vi: Add FENCE_HACK, which is useful for booting BOTW.

arm_dynarmic: Support direct page table access

.travis/macos/build.sh

@@ -2,7 +2,7 @@
 
 set -o pipefail
 
-export MACOSX_DEPLOYMENT_TARGET=10.9
+export MACOSX_DEPLOYMENT_TARGET=10.12
 export Qt5_DIR=$(brew --prefix)/opt/qt5
 export UNICORNDIR=$(pwd)/externals/unicorn
 

src/core/CMakeLists.txt

@@ -92,6 +92,8 @@ add_library(core STATIC
     hle/service/aoc/aoc_u.h
     hle/service/apm/apm.cpp
     hle/service/apm/apm.h
+    hle/service/apm/interface.cpp
+    hle/service/apm/interface.h
     hle/service/audio/audio.cpp
     hle/service/audio/audio.h
     hle/service/audio/audin_u.cpp
@@ -129,6 +131,10 @@ add_library(core STATIC
     hle/service/nvdrv/devices/nvhost_as_gpu.h
     hle/service/nvdrv/devices/nvhost_ctrl.cpp
     hle/service/nvdrv/devices/nvhost_ctrl.h
+    hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
+    hle/service/nvdrv/devices/nvhost_ctrl_gpu.h
+    hle/service/nvdrv/devices/nvhost_gpu.cpp
+    hle/service/nvdrv/devices/nvhost_gpu.h
     hle/service/nvdrv/devices/nvmap.cpp
     hle/service/nvdrv/devices/nvmap.h
     hle/service/nvdrv/interface.cpp

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -8,9 +8,12 @@
 #include <dynarmic/A64/config.h>
 #include "core/arm/dynarmic/arm_dynarmic.h"
 #include "core/core_timing.h"
+#include "core/hle/kernel/memory.h"
 #include "core/hle/kernel/svc.h"
 #include "core/memory.h"
 
+using Vector = Dynarmic::A64::Vector;
+
 class ARM_Dynarmic_Callbacks : public Dynarmic::A64::UserCallbacks {
 public:
     explicit ARM_Dynarmic_Callbacks(ARM_Dynarmic& parent) : parent(parent) {}
@@ -28,6 +31,9 @@ public:
     u64 MemoryRead64(u64 vaddr) override {
         return Memory::Read64(vaddr);
     }
+    Vector MemoryRead128(u64 vaddr) override {
+        return {Memory::Read64(vaddr), Memory::Read64(vaddr + 8)};
+    }
 
     void MemoryWrite8(u64 vaddr, u8 value) override {
         Memory::Write8(vaddr, value);
@@ -41,6 +47,10 @@ public:
     void MemoryWrite64(u64 vaddr, u64 value) override {
         Memory::Write64(vaddr, value);
     }
+    void MemoryWrite128(u64 vaddr, Vector value) override {
+        Memory::Write64(vaddr, value[0]);
+        Memory::Write64(vaddr + 8, value[1]);
+    }
 
     void InterpreterFallback(u64 pc, size_t num_instructions) override {
         ARM_Interface::ThreadContext ctx;
@@ -52,12 +62,12 @@ public:
         num_interpreted_instructions += num_instructions;
     }
 
-    void ExceptionRaised(u64 pc, Dynarmic::A64::Exception /*exception*/) override {
-        ASSERT_MSG(false, "ExceptionRaised(%" PRIx64 ")", pc);
+    void ExceptionRaised(u64 pc, Dynarmic::A64::Exception exception) override {
+        ASSERT_MSG(false, "ExceptionRaised(exception = %zu, pc = %" PRIx64 ")",
+                   static_cast<size_t>(exception), pc);
     }
 
     void CallSVC(u32 swi) override {
-        printf("svc %x\n", swi);
         Kernel::CallSVC(swi);
     }
 
@@ -75,12 +85,24 @@ public:
     ARM_Dynarmic& parent;
     size_t ticks_remaining = 0;
     size_t num_interpreted_instructions = 0;
-    u64 tpidrr0_el0 = 0;
+    u64 tpidrro_el0 = 0;
 };
 
+std::unique_ptr<Dynarmic::A64::Jit> MakeJit(const std::unique_ptr<ARM_Dynarmic_Callbacks>& cb) {
+    const auto page_table = Kernel::g_current_process->vm_manager.page_table.pointers.data();
+
+    Dynarmic::A64::UserConfig config;
+    config.callbacks = cb.get();
+    config.tpidrro_el0 = &cb->tpidrro_el0;
+    config.dczid_el0 = 4;
+    config.page_table = reinterpret_cast<void**>(page_table);
+    config.page_table_address_space_bits = Memory::ADDRESS_SPACE_BITS;
+    config.silently_mirror_page_table = false;
+    return std::make_unique<Dynarmic::A64::Jit>(config);
+}
+
 ARM_Dynarmic::ARM_Dynarmic()
-    : cb(std::make_unique<ARM_Dynarmic_Callbacks>(*this)),
-      jit(Dynarmic::A64::UserConfig{cb.get()}) {
+    : cb(std::make_unique<ARM_Dynarmic_Callbacks>(*this)), jit(MakeJit(cb)) {
     ARM_Interface::ThreadContext ctx;
     inner_unicorn.SaveContext(ctx);
     LoadContext(ctx);
@@ -94,27 +116,27 @@ void ARM_Dynarmic::MapBackingMemory(u64 address, size_t size, u8* memory,
 }
 
 void ARM_Dynarmic::SetPC(u64 pc) {
-    jit.SetPC(pc);
+    jit->SetPC(pc);
 }
 
 u64 ARM_Dynarmic::GetPC() const {
-    return jit.GetPC();
+    return jit->GetPC();
 }
 
 u64 ARM_Dynarmic::GetReg(int index) const {
-    return jit.GetRegister(index);
+    return jit->GetRegister(index);
 }
 
 void ARM_Dynarmic::SetReg(int index, u64 value) {
-    jit.SetRegister(index, value);
+    jit->SetRegister(index, value);
 }
 
 u128 ARM_Dynarmic::GetExtReg(int index) const {
-    return jit.GetVector(index);
+    return jit->GetVector(index);
 }
 
 void ARM_Dynarmic::SetExtReg(int index, u128 value) {
-    jit.SetVector(index, value);
+    jit->SetVector(index, value);
 }
 
 u32 ARM_Dynarmic::GetVFPReg(int /*index*/) const {
@@ -127,58 +149,58 @@ void ARM_Dynarmic::SetVFPReg(int /*index*/, u32 /*value*/) {
 }
 
 u32 ARM_Dynarmic::GetCPSR() const {
-    return jit.GetPstate();
+    return jit->GetPstate();
 }
 
 void ARM_Dynarmic::SetCPSR(u32 cpsr) {
-    jit.SetPstate(cpsr);
+    jit->SetPstate(cpsr);
 }
 
 u64 ARM_Dynarmic::GetTlsAddress() const {
-    return cb->tpidrr0_el0;
+    return cb->tpidrro_el0;
 }
 
 void ARM_Dynarmic::SetTlsAddress(u64 address) {
-    cb->tpidrr0_el0 = address;
+    cb->tpidrro_el0 = address;
 }
 
 void ARM_Dynarmic::ExecuteInstructions(int num_instructions) {
     cb->ticks_remaining = num_instructions;
-    jit.Run();
+    jit->Run();
     CoreTiming::AddTicks(num_instructions - cb->num_interpreted_instructions);
     cb->num_interpreted_instructions = 0;
 }
 
 void ARM_Dynarmic::SaveContext(ARM_Interface::ThreadContext& ctx) {
-    ctx.cpu_registers = jit.GetRegisters();
-    ctx.sp = jit.GetSP();
-    ctx.pc = jit.GetPC();
-    ctx.cpsr = jit.GetPstate();
-    ctx.fpu_registers = jit.GetVectors();
-    ctx.fpscr = jit.GetFpcr();
-    ctx.tls_address = cb->tpidrr0_el0;
+    ctx.cpu_registers = jit->GetRegisters();
+    ctx.sp = jit->GetSP();
+    ctx.pc = jit->GetPC();
+    ctx.cpsr = jit->GetPstate();
+    ctx.fpu_registers = jit->GetVectors();
+    ctx.fpscr = jit->GetFpcr();
+    ctx.tls_address = cb->tpidrro_el0;
 }
 
 void ARM_Dynarmic::LoadContext(const ARM_Interface::ThreadContext& ctx) {
-    jit.SetRegisters(ctx.cpu_registers);
-    jit.SetSP(ctx.sp);
-    jit.SetPC(ctx.pc);
-    jit.SetPstate(static_cast<u32>(ctx.cpsr));
-    jit.SetVectors(ctx.fpu_registers);
-    jit.SetFpcr(static_cast<u32>(ctx.fpscr));
-    cb->tpidrr0_el0 = ctx.tls_address;
+    jit->SetRegisters(ctx.cpu_registers);
+    jit->SetSP(ctx.sp);
+    jit->SetPC(ctx.pc);
+    jit->SetPstate(static_cast<u32>(ctx.cpsr));
+    jit->SetVectors(ctx.fpu_registers);
+    jit->SetFpcr(static_cast<u32>(ctx.fpscr));
+    cb->tpidrro_el0 = ctx.tls_address;
 }
 
 void ARM_Dynarmic::PrepareReschedule() {
-    if (jit.IsExecuting()) {
-        jit.HaltExecution();
+    if (jit->IsExecuting()) {
+        jit->HaltExecution();
     }
 }
 
 void ARM_Dynarmic::ClearInstructionCache() {
-    jit.ClearCache();
+    jit->ClearCache();
 }
 
 void ARM_Dynarmic::PageTableChanged() {
-    UNIMPLEMENTED();
+    jit = MakeJit(cb);
 }

src/core/arm/dynarmic/arm_dynarmic.h

@@ -45,6 +45,6 @@ public:
 private:
     friend class ARM_Dynarmic_Callbacks;
     std::unique_ptr<ARM_Dynarmic_Callbacks> cb;
-    Dynarmic::A64::Jit jit;
+    std::unique_ptr<Dynarmic::A64::Jit> jit;
     ARM_Unicorn inner_unicorn;
 };

src/core/core.cpp

@@ -154,6 +154,8 @@ System::ResultStatus System::Init(EmuWindow* emu_window, u32 system_mode) {
         break;
     }
 
