) * range_size;
+    while (true) {
+        if (const u64 rnd = f(); rnd < effective_max) {
+            return min + (rnd % range_size);
+        }
+    }
+}
+
+u64 GenerateRandomU64ForInit() {
+    static std::random_device device;
+    static std::mt19937 gen(device());
+    static std::uniform_int_distribution<u64> distribution(1, std::numeric_limits<u64>::max());
+    return distribution(gen);
+}
+} // Anonymous namespace
+
+u64 GenerateRandomRange(u64 min, u64 max) {
+    return GenerateUniformRange(min, max, GenerateRandomU64ForInit);
+}
+
+} // namespace Kernel::Memory::SystemControl

src/core/hle/kernel/memory/system_control.h

@@ -0,0 +1,13 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Kernel::Memory::SystemControl {
+
+u64 GenerateRandomRange(u64 min, u64 max);
+
+} // namespace Kernel::Memory::SystemControl

src/core/hle/kernel/mutex.cpp

@@ -0,0 +1,170 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <memory>
+#include <utility>
+#include <vector>
+
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "core/core.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/mutex.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/result.h"
+#include "core/memory.h"
+
+namespace Kernel {
+
+/// Returns the number of threads that are waiting for a mutex, and the highest priority one among
+/// those.
+static std::pair<std::shared_ptr<Thread>, u32> GetHighestPriorityMutexWaitingThread(
+    const std::shared_ptr<Thread>& current_thread, VAddr mutex_addr) {
+
+    std::shared_ptr<Thread> highest_priority_thread;
+    u32 num_waiters = 0;
+
+    for (const auto& thread : current_thread->GetMutexWaitingThreads()) {
+        if (thread->GetMutexWaitAddress() != mutex_addr)
+            continue;
+
+        ++num_waiters;
+        if (highest_priority_thread == nullptr ||
+            thread->GetPriority() < highest_priority_thread->GetPriority()) {
+            highest_priority_thread = thread;
+        }
+    }
+
+    return {highest_priority_thread, num_waiters};
+}
+
+/// Update the mutex owner field of all threads waiting on the mutex to point to the new owner.
+static void TransferMutexOwnership(VAddr mutex_addr, std::shared_ptr<Thread> current_thread,
+                                   std::shared_ptr<Thread> new_owner) {
+    current_thread->RemoveMutexWaiter(new_owner);
+    const auto threads = current_thread->GetMutexWaitingThreads();
+    for (const auto& thread : threads) {
+        if (thread->GetMutexWaitAddress() != mutex_addr)
+            continue;
+
+        ASSERT(thread->GetLockOwner() == current_thread.get());
+        current_thread->RemoveMutexWaiter(thread);
+        if (new_owner != thread)
+            new_owner->AddMutexWaiter(thread);
+    }
+}
+
+Mutex::Mutex(Core::System& system) : system{system} {}
+Mutex::~Mutex() = default;
+
+ResultCode Mutex::TryAcquire(VAddr address, Handle holding_thread_handle,
+                             Handle requesting_thread_handle) {
+    // The mutex address must be 4-byte aligned
+    if ((address % sizeof(u32)) != 0) {
+        LOG_ERROR(Kernel, "Address is not 4-byte aligned! address={:016X}", address);
+        return ERR_INVALID_ADDRESS;
+    }
+
+    auto& kernel = system.Kernel();
+    std::shared_ptr<Thread> current_thread =
+        SharedFrom(kernel.CurrentScheduler()->GetCurrentThread());
+    {
+        KScopedSchedulerLock lock(kernel);
+        // The mutex address must be 4-byte aligned
+        if ((address % sizeof(u32)) != 0) {
+            return ERR_INVALID_ADDRESS;
+        }
+
+        const auto& handle_table = kernel.CurrentProcess()->GetHandleTable();
+        std::shared_ptr<Thread> holding_thread = handle_table.Get<Thread>(holding_thread_handle);
+        std::shared_ptr<Thread> requesting_thread =
+            handle_table.Get<Thread>(requesting_thread_handle);
+
+        // TODO(Subv): It is currently unknown if it is possible to lock a mutex in behalf of
+        // another thread.
+        ASSERT(requesting_thread == current_thread);
+
+        current_thread->SetSynchronizationResults(nullptr, RESULT_SUCCESS);
+
+        const u32 addr_value = system.Memory().Read32(address);
+
+        // If the mutex isn't being held, just return success.
+        if (addr_value != (holding_thread_handle | Mutex::MutexHasWaitersFlag)) {
+            return RESULT_SUCCESS;
+        }
+
+        if (holding_thread == nullptr) {
+            return ERR_INVALID_HANDLE;
+        }
+
+        // Wait until the mutex is released
+        current_thread->SetMutexWaitAddress(address);
+        current_thread->SetWaitHandle(requesting_thread_handle);
+
+        current_thread->SetStatus(ThreadStatus::WaitMutex);
+
+        // Update the lock holder thread's priority to prevent priority inversion.
+        holding_thread->AddMutexWaiter(current_thread);
+    }
+
+    {
+        KScopedSchedulerLock lock(kernel);
+        auto* owner = current_thread->GetLockOwner();
+        if (owner != nullptr) {
+            owner->RemoveMutexWaiter(current_thread);
+        }
+    }
+    return current_thread->GetSignalingResult();
+}
+
+std::pair<ResultCode, std::shared_ptr<Thread>> Mutex::Unlock(std::shared_ptr<Thread> owner,
+                                                             VAddr address) {
+    // The mutex address must be 4-byte aligned
+    if ((address % sizeof(u32)) != 0) {
+        LOG_ERROR(Kernel, "Address is not 4-byte aligned! address={:016X}", address);
+        return {ERR_INVALID_ADDRESS, nullptr};
+    }
+
+    auto [new_owner, num_waiters] = GetHighestPriorityMutexWaitingThread(owner, address);
+    if (new_owner == nullptr) {
+        system.Memory().Write32(address, 0);
+        return {RESULT_SUCCESS, nullptr};
+    }
+    // Transfer the ownership of the mutex from the previous owner to the new one.
+    TransferMutexOwnership(address, owner, new_owner);
+    u32 mutex_value = new_owner->GetWaitHandle();
+    if (num_waiters >= 2) {
+        // Notify the guest that there are still some threads waiting for the mutex
+        mutex_value |= Mutex::MutexHasWaitersFlag;
+    }
+    new_owner->SetSynchronizationResults(nullptr, RESULT_SUCCESS);
+    new_owner->SetLockOwner(nullptr);
+    new_owner->ResumeFromWait();
+
+    system.Memory().Write32(address, mutex_value);
+    return {RESULT_SUCCESS, new_owner};
+}
+
+ResultCode Mutex::Release(VAddr address) {
+    auto& kernel = system.Kernel();
+    KScopedSchedulerLock lock(kernel);
+
+    std::shared_ptr<Thread> current_thread =
+        SharedFrom(kernel.CurrentScheduler()->GetCurrentThread());
+
+    auto [result, new_owner] = Unlock(current_thread, address);
+
+    if (result != RESULT_SUCCESS && new_owner != nullptr) {
+        new_owner->SetSynchronizationResults(nullptr, result);
+    }
+
+    return result;
+}
+
+} // namespace Kernel

src/core/hle/kernel/mutex.h

@@ -0,0 +1,42 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+
+union ResultCode;
+
+namespace Core {
+class System;
+}
+
+namespace Kernel {
+
+class Mutex final {
+public:
+    explicit Mutex(Core::System& system);
+    ~Mutex();
+
+    /// Flag that indicates that a mutex still has threads waiting for it.
+    static constexpr u32 MutexHasWaitersFlag = 0x40000000;
+    /// Mask of the bits in a mutex address value that contain the mutex owner.
+    static constexpr u32 MutexOwnerMask = 0xBFFFFFFF;
+
+    /// Attempts to acquire a mutex at the specified address.
+    ResultCode TryAcquire(VAddr address, Handle holding_thread_handle,
+                          Handle requesting_thread_handle);
+
+    /// Unlocks a mutex for owner at address
+    std::pair<ResultCode, std::shared_ptr<Thread>> Unlock(std::shared_ptr<Thread> owner,
+                                                          VAddr address);
+
+    /// Releases the mutex at the specified address.
+    ResultCode Release(VAddr address);
+
+private:
+    Core::System& system;
+};
+
+} // namespace Kernel

src/core/hle/kernel/object.cpp

@@ -0,0 +1,38 @@
+// Copyright 2018 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/object.h"
+
+namespace Kernel {
+
+Object::Object(KernelCore& kernel) : kernel{kernel}, object_id{kernel.CreateNewObjectID()} {}
+Object::~Object() = default;
+
+bool Object::IsWaitable() const {
+    switch (GetHandleType()) {
+    case HandleType::ReadableEvent:
+    case HandleType::Thread:
+    case HandleType::Process:
+    case HandleType::ServerPort:
+    case HandleType::ServerSession:
+        return true;
+
+    case HandleType::Unknown:
+    case HandleType::WritableEvent:
+    case HandleType::SharedMemory:
+    case HandleType::TransferMemory:
+    case HandleType::ResourceLimit:
+    case HandleType::ClientPort:
+    case HandleType::ClientSession:
+    case HandleType::Session:
+        return false;
+    }
+
+    UNREACHABLE();
+    return false;
+}
+
+} // namespace Kernel

src/core/hle/kernel/object.h

@@ -0,0 +1,86 @@
+// Copyright 2018 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <atomic>
+#include <memory>
+#include <string>
+
+#include "common/common_types.h"
+
+namespace Kernel {
+
+class KernelCore;
+
+using Handle = u32;
+
+enum class HandleType : u32 {
+    Unknown,
+    WritableEvent,
+    ReadableEvent,
+    SharedMemory,
+    TransferMemory,
+    Thread,
+    Process,
+    ResourceLimit,
+    ClientPort,
+    ServerPort,
+    ClientSession,
+    ServerSession,
+    Session,
+};
+
+class Object : NonCopyable, public std::enable_shared_from_this<Object> {
+public:
+    explicit Object(KernelCore& kernel);
+    virtual ~Object();
+
+    /// Returns a unique identifier for the object. For debugging purposes only.
+    u32 GetObjectId() const {
+        return object_id.load(std::memory_order_relaxed);
+    }
+
+    virtual std::string GetTypeName() const {
+        return "[BAD KERNEL OBJECT TYPE]";
+    }
+    virtual std::string GetName() const {
+        return "[UNKNOWN KERNEL OBJECT]";
+    }
+    virtual HandleType GetHandleType() const = 0;
+
+    /**
+     * Check if a thread can wait on the object
+     * @return True if a thread can wait on the object, otherwise false
+     */
+    bool IsWaitable() const;
+
+protected:
+    /// The kernel instance this object was created under.
+    KernelCore& kernel;
+
+private:
+    std::atomic<u32> object_id{0};
+};
+
+template <typename T>
+std::shared_ptr<T> SharedFrom(T* raw) {
+    if (raw == nullptr)
+        return nullptr;
+    return std::static_pointer_cast<T>(raw->shared_from_this());
+}
+
+/**
+ * Attempts to downcast the given Object pointer to a pointer to T.
+ * @return Derived pointer to the object, or `nullptr` if `object` isn't of type T.
+ */
+template <typename T>
+inline std::shared_ptr<T> DynamicObjectCast(std::shared_ptr<Object> object) {
+    if (object != nullptr && object->GetHandleType() == T::HANDLE_TYPE) {
+        return std::static_pointer_cast<T>(object);
+    }
+    return nullptr;
+}
+
+} // namespace Kernel

src/core/hle/kernel/physical_core.cpp

@@ -0,0 +1,62 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/spin_lock.h"
+#include "core/arm/cpu_interrupt_handler.h"
+#include "core/arm/dynarmic/arm_dynarmic_32.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
+#include "core/core.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/physical_core.h"
+
+namespace Kernel {
+
+PhysicalCore::PhysicalCore(std::size_t core_index, Core::System& system,
+                           Kernel::KScheduler& scheduler, Core::CPUInterrupts& interrupts)
+    : core_index{core_index}, system{system}, scheduler{scheduler},
+      interrupts{interrupts}, guard{std::make_unique<Common::SpinLock>()} {}
+
+PhysicalCore::~PhysicalCore() = default;
+
+void PhysicalCore::Initialize([[maybe_unused]] bool is_64_bit) {
+#ifdef ARCHITECTURE_x86_64
+    auto& kernel = system.Kernel();
+    if (is_64_bit) {
+        arm_interface = std::make_unique<Core::ARM_Dynarmic_64>(
+            system, interrupts, kernel.IsMulticore(), kernel.GetExclusiveMonitor(), core_index);
+    } else {
+        arm_interface = std::make_unique<Core::ARM_Dynarmic_32>(
+            system, interrupts, kernel.IsMulticore(), kernel.GetExclusiveMonitor(), core_index);
+    }
+#else
+#error Platform not supported yet.
+#endif
+}
+
+void PhysicalCore::Run() {
+    arm_interface->Run();
+}
+
+void PhysicalCore::Idle() {
+    interrupts[core_index].AwaitInterrupt();
+}
+
+bool PhysicalCore::IsInterrupted() const {
+    return interrupts[core_index].IsInterrupted();
+}
+
+void PhysicalCore::Interrupt() {
+    guard->lock();
+    interrupts[core_index].SetInterrupt(true);
+    guard->unlock();
+}
+
+void PhysicalCore::ClearInterrupt() {
+    guard->lock();
+    interrupts[core_index].SetInterrupt(false);
+    guard->unlock();
+}
+
+} // namespace Kernel

src/core/hle/kernel/physical_core.h

@@ -0,0 +1,99 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <cstddef>
+#include <memory>
+
+#include "core/arm/arm_interface.h"
+
+namespace Common {
+class SpinLock;
+}
+
+namespace Kernel {
+class KScheduler;
+} // namespace Kernel
+
+namespace Core {
+class CPUInterruptHandler;
+class ExclusiveMonitor;
+class System;
+} // namespace Core
+
+namespace Kernel {
+
+class PhysicalCore {
+public:
+    PhysicalCore(std::size_t core_index, Core::System& system, Kernel::KScheduler& scheduler,
+                 Core::CPUInterrupts& interrupts);
+    ~PhysicalCore();
+
+    PhysicalCore(const PhysicalCore&) = delete;
+    PhysicalCore& operator=(const PhysicalCore&) = delete;
+
+    PhysicalCore(PhysicalCore&&) = default;
+    PhysicalCore& operator=(PhysicalCore&&) = delete;
+
+    /// Initialize the core for the specified parameters.
+    void Initialize(bool is_64_bit);
+
+    /// Execute current jit state
+    void Run();
+
+    void Idle();
+
+    /// Interrupt this physical core.
+    void Interrupt();
+
+    /// Clear this core's interrupt
+    void ClearInterrupt();
+
+    /// Check if this core is interrupted
+    bool IsInterrupted() const;
+
+    bool IsInitialized() const {
+        return arm_interface != nullptr;
+    }
+
+    Core::ARM_Interface& ArmInterface() {
+        return *arm_interface;
+    }
+
+    const Core::ARM_Interface& ArmInterface() const {
+        return *arm_interface;
+    }
+
+    bool IsMainCore() const {
+        return core_index == 0;
+    }
+
+    bool IsSystemCore() const {
+        return core_index == 3;
+    }
+
+    std::size_t CoreIndex() const {
+        return core_index;
+    }
+
+    Kernel::KScheduler& Scheduler() {
+        return scheduler;
+    }
+
+    const Kernel::KScheduler& Scheduler() const {
+        return scheduler;
+    }
+
+private:
+    const std::size_t core_index;
+    Core::System& system;
+    Kernel::KScheduler& scheduler;
+    Core::CPUInterrupts& interrupts;
+    std::unique_ptr<Common::SpinLock> guard;
+    std::unique_ptr<Core::ARM_Interface> arm_interface;
+};
+
+} // namespace Kernel

src/core/hle/kernel/physical_memory.h

@@ -0,0 +1,24 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <vector>
+
+#include "common/alignment.h"
+
+namespace Kernel {
+
+// This encapsulation serves 2 purposes:
+// - First, to encapsulate host physical memory under a single type and set an
+// standard for managing it.
+// - Second to ensure all host backing memory used is aligned to 256 bytes due
+// to strict alignment restrictions on GPU memory.
+
+using PhysicalMemoryVector = std::vector<u8, Common::AlignmentAllocator<u8, 256>>;
+class PhysicalMemory final : public PhysicalMemoryVector {
+    using PhysicalMemoryVector::PhysicalMemoryVector;
+};
+
+} // namespace Kernel

