Merge branch 'master' into feature/savestates-2

src/core/CMakeLists.txt

@@ -72,6 +72,8 @@ add_library(core STATIC
     file_sys/delay_generator.h
     file_sys/ivfc_archive.cpp
     file_sys/ivfc_archive.h
+    file_sys/layered_fs.cpp
+    file_sys/layered_fs.h
     file_sys/ncch_container.cpp
     file_sys/ncch_container.h
     file_sys/patch.cpp
@@ -469,9 +471,17 @@ create_target_directory_groups(core)
 
 target_link_libraries(core PUBLIC common PRIVATE audio_core network video_core)
 target_link_libraries(core PUBLIC Boost::boost PRIVATE cryptopp fmt open_source_archives Boost::serialization)
+
 if (ENABLE_WEB_SERVICE)
-    target_compile_definitions(core PRIVATE -DENABLE_WEB_SERVICE)
-    target_link_libraries(core PRIVATE web_service)
+    get_directory_property(OPENSSL_LIBS
+        DIRECTORY ${PROJECT_SOURCE_DIR}/externals/libressl
+        DEFINITION OPENSSL_LIBS)
+
+    target_compile_definitions(core PRIVATE -DENABLE_WEB_SERVICE -DCPPHTTPLIB_OPENSSL_SUPPORT)
+    target_link_libraries(core PRIVATE web_service ${OPENSSL_LIBS} httplib lurlparser)
+    if (ANDROID)
+        target_link_libraries(core PRIVATE ifaddrs)
+    endif()
 endif()
 
 if (ARCHITECTURE_x86_64)

src/core/arm/arm_interface.h

@@ -10,6 +10,7 @@
 #include "common/common_types.h"
 #include "core/arm/skyeye_common/arm_regformat.h"
 #include "core/arm/skyeye_common/vfp/asm_vfp.h"
+#include "core/core_timing.h"
 
 namespace Memory {
 struct PageTable;
@@ -18,6 +19,8 @@ struct PageTable;
 /// Generic ARM11 CPU interface
 class ARM_Interface : NonCopyable {
 public:
+    explicit ARM_Interface(u32 id, std::shared_ptr<Core::Timing::Timer> timer)
+        : timer(timer), id(id){};
     virtual ~ARM_Interface() {}
 
     class ThreadContext {
@@ -222,11 +225,26 @@ public:
 
     virtual void PurgeState() = 0;
 
+    std::shared_ptr<Core::Timing::Timer> GetTimer() {
+        return timer;
+    }
+
+    u32 GetID() const {
+        return id;
+    }
+
+protected:
+    std::shared_ptr<Core::Timing::Timer> timer;
+
 private:
+    u32 id;
+
     friend class boost::serialization::access;
 
     template <class Archive>
     void save(Archive& ar, const unsigned int file_version) const {
+        ar << timer;
+        ar << id;
         auto page_table = GetPageTable();
         ar << page_table;
         for (auto i = 0; i < 15; i++) {
@@ -254,6 +272,8 @@ private:
     template <class Archive>
     void load(Archive& ar, const unsigned int file_version) {
         PurgeState();
+        ar >> timer;
+        ar >> id;
         std::shared_ptr<Memory::PageTable> page_table = nullptr;
         ar >> page_table;
         SetPageTable(page_table);

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -72,8 +72,7 @@ private:
 class DynarmicUserCallbacks final : public Dynarmic::A32::UserCallbacks {
 public:
     explicit DynarmicUserCallbacks(ARM_Dynarmic& parent)
-        : parent(parent), timing(parent.system.CoreTiming()), svc_context(parent.system),
-          memory(parent.memory) {}
+        : parent(parent), svc_context(parent.system), memory(parent.memory) {}
     ~DynarmicUserCallbacks() = default;
 
     std::uint8_t MemoryRead8(VAddr vaddr) override {
@@ -137,7 +136,7 @@ public:
                 parent.jit->HaltExecution();
                 parent.SetPC(pc);
                 Kernel::Thread* thread =
-                    parent.system.Kernel().GetThreadManager().GetCurrentThread();
+                    parent.system.Kernel().GetCurrentThreadManager().GetCurrentThread();
                 parent.SaveContext(thread->context);
                 GDBStub::Break();
                 GDBStub::SendTrap(thread, 5);
@@ -150,22 +149,23 @@ public:
     }
 
     void AddTicks(std::uint64_t ticks) override {
-        timing.AddTicks(ticks);
+        parent.GetTimer()->AddTicks(ticks);
     }
     std::uint64_t GetTicksRemaining() override {
-        s64 ticks = timing.GetDowncount();
+        s64 ticks = parent.GetTimer()->GetDowncount();
         return static_cast<u64>(ticks <= 0 ? 0 : ticks);
     }
 
     ARM_Dynarmic& parent;
-    Core::Timing& timing;
     Kernel::SVCContext svc_context;
     Memory::MemorySystem& memory;
 };
 
 ARM_Dynarmic::ARM_Dynarmic(Core::System* system, Memory::MemorySystem& memory,
-                           PrivilegeMode initial_mode)
-    : system(*system), memory(memory), cb(std::make_unique<DynarmicUserCallbacks>(*this)) {
+                           PrivilegeMode initial_mode, u32 id,
+                           std::shared_ptr<Core::Timing::Timer> timer)
+    : ARM_Interface(id, timer), system(*system), memory(memory),
+      cb(std::make_unique<DynarmicUserCallbacks>(*this)) {
     interpreter_state = std::make_shared<ARMul_State>(system, memory, initial_mode);
     SetPageTable(memory.GetCurrentPageTable());
 }

src/core/arm/dynarmic/arm_dynarmic.h

@@ -24,7 +24,8 @@ class DynarmicUserCallbacks;
 
 class ARM_Dynarmic final : public ARM_Interface {
 public:
-    ARM_Dynarmic(Core::System* system, Memory::MemorySystem& memory, PrivilegeMode initial_mode);
+    ARM_Dynarmic(Core::System* system, Memory::MemorySystem& memory, PrivilegeMode initial_mode,
+                 u32 id, std::shared_ptr<Core::Timing::Timer> timer);
     ~ARM_Dynarmic() override;
 
     void Run() override;

src/core/arm/dyncom/arm_dyncom.cpp

@@ -69,8 +69,9 @@ private:
 };
 
 ARM_DynCom::ARM_DynCom(Core::System* system, Memory::MemorySystem& memory,
-                       PrivilegeMode initial_mode)
-    : system(system) {
+                       PrivilegeMode initial_mode, u32 id,
+                       std::shared_ptr<Core::Timing::Timer> timer)
+    : ARM_Interface(id, timer), system(system) {
     state = std::make_unique<ARMul_State>(system, memory, initial_mode);
 }
 
@@ -78,7 +79,7 @@ ARM_DynCom::~ARM_DynCom() {}
 
 void ARM_DynCom::Run() {
     DEBUG_ASSERT(system != nullptr);
-    ExecuteInstructions(std::max<s64>(system->CoreTiming().GetDowncount(), 0));
+    ExecuteInstructions(std::max<s64>(timer->GetDowncount(), 0));
 }
 
 void ARM_DynCom::Step() {
@@ -156,7 +157,7 @@ void ARM_DynCom::ExecuteInstructions(u64 num_instructions) {
     state->NumInstrsToExecute = num_instructions;
     unsigned ticks_executed = InterpreterMainLoop(state.get());
     if (system != nullptr) {
-        system->CoreTiming().AddTicks(ticks_executed);
+        timer->AddTicks(ticks_executed);
     }
     state->ServeBreak();
 }

src/core/arm/dyncom/arm_dyncom.h

@@ -21,7 +21,8 @@ class MemorySystem;
 class ARM_DynCom final : public ARM_Interface {
 public:
     explicit ARM_DynCom(Core::System* system, Memory::MemorySystem& memory,
-                        PrivilegeMode initial_mode);
+                        PrivilegeMode initial_mode, u32 id,
+                        std::shared_ptr<Core::Timing::Timer> timer);
     ~ARM_DynCom() override;
 
     void Run() override;

src/core/arm/dyncom/arm_dyncom_interpreter.cpp

@@ -3865,7 +3865,7 @@ SWI_INST : {
     if (inst_base->cond == ConditionCode::AL || CondPassed(cpu, inst_base->cond)) {
         DEBUG_ASSERT(cpu->system != nullptr);
         swi_inst* const inst_cream = (swi_inst*)inst_base->component;
-        cpu->system->CoreTiming().AddTicks(num_instrs);
+        cpu->system->GetRunningCore().GetTimer()->AddTicks(num_instrs);
         cpu->NumInstrsToExecute =
             num_instrs >= cpu->NumInstrsToExecute ? 0 : cpu->NumInstrsToExecute - num_instrs;
         num_instrs = 0;

src/core/arm/skyeye_common/armstate.cpp

@@ -607,8 +607,8 @@ void ARMul_State::ServeBreak() {
     }
 
     DEBUG_ASSERT(system != nullptr);
-    Kernel::Thread* thread = system->Kernel().GetThreadManager().GetCurrentThread();
-    system->CPU().SaveContext(thread->context);
+    Kernel::Thread* thread = system->Kernel().GetCurrentThreadManager().GetCurrentThread();
+    system->GetRunningCore().SaveContext(thread->context);
 
     if (last_bkpt_hit || GDBStub::IsMemoryBreak() || GDBStub::GetCpuStepFlag()) {
         last_bkpt_hit = false;

src/core/cheats/gateway_cheat.cpp

@@ -35,7 +35,7 @@ static inline std::enable_if_t<std::is_integral_v<T>> WriteOp(const GatewayCheat
                                                               Core::System& system) {
     u32 addr = line.address + state.offset;
     write_func(addr, static_cast<T>(line.value));
-    system.CPU().InvalidateCacheRange(addr, sizeof(T));
+    system.InvalidateCacheRange(addr, sizeof(T));
 }
 
 template <typename T, typename ReadFunction, typename CompareFunc>
@@ -105,7 +105,7 @@ static inline std::enable_if_t<std::is_integral_v<T>> IncrementiveWriteOp(
     Core::System& system) {
     u32 addr = line.value + state.offset;
     write_func(addr, static_cast<T>(state.reg));
-    system.CPU().InvalidateCacheRange(addr, sizeof(T));
+    system.InvalidateCacheRange(addr, sizeof(T));
     state.offset += sizeof(T);
 }
 
@@ -143,7 +143,8 @@ static inline void PatchOp(const GatewayCheat::CheatLine& line, State& state, Co
     }
     u32 num_bytes = line.value;
     u32 addr = line.address + state.offset;
-    system.CPU().InvalidateCacheRange(addr, num_bytes);
+    system.InvalidateCacheRange(addr, num_bytes);
+
     bool first = true;
     u32 bit_offset = 0;
     if (num_bytes > 0)

src/core/core.cpp

@@ -5,6 +5,7 @@
 
 #include <fstream>
 #include <memory>
+#include <stdexcept>
 #include <utility>
 #include <boost/serialization/array.hpp>
 #include "audio_core/dsp_interface.h"
@@ -65,7 +66,8 @@ System::~System() = default;
 
 System::ResultStatus System::RunLoop(bool tight_loop) {
     status = ResultStatus::Success;
-    if (!cpu_core) {
+    if (std::any_of(cpu_cores.begin(), cpu_cores.end(),
+                    [](std::shared_ptr<ARM_Interface> ptr) { return ptr == nullptr; })) {
         return ResultStatus::ErrorNotInitialized;
     }
 
@@ -83,22 +85,73 @@ System::ResultStatus System::RunLoop(bool tight_loop) {
         }
     }
 
-    // If we don't have a currently active thread then don't execute instructions,
-    // instead advance to the next event and try to yield to the next thread
-    if (kernel->GetThreadManager().GetCurrentThread() == nullptr) {
-        LOG_TRACE(Core_ARM11, "Idling");
-        timing->Idle();
-        timing->Advance();
-        PrepareReschedule();
-    } else {
-        timing->Advance();
-        if (tight_loop) {
-            cpu_core->Run();
-        } else {
-            cpu_core->Step();
+    // All cores should have executed the same amount of ticks. If this is not the case an event was
+    // scheduled with a cycles_into_future smaller then the current downcount.
+    // So we have to get those cores to the same global time first
+    u64 global_ticks = timing->GetGlobalTicks();
+    s64 max_delay = 0;
+    std::shared_ptr<ARM_Interface> current_core_to_execute = nullptr;
+    for (auto& cpu_core : cpu_cores) {
+        if (cpu_core->GetTimer()->GetTicks() < global_ticks) {
+            s64 delay = global_ticks - cpu_core->GetTimer()->GetTicks();
+            cpu_core->GetTimer()->Advance(delay);
+            if (max_delay < delay) {
+                max_delay = delay;
+                current_core_to_execute = cpu_core;
+            }
         }
     }
 
+    if (max_delay > 0) {
+        LOG_TRACE(Core_ARM11, "Core {} running (delayed) for {} ticks",
+                  current_core_to_execute->GetID(),
+                  current_core_to_execute->GetTimer()->GetDowncount());
+        running_core = current_core_to_execute.get();
+        kernel->SetRunningCPU(current_core_to_execute);
+        if (kernel->GetCurrentThreadManager().GetCurrentThread() == nullptr) {
+            LOG_TRACE(Core_ARM11, "Core {} idling", current_core_to_execute->GetID());
+            current_core_to_execute->GetTimer()->Idle();
+            PrepareReschedule();
+        } else {
+            if (tight_loop) {
+                current_core_to_execute->Run();
+            } else {
+                current_core_to_execute->Step();
+            }
+        }
+    } else {
+        // Now all cores are at the same global time. So we will run them one after the other
+        // with a max slice that is the minimum of all max slices of all cores
+        // TODO: Make special check for idle since we can easily revert the time of idle cores
+        s64 max_slice = Timing::MAX_SLICE_LENGTH;
+        for (const auto& cpu_core : cpu_cores) {
+            max_slice = std::min(max_slice, cpu_core->GetTimer()->GetMaxSliceLength());
+        }
+        for (auto& cpu_core : cpu_cores) {
+            cpu_core->GetTimer()->Advance(max_slice);
+        }
+        for (auto& cpu_core : cpu_cores) {
+            LOG_TRACE(Core_ARM11, "Core {} running for {} ticks", cpu_core->GetID(),
+                      cpu_core->GetTimer()->GetDowncount());
+            running_core = cpu_core.get();
+            kernel->SetRunningCPU(cpu_core);
+            // If we don't have a currently active thread then don't execute instructions,
+            // instead advance to the next event and try to yield to the next thread
+            if (kernel->GetCurrentThreadManager().GetCurrentThread() == nullptr) {
+                LOG_TRACE(Core_ARM11, "Core {} idling", cpu_core->GetID());
+                cpu_core->GetTimer()->Idle();
+                PrepareReschedule();
+            } else {
+                if (tight_loop) {
+                    cpu_core->Run();
+                } else {
+                    cpu_core->Step();
+                }
+            }
+        }
+        timing->AddToGlobalTicks(max_slice);
+    }
+
     if (GDBStub::IsServerEnabled()) {
         GDBStub::SetCpuStepFlag(false);
     }
@@ -183,7 +236,9 @@ System::ResultStatus System::Load(Frontend::EmuWindow& emu_window, const std::st
     }
 
     ASSERT(system_mode.first);
-    ResultStatus init_result{Init(emu_window, *system_mode.first)};
+    auto n3ds_mode = app_loader->LoadKernelN3dsMode();
+    ASSERT(n3ds_mode.first);
+    ResultStatus init_result{Init(emu_window, *system_mode.first, *n3ds_mode.first)};
     if (init_result != ResultStatus::Success) {
         LOG_CRITICAL(Core, "Failed to initialize system (Error {})!",
                      static_cast<u32>(init_result));
@@ -235,7 +290,7 @@ System::ResultStatus System::Load(Frontend::EmuWindow& emu_window, const std::st
 }
 
 void System::PrepareReschedule() {
-    cpu_core->PrepareReschedule();
+    running_core->PrepareReschedule();
     reschedule_pending = true;
 }
 
@@ -249,31 +304,50 @@ void System::Reschedule() {
     }
 
     reschedule_pending = false;
-    kernel->GetThreadManager().Reschedule();
+    for (const auto& core : cpu_cores) {
+        LOG_TRACE(Core_ARM11, "Reschedule core {}", core->GetID());
+        kernel->GetThreadManager(core->GetID()).Reschedule();
+    }
 }
 
-System::ResultStatus System::Init(Frontend::EmuWindow& emu_window, u32 system_mode) {
+System::ResultStatus System::Init(Frontend::EmuWindow& emu_window, u32 system_mode, u8 n3ds_mode) {
     LOG_DEBUG(HW_Memory, "initialized OK");
 
+    std::size_t num_cores = 2;
+    if (Settings::values.is_new_3ds) {
+        num_cores = 4;
+    }
+
     memory = std::make_unique<Memory::MemorySystem>();
 
-    timing = std::make_unique<Timing>();
+    timing = std::make_unique<Timing>(num_cores);
 
-    kernel = std::make_unique<Kernel::KernelSystem>(*memory, *timing,
-                                                    [this] { PrepareReschedule(); }, system_mode);
+    kernel = std::make_unique<Kernel::KernelSystem>(
+        *memory, *timing, [this] { PrepareReschedule(); }, system_mode, num_cores, n3ds_mode);
 
     if (Settings::values.use_cpu_jit) {
 #ifdef ARCHITECTURE_x86_64
-        cpu_core = std::make_shared<ARM_Dynarmic>(this, *memory, USER32MODE);
+        for (std::size_t i = 0; i < num_cores; ++i) {
+            cpu_cores.push_back(
+                std::make_shared<ARM_Dynarmic>(this, *memory, USER32MODE, i, timing->GetTimer(i)));
+        }
 #else
-        cpu_core = std::make_shared<ARM_DynCom>(this, *memory, USER32MODE);
+        for (std::size_t i = 0; i < num_cores; ++i) {
+            cpu_cores.push_back(
+                std::make_shared<ARM_DynCom>(this, *memory, USER32MODE, i, timing->GetTimer(i)));
+        }
         LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
 #endif
     } else {
-        cpu_core = std::make_shared<ARM_DynCom>(this, *memory, USER32MODE);
+        for (std::size_t i = 0; i < num_cores; ++i) {
+            cpu_cores.push_back(
+                std::make_shared<ARM_DynCom>(this, *memory, USER32MODE, i, timing->GetTimer(i)));
+        }
     }
+    running_core = cpu_cores[0].get();
 
-    kernel->SetCPU(cpu_core);
+    kernel->SetCPUs(cpu_cores);
+    kernel->SetRunningCPU(cpu_cores[0]);
 
     if (Settings::values.enable_dsp_lle) {
         dsp_core = std::make_unique<AudioCore::DspLle>(*memory,
@@ -296,7 +370,7 @@ System::ResultStatus System::Init(Frontend::EmuWindow& emu_window, u32 system_mo
 
     HW::Init(*memory);
     Service::Init(*this);
-    GDBStub::Init();
+    GDBStub::DeferStart();
 
     VideoCore::ResultStatus result = VideoCore::Init(emu_window, *memory);
     if (result != VideoCore::ResultStatus::Success) {
@@ -318,6 +392,8 @@ System::ResultStatus System::Init(Frontend::EmuWindow& emu_window, u32 system_mo
 
     LOG_DEBUG(Core, "Initialized OK");
 
+    initalized = true;
+
     return ResultStatus::Success;
 }
 
@@ -421,9 +497,10 @@ void System::Shutdown() {
     perf_stats.reset();
     rpc_server.reset();
     cheat_engine.reset();
+    archive_manager.reset();
     service_manager.reset();
     dsp_core.reset();
-    cpu_core.reset();
+    cpu_cores.clear();
     kernel.reset();
     timing.reset();
     app_loader.reset();
@@ -452,11 +529,18 @@ void System::Reset() {
 
 template <class Archive>
 void System::serialize(Archive& ar, const unsigned int file_version) {
+    u32 num_cores;
+    ar& num_cores;
+    if (num_cores != this->GetNumCores()) {
+        throw std::runtime_error("Wrong N3DS mode");
+    }
     // flush on save, don't flush on load
     bool should_flush = !Archive::is_loading::value;
     Memory::RasterizerClearAll(should_flush);
     ar&* timing.get();
-    ar&* cpu_core.get();
+    for (int i = 0; i < num_cores; i++) {
+        ar&* cpu_cores[i].get();
+    }
     ar&* service_manager.get();
     ar& GPU::g_regs;
     ar& LCD::g_regs;

src/core/core.h

@@ -148,7 +148,10 @@ public:
      * @returns True if the emulated system is powered on, otherwise false.
      */
     bool IsPoweredOn() const {
-        return cpu_core != nullptr;
+        return cpu_cores.size() > 0 &&
+               std::all_of(cpu_cores.begin(), cpu_cores.end(),
+                           [](std::shared_ptr<ARM_Interface> ptr) { return ptr != nullptr; });
+        ;
     }
 
     /**
@@ -168,8 +171,29 @@ public:
      * Gets a reference to the emulated CPU.
      * @returns A reference to the emulated CPU.
      */
-    ARM_Interface& CPU() {
-        return *cpu_core;
+
+    ARM_Interface& GetRunningCore() {
+        return *running_core;
+    };
+
+    /**
+     * Gets a reference to the emulated CPU.
+     * @param core_id The id of the core requested.
+     * @returns A reference to the emulated CPU.
+     */
+
+    ARM_Interface& GetCore(u32 core_id) {
+        return *cpu_cores[core_id];
+    };
+
+    u32 GetNumCores() const {
+        return static_cast<u32>(cpu_cores.size());
+    }
+
+    void InvalidateCacheRange(u32 start_address, std::size_t length) {
+        for (const auto& cpu : cpu_cores) {
+            cpu->InvalidateCacheRange(start_address, length);
+        }
     }
 
     /**
@@ -291,7 +315,7 @@ private:
      * @param system_mode The system mode.
      * @return ResultStatus code, indicating if the operation succeeded.
      */
-    ResultStatus Init(Frontend::EmuWindow& emu_window, u32 system_mode);
+    ResultStatus Init(Frontend::EmuWindow& emu_window, u32 system_mode, u8 n3ds_mode);
 
     /// Reschedule the core emulation
     void Reschedule();
@@ -300,7 +324,8 @@ private:
     std::unique_ptr<Loader::AppLoader> app_loader;
 
     /// ARM11 CPU core
-    std::shared_ptr<ARM_Interface> cpu_core;
+    std::vector<std::shared_ptr<ARM_Interface>> cpu_cores;
+    ARM_Interface* running_core = nullptr;
 
     /// DSP core
     std::unique_ptr<AudioCore::DspInterface> dsp_core;
@@ -342,6 +367,8 @@ private:
 private:
     static System s_instance;
 