+    gpu_core = std::make_unique<Tegra::GPU>();
+
     telemetry_session = std::make_unique<Core::TelemetrySession>();
 
     CoreTiming::Init();

src/core/core.h

@@ -11,6 +11,7 @@
 #include "core/memory.h"
 #include "core/perf_stats.h"
 #include "core/telemetry_session.h"
+#include "video_core/gpu.h"
 
 class EmuWindow;
 class ARM_Interface;
@@ -102,6 +103,10 @@ public:
         return *cpu_core;
     }
 
+    Tegra::GPU& GPU() {
+        return *gpu_core;
+    }
+
     PerfStats perf_stats;
     FrameLimiter frame_limiter;
 
@@ -138,6 +143,8 @@ private:
     ///< ARM11 CPU core
     std::unique_ptr<ARM_Interface> cpu_core;
 
+    std::unique_ptr<Tegra::GPU> gpu_core;
+
     /// When true, signals that a reschedule should happen
     bool reschedule_pending{};
 

src/core/hle/kernel/mutex.cpp

@@ -70,6 +70,7 @@ ResultCode Mutex::Release(Thread* thread) {
     holding_thread->held_mutexes.erase(this);
     holding_thread->UpdatePriority();
     SetHoldingThread(nullptr);
+    SetHasWaiters(!GetWaitingThreads().empty());
     WakeupAllWaitingThreads();
     Core::System::GetInstance().PrepareReschedule();
 

src/core/hle/service/acc/acc_u0.cpp

@@ -9,6 +9,9 @@
 namespace Service {
 namespace Account {
 
+using Uid = std::array<u64, 2>;
+static constexpr Uid DEFAULT_USER_ID{0x10ull, 0x20ull};
+
 class IProfile final : public ServiceFramework<IProfile> {
 public:
     IProfile() : ServiceFramework("IProfile") {
@@ -61,6 +64,15 @@ void ACC_U0::GetUserExistence(Kernel::HLERequestContext& ctx) {
     rb.Push(true); // TODO: Check when this is supposed to return true and when not
 }
 
+void ACC_U0::ListAllUsers(Kernel::HLERequestContext& ctx) {
+    constexpr std::array<u128, 10> user_ids{DEFAULT_USER_ID};
+    const auto& output_buffer = ctx.BufferDescriptorC()[0];
+    Memory::WriteBlock(output_buffer.Address(), user_ids.data(), user_ids.size());
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_DEBUG(Service_ACC, "called");
+}
+
 void ACC_U0::GetProfile(Kernel::HLERequestContext& ctx) {
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
@@ -85,13 +97,13 @@ void ACC_U0::GetLastOpenedUser(Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service_ACC, "(STUBBED) called");
     IPC::ResponseBuilder rb{ctx, 6};
     rb.Push(RESULT_SUCCESS);
-    rb.Push<u64>(0x0);
-    rb.Push<u64>(0x0);
+    rb.PushRaw(DEFAULT_USER_ID);
 }
 
 ACC_U0::ACC_U0() : ServiceFramework("acc:u0") {
     static const FunctionInfo functions[] = {
         {1, &ACC_U0::GetUserExistence, "GetUserExistence"},
+        {2, &ACC_U0::ListAllUsers, "ListAllUsers"},
         {4, &ACC_U0::GetLastOpenedUser, "GetLastOpenedUser"},
         {5, &ACC_U0::GetProfile, "GetProfile"},
         {100, &ACC_U0::InitializeApplicationInfo, "InitializeApplicationInfo"},

src/core/hle/service/acc/acc_u0.h

@@ -28,6 +28,7 @@ public:
 
 private:
     void GetUserExistence(Kernel::HLERequestContext& ctx);
+    void ListAllUsers(Kernel::HLERequestContext& ctx);
     void GetLastOpenedUser(Kernel::HLERequestContext& ctx);
     void GetProfile(Kernel::HLERequestContext& ctx);
     void InitializeApplicationInfo(Kernel::HLERequestContext& ctx);

src/core/hle/service/am/am.cpp

@@ -45,6 +45,7 @@ ISelfController::ISelfController(std::shared_ptr<NVFlinger::NVFlinger> nvflinger
     static const FunctionInfo functions[] = {
         {1, &ISelfController::LockExit, "LockExit"},
         {2, &ISelfController::UnlockExit, "UnlockExit"},
+        {9, &ISelfController::GetLibraryAppletLaunchableEvent, "GetLibraryAppletLaunchableEvent"},
         {11, &ISelfController::SetOperationModeChangedNotification,
          "SetOperationModeChangedNotification"},
         {12, &ISelfController::SetPerformanceModeChangedNotification,
@@ -55,6 +56,9 @@ ISelfController::ISelfController(std::shared_ptr<NVFlinger::NVFlinger> nvflinger
         {40, &ISelfController::CreateManagedDisplayLayer, "CreateManagedDisplayLayer"},
     };
     RegisterHandlers(functions);
+
+    launchable_event =
+        Kernel::Event::Create(Kernel::ResetType::OneShot, "ISelfController:LaunchableEvent");
 }
 
 void ISelfController::SetFocusHandlingMode(Kernel::HLERequestContext& ctx) {
@@ -132,6 +136,16 @@ void ISelfController::UnlockExit(Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service_AM, "(STUBBED) called");
 }
 
+void ISelfController::GetLibraryAppletLaunchableEvent(Kernel::HLERequestContext& ctx) {
+    launchable_event->Signal();
+
+    IPC::ResponseBuilder rb{ctx, 2, 1};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushCopyObjects(launchable_event);
+
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+}
+
 void ISelfController::CreateManagedDisplayLayer(Kernel::HLERequestContext& ctx) {
     // TODO(Subv): Find out how AM determines the display to use, for now just create the layer
     // in the Default display.
@@ -200,7 +214,69 @@ void ICommonStateGetter::GetPerformanceMode(Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service_AM, "(STUBBED) called");
 }
 
-ILibraryAppletCreator::ILibraryAppletCreator() : ServiceFramework("ILibraryAppletCreator") {}
+class ILibraryAppletAccessor final : public ServiceFramework<ILibraryAppletAccessor> {
+public:
+    explicit ILibraryAppletAccessor() : ServiceFramework("ILibraryAppletAccessor") {
+        static const FunctionInfo functions[] = {
+            {0, &ILibraryAppletAccessor::GetAppletStateChangedEvent, "GetAppletStateChangedEvent"},
+            {1, nullptr, "IsCompleted"},
+            {10, nullptr, "Start"},
+            {20, nullptr, "RequestExit"},
+            {25, nullptr, "Terminate"},
+            {30, nullptr, "GetResult"},
+            {50, nullptr, "SetOutOfFocusApplicationSuspendingEnabled"},
+            {100, nullptr, "PushInData"},
+            {101, nullptr, "PopOutData"},
+            {102, nullptr, "PushExtraStorage"},
+            {103, nullptr, "PushInteractiveInData"},
+            {104, nullptr, "PopInteractiveOutData"},
+            {105, nullptr, "GetPopOutDataEvent"},
+            {106, nullptr, "GetPopInteractiveOutDataEvent"},
+            {120, nullptr, "NeedsToExitProcess"},
+            {120, nullptr, "GetLibraryAppletInfo"},
+            {150, nullptr, "RequestForAppletToGetForeground"},
+            {160, nullptr, "GetIndirectLayerConsumerHandle"},
+        };
+        RegisterHandlers(functions);
+
+        state_changed_event = Kernel::Event::Create(Kernel::ResetType::OneShot,
+                                                    "ILibraryAppletAccessor:StateChangedEvent");
+    }
+
+private:
+    void GetAppletStateChangedEvent(Kernel::HLERequestContext& ctx) {
+        state_changed_event->Signal();
+
+        IPC::ResponseBuilder rb{ctx, 2, 1};
+        rb.Push(RESULT_SUCCESS);
+        rb.PushCopyObjects(state_changed_event);
+
+        LOG_WARNING(Service_AM, "(STUBBED) called");
+    }
+
+    Kernel::SharedPtr<Kernel::Event> state_changed_event;
+};
+
+ILibraryAppletCreator::ILibraryAppletCreator() : ServiceFramework("ILibraryAppletCreator") {
+    static const FunctionInfo functions[] = {
+        {0, &ILibraryAppletCreator::CreateLibraryApplet, "CreateLibraryApplet"},
+        {1, nullptr, "TerminateAllLibraryApplets"},
+        {2, nullptr, "AreAnyLibraryAppletsLeft"},
+        {10, nullptr, "CreateStorage"},
+        {11, nullptr, "CreateTransferMemoryStorage"},
+        {12, nullptr, "CreateHandleStorage"},
+    };
+    RegisterHandlers(functions);
+}
+
+void ILibraryAppletCreator::CreateLibraryApplet(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+
+    rb.Push(RESULT_SUCCESS);
+    rb.PushIpcInterface<AM::ILibraryAppletAccessor>();
+
+    LOG_DEBUG(Service_AM, "called");
+}
 