src/core/hle/kernel/process.cpp

@@ -0,0 +1,454 @@
+// Copyright 2015 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <bitset>
+#include <ctime>
+#include <memory>
+#include <random>
+#include "common/alignment.h"
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "core/core.h"
+#include "core/device_memory.h"
+#include "core/file_sys/program_metadata.h"
+#include "core/hle/kernel/code_set.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/memory/memory_block_manager.h"
+#include "core/hle/kernel/memory/page_table.h"
+#include "core/hle/kernel/memory/slab_heap.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/kernel/resource_limit.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/lock.h"
+#include "core/memory.h"
+#include "core/settings.h"
+
+namespace Kernel {
+namespace {
+/**
+ * Sets up the primary application thread
+ *
+ * @param system The system instance to create the main thread under.
+ * @param owner_process The parent process for the main thread
+ * @param priority The priority to give the main thread
+ */
+void SetupMainThread(Core::System& system, Process& owner_process, u32 priority, VAddr stack_top) {
+    const VAddr entry_point = owner_process.PageTable().GetCodeRegionStart();
+    ThreadType type = THREADTYPE_USER;
+    auto thread_res = Thread::Create(system, type, "main", entry_point, priority, 0,
+                                     owner_process.GetIdealCore(), stack_top, &owner_process);
+
+    std::shared_ptr<Thread> thread = std::move(thread_res).Unwrap();
+
+    // Register 1 must be a handle to the main thread
+    const Handle thread_handle = owner_process.GetHandleTable().Create(thread).Unwrap();
+    thread->GetContext32().cpu_registers[0] = 0;
+    thread->GetContext64().cpu_registers[0] = 0;
+    thread->GetContext32().cpu_registers[1] = thread_handle;
+    thread->GetContext64().cpu_registers[1] = thread_handle;
+
+    auto& kernel = system.Kernel();
+    // Threads by default are dormant, wake up the main thread so it runs when the scheduler fires
+    {
+        KScopedSchedulerLock lock{kernel};
+        thread->SetStatus(ThreadStatus::Ready);
+    }
+}
+} // Anonymous namespace
+
+// Represents a page used for thread-local storage.
+//
+// Each TLS page contains slots that may be used by processes and threads.
+// Every process and thread is created with a slot in some arbitrary page
+// (whichever page happens to have an available slot).
+class TLSPage {
+public:
+    static constexpr std::size_t num_slot_entries =
+        Core::Memory::PAGE_SIZE / Core::Memory::TLS_ENTRY_SIZE;
+
+    explicit TLSPage(VAddr address) : base_address{address} {}
+
+    bool HasAvailableSlots() const {
+        return !is_slot_used.all();
+    }
+
+    VAddr GetBaseAddress() const {
+        return base_address;
+    }
+
+    std::optional<VAddr> ReserveSlot() {
+        for (std::size_t i = 0; i < is_slot_used.size(); i++) {
+            if (is_slot_used[i]) {
+                continue;
+            }
+
+            is_slot_used[i] = true;
+            return base_address + (i * Core::Memory::TLS_ENTRY_SIZE);
+        }
+
+        return std::nullopt;
+    }
+
+    void ReleaseSlot(VAddr address) {
+        // Ensure that all given addresses are consistent with how TLS pages
+        // are intended to be used when releasing slots.
+        ASSERT(IsWithinPage(address));
+        ASSERT((address % Core::Memory::TLS_ENTRY_SIZE) == 0);
+
+        const std::size_t index = (address - base_address) / Core::Memory::TLS_ENTRY_SIZE;
+        is_slot_used[index] = false;
+    }
+
+private:
+    bool IsWithinPage(VAddr address) const {
+        return base_address <= address && address < base_address + Core::Memory::PAGE_SIZE;
+    }
+
+    VAddr base_address;
+    std::bitset<num_slot_entries> is_slot_used;
+};
+
+std::shared_ptr<Process> Process::Create(Core::System& system, std::string name, ProcessType type) {
+    auto& kernel = system.Kernel();
+
+    std::shared_ptr<Process> process = std::make_shared<Process>(system);
+    process->name = std::move(name);
+    process->resource_limit = ResourceLimit::Create(kernel);
+    process->status = ProcessStatus::Created;
+    process->program_id = 0;
+    process->process_id = type == ProcessType::KernelInternal ? kernel.CreateNewKernelProcessID()
+                                                              : kernel.CreateNewUserProcessID();
+    process->capabilities.InitializeForMetadatalessProcess();
+
+    std::mt19937 rng(Settings::values.rng_seed.GetValue().value_or(std::time(nullptr)));
+    std::uniform_int_distribution<u64> distribution;
+    std::generate(process->random_entropy.begin(), process->random_entropy.end(),
+                  [&] { return distribution(rng); });
+
+    kernel.AppendNewProcess(process);
+    return process;
+}
+
+std::shared_ptr<ResourceLimit> Process::GetResourceLimit() const {
+    return resource_limit;
+}
+
+u64 Process::GetTotalPhysicalMemoryAvailable() const {
+    const u64 capacity{resource_limit->GetCurrentResourceValue(ResourceType::PhysicalMemory) +
+                       page_table->GetTotalHeapSize() + GetSystemResourceSize() + image_size +
+                       main_thread_stack_size};
+
+    if (capacity < memory_usage_capacity) {
+        return capacity;
+    }
+
+    return memory_usage_capacity;
+}
+
+u64 Process::GetTotalPhysicalMemoryAvailableWithoutSystemResource() const {
+    return GetTotalPhysicalMemoryAvailable() - GetSystemResourceSize();
+}
+
+u64 Process::GetTotalPhysicalMemoryUsed() const {
+    return image_size + main_thread_stack_size + page_table->GetTotalHeapSize() +
+           GetSystemResourceSize();
+}
+
+u64 Process::GetTotalPhysicalMemoryUsedWithoutSystemResource() const {
+    return GetTotalPhysicalMemoryUsed() - GetSystemResourceUsage();
+}
+
+void Process::InsertConditionVariableThread(std::shared_ptr<Thread> thread) {
+    VAddr cond_var_addr = thread->GetCondVarWaitAddress();
+    std::list<std::shared_ptr<Thread>>& thread_list = cond_var_threads[cond_var_addr];
+    auto it = thread_list.begin();
+    while (it != thread_list.end()) {
+        const std::shared_ptr<Thread> current_thread = *it;
+        if (current_thread->GetPriority() > thread->GetPriority()) {
+            thread_list.insert(it, thread);
+            return;
+        }
+        ++it;
+    }
+    thread_list.push_back(thread);
+}
+
+void Process::RemoveConditionVariableThread(std::shared_ptr<Thread> thread) {
+    VAddr cond_var_addr = thread->GetCondVarWaitAddress();
+    std::list<std::shared_ptr<Thread>>& thread_list = cond_var_threads[cond_var_addr];
+    auto it = thread_list.begin();
+    while (it != thread_list.end()) {
+        const std::shared_ptr<Thread> current_thread = *it;
+        if (current_thread.get() == thread.get()) {
+            thread_list.erase(it);
+            return;
+        }
+        ++it;
+    }
+}
+
+std::vector<std::shared_ptr<Thread>> Process::GetConditionVariableThreads(
+    const VAddr cond_var_addr) {
+    std::vector<std::shared_ptr<Thread>> result{};
+    std::list<std::shared_ptr<Thread>>& thread_list = cond_var_threads[cond_var_addr];
+    auto it = thread_list.begin();
+    while (it != thread_list.end()) {
+        std::shared_ptr<Thread> current_thread = *it;
+        result.push_back(current_thread);
+        ++it;
+    }
+    return result;
+}
+
+void Process::RegisterThread(const Thread* thread) {
+    thread_list.push_back(thread);
+}
+
+void Process::UnregisterThread(const Thread* thread) {
+    thread_list.remove(thread);
+}
+
+ResultCode Process::ClearSignalState() {
+    KScopedSchedulerLock lock(system.Kernel());
+    if (status == ProcessStatus::Exited) {
+        LOG_ERROR(Kernel, "called on a terminated process instance.");
+        return ERR_INVALID_STATE;
+    }
+
+    if (!is_signaled) {
+        LOG_ERROR(Kernel, "called on a process instance that isn't signaled.");
+        return ERR_INVALID_STATE;
+    }
+
+    is_signaled = false;
+    return RESULT_SUCCESS;
+}
+
+ResultCode Process::LoadFromMetadata(const FileSys::ProgramMetadata& metadata,
+                                     std::size_t code_size) {
+    program_id = metadata.GetTitleID();
+    ideal_core = metadata.GetMainThreadCore();
+    is_64bit_process = metadata.Is64BitProgram();
+    system_resource_size = metadata.GetSystemResourceSize();
+    image_size = code_size;
+
+    // Initialize proces address space
+    if (const ResultCode result{
+            page_table->InitializeForProcess(metadata.GetAddressSpaceType(), false, 0x8000000,
+                                             code_size, Memory::MemoryManager::Pool::Application)};
+        result.IsError()) {
+        return result;
+    }
+
+    // Map process code region
+    if (const ResultCode result{page_table->MapProcessCode(
+            page_table->GetCodeRegionStart(), code_size / Memory::PageSize,
+            Memory::MemoryState::Code, Memory::MemoryPermission::None)};
+        result.IsError()) {
+        return result;
+    }
+
+    // Initialize process capabilities
+    const auto& caps{metadata.GetKernelCapabilities()};
+    if (const ResultCode result{
+            capabilities.InitializeForUserProcess(caps.data(), caps.size(), *page_table)};
+        result.IsError()) {
+        return result;
+    }
+
+    // Set memory usage capacity
+    switch (metadata.GetAddressSpaceType()) {
+    case FileSys::ProgramAddressSpaceType::Is32Bit:
+    case FileSys::ProgramAddressSpaceType::Is36Bit:
+    case FileSys::ProgramAddressSpaceType::Is39Bit:
+        memory_usage_capacity = page_table->GetHeapRegionEnd() - page_table->GetHeapRegionStart();
+        break;
+
+    case FileSys::ProgramAddressSpaceType::Is32BitNoMap:
+        memory_usage_capacity = page_table->GetHeapRegionEnd() - page_table->GetHeapRegionStart() +
+                                page_table->GetAliasRegionEnd() - page_table->GetAliasRegionStart();
+        break;
+
+    default:
+        UNREACHABLE();
+    }
+
+    // Set initial resource limits
+    resource_limit->SetLimitValue(
+        ResourceType::PhysicalMemory,
+        kernel.MemoryManager().GetSize(Memory::MemoryManager::Pool::Application));
+    resource_limit->SetLimitValue(ResourceType::Threads, 608);
+    resource_limit->SetLimitValue(ResourceType::Events, 700);
+    resource_limit->SetLimitValue(ResourceType::TransferMemory, 128);
+    resource_limit->SetLimitValue(ResourceType::Sessions, 894);
+    ASSERT(resource_limit->Reserve(ResourceType::PhysicalMemory, code_size));
+
+    // Create TLS region
+    tls_region_address = CreateTLSRegion();
+
+    return handle_table.SetSize(capabilities.GetHandleTableSize());
+}
+
+void Process::Run(s32 main_thread_priority, u64 stack_size) {
+    AllocateMainThreadStack(stack_size);
+
+    const std::size_t heap_capacity{memory_usage_capacity - main_thread_stack_size - image_size};
+    ASSERT(!page_table->SetHeapCapacity(heap_capacity).IsError());
+
+    ChangeStatus(ProcessStatus::Running);
+
+    SetupMainThread(system, *this, main_thread_priority, main_thread_stack_top);
+    resource_limit->Reserve(ResourceType::Threads, 1);
+    resource_limit->Reserve(ResourceType::PhysicalMemory, main_thread_stack_size);
+}
+
+void Process::PrepareForTermination() {
+    ChangeStatus(ProcessStatus::Exiting);
+
+    const auto stop_threads = [this](const std::vector<std::shared_ptr<Thread>>& thread_list) {
+        for (auto& thread : thread_list) {
+            if (thread->GetOwnerProcess() != this)
+                continue;
+
+            if (thread.get() == kernel.CurrentScheduler()->GetCurrentThread())
+                continue;
+
+            // TODO(Subv): When are the other running/ready threads terminated?
+            ASSERT_MSG(thread->GetStatus() == ThreadStatus::WaitSynch,
+                       "Exiting processes with non-waiting threads is currently unimplemented");
+
+            thread->Stop();
+        }
+    };
+
+    stop_threads(system.GlobalSchedulerContext().GetThreadList());
+
+    FreeTLSRegion(tls_region_address);
+    tls_region_address = 0;
+
+    ChangeStatus(ProcessStatus::Exited);
+}
+
+/**
+ * Attempts to find a TLS page that contains a free slot for
+ * use by a thread.
+ *
+ * @returns If a page with an available slot is found, then an iterator
+ *          pointing to the page is returned. Otherwise the end iterator
+ *          is returned instead.
+ */
+static auto FindTLSPageWithAvailableSlots(std::vector<TLSPage>& tls_pages) {
+    return std::find_if(tls_pages.begin(), tls_pages.end(),
+                        [](const auto& page) { return page.HasAvailableSlots(); });
+}
+
+VAddr Process::CreateTLSRegion() {
+    KScopedSchedulerLock lock(system.Kernel());
+    if (auto tls_page_iter{FindTLSPageWithAvailableSlots(tls_pages)};
+        tls_page_iter != tls_pages.cend()) {
+        return *tls_page_iter->ReserveSlot();
+    }
+
+    Memory::Page* const tls_page_ptr{kernel.GetUserSlabHeapPages().Allocate()};
+    ASSERT(tls_page_ptr);
+
+    const VAddr start{page_table->GetKernelMapRegionStart()};
+    const VAddr size{page_table->GetKernelMapRegionEnd() - start};
+    const PAddr tls_map_addr{system.DeviceMemory().GetPhysicalAddr(tls_page_ptr)};
+    const VAddr tls_page_addr{
+        page_table
+            ->AllocateAndMapMemory(1, Memory::PageSize, true, start, size / Memory::PageSize,
+                                   Memory::MemoryState::ThreadLocal,
+                                   Memory::MemoryPermission::ReadAndWrite, tls_map_addr)
+            .ValueOr(0)};
+
+    ASSERT(tls_page_addr);
+
+    std::memset(tls_page_ptr, 0, Memory::PageSize);
+    tls_pages.emplace_back(tls_page_addr);
+
+    const auto reserve_result{tls_pages.back().ReserveSlot()};
+    ASSERT(reserve_result.has_value());
+
+    return *reserve_result;
+}
+
+void Process::FreeTLSRegion(VAddr tls_address) {
+    KScopedSchedulerLock lock(system.Kernel());
+    const VAddr aligned_address = Common::AlignDown(tls_address, Core::Memory::PAGE_SIZE);
+    auto iter =
+        std::find_if(tls_pages.begin(), tls_pages.end(), [aligned_address](const auto& page) {
+            return page.GetBaseAddress() == aligned_address;
+        });
+
+    // Something has gone very wrong if we're freeing a region
+    // with no actual page available.
+    ASSERT(iter != tls_pages.cend());
+
+    iter->ReleaseSlot(tls_address);
+}
+
+void Process::LoadModule(CodeSet code_set, VAddr base_addr) {
+    std::lock_guard lock{HLE::g_hle_lock};
+    const auto ReprotectSegment = [&](const CodeSet::Segment& segment,
+                                      Memory::MemoryPermission permission) {
+        page_table->SetCodeMemoryPermission(segment.addr + base_addr, segment.size, permission);
+    };
+
+    system.Memory().WriteBlock(*this, base_addr, code_set.memory.data(), code_set.memory.size());
+
+    ReprotectSegment(code_set.CodeSegment(), Memory::MemoryPermission::ReadAndExecute);
+    ReprotectSegment(code_set.RODataSegment(), Memory::MemoryPermission::Read);
+    ReprotectSegment(code_set.DataSegment(), Memory::MemoryPermission::ReadAndWrite);
+}
+
+Process::Process(Core::System& system)
+    : SynchronizationObject{system.Kernel()}, page_table{std::make_unique<Memory::PageTable>(
+                                                  system)},
+      handle_table{system.Kernel()}, address_arbiter{system}, mutex{system}, system{system} {}
+
+Process::~Process() = default;
+
+void Process::Acquire(Thread* thread) {
+    ASSERT_MSG(!ShouldWait(thread), "Object unavailable!");
+}
+
+bool Process::ShouldWait(const Thread* thread) const {
+    return !is_signaled;
+}
+
+void Process::ChangeStatus(ProcessStatus new_status) {
+    if (status == new_status) {
+        return;
+    }
+
+    status = new_status;
+    is_signaled = true;
+    Signal();
+}
+
+ResultCode Process::AllocateMainThreadStack(std::size_t stack_size) {
+    ASSERT(stack_size);
+
+    // The kernel always ensures that the given stack size is page aligned.
+    main_thread_stack_size = Common::AlignUp(stack_size, Memory::PageSize);
+
+    const VAddr start{page_table->GetStackRegionStart()};
+    const std::size_t size{page_table->GetStackRegionEnd() - start};
+
+    CASCADE_RESULT(main_thread_stack_top,
+                   page_table->AllocateAndMapMemory(
+                       main_thread_stack_size / Memory::PageSize, Memory::PageSize, false, start,
+                       size / Memory::PageSize, Memory::MemoryState::Stack,
+                       Memory::MemoryPermission::ReadAndWrite));
+
+    main_thread_stack_top += main_thread_stack_size;
+
+    return RESULT_SUCCESS;
+}
+
+} // namespace Kernel

src/core/hle/kernel/process.h

@@ -0,0 +1,417 @@
+// Copyright 2015 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <cstddef>
+#include <list>
+#include <string>
+#include <unordered_map>
+#include <vector>
+#include "common/common_types.h"
+#include "core/hle/kernel/address_arbiter.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/mutex.h"
+#include "core/hle/kernel/process_capability.h"
+#include "core/hle/kernel/synchronization_object.h"
+#include "core/hle/result.h"
+
+namespace Core {
+class System;
+}
+
+namespace FileSys {
+class ProgramMetadata;
+}
+
+namespace Kernel {
+
+class KernelCore;
+class ResourceLimit;
+class Thread;
+class TLSPage;
+
+struct CodeSet;
+
+namespace Memory {
+class PageTable;
+}
+
+enum class MemoryRegion : u16 {
+    APPLICATION = 1,
+    SYSTEM = 2,
+    BASE = 3,
+};
+
+/**
+ * Indicates the status of a Process instance.
+ *
+ * @note These match the values as used by kernel,
+ *       so new entries should only be added if RE
+ *       shows that a new value has been introduced.
+ */
+enum class ProcessStatus {
+    Created,
+    CreatedWithDebuggerAttached,
+    Running,
+    WaitingForDebuggerToAttach,
+    DebuggerAttached,
+    Exiting,
+    Exited,
+    DebugBreak,
+};
+
+class Process final : public SynchronizationObject {
+public:
+    explicit Process(Core::System& system);
+    ~Process() override;
+
+    enum : u64 {
+        /// Lowest allowed process ID for a kernel initial process.
+        InitialKIPIDMin = 1,
+        /// Highest allowed process ID for a kernel initial process.
+        InitialKIPIDMax = 80,
+
+        /// Lowest allowed process ID for a userland process.
+        ProcessIDMin = 81,
+        /// Highest allowed process ID for a userland process.
+        ProcessIDMax = 0xFFFFFFFFFFFFFFFF,
+    };
+
+    // Used to determine how process IDs are assigned.
+    enum class ProcessType {
+        KernelInternal,
+        Userland,
+    };
+
+    static constexpr std::size_t RANDOM_ENTROPY_SIZE = 4;
+
+    static std::shared_ptr<Process> Create(Core::System& system, std::string name,
+                                           ProcessType type);
+
+    std::string GetTypeName() const override {
+        return "Process";
+    }
+    std::string GetName() const override {
+        return name;
+    }
+
+    static constexpr HandleType HANDLE_TYPE = HandleType::Process;
+    HandleType GetHandleType() const override {
+        return HANDLE_TYPE;
+    }
+
+    /// Gets a reference to the process' page table.
+    Memory::PageTable& PageTable() {
+        return *page_table;
+    }
+
+    /// Gets const a reference to the process' page table.
+    const Memory::PageTable& PageTable() const {
+        return *page_table;
+    }
+
+    /// Gets a reference to the process' handle table.
+    HandleTable& GetHandleTable() {
+        return handle_table;
+    }
+
+    /// Gets a const reference to the process' handle table.
+    const HandleTable& GetHandleTable() const {
+        return handle_table;
+    }
+
+    /// Gets a reference to the process' address arbiter.
+    AddressArbiter& GetAddressArbiter() {
+        return address_arbiter;
+    }
+
+    /// Gets a const reference to the process' address arbiter.
+    const AddressArbiter& GetAddressArbiter() const {
+        return address_arbiter;
+    }
+
+    /// Gets a reference to the process' mutex lock.
+    Mutex& GetMutex() {
+        return mutex;
+    }
+
+    /// Gets a const reference to the process' mutex lock
+    const Mutex& GetMutex() const {
+        return mutex;
+    }
+
+    /// Gets the address to the process' dedicated TLS region.
+    VAddr GetTLSRegionAddress() const {
+        return tls_region_address;
+    }
+
+    /// Gets the current status of the process
+    ProcessStatus GetStatus() const {
+        return status;
+    }
+
+    /// Gets the unique ID that identifies this particular process.
+    u64 GetProcessID() const {
+        return process_id;
+    }
+
+    /// Gets the title ID corresponding to this process.
+    u64 GetTitleID() const {
+        return program_id;
+    }
+
+    /// Gets the resource limit descriptor for this process
+    std::shared_ptr<ResourceLimit> GetResourceLimit() const;
+
+    /// Gets the ideal CPU core ID for this process
+    u8 GetIdealCore() const {
+        return ideal_core;
+    }
+
+    /// Gets the bitmask of allowed cores that this process' threads can run on.
+    u64 GetCoreMask() const {
+        return capabilities.GetCoreMask();
+    }
+
+    /// Gets the bitmask of allowed thread priorities.
+    u64 GetPriorityMask() const {
+        return capabilities.GetPriorityMask();
+    }
+
+    /// Gets the amount of secure memory to allocate for memory management.
+    u32 GetSystemResourceSize() const {
+        return system_resource_size;
+    }
+
+    /// Gets the amount of secure memory currently in use for memory management.
+    u32 GetSystemResourceUsage() const {
+        // On hardware, this returns the amount of system resource memory that has
+        // been used by the kernel. This is problematic for Yuzu to emulate, because
+        // system resource memory is used for page tables -- and yuzu doesn't really
+        // have a way to calculate how much memory is required for page tables for
+        // the current process at any given time.
+        // TODO: Is this even worth implementing? Games may retrieve this value via
+        // an SDK function that gets used + available system resource size for debug
+        // or diagnostic purposes. However, it seems unlikely that a game would make
+        // decisions based on how much system memory is dedicated to its page tables.
+        // Is returning a value other than zero wise?
+        return 0;
+    }
+
+    /// Whether this process is an AArch64 or AArch32 process.
+    bool Is64BitProcess() const {
+        return is_64bit_process;
+    }
+
+    /// Gets the total running time of the process instance in ticks.
+    u64 GetCPUTimeTicks() const {
+        return total_process_running_time_ticks;
+    }
+
+    /// Updates the total running time, adding the given ticks to it.
+    void UpdateCPUTimeTicks(u64 ticks) {
+        total_process_running_time_ticks += ticks;
+    }
+
+    /// Gets the process schedule count, used for thread yelding
+    s64 GetScheduledCount() const {
+        return schedule_count;
+    }
+
+    /// Increments the process schedule count, used for thread yielding.
+    void IncrementScheduledCount() {
+        ++schedule_count;
+    }
+
+    /// Gets 8 bytes of random data for svcGetInfo RandomEntropy
+    u64 GetRandomEntropy(std::size_t index) const {
+        return random_entropy.at(index);
+    }
+
+    /// Retrieves the total physical memory available to this process in bytes.
+    u64 GetTotalPhysicalMemoryAvailable() const;
+
+    /// Retrieves the total physical memory available to this process in bytes,
+    /// without the size of the personal system resource heap added to it.
+    u64 GetTotalPhysicalMemoryAvailableWithoutSystemResource() const;
+
+    /// Retrieves the total physical memory used by this process in bytes.
+    u64 GetTotalPhysicalMemoryUsed() const;
+
+    /// Retrieves the total physical memory used by this process in bytes,
+    /// without the size of the personal system resource heap added to it.
+    u64 GetTotalPhysicalMemoryUsedWithoutSystemResource() const;
+
+    /// Gets the list of all threads created with this process as their owner.
+    const std::list<const Thread*>& GetThreadList() const {
+        return thread_list;
+    }
+
+    /// Insert a thread into the condition variable wait container
+    void InsertConditionVariableThread(std::shared_ptr<Thread> thread);
+
+    /// Remove a thread from the condition variable wait container
+    void RemoveConditionVariableThread(std::shared_ptr<Thread> thread);
+
+    /// Obtain all condition variable threads waiting for some address
+    std::vector<std::shared_ptr<Thread>> GetConditionVariableThreads(VAddr cond_var_addr);
+
+    /// Registers a thread as being created under this process,
+    /// adding it to this process' thread list.
+    void RegisterThread(const Thread* thread);
+
+    /// Unregisters a thread from this process, removing it
+    /// from this process' thread list.
+    void UnregisterThread(const Thread* thread);
+
+    /// Clears the signaled state of the process if and only if it's signaled.
+    ///
+    /// @pre The process must not be already terminated. If this is called on a
+    ///      terminated process, then ERR_INVALID_STATE will be returned.
+    ///
+    /// @pre The process must be in a signaled state. If this is called on a
+    ///      process instance that is not signaled, ERR_INVALID_STATE will be
+    ///      returned.
+    ResultCode ClearSignalState();
+
+    /**
+     * Loads process-specifics configuration info with metadata provided
+     * by an executable.
+     *
+     * @param metadata The provided metadata to load process specific info from.
+     *
+     * @returns RESULT_SUCCESS if all relevant metadata was able to be
+     *          loaded and parsed. Otherwise, an error code is returned.
+     */
+    ResultCode LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std::size_t code_size);
+
+    /**
+     * Starts the main application thread for this process.
+     *
+     * @param main_thread_priority The priority for the main thread.
+     * @param stack_size           The stack size for the main thread in bytes.
+     */
+    void Run(s32 main_thread_priority, u64 stack_size);
+
+    /**
+     * Prepares a process for termination by stopping all of its threads
+     * and clearing any other resources.
+     */
+    void PrepareForTermination();
+
+    void LoadModule(CodeSet code_set, VAddr base_addr);
+
+    ///////////////////////////////////////////////////////////////////////////////////////////////
+    // Thread-local storage management
+
+    // Marks the next available region as used and returns the address of the slot.
+    [[nodiscard]] VAddr CreateTLSRegion();
+
+    // Frees a used TLS slot identified by the given address
+    void FreeTLSRegion(VAddr tls_address);
+
+private:
+    /// Checks if the specified thread should wait until this process is available.
+    bool ShouldWait(const Thread* thread) const override;
+
+    /// Acquires/locks this process for the specified thread if it's available.
+    void Acquire(Thread* thread) override;
+
+    /// Changes the process status. If the status is different
+    /// from the current process status, then this will trigger
+    /// a process signal.
+    void ChangeStatus(ProcessStatus new_status);
+
+    /// Allocates the main thread stack for the process, given the stack size in bytes.
+    ResultCode AllocateMainThreadStack(std::size_t stack_size);
+
+    /// Memory manager for this process
+    std::unique_ptr<Memory::PageTable> page_table;
+
+    /// Current status of the process
+    ProcessStatus status{};
+
+    /// The ID of this process
+    u64 process_id = 0;
+
+    /// Title ID corresponding to the process
+    u64 program_id = 0;
+
+    /// Specifies additional memory to be reserved for the process's memory management by the
+    /// system. When this is non-zero, secure memory is allocated and used for page table allocation
+    /// instead of using the normal global page tables/memory block management.
+    u32 system_resource_size = 0;
+
+    /// Resource limit descriptor for this process
+    std::shared_ptr<ResourceLimit> resource_limit;
+
+    /// The ideal CPU core for this process, threads are scheduled on this core by default.
+    u8 ideal_core = 0;
+
+    /// The Thread Local Storage area is allocated as processes create threads,
+    /// each TLS area is 0x200 bytes, so one page (0x1000) is split up in 8 parts, and each part
+    /// holds the TLS for a specific thread. This vector contains which parts are in use for each
+    /// page as a bitmask.
+    /// This vector will grow as more pages are allocated for new threads.
+    std::vector<TLSPage> tls_pages;
+
+    /// Contains the parsed process capability descriptors.
+    ProcessCapabilities capabilities;
+
+    /// Whether or not this process is AArch64, or AArch32.
+    /// By default, we currently assume this is true, unless otherwise
+    /// specified by metadata provided to the process during loading.
+    bool is_64bit_process = true;
+
+    /// Total running time for the process in ticks.
+    u64 total_process_running_time_ticks = 0;
+
+    /// Per-process handle table for storing created object handles in.
+    HandleTable handle_table;
+
+    /// Per-process address arbiter.
+    AddressArbiter address_arbiter;
+
+    /// The per-process mutex lock instance used for handling various
+    /// forms of services, such as lock arbitration, and condition
+    /// variable related facilities.
+    Mutex mutex;
+
+    /// Address indicating the location of the process' dedicated TLS region.
+    VAddr tls_region_address = 0;
+
+    /// Random values for svcGetInfo RandomEntropy
+    std::array<u64, RANDOM_ENTROPY_SIZE> random_entropy{};
+
+    /// List of threads that are running with this process as their owner.
+    std::list<const Thread*> thread_list;
+
+    /// List of threads waiting for a condition variable
+    std::unordered_map<VAddr, std::list<std::shared_ptr<Thread>>> cond_var_threads;
+
+    /// Address of the top of the main thread's stack
+    VAddr main_thread_stack_top{};
+
+    /// Size of the main thread's stack
+    std::size_t main_thread_stack_size{};
+
+    /// Memory usage capacity for the process
+    std::size_t memory_usage_capacity{};
+
+    /// Process total image size
+    std::size_t image_size{};
+
+    /// Name of this process
+    std::string name;
+
+    /// Schedule count of this process
+    s64 schedule_count{};
+
+    /// System context
+    Core::System& system;
+};
+
+} // namespace Kernel