+    bool initalized = false;
+
     ResultStatus status = ResultStatus::Success;
     std::string status_details = "";
     /// Saved variables for reset
@@ -358,8 +385,16 @@ private:
     void serialize(Archive& ar, const unsigned int file_version);
 };
 
-inline ARM_Interface& CPU() {
-    return System::GetInstance().CPU();
+inline ARM_Interface& GetRunningCore() {
+    return System::GetInstance().GetRunningCore();
+}
+
+inline ARM_Interface& GetCore(u32 core_id) {
+    return System::GetInstance().GetCore(core_id);
+}
+
+inline u32 GetNumCores() {
+    return System::GetInstance().GetNumCores();
 }
 
 inline AudioCore::DspInterface& DSP() {

src/core/core_timing.cpp

@@ -14,14 +14,22 @@ namespace Core {
 Timing* Timing::deserializing = nullptr;
 
 // Sort by time, unless the times are the same, in which case sort by the order added to the queue
-bool Timing::Event::operator>(const Event& right) const {
+bool Timing::Event::operator>(const Timing::Event& right) const {
     return std::tie(time, fifo_order) > std::tie(right.time, right.fifo_order);
 }
 
-bool Timing::Event::operator<(const Event& right) const {
+bool Timing::Event::operator<(const Timing::Event& right) const {
     return std::tie(time, fifo_order) < std::tie(right.time, right.fifo_order);
 }
 
+Timing::Timing(std::size_t num_cores) {
+    timers.resize(num_cores);
+    for (std::size_t i = 0; i < num_cores; ++i) {
+        timers[i] = std::make_shared<Timer>();
+    }
+    current_timer = timers[0];
+}
+
 TimingEventType* Timing::RegisterEvent(const std::string& name, TimedCallback callback) {
     // check for existing type with same name.
     // we want event type names to remain unique so that we can use them for serialization.
@@ -34,73 +42,102 @@ TimingEventType* Timing::RegisterEvent(const std::string& name, TimedCallback ca
     return event_type;
 }
 
-Timing::~Timing() {
+void Timing::ScheduleEvent(s64 cycles_into_future, const TimingEventType* event_type, u64 userdata,
+                           std::size_t core_id) {
+    ASSERT(event_type != nullptr);
+    std::shared_ptr<Timing::Timer> timer;
+    if (core_id == std::numeric_limits<std::size_t>::max()) {
+        timer = current_timer;
+    } else {
+        ASSERT(core_id < timers.size());
+        timer = timers.at(core_id);
+    }
+
+    s64 timeout = timer->GetTicks() + cycles_into_future;
+    if (current_timer == timer) {
+        // If this event needs to be scheduled before the next advance(), force one early
+        if (!timer->is_timer_sane)
+            timer->ForceExceptionCheck(cycles_into_future);
+
+        timer->event_queue.emplace_back(
+            Event{timeout, timer->event_fifo_id++, userdata, event_type});
+        std::push_heap(timer->event_queue.begin(), timer->event_queue.end(), std::greater<>());
+    } else {
+        timer->ts_queue.Push(Event{static_cast<s64>(timer->GetTicks() + cycles_into_future), 0,
+                                   userdata, event_type});
+    }
+}
+
+void Timing::UnscheduleEvent(const TimingEventType* event_type, u64 userdata) {
+    for (auto timer : timers) {
+        auto itr = std::remove_if(
+            timer->event_queue.begin(), timer->event_queue.end(),
+            [&](const Event& e) { return e.type == event_type && e.userdata == userdata; });
+
+        // Removing random items breaks the invariant so we have to re-establish it.
+        if (itr != timer->event_queue.end()) {
+            timer->event_queue.erase(itr, timer->event_queue.end());
+            std::make_heap(timer->event_queue.begin(), timer->event_queue.end(), std::greater<>());
+        }
+    }
+    // TODO:remove events from ts_queue
+}
+
+void Timing::RemoveEvent(const TimingEventType* event_type) {
+    for (auto timer : timers) {
+        auto itr = std::remove_if(timer->event_queue.begin(), timer->event_queue.end(),
+                                  [&](const Event& e) { return e.type == event_type; });
+
+        // Removing random items breaks the invariant so we have to re-establish it.
+        if (itr != timer->event_queue.end()) {
+            timer->event_queue.erase(itr, timer->event_queue.end());
+            std::make_heap(timer->event_queue.begin(), timer->event_queue.end(), std::greater<>());
+        }
+    }
+    // TODO:remove events from ts_queue
+}
+
+void Timing::SetCurrentTimer(std::size_t core_id) {
+    current_timer = timers[core_id];
+}
+
+s64 Timing::GetTicks() const {
+    return current_timer->GetTicks();
+}
+
+s64 Timing::GetGlobalTicks() const {
+    return global_timer;
+}
+
+std::chrono::microseconds Timing::GetGlobalTimeUs() const {
+    return std::chrono::microseconds{GetTicks() * 1000000 / BASE_CLOCK_RATE_ARM11};
+}
+
+std::shared_ptr<Timing::Timer> Timing::GetTimer(std::size_t cpu_id) {
+    return timers[cpu_id];
+}
+
+Timing::Timer::~Timer() {
     MoveEvents();
 }
 
-u64 Timing::GetTicks() const {
-    u64 ticks = static_cast<u64>(global_timer);
-    if (!is_global_timer_sane) {
+u64 Timing::Timer::GetTicks() const {
+    u64 ticks = static_cast<u64>(executed_ticks);
+    if (!is_timer_sane) {
         ticks += slice_length - downcount;
     }
     return ticks;
 }
 
-void Timing::AddTicks(u64 ticks) {
+void Timing::Timer::AddTicks(u64 ticks) {
     downcount -= ticks;
 }
 
-u64 Timing::GetIdleTicks() const {
+u64 Timing::Timer::GetIdleTicks() const {
     return static_cast<u64>(idled_cycles);
 }
 
-void Timing::ScheduleEvent(s64 cycles_into_future, const TimingEventType* event_type,
-                           u64 userdata) {
-    ASSERT(event_type != nullptr);
-    s64 timeout = GetTicks() + cycles_into_future;
-
-    // If this event needs to be scheduled before the next advance(), force one early
-    if (!is_global_timer_sane)
-        ForceExceptionCheck(cycles_into_future);
-
-    event_queue.emplace_back(Event{timeout, event_fifo_id++, userdata, event_type});
-    std::push_heap(event_queue.begin(), event_queue.end(), std::greater<>());
-}
-
-void Timing::ScheduleEventThreadsafe(s64 cycles_into_future, const TimingEventType* event_type,
-                                     u64 userdata) {
-    ts_queue.Push(Event{global_timer + cycles_into_future, 0, userdata, event_type});
-}
-
-void Timing::UnscheduleEvent(const TimingEventType* event_type, u64 userdata) {
-    auto itr = std::remove_if(event_queue.begin(), event_queue.end(), [&](const Event& e) {
-        return e.type == event_type && e.userdata == userdata;
-    });
-
-    // Removing random items breaks the invariant so we have to re-establish it.
-    if (itr != event_queue.end()) {
-        event_queue.erase(itr, event_queue.end());
-        std::make_heap(event_queue.begin(), event_queue.end(), std::greater<>());
-    }
-}
-
-void Timing::RemoveEvent(const TimingEventType* event_type) {
-    auto itr = std::remove_if(event_queue.begin(), event_queue.end(),
-                              [&](const Event& e) { return e.type == event_type; });
-
-    // Removing random items breaks the invariant so we have to re-establish it.
-    if (itr != event_queue.end()) {
-        event_queue.erase(itr, event_queue.end());
-        std::make_heap(event_queue.begin(), event_queue.end(), std::greater<>());
-    }
-}
-
-void Timing::RemoveNormalAndThreadsafeEvent(const TimingEventType* event_type) {
-    MoveEvents();
-    RemoveEvent(event_type);
-}
-
-void Timing::ForceExceptionCheck(s64 cycles) {
+void Timing::Timer::ForceExceptionCheck(s64 cycles) {
     cycles = std::max<s64>(0, cycles);
     if (downcount > cycles) {
         slice_length -= downcount - cycles;
@@ -108,7 +145,7 @@ void Timing::ForceExceptionCheck(s64 cycles) {
     }
 }
 
-void Timing::MoveEvents() {
+void Timing::Timer::MoveEvents() {
     for (Event ev; ts_queue.Pop(ev);) {
         ev.fifo_order = event_fifo_id++;
         event_queue.emplace_back(std::move(ev));
@@ -116,50 +153,54 @@ void Timing::MoveEvents() {
     }
 }
 
-void Timing::Advance() {
+s64 Timing::Timer::GetMaxSliceLength() const {
+    auto next_event = std::find_if(event_queue.begin(), event_queue.end(),
+                                   [&](const Event& e) { return e.time - executed_ticks > 0; });
+    if (next_event != event_queue.end()) {
+        return next_event->time - executed_ticks;
+    }
+    return MAX_SLICE_LENGTH;
+}
+
+void Timing::Timer::Advance(s64 max_slice_length) {
     MoveEvents();
 
     s64 cycles_executed = slice_length - downcount;
-    global_timer += cycles_executed;
-    slice_length = MAX_SLICE_LENGTH;
+    idled_cycles = 0;
+    executed_ticks += cycles_executed;
+    slice_length = max_slice_length;
 
-    is_global_timer_sane = true;
+    is_timer_sane = true;
 
-    while (!event_queue.empty() && event_queue.front().time <= global_timer) {
+    while (!event_queue.empty() && event_queue.front().time <= executed_ticks) {
         Event evt = std::move(event_queue.front());
         std::pop_heap(event_queue.begin(), event_queue.end(), std::greater<>());
         event_queue.pop_back();
-        if (event_types.find(*evt.type->name) == event_types.end()) {
-            LOG_ERROR(Core, "Unknown queued event {}", *evt.type->name);
-        } else if (evt.type->callback == nullptr) {
+        if (evt.type->callback == nullptr) {
             LOG_ERROR(Core, "Event '{}' has no callback", *evt.type->name);
         }
         if (evt.type->callback != nullptr) {
-            evt.type->callback(evt.userdata, global_timer - evt.time);
+            evt.type->callback(evt.userdata, executed_ticks - evt.time);
         }
     }
 
-    is_global_timer_sane = false;
+    is_timer_sane = false;
 
     // Still events left (scheduled in the future)
     if (!event_queue.empty()) {
         slice_length = static_cast<int>(
-            std::min<s64>(event_queue.front().time - global_timer, MAX_SLICE_LENGTH));
+            std::min<s64>(event_queue.front().time - executed_ticks, max_slice_length));
     }
 
     downcount = slice_length;
 }
 
-void Timing::Idle() {
+void Timing::Timer::Idle() {
     idled_cycles += downcount;
     downcount = 0;
 }
 
-std::chrono::microseconds Timing::GetGlobalTimeUs() const {
-    return std::chrono::microseconds{GetTicks() * 1000000 / BASE_CLOCK_RATE_ARM11};
-}
-
-s64 Timing::GetDowncount() const {
+s64 Timing::Timer::GetDowncount() const {
     return downcount;
 }
 

src/core/core_timing.h

@@ -135,65 +135,10 @@ struct TimingEventType {
 };
 
 class Timing {
-public:
-    ~Timing();
-
-    /**
-     * This should only be called from the emu thread, if you are calling it any other thread, you
-     * are doing something evil
-     */
-    u64 GetTicks() const;
-    u64 GetIdleTicks() const;
-    void AddTicks(u64 ticks);
-
-    /**
-     * Returns the event_type identifier. if name is not unique, it will assert.
-     */
-    TimingEventType* RegisterEvent(const std::string& name, TimedCallback callback);
-
-    /**
-     * After the first Advance, the slice lengths and the downcount will be reduced whenever an
-     * event is scheduled earlier than the current values. Scheduling from a callback will not
-     * update the downcount until the Advance() completes.
-     */
-    void ScheduleEvent(s64 cycles_into_future, const TimingEventType* event_type, u64 userdata = 0);
-
-    /**
-     * This is to be called when outside of hle threads, such as the graphics thread, wants to
-     * schedule things to be executed on the main thread.
-     * Not that this doesn't change slice_length and thus events scheduled by this might be called
-     * with a delay of up to MAX_SLICE_LENGTH
-     */
-    void ScheduleEventThreadsafe(s64 cycles_into_future, const TimingEventType* event_type,
-                                 u64 userdata);
-
-    void UnscheduleEvent(const TimingEventType* event_type, u64 userdata);
-
-    /// We only permit one event of each type in the queue at a time.
-    void RemoveEvent(const TimingEventType* event_type);
-    void RemoveNormalAndThreadsafeEvent(const TimingEventType* event_type);
-
-    /** Advance must be called at the beginning of dispatcher loops, not the end. Advance() ends
-     * the previous timing slice and begins the next one, you must Advance from the previous
-     * slice to the current one before executing any cycles. CoreTiming starts in slice -1 so an
-     * Advance() is required to initialize the slice length before the first cycle of emulated
-     * instructions is executed.
-     */
-    void Advance();
-    void MoveEvents();
-
-    /// Pretend that the main CPU has executed enough cycles to reach the next event.
-    void Idle();
-
-    void ForceExceptionCheck(s64 cycles);
-
-    std::chrono::microseconds GetGlobalTimeUs() const;
-
-    s64 GetDowncount() const;
-
 private:
     static Timing* deserializing;
 
+public:
     struct Event {
         s64 time;
         u64 fifo_order;
@@ -229,48 +174,116 @@ private:
 
     static constexpr int MAX_SLICE_LENGTH = 20000;
 
+    class Timer {
+    public:
+        ~Timer();
+
+        s64 GetMaxSliceLength() const;
+
+        void Advance(s64 max_slice_length = MAX_SLICE_LENGTH);
+
+        void Idle();
+
+        u64 GetTicks() const;
+        u64 GetIdleTicks() const;
+
+        void AddTicks(u64 ticks);
+
+        s64 GetDowncount() const;
+
+        void ForceExceptionCheck(s64 cycles);
+
+        void MoveEvents();
+
+    private:
+        friend class Timing;
+        // The queue is a min-heap using std::make_heap/push_heap/pop_heap.
+        // We don't use std::priority_queue because we need to be able to serialize, unserialize and
+        // erase arbitrary events (RemoveEvent()) regardless of the queue order. These aren't
+        // accomodated by the standard adaptor class.
+        std::vector<Event> event_queue;
+        u64 event_fifo_id = 0;
+        // the queue for storing the events from other threads threadsafe until they will be added
+        // to the event_queue by the emu thread
+        Common::MPSCQueue<Event> ts_queue;
+        // Are we in a function that has been called from Advance()
+        // If events are sheduled from a function that gets called from Advance(),
+        // don't change slice_length and downcount.
+        // The time between CoreTiming being intialized and the first call to Advance() is
+        // considered the slice boundary between slice -1 and slice 0. Dispatcher loops must call
+        // Advance() before executing the first cycle of each slice to prepare the slice length and
+        // downcount for that slice.
+        bool is_timer_sane = true;
+
+        s64 slice_length = MAX_SLICE_LENGTH;
+        s64 downcount = MAX_SLICE_LENGTH;
+        s64 executed_ticks = 0;
+        u64 idled_cycles;
+
+        template <class Archive>
+        void serialize(Archive& ar, const unsigned int) {
+            MoveEvents();
+            ar& slice_length;
+            ar& downcount;
+            ar& event_queue;
+            ar& event_fifo_id;
+            ar& idled_cycles;
+        }
+        friend class boost::serialization::access;
+    };
+
+    explicit Timing(std::size_t num_cores);
+
+    ~Timing(){};
+
+    /**
+     * Returns the event_type identifier. if name is not unique, it will assert.
+     */
+    TimingEventType* RegisterEvent(const std::string& name, TimedCallback callback);
+
+    void ScheduleEvent(s64 cycles_into_future, const TimingEventType* event_type, u64 userdata = 0,
+                       std::size_t core_id = std::numeric_limits<std::size_t>::max());
+
+    void UnscheduleEvent(const TimingEventType* event_type, u64 userdata);
+
+    /// We only permit one event of each type in the queue at a time.
+    void RemoveEvent(const TimingEventType* event_type);
+
+    void SetCurrentTimer(std::size_t core_id);
+
+    s64 GetTicks() const;
+
+    s64 GetGlobalTicks() const;
+
+    void AddToGlobalTicks(s64 ticks) {
+        global_timer += ticks;
+    }
+
+    std::chrono::microseconds GetGlobalTimeUs() const;
+
+    std::shared_ptr<Timer> GetTimer(std::size_t cpu_id);
+
+private:
     s64 global_timer = 0;
-    s64 slice_length = MAX_SLICE_LENGTH;
-    s64 downcount = MAX_SLICE_LENGTH;
 
     // unordered_map stores each element separately as a linked list node so pointers to
     // elements remain stable regardless of rehashes/resizing.
     std::unordered_map<std::string, TimingEventType> event_types;
 
-    // The queue is a min-heap using std::make_heap/push_heap/pop_heap.
-    // We don't use std::priority_queue because we need to be able to serialize, unserialize and
-    // erase arbitrary events (RemoveEvent()) regardless of the queue order. These aren't
-    // accomodated by the standard adaptor class.
-    std::vector<Event> event_queue;
-    u64 event_fifo_id = 0;
-    // the queue for storing the events from other threads threadsafe until they will be added
-    // to the event_queue by the emu thread
-    Common::MPSCQueue<Event> ts_queue;
-    s64 idled_cycles = 0;
-
-    // Are we in a function that has been called from Advance()
-    // If events are sheduled from a function that gets called from Advance(),
-    // don't change slice_length and downcount.
-    // The time between CoreTiming being intialized and the first call to Advance() is considered
-    // the slice boundary between slice -1 and slice 0. Dispatcher loops must call Advance() before
-    // executing the first cycle of each slice to prepare the slice length and downcount for
-    // that slice.
-    bool is_global_timer_sane = true;
+    std::vector<std::shared_ptr<Timer>> timers;
+    std::shared_ptr<Timer> current_timer;
 
     template <class Archive>
     void serialize(Archive& ar, const unsigned int) {
         // event_types set during initialization of other things
         deserializing = this;
-        MoveEvents();
         ar& global_timer;
-        ar& slice_length;
-        ar& downcount;
-        ar& event_queue;
-        ar& event_fifo_id;
-        ar& idled_cycles;
+        ar& timers;
+        ar& current_timer;
         deserializing = nullptr;
     }
     friend class boost::serialization::access;
+
 };
 