 class IStorageAccessor final : public ServiceFramework<IStorageAccessor> {
 public:
@@ -270,6 +346,7 @@ private:
 IApplicationFunctions::IApplicationFunctions() : ServiceFramework("IApplicationFunctions") {
     static const FunctionInfo functions[] = {
         {1, &IApplicationFunctions::PopLaunchParameter, "PopLaunchParameter"},
+        {20, &IApplicationFunctions::EnsureSaveData, "EnsureSaveData"},
         {21, &IApplicationFunctions::GetDesiredLanguage, "GetDesiredLanguage"},
         {22, &IApplicationFunctions::SetTerminateResult, "SetTerminateResult"},
         {66, &IApplicationFunctions::InitializeGamePlayRecording, "InitializeGamePlayRecording"},
@@ -299,6 +376,12 @@ void IApplicationFunctions::PopLaunchParameter(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_AM, "called");
 }
 
+void IApplicationFunctions::EnsureSaveData(Kernel::HLERequestContext& ctx) {
+    LOG_WARNING(Service, "(STUBBED) called");
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
 void IApplicationFunctions::SetTerminateResult(Kernel::HLERequestContext& ctx) {
     // Takes an input u32 Result, no output.
     // For example, in some cases official apps use this with error 0x2A2 then uses svcBreak.

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

@@ -60,9 +60,11 @@ private:
     void SetOutOfFocusSuspendingEnabled(Kernel::HLERequestContext& ctx);
     void LockExit(Kernel::HLERequestContext& ctx);
     void UnlockExit(Kernel::HLERequestContext& ctx);
+    void GetLibraryAppletLaunchableEvent(Kernel::HLERequestContext& ctx);
     void CreateManagedDisplayLayer(Kernel::HLERequestContext& ctx);
 
     std::shared_ptr<NVFlinger::NVFlinger> nvflinger;
+    Kernel::SharedPtr<Kernel::Event> launchable_event;
 };
 
 class ICommonStateGetter final : public ServiceFramework<ICommonStateGetter> {
@@ -92,6 +94,9 @@ private:
 class ILibraryAppletCreator final : public ServiceFramework<ILibraryAppletCreator> {
 public:
     ILibraryAppletCreator();
+
+private:
+    void CreateLibraryApplet(Kernel::HLERequestContext& ctx);
 };
 
 class IApplicationFunctions final : public ServiceFramework<IApplicationFunctions> {
@@ -100,6 +105,7 @@ public:
 
 private:
     void PopLaunchParameter(Kernel::HLERequestContext& ctx);
+    void EnsureSaveData(Kernel::HLERequestContext& ctx);
     void SetTerminateResult(Kernel::HLERequestContext& ctx);
     void GetDesiredLanguage(Kernel::HLERequestContext& ctx);
     void InitializeGamePlayRecording(Kernel::HLERequestContext& ctx);

src/core/hle/service/apm/apm.cpp

@@ -5,63 +5,15 @@
 #include "common/logging/log.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/service/apm/apm.h"
+#include "core/hle/service/apm/interface.h"
 
 namespace Service {
 namespace APM {
 
 void InstallInterfaces(SM::ServiceManager& service_manager) {
-    std::make_shared<APM>()->InstallAsService(service_manager);
-}
-
-class ISession final : public ServiceFramework<ISession> {
-public:
-    ISession() : ServiceFramework("ISession") {
-        static const FunctionInfo functions[] = {
-            {0, &ISession::SetPerformanceConfiguration, "SetPerformanceConfiguration"},
-            {1, &ISession::GetPerformanceConfiguration, "GetPerformanceConfiguration"},
-        };
-        RegisterHandlers(functions);
-    }
-
-private:
-    void SetPerformanceConfiguration(Kernel::HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-
-        auto mode = static_cast<PerformanceMode>(rp.Pop<u32>());
-        u32 config = rp.Pop<u32>();
-
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(RESULT_SUCCESS);
-
-        LOG_WARNING(Service_APM, "(STUBBED) called mode=%u config=%u", static_cast<u32>(mode),
-                    config);
-    }
-
-    void GetPerformanceConfiguration(Kernel::HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-
-        auto mode = static_cast<PerformanceMode>(rp.Pop<u32>());
-
-        IPC::ResponseBuilder rb{ctx, 3};
-        rb.Push(RESULT_SUCCESS);
-        rb.Push<u32>(0); // Performance configuration
-
-        LOG_WARNING(Service_APM, "(STUBBED) called mode=%u", static_cast<u32>(mode));
-    }
-};
-
-APM::APM() : ServiceFramework("apm") {
-    static const FunctionInfo functions[] = {
-        {0x00000000, &APM::OpenSession, "OpenSession"},
-        {0x00000001, nullptr, "GetPerformanceMode"},
-    };
-    RegisterHandlers(functions);
-}
-
-void APM::OpenSession(Kernel::HLERequestContext& ctx) {
-    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-    rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISession>();
+    auto module_ = std::make_shared<Module>();
+    std::make_shared<APM>(module_, "apm")->InstallAsService(service_manager);
+    std::make_shared<APM>(module_, "apm:p")->InstallAsService(service_manager);
 }
 
 } // namespace APM

src/core/hle/service/apm/apm.h

@@ -14,13 +14,10 @@ enum class PerformanceMode : u8 {
     Docked = 1,
 };
 
-class APM final : public ServiceFramework<APM> {
+class Module final {
 public:
-    APM();
-    ~APM() = default;
-
-private:
-    void OpenSession(Kernel::HLERequestContext& ctx);
+    Module() = default;
+    ~Module() = default;
 };
 
 /// Registers all AM services with the specified service manager.

src/core/hle/service/apm/interface.cpp

@@ -0,0 +1,66 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/logging/log.h"
+#include "core/hle/ipc_helpers.h"
+#include "core/hle/service/apm/apm.h"
+#include "core/hle/service/apm/interface.h"
+
+namespace Service {
+namespace APM {
+
+class ISession final : public ServiceFramework<ISession> {
+public:
+    ISession() : ServiceFramework("ISession") {
+        static const FunctionInfo functions[] = {
+            {0, &ISession::SetPerformanceConfiguration, "SetPerformanceConfiguration"},
+            {1, &ISession::GetPerformanceConfiguration, "GetPerformanceConfiguration"},
+        };
+        RegisterHandlers(functions);
+    }
+
+private:
+    void SetPerformanceConfiguration(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+
+        auto mode = static_cast<PerformanceMode>(rp.Pop<u32>());
+        u32 config = rp.Pop<u32>();
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+
+        LOG_WARNING(Service_APM, "(STUBBED) called mode=%u config=%u", static_cast<u32>(mode),
+                    config);
+    }
+
+    void GetPerformanceConfiguration(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+
+        auto mode = static_cast<PerformanceMode>(rp.Pop<u32>());
+
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push<u32>(0); // Performance configuration
+
+        LOG_WARNING(Service_APM, "(STUBBED) called mode=%u", static_cast<u32>(mode));
+    }
+};
+
+APM::APM(std::shared_ptr<Module> apm, const char* name)
+    : ServiceFramework(name), apm(std::move(apm)) {
+    static const FunctionInfo functions[] = {
+        {0, &APM::OpenSession, "OpenSession"},
+        {1, nullptr, "GetPerformanceMode"},
+    };
+    RegisterHandlers(functions);
+}
+
+void APM::OpenSession(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushIpcInterface<ISession>();
+}
+
+} // namespace APM
+} // namespace Service

src/core/hle/service/apm/interface.h

@@ -0,0 +1,27 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/service/service.h"
+
+namespace Service {
+namespace APM {
+
+class APM final : public ServiceFramework<APM> {
+public:
+    APM(std::shared_ptr<Module> apm, const char* name);
+    ~APM() = default;
+
+private:
+    void OpenSession(Kernel::HLERequestContext& ctx);
+
+    std::shared_ptr<Module> apm;
+};
+
+/// Registers all AM services with the specified service manager.
+void InstallInterfaces(SM::ServiceManager& service_manager);
+
+} // namespace APM
+} // namespace Service

src/core/hle/service/audio/audren_u.cpp

@@ -4,6 +4,7 @@
 
 #include "common/logging/log.h"
 #include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/event.h"
 #include "core/hle/kernel/hle_ipc.h"
 #include "core/hle/service/audio/audren_u.h"
 