src/core/hle/kernel/process_capability.cpp

@@ -0,0 +1,385 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/bit_util.h"
+#include "common/logging/log.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/memory/page_table.h"
+#include "core/hle/kernel/process_capability.h"
+
+namespace Kernel {
+namespace {
+
+// clang-format off
+
+// Shift offsets for kernel capability types.
+enum : u32 {
+    CapabilityOffset_PriorityAndCoreNum = 3,
+    CapabilityOffset_Syscall            = 4,
+    CapabilityOffset_MapPhysical        = 6,
+    CapabilityOffset_MapIO              = 7,
+    CapabilityOffset_Interrupt          = 11,
+    CapabilityOffset_ProgramType        = 13,
+    CapabilityOffset_KernelVersion      = 14,
+    CapabilityOffset_HandleTableSize    = 15,
+    CapabilityOffset_Debug              = 16,
+};
+
+// Combined mask of all parameters that may be initialized only once.
+constexpr u32 InitializeOnceMask = (1U << CapabilityOffset_PriorityAndCoreNum) |
+                                   (1U << CapabilityOffset_ProgramType) |
+                                   (1U << CapabilityOffset_KernelVersion) |
+                                   (1U << CapabilityOffset_HandleTableSize) |
+                                   (1U << CapabilityOffset_Debug);
+
+// Packed kernel version indicating 10.4.0
+constexpr u32 PackedKernelVersion = 0x520000;
+
+// Indicates possible types of capabilities that can be specified.
+enum class CapabilityType : u32 {
+    Unset              = 0U,
+    PriorityAndCoreNum = (1U << CapabilityOffset_PriorityAndCoreNum) - 1,
+    Syscall            = (1U << CapabilityOffset_Syscall) - 1,
+    MapPhysical        = (1U << CapabilityOffset_MapPhysical) - 1,
+    MapIO              = (1U << CapabilityOffset_MapIO) - 1,
+    Interrupt          = (1U << CapabilityOffset_Interrupt) - 1,
+    ProgramType        = (1U << CapabilityOffset_ProgramType) - 1,
+    KernelVersion      = (1U << CapabilityOffset_KernelVersion) - 1,
+    HandleTableSize    = (1U << CapabilityOffset_HandleTableSize) - 1,
+    Debug              = (1U << CapabilityOffset_Debug) - 1,
+    Ignorable          = 0xFFFFFFFFU,
+};
+
+// clang-format on
+
+constexpr CapabilityType GetCapabilityType(u32 value) {
+    return static_cast<CapabilityType>((~value & (value + 1)) - 1);
+}
+
+u32 GetFlagBitOffset(CapabilityType type) {
+    const auto value = static_cast<u32>(type);
+    return static_cast<u32>(Common::BitSize<u32>() - Common::CountLeadingZeroes32(value));
+}
+
+} // Anonymous namespace
+
+ResultCode ProcessCapabilities::InitializeForKernelProcess(const u32* capabilities,
+                                                           std::size_t num_capabilities,
+                                                           Memory::PageTable& page_table) {
+    Clear();
+
+    // Allow all cores and priorities.
+    core_mask = 0xF;
+    priority_mask = 0xFFFFFFFFFFFFFFFF;
+    kernel_version = PackedKernelVersion;
+
+    return ParseCapabilities(capabilities, num_capabilities, page_table);
+}
+
+ResultCode ProcessCapabilities::InitializeForUserProcess(const u32* capabilities,
+                                                         std::size_t num_capabilities,
+                                                         Memory::PageTable& page_table) {
+    Clear();
+
+    return ParseCapabilities(capabilities, num_capabilities, page_table);
+}
+
+void ProcessCapabilities::InitializeForMetadatalessProcess() {
+    // Allow all cores and priorities
+    core_mask = 0xF;
+    priority_mask = 0xFFFFFFFFFFFFFFFF;
+    kernel_version = PackedKernelVersion;
+
+    // Allow all system calls and interrupts.
+    svc_capabilities.set();
+    interrupt_capabilities.set();
+
+    // Allow using the maximum possible amount of handles
+    handle_table_size = static_cast<s32>(HandleTable::MAX_COUNT);
+
+    // Allow all debugging capabilities.
+    is_debuggable = true;
+    can_force_debug = true;
+}
+
+ResultCode ProcessCapabilities::ParseCapabilities(const u32* capabilities,
+                                                  std::size_t num_capabilities,
+                                                  Memory::PageTable& page_table) {
+    u32 set_flags = 0;
+    u32 set_svc_bits = 0;
+
+    for (std::size_t i = 0; i < num_capabilities; ++i) {
+        const u32 descriptor = capabilities[i];
+        const auto type = GetCapabilityType(descriptor);
+
+        if (type == CapabilityType::MapPhysical) {
+            i++;
+
+            // The MapPhysical type uses two descriptor flags for its parameters.
+            // If there's only one, then there's a problem.
+            if (i >= num_capabilities) {
+                LOG_ERROR(Kernel, "Invalid combination! i={}", i);
+                return ERR_INVALID_COMBINATION;
+            }
+
+            const auto size_flags = capabilities[i];
+            if (GetCapabilityType(size_flags) != CapabilityType::MapPhysical) {
+                LOG_ERROR(Kernel, "Invalid capability type! size_flags={}", size_flags);
+                return ERR_INVALID_COMBINATION;
+            }
+
+            const auto result = HandleMapPhysicalFlags(descriptor, size_flags, page_table);
+            if (result.IsError()) {
+                LOG_ERROR(Kernel, "Failed to map physical flags! descriptor={}, size_flags={}",
+                          descriptor, size_flags);
+                return result;
+            }
+        } else {
+            const auto result =
+                ParseSingleFlagCapability(set_flags, set_svc_bits, descriptor, page_table);
+            if (result.IsError()) {
+                LOG_ERROR(
+                    Kernel,
+                    "Failed to parse capability flag! set_flags={}, set_svc_bits={}, descriptor={}",
+                    set_flags, set_svc_bits, descriptor);
+                return result;
+            }
+        }
+    }
+
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::ParseSingleFlagCapability(u32& set_flags, u32& set_svc_bits,
+                                                          u32 flag, Memory::PageTable& page_table) {
+    const auto type = GetCapabilityType(flag);
+
+    if (type == CapabilityType::Unset) {
+        return ERR_INVALID_CAPABILITY_DESCRIPTOR;
+    }
+
+    // Bail early on ignorable entries, as one would expect,
+    // ignorable descriptors can be ignored.
+    if (type == CapabilityType::Ignorable) {
+        return RESULT_SUCCESS;
+    }
+
+    // Ensure that the give flag hasn't already been initialized before.
+    // If it has been, then bail.
+    const u32 flag_length = GetFlagBitOffset(type);
+    const u32 set_flag = 1U << flag_length;
+    if ((set_flag & set_flags & InitializeOnceMask) != 0) {
+        LOG_ERROR(Kernel,
+                  "Attempted to initialize flags that may only be initialized once. set_flags={}",
+                  set_flags);
+        return ERR_INVALID_COMBINATION;
+    }
+    set_flags |= set_flag;
+
+    switch (type) {
+    case CapabilityType::PriorityAndCoreNum:
+        return HandlePriorityCoreNumFlags(flag);
+    case CapabilityType::Syscall:
+        return HandleSyscallFlags(set_svc_bits, flag);
+    case CapabilityType::MapIO:
+        return HandleMapIOFlags(flag, page_table);
+    case CapabilityType::Interrupt:
+        return HandleInterruptFlags(flag);
+    case CapabilityType::ProgramType:
+        return HandleProgramTypeFlags(flag);
+    case CapabilityType::KernelVersion:
+        return HandleKernelVersionFlags(flag);
+    case CapabilityType::HandleTableSize:
+        return HandleHandleTableFlags(flag);
+    case CapabilityType::Debug:
+        return HandleDebugFlags(flag);
+    default:
+        break;
+    }
+
+    LOG_ERROR(Kernel, "Invalid capability type! type={}", type);
+    return ERR_INVALID_CAPABILITY_DESCRIPTOR;
+}
+
+void ProcessCapabilities::Clear() {
+    svc_capabilities.reset();
+    interrupt_capabilities.reset();
+
+    core_mask = 0;
+    priority_mask = 0;
+
+    handle_table_size = 0;
+    kernel_version = 0;
+
+    program_type = ProgramType::SysModule;
+
+    is_debuggable = false;
+    can_force_debug = false;
+}
+
+ResultCode ProcessCapabilities::HandlePriorityCoreNumFlags(u32 flags) {
+    if (priority_mask != 0 || core_mask != 0) {
+        LOG_ERROR(Kernel, "Core or priority mask are not zero! priority_mask={}, core_mask={}",
+                  priority_mask, core_mask);
+        return ERR_INVALID_CAPABILITY_DESCRIPTOR;
+    }
+
+    const u32 core_num_min = (flags >> 16) & 0xFF;
+    const u32 core_num_max = (flags >> 24) & 0xFF;
+    if (core_num_min > core_num_max) {
+        LOG_ERROR(Kernel, "Core min is greater than core max! core_num_min={}, core_num_max={}",
+                  core_num_min, core_num_max);
+        return ERR_INVALID_COMBINATION;
+    }
+
+    const u32 priority_min = (flags >> 10) & 0x3F;
+    const u32 priority_max = (flags >> 4) & 0x3F;
+    if (priority_min > priority_max) {
+        LOG_ERROR(Kernel,
+                  "Priority min is greater than priority max! priority_min={}, priority_max={}",
+                  core_num_min, priority_max);
+        return ERR_INVALID_COMBINATION;
+    }
+
+    // The switch only has 4 usable cores.
+    if (core_num_max >= 4) {
+        LOG_ERROR(Kernel, "Invalid max cores specified! core_num_max={}", core_num_max);
+        return ERR_INVALID_PROCESSOR_ID;
+    }
+
+    const auto make_mask = [](u64 min, u64 max) {
+        const u64 range = max - min + 1;
+        const u64 mask = (1ULL << range) - 1;
+
+        return mask << min;
+    };
+
+    core_mask = make_mask(core_num_min, core_num_max);
+    priority_mask = make_mask(priority_min, priority_max);
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleSyscallFlags(u32& set_svc_bits, u32 flags) {
+    const u32 index = flags >> 29;
+    const u32 svc_bit = 1U << index;
+
+    // If we've already set this svc before, bail.
+    if ((set_svc_bits & svc_bit) != 0) {
+        return ERR_INVALID_COMBINATION;
+    }
+    set_svc_bits |= svc_bit;
+
+    const u32 svc_mask = (flags >> 5) & 0xFFFFFF;
+    for (u32 i = 0; i < 24; ++i) {
+        const u32 svc_number = index * 24 + i;
+
+        if ((svc_mask & (1U << i)) == 0) {
+            continue;
+        }
+
+        if (svc_number >= svc_capabilities.size()) {
+            LOG_ERROR(Kernel, "Process svc capability is out of range! svc_number={}", svc_number);
+            return ERR_OUT_OF_RANGE;
+        }
+
+        svc_capabilities[svc_number] = true;
+    }
+
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleMapPhysicalFlags(u32 flags, u32 size_flags,
+                                                       Memory::PageTable& page_table) {
+    // TODO(Lioncache): Implement once the memory manager can handle this.
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleMapIOFlags(u32 flags, Memory::PageTable& page_table) {
+    // TODO(Lioncache): Implement once the memory manager can handle this.
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleInterruptFlags(u32 flags) {
+    constexpr u32 interrupt_ignore_value = 0x3FF;
+    const u32 interrupt0 = (flags >> 12) & 0x3FF;
+    const u32 interrupt1 = (flags >> 22) & 0x3FF;
+
+    for (u32 interrupt : {interrupt0, interrupt1}) {
+        if (interrupt == interrupt_ignore_value) {
+            continue;
+        }
+
+        // NOTE:
+        // This should be checking a generic interrupt controller value
+        // as part of the calculation, however, given we don't currently
+        // emulate that, it's sufficient to mark every interrupt as defined.
+
+        if (interrupt >= interrupt_capabilities.size()) {
+            LOG_ERROR(Kernel, "Process interrupt capability is out of range! svc_number={}",
+                      interrupt);
+            return ERR_OUT_OF_RANGE;
+        }
+
+        interrupt_capabilities[interrupt] = true;
+    }
+
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleProgramTypeFlags(u32 flags) {
+    const u32 reserved = flags >> 17;
+    if (reserved != 0) {
+        LOG_ERROR(Kernel, "Reserved value is non-zero! reserved={}", reserved);
+        return ERR_RESERVED_VALUE;
+    }
+
+    program_type = static_cast<ProgramType>((flags >> 14) & 0b111);
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleKernelVersionFlags(u32 flags) {
+    // Yes, the internal member variable is checked in the actual kernel here.
+    // This might look odd for options that are only allowed to be initialized
+    // just once, however the kernel has a separate initialization function for
+    // kernel processes and userland processes. The kernel variant sets this
+    // member variable ahead of time.
+
+    const u32 major_version = kernel_version >> 19;
+
+    if (major_version != 0 || flags < 0x80000) {
+        LOG_ERROR(Kernel,
+                  "Kernel version is non zero or flags are too small! major_version={}, flags={}",
+                  major_version, flags);
+        return ERR_INVALID_CAPABILITY_DESCRIPTOR;
+    }
+
+    kernel_version = flags;
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleHandleTableFlags(u32 flags) {
+    const u32 reserved = flags >> 26;
+    if (reserved != 0) {
+        LOG_ERROR(Kernel, "Reserved value is non-zero! reserved={}", reserved);
+        return ERR_RESERVED_VALUE;
+    }
+
+    handle_table_size = static_cast<s32>((flags >> 16) & 0x3FF);
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleDebugFlags(u32 flags) {
+    const u32 reserved = flags >> 19;
+    if (reserved != 0) {
+        LOG_ERROR(Kernel, "Reserved value is non-zero! reserved={}", reserved);
+        return ERR_RESERVED_VALUE;
+    }
+
+    is_debuggable = (flags & 0x20000) != 0;
+    can_force_debug = (flags & 0x40000) != 0;
+    return RESULT_SUCCESS;
+}
+
+} // namespace Kernel

src/core/hle/kernel/process_capability.h

@@ -0,0 +1,266 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <bitset>
+
+#include "common/common_types.h"
+
+union ResultCode;
+
+namespace Kernel {
+
+namespace Memory {
+class PageTable;
+}
+
+/// The possible types of programs that may be indicated
+/// by the program type capability descriptor.
+enum class ProgramType {
+    SysModule,
+    Application,
+    Applet,
+};
+
+/// Handles kernel capability descriptors that are provided by
+/// application metadata. These descriptors provide information
+/// that alters certain parameters for kernel process instance
+/// that will run said application (or applet).
+///
+/// Capabilities are a sequence of flag descriptors, that indicate various
+/// configurations and constraints for a particular process.
+///
+/// Flag types are indicated by a sequence of set low bits. E.g. the
+/// types are indicated with the low bits as follows (where x indicates "don't care"):
+///
+/// - Priority and core mask   : 0bxxxxxxxxxxxx0111
+/// - Allowed service call mask: 0bxxxxxxxxxxx01111
+/// - Map physical memory      : 0bxxxxxxxxx0111111
+/// - Map IO memory            : 0bxxxxxxxx01111111
+/// - Interrupts               : 0bxxxx011111111111
+/// - Application type         : 0bxx01111111111111
+/// - Kernel version           : 0bx011111111111111
+/// - Handle table size        : 0b0111111111111111
+/// - Debugger flags           : 0b1111111111111111
+///
+/// These are essentially a bit offset subtracted by 1 to create a mask.
+/// e.g. The first entry in the above list is simply bit 3 (value 8 -> 0b1000)
+///      subtracted by one (7 -> 0b0111)
+///
+/// An example of a bit layout (using the map physical layout):
+/// <example>
+///   The MapPhysical type indicates a sequence entry pair of:
+///
+///   [initial, memory_flags], where:
+///
+///   initial:
+///     bits:
+///       7-24: Starting page to map memory at.
+///       25  : Indicates if the memory should be mapped as read only.
+///
+///   memory_flags:
+///     bits:
+///       7-20 : Number of pages to map
+///       21-25: Seems to be reserved (still checked against though)
+///       26   : Whether or not the memory being mapped is IO memory, or physical memory
+/// </example>
+///
+class ProcessCapabilities {
+public:
+    using InterruptCapabilities = std::bitset<1024>;
+    using SyscallCapabilities = std::bitset<128>;
+
+    ProcessCapabilities() = default;
+    ProcessCapabilities(const ProcessCapabilities&) = delete;
+    ProcessCapabilities(ProcessCapabilities&&) = default;
+
+    ProcessCapabilities& operator=(const ProcessCapabilities&) = delete;
+    ProcessCapabilities& operator=(ProcessCapabilities&&) = default;
+
+    /// Initializes this process capabilities instance for a kernel process.
+    ///
+    /// @param capabilities     The capabilities to parse
+    /// @param num_capabilities The number of capabilities to parse.
+    /// @param page_table       The memory manager to use for handling any mapping-related
+    ///                         operations (such as mapping IO memory, etc).
+    ///
+    /// @returns RESULT_SUCCESS if this capabilities instance was able to be initialized,
+    ///          otherwise, an error code upon failure.
+    ///
+    ResultCode InitializeForKernelProcess(const u32* capabilities, std::size_t num_capabilities,
+                                          Memory::PageTable& page_table);
+
+    /// Initializes this process capabilities instance for a userland process.
+    ///
+    /// @param capabilities     The capabilities to parse.
+    /// @param num_capabilities The total number of capabilities to parse.
+    /// @param page_table       The memory manager to use for handling any mapping-related
+    ///                         operations (such as mapping IO memory, etc).
+    ///
+    /// @returns RESULT_SUCCESS if this capabilities instance was able to be initialized,
+    ///          otherwise, an error code upon failure.
+    ///
+    ResultCode InitializeForUserProcess(const u32* capabilities, std::size_t num_capabilities,
+                                        Memory::PageTable& page_table);
+
+    /// Initializes this process capabilities instance for a process that does not
+    /// have any metadata to parse.
+    ///
+    /// This is necessary, as we allow running raw executables, and the internal
+    /// kernel process capabilities also determine what CPU cores the process is
+    /// allowed to run on, and what priorities are allowed for  threads. It also
+    /// determines the max handle table size, what the program type is, whether or
+    /// not the process can be debugged, or whether it's possible for a process to
+    /// forcibly debug another process.
+    ///
+    /// Given the above, this essentially enables all capabilities across the board
+    /// for the process. It allows the process to:
+    ///
+    /// - Run on any core
+    /// - Use any thread priority
+    /// - Use the maximum amount of handles a process is allowed to.
+    /// - Be debuggable
+    /// - Forcibly debug other processes.
+    ///
+    /// Note that this is not a behavior that the kernel allows a process to do via
+    /// a single function like this. This is yuzu-specific behavior to handle
+    /// executables with no capability descriptors whatsoever to derive behavior from.
+    /// It being yuzu-specific is why this is also not the default behavior and not
+    /// done by default in the constructor.
+    ///
+    void InitializeForMetadatalessProcess();
+
+    /// Gets the allowable core mask
+    u64 GetCoreMask() const {
+        return core_mask;
+    }
+
+    /// Gets the allowable priority mask
+    u64 GetPriorityMask() const {
+        return priority_mask;
+    }
+
+    /// Gets the SVC access permission bits
+    const SyscallCapabilities& GetServiceCapabilities() const {
+        return svc_capabilities;
+    }
+
+    /// Gets the valid interrupt bits.
+    const InterruptCapabilities& GetInterruptCapabilities() const {
+        return interrupt_capabilities;
+    }
+
+    /// Gets the program type for this process.
+    ProgramType GetProgramType() const {
+        return program_type;
+    }
+
+    /// Gets the number of total allowable handles for the process' handle table.
+    s32 GetHandleTableSize() const {
+        return handle_table_size;
+    }
+
+    /// Gets the kernel version value.
+    u32 GetKernelVersion() const {
+        return kernel_version;
+    }
+
+    /// Whether or not this process can be debugged.
+    bool IsDebuggable() const {
+        return is_debuggable;
+    }
+
+    /// Whether or not this process can forcibly debug another
+    /// process, even if that process is not considered debuggable.
+    bool CanForceDebug() const {
+        return can_force_debug;
+    }
+
+private:
+    /// Attempts to parse a given sequence of capability descriptors.
+    ///
+    /// @param capabilities     The sequence of capability descriptors to parse.
+    /// @param num_capabilities The number of descriptors within the given sequence.
+    /// @param page_table       The memory manager that will perform any memory
+    ///                         mapping if necessary.
+    ///
+    /// @return RESULT_SUCCESS if no errors occur, otherwise an error code.
+    ///
+    ResultCode ParseCapabilities(const u32* capabilities, std::size_t num_capabilities,
+                                 Memory::PageTable& page_table);
+
+    /// Attempts to parse a capability descriptor that is only represented by a
+    /// single flag set.
+    ///
+    /// @param set_flags    Running set of flags that are used to catch
+    ///                     flags being initialized more than once when they shouldn't be.
+    /// @param set_svc_bits Running set of bits representing the allowed supervisor calls mask.
+    /// @param flag         The flag to attempt to parse.
+    /// @param page_table   The memory manager that will perform any memory
+    ///                     mapping if necessary.
+    ///
+    /// @return RESULT_SUCCESS if no errors occurred, otherwise an error code.
+    ///
+    ResultCode ParseSingleFlagCapability(u32& set_flags, u32& set_svc_bits, u32 flag,
+                                         Memory::PageTable& page_table);
+
+    /// Clears the internal state of this process capability instance. Necessary,
+    /// to have a sane starting point due to us allowing running executables without
+    /// configuration metadata. We assume a process is not going to have metadata,
+    /// and if it turns out that the process does, in fact, have metadata, then
+    /// we attempt to parse it. Thus, we need this to reset data members back to
+    /// a good state.
+    ///
+    /// DO NOT ever make this a public member function. This isn't an invariant
+    /// anything external should depend upon (and if anything comes to rely on it,
+    /// you should immediately be questioning the design of that thing, not this
+    /// class. If the kernel itself can run without depending on behavior like that,
+    /// then so can yuzu).
+    ///
+    void Clear();
+
+    /// Handles flags related to the priority and core number capability flags.
+    ResultCode HandlePriorityCoreNumFlags(u32 flags);
+
+    /// Handles flags related to determining the allowable SVC mask.
+    ResultCode HandleSyscallFlags(u32& set_svc_bits, u32 flags);
+
+    /// Handles flags related to mapping physical memory pages.
+    ResultCode HandleMapPhysicalFlags(u32 flags, u32 size_flags, Memory::PageTable& page_table);
+
+    /// Handles flags related to mapping IO pages.
+    ResultCode HandleMapIOFlags(u32 flags, Memory::PageTable& page_table);
+
+    /// Handles flags related to the interrupt capability flags.
+    ResultCode HandleInterruptFlags(u32 flags);
+
+    /// Handles flags related to the program type.
+    ResultCode HandleProgramTypeFlags(u32 flags);
+
+    /// Handles flags related to the handle table size.
+    ResultCode HandleHandleTableFlags(u32 flags);
+
+    /// Handles flags related to the kernel version capability flags.
+    ResultCode HandleKernelVersionFlags(u32 flags);
+
+    /// Handles flags related to debug-specific capabilities.
+    ResultCode HandleDebugFlags(u32 flags);
+
+    SyscallCapabilities svc_capabilities;
+    InterruptCapabilities interrupt_capabilities;
+
+    u64 core_mask = 0;
+    u64 priority_mask = 0;
+
+    s32 handle_table_size = 0;
+    u32 kernel_version = 0;
+
+    ProgramType program_type = ProgramType::SysModule;
+
+    bool is_debuggable = false;
+    bool can_force_debug = false;
+};
+
+} // namespace Kernel