 } // namespace Core

src/core/file_sys/layered_fs.cpp

@@ -0,0 +1,604 @@
+// Copyright 2020 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cstring>
+#include "common/alignment.h"
+#include "common/assert.h"
+#include "common/common_paths.h"
+#include "common/file_util.h"
+#include "common/string_util.h"
+#include "common/swap.h"
+#include "core/file_sys/layered_fs.h"
+#include "core/file_sys/patch.h"
+
+namespace FileSys {
+
+struct FileRelocationInfo {
+    int type;                      // 0 - none, 1 - replaced / created, 2 - patched, 3 - removed
+    u64 original_offset;           // Type 0. Offset is absolute
+    std::string replace_file_path; // Type 1
+    std::vector<u8> patched_file;  // Type 2
+    u64 size;                      // Relocated file size
+};
+struct LayeredFS::File {
+    std::string name;
+    std::string path;
+    FileRelocationInfo relocation{};
+    Directory* parent;
+};
+
+struct DirectoryMetadata {
+    u32_le parent_directory_offset;
+    u32_le next_sibling_offset;
+    u32_le first_child_directory_offset;
+    u32_le first_file_offset;
+    u32_le hash_bucket_next;
+    u32_le name_length;
+    // Followed by a name of name length (aligned up to 4)
+};
+static_assert(sizeof(DirectoryMetadata) == 0x18, "Size of DirectoryMetadata is not correct");
+
+struct FileMetadata {
+    u32_le parent_directory_offset;
+    u32_le next_sibling_offset;
+    u64_le file_data_offset;
+    u64_le file_data_length;
+    u32_le hash_bucket_next;
+    u32_le name_length;
+    // Followed by a name of name length (aligned up to 4)
+};
+static_assert(sizeof(FileMetadata) == 0x20, "Size of FileMetadata is not correct");
+
+LayeredFS::LayeredFS(std::shared_ptr<RomFSReader> romfs_, std::string patch_path_,
+                     std::string patch_ext_path_, bool load_relocations)
+    : romfs(std::move(romfs_)), patch_path(std::move(patch_path_)),
+      patch_ext_path(std::move(patch_ext_path_)) {
+
+    romfs->ReadFile(0, sizeof(header), reinterpret_cast<u8*>(&header));
+
+    ASSERT_MSG(header.header_length == sizeof(header), "Header size is incorrect");
+
+    // TODO: is root always the first directory in table?
+    root.parent = &root;
+    LoadDirectory(root, 0);
+
+    if (load_relocations) {
+        LoadRelocations();
+        LoadExtRelocations();
+    }
+
+    RebuildMetadata();
+}
+
+LayeredFS::~LayeredFS() = default;
+
+void LayeredFS::LoadDirectory(Directory& current, u32 offset) {
+    DirectoryMetadata metadata;
+    romfs->ReadFile(header.directory_metadata_table.offset + offset, sizeof(metadata),
+                    reinterpret_cast<u8*>(&metadata));
+
+    current.name = ReadName(header.directory_metadata_table.offset + offset + sizeof(metadata),
+                            metadata.name_length);
+    current.path = current.parent->path + current.name + DIR_SEP;
+    directory_path_map.emplace(current.path, &current);
+
+    if (metadata.first_file_offset != 0xFFFFFFFF) {
+        LoadFile(current, metadata.first_file_offset);
+    }
+
+    if (metadata.first_child_directory_offset != 0xFFFFFFFF) {
+        auto child = std::make_unique<Directory>();
+        auto& directory = *child;
+        directory.parent = &current;
+        current.directories.emplace_back(std::move(child));
+        LoadDirectory(directory, metadata.first_child_directory_offset);
+    }
+
+    if (metadata.next_sibling_offset != 0xFFFFFFFF) {
+        auto sibling = std::make_unique<Directory>();
+        auto& directory = *sibling;
+        directory.parent = current.parent;
+        current.parent->directories.emplace_back(std::move(sibling));
+        LoadDirectory(directory, metadata.next_sibling_offset);
+    }
+}
+
+void LayeredFS::LoadFile(Directory& parent, u32 offset) {
+    FileMetadata metadata;
+    romfs->ReadFile(header.file_metadata_table.offset + offset, sizeof(metadata),
+                    reinterpret_cast<u8*>(&metadata));
+
+    auto file = std::make_unique<File>();
+    file->name = ReadName(header.file_metadata_table.offset + offset + sizeof(metadata),
+                          metadata.name_length);
+    file->path = parent.path + file->name;
+    file->relocation.original_offset = header.file_data_offset + metadata.file_data_offset;
+    file->relocation.size = metadata.file_data_length;
+    file->parent = &parent;
+
+    file_path_map.emplace(file->path, file.get());
+    parent.files.emplace_back(std::move(file));
+
+    if (metadata.next_sibling_offset != 0xFFFFFFFF) {
+        LoadFile(parent, metadata.next_sibling_offset);
+    }
+}
+
+std::string LayeredFS::ReadName(u32 offset, u32 name_length) {
+    std::vector<u16_le> buffer(name_length / sizeof(u16_le));
+    romfs->ReadFile(offset, name_length, reinterpret_cast<u8*>(buffer.data()));
+
+    std::u16string name(buffer.size(), 0);
+    std::transform(buffer.begin(), buffer.end(), name.begin(), [](u16_le character) {
+        return static_cast<char16_t>(static_cast<u16>(character));
+    });
+    return Common::UTF16ToUTF8(name);
+}
+
+void LayeredFS::LoadRelocations() {
+    if (!FileUtil::Exists(patch_path)) {
+        return;
+    }
+
+    const FileUtil::DirectoryEntryCallable callback = [this,
+                                                       &callback](u64* /*num_entries_out*/,
+                                                                  const std::string& directory,
+                                                                  const std::string& virtual_name) {
+        auto* parent = directory_path_map.at(directory.substr(patch_path.size() - 1));
+
+        if (FileUtil::IsDirectory(directory + virtual_name + DIR_SEP)) {
+            const auto path = (directory + virtual_name + DIR_SEP).substr(patch_path.size() - 1);
+            if (!directory_path_map.count(path)) { // Add this directory
+                auto directory = std::make_unique<Directory>();
+                directory->name = virtual_name;
+                directory->path = path;
+                directory->parent = parent;
+                directory_path_map.emplace(path, directory.get());
+                parent->directories.emplace_back(std::move(directory));
+                LOG_INFO(Service_FS, "LayeredFS created directory {}", path);
+            }
+            return FileUtil::ForeachDirectoryEntry(nullptr, directory + virtual_name + DIR_SEP,
+                                                   callback);
+        }
+
+        const auto path = (directory + virtual_name).substr(patch_path.size() - 1);
+        if (!file_path_map.count(path)) { // Newly created file
+            auto file = std::make_unique<File>();
+            file->name = virtual_name;
+            file->path = path;
+            file->parent = parent;
+            file_path_map.emplace(path, file.get());
+            parent->files.emplace_back(std::move(file));
+            LOG_INFO(Service_FS, "LayeredFS created file {}", path);
+        }
+
+        auto* file = file_path_map.at(path);
+        file->relocation.type = 1;
+        file->relocation.replace_file_path = directory + virtual_name;
+        file->relocation.size = FileUtil::GetSize(directory + virtual_name);
+        LOG_INFO(Service_FS, "LayeredFS replacement file in use for {}", path);
+        return true;
+    };
+
+    FileUtil::ForeachDirectoryEntry(nullptr, patch_path, callback);
+}
+
+void LayeredFS::LoadExtRelocations() {
+    if (!FileUtil::Exists(patch_ext_path)) {
+        return;
+    }
+
+    if (patch_ext_path.back() == '/' || patch_ext_path.back() == '\\') {
+        // ScanDirectoryTree expects a path without trailing '/'
+        patch_ext_path.erase(patch_ext_path.size() - 1, 1);
+    }
+
+    FileUtil::FSTEntry result;
+    FileUtil::ScanDirectoryTree(patch_ext_path, result, 256);
+
+    for (const auto& entry : result.children) {
+        if (FileUtil::IsDirectory(entry.physicalName)) {
+            continue;
+        }
+
+        const auto path = entry.physicalName.substr(patch_ext_path.size());
+        if (path.size() >= 5 && path.substr(path.size() - 5) == ".stub") {
+            // Remove the corresponding file if exists
+            const auto file_path = path.substr(0, path.size() - 5);
+            if (file_path_map.count(file_path)) {
+                auto& file = *file_path_map[file_path];
+                file.relocation.type = 3;
+                file.relocation.size = 0;
+                file_path_map.erase(file_path);
+                LOG_INFO(Service_FS, "LayeredFS removed file {}", file_path);
+            } else {
+                LOG_WARNING(Service_FS, "LayeredFS file for stub {} not found", path);
+            }
+        } else if (path.size() >= 4) {
+            const auto extension = path.substr(path.size() - 4);
+            if (extension != ".ips" && extension != ".bps") {
+                LOG_WARNING(Service_FS, "LayeredFS unknown ext file {}", path);
+            }
+
+            const auto file_path = path.substr(0, path.size() - 4);
+            if (!file_path_map.count(file_path)) {
+                LOG_WARNING(Service_FS, "LayeredFS original file for patch {} not found", path);
+                continue;
+            }
+
+            FileUtil::IOFile patch_file(entry.physicalName, "rb");
+            if (!patch_file) {
+                LOG_ERROR(Service_FS, "LayeredFS Could not open file {}", entry.physicalName);
+                continue;
+            }
+
+            const auto size = patch_file.GetSize();
+            std::vector<u8> patch(size);
+            if (patch_file.ReadBytes(patch.data(), size) != size) {
+                LOG_ERROR(Service_FS, "LayeredFS Could not read file {}", entry.physicalName);
+                continue;
+            }
+
+            auto& file = *file_path_map[file_path];
+            std::vector<u8> buffer(file.relocation.size); // Original size
+            romfs->ReadFile(file.relocation.original_offset, buffer.size(), buffer.data());
+
+            bool ret = false;
+            if (extension == ".ips") {
+                ret = Patch::ApplyIpsPatch(patch, buffer);
+            } else {
+                ret = Patch::ApplyBpsPatch(patch, buffer);
+            }
+
+            if (ret) {
+                LOG_INFO(Service_FS, "LayeredFS patched file {}", file_path);
+
+                file.relocation.type = 2;
+                file.relocation.size = buffer.size();
+                file.relocation.patched_file = std::move(buffer);
+            } else {
+                LOG_ERROR(Service_FS, "LayeredFS failed to patch file {}", file_path);
+            }
+        } else {
+            LOG_WARNING(Service_FS, "LayeredFS unknown ext file {}", path);
+        }
+    }
+}
+
+std::size_t GetNameSize(const std::string& name) {
+    std::u16string u16name = Common::UTF8ToUTF16(name);
+    return Common::AlignUp(u16name.size() * 2, 4);
+}
+
+void LayeredFS::PrepareBuildDirectory(Directory& current) {
+    directory_metadata_offset_map.emplace(&current, current_directory_offset);
+    directory_list.emplace_back(&current);
+    current_directory_offset += sizeof(DirectoryMetadata) + GetNameSize(current.name);
+}
+
+void LayeredFS::PrepareBuildFile(File& current) {
+    if (current.relocation.type == 3) { // Deleted files are not counted
+        return;
+    }
+    file_metadata_offset_map.emplace(&current, current_file_offset);
+    file_list.emplace_back(&current);
+    current_file_offset += sizeof(FileMetadata) + GetNameSize(current.name);
+}
+
+void LayeredFS::PrepareBuild(Directory& current) {
+    for (const auto& child : current.files) {
+        PrepareBuildFile(*child);
+    }
+
+    for (const auto& child : current.directories) {
+        PrepareBuildDirectory(*child);
+    }
+
+    for (const auto& child : current.directories) {
+        PrepareBuild(*child);
+    }
+}
+
+// Implementation from 3dbrew
+u32 CalcHash(const std::string& name, u32 parent_offset) {
+    u32 hash = parent_offset ^ 123456789;
+    std::u16string u16name = Common::UTF8ToUTF16(name);
+    for (char16_t c : u16name) {
+        hash = (hash >> 5) | (hash << 27);
+        hash ^= static_cast<u16>(c);
+    }
+    return hash;
+}
+
+std::size_t WriteName(u8* dest, std::u16string name) {
+    const auto buffer_size = Common::AlignUp(name.size() * 2, 4);
+    std::vector<u16_le> buffer(buffer_size / 2);
+    std::transform(name.begin(), name.end(), buffer.begin(), [](char16_t character) {
+        return static_cast<u16_le>(static_cast<u16>(character));
+    });
+    std::memcpy(dest, buffer.data(), buffer_size);
+
+    return buffer_size;
+}
+
+void LayeredFS::BuildDirectories() {
+    directory_metadata_table.resize(current_directory_offset, 0xFF);
+
+    std::size_t written = 0;
+    for (const auto& directory : directory_list) {
+        DirectoryMetadata metadata;
+        std::memset(&metadata, 0xFF, sizeof(metadata));
+        metadata.parent_directory_offset = directory_metadata_offset_map.at(directory->parent);
+
+        if (directory->parent != directory) {
+            bool flag = false;
+            for (const auto& sibling : directory->parent->directories) {
+                if (flag) {
+                    metadata.next_sibling_offset = directory_metadata_offset_map.at(sibling.get());
+                    break;
+                } else if (sibling.get() == directory) {
+                    flag = true;
+                }
+            }
+        }
+
+        if (!directory->directories.empty()) {
+            metadata.first_child_directory_offset =
+                directory_metadata_offset_map.at(directory->directories.front().get());
+        }
+
+        if (!directory->files.empty()) {
+            metadata.first_file_offset =
+                file_metadata_offset_map.at(directory->files.front().get());
+        }
+
+        const auto bucket = CalcHash(directory->name, metadata.parent_directory_offset) %
+                            directory_hash_table.size();
+        metadata.hash_bucket_next = directory_hash_table[bucket];
+        directory_hash_table[bucket] = directory_metadata_offset_map.at(directory);
+
+        // Write metadata and name
+        std::u16string u16name = Common::UTF8ToUTF16(directory->name);
+        metadata.name_length = u16name.size() * 2;
+
+        std::memcpy(directory_metadata_table.data() + written, &metadata, sizeof(metadata));
+        written += sizeof(metadata);
+
+        written += WriteName(directory_metadata_table.data() + written, u16name);
+    }
+
+    ASSERT_MSG(written == directory_metadata_table.size(),
+               "Calculated size for directory metadata table is wrong");
+}
+
+void LayeredFS::BuildFiles() {
+    file_metadata_table.resize(current_file_offset, 0xFF);
+
+    std::size_t written = 0;
+    for (const auto& file : file_list) {
+        FileMetadata metadata;
+        std::memset(&metadata, 0xFF, sizeof(metadata));
+
+        metadata.parent_directory_offset = directory_metadata_offset_map.at(file->parent);
+
+        bool flag = false;
+        for (const auto& sibling : file->parent->files) {
+            if (sibling->relocation.type == 3) { // removed file
+                continue;
+            }
+            if (flag) {
+                metadata.next_sibling_offset = file_metadata_offset_map.at(sibling.get());
+                break;
+            } else if (sibling.get() == file) {
+                flag = true;
+            }
+        }
+
+        metadata.file_data_offset = current_data_offset;
+        metadata.file_data_length = file->relocation.size;
+        current_data_offset += Common::AlignUp(metadata.file_data_length, 16);
+        if (metadata.file_data_length != 0) {
+            data_offset_map.emplace(metadata.file_data_offset, file);
+        }
+
+        const auto bucket =
+            CalcHash(file->name, metadata.parent_directory_offset) % file_hash_table.size();
+        metadata.hash_bucket_next = file_hash_table[bucket];
+        file_hash_table[bucket] = file_metadata_offset_map.at(file);
+
+        // Write metadata and name
+        std::u16string u16name = Common::UTF8ToUTF16(file->name);
+        metadata.name_length = u16name.size() * 2;
+
+        std::memcpy(file_metadata_table.data() + written, &metadata, sizeof(metadata));
+        written += sizeof(metadata);
+
+        written += WriteName(file_metadata_table.data() + written, u16name);
+    }
+
+    ASSERT_MSG(written == file_metadata_table.size(),
+               "Calculated size for file metadata table is wrong");
+}
+
+// Implementation from 3dbrew
+std::size_t GetHashTableSize(std::size_t entry_count) {
+    if (entry_count < 3) {
+        return 3;
+    } else if (entry_count < 19) {
+        return entry_count | 1;
+    } else {
+        std::size_t count = entry_count;
+        while (count % 2 == 0 || count % 3 == 0 || count % 5 == 0 || count % 7 == 0 ||
+               count % 11 == 0 || count % 13 == 0 || count % 17 == 0) {
+            count++;
+        }
+        return count;
+    }
+}
+
+void LayeredFS::RebuildMetadata() {
+    PrepareBuildDirectory(root);
+    PrepareBuild(root);
+
+    directory_hash_table.resize(GetHashTableSize(directory_list.size()), 0xFFFFFFFF);
+    file_hash_table.resize(GetHashTableSize(file_list.size()), 0xFFFFFFFF);
+
+    BuildDirectories();
+    BuildFiles();
+
+    // Create header
+    RomFSHeader header;
+    header.header_length = sizeof(header);
+    header.directory_hash_table = {
+        /*offset*/ sizeof(header),
+        /*length*/ static_cast<u32_le>(directory_hash_table.size() * sizeof(u32_le))};
+    header.directory_metadata_table = {
+        /*offset*/
+        header.directory_hash_table.offset + header.directory_hash_table.length,
+        /*length*/ static_cast<u32_le>(directory_metadata_table.size())};
+    header.file_hash_table = {
+        /*offset*/
+        header.directory_metadata_table.offset + header.directory_metadata_table.length,
+        /*length*/ static_cast<u32_le>(file_hash_table.size() * sizeof(u32_le))};
+    header.file_metadata_table = {/*offset*/ header.file_hash_table.offset +
+                                      header.file_hash_table.length,
+                                  /*length*/ static_cast<u32_le>(file_metadata_table.size())};
+    header.file_data_offset =
+        Common::AlignUp(header.file_metadata_table.offset + header.file_metadata_table.length, 16);
+
+    // Write hash table and metadata table
+    metadata.resize(header.file_data_offset);
+    std::memcpy(metadata.data(), &header, header.header_length);
+    std::memcpy(metadata.data() + header.directory_hash_table.offset, directory_hash_table.data(),
+                header.directory_hash_table.length);
+    std::memcpy(metadata.data() + header.directory_metadata_table.offset,
+                directory_metadata_table.data(), header.directory_metadata_table.length);
+    std::memcpy(metadata.data() + header.file_hash_table.offset, file_hash_table.data(),
+                header.file_hash_table.length);
+    std::memcpy(metadata.data() + header.file_metadata_table.offset, file_metadata_table.data(),
+                header.file_metadata_table.length);
+}
+
+std::size_t LayeredFS::GetSize() const {
+    return metadata.size() + current_data_offset;
+}
+
+std::size_t LayeredFS::ReadFile(std::size_t offset, std::size_t length, u8* buffer) {
+    ASSERT_MSG(offset + length <= GetSize(), "Out of bound");
+
+    std::size_t read_size = 0;
+    if (offset < metadata.size()) {
+        // First read the metadata
+        const auto to_read = std::min(metadata.size() - offset, length);
+        std::memcpy(buffer, metadata.data() + offset, to_read);
+        read_size += to_read;
+        offset = 0;
+    } else {
+        offset -= metadata.size();
+    }
+
+    // Read files
+    auto current = (--data_offset_map.upper_bound(offset));
+    while (read_size < length) {
+        const auto relative_offset = offset - current->first;
+        std::size_t to_read{};
+        if (current->second->relocation.size > relative_offset) {
+            to_read = std::min<std::size_t>(current->second->relocation.size - relative_offset,
+                                            length - read_size);
+        }
+        const auto alignment =
+            std::min<std::size_t>(Common::AlignUp(current->second->relocation.size, 16) -
+                                      relative_offset,
+                                  length - read_size) -
+            to_read;
+
+        // Read the file in different ways depending on relocation type
+        auto& relocation = current->second->relocation;
+        if (relocation.type == 0) { // none
+            romfs->ReadFile(relocation.original_offset + relative_offset, to_read,
+                            buffer + read_size);
+        } else if (relocation.type == 1) { // replace
+            FileUtil::IOFile replace_file(relocation.replace_file_path, "rb");
+            if (replace_file) {
+                replace_file.Seek(relative_offset, SEEK_SET);
+                replace_file.ReadBytes(buffer + read_size, to_read);
+            } else {
+                LOG_ERROR(Service_FS, "Could not open replacement file for {}",
+                          current->second->path);
+            }
+        } else if (relocation.type == 2) { // patch
+            std::memcpy(buffer + read_size, relocation.patched_file.data() + relative_offset,
+                        to_read);
+        } else {
+            UNREACHABLE();
+        }
+
+        std::memset(buffer + read_size + to_read, 0, alignment);
+
+        read_size += to_read + alignment;
+        offset += to_read + alignment;
+        current++;
+    }
+
+    return read_size;
+}
+
+bool LayeredFS::ExtractDirectory(Directory& current, const std::string& target_path) {
+    if (!FileUtil::CreateFullPath(target_path + current.path)) {
+        LOG_ERROR(Service_FS, "Could not create path {}", target_path + current.path);
+        return false;
+    }
+
+    constexpr std::size_t BufferSize = 0x10000;
+    std::array<u8, BufferSize> buffer;
+    for (const auto& file : current.files) {
+        // Extract file
+        const auto path = target_path + file->path;
+        LOG_INFO(Service_FS, "Extracting {} to {}", file->path, path);
+
+        FileUtil::IOFile target_file(path, "wb");
+        if (!target_file) {
+            LOG_ERROR(Service_FS, "Could not open file {}", path);
+            return false;
+        }
+
+        std::size_t written = 0;
+        while (written < file->relocation.size) {
+            const auto to_read =
+                std::min<std::size_t>(buffer.size(), file->relocation.size - written);
+            if (romfs->ReadFile(file->relocation.original_offset + written, to_read,
+                                buffer.data()) != to_read) {
+                LOG_ERROR(Service_FS, "Could not read from RomFS");
+                return false;
+            }
+
+            if (target_file.WriteBytes(buffer.data(), to_read) != to_read) {
+                LOG_ERROR(Service_FS, "Could not write to file {}", path);
+                return false;
+            }
+
+            written += to_read;
+        }
+    }
+
+    for (const auto& directory : current.directories) {
+        if (!ExtractDirectory(*directory, target_path)) {
+            return false;
+        }
+    }
+
+    return true;
+}
+
+bool LayeredFS::DumpRomFS(const std::string& target_path) {
+    std::string path = target_path;
+    if (path.back() == '/' || path.back() == '\\') {
+        path.erase(path.size() - 1, 1);
+    }
+
+    return ExtractDirectory(root, path);
+}
+
+} // namespace FileSys

src/core/file_sys/layered_fs.h

@@ -0,0 +1,123 @@
+// Copyright 2020 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <map>
+#include <memory>
+#include <string>
+#include <unordered_map>
+#include <vector>
+#include "common/common_types.h"
+#include "common/swap.h"
+#include "core/file_sys/romfs_reader.h"
+
+namespace FileSys {
+
+struct RomFSHeader {
+    struct Descriptor {
+        u32_le offset;
+        u32_le length;
+    };
+    u32_le header_length;
+    Descriptor directory_hash_table;
+    Descriptor directory_metadata_table;
+    Descriptor file_hash_table;
+    Descriptor file_metadata_table;
+    u32_le file_data_offset;
+};
+static_assert(sizeof(RomFSHeader) == 0x28, "Size of RomFSHeader is not correct");
+
+/**
+ * LayeredFS implementation. This basically adds a layer to another RomFSReader.
+ *
+ * patch_path: Path for RomFS replacements. Files present in this path replace or create
+ * corresponding files in RomFS.
+ * patch_ext_path: Path for RomFS extensions. Files present in this path:
+ *  - When with an extension of ".stub", remove the corresponding file in the RomFS.
+ *  - When with an extension of ".ips" or ".bps", patch the file in the RomFS.
+ */
+class LayeredFS : public RomFSReader {
+public:
+    explicit LayeredFS(std::shared_ptr<RomFSReader> romfs, std::string patch_path,
+                       std::string patch_ext_path, bool load_relocations = true);
+    ~LayeredFS() override;
+
+    std::size_t GetSize() const override;
+    std::size_t ReadFile(std::size_t offset, std::size_t length, u8* buffer) override;
+
+    bool DumpRomFS(const std::string& target_path);
+
+private:
+    struct File;
+    struct Directory {
+        std::string name;
+        std::string path; // with trailing '/'
+        std::vector<std::unique_ptr<File>> files;
+        std::vector<std::unique_ptr<Directory>> directories;
+        Directory* parent;
+    };
+
+    std::string ReadName(u32 offset, u32 name_length);
+
+    // Loads the current directory, then its siblings, and then its children.
+    void LoadDirectory(Directory& current, u32 offset);
+
+    // Load the file at offset, and then its siblings.
+    void LoadFile(Directory& parent, u32 offset);
+
+    // Load replace/create relocations
+    void LoadRelocations();
+
+    // Load patch/remove relocations
+    void LoadExtRelocations();
+
+    // Calculate the offset of a single directory add it to the map and list of directories
+    void PrepareBuildDirectory(Directory& current);
+
+    // Calculate the offset of a single file add it to the map and list of files
+    void PrepareBuildFile(File& current);
+
+    // Recursively generate a sequence of files and directories and their offsets for all
+    // children of current. (The current directory itself is not handled.)
+    void PrepareBuild(Directory& current);
+
+    void BuildDirectories();
+    void BuildFiles();
+
+    // Recursively extract a directory and all its contents to target_path
+    // target_path should be without trailing '/'.
+    bool ExtractDirectory(Directory& current, const std::string& target_path);
+
+    void RebuildMetadata();
+
+    std::shared_ptr<RomFSReader> romfs;
+    std::string patch_path;
+    std::string patch_ext_path;
+
+    RomFSHeader header;
+    Directory root;
+    std::unordered_map<std::string, File*> file_path_map;
+    std::unordered_map<std::string, Directory*> directory_path_map;
+    std::map<u64, File*> data_offset_map; // assigned data offset -> file
+    std::vector<u8> metadata;             // Includes header, hash table and metadata
+
+    // Used for rebuilding header
+    std::vector<u32_le> directory_hash_table;
+    std::vector<u32_le> file_hash_table;
+
+    std::unordered_map<Directory*, u32>
+        directory_metadata_offset_map;        // directory -> metadata offset
+    std::vector<Directory*> directory_list;   // sequence of directories to be written to metadata
+    u64 current_directory_offset{};           // current directory metadata offset
+    std::vector<u8> directory_metadata_table; // rebuilt directory metadata table
+
+    std::unordered_map<File*, u32> file_metadata_offset_map; // file -> metadata offset
+    std::vector<File*> file_list;        // sequence of files to be written to metadata
+    u64 current_file_offset{};           // current file metadata offset
+    std::vector<u8> file_metadata_table; // rebuilt file metadata table
+    u64 current_data_offset{};           // current assigned data offset
+};
+
+} // namespace FileSys

src/core/file_sys/ncch_container.cpp

@@ -11,6 +11,7 @@
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "core/core.h"
+#include "core/file_sys/layered_fs.h"
 #include "core/file_sys/ncch_container.h"
 #include "core/file_sys/patch.h"
 #include "core/file_sys/seed_db.h"
@@ -25,6 +26,14 @@ namespace FileSys {
 static const int kMaxSections = 8;   ///< Maximum number of sections (files) in an ExeFs
 static const int kBlockSize = 0x200; ///< Size of ExeFS blocks (in bytes)
 