@@ -18,27 +19,143 @@ public:
             {0x1, nullptr, "GetAudioRendererSampleCount"},
             {0x2, nullptr, "GetAudioRendererMixBufferCount"},
             {0x3, nullptr, "GetAudioRendererState"},
-            {0x4, nullptr, "RequestUpdateAudioRenderer"},
-            {0x5, nullptr, "StartAudioRenderer"},
-            {0x6, nullptr, "StopAudioRenderer"},
-            {0x7, nullptr, "QuerySystemEvent"},
+            {0x4, &IAudioRenderer::RequestUpdateAudioRenderer, "RequestUpdateAudioRenderer"},
+            {0x5, &IAudioRenderer::StartAudioRenderer, "StartAudioRenderer"},
+            {0x6, &IAudioRenderer::StopAudioRenderer, "StopAudioRenderer"},
+            {0x7, &IAudioRenderer::QuerySystemEvent, "QuerySystemEvent"},
             {0x8, nullptr, "SetAudioRendererRenderingTimeLimit"},
             {0x9, nullptr, "GetAudioRendererRenderingTimeLimit"},
         };
         RegisterHandlers(functions);
+
+        system_event =
+            Kernel::Event::Create(Kernel::ResetType::OneShot, "IAudioRenderer:SystemEvent");
     }
     ~IAudioRenderer() = default;
+
+private:
+    void RequestUpdateAudioRenderer(Kernel::HLERequestContext& ctx) {
+        AudioRendererResponseData response_data = {0};
+
+        response_data.section_0_size =
+            response_data.state_entries.size() * sizeof(AudioRendererStateEntry);
+        response_data.section_1_size = response_data.section_1.size();
+        response_data.section_2_size = response_data.section_2.size();
+        response_data.section_3_size = response_data.section_3.size();
+        response_data.section_4_size = response_data.section_4.size();
+        response_data.section_5_size = response_data.section_5.size();
+        response_data.total_size = sizeof(AudioRendererResponseData);
+
+        for (unsigned i = 0; i < response_data.state_entries.size(); i++) {
+            // 4 = Busy and 5 = Ready?
+            response_data.state_entries[i].state = 5;
+        }
+
+        auto& buffer = ctx.BufferDescriptorB()[0];
+
+        Memory::WriteBlock(buffer.Address(), &response_data, response_data.total_size);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+
+        rb.Push(RESULT_SUCCESS);
+
+        LOG_WARNING(Service_Audio, "(STUBBED) called");
+    }
+
+    void StartAudioRenderer(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2};
+
+        rb.Push(RESULT_SUCCESS);
+
+        LOG_WARNING(Service_Audio, "(STUBBED) called");
+    }
+
+    void StopAudioRenderer(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2};
+
+        rb.Push(RESULT_SUCCESS);
+
+        LOG_WARNING(Service_Audio, "(STUBBED) called");
+    }
+
+    void QuerySystemEvent(Kernel::HLERequestContext& ctx) {
+        // system_event->Signal();
+
+        IPC::ResponseBuilder rb{ctx, 2, 1};
+        rb.Push(RESULT_SUCCESS);
+        rb.PushCopyObjects(system_event);
+
+        LOG_WARNING(Service_Audio, "(STUBBED) called");
+    }
+
+    struct AudioRendererStateEntry {
+        u32_le state;
+        u32_le unknown_4;
+        u32_le unknown_8;
+        u32_le unknown_c;
+    };
+    static_assert(sizeof(AudioRendererStateEntry) == 0x10,
+                  "AudioRendererStateEntry has wrong size");
+
+    struct AudioRendererResponseData {
+        u32_le unknown_0;
+        u32_le section_5_size;
+        u32_le section_0_size;
+        u32_le section_1_size;
+        u32_le unknown_10;
+        u32_le section_2_size;
+        u32_le unknown_18;
+        u32_le section_3_size;
+        u32_le section_4_size;
+        u32_le unknown_24;
+        u32_le unknown_28;
+        u32_le unknown_2c;
+        u32_le unknown_30;
+        u32_le unknown_34;
+        u32_le unknown_38;
+        u32_le total_size;
+
+        std::array<AudioRendererStateEntry, 0x18e> state_entries;
+
+        std::array<u8, 0x600> section_1;
+        std::array<u8, 0xe0> section_2;
+        std::array<u8, 0x20> section_3;
+        std::array<u8, 0x10> section_4;
+        std::array<u8, 0xb0> section_5;
+    };
+    static_assert(sizeof(AudioRendererResponseData) == 0x20e0,
+                  "AudioRendererResponseData has wrong size");
+
+    Kernel::SharedPtr<Kernel::Event> system_event;
 };
 
 AudRenU::AudRenU() : ServiceFramework("audren:u") {
     static const FunctionInfo functions[] = {
-        {0x00000000, nullptr, "OpenAudioRenderer"},
-        {0x00000001, nullptr, "GetAudioRendererWorkBufferSize"},
+        {0x00000000, &AudRenU::OpenAudioRenderer, "OpenAudioRenderer"},
+        {0x00000001, &AudRenU::GetAudioRendererWorkBufferSize, "GetAudioRendererWorkBufferSize"},
         {0x00000002, nullptr, "GetAudioRenderersProcessMasterVolume"},
         {0x00000003, nullptr, "SetAudioRenderersProcessMasterVolume"},
     };
     RegisterHandlers(functions);
 }
 
+void AudRenU::OpenAudioRenderer(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+
+    rb.Push(RESULT_SUCCESS);
+    rb.PushIpcInterface<Audio::IAudioRenderer>();
+
+    LOG_DEBUG(Service_Audio, "called");
+}
+
+void AudRenU::GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 4};
+
+    rb.Push(RESULT_SUCCESS);
+    rb.Push<u64>(0x1000);
+
+    LOG_WARNING(Service_Audio, "called");
+}
+
 } // namespace Audio
 } // namespace Service

src/core/hle/service/audio/audren_u.h

@@ -17,6 +17,10 @@ class AudRenU final : public ServiceFramework<AudRenU> {
 public:
     explicit AudRenU();
     ~AudRenU() = default;
+
+private:
+    void OpenAudioRenderer(Kernel::HLERequestContext& ctx);
+    void GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx);
 };
 
 } // namespace Audio

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

@@ -70,7 +70,9 @@ private:
 FSP_SRV::FSP_SRV() : ServiceFramework("fsp-srv") {
     static const FunctionInfo functions[] = {
         {1, &FSP_SRV::Initalize, "Initalize"},
+        {18, &FSP_SRV::MountSdCard, "MountSdCard"},
         {200, &FSP_SRV::OpenDataStorageByCurrentProcess, "OpenDataStorageByCurrentProcess"},
+        {202, nullptr, "OpenDataStorageByDataId"},
         {203, &FSP_SRV::OpenRomStorage, "OpenRomStorage"},
         {1005, &FSP_SRV::GetGlobalAccessLogMode, "GetGlobalAccessLogMode"},
     };
@@ -95,6 +97,13 @@ void FSP_SRV::Initalize(Kernel::HLERequestContext& ctx) {
     rb.Push(RESULT_SUCCESS);
 }
 
+void FSP_SRV::MountSdCard(Kernel::HLERequestContext& ctx) {
+    LOG_WARNING(Service_FS, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
 void FSP_SRV::GetGlobalAccessLogMode(Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service_FS, "(STUBBED) called");
 

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

@@ -23,6 +23,7 @@ private:
     void TryLoadRomFS();
 
     void Initalize(Kernel::HLERequestContext& ctx);
+    void MountSdCard(Kernel::HLERequestContext& ctx);
     void GetGlobalAccessLogMode(Kernel::HLERequestContext& ctx);
     void OpenDataStorageByCurrentProcess(Kernel::HLERequestContext& ctx);
     void OpenRomStorage(Kernel::HLERequestContext& ctx);

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

@@ -9,6 +9,7 @@
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/client_session.h"
+#include "core/hle/kernel/event.h"
 #include "core/hle/kernel/shared_memory.h"
 #include "core/hle/service/hid/hid.h"
 #include "core/hle/service/service.h"
@@ -174,21 +175,29 @@ public:
         static const FunctionInfo functions[] = {
             {0, &Hid::CreateAppletResource, "CreateAppletResource"},
             {1, &Hid::ActivateDebugPad, "ActivateDebugPad"},
-            {11, nullptr, "ActivateTouchScreen"},
+            {11, &Hid::ActivateTouchScreen, "ActivateTouchScreen"},
             {66, &Hid::StartSixAxisSensor, "StartSixAxisSensor"},
             {100, &Hid::SetSupportedNpadStyleSet, "SetSupportedNpadStyleSet"},
             {102, &Hid::SetSupportedNpadIdType, "SetSupportedNpadIdType"},
             {103, &Hid::ActivateNpad, "ActivateNpad"},
-            {120, nullptr, "SetNpadJoyHoldType"},
+            {106, &Hid::AcquireNpadStyleSetUpdateEventHandle,
+             "AcquireNpadStyleSetUpdateEventHandle"},
+            {120, &Hid::SetNpadJoyHoldType, "SetNpadJoyHoldType"},
+            {121, &Hid::GetNpadJoyHoldType, "GetNpadJoyHoldType"},
             {124, nullptr, "SetNpadJoyAssignmentModeDual"},
+            {128, &Hid::SetNpadHandheldActivationMode, "SetNpadHandheldActivationMode"},
             {203, &Hid::CreateActiveVibrationDeviceList, "CreateActiveVibrationDeviceList"},
         };
         RegisterHandlers(functions);
+
+        event = Kernel::Event::Create(Kernel::ResetType::OneShot, "hid:EventHandle");
     }
     ~Hid() = default;
 
 private:
     std::shared_ptr<IAppletResource> applet_resource;
+    u32 joy_hold_type{0};
+    Kernel::SharedPtr<Kernel::Event> event;
 
     void CreateAppletResource(Kernel::HLERequestContext& ctx) {
         if (applet_resource == nullptr) {
@@ -207,6 +216,12 @@ private:
         LOG_WARNING(Service_HID, "(STUBBED) called");
     }
 