src/core/hle/kernel/readable_event.cpp

@@ -0,0 +1,54 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/readable_event.h"
+#include "core/hle/kernel/thread.h"
+
+namespace Kernel {
+
+ReadableEvent::ReadableEvent(KernelCore& kernel) : SynchronizationObject{kernel} {}
+ReadableEvent::~ReadableEvent() = default;
+
+bool ReadableEvent::ShouldWait(const Thread* thread) const {
+    return !is_signaled;
+}
+
+void ReadableEvent::Acquire(Thread* thread) {
+    ASSERT_MSG(IsSignaled(), "object unavailable!");
+}
+
+void ReadableEvent::Signal() {
+    if (is_signaled) {
+        return;
+    }
+
+    is_signaled = true;
+    SynchronizationObject::Signal();
+}
+
+void ReadableEvent::Clear() {
+    is_signaled = false;
+}
+
+ResultCode ReadableEvent::Reset() {
+    KScopedSchedulerLock lock(kernel);
+    if (!is_signaled) {
+        LOG_TRACE(Kernel, "Handle is not signaled! object_id={}, object_type={}, object_name={}",
+                  GetObjectId(), GetTypeName(), GetName());
+        return ERR_INVALID_STATE;
+    }
+
+    Clear();
+
+    return RESULT_SUCCESS;
+}
+
+} // namespace Kernel

src/core/hle/kernel/readable_event.h

@@ -0,0 +1,57 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/synchronization_object.h"
+
+union ResultCode;
+
+namespace Kernel {
+
+class KernelCore;
+class WritableEvent;
+
+class ReadableEvent final : public SynchronizationObject {
+    friend class WritableEvent;
+
+public:
+    ~ReadableEvent() override;
+
+    std::string GetTypeName() const override {
+        return "ReadableEvent";
+    }
+    std::string GetName() const override {
+        return name;
+    }
+
+    static constexpr HandleType HANDLE_TYPE = HandleType::ReadableEvent;
+    HandleType GetHandleType() const override {
+        return HANDLE_TYPE;
+    }
+
+    bool ShouldWait(const Thread* thread) const override;
+    void Acquire(Thread* thread) override;
+
+    /// Unconditionally clears the readable event's state.
+    void Clear();
+
+    /// Clears the readable event's state if and only if it
+    /// has already been signaled.
+    ///
+    /// @pre The event must be in a signaled state. If this event
+    ///      is in an unsignaled state and this function is called,
+    ///      then ERR_INVALID_STATE will be returned.
+    ResultCode Reset();
+
+    void Signal() override;
+
+private:
+    explicit ReadableEvent(KernelCore& kernel);
+
+    std::string name; ///< Name of event (optional)
+};
+
+} // namespace Kernel

src/core/hle/kernel/resource_limit.cpp

@@ -0,0 +1,73 @@
+// Copyright 2015 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/resource_limit.h"
+#include "core/hle/result.h"
+
+namespace Kernel {
+namespace {
+constexpr std::size_t ResourceTypeToIndex(ResourceType type) {
+    return static_cast<std::size_t>(type);
+}
+} // Anonymous namespace
+
+ResourceLimit::ResourceLimit(KernelCore& kernel) : Object{kernel} {}
+ResourceLimit::~ResourceLimit() = default;
+
+bool ResourceLimit::Reserve(ResourceType resource, s64 amount) {
+    return Reserve(resource, amount, 10000000000);
+}
+
+bool ResourceLimit::Reserve(ResourceType resource, s64 amount, u64 timeout) {
+    const std::size_t index{ResourceTypeToIndex(resource)};
+
+    s64 new_value = current[index] + amount;
+    if (new_value > limit[index] && available[index] + amount <= limit[index]) {
+        // TODO(bunnei): This is wrong for multicore, we should wait the calling thread for timeout
+        new_value = current[index] + amount;
+    }
+
+    if (new_value <= limit[index]) {
+        current[index] = new_value;
+        return true;
+    }
+    return false;
+}
+
+void ResourceLimit::Release(ResourceType resource, u64 amount) {
+    Release(resource, amount, amount);
+}
+
+void ResourceLimit::Release(ResourceType resource, u64 used_amount, u64 available_amount) {
+    const std::size_t index{ResourceTypeToIndex(resource)};
+
+    current[index] -= used_amount;
+    available[index] -= available_amount;
+}
+
+std::shared_ptr<ResourceLimit> ResourceLimit::Create(KernelCore& kernel) {
+    return std::make_shared<ResourceLimit>(kernel);
+}
+
+s64 ResourceLimit::GetCurrentResourceValue(ResourceType resource) const {
+    return limit.at(ResourceTypeToIndex(resource)) - current.at(ResourceTypeToIndex(resource));
+}
+
+s64 ResourceLimit::GetMaxResourceValue(ResourceType resource) const {
+    return limit.at(ResourceTypeToIndex(resource));
+}
+
+ResultCode ResourceLimit::SetLimitValue(ResourceType resource, s64 value) {
+    const std::size_t index{ResourceTypeToIndex(resource)};
+    if (current[index] <= value) {
+        limit[index] = value;
+        return RESULT_SUCCESS;
+    } else {
+        LOG_ERROR(Kernel, "Limit value is too large! resource={}, value={}, index={}", resource,
+                  value, index);
+        return ERR_INVALID_STATE;
+    }
+}
+} // namespace Kernel

src/core/hle/kernel/resource_limit.h

@@ -0,0 +1,104 @@
+// Copyright 2015 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <memory>
+
+#include "common/common_types.h"
+#include "core/hle/kernel/object.h"
+
+union ResultCode;
+
+namespace Kernel {
+
+class KernelCore;
+
+enum class ResourceType : u32 {
+    PhysicalMemory,
+    Threads,
+    Events,
+    TransferMemory,
+    Sessions,
+
+    // Used as a count, not an actual type.
+    ResourceTypeCount
+};
+
+constexpr bool IsValidResourceType(ResourceType type) {
+    return type < ResourceType::ResourceTypeCount;
+}
+
+class ResourceLimit final : public Object {
+public:
+    explicit ResourceLimit(KernelCore& kernel);
+    ~ResourceLimit() override;
+
+    /// Creates a resource limit object.
+    static std::shared_ptr<ResourceLimit> Create(KernelCore& kernel);
+
+    std::string GetTypeName() const override {
+        return "ResourceLimit";
+    }
+    std::string GetName() const override {
+        return GetTypeName();
+    }
+
+    static constexpr HandleType HANDLE_TYPE = HandleType::ResourceLimit;
+    HandleType GetHandleType() const override {
+        return HANDLE_TYPE;
+    }
+
+    bool Reserve(ResourceType resource, s64 amount);
+    bool Reserve(ResourceType resource, s64 amount, u64 timeout);
+    void Release(ResourceType resource, u64 amount);
+    void Release(ResourceType resource, u64 used_amount, u64 available_amount);
+
+    /**
+     * Gets the current value for the specified resource.
+     * @param resource Requested resource type
+     * @returns The current value of the resource type
+     */
+    s64 GetCurrentResourceValue(ResourceType resource) const;
+
+    /**
+     * Gets the max value for the specified resource.
+     * @param resource Requested resource type
+     * @returns The max value of the resource type
+     */
+    s64 GetMaxResourceValue(ResourceType resource) const;
+
+    /**
+     * Sets the limit value for a given resource type.
+     *
+     * @param resource The resource type to apply the limit to.
+     * @param value    The limit to apply to the given resource type.
+     *
+     * @return A result code indicating if setting the limit value
+     *         was successful or not.
+     *
+     * @note The supplied limit value *must* be greater than or equal to
+     *       the current resource value for the given resource type,
+     *       otherwise ERR_INVALID_STATE will be returned.
+     */
+    ResultCode SetLimitValue(ResourceType resource, s64 value);
+
+private:
+    // TODO(Subv): Increment resource limit current values in their respective Kernel::T::Create
+    // functions
+    //
+    // Currently we have no way of distinguishing if a Create was called by the running application,
+    // or by a service module. Approach this once we have separated the service modules into their
+    // own processes
+
+    using ResourceArray =
+        std::array<s64, static_cast<std::size_t>(ResourceType::ResourceTypeCount)>;
+
+    ResourceArray limit{};
+    ResourceArray current{};
+    ResourceArray available{};
+};
+
+} // namespace Kernel

src/core/hle/kernel/server_port.cpp

@@ -0,0 +1,60 @@
+// Copyright 2016 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <tuple>
+#include "common/assert.h"
+#include "core/hle/kernel/client_port.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/server_port.h"
+#include "core/hle/kernel/server_session.h"
+#include "core/hle/kernel/thread.h"
+
+namespace Kernel {
+
+ServerPort::ServerPort(KernelCore& kernel) : SynchronizationObject{kernel} {}
+ServerPort::~ServerPort() = default;
+
+ResultVal<std::shared_ptr<ServerSession>> ServerPort::Accept() {
+    if (pending_sessions.empty()) {
+        return ERR_NOT_FOUND;
+    }
+
+    auto session = std::move(pending_sessions.back());
+    pending_sessions.pop_back();
+    return MakeResult(std::move(session));
+}
+
+void ServerPort::AppendPendingSession(std::shared_ptr<ServerSession> pending_session) {
+    pending_sessions.push_back(std::move(pending_session));
+}
+
+bool ServerPort::ShouldWait(const Thread* thread) const {
+    // If there are no pending sessions, we wait until a new one is added.
+    return pending_sessions.empty();
+}
+
+void ServerPort::Acquire(Thread* thread) {
+    ASSERT_MSG(!ShouldWait(thread), "object unavailable!");
+}
+
+bool ServerPort::IsSignaled() const {
+    return !pending_sessions.empty();
+}
+
+ServerPort::PortPair ServerPort::CreatePortPair(KernelCore& kernel, u32 max_sessions,
+                                                std::string name) {
+    std::shared_ptr<ServerPort> server_port = std::make_shared<ServerPort>(kernel);
+    std::shared_ptr<ClientPort> client_port = std::make_shared<ClientPort>(kernel);
+
+    server_port->name = name + "_Server";
+    client_port->name = name + "_Client";
+    client_port->server_port = server_port;
+    client_port->max_sessions = max_sessions;
+    client_port->active_sessions = 0;
+
+    return std::make_pair(std::move(server_port), std::move(client_port));
+}
+
+} // namespace Kernel

src/core/hle/kernel/server_port.h

@@ -0,0 +1,99 @@
+// Copyright 2016 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <string>
+#include <utility>
+#include <vector>
+#include "common/common_types.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/synchronization_object.h"
+#include "core/hle/result.h"
+
+namespace Kernel {
+
+class ClientPort;
+class KernelCore;
+class ServerSession;
+class SessionRequestHandler;
+
+class ServerPort final : public SynchronizationObject {
+public:
+    explicit ServerPort(KernelCore& kernel);
+    ~ServerPort() override;
+
+    using HLEHandler = std::shared_ptr<SessionRequestHandler>;
+    using PortPair = std::pair<std::shared_ptr<ServerPort>, std::shared_ptr<ClientPort>>;
+
+    /**
+     * Creates a pair of ServerPort and an associated ClientPort.
+     *
+     * @param kernel The kernel instance to create the port pair under.
+     * @param max_sessions Maximum number of sessions to the port
+     * @param name Optional name of the ports
+     * @return The created port tuple
+     */
+    static PortPair CreatePortPair(KernelCore& kernel, u32 max_sessions,
+                                   std::string name = "UnknownPort");
+
+    std::string GetTypeName() const override {
+        return "ServerPort";
+    }
+    std::string GetName() const override {
+        return name;
+    }
+
+    static constexpr HandleType HANDLE_TYPE = HandleType::ServerPort;
+    HandleType GetHandleType() const override {
+        return HANDLE_TYPE;
+    }
+
+    /**
+     * Accepts a pending incoming connection on this port. If there are no pending sessions, will
+     * return ERR_NO_PENDING_SESSIONS.
+     */
+    ResultVal<std::shared_ptr<ServerSession>> Accept();
+
+    /// Whether or not this server port has an HLE handler available.
+    bool HasHLEHandler() const {
+        return hle_handler != nullptr;
+    }
+
+    /// Gets the HLE handler for this port.
+    HLEHandler GetHLEHandler() const {
+        return hle_handler;
+    }
+
+    /**
+     * Sets the HLE handler template for the port. ServerSessions crated by connecting to this port
+     * will inherit a reference to this handler.
+     */
+    void SetHleHandler(HLEHandler hle_handler_) {
+        hle_handler = std::move(hle_handler_);
+    }
+
+    /// Appends a ServerSession to the collection of ServerSessions
+    /// waiting to be accepted by this port.
+    void AppendPendingSession(std::shared_ptr<ServerSession> pending_session);
+
+    bool ShouldWait(const Thread* thread) const override;
+    void Acquire(Thread* thread) override;
+
+    bool IsSignaled() const override;
+
+private:
+    /// ServerSessions waiting to be accepted by the port
+    std::vector<std::shared_ptr<ServerSession>> pending_sessions;
+
+    /// This session's HLE request handler template (optional)
+    /// ServerSessions created from this port inherit a reference to this handler.
+    HLEHandler hle_handler;
+
+    /// Name of the port (optional)
+    std::string name;
+};
+
+} // namespace Kernel

src/core/hle/kernel/server_session.cpp

@@ -0,0 +1,189 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <tuple>
+#include <utility>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "common/logging/log.h"
+#include "core/core_timing.h"
+#include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/client_port.h"
+#include "core/hle/kernel/client_session.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/hle_ipc.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/kernel/server_session.h"
+#include "core/hle/kernel/session.h"
+#include "core/hle/kernel/thread.h"
+#include "core/memory.h"
+
+namespace Kernel {
+
+ServerSession::ServerSession(KernelCore& kernel) : SynchronizationObject{kernel} {}
+
+ServerSession::~ServerSession() {
+    kernel.ReleaseServiceThread(service_thread);
+}
+
+ResultVal<std::shared_ptr<ServerSession>> ServerSession::Create(KernelCore& kernel,
+                                                                std::shared_ptr<Session> parent,
+                                                                std::string name) {
+    std::shared_ptr<ServerSession> session{std::make_shared<ServerSession>(kernel)};
+
+    session->name = std::move(name);
+    session->parent = std::move(parent);
+    session->service_thread = kernel.CreateServiceThread(session->name);
+
+    return MakeResult(std::move(session));
+}
+
+bool ServerSession::ShouldWait(const Thread* thread) const {
+    // Closed sessions should never wait, an error will be returned from svcReplyAndReceive.
+    if (!parent->Client()) {
+        return false;
+    }
+
+    // Wait if we have no pending requests, or if we're currently handling a request.
+    return pending_requesting_threads.empty() || currently_handling != nullptr;
+}
+
+bool ServerSession::IsSignaled() const {
+    // Closed sessions should never wait, an error will be returned from svcReplyAndReceive.
+    if (!parent->Client()) {
+        return true;
+    }
+
+    // Wait if we have no pending requests, or if we're currently handling a request.
+    return !pending_requesting_threads.empty() && currently_handling == nullptr;
+}
+
+void ServerSession::Acquire(Thread* thread) {
+    ASSERT_MSG(!ShouldWait(thread), "object unavailable!");
+    // We are now handling a request, pop it from the stack.
+    // TODO(Subv): What happens if the client endpoint is closed before any requests are made?
+    ASSERT(!pending_requesting_threads.empty());
+    currently_handling = pending_requesting_threads.back();
+    pending_requesting_threads.pop_back();
+}
+
+void ServerSession::ClientDisconnected() {
+    // We keep a shared pointer to the hle handler to keep it alive throughout
+    // the call to ClientDisconnected, as ClientDisconnected invalidates the
+    // hle_handler member itself during the course of the function executing.
+    std::shared_ptr<SessionRequestHandler> handler = hle_handler;
+    if (handler) {
+        // Note that after this returns, this server session's hle_handler is
+        // invalidated (set to null).
+        handler->ClientDisconnected(SharedFrom(this));
+    }
+
+    // Clean up the list of client threads with pending requests, they are unneeded now that the
+    // client endpoint is closed.
+    pending_requesting_threads.clear();
+    currently_handling = nullptr;
+}
+
+void ServerSession::AppendDomainRequestHandler(std::shared_ptr<SessionRequestHandler> handler) {
+    domain_request_handlers.push_back(std::move(handler));
+}
+
+std::size_t ServerSession::NumDomainRequestHandlers() const {
+    return domain_request_handlers.size();
+}
+
+ResultCode ServerSession::HandleDomainSyncRequest(Kernel::HLERequestContext& context) {
+    if (!context.HasDomainMessageHeader()) {
+        return RESULT_SUCCESS;
+    }
+
+    // Set domain handlers in HLE context, used for domain objects (IPC interfaces) as inputs
+    context.SetDomainRequestHandlers(domain_request_handlers);
+
+    // If there is a DomainMessageHeader, then this is CommandType "Request"
+    const auto& domain_message_header = context.GetDomainMessageHeader();
+    const u32 object_id{domain_message_header.object_id};
+    switch (domain_message_header.command) {
+    case IPC::DomainMessageHeader::CommandType::SendMessage:
+        if (object_id > domain_request_handlers.size()) {
+            LOG_CRITICAL(IPC,
+                         "object_id {} is too big! This probably means a recent service call "
+                         "to {} needed to return a new interface!",
+                         object_id, name);
+            UNREACHABLE();
+            return RESULT_SUCCESS; // Ignore error if asserts are off
+        }
+        return domain_request_handlers[object_id - 1]->HandleSyncRequest(context);
+
+    case IPC::DomainMessageHeader::CommandType::CloseVirtualHandle: {
+        LOG_DEBUG(IPC, "CloseVirtualHandle, object_id=0x{:08X}", object_id);
+
+        domain_request_handlers[object_id - 1] = nullptr;
+
+        IPC::ResponseBuilder rb{context, 2};
+        rb.Push(RESULT_SUCCESS);
+        return RESULT_SUCCESS;
+    }
+    }
+
+    LOG_CRITICAL(IPC, "Unknown domain command={}", domain_message_header.command.Value());
+    ASSERT(false);
+    return RESULT_SUCCESS;
+}
+
+ResultCode ServerSession::QueueSyncRequest(std::shared_ptr<Thread> thread,
+                                           Core::Memory::Memory& memory) {
+    u32* cmd_buf{reinterpret_cast<u32*>(memory.GetPointer(thread->GetTLSAddress()))};
+    auto context =
+        std::make_shared<HLERequestContext>(kernel, memory, SharedFrom(this), std::move(thread));
+
+    context->PopulateFromIncomingCommandBuffer(kernel.CurrentProcess()->GetHandleTable(), cmd_buf);
+
+    if (auto strong_ptr = service_thread.lock()) {
+        strong_ptr->QueueSyncRequest(*this, std::move(context));
+        return RESULT_SUCCESS;
+    }
+
+    return RESULT_SUCCESS;
+}
+
+ResultCode ServerSession::CompleteSyncRequest(HLERequestContext& context) {
+    ResultCode result = RESULT_SUCCESS;
+    // If the session has been converted to a domain, handle the domain request
+    if (IsDomain() && context.HasDomainMessageHeader()) {
+        result = HandleDomainSyncRequest(context);
+        // If there is no domain header, the regular session handler is used
+    } else if (hle_handler != nullptr) {
+        // If this ServerSession has an associated HLE handler, forward the request to it.
+        result = hle_handler->HandleSyncRequest(context);
+    }
+
+    if (convert_to_domain) {
+        ASSERT_MSG(IsSession(), "ServerSession is already a domain instance.");
+        domain_request_handlers = {hle_handler};
+        convert_to_domain = false;
+    }
+
+    // Some service requests require the thread to block
+    {
+        KScopedSchedulerLock lock(kernel);
+        if (!context.IsThreadWaiting()) {
+            context.GetThread().ResumeFromWait();
+            context.GetThread().SetSynchronizationResults(nullptr, result);
+        }
+    }
+
+    return result;
+}
+
+ResultCode ServerSession::HandleSyncRequest(std::shared_ptr<Thread> thread,
+                                            Core::Memory::Memory& memory,
+                                            Core::Timing::CoreTiming& core_timing) {
+    return QueueSyncRequest(std::move(thread), memory);
+}
+
+} // namespace Kernel