+u64 GetModId(u64 program_id) {
+    constexpr u64 UPDATE_MASK = 0x0000000e'00000000;
+    if ((program_id & 0x000000ff'00000000) == UPDATE_MASK) { // Apply the mods to updates
+        return program_id & ~UPDATE_MASK;
+    }
+    return program_id;
+}
+
 /**
  * Get the decompressed size of an LZSS compressed ExeFS file
  * @param buffer Buffer of compressed file
@@ -303,8 +312,22 @@ Loader::ResultStatus NCCHContainer::Load() {
                 }
             }
 
-            FileUtil::IOFile exheader_override_file{filepath + ".exheader", "rb"};
-            const bool has_exheader_override = read_exheader(exheader_override_file);
+            const auto mods_path =
+                fmt::format("{}mods/{:016X}/", FileUtil::GetUserPath(FileUtil::UserPath::LoadDir),
+                            GetModId(ncch_header.program_id));
+            const std::array<std::string, 2> exheader_override_paths{{
+                mods_path + "exheader.bin",
+                filepath + ".exheader",
+            }};
+
+            bool has_exheader_override = false;
+            for (const auto& path : exheader_override_paths) {
+                FileUtil::IOFile exheader_override_file{path, "rb"};
+                if (read_exheader(exheader_override_file)) {
+                    has_exheader_override = true;
+                    break;
+                }
+            }
             if (has_exheader_override) {
                 if (exheader_header.system_info.jump_id !=
                     exheader_header.arm11_system_local_caps.program_id) {
@@ -512,7 +535,15 @@ Loader::ResultStatus NCCHContainer::ApplyCodePatch(std::vector<u8>& code) const
         std::string path;
         bool (*patch_fn)(const std::vector<u8>& patch, std::vector<u8>& code);
     };
-    const std::array<PatchLocation, 2> patch_paths{{
+
+    const auto mods_path =
+        fmt::format("{}mods/{:016X}/", FileUtil::GetUserPath(FileUtil::UserPath::LoadDir),
+                    GetModId(ncch_header.program_id));
+    const std::array<PatchLocation, 6> patch_paths{{
+        {mods_path + "exefs/code.ips", Patch::ApplyIpsPatch},
+        {mods_path + "exefs/code.bps", Patch::ApplyBpsPatch},
+        {mods_path + "code.ips", Patch::ApplyIpsPatch},
+        {mods_path + "code.bps", Patch::ApplyBpsPatch},
         {filepath + ".exefsdir/code.ips", Patch::ApplyIpsPatch},
         {filepath + ".exefsdir/code.bps", Patch::ApplyBpsPatch},
     }};
@@ -551,23 +582,34 @@ Loader::ResultStatus NCCHContainer::LoadOverrideExeFSSection(const char* name,
     else
         return Loader::ResultStatus::Error;
 
-    std::string section_override = filepath + ".exefsdir/" + override_name;
-    FileUtil::IOFile section_file(section_override, "rb");
+    const auto mods_path =
+        fmt::format("{}mods/{:016X}/", FileUtil::GetUserPath(FileUtil::UserPath::LoadDir),
+                    GetModId(ncch_header.program_id));
+    const std::array<std::string, 3> override_paths{{
+        mods_path + "exefs/" + override_name,
+        mods_path + override_name,
+        filepath + ".exefsdir/" + override_name,
+    }};
 
-    if (section_file.IsOpen()) {
-        auto section_size = section_file.GetSize();
-        buffer.resize(section_size);
+    for (const auto& path : override_paths) {
+        FileUtil::IOFile section_file(path, "rb");
 
-        section_file.Seek(0, SEEK_SET);
-        if (section_file.ReadBytes(&buffer[0], section_size) == section_size) {
-            LOG_WARNING(Service_FS, "File {} overriding built-in ExeFS file", section_override);
-            return Loader::ResultStatus::Success;
+        if (section_file.IsOpen()) {
+            auto section_size = section_file.GetSize();
+            buffer.resize(section_size);
+
+            section_file.Seek(0, SEEK_SET);
+            if (section_file.ReadBytes(&buffer[0], section_size) == section_size) {
+                LOG_WARNING(Service_FS, "File {} overriding built-in ExeFS file", path);
+                return Loader::ResultStatus::Success;
+            }
         }
     }
     return Loader::ResultStatus::ErrorNotUsed;
 }
 
-Loader::ResultStatus NCCHContainer::ReadRomFS(std::shared_ptr<RomFSReader>& romfs_file) {
+Loader::ResultStatus NCCHContainer::ReadRomFS(std::shared_ptr<RomFSReader>& romfs_file,
+                                              bool use_layered_fs) {
     Loader::ResultStatus result = Load();
     if (result != Loader::ResultStatus::Success)
         return result;
@@ -597,14 +639,43 @@ Loader::ResultStatus NCCHContainer::ReadRomFS(std::shared_ptr<RomFSReader>& romf
     if (!romfs_file_inner.IsOpen())
         return Loader::ResultStatus::Error;
 
+    std::shared_ptr<RomFSReader> direct_romfs;
     if (is_encrypted) {
-        romfs_file = std::make_shared<RomFSReader>(std::move(romfs_file_inner), romfs_offset,
-                                                   romfs_size, secondary_key, romfs_ctr, 0x1000);
+        direct_romfs =
+            std::make_shared<DirectRomFSReader>(std::move(romfs_file_inner), romfs_offset,
+                                                romfs_size, secondary_key, romfs_ctr, 0x1000);
     } else {
-        romfs_file =
-            std::make_shared<RomFSReader>(std::move(romfs_file_inner), romfs_offset, romfs_size);
+        direct_romfs = std::make_shared<DirectRomFSReader>(std::move(romfs_file_inner),
+                                                           romfs_offset, romfs_size);
     }
 
+    const auto path =
+        fmt::format("{}mods/{:016X}/", FileUtil::GetUserPath(FileUtil::UserPath::LoadDir),
+                    GetModId(ncch_header.program_id));
+    if (use_layered_fs &&
+        (FileUtil::Exists(path + "romfs/") || FileUtil::Exists(path + "romfs_ext/"))) {
+
+        romfs_file = std::make_shared<LayeredFS>(std::move(direct_romfs), path + "romfs/",
+                                                 path + "romfs_ext/");
+    } else {
+        romfs_file = std::move(direct_romfs);
+    }
+
+    return Loader::ResultStatus::Success;
+}
+
+Loader::ResultStatus NCCHContainer::DumpRomFS(const std::string& target_path) {
+    std::shared_ptr<RomFSReader> direct_romfs;
+    Loader::ResultStatus result = ReadRomFS(direct_romfs, false);
+    if (result != Loader::ResultStatus::Success)
+        return result;
+
+    std::shared_ptr<LayeredFS> layered_fs =
+        std::make_shared<LayeredFS>(std::move(direct_romfs), "", "", false);
+
+    if (!layered_fs->DumpRomFS(target_path)) {
+        return Loader::ResultStatus::Error;
+    }
     return Loader::ResultStatus::Success;
 }
 
@@ -614,9 +685,10 @@ Loader::ResultStatus NCCHContainer::ReadOverrideRomFS(std::shared_ptr<RomFSReade
     if (FileUtil::Exists(split_filepath)) {
         FileUtil::IOFile romfs_file_inner(split_filepath, "rb");
         if (romfs_file_inner.IsOpen()) {
-            LOG_WARNING(Service_FS, "File {} overriding built-in RomFS", split_filepath);
-            romfs_file = std::make_shared<RomFSReader>(std::move(romfs_file_inner), 0,
-                                                       romfs_file_inner.GetSize());
+            LOG_WARNING(Service_FS, "File {} overriding built-in RomFS; LayeredFS not enabled",
+                        split_filepath);
+            romfs_file = std::make_shared<DirectRomFSReader>(std::move(romfs_file_inner), 0,
+                                                             romfs_file_inner.GetSize());
             return Loader::ResultStatus::Success;
         }
     }

src/core/file_sys/ncch_container.h

@@ -149,7 +149,8 @@ struct ExHeader_StorageInfo {
 struct ExHeader_ARM11_SystemLocalCaps {
     u64_le program_id;
     u32_le core_version;
-    u8 reserved_flags[2];
+    u8 reserved_flag;
+    u8 n3ds_mode;
     union {
         u8 flags0;
         BitField<0, 2, u8> ideal_processor;
@@ -247,7 +248,15 @@ public:
      * @param size The size of the romfs
      * @return ResultStatus result of function
      */
-    Loader::ResultStatus ReadRomFS(std::shared_ptr<RomFSReader>& romfs_file);
+    Loader::ResultStatus ReadRomFS(std::shared_ptr<RomFSReader>& romfs_file,
+                                   bool use_layered_fs = true);
+
+    /**
+     * Dump the RomFS of the NCCH container to the user folder.
+     * @param target_path target path to dump to
+     * @return ResultStatus result of function.
+     */
+    Loader::ResultStatus DumpRomFS(const std::string& target_path);
 
     /**
      * Get the override RomFS of the NCCH container

src/core/file_sys/romfs_reader.cpp

@@ -5,7 +5,7 @@
 
 namespace FileSys {
 
-std::size_t RomFSReader::ReadFile(std::size_t offset, std::size_t length, u8* buffer) {
+std::size_t DirectRomFSReader::ReadFile(std::size_t offset, std::size_t length, u8* buffer) {
     if (length == 0)
         return 0; // Crypto++ does not like zero size buffer
     file.Seek(file_offset + offset, SEEK_SET);

src/core/file_sys/romfs_reader.h

@@ -7,23 +7,39 @@
 
 namespace FileSys {
 
+/**
+ * Interface for reading RomFS data.
+ */
 class RomFSReader {
 public:
-    RomFSReader(FileUtil::IOFile&& file, std::size_t file_offset, std::size_t data_size)
+    virtual ~RomFSReader() = default;
+
+    virtual std::size_t GetSize() const = 0;
+    virtual std::size_t ReadFile(std::size_t offset, std::size_t length, u8* buffer) = 0;
+};
+
+/**
+ * A RomFS reader that directly reads the RomFS file.
+ */
+class DirectRomFSReader : public RomFSReader {
+public:
+    DirectRomFSReader(FileUtil::IOFile&& file, std::size_t file_offset, std::size_t data_size)
         : is_encrypted(false), file(std::move(file)), file_offset(file_offset),
           data_size(data_size) {}
 
-    RomFSReader(FileUtil::IOFile&& file, std::size_t file_offset, std::size_t data_size,
-                const std::array<u8, 16>& key, const std::array<u8, 16>& ctr,
-                std::size_t crypto_offset)
+    DirectRomFSReader(FileUtil::IOFile&& file, std::size_t file_offset, std::size_t data_size,
+                      const std::array<u8, 16>& key, const std::array<u8, 16>& ctr,
+                      std::size_t crypto_offset)
         : is_encrypted(true), file(std::move(file)), key(key), ctr(ctr), file_offset(file_offset),
           crypto_offset(crypto_offset), data_size(data_size) {}
 
-    std::size_t GetSize() const {
+    ~DirectRomFSReader() override = default;
+
+    std::size_t GetSize() const override {
         return data_size;
     }
 
-    std::size_t ReadFile(std::size_t offset, std::size_t length, u8* buffer);
+    std::size_t ReadFile(std::size_t offset, std::size_t length, u8* buffer) override;
 
 private:
     bool is_encrypted;
@@ -34,7 +50,7 @@ private:
     u64 crypto_offset;
     u64 data_size;
 
-    RomFSReader() = default;
+    DirectRomFSReader() = default;
 
     template <class Archive>
     void serialize(Archive& ar, const unsigned int) {

src/core/gdbstub/gdbstub.cpp

@@ -121,6 +121,7 @@ constexpr char target_xml[] =
 )";
 
 int gdbserver_socket = -1;
+bool defer_start = false;
 
 u8 command_buffer[GDB_BUFFER_SIZE];
 u32 command_length;
@@ -160,10 +161,14 @@ BreakpointMap breakpoints_write;
 } // Anonymous namespace
 
 static Kernel::Thread* FindThreadById(int id) {
-    const auto& threads = Core::System::GetInstance().Kernel().GetThreadManager().GetThreadList();
-    for (auto& thread : threads) {
-        if (thread->GetThreadId() == static_cast<u32>(id)) {
-            return thread.get();
+    u32 num_cores = Core::GetNumCores();
+    for (u32 i = 0; i < num_cores; ++i) {
+        const auto& threads =
+            Core::System::GetInstance().Kernel().GetThreadManager(i).GetThreadList();
+        for (auto& thread : threads) {
+            if (thread->GetThreadId() == static_cast<u32>(id)) {
+                return thread.get();
+            }
         }
     }
     return nullptr;
@@ -414,7 +419,10 @@ static void RemoveBreakpoint(BreakpointType type, VAddr addr) {
         Core::System::GetInstance().Memory().WriteBlock(
             *Core::System::GetInstance().Kernel().GetCurrentProcess(), bp->second.addr,
             bp->second.inst.data(), bp->second.inst.size());
-        Core::CPU().ClearInstructionCache();
+        u32 num_cores = Core::GetNumCores();
+        for (u32 i = 0; i < num_cores; ++i) {
+            Core::GetCore(i).ClearInstructionCache();
+        }
     }
     p.erase(addr);
 }
@@ -540,10 +548,13 @@ static void HandleQuery() {
         SendReply(target_xml);
     } else if (strncmp(query, "fThreadInfo", strlen("fThreadInfo")) == 0) {
         std::string val = "m";
-        const auto& threads =
-            Core::System::GetInstance().Kernel().GetThreadManager().GetThreadList();
-        for (const auto& thread : threads) {
-            val += fmt::format("{:x},", thread->GetThreadId());
+        u32 num_cores = Core::GetNumCores();
+        for (u32 i = 0; i < num_cores; ++i) {
+            const auto& threads =
+                Core::System::GetInstance().Kernel().GetThreadManager(i).GetThreadList();
+            for (const auto& thread : threads) {
+                val += fmt::format("{:x},", thread->GetThreadId());
+            }
         }
         val.pop_back();
         SendReply(val.c_str());
@@ -553,11 +564,14 @@ static void HandleQuery() {
         std::string buffer;
         buffer += "l<?xml version=\"1.0\"?>";
         buffer += "<threads>";
-        const auto& threads =
-            Core::System::GetInstance().Kernel().GetThreadManager().GetThreadList();
-        for (const auto& thread : threads) {
-            buffer += fmt::format(R"*(<thread id="{:x}" name="Thread {:x}"></thread>)*",
-                                  thread->GetThreadId(), thread->GetThreadId());
+        u32 num_cores = Core::GetNumCores();
+        for (u32 i = 0; i < num_cores; ++i) {
+            const auto& threads =
+                Core::System::GetInstance().Kernel().GetThreadManager(i).GetThreadList();
+            for (const auto& thread : threads) {
+                buffer += fmt::format(R"*(<thread id="{:x}" name="Thread {:x}"></thread>)*",
+                                      thread->GetThreadId(), thread->GetThreadId());
+            }
         }
         buffer += "</threads>";
         SendReply(buffer.c_str());
@@ -619,9 +633,9 @@ static void SendSignal(Kernel::Thread* thread, u32 signal, bool full = true) {
     if (full) {
 
         buffer = fmt::format("T{:02x}{:02x}:{:08x};{:02x}:{:08x};{:02x}:{:08x}", latest_signal,
-                             PC_REGISTER, htonl(Core::CPU().GetPC()), SP_REGISTER,
-                             htonl(Core::CPU().GetReg(SP_REGISTER)), LR_REGISTER,
-                             htonl(Core::CPU().GetReg(LR_REGISTER)));
+                             PC_REGISTER, htonl(Core::GetRunningCore().GetPC()), SP_REGISTER,
+                             htonl(Core::GetRunningCore().GetReg(SP_REGISTER)), LR_REGISTER,
+                             htonl(Core::GetRunningCore().GetReg(LR_REGISTER)));
     } else {
         buffer = fmt::format("T{:02x}", latest_signal);
     }
@@ -782,7 +796,7 @@ static void WriteRegister() {
         return SendReply("E01");
     }
 
-    Core::CPU().LoadContext(current_thread->context);
+    Core::GetRunningCore().LoadContext(current_thread->context);
 
     SendReply("OK");
 }
@@ -812,7 +826,7 @@ static void WriteRegisters() {
         }
     }
 
-    Core::CPU().LoadContext(current_thread->context);
+    Core::GetRunningCore().LoadContext(current_thread->context);
 
     SendReply("OK");
 }
@@ -869,7 +883,7 @@ static void WriteMemory() {
     GdbHexToMem(data.data(), len_pos + 1, len);
     Core::System::GetInstance().Memory().WriteBlock(
         *Core::System::GetInstance().Kernel().GetCurrentProcess(), addr, data.data(), len);
-    Core::CPU().ClearInstructionCache();
+    Core::GetRunningCore().ClearInstructionCache();
     SendReply("OK");
 }
 
@@ -883,12 +897,12 @@ void Break(bool is_memory_break) {
 static void Step() {
     if (command_length > 1) {
         RegWrite(PC_REGISTER, GdbHexToInt(command_buffer + 1), current_thread);
-        Core::CPU().LoadContext(current_thread->context);
+        Core::GetRunningCore().LoadContext(current_thread->context);
     }
     step_loop = true;
     halt_loop = true;
     send_trap = true;
-    Core::CPU().ClearInstructionCache();
+    Core::GetRunningCore().ClearInstructionCache();
 }
 
 bool IsMemoryBreak() {
@@ -904,7 +918,7 @@ static void Continue() {
     memory_break = false;
     step_loop = false;
     halt_loop = false;
-    Core::CPU().ClearInstructionCache();
+    Core::GetRunningCore().ClearInstructionCache();
 }
 
 /**
@@ -930,7 +944,7 @@ static bool CommitBreakpoint(BreakpointType type, VAddr addr, u32 len) {
         Core::System::GetInstance().Memory().WriteBlock(
             *Core::System::GetInstance().Kernel().GetCurrentProcess(), addr, btrap.data(),
             btrap.size());
-        Core::CPU().ClearInstructionCache();
+        Core::GetRunningCore().ClearInstructionCache();
     }
     p.insert({addr, breakpoint});
 
@@ -1030,6 +1044,9 @@ static void RemoveBreakpoint() {
 
 void HandlePacket() {
     if (!IsConnected()) {
+        if (defer_start) {
+            ToggleServer(true);
+        }
         return;
     }
 
@@ -1120,6 +1137,10 @@ void ToggleServer(bool status) {
     }
 }
 
+void DeferStart() {
+    defer_start = true;
+}
+
 static void Init(u16 port) {
     if (!server_enabled) {
         // Set the halt loop to false in case the user enabled the gdbstub mid-execution.
@@ -1203,6 +1224,7 @@ void Shutdown() {
     if (!server_enabled) {
         return;
     }
+    defer_start = false;
 
     LOG_INFO(Debug_GDBStub, "Stopping GDB ...");
     if (gdbserver_socket != -1) {

src/core/gdbstub/gdbstub.h

@@ -42,6 +42,13 @@ void ToggleServer(bool status);
 /// Start the gdbstub server.
 void Init();
 
+/**
+ * Defer initialization of the gdbstub to the first packet processing functions.
+ * This avoids a case where the gdbstub thread is frozen after initialization
+ * and fails to respond in time to packets.
+ */
+void DeferStart();
+
 /// Stop gdbstub server.
 void Shutdown();
 

src/core/hle/kernel/handle_table.cpp

@@ -83,7 +83,7 @@ bool HandleTable::IsValid(Handle handle) const {
 
 std::shared_ptr<Object> HandleTable::GetGeneric(Handle handle) const {
     if (handle == CurrentThread) {
-        return SharedFrom(kernel.GetThreadManager().GetCurrentThread());
+        return SharedFrom(kernel.GetCurrentThreadManager().GetCurrentThread());
     } else if (handle == CurrentProcess) {
         return kernel.GetCurrentProcess();
     }

src/core/hle/kernel/kernel.cpp

@@ -20,22 +20,30 @@ namespace Kernel {
 
 /// Initialize the kernel
 KernelSystem::KernelSystem(Memory::MemorySystem& memory, Core::Timing& timing,
-                           std::function<void()> prepare_reschedule_callback, u32 system_mode)
+                           std::function<void()> prepare_reschedule_callback, u32 system_mode,
+                           u32 num_cores, u8 n3ds_mode)
     : memory(memory), timing(timing),
       prepare_reschedule_callback(std::move(prepare_reschedule_callback)) {
     for (auto i = 0; i < memory_regions.size(); i++) {
         memory_regions[i] = std::make_shared<MemoryRegionInfo>();
     }
-    MemoryInit(system_mode);
+    MemoryInit(system_mode, n3ds_mode);
 
     resource_limits = std::make_unique<ResourceLimitList>(*this);
-    thread_manager = std::make_unique<ThreadManager>(*this);
+    for (u32 core_id = 0; core_id < num_cores; ++core_id) {
+        thread_managers.push_back(std::make_unique<ThreadManager>(*this, core_id));
+    }
     timer_manager = std::make_unique<TimerManager>(timing);
     ipc_recorder = std::make_unique<IPCDebugger::Recorder>();
+    stored_processes.assign(num_cores, nullptr);
+
+    next_thread_id = 1;
 }
 
 /// Shutdown the kernel
-KernelSystem::~KernelSystem() = default;
+KernelSystem::~KernelSystem() {
+    ResetThreadIDs();
+};
 
 ResourceLimitList& KernelSystem::ResourceLimit() {
     return *resource_limits;
@@ -58,6 +66,15 @@ void KernelSystem::SetCurrentProcess(std::shared_ptr<Process> process) {
     SetCurrentMemoryPageTable(process->vm_manager.page_table);
 }
 
+void KernelSystem::SetCurrentProcessForCPU(std::shared_ptr<Process> process, u32 core_id) {
+    if (current_cpu->GetID() == core_id) {
+        current_process = process;
+        SetCurrentMemoryPageTable(process->vm_manager.page_table);
+    } else {
+        stored_processes[core_id] = process;
+    }
+}
+
 void KernelSystem::SetCurrentMemoryPageTable(std::shared_ptr<Memory::PageTable> page_table) {
     memory.SetCurrentPageTable(page_table);
     if (current_cpu != nullptr) {
@@ -65,17 +82,39 @@ void KernelSystem::SetCurrentMemoryPageTable(std::shared_ptr<Memory::PageTable>
     }
 }
 