+    void ActivateTouchScreen(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+        LOG_WARNING(Service_HID, "(STUBBED) called");
+    }
+
     void StartSixAxisSensor(Kernel::HLERequestContext& ctx) {
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);
@@ -231,6 +246,32 @@ private:
         LOG_WARNING(Service_HID, "(STUBBED) called");
     }
 
+    void AcquireNpadStyleSetUpdateEventHandle(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2, 1};
+        rb.Push(RESULT_SUCCESS);
+        rb.PushCopyObjects(event);
+        LOG_WARNING(Service_HID, "(STUBBED) called");
+    }
+
+    void SetNpadJoyHoldType(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+        LOG_WARNING(Service_HID, "(STUBBED) called");
+    }
+
+    void GetNpadJoyHoldType(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push(joy_hold_type);
+        LOG_WARNING(Service_HID, "(STUBBED) called");
+    }
+
+    void SetNpadHandheldActivationMode(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+        LOG_WARNING(Service_HID, "(STUBBED) called");
+    }
+
     void CreateActiveVibrationDeviceList(Kernel::HLERequestContext& ctx) {
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);

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

@@ -5,7 +5,9 @@
 #pragma once
 
 #include <vector>
+#include "common/bit_field.h"
 #include "common/common_types.h"
+#include "common/swap.h"
 
 namespace Service {
 namespace Nvidia {
@@ -17,6 +19,14 @@ class nvdevice {
 public:
     nvdevice() = default;
     virtual ~nvdevice() = default;
+    union Ioctl {
+        u32_le raw;
+        BitField<0, 8, u32_le> cmd;
+        BitField<8, 8, u32_le> group;
+        BitField<16, 14, u32_le> length;
+        BitField<30, 1, u32_le> is_in;
+        BitField<31, 1, u32_le> is_out;
+    };
 
     /**
      * Handles an ioctl request.
@@ -25,7 +35,7 @@ public:
      * @param output A buffer where the output data will be written to.
      * @returns The result code of the ioctl.
      */
-    virtual u32 ioctl(u32 command, const std::vector<u8>& input, std::vector<u8>& output) = 0;
+    virtual u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) = 0;
 };
 
 } // namespace Devices

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

@@ -14,21 +14,23 @@ namespace Service {
 namespace Nvidia {
 namespace Devices {
 
-u32 nvdisp_disp0::ioctl(u32 command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvdisp_disp0::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
     UNIMPLEMENTED();
     return 0;
 }
 
 void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u32 height,
-                        u32 stride) {
+                        u32 stride, NVFlinger::BufferQueue::BufferTransformFlags transform) {
     VAddr addr = nvmap_dev->GetObjectAddress(buffer_handle);
     LOG_WARNING(Service,
                 "Drawing from address %llx offset %08X Width %u Height %u Stride %u Format %u",
                 addr, offset, width, height, stride, format);
 
     using PixelFormat = RendererBase::FramebufferInfo::PixelFormat;
+    using Flags = NVFlinger::BufferQueue::BufferTransformFlags;
+    const bool flip_vertical = static_cast<u32>(transform) & static_cast<u32>(Flags::FlipV);
     const RendererBase::FramebufferInfo framebuffer_info{
-        addr, offset, width, height, stride, static_cast<PixelFormat>(format)};
+        addr, offset, width, height, stride, static_cast<PixelFormat>(format), flip_vertical};
 
     Core::System::GetInstance().perf_stats.EndGameFrame();
     VideoCore::g_renderer->SwapBuffers(framebuffer_info);

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

@@ -8,6 +8,7 @@
 #include <vector>
 #include "common/common_types.h"
 #include "core/hle/service/nvdrv/devices/nvdevice.h"
+#include "core/hle/service/nvflinger/buffer_queue.h"
 
 namespace Service {
 namespace Nvidia {
@@ -20,10 +21,11 @@ public:
     nvdisp_disp0(std::shared_ptr<nvmap> nvmap_dev) : nvdevice(), nvmap_dev(std::move(nvmap_dev)) {}
     ~nvdisp_disp0() = default;
 
-    u32 ioctl(u32 command, const std::vector<u8>& input, std::vector<u8>& output) override;
+    u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
 
     /// Performs a screen flip, drawing the buffer pointed to by the handle.
-    void flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u32 height, u32 stride);
+    void flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u32 height, u32 stride,
+              NVFlinger::BufferQueue::BufferTransformFlags transform);
 
 private:
     std::shared_ptr<nvmap> nvmap_dev;

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

@@ -4,14 +4,112 @@
 
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/hle/service/nvdrv/devices/nvhost_as_gpu.h"
+#include "core/hle/service/nvdrv/devices/nvmap.h"
 
 namespace Service {
 namespace Nvidia {
 namespace Devices {
 
-u32 nvhost_as_gpu::ioctl(u32 command, const std::vector<u8>& input, std::vector<u8>& output) {
-    UNIMPLEMENTED();
+u32 nvhost_as_gpu::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+    LOG_DEBUG(Service_NVDRV, "called, command=0x%08x, input_size=0x%llx, output_size=0x%llx",
+              command, input.size(), output.size());
+
+    switch (static_cast<IoctlCommand>(command.raw)) {
+    case IoctlCommand::IocInitalizeExCommand:
+        return InitalizeEx(input, output);
+    case IoctlCommand::IocAllocateSpaceCommand:
+        return AllocateSpace(input, output);
+    case IoctlCommand::IocMapBufferExCommand:
+        return MapBufferEx(input, output);
+    case IoctlCommand::IocBindChannelCommand:
+        return BindChannel(input, output);
+    case IoctlCommand::IocGetVaRegionsCommand:
+        return GetVARegions(input, output);
+    }
+    return 0;
+}
+
+u32 nvhost_as_gpu::InitalizeEx(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlInitalizeEx params{};
+    std::memcpy(&params, input.data(), input.size());
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called, big_page_size=0x%x", params.big_page_size);
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_as_gpu::AllocateSpace(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlAllocSpace params{};
+    std::memcpy(&params, input.data(), input.size());
+    LOG_DEBUG(Service_NVDRV, "called, pages=%x, page_size=%x, flags=%x", params.pages,
+              params.page_size, params.flags);
+
+    auto& gpu = Core::System::GetInstance().GPU();
+    const u64 size{static_cast<u64>(params.pages) * static_cast<u64>(params.page_size)};
+    if (params.flags & 1) {
+        params.offset = gpu.memory_manager->AllocateSpace(params.offset, size, 1);
+    } else {
+        params.offset = gpu.memory_manager->AllocateSpace(size, params.align);
+    }
+
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_as_gpu::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlMapBufferEx params{};
+    std::memcpy(&params, input.data(), input.size());
+
+    LOG_DEBUG(Service_NVDRV,
+              "called, flags=%x, nvmap_handle=%x, buffer_offset=%lx, mapping_size=%lx, offset=%lx",
+              params.flags, params.nvmap_handle, params.buffer_offset, params.mapping_size,
+              params.offset);
+
+    if (!params.nvmap_handle) {
+        return 0;
+    }
+
+    auto object = nvmap_dev->GetObject(params.nvmap_handle);
+    ASSERT(object);
+
+    auto& gpu = Core::System::GetInstance().GPU();
+
+    if (params.flags & 1) {
+        params.offset = gpu.memory_manager->MapBufferEx(object->addr, params.offset, object->size);
+    } else {
+        params.offset = gpu.memory_manager->MapBufferEx(object->addr, object->size);
+    }
+
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_as_gpu::BindChannel(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlBindChannel params{};
+    std::memcpy(&params, input.data(), input.size());
+    LOG_DEBUG(Service_NVDRV, "called, fd=%x", params.fd);
+    channel = params.fd;
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_as_gpu::GetVARegions(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlGetVaRegions params{};
+    std::memcpy(&params, input.data(), input.size());
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called, buf_addr=%lx, buf_size=%x", params.buf_addr,
+                params.buf_size);
+
+    params.buf_size = 0x30;
+    params.regions[0].offset = 0x04000000;
+    params.regions[0].page_size = 0x1000;
+    params.regions[0].pages = 0x3fbfff;
+
+    params.regions[1].offset = 0x04000000;
+    params.regions[1].page_size = 0x10000;
+    params.regions[1].pages = 0x1bffff;
+    // TODO(ogniK): This probably can stay stubbed but should add support way way later
+    std::memcpy(output.data(), &params, output.size());
     return 0;
 }
 