src/core/hle/kernel/server_session.h

@@ -0,0 +1,173 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <string>
+#include <utility>
+#include <vector>
+
+#include "common/threadsafe_queue.h"
+#include "core/hle/kernel/service_thread.h"
+#include "core/hle/kernel/synchronization_object.h"
+#include "core/hle/result.h"
+
+namespace Core::Memory {
+class Memory;
+}
+
+namespace Core::Timing {
+class CoreTiming;
+struct EventType;
+} // namespace Core::Timing
+
+namespace Kernel {
+
+class HLERequestContext;
+class KernelCore;
+class Session;
+class SessionRequestHandler;
+class Thread;
+
+/**
+ * Kernel object representing the server endpoint of an IPC session. Sessions are the basic CTR-OS
+ * primitive for communication between different processes, and are used to implement service calls
+ * to the various system services.
+ *
+ * To make a service call, the client must write the command header and parameters to the buffer
+ * located at offset 0x80 of the TLS (Thread-Local Storage) area, then execute a SendSyncRequest
+ * SVC call with its ClientSession handle. The kernel will read the command header, using it to
+ * marshall the parameters to the process at the server endpoint of the session.
+ * After the server replies to the request, the response is marshalled back to the caller's
+ * TLS buffer and control is transferred back to it.
+ */
+class ServerSession final : public SynchronizationObject {
+    friend class ServiceThread;
+
+public:
+    explicit ServerSession(KernelCore& kernel);
+    ~ServerSession() override;
+
+    friend class Session;
+
+    static ResultVal<std::shared_ptr<ServerSession>> Create(KernelCore& kernel,
+                                                            std::shared_ptr<Session> parent,
+                                                            std::string name = "Unknown");
+
+    std::string GetTypeName() const override {
+        return "ServerSession";
+    }
+
+    std::string GetName() const override {
+        return name;
+    }
+
+    static constexpr HandleType HANDLE_TYPE = HandleType::ServerSession;
+    HandleType GetHandleType() const override {
+        return HANDLE_TYPE;
+    }
+
+    Session* GetParent() {
+        return parent.get();
+    }
+
+    const Session* GetParent() const {
+        return parent.get();
+    }
+
+    bool IsSignaled() const override;
+
+    /**
+     * Sets the HLE handler for the session. This handler will be called to service IPC requests
+     * instead of the regular IPC machinery. (The regular IPC machinery is currently not
+     * implemented.)
+     */
+    void SetHleHandler(std::shared_ptr<SessionRequestHandler> hle_handler_) {
+        hle_handler = std::move(hle_handler_);
+    }
+
+    /**
+     * Handle a sync request from the emulated application.
+     *
+     * @param thread      Thread that initiated the request.
+     * @param memory      Memory context to handle the sync request under.
+     * @param core_timing Core timing context to schedule the request event under.
+     *
+     * @returns ResultCode from the operation.
+     */
+    ResultCode HandleSyncRequest(std::shared_ptr<Thread> thread, Core::Memory::Memory& memory,
+                                 Core::Timing::CoreTiming& core_timing);
+
+    bool ShouldWait(const Thread* thread) const override;
+
+    void Acquire(Thread* thread) override;
+
+    /// Called when a client disconnection occurs.
+    void ClientDisconnected();
+
+    /// Adds a new domain request handler to the collection of request handlers within
+    /// this ServerSession instance.
+    void AppendDomainRequestHandler(std::shared_ptr<SessionRequestHandler> handler);
+
+    /// Retrieves the total number of domain request handlers that have been
+    /// appended to this ServerSession instance.
+    std::size_t NumDomainRequestHandlers() const;
+
+    /// Returns true if the session has been converted to a domain, otherwise False
+    bool IsDomain() const {
+        return !IsSession();
+    }
+
+    /// Returns true if this session has not been converted to a domain, otherwise false.
+    bool IsSession() const {
+        return domain_request_handlers.empty();
+    }
+
+    /// Converts the session to a domain at the end of the current command
+    void ConvertToDomain() {
+        convert_to_domain = true;
+    }
+
+private:
+    /// Queues a sync request from the emulated application.
+    ResultCode QueueSyncRequest(std::shared_ptr<Thread> thread, Core::Memory::Memory& memory);
+
+    /// Completes a sync request from the emulated application.
+    ResultCode CompleteSyncRequest(HLERequestContext& context);
+
+    /// Handles a SyncRequest to a domain, forwarding the request to the proper object or closing an
+    /// object handle.
+    ResultCode HandleDomainSyncRequest(Kernel::HLERequestContext& context);
+
+    /// The parent session, which links to the client endpoint.
+    std::shared_ptr<Session> parent;
+
+    /// This session's HLE request handler (applicable when not a domain)
+    std::shared_ptr<SessionRequestHandler> hle_handler;
+
+    /// This is the list of domain request handlers (after conversion to a domain)
+    std::vector<std::shared_ptr<SessionRequestHandler>> domain_request_handlers;
+
+    /// List of threads that are pending a response after a sync request. This list is processed in
+    /// a LIFO manner, thus, the last request will be dispatched first.
+    /// TODO(Subv): Verify if this is indeed processed in LIFO using a hardware test.
+    std::vector<std::shared_ptr<Thread>> pending_requesting_threads;
+
+    /// Thread whose request is currently being handled. A request is considered "handled" when a
+    /// response is sent via svcReplyAndReceive.
+    /// TODO(Subv): Find a better name for this.
+    std::shared_ptr<Thread> currently_handling;
+
+    /// When set to True, converts the session to a domain at the end of the command
+    bool convert_to_domain{};
+
+    /// The name of this session (optional)
+    std::string name;
+
+    /// Thread to dispatch service requests
+    std::weak_ptr<ServiceThread> service_thread;
+};
+
+} // namespace Kernel

src/core/hle/kernel/service_thread.cpp

@@ -0,0 +1,106 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <condition_variable>
+#include <functional>
+#include <mutex>
+#include <thread>
+#include <vector>
+#include <queue>
+
+#include "common/assert.h"
+#include "common/scope_exit.h"
+#include "common/thread.h"
+#include "core/core.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/server_session.h"
+#include "core/hle/kernel/service_thread.h"
+#include "core/hle/lock.h"
+#include "video_core/renderer_base.h"
+
+namespace Kernel {
+
+class ServiceThread::Impl final {
+public:
+    explicit Impl(KernelCore& kernel, std::size_t num_threads, const std::string& name);
+    ~Impl();
+
+    void QueueSyncRequest(ServerSession& session, std::shared_ptr<HLERequestContext>&& context);
+
+private:
+    std::vector<std::thread> threads;
+    std::queue<std::function<void()>> requests;
+    std::mutex queue_mutex;
+    std::condition_variable condition;
+    const std::string service_name;
+    bool stop{};
+};
+
+ServiceThread::Impl::Impl(KernelCore& kernel, std::size_t num_threads, const std::string& name)
+    : service_name{name} {
+
+    for (std::size_t i = 0; i < num_threads; ++i)
+        threads.emplace_back([this, &kernel] {
+            Common::SetCurrentThreadName(std::string{"Hle_" + service_name}.c_str());
+
+            // Wait for first request before trying to acquire a render context
+            {
+                std::unique_lock lock{queue_mutex};
+                condition.wait(lock, [this] { return stop || !requests.empty(); });
+            }
+
+            kernel.RegisterHostThread();
+
+            while (true) {
+                std::function<void()> task;
+
+                {
+                    std::unique_lock lock{queue_mutex};
+                    condition.wait(lock, [this] { return stop || !requests.empty(); });
+                    if (stop || requests.empty()) {
+                        return;
+                    }
+                    task = std::move(requests.front());
+                    requests.pop();
+                }
+
+                task();
+            }
+        });
+}
+
+void ServiceThread::Impl::QueueSyncRequest(ServerSession& session,
+                                           std::shared_ptr<HLERequestContext>&& context) {
+    {
+        std::unique_lock lock{queue_mutex};
+        requests.emplace([session{SharedFrom(&session)}, context{std::move(context)}]() {
+            session->CompleteSyncRequest(*context);
+            return;
+        });
+    }
+    condition.notify_one();
+}
+
+ServiceThread::Impl::~Impl() {
+    {
+        std::unique_lock lock{queue_mutex};
+        stop = true;
+    }
+    condition.notify_all();
+    for (std::thread& thread : threads) {
+        thread.join();
+    }
+}
+
+ServiceThread::ServiceThread(KernelCore& kernel, std::size_t num_thread, const std::string& name)
+    : impl{std::make_unique<Impl>(kernel, num_thread, name)} {}
+
+ServiceThread::~ServiceThread() = default;
+
+void ServiceThread::QueueSyncRequest(ServerSession& session,
+                                     std::shared_ptr<HLERequestContext>&& context) {
+    impl->QueueSyncRequest(session, std::move(context));
+}
+
+} // namespace Kernel

src/core/hle/kernel/service_thread.h

@@ -0,0 +1,28 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <string>
+
+namespace Kernel {
+
+class HLERequestContext;
+class KernelCore;
+class ServerSession;
+
+class ServiceThread final {
+public:
+    explicit ServiceThread(KernelCore& kernel, std::size_t num_threads, const std::string& name);
+    ~ServiceThread();
+
+    void QueueSyncRequest(ServerSession& session, std::shared_ptr<HLERequestContext>&& context);
+
+private:
+    class Impl;
+    std::unique_ptr<Impl> impl;
+};
+
+} // namespace Kernel

src/core/hle/kernel/session.cpp

@@ -0,0 +1,41 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "core/hle/kernel/client_session.h"
+#include "core/hle/kernel/server_session.h"
+#include "core/hle/kernel/session.h"
+
+namespace Kernel {
+
+Session::Session(KernelCore& kernel) : SynchronizationObject{kernel} {}
+Session::~Session() = default;
+
+Session::SessionPair Session::Create(KernelCore& kernel, std::string name) {
+    auto session{std::make_shared<Session>(kernel)};
+    auto client_session{Kernel::ClientSession::Create(kernel, session, name + "_Client").Unwrap()};
+    auto server_session{Kernel::ServerSession::Create(kernel, session, name + "_Server").Unwrap()};
+
+    session->name = std::move(name);
+    session->client = client_session;
+    session->server = server_session;
+
+    return std::make_pair(std::move(client_session), std::move(server_session));
+}
+
+bool Session::ShouldWait(const Thread* thread) const {
+    UNIMPLEMENTED();
+    return {};
+}
+
+bool Session::IsSignaled() const {
+    UNIMPLEMENTED();
+    return true;
+}
+
+void Session::Acquire(Thread* thread) {
+    UNIMPLEMENTED();
+}
+
+} // namespace Kernel

src/core/hle/kernel/session.h

@@ -0,0 +1,66 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <string>
+#include <utility>
+
+#include "core/hle/kernel/synchronization_object.h"
+
+namespace Kernel {
+
+class ClientSession;
+class ServerSession;
+
+/**
+ * Parent structure to link the client and server endpoints of a session with their associated
+ * client port.
+ */
+class Session final : public SynchronizationObject {
+public:
+    explicit Session(KernelCore& kernel);
+    ~Session() override;
+
+    using SessionPair = std::pair<std::shared_ptr<ClientSession>, std::shared_ptr<ServerSession>>;
+
+    static SessionPair Create(KernelCore& kernel, std::string name = "Unknown");
+
+    std::string GetName() const override {
+        return name;
+    }
+
+    static constexpr HandleType HANDLE_TYPE = HandleType::Session;
+    HandleType GetHandleType() const override {
+        return HANDLE_TYPE;
+    }
+
+    bool ShouldWait(const Thread* thread) const override;
+
+    bool IsSignaled() const override;
+
+    void Acquire(Thread* thread) override;
+
+    std::shared_ptr<ClientSession> Client() {
+        if (auto result{client.lock()}) {
+            return result;
+        }
+        return {};
+    }
+
+    std::shared_ptr<ServerSession> Server() {
+        if (auto result{server.lock()}) {
+            return result;
+        }
+        return {};
+    }
+
+private:
+    std::string name;
+    std::weak_ptr<ClientSession> client;
+    std::weak_ptr<ServerSession> server;
+};
+
+} // namespace Kernel

src/core/hle/kernel/shared_memory.cpp

@@ -0,0 +1,57 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "core/core.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/memory/page_table.h"
+#include "core/hle/kernel/shared_memory.h"
+
+namespace Kernel {
+
+SharedMemory::SharedMemory(KernelCore& kernel, Core::DeviceMemory& device_memory)
+    : Object{kernel}, device_memory{device_memory} {}
+
+SharedMemory::~SharedMemory() = default;
+
+std::shared_ptr<SharedMemory> SharedMemory::Create(
+    KernelCore& kernel, Core::DeviceMemory& device_memory, Process* owner_process,
+    Memory::PageLinkedList&& page_list, Memory::MemoryPermission owner_permission,
+    Memory::MemoryPermission user_permission, PAddr physical_address, std::size_t size,
+    std::string name) {
+
+    std::shared_ptr<SharedMemory> shared_memory{
+        std::make_shared<SharedMemory>(kernel, device_memory)};
+
+    shared_memory->owner_process = owner_process;
+    shared_memory->page_list = std::move(page_list);
+    shared_memory->owner_permission = owner_permission;
+    shared_memory->user_permission = user_permission;
+    shared_memory->physical_address = physical_address;
+    shared_memory->size = size;
+    shared_memory->name = name;
+
+    return shared_memory;
+}
+
+ResultCode SharedMemory::Map(Process& target_process, VAddr address, std::size_t size,
+                             Memory::MemoryPermission permissions) {
+    const u64 page_count{(size + Memory::PageSize - 1) / Memory::PageSize};
+
+    if (page_list.GetNumPages() != page_count) {
+        UNIMPLEMENTED_MSG("Page count does not match");
+    }
+
+    const Memory::MemoryPermission expected =
+        &target_process == owner_process ? owner_permission : user_permission;
+
+    if (permissions != expected) {
+        UNIMPLEMENTED_MSG("Permission does not match");
+    }
+
+    return target_process.PageTable().MapPages(address, page_list, Memory::MemoryState::Shared,
+                                               permissions);
+}
+
+} // namespace Kernel

src/core/hle/kernel/shared_memory.h

@@ -0,0 +1,85 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <string>
+
+#include "common/common_types.h"
+#include "core/device_memory.h"
+#include "core/hle/kernel/memory/memory_block.h"
+#include "core/hle/kernel/memory/page_linked_list.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/result.h"
+
+namespace Kernel {
+
+class KernelCore;
+
+class SharedMemory final : public Object {
+public:
+    explicit SharedMemory(KernelCore& kernel, Core::DeviceMemory& device_memory);
+    ~SharedMemory() override;
+
+    static std::shared_ptr<SharedMemory> Create(
+        KernelCore& kernel, Core::DeviceMemory& device_memory, Process* owner_process,
+        Memory::PageLinkedList&& page_list, Memory::MemoryPermission owner_permission,
+        Memory::MemoryPermission user_permission, PAddr physical_address, std::size_t size,
+        std::string name);
+
+    std::string GetTypeName() const override {
+        return "SharedMemory";
+    }
+
+    std::string GetName() const override {
+        return name;
+    }
+
+    static constexpr HandleType HANDLE_TYPE = HandleType::SharedMemory;
+    HandleType GetHandleType() const override {
+        return HANDLE_TYPE;
+    }
+
+    /**
+     * Maps a shared memory block to an address in the target process' address space
+     * @param target_process Process on which to map the memory block
+     * @param address Address in system memory to map shared memory block to
+     * @param size Size of the shared memory block to map
+     * @param permissions Memory block map permissions (specified by SVC field)
+     */
+    ResultCode Map(Process& target_process, VAddr address, std::size_t size,
+                   Memory::MemoryPermission permissions);
+
+    /**
+     * Gets a pointer to the shared memory block
+     * @param offset Offset from the start of the shared memory block to get pointer
+     * @return A pointer to the shared memory block from the specified offset
+     */
+    u8* GetPointer(std::size_t offset = 0) {
+        return device_memory.GetPointer(physical_address + offset);
+    }
+
+    /**
+     * Gets a pointer to the shared memory block
+     * @param offset Offset from the start of the shared memory block to get pointer
+     * @return A pointer to the shared memory block from the specified offset
+     */
+    const u8* GetPointer(std::size_t offset = 0) const {
+        return device_memory.GetPointer(physical_address + offset);
+    }
+
+private:
+    Core::DeviceMemory& device_memory;
+    Process* owner_process{};
+    Memory::PageLinkedList page_list;
+    Memory::MemoryPermission owner_permission{};
+    Memory::MemoryPermission user_permission{};
+    PAddr physical_address{};
+    std::size_t size{};
+    std::string name;
+};
+
+} // namespace Kernel

src/core/hle/kernel/svc.h

@@ -0,0 +1,17 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Core {
+class System;
+}
+
+namespace Kernel::Svc {
+
+void Call(Core::System& system, u32 immediate);
+
+} // namespace Kernel::Svc

src/core/hle/kernel/svc_types.h

@@ -0,0 +1,68 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+
+namespace Kernel::Svc {
+
+enum class MemoryState : u32 {
+    Free = 0x00,
+    Io = 0x01,
+    Static = 0x02,
+    Code = 0x03,
+    CodeData = 0x04,
+    Normal = 0x05,
+    Shared = 0x06,
+    Alias = 0x07,
+    AliasCode = 0x08,
+    AliasCodeData = 0x09,
+    Ipc = 0x0A,
+    Stack = 0x0B,
+    ThreadLocal = 0x0C,
+    Transfered = 0x0D,
+    SharedTransfered = 0x0E,
+    SharedCode = 0x0F,
+    Inaccessible = 0x10,
+    NonSecureIpc = 0x11,
+    NonDeviceIpc = 0x12,
+    Kernel = 0x13,
+    GeneratedCode = 0x14,
+    CodeOut = 0x15,
+};
+DECLARE_ENUM_FLAG_OPERATORS(MemoryState);
+
+enum class MemoryAttribute : u32 {
+    Locked = (1 << 0),
+    IpcLocked = (1 << 1),
+    DeviceShared = (1 << 2),
+    Uncached = (1 << 3),
+};
+DECLARE_ENUM_FLAG_OPERATORS(MemoryAttribute);
+
+enum class MemoryPermission : u32 {
+    None = (0 << 0),
+    Read = (1 << 0),
+    Write = (1 << 1),
+    Execute = (1 << 2),
+    ReadWrite = Read | Write,
+    ReadExecute = Read | Execute,
+    DontCare = (1 << 28),
+};
+DECLARE_ENUM_FLAG_OPERATORS(MemoryPermission);
+
+struct MemoryInfo {
+    u64 addr{};
+    u64 size{};
+    MemoryState state{};
+    MemoryAttribute attr{};
+    MemoryPermission perm{};
+    u32 ipc_refcount{};
+    u32 device_refcount{};
+    u32 padding{};
+};
+
+} // namespace Kernel::Svc