-void KernelSystem::SetCPU(std::shared_ptr<ARM_Interface> cpu) {
+void KernelSystem::SetCPUs(std::vector<std::shared_ptr<ARM_Interface>> cpus) {
+    ASSERT(cpus.size() == thread_managers.size());
+    u32 i = 0;
+    for (const auto& cpu : cpus) {
+        thread_managers[i++]->SetCPU(*cpu);
+    }
+}
+
+void KernelSystem::SetRunningCPU(std::shared_ptr<ARM_Interface> cpu) {
+    if (current_process) {
+        stored_processes[current_cpu->GetID()] = current_process;
+    }
     current_cpu = cpu;
-    thread_manager->SetCPU(*cpu);
+    timing.SetCurrentTimer(cpu->GetID());
+    if (stored_processes[current_cpu->GetID()]) {
+        SetCurrentProcess(stored_processes[current_cpu->GetID()]);
+    }
 }
 
-ThreadManager& KernelSystem::GetThreadManager() {
-    return *thread_manager;
+ThreadManager& KernelSystem::GetThreadManager(u32 core_id) {
+    return *thread_managers[core_id];
 }
 
-const ThreadManager& KernelSystem::GetThreadManager() const {
-    return *thread_manager;
+const ThreadManager& KernelSystem::GetThreadManager(u32 core_id) const {
+    return *thread_managers[core_id];
+}
+
+ThreadManager& KernelSystem::GetCurrentThreadManager() {
+    return *thread_managers[current_cpu->GetID()];
+}
+
+const ThreadManager& KernelSystem::GetCurrentThreadManager() const {
+    return *thread_managers[current_cpu->GetID()];
 }
 
 TimerManager& KernelSystem::GetTimerManager() {
@@ -106,6 +145,14 @@ void KernelSystem::AddNamedPort(std::string name, std::shared_ptr<ClientPort> po
     named_ports.emplace(std::move(name), std::move(port));
 }
 
+u32 KernelSystem::NewThreadId() {
+    return next_thread_id++;
+}
+
+void KernelSystem::ResetThreadIDs() {
+    next_thread_id = 0;
+}
+
 template <class Archive>
 void KernelSystem::serialize(Archive& ar, const unsigned int file_version) {
     ar& memory_regions;
@@ -118,9 +165,14 @@ void KernelSystem::serialize(Archive& ar, const unsigned int file_version) {
     ar& next_process_id;
     ar& process_list;
     ar& current_process;
-    ar&* thread_manager.get();
+    // NB: core count checked in 'core'
+    for (auto& thread_manager : thread_managers) {
+        ar&* thread_manager.get();
+    }
     ar& config_mem_handler;
     ar& shared_page_handler;
+    ar& stored_processes;
+    ar& next_thread_id;
     // Deliberately don't include debugger info to allow debugging through loads
 }
 

src/core/hle/kernel/kernel.h

@@ -88,7 +88,8 @@ enum class MemoryRegion : u16 {
 class KernelSystem {
 public:
     explicit KernelSystem(Memory::MemorySystem& memory, Core::Timing& timing,
-                          std::function<void()> prepare_reschedule_callback, u32 system_mode);
+                          std::function<void()> prepare_reschedule_callback, u32 system_mode,
+                          u32 num_cores, u8 n3ds_mode);
     ~KernelSystem();
 
     using PortPair = std::pair<std::shared_ptr<ServerPort>, std::shared_ptr<ClientPort>>;
@@ -214,13 +215,19 @@ public:
 
     std::shared_ptr<Process> GetCurrentProcess() const;
     void SetCurrentProcess(std::shared_ptr<Process> process);
+    void SetCurrentProcessForCPU(std::shared_ptr<Process> process, u32 core_id);
 
     void SetCurrentMemoryPageTable(std::shared_ptr<Memory::PageTable> page_table);
 
-    void SetCPU(std::shared_ptr<ARM_Interface> cpu);
+    void SetCPUs(std::vector<std::shared_ptr<ARM_Interface>> cpu);
 
-    ThreadManager& GetThreadManager();
-    const ThreadManager& GetThreadManager() const;
+    void SetRunningCPU(std::shared_ptr<ARM_Interface> cpu);
+
+    ThreadManager& GetThreadManager(u32 core_id);
+    const ThreadManager& GetThreadManager(u32 core_id) const;
+
+    ThreadManager& GetCurrentThreadManager();
+    const ThreadManager& GetCurrentThreadManager() const;
 
     TimerManager& GetTimerManager();
     const TimerManager& GetTimerManager() const;
@@ -246,6 +253,10 @@ public:
         prepare_reschedule_callback();
     }
 
+    u32 NewThreadId();
+
+    void ResetThreadIDs();
+
     /// Map of named ports managed by the kernel, which can be retrieved using the ConnectToPort
     std::unordered_map<std::string, std::shared_ptr<ClientPort>> named_ports;
 
@@ -256,7 +267,7 @@ public:
     Core::Timing& timing;
 
 private:
-    void MemoryInit(u32 mem_type);
+    void MemoryInit(u32 mem_type, u8 n3ds_mode);
 
     std::function<void()> prepare_reschedule_callback;
 
@@ -280,14 +291,17 @@ private:
     std::vector<std::shared_ptr<Process>> process_list;
 
     std::shared_ptr<Process> current_process;
+    std::vector<std::shared_ptr<Process>> stored_processes;
 
-    std::unique_ptr<ThreadManager> thread_manager;
+    std::vector<std::unique_ptr<ThreadManager>> thread_managers;
 
     std::shared_ptr<ConfigMem::Handler> config_mem_handler;
     std::shared_ptr<SharedPage::Handler> shared_page_handler;
 
     std::unique_ptr<IPCDebugger::Recorder> ipc_recorder;
 
+    u32 next_thread_id;
+
     friend class boost::serialization::access;
     template <class Archive>
     void serialize(Archive& ar, const unsigned int file_version);

src/core/hle/kernel/memory.cpp

@@ -19,6 +19,7 @@
 #include "core/hle/kernel/vm_manager.h"
 #include "core/hle/result.h"
 #include "core/memory.h"
+#include "core/settings.h"
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 
@@ -40,11 +41,32 @@ static const u32 memory_region_sizes[8][3] = {
     {0x0B200000, 0x02E00000, 0x02000000}, // 7
 };
 
-void KernelSystem::MemoryInit(u32 mem_type) {
-    // TODO(yuriks): On the n3DS, all o3DS configurations (<=5) are forced to 6 instead.
-    ASSERT_MSG(mem_type <= 5, "New 3DS memory configuration aren't supported yet!");
+namespace MemoryMode {
+enum N3DSMode : u8 {
+    Mode6 = 1,
+    Mode7 = 2,
+    Mode6_2 = 3,
+};
+}
+
+void KernelSystem::MemoryInit(u32 mem_type, u8 n3ds_mode) {
     ASSERT(mem_type != 1);
 
+    const bool is_new_3ds = Settings::values.is_new_3ds;
+    u32 reported_mem_type = mem_type;
+    if (is_new_3ds) {
+        if (n3ds_mode == MemoryMode::Mode6 || n3ds_mode == MemoryMode::Mode6_2) {
+            mem_type = 6;
+            reported_mem_type = 6;
+        } else if (n3ds_mode == MemoryMode::Mode7) {
+            mem_type = 7;
+            reported_mem_type = 7;
+        } else {
+            // On the N3ds, all O3ds configurations (<=5) are forced to 6 instead.
+            mem_type = 6;
+        }
+    }
+
     // The kernel allocation regions (APPLICATION, SYSTEM and BASE) are laid out in sequence, with
     // the sizes specified in the memory_region_sizes table.
     VAddr base = 0;
@@ -55,14 +77,12 @@ void KernelSystem::MemoryInit(u32 mem_type) {
     }
 
     // We must've allocated the entire FCRAM by the end
-    ASSERT(base == Memory::FCRAM_SIZE);
+    ASSERT(base == (is_new_3ds ? Memory::FCRAM_N3DS_SIZE : Memory::FCRAM_SIZE));
 
     config_mem_handler = std::make_shared<ConfigMem::Handler>();
     auto& config_mem = config_mem_handler->GetConfigMem();
-    config_mem.app_mem_type = mem_type;
-    // app_mem_malloc does not always match the configured size for memory_region[0]: in case the
-    // n3DS type override is in effect it reports the size the game expects, not the real one.
-    config_mem.app_mem_alloc = memory_region_sizes[mem_type][0];
+    config_mem.app_mem_type = reported_mem_type;
+    config_mem.app_mem_alloc = memory_region_sizes[reported_mem_type][0];
     config_mem.sys_mem_alloc = memory_regions[1]->size;
     config_mem.base_mem_alloc = memory_regions[2]->size;
 

src/core/hle/kernel/mutex.cpp

@@ -39,7 +39,7 @@ std::shared_ptr<Mutex> KernelSystem::CreateMutex(bool initial_locked, std::strin
 
     // Acquire mutex with current thread if initialized as locked
     if (initial_locked)
-        mutex->Acquire(thread_manager->GetCurrentThread());
+        mutex->Acquire(thread_managers[current_cpu->GetID()]->GetCurrentThread());
 
     return mutex;
 }

src/core/hle/kernel/shared_page.cpp

@@ -70,7 +70,7 @@ Handler::Handler(Core::Timing& timing) : timing(timing) {
     using namespace std::placeholders;
     update_time_event = timing.RegisterEvent("SharedPage::UpdateTimeCallback",
                                              std::bind(&Handler::UpdateTimeCallback, this, _1, _2));
-    timing.ScheduleEvent(0, update_time_event);
+    timing.ScheduleEvent(0, update_time_event, 0, 0);
 
     float slidestate = Settings::values.factor_3d / 100.0f;
     shared_page.sliderstate_3d = static_cast<float_le>(slidestate);

src/core/hle/kernel/svc.cpp

@@ -280,12 +280,12 @@ void SVC::ExitProcess() {
     current_process->status = ProcessStatus::Exited;
 
     // Stop all the process threads that are currently waiting for objects.
-    auto& thread_list = kernel.GetThreadManager().GetThreadList();
+    auto& thread_list = kernel.GetCurrentThreadManager().GetThreadList();
     for (auto& thread : thread_list) {
         if (thread->owner_process != current_process)
             continue;
 
-        if (thread.get() == kernel.GetThreadManager().GetCurrentThread())
+        if (thread.get() == kernel.GetCurrentThreadManager().GetCurrentThread())
             continue;
 
         // TODO(Subv): When are the other running/ready threads terminated?
@@ -297,7 +297,7 @@ void SVC::ExitProcess() {
     }
 
     // Kill the current thread
-    kernel.GetThreadManager().GetCurrentThread()->Stop();
+    kernel.GetCurrentThreadManager().GetCurrentThread()->Stop();
 
     system.PrepareReschedule();
 }
@@ -388,7 +388,7 @@ ResultCode SVC::SendSyncRequest(Handle handle) {
 
     system.PrepareReschedule();
 
-    auto thread = SharedFrom(kernel.GetThreadManager().GetCurrentThread());
+    auto thread = SharedFrom(kernel.GetCurrentThreadManager().GetCurrentThread());
 
     if (kernel.GetIPCRecorder().IsEnabled()) {
         kernel.GetIPCRecorder().RegisterRequest(session, thread);
@@ -476,7 +476,7 @@ private:
 /// Wait for a handle to synchronize, timeout after the specified nanoseconds
 ResultCode SVC::WaitSynchronization1(Handle handle, s64 nano_seconds) {
     auto object = kernel.GetCurrentProcess()->handle_table.Get<WaitObject>(handle);
-    Thread* thread = kernel.GetThreadManager().GetCurrentThread();
+    Thread* thread = kernel.GetCurrentThreadManager().GetCurrentThread();
 
     if (object == nullptr)
         return ERR_INVALID_HANDLE;
@@ -514,7 +514,7 @@ ResultCode SVC::WaitSynchronization1(Handle handle, s64 nano_seconds) {
 /// Wait for the given handles to synchronize, timeout after the specified nanoseconds
 ResultCode SVC::WaitSynchronizationN(s32* out, VAddr handles_address, s32 handle_count,
                                      bool wait_all, s64 nano_seconds) {
-    Thread* thread = kernel.GetThreadManager().GetCurrentThread();
+    Thread* thread = kernel.GetCurrentThreadManager().GetCurrentThread();
 
     if (!Memory::IsValidVirtualAddress(*kernel.GetCurrentProcess(), handles_address))
         return ERR_INVALID_POINTER;
@@ -684,7 +684,7 @@ ResultCode SVC::ReplyAndReceive(s32* index, VAddr handles_address, s32 handle_co
 
     // We are also sending a command reply.
     // Do not send a reply if the command id in the command buffer is 0xFFFF.
-    Thread* thread = kernel.GetThreadManager().GetCurrentThread();
+    Thread* thread = kernel.GetCurrentThreadManager().GetCurrentThread();
     u32 cmd_buff_header = memory.Read32(thread->GetCommandBufferAddress());
     IPC::Header header{cmd_buff_header};
     if (reply_target != 0 && header.command_id != 0xFFFF) {
@@ -791,7 +791,7 @@ ResultCode SVC::ArbitrateAddress(Handle handle, u32 address, u32 type, u32 value
         return ERR_INVALID_HANDLE;
 
     auto res =
-        arbiter->ArbitrateAddress(SharedFrom(kernel.GetThreadManager().GetCurrentThread()),
+        arbiter->ArbitrateAddress(SharedFrom(kernel.GetCurrentThreadManager().GetCurrentThread()),
                                   static_cast<ArbitrationType>(type), address, value, nanoseconds);
 
     // TODO(Subv): Identify in which specific cases this call should cause a reschedule.
@@ -912,14 +912,19 @@ ResultCode SVC::CreateThread(Handle* out_handle, u32 entry_point, u32 arg, VAddr
         break;
     case ThreadProcessorIdAll:
         LOG_INFO(Kernel_SVC,
-                 "Newly created thread is allowed to be run in any Core, unimplemented.");
+                 "Newly created thread is allowed to be run in any Core, for now run in core 0.");
+        processor_id = ThreadProcessorId0;
         break;
     case ThreadProcessorId1:
-        LOG_ERROR(Kernel_SVC,
-                  "Newly created thread must run in the SysCore (Core1), unimplemented.");
+    case ThreadProcessorId2:
+    case ThreadProcessorId3:
+        // TODO: Check and log for: When processorid==0x2 and the process is not a BASE mem-region
+        // process, exheader kernel-flags bitmask 0x2000 must be set (otherwise error 0xD9001BEA is
+        // returned). When processorid==0x3 and the process is not a BASE mem-region process, error
+        // 0xD9001BEA is returned. These are the only restriction checks done by the kernel for
+        // processorid.
         break;
     default:
-        // TODO(bunnei): Implement support for other processor IDs
         ASSERT_MSG(false, "Unsupported thread processor ID: {}", processor_id);
         break;
     }
@@ -945,9 +950,9 @@ ResultCode SVC::CreateThread(Handle* out_handle, u32 entry_point, u32 arg, VAddr
 
 /// Called when a thread exits
 void SVC::ExitThread() {
-    LOG_TRACE(Kernel_SVC, "called, pc=0x{:08X}", system.CPU().GetPC());
+    LOG_TRACE(Kernel_SVC, "called, pc=0x{:08X}", system.GetRunningCore().GetPC());
 
-    kernel.GetThreadManager().ExitCurrentThread();
+    kernel.GetCurrentThreadManager().ExitCurrentThread();
     system.PrepareReschedule();
 }
 
@@ -993,7 +998,7 @@ ResultCode SVC::SetThreadPriority(Handle handle, u32 priority) {
 /// Create a mutex
 ResultCode SVC::CreateMutex(Handle* out_handle, u32 initial_locked) {
     std::shared_ptr<Mutex> mutex = kernel.CreateMutex(initial_locked != 0);
-    mutex->name = fmt::format("mutex-{:08x}", system.CPU().GetReg(14));
+    mutex->name = fmt::format("mutex-{:08x}", system.GetRunningCore().GetReg(14));
     CASCADE_RESULT(*out_handle, kernel.GetCurrentProcess()->handle_table.Create(std::move(mutex)));
 
     LOG_TRACE(Kernel_SVC, "called initial_locked={} : created handle=0x{:08X}",
@@ -1010,7 +1015,7 @@ ResultCode SVC::ReleaseMutex(Handle handle) {
     if (mutex == nullptr)
         return ERR_INVALID_HANDLE;
 
-    return mutex->Release(kernel.GetThreadManager().GetCurrentThread());
+    return mutex->Release(kernel.GetCurrentThreadManager().GetCurrentThread());
 }
 
 /// Get the ID of the specified process
@@ -1060,7 +1065,7 @@ ResultCode SVC::GetThreadId(u32* thread_id, Handle handle) {
 ResultCode SVC::CreateSemaphore(Handle* out_handle, s32 initial_count, s32 max_count) {
     CASCADE_RESULT(std::shared_ptr<Semaphore> semaphore,
                    kernel.CreateSemaphore(initial_count, max_count));
-    semaphore->name = fmt::format("semaphore-{:08x}", system.CPU().GetReg(14));
+    semaphore->name = fmt::format("semaphore-{:08x}", system.GetRunningCore().GetReg(14));
     CASCADE_RESULT(*out_handle,
                    kernel.GetCurrentProcess()->handle_table.Create(std::move(semaphore)));
 
@@ -1130,8 +1135,9 @@ ResultCode SVC::QueryMemory(MemoryInfo* memory_info, PageInfo* page_info, u32 ad
 
 /// Create an event
 ResultCode SVC::CreateEvent(Handle* out_handle, u32 reset_type) {
-    std::shared_ptr<Event> evt = kernel.CreateEvent(
-        static_cast<ResetType>(reset_type), fmt::format("event-{:08x}", system.CPU().GetReg(14)));
+    std::shared_ptr<Event> evt =
+        kernel.CreateEvent(static_cast<ResetType>(reset_type),
+                           fmt::format("event-{:08x}", system.GetRunningCore().GetReg(14)));
     CASCADE_RESULT(*out_handle, kernel.GetCurrentProcess()->handle_table.Create(std::move(evt)));
 
     LOG_TRACE(Kernel_SVC, "called reset_type=0x{:08X} : created handle=0x{:08X}", reset_type,
@@ -1173,8 +1179,9 @@ ResultCode SVC::ClearEvent(Handle handle) {
 
 /// Creates a timer
 ResultCode SVC::CreateTimer(Handle* out_handle, u32 reset_type) {
-    std::shared_ptr<Timer> timer = kernel.CreateTimer(
-        static_cast<ResetType>(reset_type), fmt ::format("timer-{:08x}", system.CPU().GetReg(14)));
+    std::shared_ptr<Timer> timer =
+        kernel.CreateTimer(static_cast<ResetType>(reset_type),
+                           fmt ::format("timer-{:08x}", system.GetRunningCore().GetReg(14)));
     CASCADE_RESULT(*out_handle, kernel.GetCurrentProcess()->handle_table.Create(std::move(timer)));
 
     LOG_TRACE(Kernel_SVC, "called reset_type=0x{:08X} : created handle=0x{:08X}", reset_type,
@@ -1228,7 +1235,7 @@ ResultCode SVC::CancelTimer(Handle handle) {
 void SVC::SleepThread(s64 nanoseconds) {
     LOG_TRACE(Kernel_SVC, "called nanoseconds={}", nanoseconds);
 
-    ThreadManager& thread_manager = kernel.GetThreadManager();
+    ThreadManager& thread_manager = kernel.GetCurrentThreadManager();
 
     // Don't attempt to yield execution if there are no available threads to run,
     // this way we avoid a useless reschedule to the idle thread.
@@ -1246,10 +1253,11 @@ void SVC::SleepThread(s64 nanoseconds) {
 
 /// This returns the total CPU ticks elapsed since the CPU was powered-on
 s64 SVC::GetSystemTick() {
-    s64 result = system.CoreTiming().GetTicks();
+    // TODO: Use globalTicks here?
+    s64 result = system.GetRunningCore().GetTimer()->GetTicks();
     // Advance time to defeat dumb games (like Cubic Ninja) that busy-wait for the frame to end.
     // Measured time between two calls on a 9.2 o3DS with Ninjhax 1.1b
-    system.CoreTiming().AddTicks(150);
+    system.GetRunningCore().GetTimer()->AddTicks(150);
     return result;
 }
 
@@ -1611,11 +1619,11 @@ void SVC::CallSVC(u32 immediate) {
 SVC::SVC(Core::System& system) : system(system), kernel(system.Kernel()), memory(system.Memory()) {}
 
 u32 SVC::GetReg(std::size_t n) {
-    return system.CPU().GetReg(static_cast<int>(n));
+    return system.GetRunningCore().GetReg(static_cast<int>(n));
 }
 
 void SVC::SetReg(std::size_t n, u32 value) {
-    system.CPU().SetReg(static_cast<int>(n), value);
+    system.GetRunningCore().SetReg(static_cast<int>(n), value);
 }
 
 SVCContext::SVCContext(Core::System& system) : impl(std::make_unique<SVC>(system)) {}

src/core/hle/kernel/thread.cpp

@@ -62,13 +62,10 @@ void Thread::Acquire(Thread* thread) {
     ASSERT_MSG(!ShouldWait(thread), "object unavailable!");
 }
 
-u32 ThreadManager::NewThreadId() {
-    return next_thread_id++;
-}
-
-Thread::Thread(KernelSystem& kernel)
-    : WaitObject(kernel), context(kernel.GetThreadManager().NewContext()),
-      thread_manager(kernel.GetThreadManager()) {}
+Thread::Thread(KernelSystem& kernel, u32 core_id)
+    : WaitObject(kernel), context(kernel.GetThreadManager(core_id).NewContext()),
+      core_id(core_id),
+      thread_manager(kernel.GetThreadManager(core_id)) {}
 Thread::~Thread() {}
 
 Thread* ThreadManager::GetCurrentThread() const {
@@ -113,7 +110,7 @@ void ThreadManager::SwitchContext(Thread* new_thread) {
 
     // Save context for previous thread
     if (previous_thread) {
-        previous_thread->last_running_ticks = timing.GetTicks();
+        previous_thread->last_running_ticks = timing.GetGlobalTicks();
         cpu->SaveContext(previous_thread->context);
 
         if (previous_thread->status == ThreadStatus::Running) {
@@ -140,7 +137,7 @@ void ThreadManager::SwitchContext(Thread* new_thread) {
         new_thread->status = ThreadStatus::Running;
 
         if (previous_process != current_thread->owner_process) {
-            kernel.SetCurrentProcess(current_thread->owner_process);
+            kernel.SetCurrentProcessForCPU(current_thread->owner_process, cpu->GetID());
         }
 
         cpu->LoadContext(new_thread->context);
@@ -153,7 +150,7 @@ void ThreadManager::SwitchContext(Thread* new_thread) {
 }
 