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

@@ -4,20 +4,100 @@
 
 #pragma once
 
+#include <memory>
+#include <utility>
 #include <vector>
 #include "common/common_types.h"
+#include "common/swap.h"
 #include "core/hle/service/nvdrv/devices/nvdevice.h"
 
 namespace Service {
 namespace Nvidia {
 namespace Devices {
 
+class nvmap;
+
 class nvhost_as_gpu final : public nvdevice {
 public:
-    nvhost_as_gpu() = default;
+    nvhost_as_gpu(std::shared_ptr<nvmap> nvmap_dev) : nvmap_dev(std::move(nvmap_dev)) {}
     ~nvhost_as_gpu() override = default;
 
-    u32 ioctl(u32 command, const std::vector<u8>& input, std::vector<u8>& output) override;
+    u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
+
+private:
+    enum class IoctlCommand : u32_le {
+        IocInitalizeExCommand = 0x40284109,
+        IocAllocateSpaceCommand = 0xC0184102,
+        IocMapBufferExCommand = 0xC0284106,
+        IocBindChannelCommand = 0x40044101,
+        IocGetVaRegionsCommand = 0xC0404108,
+    };
+
+    struct IoctlInitalizeEx {
+        u32_le big_page_size; // depends on GPU's available_big_page_sizes; 0=default
+        s32_le as_fd;         // ignored; passes 0
+        u32_le flags;         // passes 0
+        u32_le reserved;      // ignored; passes 0
+        u64_le unk0;
+        u64_le unk1;
+        u64_le unk2;
+    };
+    static_assert(sizeof(IoctlInitalizeEx) == 40, "IoctlInitalizeEx is incorrect size");
+
+    struct IoctlAllocSpace {
+        u32_le pages;
+        u32_le page_size;
+        u32_le flags;
+        INSERT_PADDING_WORDS(1);
+        union {
+            u64_le offset;
+            u64_le align;
+        };
+    };
+    static_assert(sizeof(IoctlAllocSpace) == 24, "IoctlInitalizeEx is incorrect size");
+
+    struct IoctlMapBufferEx {
+        u32_le flags; // bit0: fixed_offset, bit2: cacheable
+        u32_le kind;  // -1 is default
+        u32_le nvmap_handle;
+        u32_le page_size; // 0 means don't care
+        u64_le buffer_offset;
+        u64_le mapping_size;
+        u64_le offset;
+    };
+    static_assert(sizeof(IoctlMapBufferEx) == 40, "IoctlMapBufferEx is incorrect size");
+
+    struct IoctlBindChannel {
+        u32_le fd;
+    };
+    static_assert(sizeof(IoctlBindChannel) == 4, "IoctlBindChannel is incorrect size");
+
+    struct IoctlVaRegion {
+        u64_le offset;
+        u32_le page_size;
+        INSERT_PADDING_WORDS(1);
+        u64_le pages;
+    };
+    static_assert(sizeof(IoctlVaRegion) == 24, "IoctlVaRegion is incorrect size");
+
+    struct IoctlGetVaRegions {
+        u64_le buf_addr; // (contained output user ptr on linux, ignored)
+        u32_le buf_size; // forced to 2*sizeof(struct va_region)
+        u32_le reserved;
+        IoctlVaRegion regions[2];
+    };
+    static_assert(sizeof(IoctlGetVaRegions) == 16 + sizeof(IoctlVaRegion) * 2,
+                  "IoctlGetVaRegions is incorrect size");
+
+    u32 channel{};
+
+    u32 InitalizeEx(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 AllocateSpace(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 BindChannel(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetVARegions(const std::vector<u8>& input, std::vector<u8>& output);
+
+    std::shared_ptr<nvmap> nvmap_dev;
 };
 
 } // namespace Devices

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

@@ -10,12 +10,12 @@ namespace Service {
 namespace Nvidia {
 namespace Devices {
 
-u32 nvhost_ctrl::ioctl(u32 command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvhost_ctrl::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
     LOG_DEBUG(Service_NVDRV, "called, command=0x%08x, input_size=0x%lx, output_size=0x%lx", command,
               input.size(), output.size());
 
-    switch (command) {
-    case IocGetConfigCommand:
+    switch (static_cast<IoctlCommand>(command.raw)) {
+    case IoctlCommand::IocGetConfigCommand:
         return NvOsGetConfigU32(input, output);
     }
     UNIMPLEMENTED();
@@ -23,19 +23,23 @@ u32 nvhost_ctrl::ioctl(u32 command, const std::vector<u8>& input, std::vector<u8
 }
 
 u32 nvhost_ctrl::NvOsGetConfigU32(const std::vector<u8>& input, std::vector<u8>& output) {
-    IocGetConfigParams params;
+    IocGetConfigParams params{};
     std::memcpy(&params, input.data(), sizeof(params));
     LOG_DEBUG(Service_NVDRV, "called, setting=%s!%s", params.domain_str.data(),
               params.param_str.data());
 
     if (!strcmp(params.domain_str.data(), "nv")) {
         if (!strcmp(params.param_str.data(), "NV_MEMORY_PROFILER")) {
-            params.config_str[0] = '1';
+            params.config_str[0] = '0';
+        } else if (!strcmp(params.param_str.data(), "NVN_THROUGH_OPENGL")) {
+            params.config_str[0] = '0';
+        } else if (!strcmp(params.param_str.data(), "NVRM_GPU_PREVENT_USE")) {
+            params.config_str[0] = '0';
         } else {
-            UNIMPLEMENTED();
+            params.config_str[0] = '0';
         }
     } else {
-        UNIMPLEMENTED();
+        UNIMPLEMENTED(); // unknown domain? Only nv has been seen so far on hardware
     }
     std::memcpy(output.data(), &params, sizeof(params));
     return 0;

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

@@ -20,10 +20,10 @@ public:
     nvhost_ctrl() = default;
     ~nvhost_ctrl() override = default;
 
-    u32 ioctl(u32 command, const std::vector<u8>& input, std::vector<u8>& output) override;
+    u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
 
 private:
-    enum IoctlCommands {
+    enum class IoctlCommand : u32_le {
         IocSyncptReadCommand = 0xC0080014,
         IocSyncptIncrCommand = 0x40040015,
         IocSyncptWaitCommand = 0xC00C0016,
@@ -39,6 +39,7 @@ private:
         std::array<char, 0x41> param_str;
         std::array<char, 0x101> config_str;
     };
+    static_assert(sizeof(IocGetConfigParams) == 387, "IocGetConfigParams is incorrect size");
 
     u32 NvOsGetConfigU32(const std::vector<u8>& input, std::vector<u8>& output);
 };