src/core/hle/kernel/svc_wrap.h

@@ -0,0 +1,552 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+#include "core/arm/arm_interface.h"
+#include "core/core.h"
+#include "core/hle/result.h"
+
+namespace Kernel {
+
+static inline u64 Param(const Core::System& system, int n) {
+    return system.CurrentArmInterface().GetReg(n);
+}
+
+static inline u32 Param32(const Core::System& system, int n) {
+    return static_cast<u32>(system.CurrentArmInterface().GetReg(n));
+}
+
+/**
+ * HLE a function return from the current ARM userland process
+ * @param system System context
+ * @param result Result to return
+ */
+static inline void FuncReturn(Core::System& system, u64 result) {
+    system.CurrentArmInterface().SetReg(0, result);
+}
+
+static inline void FuncReturn32(Core::System& system, u32 result) {
+    system.CurrentArmInterface().SetReg(0, (u64)result);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Function wrappers that return type ResultCode
+
+template <ResultCode func(Core::System&, u64)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, Param(system, 0)).raw);
+}
+
+template <ResultCode func(Core::System&, u64, u64)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, Param(system, 0), Param(system, 1)).raw);
+}
+
+template <ResultCode func(Core::System&, u32)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0))).raw);
+}
+
+template <ResultCode func(Core::System&, u32, u32)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(
+        system,
+        func(system, static_cast<u32>(Param(system, 0)), static_cast<u32>(Param(system, 1))).raw);
+}
+
+template <ResultCode func(Core::System&, u32, u64, u64, u64)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1),
+                            Param(system, 2), Param(system, 3))
+                           .raw);
+}
+
+template <ResultCode func(Core::System&, u32*)>
+void SvcWrap64(Core::System& system) {
+    u32 param = 0;
+    const u32 retval = func(system, &param).raw;
+    system.CurrentArmInterface().SetReg(1, param);
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, u32*, u32)>
+void SvcWrap64(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, &param_1, static_cast<u32>(Param(system, 1))).raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, u32*, u32*)>
+void SvcWrap64(Core::System& system) {
+    u32 param_1 = 0;
+    u32 param_2 = 0;
+    const u32 retval = func(system, &param_1, &param_2).raw;
+
+    auto& arm_interface = system.CurrentArmInterface();
+    arm_interface.SetReg(1, param_1);
+    arm_interface.SetReg(2, param_2);
+
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, u32*, u64)>
+void SvcWrap64(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, &param_1, Param(system, 1)).raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, u32*, u64, u32)>
+void SvcWrap64(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval =
+        func(system, &param_1, Param(system, 1), static_cast<u32>(Param(system, 2))).raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, u64*, u32)>
+void SvcWrap64(Core::System& system) {
+    u64 param_1 = 0;
+    const u32 retval = func(system, &param_1, static_cast<u32>(Param(system, 1))).raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, u64, u32)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, Param(system, 0), static_cast<u32>(Param(system, 1))).raw);
+}
+
+template <ResultCode func(Core::System&, u64*, u64)>
+void SvcWrap64(Core::System& system) {
+    u64 param_1 = 0;
+    const u32 retval = func(system, &param_1, Param(system, 1)).raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, u64*, u32, u32)>
+void SvcWrap64(Core::System& system) {
+    u64 param_1 = 0;
+    const u32 retval = func(system, &param_1, static_cast<u32>(Param(system, 1)),
+                            static_cast<u32>(Param(system, 2)))
+                           .raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, u32, u64)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1)).raw);
+}
+
+template <ResultCode func(Core::System&, u32, u32, u64)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)),
+                            static_cast<u32>(Param(system, 1)), Param(system, 2))
+                           .raw);
+}
+
+template <ResultCode func(Core::System&, u32, u32*, u64*)>
+void SvcWrap64(Core::System& system) {
+    u32 param_1 = 0;
+    u64 param_2 = 0;
+    const ResultCode retval = func(system, static_cast<u32>(Param(system, 2)), &param_1, &param_2);
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    system.CurrentArmInterface().SetReg(2, param_2);
+    FuncReturn(system, retval.raw);
+}
+
+template <ResultCode func(Core::System&, u64, u64, u32, u32)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, Param(system, 0), Param(system, 1),
+                            static_cast<u32>(Param(system, 2)), static_cast<u32>(Param(system, 3)))
+                           .raw);
+}
+
+template <ResultCode func(Core::System&, u64, u64, u32, u64)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, Param(system, 0), Param(system, 1),
+                            static_cast<u32>(Param(system, 2)), Param(system, 3))
+                           .raw);
+}
+
+template <ResultCode func(Core::System&, u32, u64, u32)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1),
+                            static_cast<u32>(Param(system, 2)))
+                           .raw);
+}
+
+template <ResultCode func(Core::System&, u64, u64, u64)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, Param(system, 0), Param(system, 1), Param(system, 2)).raw);
+}
+
+template <ResultCode func(Core::System&, u64, u64, u32)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(
+        system,
+        func(system, Param(system, 0), Param(system, 1), static_cast<u32>(Param(system, 2))).raw);
+}
+
+template <ResultCode func(Core::System&, u32, u64, u64, u32)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1),
+                            Param(system, 2), static_cast<u32>(Param(system, 3)))
+                           .raw);
+}
+
+template <ResultCode func(Core::System&, u32, u64, u64)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(
+        system,
+        func(system, static_cast<u32>(Param(system, 0)), Param(system, 1), Param(system, 2)).raw);
+}
+
+template <ResultCode func(Core::System&, u32*, u64, u64, s64)>
+void SvcWrap64(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, &param_1, Param(system, 1), static_cast<u32>(Param(system, 2)),
+                            static_cast<s64>(Param(system, 3)))
+                           .raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, u64, u64, u32, s64)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, Param(system, 0), Param(system, 1),
+                            static_cast<u32>(Param(system, 2)), static_cast<s64>(Param(system, 3)))
+                           .raw);
+}
+
+template <ResultCode func(Core::System&, u64*, u64, u64, u64)>
+void SvcWrap64(Core::System& system) {
+    u64 param_1 = 0;
+    const u32 retval =
+        func(system, &param_1, Param(system, 1), Param(system, 2), Param(system, 3)).raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, u32*, u64, u64, u64, u32, s32)>
+void SvcWrap64(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, &param_1, Param(system, 1), Param(system, 2), Param(system, 3),
+                            static_cast<u32>(Param(system, 4)), static_cast<s32>(Param(system, 5)))
+                           .raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, u32*, u64, u64, u32)>
+void SvcWrap64(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, &param_1, Param(system, 1), Param(system, 2),
+                            static_cast<u32>(Param(system, 3)))
+                           .raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, Handle*, u64, u32, u32)>
+void SvcWrap64(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, &param_1, Param(system, 1), static_cast<u32>(Param(system, 2)),
+                            static_cast<u32>(Param(system, 3)))
+                           .raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+template <ResultCode func(Core::System&, u64, u32, s32, s64)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, Param(system, 0), static_cast<u32>(Param(system, 1)),
+                            static_cast<s32>(Param(system, 2)), static_cast<s64>(Param(system, 3)))
+                           .raw);
+}
+
+template <ResultCode func(Core::System&, u64, u32, s32, s32)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system, Param(system, 0), static_cast<u32>(Param(system, 1)),
+                            static_cast<s32>(Param(system, 2)), static_cast<s32>(Param(system, 3)))
+                           .raw);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Function wrappers that return type u32
+
+template <u32 func(Core::System&)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system));
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Function wrappers that return type u64
+
+template <u64 func(Core::System&)>
+void SvcWrap64(Core::System& system) {
+    FuncReturn(system, func(system));
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+/// Function wrappers that return type void
+
+template <void func(Core::System&)>
+void SvcWrap64(Core::System& system) {
+    func(system);
+}
+
+template <void func(Core::System&, u32)>
+void SvcWrap64(Core::System& system) {
+    func(system, static_cast<u32>(Param(system, 0)));
+}
+
+template <void func(Core::System&, u32, u64, u64, u64)>
+void SvcWrap64(Core::System& system) {
+    func(system, static_cast<u32>(Param(system, 0)), Param(system, 1), Param(system, 2),
+         Param(system, 3));
+}
+
+template <void func(Core::System&, s64)>
+void SvcWrap64(Core::System& system) {
+    func(system, static_cast<s64>(Param(system, 0)));
+}
+
+template <void func(Core::System&, u64, s32)>
+void SvcWrap64(Core::System& system) {
+    func(system, Param(system, 0), static_cast<s32>(Param(system, 1)));
+}
+
+template <void func(Core::System&, u64, u64)>
+void SvcWrap64(Core::System& system) {
+    func(system, Param(system, 0), Param(system, 1));
+}
+
+template <void func(Core::System&, u64, u64, u64)>
+void SvcWrap64(Core::System& system) {
+    func(system, Param(system, 0), Param(system, 1), Param(system, 2));
+}
+
+template <void func(Core::System&, u32, u64, u64)>
+void SvcWrap64(Core::System& system) {
+    func(system, static_cast<u32>(Param(system, 0)), Param(system, 1), Param(system, 2));
+}
+
+// Used by QueryMemory32, ArbitrateLock32
+template <ResultCode func(Core::System&, u32, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    FuncReturn32(system,
+                 func(system, Param32(system, 0), Param32(system, 1), Param32(system, 2)).raw);
+}
+
+// Used by Break32
+template <void func(Core::System&, u32, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    func(system, Param32(system, 0), Param32(system, 1), Param32(system, 2));
+}
+
+// Used by ExitProcess32, ExitThread32
+template <void func(Core::System&)>
+void SvcWrap32(Core::System& system) {
+    func(system);
+}
+
+// Used by GetCurrentProcessorNumber32
+template <u32 func(Core::System&)>
+void SvcWrap32(Core::System& system) {
+    FuncReturn32(system, func(system));
+}
+
+// Used by SleepThread32
+template <void func(Core::System&, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    func(system, Param32(system, 0), Param32(system, 1));
+}
+
+// Used by CreateThread32
+template <ResultCode func(Core::System&, Handle*, u32, u32, u32, u32, s32)>
+void SvcWrap32(Core::System& system) {
+    Handle param_1 = 0;
+
+    const u32 retval = func(system, &param_1, Param32(system, 0), Param32(system, 1),
+                            Param32(system, 2), Param32(system, 3), Param32(system, 4))
+                           .raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+// Used by GetInfo32
+template <ResultCode func(Core::System&, u32*, u32*, u32, u32, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    u32 param_2 = 0;
+
+    const u32 retval = func(system, &param_1, &param_2, Param32(system, 0), Param32(system, 1),
+                            Param32(system, 2), Param32(system, 3))
+                           .raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    system.CurrentArmInterface().SetReg(2, param_2);
+    FuncReturn(system, retval);
+}
+
+// Used by GetThreadPriority32, ConnectToNamedPort32
+template <ResultCode func(Core::System&, u32*, u32)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, &param_1, Param32(system, 1)).raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+// Used by GetThreadId32
+template <ResultCode func(Core::System&, u32*, u32*, u32)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    u32 param_2 = 0;
+
+    const u32 retval = func(system, &param_1, &param_2, Param32(system, 1)).raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    system.CurrentArmInterface().SetReg(2, param_2);
+    FuncReturn(system, retval);
+}
+
+// Used by GetSystemTick32
+template <void func(Core::System&, u32*, u32*)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    u32 param_2 = 0;
+
+    func(system, &param_1, &param_2);
+    system.CurrentArmInterface().SetReg(0, param_1);
+    system.CurrentArmInterface().SetReg(1, param_2);
+}
+
+// Used by CreateEvent32
+template <ResultCode func(Core::System&, Handle*, Handle*)>
+void SvcWrap32(Core::System& system) {
+    Handle param_1 = 0;
+    Handle param_2 = 0;
+
+    const u32 retval = func(system, &param_1, &param_2).raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    system.CurrentArmInterface().SetReg(2, param_2);
+    FuncReturn(system, retval);
+}
+
+// Used by GetThreadId32
+template <ResultCode func(Core::System&, Handle, u32*, u32*, u32*)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    u32 param_2 = 0;
+    u32 param_3 = 0;
+
+    const u32 retval = func(system, Param32(system, 2), &param_1, &param_2, &param_3).raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    system.CurrentArmInterface().SetReg(2, param_2);
+    system.CurrentArmInterface().SetReg(3, param_3);
+    FuncReturn(system, retval);
+}
+
+// Used by SignalProcessWideKey32
+template <void func(Core::System&, u32, s32)>
+void SvcWrap32(Core::System& system) {
+    func(system, static_cast<u32>(Param(system, 0)), static_cast<s32>(Param(system, 1)));
+}
+
+// Used by SetThreadPriority32
+template <ResultCode func(Core::System&, Handle, u32)>
+void SvcWrap32(Core::System& system) {
+    const u32 retval =
+        func(system, static_cast<Handle>(Param(system, 0)), static_cast<u32>(Param(system, 1))).raw;
+    FuncReturn(system, retval);
+}
+
+// Used by SetThreadCoreMask32
+template <ResultCode func(Core::System&, Handle, u32, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    const u32 retval =
+        func(system, static_cast<Handle>(Param(system, 0)), static_cast<u32>(Param(system, 1)),
+             static_cast<u32>(Param(system, 2)), static_cast<u32>(Param(system, 3)))
+            .raw;
+    FuncReturn(system, retval);
+}
+
+// Used by WaitProcessWideKeyAtomic32
+template <ResultCode func(Core::System&, u32, u32, Handle, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    const u32 retval =
+        func(system, static_cast<u32>(Param(system, 0)), static_cast<u32>(Param(system, 1)),
+             static_cast<Handle>(Param(system, 2)), static_cast<u32>(Param(system, 3)),
+             static_cast<u32>(Param(system, 4)))
+            .raw;
+    FuncReturn(system, retval);
+}
+
+// Used by WaitForAddress32
+template <ResultCode func(Core::System&, u32, u32, s32, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    const u32 retval = func(system, static_cast<u32>(Param(system, 0)),
+                            static_cast<u32>(Param(system, 1)), static_cast<s32>(Param(system, 2)),
+                            static_cast<u32>(Param(system, 3)), static_cast<u32>(Param(system, 4)))
+                           .raw;
+    FuncReturn(system, retval);
+}
+
+// Used by SignalToAddress32
+template <ResultCode func(Core::System&, u32, u32, s32, s32)>
+void SvcWrap32(Core::System& system) {
+    const u32 retval =
+        func(system, static_cast<u32>(Param(system, 0)), static_cast<u32>(Param(system, 1)),
+             static_cast<s32>(Param(system, 2)), static_cast<s32>(Param(system, 3)))
+            .raw;
+    FuncReturn(system, retval);
+}
+
+// Used by SendSyncRequest32, ArbitrateUnlock32
+template <ResultCode func(Core::System&, u32)>
+void SvcWrap32(Core::System& system) {
+    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0))).raw);
+}
+
+// Used by CreateTransferMemory32
+template <ResultCode func(Core::System&, Handle*, u32, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    Handle handle = 0;
+    const u32 retval =
+        func(system, &handle, Param32(system, 1), Param32(system, 2), Param32(system, 3)).raw;
+    system.CurrentArmInterface().SetReg(1, handle);
+    FuncReturn(system, retval);
+}
+
+// Used by WaitSynchronization32
+template <ResultCode func(Core::System&, u32, u32, s32, u32, Handle*)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, Param32(system, 0), Param32(system, 1), Param32(system, 2),
+                            Param32(system, 3), &param_1)
+                           .raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+} // namespace Kernel

src/core/hle/kernel/synchronization.cpp

@@ -0,0 +1,116 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/core.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/synchronization.h"
+#include "core/hle/kernel/synchronization_object.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/time_manager.h"
+
+namespace Kernel {
+
+Synchronization::Synchronization(Core::System& system) : system{system} {}
+
+void Synchronization::SignalObject(SynchronizationObject& obj) const {
+    auto& kernel = system.Kernel();
+    KScopedSchedulerLock lock(kernel);
+    if (obj.IsSignaled()) {
+        for (auto thread : obj.GetWaitingThreads()) {
+            if (thread->GetSchedulingStatus() == ThreadSchedStatus::Paused) {
+                if (thread->GetStatus() != ThreadStatus::WaitHLEEvent) {
+                    ASSERT(thread->GetStatus() == ThreadStatus::WaitSynch);
+                    ASSERT(thread->IsWaitingSync());
+                }
+                thread->SetSynchronizationResults(&obj, RESULT_SUCCESS);
+                thread->ResumeFromWait();
+            }
+        }
+        obj.ClearWaitingThreads();
+    }
+}
+
+std::pair<ResultCode, Handle> Synchronization::WaitFor(
+    std::vector<std::shared_ptr<SynchronizationObject>>& sync_objects, s64 nano_seconds) {
+    auto& kernel = system.Kernel();
+    auto* const thread = kernel.CurrentScheduler()->GetCurrentThread();
+    Handle event_handle = InvalidHandle;
+    {
+        KScopedSchedulerLockAndSleep lock(kernel, event_handle, thread, nano_seconds);
+        const auto itr =
+            std::find_if(sync_objects.begin(), sync_objects.end(),
+                         [thread](const std::shared_ptr<SynchronizationObject>& object) {
+                             return object->IsSignaled();
+                         });
+
+        if (itr != sync_objects.end()) {
+            // We found a ready object, acquire it and set the result value
+            SynchronizationObject* object = itr->get();
+            object->Acquire(thread);
+            const u32 index = static_cast<s32>(std::distance(sync_objects.begin(), itr));
+            lock.CancelSleep();
+            return {RESULT_SUCCESS, index};
+        }
+
+        if (nano_seconds == 0) {
+            lock.CancelSleep();
+            return {RESULT_TIMEOUT, InvalidHandle};
+        }
+
+        if (thread->IsPendingTermination()) {
+            lock.CancelSleep();
+            return {ERR_THREAD_TERMINATING, InvalidHandle};
+        }
+
+        if (thread->IsSyncCancelled()) {
+            thread->SetSyncCancelled(false);
+            lock.CancelSleep();
+            return {ERR_SYNCHRONIZATION_CANCELED, InvalidHandle};
+        }
+
+        for (auto& object : sync_objects) {
+            object->AddWaitingThread(SharedFrom(thread));
+        }
+
+        thread->SetSynchronizationObjects(&sync_objects);
+        thread->SetSynchronizationResults(nullptr, RESULT_TIMEOUT);
+        thread->SetStatus(ThreadStatus::WaitSynch);
+        thread->SetWaitingSync(true);
+    }
+    thread->SetWaitingSync(false);
+
+    if (event_handle != InvalidHandle) {
+        auto& time_manager = kernel.TimeManager();
+        time_manager.UnscheduleTimeEvent(event_handle);
+    }
+
+    {
+        KScopedSchedulerLock lock(kernel);
+        ResultCode signaling_result = thread->GetSignalingResult();
+        SynchronizationObject* signaling_object = thread->GetSignalingObject();
+        thread->SetSynchronizationObjects(nullptr);
+        auto shared_thread = SharedFrom(thread);
+        for (auto& obj : sync_objects) {
+            obj->RemoveWaitingThread(shared_thread);
+        }
+        if (signaling_object != nullptr) {
+            const auto itr = std::find_if(
+                sync_objects.begin(), sync_objects.end(),
+                [signaling_object](const std::shared_ptr<SynchronizationObject>& object) {
+                    return object.get() == signaling_object;
+                });
+            ASSERT(itr != sync_objects.end());
+            signaling_object->Acquire(thread);
+            const u32 index = static_cast<s32>(std::distance(sync_objects.begin(), itr));
+            return {signaling_result, index};
+        }
+        return {signaling_result, -1};
+    }
+}
+
+} // namespace Kernel

src/core/hle/kernel/synchronization.h

@@ -0,0 +1,44 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <utility>
+#include <vector>
+
+#include "core/hle/kernel/object.h"
+#include "core/hle/result.h"
+
+namespace Core {
+class System;
+} // namespace Core
+
+namespace Kernel {
+
+class SynchronizationObject;
+
+/**
+ * The 'Synchronization' class is an interface for handling synchronization methods
+ * used by Synchronization objects and synchronization SVCs. This centralizes processing of
+ * such
+ */
+class Synchronization {
+public:
+    explicit Synchronization(Core::System& system);
+
+    /// Signals a synchronization object, waking up all its waiting threads
+    void SignalObject(SynchronizationObject& obj) const;
+
+    /// Tries to see if waiting for any of the sync_objects is necessary, if not
+    /// it returns Success and the handle index of the signaled sync object. In
+    /// case not, the current thread will be locked and wait for nano_seconds or
+    /// for a synchronization object to signal.
+    std::pair<ResultCode, Handle> WaitFor(
+        std::vector<std::shared_ptr<SynchronizationObject>>& sync_objects, s64 nano_seconds);
+
+private:
+    Core::System& system;
+};
+} // namespace Kernel

src/core/hle/kernel/synchronization_object.cpp

@@ -0,0 +1,49 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "common/logging/log.h"
+#include "core/core.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/kernel/synchronization.h"
+#include "core/hle/kernel/synchronization_object.h"
+#include "core/hle/kernel/thread.h"
+
+namespace Kernel {
+
+SynchronizationObject::SynchronizationObject(KernelCore& kernel) : Object{kernel} {}
+SynchronizationObject::~SynchronizationObject() = default;
+
+void SynchronizationObject::Signal() {
+    kernel.Synchronization().SignalObject(*this);
+}
+
+void SynchronizationObject::AddWaitingThread(std::shared_ptr<Thread> thread) {
+    auto itr = std::find(waiting_threads.begin(), waiting_threads.end(), thread);
+    if (itr == waiting_threads.end())
+        waiting_threads.push_back(std::move(thread));
+}
+
+void SynchronizationObject::RemoveWaitingThread(std::shared_ptr<Thread> thread) {
+    auto itr = std::find(waiting_threads.begin(), waiting_threads.end(), thread);
+    // If a thread passed multiple handles to the same object,
+    // the kernel might attempt to remove the thread from the object's
+    // waiting threads list multiple times.
+    if (itr != waiting_threads.end())
+        waiting_threads.erase(itr);
+}
+
+void SynchronizationObject::ClearWaitingThreads() {
+    waiting_threads.clear();
+}
+
+const std::vector<std::shared_ptr<Thread>>& SynchronizationObject::GetWaitingThreads() const {
+    return waiting_threads;
+}
+
+} // namespace Kernel

src/core/hle/kernel/synchronization_object.h

@@ -0,0 +1,77 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <atomic>
+#include <memory>
+#include <vector>
+
+#include "core/hle/kernel/object.h"
+
+namespace Kernel {
+
+class KernelCore;
+class Synchronization;
+class Thread;
+
+/// Class that represents a Kernel object that a thread can be waiting on
+class SynchronizationObject : public Object {
+public:
+    explicit SynchronizationObject(KernelCore& kernel);
+    ~SynchronizationObject() override;
+
+    /**
+     * Check if the specified thread should wait until the object is available
+     * @param thread The thread about which we're deciding.
+     * @return True if the current thread should wait due to this object being unavailable
+     */
+    virtual bool ShouldWait(const Thread* thread) const = 0;
+
+    /// Acquire/lock the object for the specified thread if it is available
+    virtual void Acquire(Thread* thread) = 0;
+
+    /// Signal this object
+    virtual void Signal();
+
+    virtual bool IsSignaled() const {
+        return is_signaled;
+    }
+
+    /**
+     * Add a thread to wait on this object
+     * @param thread Pointer to thread to add
+     */
+    void AddWaitingThread(std::shared_ptr<Thread> thread);
+
+    /**
+     * Removes a thread from waiting on this object (e.g. if it was resumed already)
+     * @param thread Pointer to thread to remove
+     */
+    void RemoveWaitingThread(std::shared_ptr<Thread> thread);
+
+    /// Get a const reference to the waiting threads list for debug use
+    const std::vector<std::shared_ptr<Thread>>& GetWaitingThreads() const;
+
+    void ClearWaitingThreads();
+
+protected:
+    std::atomic_bool is_signaled{}; // Tells if this sync object is signaled
+
+private:
+    /// Threads waiting for this object to become available
+    std::vector<std::shared_ptr<Thread>> waiting_threads;
+};
+
+// Specialization of DynamicObjectCast for SynchronizationObjects
+template <>
+inline std::shared_ptr<SynchronizationObject> DynamicObjectCast<SynchronizationObject>(
+    std::shared_ptr<Object> object) {
+    if (object != nullptr && object->IsWaitable()) {
+        return std::static_pointer_cast<SynchronizationObject>(object);
+    }
+    return nullptr;
+}
+
+} // namespace Kernel