 Thread* ThreadManager::PopNextReadyThread() {
-    Thread* next;
+    Thread* next = nullptr;
     Thread* thread = GetCurrentThread();
 
     if (thread && thread->status == ThreadStatus::Running) {
@@ -337,22 +334,22 @@ ResultVal<std::shared_ptr<Thread>> KernelSystem::CreateThread(
                           ErrorSummary::InvalidArgument, ErrorLevel::Permanent);
     }
 
-    auto thread{std::make_shared<Thread>(*this)};
+    auto thread{std::make_shared<Thread>(*this, processor_id)};
 
-    thread_manager->thread_list.push_back(thread);
-    thread_manager->ready_queue.prepare(priority);
+    thread_managers[processor_id]->thread_list.push_back(thread);
+    thread_managers[processor_id]->ready_queue.prepare(priority);
 
-    thread->thread_id = thread_manager->NewThreadId();
+    thread->thread_id = NewThreadId();
     thread->status = ThreadStatus::Dormant;
     thread->entry_point = entry_point;
     thread->stack_top = stack_top;
     thread->nominal_priority = thread->current_priority = priority;
-    thread->last_running_ticks = timing.GetTicks();
+    thread->last_running_ticks = timing.GetGlobalTicks();
     thread->processor_id = processor_id;
     thread->wait_objects.clear();
     thread->wait_address = 0;
     thread->name = std::move(name);
-    thread_manager->wakeup_callback_table[thread->thread_id] = thread.get();
+    thread_managers[processor_id]->wakeup_callback_table[thread->thread_id] = thread.get();
     thread->owner_process = owner_process;
 
     // Find the next available TLS index, and mark it as used
@@ -397,7 +394,7 @@ ResultVal<std::shared_ptr<Thread>> KernelSystem::CreateThread(
     // to initialize the context
     ResetThreadContext(thread->context, stack_top, entry_point, arg);
 
-    thread_manager->ready_queue.push_back(thread->current_priority, thread.get());
+    thread_managers[processor_id]->ready_queue.push_back(thread->current_priority, thread.get());
     thread->status = ThreadStatus::Ready;
 
     return MakeResult<std::shared_ptr<Thread>>(std::move(thread));
@@ -463,6 +460,9 @@ void ThreadManager::Reschedule() {
         LOG_TRACE(Kernel, "context switch {} -> idle", cur->GetObjectId());
     } else if (next) {
         LOG_TRACE(Kernel, "context switch idle -> {}", next->GetObjectId());
+    } else {
+        LOG_TRACE(Kernel, "context switch idle -> idle, do nothing");
+        return;
     }
 
     SwitchContext(next);
@@ -489,11 +489,10 @@ VAddr Thread::GetCommandBufferAddress() const {
     return GetTLSAddress() + command_header_offset;
 }
 
-ThreadManager::ThreadManager(Kernel::KernelSystem& kernel) : kernel(kernel) {
-    ThreadWakeupEventType =
-        kernel.timing.RegisterEvent("ThreadWakeupCallback", [this](u64 thread_id, s64 cycle_late) {
-            ThreadWakeupCallback(thread_id, cycle_late);
-        });
+ThreadManager::ThreadManager(Kernel::KernelSystem& kernel, u32 core_id) : kernel(kernel) {
+    ThreadWakeupEventType = kernel.timing.RegisterEvent(
+        "ThreadWakeupCallback_" + std::to_string(core_id),
+        [this](u64 thread_id, s64 cycle_late) { ThreadWakeupCallback(thread_id, cycle_late); });
 }
 
 ThreadManager::~ThreadManager() {

src/core/hle/kernel/thread.h

@@ -38,7 +38,9 @@ enum ThreadProcessorId : s32 {
     ThreadProcessorIdAll = -1,     ///< Run thread on either core
     ThreadProcessorId0 = 0,        ///< Run thread on core 0 (AppCore)
     ThreadProcessorId1 = 1,        ///< Run thread on core 1 (SysCore)
-    ThreadProcessorIdMax = 2,      ///< Processor ID must be less than this
+    ThreadProcessorId2 = 2,        ///< Run thread on core 2 (additional n3ds core)
+    ThreadProcessorId3 = 3,        ///< Run thread on core 3 (additional n3ds core)
+    ThreadProcessorIdMax = 4,      ///< Processor ID must be less than this
 };
 
 enum class ThreadStatus {
@@ -75,15 +77,9 @@ private:
 
 class ThreadManager {
 public:
-    explicit ThreadManager(Kernel::KernelSystem& kernel);
+    explicit ThreadManager(Kernel::KernelSystem& kernel, u32 core_id);
     ~ThreadManager();
 
-    /**
-     * Creates a new thread ID
-     * @return The new thread ID
-     */
-    u32 NewThreadId();
-
     /**
      * Gets the current thread
      */
@@ -150,7 +146,6 @@ private:
     Kernel::KernelSystem& kernel;
     ARM_Interface* cpu;
 
-    u32 next_thread_id = 1;
     std::shared_ptr<Thread> current_thread;
     Common::ThreadQueueList<Thread*, ThreadPrioLowest + 1> ready_queue;
     std::unordered_map<u64, Thread*> wakeup_callback_table;
@@ -167,7 +162,6 @@ private:
     friend class boost::serialization::access;
     template <class Archive>
     void serialize(Archive& ar, const unsigned int file_version) {
-        ar& next_thread_id;
         ar& current_thread;
         ar& ready_queue;
         ar& wakeup_callback_table;
@@ -177,7 +171,7 @@ private:
 
 class Thread final : public WaitObject {
 public:
-    explicit Thread(KernelSystem&);
+    explicit Thread(KernelSystem&, u32 core_id);
     ~Thread() override;
 
     std::string GetName() const override {
@@ -329,6 +323,8 @@ public:
     // available. In case of a timeout, the object will be nullptr.
     std::shared_ptr<WakeupCallback> wakeup_callback;
 
+    const u32 core_id;
+
 private:
     ThreadManager& thread_manager;
 
@@ -351,4 +347,20 @@ std::shared_ptr<Thread> SetupMainThread(KernelSystem& kernel, u32 entry_point, u
 } // namespace Kernel
 
 BOOST_CLASS_EXPORT_KEY(Kernel::Thread)
-CONSTRUCT_KERNEL_OBJECT(Kernel::Thread)
+
+namespace boost::serialization {
+
+template <class Archive>
+inline void save_construct_data(Archive& ar, const Kernel::Thread* t,
+                                const unsigned int file_version) {
+    ar << t->core_id;
+}
+
+template <class Archive>
+inline void load_construct_data(Archive& ar, Kernel::Thread* t, const unsigned int file_version) {
+    u32 core_id;
+    ar >> core_id;
+    ::new (t) Kernel::Thread(Core::Global<Kernel::KernelSystem>(), core_id);
+}
+
+} // namespace boost::serialization

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

@@ -6,9 +6,7 @@
 
 #include <array>
 #include <atomic>
-#ifndef _MSC_VER
 #include <cstddef>
-#endif
 #include <memory>
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
@@ -177,10 +175,6 @@ struct GyroscopeCalibrateParam {
     } x, y, z;
 };
 
-// TODO: MSVC does not support using offsetof() on non-static data members even though this
-//       is technically allowed since C++11. This macro should be enabled once MSVC adds
-//       support for that.
-#ifndef _MSC_VER
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
     static_assert(offsetof(SharedMem, field_name) == position * 4,                                 \
                   "Field " #field_name " has invalid position")
@@ -189,7 +183,6 @@ ASSERT_REG_POSITION(pad.index_reset_ticks, 0x0);
 ASSERT_REG_POSITION(touch.index_reset_ticks, 0x2A);
 
 #undef ASSERT_REG_POSITION
-#endif // !defined(_MSC_VER)
 
 struct DirectionState {
     bool up;

src/core/hle/service/http_c.cpp

@@ -2,9 +2,14 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <atomic>
+#ifdef ENABLE_WEB_SERVICE
+#include <LUrlParser.h>
+#endif
 #include <cryptopp/aes.h>
 #include <cryptopp/modes.h>
 #include "common/archives.h"
+#include "common/assert.h"
 #include "core/core.h"
 #include "core/file_sys/archive_ncch.h"
 #include "core/file_sys/file_backend.h"
@@ -52,6 +57,82 @@ const ResultCode ERROR_WRONG_CERT_HANDLE = // 0xD8A0A0C9
 const ResultCode ERROR_CERT_ALREADY_SET = // 0xD8A0A03D
     ResultCode(61, ErrorModule::HTTP, ErrorSummary::InvalidState, ErrorLevel::Permanent);
 
+void Context::MakeRequest() {
+    ASSERT(state == RequestState::NotStarted);
+
+#ifdef ENABLE_WEB_SERVICE
+    LUrlParser::clParseURL parsedUrl = LUrlParser::clParseURL::ParseURL(url);
+    int port;
+    std::unique_ptr<httplib::Client> client;
+    if (parsedUrl.m_Scheme == "http") {
+        if (!parsedUrl.GetPort(&port)) {
+            port = 80;
+        }
+        // TODO(B3N30): Support for setting timeout
+        // Figure out what the default timeout on 3DS is
+        client = std::make_unique<httplib::Client>(parsedUrl.m_Host.c_str(), port);
+    } else {
+        if (!parsedUrl.GetPort(&port)) {
+            port = 443;
+        }
+        // TODO(B3N30): Support for setting timeout
+        // Figure out what the default timeout on 3DS is
+
+        auto ssl_client = std::make_unique<httplib::SSLClient>(parsedUrl.m_Host, port);
+        SSL_CTX* ctx = ssl_client->ssl_context();
+        client = std::move(ssl_client);
+
+        if (auto client_cert = ssl_config.client_cert_ctx.lock()) {
+            SSL_CTX_use_certificate_ASN1(ctx, client_cert->certificate.size(),
+                                         client_cert->certificate.data());
+            SSL_CTX_use_PrivateKey_ASN1(EVP_PKEY_RSA, ctx, client_cert->private_key.data(),
+                                        client_cert->private_key.size());
+        }
+
+        // TODO(B3N30): Check for SSLOptions-Bits and set the verify method accordingly
+        // https://www.3dbrew.org/wiki/SSL_Services#SSLOpt
+        // Hack: Since for now RootCerts are not implemented we set the VerifyMode to None.
+        SSL_CTX_set_verify(ctx, SSL_VERIFY_NONE, NULL);
+    }
+
+    state = RequestState::InProgress;
+
+    static const std::unordered_map<RequestMethod, std::string> request_method_strings{
+        {RequestMethod::Get, "GET"},       {RequestMethod::Post, "POST"},
+        {RequestMethod::Head, "HEAD"},     {RequestMethod::Put, "PUT"},
+        {RequestMethod::Delete, "DELETE"}, {RequestMethod::PostEmpty, "POST"},
+        {RequestMethod::PutEmpty, "PUT"},
+    };
+
+    httplib::Request request;
+    request.method = request_method_strings.at(method);
+    request.path = url;
+    // TODO(B3N30): Add post data body
+    request.progress = [this](u64 current, u64 total) -> bool {
+        // TODO(B3N30): Is there a state that shows response header are available
+        current_download_size_bytes = current;
+        total_download_size_bytes = total;
+        return true;
+    };
+
+    for (const auto& header : headers) {
+        request.headers.emplace(header.name, header.value);
+    }
+
+    if (!client->send(request, response)) {
+        LOG_ERROR(Service_HTTP, "Request failed");
+        state = RequestState::TimedOut;
+    } else {
+        LOG_DEBUG(Service_HTTP, "Request successful");
+        // TODO(B3N30): Verify this state on HW
+        state = RequestState::ReadyToDownloadContent;
+    }
+#else
+    LOG_ERROR(Service_HTTP, "Tried to make request but WebServices is not enabled in this build");
+    state = RequestState::TimedOut;
+#endif
+}
+
 void HTTP_C::Initialize(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp(ctx, 0x1, 1, 4);
     const u32 shmem_size = rp.Pop<u32>();
@@ -156,7 +237,15 @@ void HTTP_C::BeginRequest(Kernel::HLERequestContext& ctx) {
     auto itr = contexts.find(context_handle);
     ASSERT(itr != contexts.end());
 
-    // TODO(B3N30): Make the request
+    // On a 3DS BeginRequest and BeginRequestAsync will push the Request to a worker queue.
+    // You can only enqueue 8 requests at the same time.
+    // trying to enqueue any more will either fail (BeginRequestAsync), or block (BeginRequest)
+    // Note that you only can have 8 Contexts at a time. So this difference shouldn't matter
+    // Then there are 3? worker threads that pop the requests from the queue and send them
+    // For now make every request async in it's own thread.
+
+    itr->second.request_future =
+        std::async(std::launch::async, &Context::MakeRequest, std::ref(itr->second));
 
     IPC::RequestBuilder rb = rp.MakeBuilder(1, 0);
     rb.Push(RESULT_SUCCESS);
@@ -201,7 +290,15 @@ void HTTP_C::BeginRequestAsync(Kernel::HLERequestContext& ctx) {
     auto itr = contexts.find(context_handle);
     ASSERT(itr != contexts.end());
 
-    // TODO(B3N30): Make the request
+    // On a 3DS BeginRequest and BeginRequestAsync will push the Request to a worker queue.
+    // You can only enqueue 8 requests at the same time.
+    // trying to enqueue any more will either fail (BeginRequestAsync), or block (BeginRequest)
+    // Note that you only can have 8 Contexts at a time. So this difference shouldn't matter
+    // Then there are 3? worker threads that pop the requests from the queue and send them
+    // For now make every request async in it's own thread.
+
+    itr->second.request_future =
+        std::async(std::launch::async, &Context::MakeRequest, std::ref(itr->second));
 
     IPC::RequestBuilder rb = rp.MakeBuilder(1, 0);
     rb.Push(RESULT_SUCCESS);
@@ -264,7 +361,7 @@ void HTTP_C::CreateContext(Kernel::HLERequestContext& ctx) {
         return;
     }
 
-    contexts.emplace(++context_counter, Context());
+    contexts.try_emplace(++context_counter);
     contexts[context_counter].url = std::move(url);
     contexts[context_counter].method = method;
     contexts[context_counter].state = RequestState::NotStarted;
@@ -311,10 +408,9 @@ void HTTP_C::CloseContext(Kernel::HLERequestContext& ctx) {
     }
 
     // TODO(Subv): What happens if you try to close a context that's currently being used?
-    ASSERT(itr->second.state == RequestState::NotStarted);
-
     // TODO(Subv): Make sure that only the session that created the context can close it.
 
+    // Note that this will block if a request is still in progress
     contexts.erase(itr);
     session_data->num_http_contexts--;
 

src/core/hle/service/http_c.h

@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <future>
 #include <memory>
 #include <string>
 #include <unordered_map>
@@ -15,6 +16,12 @@
 #include <boost/serialization/unordered_map.hpp>
 #include <boost/serialization/vector.hpp>
 #include <boost/serialization/weak_ptr.hpp>
+#ifdef ENABLE_WEB_SERVICE
+#if defined(__ANDROID__)
+#include <ifaddrs.h>
+#endif
+#include <httplib.h>
+#endif
 #include "core/hle/kernel/shared_memory.h"
 #include "core/hle/service/service.h"
 
@@ -113,8 +120,7 @@ public:
     Context(const Context&) = delete;
     Context& operator=(const Context&) = delete;
 
-    Context(Context&& other) = default;
-    Context& operator=(Context&&) = default;
+    void MakeRequest();
 
     struct Proxy {
         std::string url;
@@ -195,14 +201,21 @@ public:
     u32 session_id;
     std::string url;
     RequestMethod method;
-    RequestState state = RequestState::NotStarted;
-    boost::optional<Proxy> proxy;
-    boost::optional<BasicAuth> basic_auth;
+    std::atomic<RequestState> state = RequestState::NotStarted;
+    std::optional<Proxy> proxy;
+    std::optional<BasicAuth> basic_auth;
     SSLConfig ssl_config{};
     u32 socket_buffer_size;
     std::vector<RequestHeader> headers;
     std::vector<PostData> post_data;
 
+    std::future<void> request_future;
+    std::atomic<u64> current_download_size_bytes;
+    std::atomic<u64> total_download_size_bytes;
+#ifdef ENABLE_WEB_SERVICE
+    httplib::Response response;
+#endif
+
 private:
     template <class Archive>
     void serialize(Archive& ar, const unsigned int) {
@@ -219,6 +232,7 @@ private:
         ar& post_data;
     }
     friend class boost::serialization::access;
+
 };
 
 struct SessionData : public Kernel::SessionRequestHandler::SessionDataBase {

src/core/hle/service/ldr_ro/cro_helper.cpp

@@ -55,7 +55,7 @@ VAddr CROHelper::SegmentTagToAddress(SegmentTag segment_tag) const {
         return 0;
 
     SegmentEntry entry;
-    GetEntry(memory, segment_tag.segment_index, entry);
+    GetEntry(system.Memory(), segment_tag.segment_index, entry);
 
     if (segment_tag.offset_into_segment >= entry.size)
         return 0;
@@ -71,12 +71,12 @@ ResultCode CROHelper::ApplyRelocation(VAddr target_address, RelocationType reloc
         break;
     case RelocationType::AbsoluteAddress:
     case RelocationType::AbsoluteAddress2:
-        memory.Write32(target_address, symbol_address + addend);
-        cpu.InvalidateCacheRange(target_address, sizeof(u32));
+        system.Memory().Write32(target_address, symbol_address + addend);
+        system.InvalidateCacheRange(target_address, sizeof(u32));
         break;
     case RelocationType::RelativeAddress:
-        memory.Write32(target_address, symbol_address + addend - target_future_address);
-        cpu.InvalidateCacheRange(target_address, sizeof(u32));
+        system.Memory().Write32(target_address, symbol_address + addend - target_future_address);
+        system.InvalidateCacheRange(target_address, sizeof(u32));
         break;
     case RelocationType::ThumbBranch:
     case RelocationType::ArmBranch:
@@ -98,8 +98,8 @@ ResultCode CROHelper::ClearRelocation(VAddr target_address, RelocationType reloc
     case RelocationType::AbsoluteAddress:
     case RelocationType::AbsoluteAddress2:
     case RelocationType::RelativeAddress:
-        memory.Write32(target_address, 0);
-        cpu.InvalidateCacheRange(target_address, sizeof(u32));
+        system.Memory().Write32(target_address, 0);
+        system.InvalidateCacheRange(target_address, sizeof(u32));
         break;
     case RelocationType::ThumbBranch:
     case RelocationType::ArmBranch:
@@ -121,7 +121,8 @@ ResultCode CROHelper::ApplyRelocationBatch(VAddr batch, u32 symbol_address, bool
     VAddr relocation_address = batch;
     while (true) {
         RelocationEntry relocation;
-        memory.ReadBlock(process, relocation_address, &relocation, sizeof(RelocationEntry));
+        system.Memory().ReadBlock(process, relocation_address, &relocation,
+                                  sizeof(RelocationEntry));
 
         VAddr relocation_target = SegmentTagToAddress(relocation.target_position);
         if (relocation_target == 0) {
@@ -142,9 +143,9 @@ ResultCode CROHelper::ApplyRelocationBatch(VAddr batch, u32 symbol_address, bool
     }
 
     RelocationEntry relocation;
-    memory.ReadBlock(process, batch, &relocation, sizeof(RelocationEntry));
+    system.Memory().ReadBlock(process, batch, &relocation, sizeof(RelocationEntry));
     relocation.is_batch_resolved = reset ? 0 : 1;
-    memory.WriteBlock(process, batch, &relocation, sizeof(RelocationEntry));
+    system.Memory().WriteBlock(process, batch, &relocation, sizeof(RelocationEntry));
     return RESULT_SUCCESS;
 }
 
@@ -154,13 +155,13 @@ VAddr CROHelper::FindExportNamedSymbol(const std::string& name) const {
 
     std::size_t len = name.size();
     ExportTreeEntry entry;
-    GetEntry(memory, 0, entry);
+    GetEntry(system.Memory(), 0, entry);
     ExportTreeEntry::Child next;
     next.raw = entry.left.raw;
     u32 found_id;
 
     while (true) {
-        GetEntry(memory, next.next_index, entry);
+        GetEntry(system.Memory(), next.next_index, entry);
 
         if (next.is_end) {
             found_id = entry.export_table_index;
@@ -186,9 +187,9 @@ VAddr CROHelper::FindExportNamedSymbol(const std::string& name) const {
 
     u32 export_strings_size = GetField(ExportStringsSize);
     ExportNamedSymbolEntry symbol_entry;
-    GetEntry(memory, found_id, symbol_entry);
+    GetEntry(system.Memory(), found_id, symbol_entry);
 
-    if (memory.ReadCString(symbol_entry.name_offset, export_strings_size) != name)
+    if (system.Memory().ReadCString(symbol_entry.name_offset, export_strings_size) != name)
         return 0;
 
     return SegmentTagToAddress(symbol_entry.symbol_position);
@@ -279,7 +280,7 @@ ResultVal<VAddr> CROHelper::RebaseSegmentTable(u32 cro_size, VAddr data_segment_
     u32 segment_num = GetField(SegmentNum);
     for (u32 i = 0; i < segment_num; ++i) {
         SegmentEntry segment;
-        GetEntry(memory, i, segment);
+        GetEntry(system.Memory(), i, segment);
         if (segment.type == SegmentType::Data) {
             if (segment.size != 0) {
                 if (segment.size > data_segment_size)
@@ -298,7 +299,7 @@ ResultVal<VAddr> CROHelper::RebaseSegmentTable(u32 cro_size, VAddr data_segment_
             if (segment.offset > module_address + cro_size)
                 return CROFormatError(0x19);
         }
-        SetEntry(memory, i, segment);
+        SetEntry(system.Memory(), i, segment);
     }
     return MakeResult<u32>(prev_data_segment + module_address);
 }
@@ -310,7 +311,7 @@ ResultCode CROHelper::RebaseExportNamedSymbolTable() {
     u32 export_named_symbol_num = GetField(ExportNamedSymbolNum);
     for (u32 i = 0; i < export_named_symbol_num; ++i) {
         ExportNamedSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
 
         if (entry.name_offset != 0) {
             entry.name_offset += module_address;
@@ -320,7 +321,7 @@ ResultCode CROHelper::RebaseExportNamedSymbolTable() {
             }
         }
 
-        SetEntry(memory, i, entry);
+        SetEntry(system.Memory(), i, entry);
     }
     return RESULT_SUCCESS;
 }
@@ -329,7 +330,7 @@ ResultCode CROHelper::VerifyExportTreeTable() const {
     u32 tree_num = GetField(ExportTreeNum);
     for (u32 i = 0; i < tree_num; ++i) {
         ExportTreeEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
 
         if (entry.left.next_index >= tree_num || entry.right.next_index >= tree_num) {
             return CROFormatError(0x11);
@@ -353,7 +354,7 @@ ResultCode CROHelper::RebaseImportModuleTable() {
     u32 module_num = GetField(ImportModuleNum);
     for (u32 i = 0; i < module_num; ++i) {
         ImportModuleEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
 
         if (entry.name_offset != 0) {
             entry.name_offset += module_address;
@@ -379,7 +380,7 @@ ResultCode CROHelper::RebaseImportModuleTable() {
             }
         }
 