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

@@ -0,0 +1,126 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h"
+
+namespace Service {
+namespace Nvidia {
+namespace Devices {
+
+u32 nvhost_ctrl_gpu::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+    LOG_DEBUG(Service_NVDRV, "called, command=0x%08x, input_size=0x%llx, output_size=0x%llx",
+              command, input.size(), output.size());
+
+    switch (static_cast<IoctlCommand>(command.raw)) {
+    case IoctlCommand::IocGetCharacteristicsCommand:
+        return GetCharacteristics(input, output);
+    case IoctlCommand::IocGetTPCMasksCommand:
+        return GetTPCMasks(input, output);
+    case IoctlCommand::IocGetActiveSlotMaskCommand:
+        return GetActiveSlotMask(input, output);
+    case IoctlCommand::IocZcullGetCtxSizeCommand:
+        return ZCullGetCtxSize(input, output);
+    case IoctlCommand::IocZcullGetInfo:
+        return ZCullGetInfo(input, output);
+    }
+    UNIMPLEMENTED();
+    return 0;
+}
+
+u32 nvhost_ctrl_gpu::GetCharacteristics(const std::vector<u8>& input, std::vector<u8>& output) {
+    LOG_DEBUG(Service_NVDRV, "called");
+    IoctlCharacteristics params{};
+    std::memcpy(&params, input.data(), input.size());
+    params.gc.arch = 0x120;
+    params.gc.impl = 0xb;
+    params.gc.rev = 0xa1;
+    params.gc.num_gpc = 0x1;
+    params.gc.l2_cache_size = 0x40000;
+    params.gc.on_board_video_memory_size = 0x0;
+    params.gc.num_tpc_per_gpc = 0x2;
+    params.gc.bus_type = 0x20;
+    params.gc.big_page_size = 0x20000;
+    params.gc.compression_page_size = 0x20000;
+    params.gc.pde_coverage_bit_count = 0x1B;
+    params.gc.available_big_page_sizes = 0x30000;
+    params.gc.gpc_mask = 0x1;
+    params.gc.sm_arch_sm_version = 0x503;
+    params.gc.sm_arch_spa_version = 0x503;
+    params.gc.sm_arch_warp_count = 0x80;
+    params.gc.gpu_va_bit_count = 0x28;
+    params.gc.reserved = 0x0;
+    params.gc.flags = 0x55;
+    params.gc.twod_class = 0x902D;
+    params.gc.threed_class = 0xB197;
+    params.gc.compute_class = 0xB1C0;
+    params.gc.gpfifo_class = 0xB06F;
+    params.gc.inline_to_memory_class = 0xA140;
+    params.gc.dma_copy_class = 0xB0B5;
+    params.gc.max_fbps_count = 0x1;
+    params.gc.fbp_en_mask = 0x0;
+    params.gc.max_ltc_per_fbp = 0x2;
+    params.gc.max_lts_per_ltc = 0x1;
+    params.gc.max_tex_per_tpc = 0x0;
+    params.gc.max_gpc_count = 0x1;
+    params.gc.rop_l2_en_mask_0 = 0x21D70;
+    params.gc.rop_l2_en_mask_1 = 0x0;
+    params.gc.chipname = 0x6230326D67;
+    params.gc.gr_compbit_store_base_hw = 0x0;
+    params.gpu_characteristics_buf_size = 0xA0;
+    params.gpu_characteristics_buf_addr = 0xdeadbeef; // Cannot be 0 (UNUSED)
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_ctrl_gpu::GetTPCMasks(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlGpuGetTpcMasksArgs params{};
+    std::memcpy(&params, input.data(), input.size());
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called, mask=0x%x, mask_buf_addr=0x%lx",
+                params.mask_buf_size, params.mask_buf_addr);
+    std::memcpy(output.data(), &params, sizeof(params));
+    return 0;
+}
+
+u32 nvhost_ctrl_gpu::GetActiveSlotMask(const std::vector<u8>& input, std::vector<u8>& output) {
+    LOG_DEBUG(Service_NVDRV, "called");
+    IoctlActiveSlotMask params{};
+    std::memcpy(&params, input.data(), input.size());
+    params.slot = 0x07;
+    params.mask = 0x01;
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_ctrl_gpu::ZCullGetCtxSize(const std::vector<u8>& input, std::vector<u8>& output) {
+    LOG_DEBUG(Service_NVDRV, "called");
+    IoctlZcullGetCtxSize params{};
+    std::memcpy(&params, input.data(), input.size());
+    params.size = 0x1;
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_ctrl_gpu::ZCullGetInfo(const std::vector<u8>& input, std::vector<u8>& output) {
+    LOG_DEBUG(Service_NVDRV, "called");
+    IoctlNvgpuGpuZcullGetInfoArgs params{};
+    std::memcpy(&params, input.data(), input.size());
+    params.width_align_pixels = 0x20;
+    params.height_align_pixels = 0x20;
+    params.pixel_squares_by_aliquots = 0x400;
+    params.aliquot_total = 0x800;
+    params.region_byte_multiplier = 0x20;
+    params.region_header_size = 0x20;
+    params.subregion_header_size = 0xc0;
+    params.subregion_width_align_pixels = 0x20;
+    params.subregion_height_align_pixels = 0x40;
+    params.subregion_count = 0x10;
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+} // namespace Devices
+} // namespace Nvidia
+} // namespace Service

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

@@ -0,0 +1,130 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <vector>
+#include "common/common_types.h"
+#include "common/swap.h"
+#include "core/hle/service/nvdrv/devices/nvdevice.h"
+
+namespace Service {
+namespace Nvidia {
+namespace Devices {
+
+class nvhost_ctrl_gpu final : public nvdevice {
+public:
+    nvhost_ctrl_gpu() = default;
+    ~nvhost_ctrl_gpu() override = default;
+
+    u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
+
+private:
+    enum class IoctlCommand : u32_le {
+        IocGetCharacteristicsCommand = 0xC0B04705,
+        IocGetTPCMasksCommand = 0xC0184706,
+        IocGetActiveSlotMaskCommand = 0x80084714,
+        IocZcullGetCtxSizeCommand = 0x80044701,
+        IocZcullGetInfo = 0x80284702,
+    };
+
+    struct IoctlGpuCharacteristics {
+        u32_le arch;                       // 0x120 (NVGPU_GPU_ARCH_GM200)
+        u32_le impl;                       // 0xB (NVGPU_GPU_IMPL_GM20B)
+        u32_le rev;                        // 0xA1 (Revision A1)
+        u32_le num_gpc;                    // 0x1
+        u64_le l2_cache_size;              // 0x40000
+        u64_le on_board_video_memory_size; // 0x0 (not used)
+        u32_le num_tpc_per_gpc;            // 0x2
+        u32_le bus_type;                   // 0x20 (NVGPU_GPU_BUS_TYPE_AXI)
+        u32_le big_page_size;              // 0x20000
+        u32_le compression_page_size;      // 0x20000
+        u32_le pde_coverage_bit_count;     // 0x1B
+        u32_le available_big_page_sizes;   // 0x30000
+        u32_le gpc_mask;                   // 0x1
+        u32_le sm_arch_sm_version;         // 0x503 (Maxwell Generation 5.0.3?)
+        u32_le sm_arch_spa_version;        // 0x503 (Maxwell Generation 5.0.3?)
+        u32_le sm_arch_warp_count;         // 0x80
+        u32_le gpu_va_bit_count;           // 0x28
+        u32_le reserved;                   // NULL
+        u64_le flags;                      // 0x55
+        u32_le twod_class;                 // 0x902D (FERMI_TWOD_A)
+        u32_le threed_class;               // 0xB197 (MAXWELL_B)
+        u32_le compute_class;              // 0xB1C0 (MAXWELL_COMPUTE_B)
+        u32_le gpfifo_class;               // 0xB06F (MAXWELL_CHANNEL_GPFIFO_A)
+        u32_le inline_to_memory_class;     // 0xA140 (KEPLER_INLINE_TO_MEMORY_B)
+        u32_le dma_copy_class;             // 0xB0B5 (MAXWELL_DMA_COPY_A)
+        u32_le max_fbps_count;             // 0x1
+        u32_le fbp_en_mask;                // 0x0 (disabled)
+        u32_le max_ltc_per_fbp;            // 0x2
+        u32_le max_lts_per_ltc;            // 0x1
+        u32_le max_tex_per_tpc;            // 0x0 (not supported)
+        u32_le max_gpc_count;              // 0x1
+        u32_le rop_l2_en_mask_0;           // 0x21D70 (fuse_status_opt_rop_l2_fbp_r)
+        u32_le rop_l2_en_mask_1;           // 0x0
+        u64_le chipname;                   // 0x6230326D67 ("gm20b")
+        u64_le gr_compbit_store_base_hw;   // 0x0 (not supported)
+    };
+    static_assert(sizeof(IoctlGpuCharacteristics) == 160,
+                  "IoctlGpuCharacteristics is incorrect size");
+
+    struct IoctlCharacteristics {
+        u64_le gpu_characteristics_buf_size; // must not be NULL, but gets overwritten with
+                                             // 0xA0=max_size
+        u64_le gpu_characteristics_buf_addr; // ignored, but must not be NULL
+        IoctlGpuCharacteristics gc;
+    };
+    static_assert(sizeof(IoctlCharacteristics) == 16 + sizeof(IoctlGpuCharacteristics),
+                  "IoctlCharacteristics is incorrect size");
+
+    struct IoctlGpuGetTpcMasksArgs {
+        /// [in]  TPC mask buffer size reserved by userspace. Should be at least
+        /// sizeof(__u32) * fls(gpc_mask) to receive TPC mask for each GPC.
+        /// [out] full kernel buffer size
+        u32_le mask_buf_size;
+        u32_le reserved;
+
+        /// [in]  pointer to TPC mask buffer. It will receive one 32-bit TPC mask per GPC or 0 if
+        /// GPC is not enabled or not present. This parameter is ignored if mask_buf_size is 0.
+        u64_le mask_buf_addr;
+        u64_le unk; // Nintendo add this?
+    };
+    static_assert(sizeof(IoctlGpuGetTpcMasksArgs) == 24,
+                  "IoctlGpuGetTpcMasksArgs is incorrect size");
+
+    struct IoctlActiveSlotMask {
+        u32_le slot; // always 0x07
+        u32_le mask;
+    };
+    static_assert(sizeof(IoctlActiveSlotMask) == 8, "IoctlActiveSlotMask is incorrect size");
+
+    struct IoctlZcullGetCtxSize {
+        u32_le size;
+    };
+    static_assert(sizeof(IoctlZcullGetCtxSize) == 4, "IoctlZcullGetCtxSize is incorrect size");
+
+    struct IoctlNvgpuGpuZcullGetInfoArgs {
+        u32_le width_align_pixels;
+        u32_le height_align_pixels;
+        u32_le pixel_squares_by_aliquots;
+        u32_le aliquot_total;
+        u32_le region_byte_multiplier;
+        u32_le region_header_size;
+        u32_le subregion_header_size;
+        u32_le subregion_width_align_pixels;
+        u32_le subregion_height_align_pixels;
+        u32_le subregion_count;
+    };
+    static_assert(sizeof(IoctlNvgpuGpuZcullGetInfoArgs) == 40,
+                  "IoctlNvgpuGpuZcullGetInfoArgs is incorrect size");
+
+    u32 GetCharacteristics(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetTPCMasks(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetActiveSlotMask(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 ZCullGetCtxSize(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 ZCullGetInfo(const std::vector<u8>& input, std::vector<u8>& output);
+};
+} // namespace Devices
+} // namespace Nvidia
+} // namespace Service