src/core/hle/kernel/thread.cpp

@@ -0,0 +1,478 @@
+// Copyright 2014 Citra Emulator Project / PPSSPP Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cinttypes>
+#include <optional>
+#include <vector>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "common/fiber.h"
+#include "common/logging/log.h"
+#include "common/thread_queue_list.h"
+#include "core/core.h"
+#include "core/cpu_manager.h"
+#include "core/hardware_properties.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/time_manager.h"
+#include "core/hle/result.h"
+#include "core/memory.h"
+
+#ifdef ARCHITECTURE_x86_64
+#include "core/arm/dynarmic/arm_dynarmic_32.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
+#endif
+
+namespace Kernel {
+
+bool Thread::ShouldWait(const Thread* thread) const {
+    return status != ThreadStatus::Dead;
+}
+
+bool Thread::IsSignaled() const {
+    return status == ThreadStatus::Dead;
+}
+
+void Thread::Acquire(Thread* thread) {
+    ASSERT_MSG(!ShouldWait(thread), "object unavailable!");
+}
+
+Thread::Thread(KernelCore& kernel) : SynchronizationObject{kernel} {}
+Thread::~Thread() = default;
+
+void Thread::Stop() {
+    {
+        KScopedSchedulerLock lock(kernel);
+        SetStatus(ThreadStatus::Dead);
+        Signal();
+        kernel.GlobalHandleTable().Close(global_handle);
+
+        if (owner_process) {
+            owner_process->UnregisterThread(this);
+
+            // Mark the TLS slot in the thread's page as free.
+            owner_process->FreeTLSRegion(tls_address);
+        }
+        has_exited = true;
+    }
+    global_handle = 0;
+}
+
+void Thread::ResumeFromWait() {
+    KScopedSchedulerLock lock(kernel);
+    switch (status) {
+    case ThreadStatus::Paused:
+    case ThreadStatus::WaitSynch:
+    case ThreadStatus::WaitHLEEvent:
+    case ThreadStatus::WaitSleep:
+    case ThreadStatus::WaitIPC:
+    case ThreadStatus::WaitMutex:
+    case ThreadStatus::WaitCondVar:
+    case ThreadStatus::WaitArb:
+    case ThreadStatus::Dormant:
+        break;
+
+    case ThreadStatus::Ready:
+        // The thread's wakeup callback must have already been cleared when the thread was first
+        // awoken.
+        ASSERT(hle_callback == nullptr);
+        // If the thread is waiting on multiple wait objects, it might be awoken more than once
+        // before actually resuming. We can ignore subsequent wakeups if the thread status has
+        // already been set to ThreadStatus::Ready.
+        return;
+    case ThreadStatus::Dead:
+        // This should never happen, as threads must complete before being stopped.
+        DEBUG_ASSERT_MSG(false, "Thread with object id {} cannot be resumed because it's DEAD.",
+                         GetObjectId());
+        return;
+    }
+
+    SetStatus(ThreadStatus::Ready);
+}
+
+void Thread::OnWakeUp() {
+    KScopedSchedulerLock lock(kernel);
+    SetStatus(ThreadStatus::Ready);
+}
+
+ResultCode Thread::Start() {
+    KScopedSchedulerLock lock(kernel);
+    SetStatus(ThreadStatus::Ready);
+    return RESULT_SUCCESS;
+}
+
+void Thread::CancelWait() {
+    KScopedSchedulerLock lock(kernel);
+    if (GetSchedulingStatus() != ThreadSchedStatus::Paused || !is_waiting_on_sync) {
+        is_sync_cancelled = true;
+        return;
+    }
+    // TODO(Blinkhawk): Implement cancel of server session
+    is_sync_cancelled = false;
+    SetSynchronizationResults(nullptr, ERR_SYNCHRONIZATION_CANCELED);
+    SetStatus(ThreadStatus::Ready);
+}
+
+static void ResetThreadContext32(Core::ARM_Interface::ThreadContext32& context, u32 stack_top,
+                                 u32 entry_point, u32 arg) {
+    context = {};
+    context.cpu_registers[0] = arg;
+    context.cpu_registers[15] = entry_point;
+    context.cpu_registers[13] = stack_top;
+}
+
+static void ResetThreadContext64(Core::ARM_Interface::ThreadContext64& context, VAddr stack_top,
+                                 VAddr entry_point, u64 arg) {
+    context = {};
+    context.cpu_registers[0] = arg;
+    context.pc = entry_point;
+    context.sp = stack_top;
+    // TODO(merry): Perform a hardware test to determine the below value.
+    context.fpcr = 0;
+}
+
+std::shared_ptr<Common::Fiber>& Thread::GetHostContext() {
+    return host_context;
+}
+
+ResultVal<std::shared_ptr<Thread>> Thread::Create(Core::System& system, ThreadType type_flags,
+                                                  std::string name, VAddr entry_point, u32 priority,
+                                                  u64 arg, s32 processor_id, VAddr stack_top,
+                                                  Process* owner_process) {
+    std::function<void(void*)> init_func = Core::CpuManager::GetGuestThreadStartFunc();
+    void* init_func_parameter = system.GetCpuManager().GetStartFuncParamater();
+    return Create(system, type_flags, name, entry_point, priority, arg, processor_id, stack_top,
+                  owner_process, std::move(init_func), init_func_parameter);
+}
+
+ResultVal<std::shared_ptr<Thread>> Thread::Create(Core::System& system, ThreadType type_flags,
+                                                  std::string name, VAddr entry_point, u32 priority,
+                                                  u64 arg, s32 processor_id, VAddr stack_top,
+                                                  Process* owner_process,
+                                                  std::function<void(void*)>&& thread_start_func,
+                                                  void* thread_start_parameter) {
+    auto& kernel = system.Kernel();
+    // Check if priority is in ranged. Lowest priority -> highest priority id.
+    if (priority > THREADPRIO_LOWEST && ((type_flags & THREADTYPE_IDLE) == 0)) {
+        LOG_ERROR(Kernel_SVC, "Invalid thread priority: {}", priority);
+        return ERR_INVALID_THREAD_PRIORITY;
+    }
+
+    if (processor_id > THREADPROCESSORID_MAX) {
+        LOG_ERROR(Kernel_SVC, "Invalid processor id: {}", processor_id);
+        return ERR_INVALID_PROCESSOR_ID;
+    }
+
+    if (owner_process) {
+        if (!system.Memory().IsValidVirtualAddress(*owner_process, entry_point)) {
+            LOG_ERROR(Kernel_SVC, "(name={}): invalid entry {:016X}", name, entry_point);
+            // TODO (bunnei): Find the correct error code to use here
+            return RESULT_UNKNOWN;
+        }
+    }
+
+    std::shared_ptr<Thread> thread = std::make_shared<Thread>(kernel);
+
+    thread->thread_id = kernel.CreateNewThreadID();
+    thread->status = ThreadStatus::Dormant;
+    thread->entry_point = entry_point;
+    thread->stack_top = stack_top;
+    thread->disable_count = 1;
+    thread->tpidr_el0 = 0;
+    thread->nominal_priority = thread->current_priority = priority;
+    thread->schedule_count = -1;
+    thread->last_scheduled_tick = 0;
+    thread->processor_id = processor_id;
+    thread->ideal_core = processor_id;
+    thread->affinity_mask.SetAffinity(processor_id, true);
+    thread->wait_objects = nullptr;
+    thread->mutex_wait_address = 0;
+    thread->condvar_wait_address = 0;
+    thread->wait_handle = 0;
+    thread->name = std::move(name);
+    thread->global_handle = kernel.GlobalHandleTable().Create(thread).Unwrap();
+    thread->owner_process = owner_process;
+    thread->type = type_flags;
+    if ((type_flags & THREADTYPE_IDLE) == 0) {
+        auto& scheduler = kernel.GlobalSchedulerContext();
+        scheduler.AddThread(thread);
+    }
+    if (owner_process) {
+        thread->tls_address = thread->owner_process->CreateTLSRegion();
+        thread->owner_process->RegisterThread(thread.get());
+    } else {
+        thread->tls_address = 0;
+    }
+
+    // TODO(peachum): move to ScheduleThread() when scheduler is added so selected core is used
+    // to initialize the context
+    if ((type_flags & THREADTYPE_HLE) == 0) {
+        ResetThreadContext32(thread->context_32, static_cast<u32>(stack_top),
+                             static_cast<u32>(entry_point), static_cast<u32>(arg));
+        ResetThreadContext64(thread->context_64, stack_top, entry_point, arg);
+    }
+    thread->host_context =
+        std::make_shared<Common::Fiber>(std::move(thread_start_func), thread_start_parameter);
+
+    return MakeResult<std::shared_ptr<Thread>>(std::move(thread));
+}
+
+void Thread::SetPriority(u32 priority) {
+    KScopedSchedulerLock lock(kernel);
+    ASSERT_MSG(priority <= THREADPRIO_LOWEST && priority >= THREADPRIO_HIGHEST,
+               "Invalid priority value.");
+    nominal_priority = priority;
+    UpdatePriority();
+}
+
+void Thread::SetSynchronizationResults(SynchronizationObject* object, ResultCode result) {
+    signaling_object = object;
+    signaling_result = result;
+}
+
+s32 Thread::GetSynchronizationObjectIndex(std::shared_ptr<SynchronizationObject> object) const {
+    ASSERT_MSG(!wait_objects->empty(), "Thread is not waiting for anything");
+    const auto match = std::find(wait_objects->rbegin(), wait_objects->rend(), object);
+    return static_cast<s32>(std::distance(match, wait_objects->rend()) - 1);
+}
+
+VAddr Thread::GetCommandBufferAddress() const {
+    // Offset from the start of TLS at which the IPC command buffer begins.
+    constexpr u64 command_header_offset = 0x80;
+    return GetTLSAddress() + command_header_offset;
+}
+
+void Thread::SetStatus(ThreadStatus new_status) {
+    if (new_status == status) {
+        return;
+    }
+
+    switch (new_status) {
+    case ThreadStatus::Ready:
+        SetSchedulingStatus(ThreadSchedStatus::Runnable);
+        break;
+    case ThreadStatus::Dormant:
+        SetSchedulingStatus(ThreadSchedStatus::None);
+        break;
+    case ThreadStatus::Dead:
+        SetSchedulingStatus(ThreadSchedStatus::Exited);
+        break;
+    default:
+        SetSchedulingStatus(ThreadSchedStatus::Paused);
+        break;
+    }
+
+    status = new_status;
+}
+
+void Thread::AddMutexWaiter(std::shared_ptr<Thread> thread) {
+    if (thread->lock_owner.get() == this) {
+        // If the thread is already waiting for this thread to release the mutex, ensure that the
+        // waiters list is consistent and return without doing anything.
+        const auto iter = std::find(wait_mutex_threads.begin(), wait_mutex_threads.end(), thread);
+        ASSERT(iter != wait_mutex_threads.end());
+        return;
+    }
+
+    // A thread can't wait on two different mutexes at the same time.
+    ASSERT(thread->lock_owner == nullptr);
+
+    // Ensure that the thread is not already in the list of mutex waiters
+    const auto iter = std::find(wait_mutex_threads.begin(), wait_mutex_threads.end(), thread);
+    ASSERT(iter == wait_mutex_threads.end());
+
+    // Keep the list in an ordered fashion
+    const auto insertion_point = std::find_if(
+        wait_mutex_threads.begin(), wait_mutex_threads.end(),
+        [&thread](const auto& entry) { return entry->GetPriority() > thread->GetPriority(); });
+    wait_mutex_threads.insert(insertion_point, thread);
+    thread->lock_owner = SharedFrom(this);
+
+    UpdatePriority();
+}
+
+void Thread::RemoveMutexWaiter(std::shared_ptr<Thread> thread) {
+    ASSERT(thread->lock_owner.get() == this);
+
+    // Ensure that the thread is in the list of mutex waiters
+    const auto iter = std::find(wait_mutex_threads.begin(), wait_mutex_threads.end(), thread);
+    ASSERT(iter != wait_mutex_threads.end());
+
+    wait_mutex_threads.erase(iter);
+
+    thread->lock_owner = nullptr;
+    UpdatePriority();
+}
+
+void Thread::UpdatePriority() {
+    // If any of the threads waiting on the mutex have a higher priority
+    // (taking into account priority inheritance), then this thread inherits
+    // that thread's priority.
+    u32 new_priority = nominal_priority;
+    if (!wait_mutex_threads.empty()) {
+        if (wait_mutex_threads.front()->current_priority < new_priority) {
+            new_priority = wait_mutex_threads.front()->current_priority;
+        }
+    }
+
+    if (new_priority == current_priority) {
+        return;
+    }
+
+    if (GetStatus() == ThreadStatus::WaitCondVar) {
+        owner_process->RemoveConditionVariableThread(SharedFrom(this));
+    }
+
+    SetCurrentPriority(new_priority);
+
+    if (GetStatus() == ThreadStatus::WaitCondVar) {
+        owner_process->InsertConditionVariableThread(SharedFrom(this));
+    }
+
+    if (!lock_owner) {
+        return;
+    }
+
+    // Ensure that the thread is within the correct location in the waiting list.
+    auto old_owner = lock_owner;
+    lock_owner->RemoveMutexWaiter(SharedFrom(this));
+    old_owner->AddMutexWaiter(SharedFrom(this));
+
+    // Recursively update the priority of the thread that depends on the priority of this one.
+    lock_owner->UpdatePriority();
+}
+
+bool Thread::AllSynchronizationObjectsReady() const {
+    return std::none_of(wait_objects->begin(), wait_objects->end(),
+                        [this](const std::shared_ptr<SynchronizationObject>& object) {
+                            return object->ShouldWait(this);
+                        });
+}
+
+bool Thread::InvokeHLECallback(std::shared_ptr<Thread> thread) {
+    ASSERT(hle_callback);
+    return hle_callback(std::move(thread));
+}
+
+ResultCode Thread::SetActivity(ThreadActivity value) {
+    KScopedSchedulerLock lock(kernel);
+
+    auto sched_status = GetSchedulingStatus();
+
+    if (sched_status != ThreadSchedStatus::Runnable && sched_status != ThreadSchedStatus::Paused) {
+        return ERR_INVALID_STATE;
+    }
+
+    if (IsPendingTermination()) {
+        return RESULT_SUCCESS;
+    }
+
+    if (value == ThreadActivity::Paused) {
+        if ((pausing_state & static_cast<u32>(ThreadSchedFlags::ThreadPauseFlag)) != 0) {
+            return ERR_INVALID_STATE;
+        }
+        AddSchedulingFlag(ThreadSchedFlags::ThreadPauseFlag);
+    } else {
+        if ((pausing_state & static_cast<u32>(ThreadSchedFlags::ThreadPauseFlag)) == 0) {
+            return ERR_INVALID_STATE;
+        }
+        RemoveSchedulingFlag(ThreadSchedFlags::ThreadPauseFlag);
+    }
+    return RESULT_SUCCESS;
+}
+
+ResultCode Thread::Sleep(s64 nanoseconds) {
+    Handle event_handle{};
+    {
+        KScopedSchedulerLockAndSleep lock(kernel, event_handle, this, nanoseconds);
+        SetStatus(ThreadStatus::WaitSleep);
+    }
+
+    if (event_handle != InvalidHandle) {
+        auto& time_manager = kernel.TimeManager();
+        time_manager.UnscheduleTimeEvent(event_handle);
+    }
+    return RESULT_SUCCESS;
+}
+
+void Thread::AddSchedulingFlag(ThreadSchedFlags flag) {
+    const u32 old_state = scheduling_state;
+    pausing_state |= static_cast<u32>(flag);
+    const u32 base_scheduling = static_cast<u32>(GetSchedulingStatus());
+    scheduling_state = base_scheduling | pausing_state;
+    KScheduler::OnThreadStateChanged(kernel, this, old_state);
+}
+
+void Thread::RemoveSchedulingFlag(ThreadSchedFlags flag) {
+    const u32 old_state = scheduling_state;
+    pausing_state &= ~static_cast<u32>(flag);
+    const u32 base_scheduling = static_cast<u32>(GetSchedulingStatus());
+    scheduling_state = base_scheduling | pausing_state;
+    KScheduler::OnThreadStateChanged(kernel, this, old_state);
+}
+
+void Thread::SetSchedulingStatus(ThreadSchedStatus new_status) {
+    const u32 old_state = scheduling_state;
+    scheduling_state = (scheduling_state & static_cast<u32>(ThreadSchedMasks::HighMask)) |
+                       static_cast<u32>(new_status);
+    KScheduler::OnThreadStateChanged(kernel, this, old_state);
+}
+
+void Thread::SetCurrentPriority(u32 new_priority) {
+    const u32 old_priority = std::exchange(current_priority, new_priority);
+    KScheduler::OnThreadPriorityChanged(kernel, this, kernel.CurrentScheduler()->GetCurrentThread(),
+                                        old_priority);
+}
+
+ResultCode Thread::SetCoreAndAffinityMask(s32 new_core, u64 new_affinity_mask) {
+    KScopedSchedulerLock lock(kernel);
+    const auto HighestSetCore = [](u64 mask, u32 max_cores) {
+        for (s32 core = static_cast<s32>(max_cores - 1); core >= 0; core--) {
+            if (((mask >> core) & 1) != 0) {
+                return core;
+            }
+        }
+        return -1;
+    };
+
+    const bool use_override = affinity_override_count != 0;
+    if (new_core == THREADPROCESSORID_DONT_UPDATE) {
+        new_core = use_override ? ideal_core_override : ideal_core;
+        if ((new_affinity_mask & (1ULL << new_core)) == 0) {
+            LOG_ERROR(Kernel, "New affinity mask is incorrect! new_core={}, new_affinity_mask={}",
+                      new_core, new_affinity_mask);
+            return ERR_INVALID_COMBINATION;
+        }
+    }
+    if (use_override) {
+        ideal_core_override = new_core;
+    } else {
+        const auto old_affinity_mask = affinity_mask;
+        affinity_mask.SetAffinityMask(new_affinity_mask);
+        ideal_core = new_core;
+        if (old_affinity_mask.GetAffinityMask() != new_affinity_mask) {
+            const s32 old_core = processor_id;
+            if (processor_id >= 0 && !affinity_mask.GetAffinity(processor_id)) {
+                if (static_cast<s32>(ideal_core) < 0) {
+                    processor_id = HighestSetCore(affinity_mask.GetAffinityMask(),
+                                                  Core::Hardware::NUM_CPU_CORES);
+                } else {
+                    processor_id = ideal_core;
+                }
+            }
+            KScheduler::OnThreadAffinityMaskChanged(kernel, this, old_affinity_mask, old_core);
+        }
+    }
+    return RESULT_SUCCESS;
+}
+
+} // namespace Kernel