-        SetEntry(memory, i, entry);
+        SetEntry(system.Memory(), i, entry);
     }
     return RESULT_SUCCESS;
 }
@@ -395,7 +396,7 @@ ResultCode CROHelper::RebaseImportNamedSymbolTable() {
     u32 num = GetField(ImportNamedSymbolNum);
     for (u32 i = 0; i < num; ++i) {
         ImportNamedSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
 
         if (entry.name_offset != 0) {
             entry.name_offset += module_address;
@@ -413,7 +414,7 @@ ResultCode CROHelper::RebaseImportNamedSymbolTable() {
             }
         }
 
-        SetEntry(memory, i, entry);
+        SetEntry(system.Memory(), i, entry);
     }
     return RESULT_SUCCESS;
 }
@@ -427,7 +428,7 @@ ResultCode CROHelper::RebaseImportIndexedSymbolTable() {
     u32 num = GetField(ImportIndexedSymbolNum);
     for (u32 i = 0; i < num; ++i) {
         ImportIndexedSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
 
         if (entry.relocation_batch_offset != 0) {
             entry.relocation_batch_offset += module_address;
@@ -437,7 +438,7 @@ ResultCode CROHelper::RebaseImportIndexedSymbolTable() {
             }
         }
 
-        SetEntry(memory, i, entry);
+        SetEntry(system.Memory(), i, entry);
     }
     return RESULT_SUCCESS;
 }
@@ -451,7 +452,7 @@ ResultCode CROHelper::RebaseImportAnonymousSymbolTable() {
     u32 num = GetField(ImportAnonymousSymbolNum);
     for (u32 i = 0; i < num; ++i) {
         ImportAnonymousSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
 
         if (entry.relocation_batch_offset != 0) {
             entry.relocation_batch_offset += module_address;
@@ -461,7 +462,7 @@ ResultCode CROHelper::RebaseImportAnonymousSymbolTable() {
             }
         }
 
-        SetEntry(memory, i, entry);
+        SetEntry(system.Memory(), i, entry);
     }
     return RESULT_SUCCESS;
 }
@@ -476,14 +477,14 @@ ResultCode CROHelper::ResetExternalRelocations() {
     ExternalRelocationEntry relocation;
 
     // Verifies that the last relocation is the end of a batch
-    GetEntry(memory, external_relocation_num - 1, relocation);
+    GetEntry(system.Memory(), external_relocation_num - 1, relocation);
     if (!relocation.is_batch_end) {
         return CROFormatError(0x12);
     }
 
     bool batch_begin = true;
     for (u32 i = 0; i < external_relocation_num; ++i) {
-        GetEntry(memory, i, relocation);
+        GetEntry(system.Memory(), i, relocation);
         VAddr relocation_target = SegmentTagToAddress(relocation.target_position);
 
         if (relocation_target == 0) {
@@ -500,7 +501,7 @@ ResultCode CROHelper::ResetExternalRelocations() {
         if (batch_begin) {
             // resets to unresolved state
             relocation.is_batch_resolved = 0;
-            SetEntry(memory, i, relocation);
+            SetEntry(system.Memory(), i, relocation);
         }
 
         // if current is an end, then the next is a beginning
@@ -516,7 +517,7 @@ ResultCode CROHelper::ClearExternalRelocations() {
 
     bool batch_begin = true;
     for (u32 i = 0; i < external_relocation_num; ++i) {
-        GetEntry(memory, i, relocation);
+        GetEntry(system.Memory(), i, relocation);
         VAddr relocation_target = SegmentTagToAddress(relocation.target_position);
 
         if (relocation_target == 0) {
@@ -532,7 +533,7 @@ ResultCode CROHelper::ClearExternalRelocations() {
         if (batch_begin) {
             // resets to unresolved state
             relocation.is_batch_resolved = 0;
-            SetEntry(memory, i, relocation);
+            SetEntry(system.Memory(), i, relocation);
         }
 
         // if current is an end, then the next is a beginning
@@ -548,13 +549,13 @@ ResultCode CROHelper::ApplyStaticAnonymousSymbolToCRS(VAddr crs_address) {
         static_relocation_table_offset +
         GetField(StaticRelocationNum) * sizeof(StaticRelocationEntry);
 
-    CROHelper crs(crs_address, process, memory, cpu);
+    CROHelper crs(crs_address, process, system);
     u32 offset_export_num = GetField(StaticAnonymousSymbolNum);
     LOG_INFO(Service_LDR, "CRO \"{}\" exports {} static anonymous symbols", ModuleName(),
              offset_export_num);
     for (u32 i = 0; i < offset_export_num; ++i) {
         StaticAnonymousSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
         u32 batch_address = entry.relocation_batch_offset + module_address;
 
         if (batch_address < static_relocation_table_offset ||
@@ -579,7 +580,7 @@ ResultCode CROHelper::ApplyInternalRelocations(u32 old_data_segment_address) {
     u32 internal_relocation_num = GetField(InternalRelocationNum);
     for (u32 i = 0; i < internal_relocation_num; ++i) {
         InternalRelocationEntry relocation;
-        GetEntry(memory, i, relocation);
+        GetEntry(system.Memory(), i, relocation);
         VAddr target_addressB = SegmentTagToAddress(relocation.target_position);
         if (target_addressB == 0) {
             return CROFormatError(0x15);
@@ -587,7 +588,7 @@ ResultCode CROHelper::ApplyInternalRelocations(u32 old_data_segment_address) {
 
         VAddr target_address;
         SegmentEntry target_segment;
-        GetEntry(memory, relocation.target_position.segment_index, target_segment);
+        GetEntry(system.Memory(), relocation.target_position.segment_index, target_segment);
 
         if (target_segment.type == SegmentType::Data) {
             // If the relocation is to the .data segment, we need to relocate it in the old buffer
@@ -602,7 +603,7 @@ ResultCode CROHelper::ApplyInternalRelocations(u32 old_data_segment_address) {
         }
 
         SegmentEntry symbol_segment;
-        GetEntry(memory, relocation.symbol_segment, symbol_segment);
+        GetEntry(system.Memory(), relocation.symbol_segment, symbol_segment);
         LOG_TRACE(Service_LDR, "Internally relocates 0x{:08X} with 0x{:08X}", target_address,
                   symbol_segment.offset);
         ResultCode result = ApplyRelocation(target_address, relocation.type, relocation.addend,
@@ -619,7 +620,7 @@ ResultCode CROHelper::ClearInternalRelocations() {
     u32 internal_relocation_num = GetField(InternalRelocationNum);
     for (u32 i = 0; i < internal_relocation_num; ++i) {
         InternalRelocationEntry relocation;
-        GetEntry(memory, i, relocation);
+        GetEntry(system.Memory(), i, relocation);
         VAddr target_address = SegmentTagToAddress(relocation.target_position);
 
         if (target_address == 0) {
@@ -639,13 +640,13 @@ void CROHelper::UnrebaseImportAnonymousSymbolTable() {
     u32 num = GetField(ImportAnonymousSymbolNum);
     for (u32 i = 0; i < num; ++i) {
         ImportAnonymousSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
 
         if (entry.relocation_batch_offset != 0) {
             entry.relocation_batch_offset -= module_address;
         }
 
-        SetEntry(memory, i, entry);
+        SetEntry(system.Memory(), i, entry);
     }
 }
 
@@ -653,13 +654,13 @@ void CROHelper::UnrebaseImportIndexedSymbolTable() {
     u32 num = GetField(ImportIndexedSymbolNum);
     for (u32 i = 0; i < num; ++i) {
         ImportIndexedSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
 
         if (entry.relocation_batch_offset != 0) {
             entry.relocation_batch_offset -= module_address;
         }
 
-        SetEntry(memory, i, entry);
+        SetEntry(system.Memory(), i, entry);
     }
 }
 
@@ -667,7 +668,7 @@ void CROHelper::UnrebaseImportNamedSymbolTable() {
     u32 num = GetField(ImportNamedSymbolNum);
     for (u32 i = 0; i < num; ++i) {
         ImportNamedSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
 
         if (entry.name_offset != 0) {
             entry.name_offset -= module_address;
@@ -677,7 +678,7 @@ void CROHelper::UnrebaseImportNamedSymbolTable() {
             entry.relocation_batch_offset -= module_address;
         }
 
-        SetEntry(memory, i, entry);
+        SetEntry(system.Memory(), i, entry);
     }
 }
 
@@ -685,7 +686,7 @@ void CROHelper::UnrebaseImportModuleTable() {
     u32 module_num = GetField(ImportModuleNum);
     for (u32 i = 0; i < module_num; ++i) {
         ImportModuleEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
 
         if (entry.name_offset != 0) {
             entry.name_offset -= module_address;
@@ -699,7 +700,7 @@ void CROHelper::UnrebaseImportModuleTable() {
             entry.import_anonymous_symbol_table_offset -= module_address;
         }
 
-        SetEntry(memory, i, entry);
+        SetEntry(system.Memory(), i, entry);
     }
 }
 
@@ -707,13 +708,13 @@ void CROHelper::UnrebaseExportNamedSymbolTable() {
     u32 export_named_symbol_num = GetField(ExportNamedSymbolNum);
     for (u32 i = 0; i < export_named_symbol_num; ++i) {
         ExportNamedSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
 
         if (entry.name_offset != 0) {
             entry.name_offset -= module_address;
         }
 
-        SetEntry(memory, i, entry);
+        SetEntry(system.Memory(), i, entry);
     }
 }
 
@@ -721,7 +722,7 @@ void CROHelper::UnrebaseSegmentTable() {
     u32 segment_num = GetField(SegmentNum);
     for (u32 i = 0; i < segment_num; ++i) {
         SegmentEntry segment;
-        GetEntry(memory, i, segment);
+        GetEntry(system.Memory(), i, segment);
 
         if (segment.type == SegmentType::BSS) {
             segment.offset = 0;
@@ -729,7 +730,7 @@ void CROHelper::UnrebaseSegmentTable() {
             segment.offset -= module_address;
         }
 
-        SetEntry(memory, i, segment);
+        SetEntry(system.Memory(), i, segment);
     }
 }
 
@@ -751,17 +752,17 @@ ResultCode CROHelper::ApplyImportNamedSymbol(VAddr crs_address) {
     u32 symbol_import_num = GetField(ImportNamedSymbolNum);
     for (u32 i = 0; i < symbol_import_num; ++i) {
         ImportNamedSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
         VAddr relocation_addr = entry.relocation_batch_offset;
         ExternalRelocationEntry relocation_entry;
-        memory.ReadBlock(process, relocation_addr, &relocation_entry,
-                         sizeof(ExternalRelocationEntry));
+        system.Memory().ReadBlock(process, relocation_addr, &relocation_entry,
+                                  sizeof(ExternalRelocationEntry));
 
         if (!relocation_entry.is_batch_resolved) {
             ResultCode result = ForEachAutoLinkCRO(
-                process, memory, cpu, crs_address, [&](CROHelper source) -> ResultVal<bool> {
+                process, system, crs_address, [&](CROHelper source) -> ResultVal<bool> {
                     std::string symbol_name =
-                        memory.ReadCString(entry.name_offset, import_strings_size);
+                        system.Memory().ReadCString(entry.name_offset, import_strings_size);
                     u32 symbol_address = source.FindExportNamedSymbol(symbol_name);
 
                     if (symbol_address != 0) {
@@ -794,11 +795,11 @@ ResultCode CROHelper::ResetImportNamedSymbol() {
     u32 symbol_import_num = GetField(ImportNamedSymbolNum);
     for (u32 i = 0; i < symbol_import_num; ++i) {
         ImportNamedSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
         VAddr relocation_addr = entry.relocation_batch_offset;
         ExternalRelocationEntry relocation_entry;
-        memory.ReadBlock(process, relocation_addr, &relocation_entry,
-                         sizeof(ExternalRelocationEntry));
+        system.Memory().ReadBlock(process, relocation_addr, &relocation_entry,
+                                  sizeof(ExternalRelocationEntry));
 
         ResultCode result = ApplyRelocationBatch(relocation_addr, unresolved_symbol, true);
         if (result.IsError()) {
@@ -815,11 +816,11 @@ ResultCode CROHelper::ResetImportIndexedSymbol() {
     u32 import_num = GetField(ImportIndexedSymbolNum);
     for (u32 i = 0; i < import_num; ++i) {
         ImportIndexedSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
         VAddr relocation_addr = entry.relocation_batch_offset;
         ExternalRelocationEntry relocation_entry;
-        memory.ReadBlock(process, relocation_addr, &relocation_entry,
-                         sizeof(ExternalRelocationEntry));
+        system.Memory().ReadBlock(process, relocation_addr, &relocation_entry,
+                                  sizeof(ExternalRelocationEntry));
 
         ResultCode result = ApplyRelocationBatch(relocation_addr, unresolved_symbol, true);
         if (result.IsError()) {
@@ -836,11 +837,11 @@ ResultCode CROHelper::ResetImportAnonymousSymbol() {
     u32 import_num = GetField(ImportAnonymousSymbolNum);
     for (u32 i = 0; i < import_num; ++i) {
         ImportAnonymousSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
         VAddr relocation_addr = entry.relocation_batch_offset;
         ExternalRelocationEntry relocation_entry;
-        memory.ReadBlock(process, relocation_addr, &relocation_entry,
-                         sizeof(ExternalRelocationEntry));
+        system.Memory().ReadBlock(process, relocation_addr, &relocation_entry,
+                                  sizeof(ExternalRelocationEntry));
 
         ResultCode result = ApplyRelocationBatch(relocation_addr, unresolved_symbol, true);
         if (result.IsError()) {
@@ -857,19 +858,20 @@ ResultCode CROHelper::ApplyModuleImport(VAddr crs_address) {
     u32 import_module_num = GetField(ImportModuleNum);
     for (u32 i = 0; i < import_module_num; ++i) {
         ImportModuleEntry entry;
-        GetEntry(memory, i, entry);
-        std::string want_cro_name = memory.ReadCString(entry.name_offset, import_strings_size);
+        GetEntry(system.Memory(), i, entry);
+        std::string want_cro_name =
+            system.Memory().ReadCString(entry.name_offset, import_strings_size);
 
         ResultCode result = ForEachAutoLinkCRO(
-            process, memory, cpu, crs_address, [&](CROHelper source) -> ResultVal<bool> {
+            process, system, crs_address, [&](CROHelper source) -> ResultVal<bool> {
                 if (want_cro_name == source.ModuleName()) {
                     LOG_INFO(Service_LDR, "CRO \"{}\" imports {} indexed symbols from \"{}\"",
                              ModuleName(), entry.import_indexed_symbol_num, source.ModuleName());
                     for (u32 j = 0; j < entry.import_indexed_symbol_num; ++j) {
                         ImportIndexedSymbolEntry im;
-                        entry.GetImportIndexedSymbolEntry(process, memory, j, im);
+                        entry.GetImportIndexedSymbolEntry(process, system.Memory(), j, im);
                         ExportIndexedSymbolEntry ex;
-                        source.GetEntry(memory, im.index, ex);
+                        source.GetEntry(system.Memory(), im.index, ex);
                         u32 symbol_address = source.SegmentTagToAddress(ex.symbol_position);
                         LOG_TRACE(Service_LDR, "    Imports 0x{:08X}", symbol_address);
                         ResultCode result =
@@ -884,7 +886,7 @@ ResultCode CROHelper::ApplyModuleImport(VAddr crs_address) {
                              ModuleName(), entry.import_anonymous_symbol_num, source.ModuleName());
                     for (u32 j = 0; j < entry.import_anonymous_symbol_num; ++j) {
                         ImportAnonymousSymbolEntry im;
-                        entry.GetImportAnonymousSymbolEntry(process, memory, j, im);
+                        entry.GetImportAnonymousSymbolEntry(process, system.Memory(), j, im);
                         u32 symbol_address = source.SegmentTagToAddress(im.symbol_position);
                         LOG_TRACE(Service_LDR, "    Imports 0x{:08X}", symbol_address);
                         ResultCode result =
@@ -913,15 +915,15 @@ ResultCode CROHelper::ApplyExportNamedSymbol(CROHelper target) {
     u32 target_symbol_import_num = target.GetField(ImportNamedSymbolNum);
     for (u32 i = 0; i < target_symbol_import_num; ++i) {
         ImportNamedSymbolEntry entry;
-        target.GetEntry(memory, i, entry);
+        target.GetEntry(system.Memory(), i, entry);
         VAddr relocation_addr = entry.relocation_batch_offset;
         ExternalRelocationEntry relocation_entry;
-        memory.ReadBlock(process, relocation_addr, &relocation_entry,
-                         sizeof(ExternalRelocationEntry));
+        system.Memory().ReadBlock(process, relocation_addr, &relocation_entry,
+                                  sizeof(ExternalRelocationEntry));
 
         if (!relocation_entry.is_batch_resolved) {
             std::string symbol_name =
-                memory.ReadCString(entry.name_offset, target_import_strings_size);
+                system.Memory().ReadCString(entry.name_offset, target_import_strings_size);
             u32 symbol_address = FindExportNamedSymbol(symbol_name);
             if (symbol_address != 0) {
                 LOG_TRACE(Service_LDR, "    exports symbol \"{}\"", symbol_name);
@@ -944,15 +946,15 @@ ResultCode CROHelper::ResetExportNamedSymbol(CROHelper target) {
     u32 target_symbol_import_num = target.GetField(ImportNamedSymbolNum);
     for (u32 i = 0; i < target_symbol_import_num; ++i) {
         ImportNamedSymbolEntry entry;
-        target.GetEntry(memory, i, entry);
+        target.GetEntry(system.Memory(), i, entry);
         VAddr relocation_addr = entry.relocation_batch_offset;
         ExternalRelocationEntry relocation_entry;
-        memory.ReadBlock(process, relocation_addr, &relocation_entry,
-                         sizeof(ExternalRelocationEntry));
+        system.Memory().ReadBlock(process, relocation_addr, &relocation_entry,
+                                  sizeof(ExternalRelocationEntry));
 
         if (relocation_entry.is_batch_resolved) {
             std::string symbol_name =
-                memory.ReadCString(entry.name_offset, target_import_strings_size);
+                system.Memory().ReadCString(entry.name_offset, target_import_strings_size);
             u32 symbol_address = FindExportNamedSymbol(symbol_name);
             if (symbol_address != 0) {
                 LOG_TRACE(Service_LDR, "    unexports symbol \"{}\"", symbol_name);
@@ -974,18 +976,19 @@ ResultCode CROHelper::ApplyModuleExport(CROHelper target) {
     u32 target_import_module_num = target.GetField(ImportModuleNum);
     for (u32 i = 0; i < target_import_module_num; ++i) {
         ImportModuleEntry entry;
-        target.GetEntry(memory, i, entry);
+        target.GetEntry(system.Memory(), i, entry);
 
-        if (memory.ReadCString(entry.name_offset, target_import_string_size) != module_name)
+        if (system.Memory().ReadCString(entry.name_offset, target_import_string_size) !=
+            module_name)
             continue;
 
         LOG_INFO(Service_LDR, "CRO \"{}\" exports {} indexed symbols to \"{}\"", module_name,
                  entry.import_indexed_symbol_num, target.ModuleName());
         for (u32 j = 0; j < entry.import_indexed_symbol_num; ++j) {
             ImportIndexedSymbolEntry im;
-            entry.GetImportIndexedSymbolEntry(process, memory, j, im);
+            entry.GetImportIndexedSymbolEntry(process, system.Memory(), j, im);
             ExportIndexedSymbolEntry ex;
-            GetEntry(memory, im.index, ex);
+            GetEntry(system.Memory(), im.index, ex);
             u32 symbol_address = SegmentTagToAddress(ex.symbol_position);
             LOG_TRACE(Service_LDR, "    exports symbol 0x{:08X}", symbol_address);
             ResultCode result =
@@ -1000,7 +1003,7 @@ ResultCode CROHelper::ApplyModuleExport(CROHelper target) {
                  entry.import_anonymous_symbol_num, target.ModuleName());
         for (u32 j = 0; j < entry.import_anonymous_symbol_num; ++j) {
             ImportAnonymousSymbolEntry im;
-            entry.GetImportAnonymousSymbolEntry(process, memory, j, im);
+            entry.GetImportAnonymousSymbolEntry(process, system.Memory(), j, im);
             u32 symbol_address = SegmentTagToAddress(im.symbol_position);
             LOG_TRACE(Service_LDR, "    exports symbol 0x{:08X}", symbol_address);
             ResultCode result =
@@ -1023,16 +1026,17 @@ ResultCode CROHelper::ResetModuleExport(CROHelper target) {
     u32 target_import_module_num = target.GetField(ImportModuleNum);
     for (u32 i = 0; i < target_import_module_num; ++i) {
         ImportModuleEntry entry;
-        target.GetEntry(memory, i, entry);
+        target.GetEntry(system.Memory(), i, entry);
 
-        if (memory.ReadCString(entry.name_offset, target_import_string_size) != module_name)
+        if (system.Memory().ReadCString(entry.name_offset, target_import_string_size) !=
+            module_name)
             continue;
 
         LOG_DEBUG(Service_LDR, "CRO \"{}\" unexports indexed symbols to \"{}\"", module_name,
                   target.ModuleName());
         for (u32 j = 0; j < entry.import_indexed_symbol_num; ++j) {
             ImportIndexedSymbolEntry im;
-            entry.GetImportIndexedSymbolEntry(process, memory, j, im);
+            entry.GetImportIndexedSymbolEntry(process, system.Memory(), j, im);
             ResultCode result =
                 target.ApplyRelocationBatch(im.relocation_batch_offset, unresolved_symbol, true);
             if (result.IsError()) {
@@ -1045,7 +1049,7 @@ ResultCode CROHelper::ResetModuleExport(CROHelper target) {
                   target.ModuleName());
         for (u32 j = 0; j < entry.import_anonymous_symbol_num; ++j) {
             ImportAnonymousSymbolEntry im;
-            entry.GetImportAnonymousSymbolEntry(process, memory, j, im);
+            entry.GetImportAnonymousSymbolEntry(process, system.Memory(), j, im);
             ResultCode result =
                 target.ApplyRelocationBatch(im.relocation_batch_offset, unresolved_symbol, true);
             if (result.IsError()) {
@@ -1063,15 +1067,16 @@ ResultCode CROHelper::ApplyExitRelocations(VAddr crs_address) {
     u32 symbol_import_num = GetField(ImportNamedSymbolNum);
     for (u32 i = 0; i < symbol_import_num; ++i) {
         ImportNamedSymbolEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
         VAddr relocation_addr = entry.relocation_batch_offset;
         ExternalRelocationEntry relocation_entry;
-        memory.ReadBlock(process, relocation_addr, &relocation_entry,
-                         sizeof(ExternalRelocationEntry));
+        system.Memory().ReadBlock(process, relocation_addr, &relocation_entry,
+                                  sizeof(ExternalRelocationEntry));
 