src/core/hle/kernel/thread.h

@@ -0,0 +1,731 @@
+// Copyright 2014 Citra Emulator Project / PPSSPP Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <functional>
+#include <string>
+#include <utility>
+#include <vector>
+
+#include "common/common_types.h"
+#include "common/spin_lock.h"
+#include "core/arm/arm_interface.h"
+#include "core/hle/kernel/k_affinity_mask.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/synchronization_object.h"
+#include "core/hle/result.h"
+
+namespace Common {
+class Fiber;
+}
+
+namespace Core {
+class ARM_Interface;
+class System;
+} // namespace Core
+
+namespace Kernel {
+
+class GlobalSchedulerContext;
+class KernelCore;
+class Process;
+class KScheduler;
+
+enum ThreadPriority : u32 {
+    THREADPRIO_HIGHEST = 0,            ///< Highest thread priority
+    THREADPRIO_MAX_CORE_MIGRATION = 2, ///< Highest priority for a core migration
+    THREADPRIO_USERLAND_MAX = 24,      ///< Highest thread priority for userland apps
+    THREADPRIO_DEFAULT = 44,           ///< Default thread priority for userland apps
+    THREADPRIO_LOWEST = 63,            ///< Lowest thread priority
+    THREADPRIO_COUNT = 64,             ///< Total number of possible thread priorities.
+};
+
+enum ThreadType : u32 {
+    THREADTYPE_USER = 0x1,
+    THREADTYPE_KERNEL = 0x2,
+    THREADTYPE_HLE = 0x4,
+    THREADTYPE_IDLE = 0x8,
+    THREADTYPE_SUSPEND = 0x10,
+};
+
+enum ThreadProcessorId : s32 {
+    /// Indicates that no particular processor core is preferred.
+    THREADPROCESSORID_DONT_CARE = -1,
+
+    /// Run thread on the ideal core specified by the process.
+    THREADPROCESSORID_IDEAL = -2,
+
+    /// Indicates that the preferred processor ID shouldn't be updated in
+    /// a core mask setting operation.
+    THREADPROCESSORID_DONT_UPDATE = -3,
+
+    THREADPROCESSORID_0 = 0,   ///< Run thread on core 0
+    THREADPROCESSORID_1 = 1,   ///< Run thread on core 1
+    THREADPROCESSORID_2 = 2,   ///< Run thread on core 2
+    THREADPROCESSORID_3 = 3,   ///< Run thread on core 3
+    THREADPROCESSORID_MAX = 4, ///< Processor ID must be less than this
+
+    /// Allowed CPU mask
+    THREADPROCESSORID_DEFAULT_MASK = (1 << THREADPROCESSORID_0) | (1 << THREADPROCESSORID_1) |
+                                     (1 << THREADPROCESSORID_2) | (1 << THREADPROCESSORID_3)
+};
+
+enum class ThreadStatus {
+    Ready,        ///< Ready to run
+    Paused,       ///< Paused by SetThreadActivity or debug
+    WaitHLEEvent, ///< Waiting for hle event to finish
+    WaitSleep,    ///< Waiting due to a SleepThread SVC
+    WaitIPC,      ///< Waiting for the reply from an IPC request
+    WaitSynch,    ///< Waiting due to WaitSynchronization
+    WaitMutex,    ///< Waiting due to an ArbitrateLock svc
+    WaitCondVar,  ///< Waiting due to an WaitProcessWideKey svc
+    WaitArb,      ///< Waiting due to a SignalToAddress/WaitForAddress svc
+    Dormant,      ///< Created but not yet made ready
+    Dead          ///< Run to completion, or forcefully terminated
+};
+
+enum class ThreadWakeupReason {
+    Signal, // The thread was woken up by WakeupAllWaitingThreads due to an object signal.
+    Timeout // The thread was woken up due to a wait timeout.
+};
+
+enum class ThreadActivity : u32 {
+    Normal = 0,
+    Paused = 1,
+};
+
+enum class ThreadSchedStatus : u32 {
+    None = 0,
+    Paused = 1,
+    Runnable = 2,
+    Exited = 3,
+};
+
+enum class ThreadSchedFlags : u32 {
+    ProcessPauseFlag = 1 << 4,
+    ThreadPauseFlag = 1 << 5,
+    ProcessDebugPauseFlag = 1 << 6,
+    KernelInitPauseFlag = 1 << 8,
+};
+
+enum class ThreadSchedMasks : u32 {
+    LowMask = 0x000f,
+    HighMask = 0xfff0,
+    ForcePauseMask = 0x0070,
+};
+
+class Thread final : public SynchronizationObject {
+public:
+    explicit Thread(KernelCore& kernel);
+    ~Thread() override;
+
+    using MutexWaitingThreads = std::vector<std::shared_ptr<Thread>>;
+
+    using ThreadContext32 = Core::ARM_Interface::ThreadContext32;
+    using ThreadContext64 = Core::ARM_Interface::ThreadContext64;
+
+    using ThreadSynchronizationObjects = std::vector<std::shared_ptr<SynchronizationObject>>;
+
+    using HLECallback = std::function<bool(std::shared_ptr<Thread> thread)>;
+
+    /**
+     * Creates and returns a new thread. The new thread is immediately scheduled
+     * @param system The instance of the whole system
+     * @param name The friendly name desired for the thread
+     * @param entry_point The address at which the thread should start execution
+     * @param priority The thread's priority
+     * @param arg User data to pass to the thread
+     * @param processor_id The ID(s) of the processors on which the thread is desired to be run
+     * @param stack_top The address of the thread's stack top
+     * @param owner_process The parent process for the thread, if null, it's a kernel thread
+     * @return A shared pointer to the newly created thread
+     */
+    static ResultVal<std::shared_ptr<Thread>> Create(Core::System& system, ThreadType type_flags,
+                                                     std::string name, VAddr entry_point,
+                                                     u32 priority, u64 arg, s32 processor_id,
+                                                     VAddr stack_top, Process* owner_process);
+
+    /**
+     * Creates and returns a new thread. The new thread is immediately scheduled
+     * @param system The instance of the whole system
+     * @param name The friendly name desired for the thread
+     * @param entry_point The address at which the thread should start execution
+     * @param priority The thread's priority
+     * @param arg User data to pass to the thread
+     * @param processor_id The ID(s) of the processors on which the thread is desired to be run
+     * @param stack_top The address of the thread's stack top
+     * @param owner_process The parent process for the thread, if null, it's a kernel thread
+     * @param thread_start_func The function where the host context will start.
+     * @param thread_start_parameter The parameter which will passed to host context on init
+     * @return A shared pointer to the newly created thread
+     */
+    static ResultVal<std::shared_ptr<Thread>> Create(Core::System& system, ThreadType type_flags,
+                                                     std::string name, VAddr entry_point,
+                                                     u32 priority, u64 arg, s32 processor_id,
+                                                     VAddr stack_top, Process* owner_process,
+                                                     std::function<void(void*)>&& thread_start_func,
+                                                     void* thread_start_parameter);
+
+    std::string GetName() const override {
+        return name;
+    }
+
+    void SetName(std::string new_name) {
+        name = std::move(new_name);
+    }
+
+    std::string GetTypeName() const override {
+        return "Thread";
+    }
+
+    static constexpr HandleType HANDLE_TYPE = HandleType::Thread;
+    HandleType GetHandleType() const override {
+        return HANDLE_TYPE;
+    }
+
+    bool ShouldWait(const Thread* thread) const override;
+    void Acquire(Thread* thread) override;
+    bool IsSignaled() const override;
+
+    /**
+     * Gets the thread's current priority
+     * @return The current thread's priority
+     */
+    u32 GetPriority() const {
+        return current_priority;
+    }
+
+    /**
+     * Gets the thread's nominal priority.
+     * @return The current thread's nominal priority.
+     */
+    u32 GetNominalPriority() const {
+        return nominal_priority;
+    }
+
+    /**
+     * Sets the thread's current priority
+     * @param priority The new priority
+     */
+    void SetPriority(u32 priority);
+
+    /// Adds a thread to the list of threads that are waiting for a lock held by this thread.
+    void AddMutexWaiter(std::shared_ptr<Thread> thread);
+
+    /// Removes a thread from the list of threads that are waiting for a lock held by this thread.
+    void RemoveMutexWaiter(std::shared_ptr<Thread> thread);
+
+    /// Recalculates the current priority taking into account priority inheritance.
+    void UpdatePriority();
+
+    /// Changes the core that the thread is running or scheduled to run on.
+    ResultCode SetCoreAndAffinityMask(s32 new_core, u64 new_affinity_mask);
+
+    /**
+     * Gets the thread's thread ID
+     * @return The thread's ID
+     */
+    u64 GetThreadID() const {
+        return thread_id;
+    }
+
+    /// Resumes a thread from waiting
+    void ResumeFromWait();
+
+    void OnWakeUp();
+
+    ResultCode Start();
+
+    /// Cancels a waiting operation that this thread may or may not be within.
+    ///
+    /// When the thread is within a waiting state, this will set the thread's
+    /// waiting result to signal a canceled wait. The function will then resume
+    /// this thread.
+    ///
+    void CancelWait();
+
+    void SetSynchronizationResults(SynchronizationObject* object, ResultCode result);
+
+    SynchronizationObject* GetSignalingObject() const {
+        return signaling_object;
+    }
+
+    ResultCode GetSignalingResult() const {
+        return signaling_result;
+    }
+
+    /**
+     * Retrieves the index that this particular object occupies in the list of objects
+     * that the thread passed to WaitSynchronization, starting the search from the last element.
+     *
+     * It is used to set the output index of WaitSynchronization when the thread is awakened.
+     *
+     * When a thread wakes up due to an object signal, the kernel will use the index of the last
+     * matching object in the wait objects list in case of having multiple instances of the same
+     * object in the list.
+     *
+     * @param object Object to query the index of.
+     */
+    s32 GetSynchronizationObjectIndex(std::shared_ptr<SynchronizationObject> object) const;
+
+    /**
+     * Stops a thread, invalidating it from further use
+     */
+    void Stop();
+
+    /*
+     * Returns the Thread Local Storage address of the current thread
+     * @returns VAddr of the thread's TLS
+     */
+    VAddr GetTLSAddress() const {
+        return tls_address;
+    }
+
+    /*
+     * Returns the value of the TPIDR_EL0 Read/Write system register for this thread.
+     * @returns The value of the TPIDR_EL0 register.
+     */
+    u64 GetTPIDR_EL0() const {
+        return tpidr_el0;
+    }
+
+    /// Sets the value of the TPIDR_EL0 Read/Write system register for this thread.
+    void SetTPIDR_EL0(u64 value) {
+        tpidr_el0 = value;
+    }
+
+    /*
+     * Returns the address of the current thread's command buffer, located in the TLS.
+     * @returns VAddr of the thread's command buffer.
+     */
+    VAddr GetCommandBufferAddress() const;
+
+    ThreadContext32& GetContext32() {
+        return context_32;
+    }
+
+    const ThreadContext32& GetContext32() const {
+        return context_32;
+    }
+
+    ThreadContext64& GetContext64() {
+        return context_64;
+    }
+
+    const ThreadContext64& GetContext64() const {
+        return context_64;
+    }
+
+    bool IsHLEThread() const {
+        return (type & THREADTYPE_HLE) != 0;
+    }
+
+    bool IsSuspendThread() const {
+        return (type & THREADTYPE_SUSPEND) != 0;
+    }
+
+    bool IsIdleThread() const {
+        return (type & THREADTYPE_IDLE) != 0;
+    }
+
+    bool WasRunning() const {
+        return was_running;
+    }
+
+    void SetWasRunning(bool value) {
+        was_running = value;
+    }
+
+    std::shared_ptr<Common::Fiber>& GetHostContext();
+
+    ThreadStatus GetStatus() const {
+        return status;
+    }
+
+    void SetStatus(ThreadStatus new_status);
+
+    s64 GetLastScheduledTick() const {
+        return this->last_scheduled_tick;
+    }
+
+    void SetLastScheduledTick(s64 tick) {
+        this->last_scheduled_tick = tick;
+    }
+
+    u64 GetTotalCPUTimeTicks() const {
+        return total_cpu_time_ticks;
+    }
+
+    void UpdateCPUTimeTicks(u64 ticks) {
+        total_cpu_time_ticks += ticks;
+    }
+
+    s32 GetProcessorID() const {
+        return processor_id;
+    }
+
+    s32 GetActiveCore() const {
+        return GetProcessorID();
+    }
+
+    void SetProcessorID(s32 new_core) {
+        processor_id = new_core;
+    }
+
+    void SetActiveCore(s32 new_core) {
+        processor_id = new_core;
+    }
+
+    Process* GetOwnerProcess() {
+        return owner_process;
+    }
+
+    const Process* GetOwnerProcess() const {
+        return owner_process;
+    }
+
+    const ThreadSynchronizationObjects& GetSynchronizationObjects() const {
+        return *wait_objects;
+    }
+
+    void SetSynchronizationObjects(ThreadSynchronizationObjects* objects) {
+        wait_objects = objects;
+    }
+
+    void ClearSynchronizationObjects() {
+        for (const auto& waiting_object : *wait_objects) {
+            waiting_object->RemoveWaitingThread(SharedFrom(this));
+        }
+        wait_objects->clear();
+    }
+
+    /// Determines whether all the objects this thread is waiting on are ready.
+    bool AllSynchronizationObjectsReady() const;
+
+    const MutexWaitingThreads& GetMutexWaitingThreads() const {
+        return wait_mutex_threads;
+    }
+
+    Thread* GetLockOwner() const {
+        return lock_owner.get();
+    }
+
+    void SetLockOwner(std::shared_ptr<Thread> owner) {
+        lock_owner = std::move(owner);
+    }
+
+    VAddr GetCondVarWaitAddress() const {
+        return condvar_wait_address;
+    }
+
+    void SetCondVarWaitAddress(VAddr address) {
+        condvar_wait_address = address;
+    }
+
+    VAddr GetMutexWaitAddress() const {
+        return mutex_wait_address;
+    }
+
+    void SetMutexWaitAddress(VAddr address) {
+        mutex_wait_address = address;
+    }
+
+    Handle GetWaitHandle() const {
+        return wait_handle;
+    }
+
+    void SetWaitHandle(Handle handle) {
+        wait_handle = handle;
+    }
+
+    VAddr GetArbiterWaitAddress() const {
+        return arb_wait_address;
+    }
+
+    void SetArbiterWaitAddress(VAddr address) {
+        arb_wait_address = address;
+    }
+
+    bool HasHLECallback() const {
+        return hle_callback != nullptr;
+    }
+
+    void SetHLECallback(HLECallback callback) {
+        hle_callback = std::move(callback);
+    }
+
+    void SetHLETimeEvent(Handle time_event) {
+        hle_time_event = time_event;
+    }
+
+    void SetHLESyncObject(SynchronizationObject* object) {
+        hle_object = object;
+    }
+
+    Handle GetHLETimeEvent() const {
+        return hle_time_event;
+    }
+
+    SynchronizationObject* GetHLESyncObject() const {
+        return hle_object;
+    }
+
+    void InvalidateHLECallback() {
+        SetHLECallback(nullptr);
+    }
+
+    bool InvokeHLECallback(std::shared_ptr<Thread> thread);
+
+    u32 GetIdealCore() const {
+        return ideal_core;
+    }
+
+    const KAffinityMask& GetAffinityMask() const {
+        return affinity_mask;
+    }
+
+    ResultCode SetActivity(ThreadActivity value);
+
+    /// Sleeps this thread for the given amount of nanoseconds.
+    ResultCode Sleep(s64 nanoseconds);
+
+    s64 GetYieldScheduleCount() const {
+        return this->schedule_count;
+    }
+
+    void SetYieldScheduleCount(s64 count) {
+        this->schedule_count = count;
+    }
+
+    ThreadSchedStatus GetSchedulingStatus() const {
+        return static_cast<ThreadSchedStatus>(scheduling_state &
+                                              static_cast<u32>(ThreadSchedMasks::LowMask));
+    }
+
+    bool IsRunnable() const {
+        return scheduling_state == static_cast<u32>(ThreadSchedStatus::Runnable);
+    }
+
+    bool IsRunning() const {
+        return is_running;
+    }
+
+    void SetIsRunning(bool value) {
+        is_running = value;
+    }
+
+    bool IsSyncCancelled() const {
+        return is_sync_cancelled;
+    }
+
+    void SetSyncCancelled(bool value) {
+        is_sync_cancelled = value;
+    }
+
+    Handle GetGlobalHandle() const {
+        return global_handle;
+    }
+
+    bool IsWaitingForArbitration() const {
+        return waiting_for_arbitration;
+    }
+
+    void WaitForArbitration(bool set) {
+        waiting_for_arbitration = set;
+    }
+
+    bool IsWaitingSync() const {
+        return is_waiting_on_sync;
+    }
+
+    void SetWaitingSync(bool is_waiting) {
+        is_waiting_on_sync = is_waiting;
+    }
+
+    bool IsPendingTermination() const {
+        return will_be_terminated || GetSchedulingStatus() == ThreadSchedStatus::Exited;
+    }
+
+    bool IsPaused() const {
+        return pausing_state != 0;
+    }
+
+    bool IsContinuousOnSVC() const {
+        return is_continuous_on_svc;
+    }
+
+    void SetContinuousOnSVC(bool is_continuous) {
+        is_continuous_on_svc = is_continuous;
+    }
+
+    bool IsPhantomMode() const {
+        return is_phantom_mode;
+    }
+
+    void SetPhantomMode(bool phantom) {
+        is_phantom_mode = phantom;
+    }
+
+    bool HasExited() const {
+        return has_exited;
+    }
+
+    class QueueEntry {
+    public:
+        constexpr QueueEntry() = default;
+
+        constexpr void Initialize() {
+            this->prev = nullptr;
+            this->next = nullptr;
+        }
+
+        constexpr Thread* GetPrev() const {
+            return this->prev;
+        }
+        constexpr Thread* GetNext() const {
+            return this->next;
+        }
+        constexpr void SetPrev(Thread* thread) {
+            this->prev = thread;
+        }
+        constexpr void SetNext(Thread* thread) {
+            this->next = thread;
+        }
+
+    private:
+        Thread* prev{};
+        Thread* next{};
+    };
+
+    QueueEntry& GetPriorityQueueEntry(s32 core) {
+        return this->per_core_priority_queue_entry[core];
+    }
+
+    const QueueEntry& GetPriorityQueueEntry(s32 core) const {
+        return this->per_core_priority_queue_entry[core];
+    }
+
+    s32 GetDisableDispatchCount() const {
+        return disable_count;
+    }
+
+    void DisableDispatch() {
+        ASSERT(GetDisableDispatchCount() >= 0);
+        disable_count++;
+    }
+
+    void EnableDispatch() {
+        ASSERT(GetDisableDispatchCount() > 0);
+        disable_count--;
+    }
+
+private:
+    friend class GlobalSchedulerContext;
+    friend class KScheduler;
+    friend class Process;
+
+    void SetSchedulingStatus(ThreadSchedStatus new_status);
+    void AddSchedulingFlag(ThreadSchedFlags flag);
+    void RemoveSchedulingFlag(ThreadSchedFlags flag);
+
+    void SetCurrentPriority(u32 new_priority);
+
+    Common::SpinLock context_guard{};
+    ThreadContext32 context_32{};
+    ThreadContext64 context_64{};
+    std::shared_ptr<Common::Fiber> host_context{};
+
+    ThreadStatus status = ThreadStatus::Dormant;
+    u32 scheduling_state = 0;
+
+    u64 thread_id = 0;
+
+    VAddr entry_point = 0;
+    VAddr stack_top = 0;
+    std::atomic_int disable_count = 0;
+
+    ThreadType type;
+
+    /// Nominal thread priority, as set by the emulated application.
+    /// The nominal priority is the thread priority without priority
+    /// inheritance taken into account.
+    u32 nominal_priority = 0;
+
+    /// Current thread priority. This may change over the course of the
+    /// thread's lifetime in order to facilitate priority inheritance.
+    u32 current_priority = 0;
+
+    u64 total_cpu_time_ticks = 0; ///< Total CPU running ticks.
+    s64 schedule_count{};
+    s64 last_scheduled_tick{};
+
+    s32 processor_id = 0;
+
+    VAddr tls_address = 0; ///< Virtual address of the Thread Local Storage of the thread
+    u64 tpidr_el0 = 0;     ///< TPIDR_EL0 read/write system register.
+
+    /// Process that owns this thread
+    Process* owner_process;
+
+    /// Objects that the thread is waiting on, in the same order as they were
+    /// passed to WaitSynchronization.
+    ThreadSynchronizationObjects* wait_objects;
+
+    SynchronizationObject* signaling_object;
+    ResultCode signaling_result{RESULT_SUCCESS};
+
+    /// List of threads that are waiting for a mutex that is held by this thread.
+    MutexWaitingThreads wait_mutex_threads;
+
+    /// Thread that owns the lock that this thread is waiting for.
+    std::shared_ptr<Thread> lock_owner;
+
+    /// If waiting on a ConditionVariable, this is the ConditionVariable address
+    VAddr condvar_wait_address = 0;
+    /// If waiting on a Mutex, this is the mutex address
+    VAddr mutex_wait_address = 0;
+    /// The handle used to wait for the mutex.
+    Handle wait_handle = 0;
+
+    /// If waiting for an AddressArbiter, this is the address being waited on.
+    VAddr arb_wait_address{0};
+    bool waiting_for_arbitration{};
+
+    /// Handle used as userdata to reference this object when inserting into the CoreTiming queue.
+    Handle global_handle = 0;
+
+    /// Callback for HLE Events
+    HLECallback hle_callback;
+    Handle hle_time_event;
+    SynchronizationObject* hle_object;
+
+    KScheduler* scheduler = nullptr;
+
+    std::array<QueueEntry, Core::Hardware::NUM_CPU_CORES> per_core_priority_queue_entry{};
+
+    u32 ideal_core{0xFFFFFFFF};
+    KAffinityMask affinity_mask{};
+
+    s32 ideal_core_override = -1;
+    u32 affinity_override_count = 0;
+
+    u32 pausing_state = 0;
+    bool is_running = false;
+    bool is_waiting_on_sync = false;
+    bool is_sync_cancelled = false;
+
+    bool is_continuous_on_svc = false;
+
+    bool will_be_terminated = false;
+    bool is_phantom_mode = false;
+    bool has_exited = false;
+
+    bool was_running = false;
+
+    std::string name;
+};
+
+} // namespace Kernel

src/core/hle/kernel/time_manager.cpp

@@ -0,0 +1,70 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "core/core.h"
+#include "core/core_timing.h"
+#include "core/core_timing_util.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/time_manager.h"
+
+namespace Kernel {
+
+TimeManager::TimeManager(Core::System& system_) : system{system_} {
+    time_manager_event_type = Core::Timing::CreateEvent(
+        "Kernel::TimeManagerCallback",
+        [this](std::uintptr_t thread_handle, std::chrono::nanoseconds) {
+            const KScopedSchedulerLock lock(system.Kernel());
+            const auto proper_handle = static_cast<Handle>(thread_handle);
+
+            std::shared_ptr<Thread> thread;
+            {
+                std::lock_guard lock{mutex};
+                if (cancelled_events[proper_handle]) {
+                    return;
+                }
+                thread = system.Kernel().RetrieveThreadFromGlobalHandleTable(proper_handle);
+            }
+
+            if (thread) {
+                // Thread can be null if process has exited
+                thread->OnWakeUp();
+            }
+        });
+}
+
+void TimeManager::ScheduleTimeEvent(Handle& event_handle, Thread* timetask, s64 nanoseconds) {
+    std::lock_guard lock{mutex};
+    event_handle = timetask->GetGlobalHandle();
+    if (nanoseconds > 0) {
+        ASSERT(timetask);
+        ASSERT(timetask->GetStatus() != ThreadStatus::Ready);
+        ASSERT(timetask->GetStatus() != ThreadStatus::WaitMutex);
+        system.CoreTiming().ScheduleEvent(std::chrono::nanoseconds{nanoseconds},
+                                          time_manager_event_type, event_handle);
+    } else {
+        event_handle = InvalidHandle;
+    }
+    cancelled_events[event_handle] = false;
+}
+
+void TimeManager::UnscheduleTimeEvent(Handle event_handle) {
+    std::lock_guard lock{mutex};
+    if (event_handle == InvalidHandle) {
+        return;
+    }
+    system.CoreTiming().UnscheduleEvent(time_manager_event_type, event_handle);
+    cancelled_events[event_handle] = true;
+}
+
+void TimeManager::CancelTimeEvent(Thread* time_task) {
+    std::lock_guard lock{mutex};
+    const Handle event_handle = time_task->GetGlobalHandle();
+    UnscheduleTimeEvent(event_handle);
+}
+
+} // namespace Kernel

src/core/hle/kernel/time_manager.h

@@ -0,0 +1,49 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <mutex>
+#include <unordered_map>
+
+#include "core/hle/kernel/object.h"
+
+namespace Core {
+class System;
+} // namespace Core
+
+namespace Core::Timing {
+struct EventType;
+} // namespace Core::Timing
+
+namespace Kernel {
+
+class Thread;
+
+/**
+ * The `TimeManager` takes care of scheduling time events on threads and executes their TimeUp
+ * method when the event is triggered.
+ */
+class TimeManager {
+public:
+    explicit TimeManager(Core::System& system);
+
+    /// Schedule a time event on `timetask` thread that will expire in 'nanoseconds'
+    /// returns a non-invalid handle in `event_handle` if correctly scheduled
+    void ScheduleTimeEvent(Handle& event_handle, Thread* timetask, s64 nanoseconds);
+
+    /// Unschedule an existing time event
+    void UnscheduleTimeEvent(Handle event_handle);
+
+    void CancelTimeEvent(Thread* time_task);
+
+private:
+    Core::System& system;
+    std::shared_ptr<Core::Timing::EventType> time_manager_event_type;
+    std::unordered_map<Handle, bool> cancelled_events;
+    std::mutex mutex;
+};
+
+} // namespace Kernel

src/core/hle/kernel/transfer_memory.cpp

@@ -0,0 +1,49 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/memory/page_table.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/kernel/transfer_memory.h"
+#include "core/hle/result.h"
+#include "core/memory.h"
+
+namespace Kernel {
+
+TransferMemory::TransferMemory(KernelCore& kernel, Core::Memory::Memory& memory)
+    : Object{kernel}, memory{memory} {}
+
+TransferMemory::~TransferMemory() {
+    // Release memory region when transfer memory is destroyed
+    Reset();
+}
+
+std::shared_ptr<TransferMemory> TransferMemory::Create(KernelCore& kernel,
+                                                       Core::Memory::Memory& memory,
+                                                       VAddr base_address, std::size_t size,
+                                                       Memory::MemoryPermission permissions) {
+    std::shared_ptr<TransferMemory> transfer_memory{
+        std::make_shared<TransferMemory>(kernel, memory)};
+
+    transfer_memory->base_address = base_address;
+    transfer_memory->size = size;
+    transfer_memory->owner_permissions = permissions;
+    transfer_memory->owner_process = kernel.CurrentProcess();
+
+    return transfer_memory;
+}
+
+const u8* TransferMemory::GetPointer() const {
+    return memory.GetPointer(base_address);
+}
+
+ResultCode TransferMemory::Reserve() {
+    return owner_process->PageTable().ReserveTransferMemory(base_address, size, owner_permissions);
+}
+
+ResultCode TransferMemory::Reset() {
+    return owner_process->PageTable().ResetTransferMemory(base_address, size);
+}
+
+} // namespace Kernel

src/core/hle/kernel/transfer_memory.h

@@ -0,0 +1,91 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+
+#include "core/hle/kernel/memory/memory_block.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/physical_memory.h"
+
+union ResultCode;
+
+namespace Core::Memory {
+class Memory;
+}
+
+namespace Kernel {
+
+class KernelCore;
+class Process;
+
+/// Defines the interface for transfer memory objects.
+///
+/// Transfer memory is typically used for the purpose of
+/// transferring memory between separate process instances,
+/// thus the name.
+///
+class TransferMemory final : public Object {
+public:
+    explicit TransferMemory(KernelCore& kernel, Core::Memory::Memory& memory);
+    ~TransferMemory() override;
+
+    static constexpr HandleType HANDLE_TYPE = HandleType::TransferMemory;
+
+    static std::shared_ptr<TransferMemory> Create(KernelCore& kernel, Core::Memory::Memory& memory,
+                                                  VAddr base_address, std::size_t size,
+                                                  Memory::MemoryPermission permissions);
+
+    TransferMemory(const TransferMemory&) = delete;
+    TransferMemory& operator=(const TransferMemory&) = delete;
+
+    TransferMemory(TransferMemory&&) = delete;
+    TransferMemory& operator=(TransferMemory&&) = delete;
+
+    std::string GetTypeName() const override {
+        return "TransferMemory";
+    }
+
+    std::string GetName() const override {
+        return GetTypeName();
+    }
+
+    HandleType GetHandleType() const override {
+        return HANDLE_TYPE;
+    }
+
+    /// Gets a pointer to the backing block of this instance.
+    const u8* GetPointer() const;
+
+    /// Gets the size of the memory backing this instance in bytes.
+    constexpr std::size_t GetSize() const {
+        return size;
+    }
+
+    /// Reserves the region to be used for the transfer memory, called after the transfer memory is
+    /// created.
+    ResultCode Reserve();
+
+    /// Resets the region previously used for the transfer memory, called after the transfer memory
+    /// is closed.
+    ResultCode Reset();
+
+private:
+    /// The base address for the memory managed by this instance.
+    VAddr base_address{};
+
+    /// Size of the memory, in bytes, that this instance manages.
+    std::size_t size{};
+
+    /// The memory permissions that are applied to this instance.
+    Memory::MemoryPermission owner_permissions{};
+
+    /// The process that this transfer memory instance was created under.
+    Process* owner_process{};
+
+    Core::Memory::Memory& memory;
+};
+
+} // namespace Kernel

src/core/hle/kernel/writable_event.cpp

@@ -0,0 +1,45 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include "common/assert.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/readable_event.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/writable_event.h"
+
+namespace Kernel {
+
+WritableEvent::WritableEvent(KernelCore& kernel) : Object{kernel} {}
+WritableEvent::~WritableEvent() = default;
+
+EventPair WritableEvent::CreateEventPair(KernelCore& kernel, std::string name) {
+    std::shared_ptr<WritableEvent> writable_event(new WritableEvent(kernel));
+    std::shared_ptr<ReadableEvent> readable_event(new ReadableEvent(kernel));
+
+    writable_event->name = name + ":Writable";
+    writable_event->readable = readable_event;
+    readable_event->name = name + ":Readable";
+
+    return {std::move(readable_event), std::move(writable_event)};
+}
+
+std::shared_ptr<ReadableEvent> WritableEvent::GetReadableEvent() const {
+    return readable;
+}
+
+void WritableEvent::Signal() {
+    readable->Signal();
+}
+
+void WritableEvent::Clear() {
+    readable->Clear();
+}
+
+bool WritableEvent::IsSignaled() const {
+    return readable->IsSignaled();
+}
+
+} // namespace Kernel

src/core/hle/kernel/writable_event.h

@@ -0,0 +1,59 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+
+#include "core/hle/kernel/object.h"
+
+namespace Kernel {
+
+class KernelCore;
+class ReadableEvent;
+class WritableEvent;
+
+struct EventPair {
+    std::shared_ptr<ReadableEvent> readable;
+    std::shared_ptr<WritableEvent> writable;
+};
+
+class WritableEvent final : public Object {
+public:
+    ~WritableEvent() override;
+
+    /**
+     * Creates an event
+     * @param kernel The kernel instance to create this event under.
+     * @param name Optional name of event
+     */
+    static EventPair CreateEventPair(KernelCore& kernel, std::string name = "Unknown");
+
+    std::string GetTypeName() const override {
+        return "WritableEvent";
+    }
+    std::string GetName() const override {
+        return name;
+    }
+
+    static constexpr HandleType HANDLE_TYPE = HandleType::WritableEvent;
+    HandleType GetHandleType() const override {
+        return HANDLE_TYPE;
+    }
+
+    std::shared_ptr<ReadableEvent> GetReadableEvent() const;
+
+    void Signal();
+    void Clear();
+    bool IsSignaled() const;
+
+private:
+    explicit WritableEvent(KernelCore& kernel);
+
+    std::shared_ptr<ReadableEvent> readable;
+
+    std::string name; ///< Name of event (optional)
+};
+
+} // namespace Kernel