-        if (memory.ReadCString(entry.name_offset, import_strings_size) == "__aeabi_atexit") {
+        if (system.Memory().ReadCString(entry.name_offset, import_strings_size) ==
+            "__aeabi_atexit") {
             ResultCode result = ForEachAutoLinkCRO(
-                process, memory, cpu, crs_address, [&](CROHelper source) -> ResultVal<bool> {
+                process, system, crs_address, [&](CROHelper source) -> ResultVal<bool> {
                     u32 symbol_address = source.FindExportNamedSymbol("nnroAeabiAtexit_");
 
                     if (symbol_address != 0) {
@@ -1126,7 +1131,8 @@ ResultCode CROHelper::Rebase(VAddr crs_address, u32 cro_size, VAddr data_segment
         return result;
     }
 
-    result = VerifyStringTableLength(memory, GetField(ModuleNameOffset), GetField(ModuleNameSize));
+    result = VerifyStringTableLength(system.Memory(), GetField(ModuleNameOffset),
+                                     GetField(ModuleNameSize));
     if (result.IsError()) {
         LOG_ERROR(Service_LDR, "Error verifying module name {:08X}", result.raw);
         return result;
@@ -1155,8 +1161,8 @@ ResultCode CROHelper::Rebase(VAddr crs_address, u32 cro_size, VAddr data_segment
         return result;
     }
 
-    result =
-        VerifyStringTableLength(memory, GetField(ExportStringsOffset), GetField(ExportStringsSize));
+    result = VerifyStringTableLength(system.Memory(), GetField(ExportStringsOffset),
+                                     GetField(ExportStringsSize));
     if (result.IsError()) {
         LOG_ERROR(Service_LDR, "Error verifying export strings {:08X}", result.raw);
         return result;
@@ -1192,8 +1198,8 @@ ResultCode CROHelper::Rebase(VAddr crs_address, u32 cro_size, VAddr data_segment
         return result;
     }
 
-    result =
-        VerifyStringTableLength(memory, GetField(ImportStringsOffset), GetField(ImportStringsSize));
+    result = VerifyStringTableLength(system.Memory(), GetField(ImportStringsOffset),
+                                     GetField(ImportStringsSize));
     if (result.IsError()) {
         LOG_ERROR(Service_LDR, "Error verifying import strings {:08X}", result.raw);
         return result;
@@ -1266,11 +1272,11 @@ ResultCode CROHelper::Link(VAddr crs_address, bool link_on_load_bug_fix) {
             // so we do the same
             if (GetField(SegmentNum) >= 2) { // means we have .data segment
                 SegmentEntry entry;
-                GetEntry(memory, 2, entry);
+                GetEntry(system.Memory(), 2, entry);
                 ASSERT(entry.type == SegmentType::Data);
                 data_segment_address = entry.offset;
                 entry.offset = GetField(DataOffset);
-                SetEntry(memory, 2, entry);
+                SetEntry(system.Memory(), 2, entry);
             }
         }
         SCOPE_EXIT({
@@ -1278,9 +1284,9 @@ ResultCode CROHelper::Link(VAddr crs_address, bool link_on_load_bug_fix) {
             if (link_on_load_bug_fix) {
                 if (GetField(SegmentNum) >= 2) {
                     SegmentEntry entry;
-                    GetEntry(memory, 2, entry);
+                    GetEntry(system.Memory(), 2, entry);
                     entry.offset = data_segment_address;
-                    SetEntry(memory, 2, entry);
+                    SetEntry(system.Memory(), 2, entry);
                 }
             }
         });
@@ -1301,7 +1307,7 @@ ResultCode CROHelper::Link(VAddr crs_address, bool link_on_load_bug_fix) {
     }
 
     // Exports symbols to other modules
-    result = ForEachAutoLinkCRO(process, memory, cpu, crs_address,
+    result = ForEachAutoLinkCRO(process, system, crs_address,
                                 [this](CROHelper target) -> ResultVal<bool> {
                                     ResultCode result = ApplyExportNamedSymbol(target);
                                     if (result.IsError())
@@ -1346,7 +1352,7 @@ ResultCode CROHelper::Unlink(VAddr crs_address) {
 
     // Resets all symbols in other modules imported from this module
     // Note: the RO service seems only searching in auto-link modules
-    result = ForEachAutoLinkCRO(process, memory, cpu, crs_address,
+    result = ForEachAutoLinkCRO(process, system, crs_address,
                                 [this](CROHelper target) -> ResultVal<bool> {
                                     ResultCode result = ResetExportNamedSymbol(target);
                                     if (result.IsError())
@@ -1387,13 +1393,13 @@ void CROHelper::InitCRS() {
 }
 
 void CROHelper::Register(VAddr crs_address, bool auto_link) {
-    CROHelper crs(crs_address, process, memory, cpu);
-    CROHelper head(auto_link ? crs.NextModule() : crs.PreviousModule(), process, memory, cpu);
+    CROHelper crs(crs_address, process, system);
+    CROHelper head(auto_link ? crs.NextModule() : crs.PreviousModule(), process, system);
 
     if (head.module_address) {
         // there are already CROs registered
         // register as the new tail
-        CROHelper tail(head.PreviousModule(), process, memory, cpu);
+        CROHelper tail(head.PreviousModule(), process, system);
 
         // link with the old tail
         ASSERT(tail.NextModule() == 0);
@@ -1419,11 +1425,11 @@ void CROHelper::Register(VAddr crs_address, bool auto_link) {
 }
 
 void CROHelper::Unregister(VAddr crs_address) {
-    CROHelper crs(crs_address, process, memory, cpu);
-    CROHelper next_head(crs.NextModule(), process, memory, cpu);
-    CROHelper previous_head(crs.PreviousModule(), process, memory, cpu);
-    CROHelper next(NextModule(), process, memory, cpu);
-    CROHelper previous(PreviousModule(), process, memory, cpu);
+    CROHelper crs(crs_address, process, system);
+    CROHelper next_head(crs.NextModule(), process, system);
+    CROHelper previous_head(crs.PreviousModule(), process, system);
+    CROHelper next(NextModule(), process, system);
+    CROHelper previous(PreviousModule(), process, system);
 
     if (module_address == next_head.module_address ||
         module_address == previous_head.module_address) {
@@ -1517,7 +1523,7 @@ std::tuple<VAddr, u32> CROHelper::GetExecutablePages() const {
     u32 segment_num = GetField(SegmentNum);
     for (u32 i = 0; i < segment_num; ++i) {
         SegmentEntry entry;
-        GetEntry(memory, i, entry);
+        GetEntry(system.Memory(), i, entry);
         if (entry.type == SegmentType::Code && entry.size != 0) {
             VAddr begin = Common::AlignDown(entry.offset, Memory::PAGE_SIZE);
             VAddr end = Common::AlignUp(entry.offset + entry.size, Memory::PAGE_SIZE);

src/core/hle/service/ldr_ro/cro_helper.h

@@ -33,12 +33,11 @@ static constexpr u32 CRO_HASH_SIZE = 0x80;
 class CROHelper final {
 public:
     // TODO (wwylele): pass in the process handle for memory access
-    explicit CROHelper(VAddr cro_address, Kernel::Process& process, Memory::MemorySystem& memory,
-                       ARM_Interface& cpu)
-        : module_address(cro_address), process(process), memory(memory), cpu(cpu) {}
+    explicit CROHelper(VAddr cro_address, Kernel::Process& process, Core::System& system)
+        : module_address(cro_address), process(process), system(system) {}
 
     std::string ModuleName() const {
-        return memory.ReadCString(GetField(ModuleNameOffset), GetField(ModuleNameSize));
+        return system.Memory().ReadCString(GetField(ModuleNameOffset), GetField(ModuleNameSize));
     }
 
     u32 GetFileSize() const {
@@ -144,8 +143,7 @@ public:
 private:
     const VAddr module_address; ///< the virtual address of this module
     Kernel::Process& process;   ///< the owner process of this module
-    Memory::MemorySystem& memory;
-    ARM_Interface& cpu;
+    Core::System& system;
 
     /**
      * Each item in this enum represents a u32 field in the header begin from address+0x80,
@@ -403,11 +401,11 @@ private:
     }
 
     u32 GetField(HeaderField field) const {
-        return memory.Read32(Field(field));
+        return system.Memory().Read32(Field(field));
     }
 
     void SetField(HeaderField field, u32 value) {
-        memory.Write32(Field(field), value);
+        system.Memory().Write32(Field(field), value);
     }
 
     /**
@@ -474,12 +472,11 @@ private:
      *         otherwise error code of the last iteration.
      */
     template <typename FunctionObject>
-    static ResultCode ForEachAutoLinkCRO(Kernel::Process& process, Memory::MemorySystem& memory,
-                                         ARM_Interface& cpu, VAddr crs_address,
-                                         FunctionObject func) {
+    static ResultCode ForEachAutoLinkCRO(Kernel::Process& process, Core::System& system,
+                                         VAddr crs_address, FunctionObject func) {
         VAddr current = crs_address;
         while (current != 0) {
-            CROHelper cro(current, process, memory, cpu);
+            CROHelper cro(current, process, system);
             CASCADE_RESULT(bool next, func(cro));
             if (!next)
                 break;

src/core/hle/service/ldr_ro/ldr_ro.cpp

@@ -120,7 +120,7 @@ void RO::Initialize(Kernel::HLERequestContext& ctx) {
         return;
     }
 
-    CROHelper crs(crs_address, *process, system.Memory(), system.CPU());
+    CROHelper crs(crs_address, *process, system);
     crs.InitCRS();
 
     result = crs.Rebase(0, crs_size, 0, 0, 0, 0, true);
@@ -254,7 +254,7 @@ void RO::LoadCRO(Kernel::HLERequestContext& ctx, bool link_on_load_bug_fix) {
         return;
     }
 
-    CROHelper cro(cro_address, *process, system.Memory(), system.CPU());
+    CROHelper cro(cro_address, *process, system);
 
     result = cro.VerifyHash(cro_size, crr_address);
     if (result.IsError()) {
@@ -318,7 +318,7 @@ void RO::LoadCRO(Kernel::HLERequestContext& ctx, bool link_on_load_bug_fix) {
         }
     }
 
-    system.CPU().InvalidateCacheRange(cro_address, cro_size);
+    system.InvalidateCacheRange(cro_address, cro_size);
 
     LOG_INFO(Service_LDR, "CRO \"{}\" loaded at 0x{:08X}, fixed_end=0x{:08X}", cro.ModuleName(),
              cro_address, cro_address + fix_size);
@@ -336,7 +336,7 @@ void RO::UnloadCRO(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_LDR, "called, cro_address=0x{:08X}, zero={}, cro_buffer_ptr=0x{:08X}",
               cro_address, zero, cro_buffer_ptr);
 
-    CROHelper cro(cro_address, *process, system.Memory(), system.CPU());
+    CROHelper cro(cro_address, *process, system);
 
     IPC::RequestBuilder rb = rp.MakeBuilder(1, 0);
 
@@ -391,7 +391,7 @@ void RO::UnloadCRO(Kernel::HLERequestContext& ctx) {
         LOG_ERROR(Service_LDR, "Error unmapping CRO {:08X}", result.raw);
     }
 
-    system.CPU().InvalidateCacheRange(cro_address, fixed_size);
+    system.InvalidateCacheRange(cro_address, fixed_size);
 
     rb.Push(result);
 }
@@ -403,7 +403,7 @@ void RO::LinkCRO(Kernel::HLERequestContext& ctx) {
 
     LOG_DEBUG(Service_LDR, "called, cro_address=0x{:08X}", cro_address);
 
-    CROHelper cro(cro_address, *process, system.Memory(), system.CPU());
+    CROHelper cro(cro_address, *process, system);
 
     IPC::RequestBuilder rb = rp.MakeBuilder(1, 0);
 
@@ -443,7 +443,7 @@ void RO::UnlinkCRO(Kernel::HLERequestContext& ctx) {
 
     LOG_DEBUG(Service_LDR, "called, cro_address=0x{:08X}", cro_address);
 
-    CROHelper cro(cro_address, *process, system.Memory(), system.CPU());
+    CROHelper cro(cro_address, *process, system);
 
     IPC::RequestBuilder rb = rp.MakeBuilder(1, 0);
 
@@ -492,7 +492,7 @@ void RO::Shutdown(Kernel::HLERequestContext& ctx) {
         return;
     }
 
-    CROHelper crs(slot->loaded_crs, *process, system.Memory(), system.CPU());
+    CROHelper crs(slot->loaded_crs, *process, system);
     crs.Unrebase(true);
 
     ResultCode result = RESULT_SUCCESS;

src/core/hw/gpu.cpp

@@ -402,8 +402,8 @@ inline void Write(u32 addr, const T data) {
     switch (index) {
 
     // Memory fills are triggered once the fill value is written.
-    case GPU_REG_INDEX_WORKAROUND(memory_fill_config[0].trigger, 0x00004 + 0x3):
-    case GPU_REG_INDEX_WORKAROUND(memory_fill_config[1].trigger, 0x00008 + 0x3): {
+    case GPU_REG_INDEX(memory_fill_config[0].trigger):
+    case GPU_REG_INDEX(memory_fill_config[1].trigger): {
         const bool is_second_filler = (index != GPU_REG_INDEX(memory_fill_config[0].trigger));
         auto& config = g_regs.memory_fill_config[is_second_filler];
 

src/core/hw/gpu.h

@@ -22,41 +22,15 @@ namespace GPU {
 constexpr float SCREEN_REFRESH_RATE = 60;
 
 // Returns index corresponding to the Regs member labeled by field_name
-// TODO: Due to Visual studio bug 209229, offsetof does not return constant expressions
-//       when used with array elements (e.g. GPU_REG_INDEX(memory_fill_config[0])).
-//       For details cf.
-//       https://connect.microsoft.com/VisualStudio/feedback/details/209229/offsetof-does-not-produce-a-constant-expression-for-array-members
-//       Hopefully, this will be fixed sometime in the future.
-//       For lack of better alternatives, we currently hardcode the offsets when constant
-//       expressions are needed via GPU_REG_INDEX_WORKAROUND (on sane compilers, static_asserts
-//       will then make sure the offsets indeed match the automatically calculated ones).
 #define GPU_REG_INDEX(field_name) (offsetof(GPU::Regs, field_name) / sizeof(u32))
-#if defined(_MSC_VER)
-#define GPU_REG_INDEX_WORKAROUND(field_name, backup_workaround_index) (backup_workaround_index)
-#else
-// NOTE: Yeah, hacking in a static_assert here just to workaround the lacking MSVC compiler
-//       really is this annoying. This macro just forwards its first argument to GPU_REG_INDEX
-//       and then performs a (no-op) cast to std::size_t iff the second argument matches the
-//       expected field offset. Otherwise, the compiler will fail to compile this code.
-#define GPU_REG_INDEX_WORKAROUND(field_name, backup_workaround_index)                              \
-    ((typename std::enable_if<backup_workaround_index == GPU_REG_INDEX(field_name),                \
-                              std::size_t>::type) GPU_REG_INDEX(field_name))
-#endif
 
 // MMIO region 0x1EFxxxxx
 struct Regs {
 
 // helper macro to make sure the defined structures are of the expected size.
-#if defined(_MSC_VER)
-// TODO: MSVC does not support using sizeof() on non-static data members even though this
-//       is technically allowed since C++11. This macro should be enabled once MSVC adds
-//       support for that.
-#define ASSERT_MEMBER_SIZE(name, size_in_bytes)
-#else
 #define ASSERT_MEMBER_SIZE(name, size_in_bytes)                                                    \
     static_assert(sizeof(name) == size_in_bytes,                                                   \
                   "Structure size and register block length don't match")
-#endif
 
     // Components are laid out in reverse byte order, most significant bits first.
     enum class PixelFormat : u32 {
@@ -307,10 +281,6 @@ private:
 };
 static_assert(std::is_standard_layout<Regs>::value, "Structure does not use standard layout");
 
-// TODO: MSVC does not support using offsetof() on non-static data members even though this
-//       is technically allowed since C++11. This macro should be enabled once MSVC adds
-//       support for that.
-#ifndef _MSC_VER
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
     static_assert(offsetof(Regs, field_name) == position * 4,                                      \
                   "Field " #field_name " has invalid position")
@@ -323,7 +293,6 @@ ASSERT_REG_POSITION(display_transfer_config, 0x00300);
 ASSERT_REG_POSITION(command_processor_config, 0x00638);
 
 #undef ASSERT_REG_POSITION
-#endif // !defined(_MSC_VER)
 
 // The total number of registers is chosen arbitrarily, but let's make sure it's not some odd value
 // anyway.

src/core/loader/3dsx.cpp

@@ -309,8 +309,8 @@ ResultStatus AppLoader_THREEDSX::ReadRomFS(std::shared_ptr<FileSys::RomFSReader>
         if (!romfs_file_inner.IsOpen())
             return ResultStatus::Error;
 
-        romfs_file = std::make_shared<FileSys::RomFSReader>(std::move(romfs_file_inner),
-                                                            romfs_offset, romfs_size);
+        romfs_file = std::make_shared<FileSys::DirectRomFSReader>(std::move(romfs_file_inner),
+                                                                  romfs_offset, romfs_size);
 
         return ResultStatus::Success;
     }

src/core/loader/loader.h

@@ -105,13 +105,22 @@ public:
      * Loads the system mode that this application needs.
      * This function defaults to 2 (96MB allocated to the application) if it can't read the
      * information.
-     * @returns A pair with the optional system mode, and and the status.
+     * @returns A pair with the optional system mode, and the status.
      */
     virtual std::pair<std::optional<u32>, ResultStatus> LoadKernelSystemMode() {
         // 96MB allocated to the application.
         return std::make_pair(2, ResultStatus::Success);
     }
 
+    /**
+     * Loads the N3ds mode that this application uses.
+     * It defaults to 0 (O3DS default) if it can't read the information.
+     * @returns A pair with the optional N3ds mode, and the status.
+     */
+    virtual std::pair<std::optional<u8>, ResultStatus> LoadKernelN3dsMode() {
+        return std::make_pair(0, ResultStatus::Success);
+    }
+
     /**
      * Get whether this application is executable.
      * @param out_executable Reference to store the executable flag into.
@@ -186,6 +195,15 @@ public:
         return ResultStatus::ErrorNotImplemented;
     }
 
+    /**
+     * Dump the RomFS of the applciation
+     * @param target_path The target path to dump to
+     * @return ResultStatus result of function
+     */
+    virtual ResultStatus DumpRomFS(const std::string& target_path) {
+        return ResultStatus::ErrorNotImplemented;
+    }
+
     /**
      * Get the update RomFS of the application
      * Since the RomFS can be huge, we return a file reference instead of copying to a buffer
@@ -196,6 +214,15 @@ public:
         return ResultStatus::ErrorNotImplemented;
     }
 
+    /**
+     * Dump the update RomFS of the applciation
+     * @param target_path The target path to dump to
+     * @return ResultStatus result of function
+     */
+    virtual ResultStatus DumpUpdateRomFS(const std::string& target_path) {
+        return ResultStatus::ErrorNotImplemented;
+    }
+
     /**
      * Get the title of the application
      * @param title Reference to store the application title into

src/core/loader/ncch.cpp

@@ -61,6 +61,19 @@ std::pair<std::optional<u32>, ResultStatus> AppLoader_NCCH::LoadKernelSystemMode
                           ResultStatus::Success);
 }
 
+std::pair<std::optional<u8>, ResultStatus> AppLoader_NCCH::LoadKernelN3dsMode() {
+    if (!is_loaded) {
+        ResultStatus res = base_ncch.Load();
+        if (res != ResultStatus::Success) {
+            return std::make_pair(std::optional<u8>{}, res);
+        }
+    }
+
+    // Set the system mode as the one from the exheader.
+    return std::make_pair(overlay_ncch->exheader_header.arm11_system_local_caps.n3ds_mode,
+                          ResultStatus::Success);
+}
+
 ResultStatus AppLoader_NCCH::LoadExec(std::shared_ptr<Kernel::Process>& process) {
     using Kernel::CodeSet;
 
@@ -254,6 +267,18 @@ ResultStatus AppLoader_NCCH::ReadUpdateRomFS(std::shared_ptr<FileSys::RomFSReade
     return ResultStatus::Success;
 }
 
+ResultStatus AppLoader_NCCH::DumpRomFS(const std::string& target_path) {
+    return base_ncch.DumpRomFS(target_path);
+}
+
+ResultStatus AppLoader_NCCH::DumpUpdateRomFS(const std::string& target_path) {
+    u64 program_id;
+    ReadProgramId(program_id);
+    update_ncch.OpenFile(
+        Service::AM::GetTitleContentPath(Service::FS::MediaType::SDMC, program_id | UPDATE_MASK));
+    return update_ncch.DumpRomFS(target_path);
+}
+
 ResultStatus AppLoader_NCCH::ReadTitle(std::string& title) {
     std::vector<u8> data;
     Loader::SMDH smdh;

src/core/loader/ncch.h

@@ -41,6 +41,8 @@ public:
      */
     std::pair<std::optional<u32>, ResultStatus> LoadKernelSystemMode() override;
 
+    std::pair<std::optional<u8>, ResultStatus> LoadKernelN3dsMode() override;
+
     ResultStatus IsExecutable(bool& out_executable) override;
 
     ResultStatus ReadCode(std::vector<u8>& buffer) override;
@@ -59,6 +61,10 @@ public:
 
     ResultStatus ReadUpdateRomFS(std::shared_ptr<FileSys::RomFSReader>& romfs_file) override;
 
+    ResultStatus DumpRomFS(const std::string& target_path) override;
+
+    ResultStatus DumpUpdateRomFS(const std::string& target_path) override;
+
     ResultStatus ReadTitle(std::string& title) override;
 
 private:

src/core/rpc/rpc_server.cpp

@@ -46,7 +46,9 @@ void RPCServer::HandleWriteMemory(Packet& packet, u32 address, const u8* data, u
         Core::System::GetInstance().Memory().WriteBlock(
             *Core::System::GetInstance().Kernel().GetCurrentProcess(), address, data, data_size);
         // If the memory happens to be executable code, make sure the changes become visible
-        Core::CPU().InvalidateCacheRange(address, data_size);
+
+        // Is current core correct here?
+        Core::System::GetInstance().InvalidateCacheRange(address, data_size);
     }
     packet.SetPacketDataSize(0);
     packet.SendReply();