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

@@ -0,0 +1,145 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <map>
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "core/core.h"
+#include "core/hle/service/nvdrv/devices/nvhost_gpu.h"
+
+namespace Service {
+namespace Nvidia {
+namespace Devices {
+
+u32 nvhost_gpu::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+    LOG_DEBUG(Service_NVDRV, "called, command=0x%08x, input_size=0x%llx, output_size=0x%llx",
+              command, input.size(), output.size());
+
+    switch (static_cast<IoctlCommand>(command.raw)) {
+    case IoctlCommand::IocSetNVMAPfdCommand:
+        return SetNVMAPfd(input, output);
+    case IoctlCommand::IocSetClientDataCommand:
+        return SetClientData(input, output);
+    case IoctlCommand::IocGetClientDataCommand:
+        return GetClientData(input, output);
+    case IoctlCommand::IocZCullBind:
+        return ZCullBind(input, output);
+    case IoctlCommand::IocSetErrorNotifierCommand:
+        return SetErrorNotifier(input, output);
+    case IoctlCommand::IocChannelSetPriorityCommand:
+        return SetChannelPriority(input, output);
+    case IoctlCommand::IocAllocGPFIFOEx2Command:
+        return AllocGPFIFOEx2(input, output);
+    case IoctlCommand::IocAllocObjCtxCommand:
+        return AllocateObjectContext(input, output);
+    }
+
+    if (command.group == NVGPU_IOCTL_MAGIC) {
+        if (command.cmd == NVGPU_IOCTL_CHANNEL_SUBMIT_GPFIFO) {
+            return SubmitGPFIFO(input, output);
+        }
+    }
+
+    UNIMPLEMENTED();
+    return 0;
+};
+
+u32 nvhost_gpu::SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlSetNvmapFD params{};
+    std::memcpy(&params, input.data(), input.size());
+    LOG_DEBUG(Service_NVDRV, "called, fd=%x", params.nvmap_fd);
+    nvmap_fd = params.nvmap_fd;
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_gpu::SetClientData(const std::vector<u8>& input, std::vector<u8>& output) {
+    LOG_DEBUG(Service_NVDRV, "called");
+    IoctlClientData params{};
+    std::memcpy(&params, input.data(), input.size());
+    user_data = params.data;
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_gpu::GetClientData(const std::vector<u8>& input, std::vector<u8>& output) {
+    LOG_DEBUG(Service_NVDRV, "called");
+    IoctlClientData params{};
+    std::memcpy(&params, input.data(), input.size());
+    params.data = user_data;
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_gpu::ZCullBind(const std::vector<u8>& input, std::vector<u8>& output) {
+    std::memcpy(&zcull_params, input.data(), input.size());
+    LOG_DEBUG(Service_NVDRV, "called, gpu_va=%lx, mode=%x", zcull_params.gpu_va, zcull_params.mode);
+    std::memcpy(output.data(), &zcull_params, output.size());
+    return 0;
+}
+
+u32 nvhost_gpu::SetErrorNotifier(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlSetErrorNotifier params{};
+    std::memcpy(&params, input.data(), input.size());
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called, offset=%lx, size=%lx, mem=%x", params.offset,
+                params.size, params.mem);
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_gpu::SetChannelPriority(const std::vector<u8>& input, std::vector<u8>& output) {
+    std::memcpy(&channel_priority, input.data(), input.size());
+    LOG_DEBUG(Service_NVDRV, "(STUBBED) called, priority=%x", channel_priority);
+    std::memcpy(output.data(), &channel_priority, output.size());
+    return 0;
+}
+
+u32 nvhost_gpu::AllocGPFIFOEx2(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlAllocGpfifoEx2 params{};
+    std::memcpy(&params, input.data(), input.size());
+    LOG_WARNING(Service_NVDRV,
+                "(STUBBED) called, num_entries=%x, flags=%x, unk0=%x, unk1=%x, unk2=%x, unk3=%x",
+                params.num_entries, params.flags, params.unk0, params.unk1, params.unk2,
+                params.unk3);
+    params.fence_out.id = 0;
+    params.fence_out.value = 0;
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_gpu::AllocateObjectContext(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlAllocObjCtx params{};
+    std::memcpy(&params, input.data(), input.size());
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called, class_num=%x, flags=%x", params.class_num,
+                params.flags);
+    params.obj_id = 0x0;
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& output) {
+    if (input.size() < sizeof(IoctlSubmitGpfifo))
+        UNIMPLEMENTED();
+    IoctlSubmitGpfifo params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called, gpfifo=%lx, num_entries=%x, flags=%x",
+                params.gpfifo, params.num_entries, params.flags);
+
+    auto entries = std::vector<IoctlGpfifoEntry>();
+    entries.resize(params.num_entries);
+    std::memcpy(&entries[0], &input.data()[sizeof(IoctlSubmitGpfifo)],
+                params.num_entries * sizeof(IoctlGpfifoEntry));
+    for (auto entry : entries) {
+        VAddr va_addr = entry.Address();
+        Core::System::GetInstance().GPU().ProcessCommandList(va_addr, entry.sz);
+    }
+    params.fence_out.id = 0;
+    params.fence_out.value = 0;
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+} // namespace Devices
+} // namespace Nvidia
+} // namespace Service

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

@@ -0,0 +1,144 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <vector>
+#include "common/common_types.h"
+#include "common/swap.h"
+#include "core/hle/service/nvdrv/devices/nvdevice.h"
+
+namespace Service {
+namespace Nvidia {
+namespace Devices {
+
+class nvmap;
+constexpr u32 NVGPU_IOCTL_MAGIC('H');
+constexpr u32 NVGPU_IOCTL_CHANNEL_SUBMIT_GPFIFO(0x8);
+
+class nvhost_gpu final : public nvdevice {
+public:
+    nvhost_gpu(std::shared_ptr<nvmap> nvmap_dev) : nvmap_dev(std::move(nvmap_dev)) {}
+    ~nvhost_gpu() override = default;
+
+    u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
+
+private:
+    enum class IoctlCommand : u32_le {
+        IocSetNVMAPfdCommand = 0x40044801,
+        IocSetClientDataCommand = 0x40084714,
+        IocGetClientDataCommand = 0x80084715,
+        IocZCullBind = 0xc010480b,
+        IocSetErrorNotifierCommand = 0xC018480C,
+        IocChannelSetPriorityCommand = 0x4004480D,
+        IocAllocGPFIFOEx2Command = 0xC020481A,
+        IocAllocObjCtxCommand = 0xC0104809,
+    };
+
+    enum class CtxObjects : u32_le {
+        Ctx2D = 0x902D,
+        Ctx3D = 0xB197,
+        CtxCompute = 0xB1C0,
+        CtxKepler = 0xA140,
+        CtxDMA = 0xB0B5,
+        CtxChannelGPFIFO = 0xB06F,
+    };
+
+    struct IoctlSetNvmapFD {
+        u32_le nvmap_fd;
+    };
+    static_assert(sizeof(IoctlSetNvmapFD) == 4, "IoctlSetNvmapFD is incorrect size");
+
+    struct IoctlClientData {
+        u64_le data;
+    };
+    static_assert(sizeof(IoctlClientData) == 8, "IoctlClientData is incorrect size");
+
+    struct IoctlZCullBind {
+        u64_le gpu_va;
+        u32_le mode; // 0=global, 1=no_ctxsw, 2=separate_buffer, 3=part_of_regular_buf
+        INSERT_PADDING_WORDS(1);
+    };
+    static_assert(sizeof(IoctlZCullBind) == 16, "IoctlZCullBind is incorrect size");
+
+    struct IoctlSetErrorNotifier {
+        u64_le offset;
+        u64_le size;
+        u32_le mem; // nvmap object handle
+        INSERT_PADDING_WORDS(1);
+    };
+    static_assert(sizeof(IoctlSetErrorNotifier) == 24, "IoctlSetErrorNotifier is incorrect size");
+
+    struct IoctlFence {
+        u32_le id;
+        u32_le value;
+    };
+    static_assert(sizeof(IoctlFence) == 8, "IoctlFence is incorrect size");
+
+    struct IoctlAllocGpfifoEx2 {
+        u32_le num_entries;   // in
+        u32_le flags;         // in
+        u32_le unk0;          // in (1 works)
+        IoctlFence fence_out; // out
+        u32_le unk1;          // in
+        u32_le unk2;          // in
+        u32_le unk3;          // in
+    };
+    static_assert(sizeof(IoctlAllocGpfifoEx2) == 32, "IoctlAllocGpfifoEx2 is incorrect size");
+
+    struct IoctlAllocObjCtx {
+        u32_le class_num; // 0x902D=2d, 0xB197=3d, 0xB1C0=compute, 0xA140=kepler, 0xB0B5=DMA,
+                          // 0xB06F=channel_gpfifo
+        u32_le flags;
+        u64_le obj_id; // (ignored) used for FREE_OBJ_CTX ioctl, which is not supported
+    };
+    static_assert(sizeof(IoctlAllocObjCtx) == 16, "IoctlAllocObjCtx is incorrect size");
+
+    struct IoctlGpfifoEntry {
+        u32_le entry0; // gpu_va_lo
+        union {
+            u32_le entry1; // gpu_va_hi | (unk_0x02 << 0x08) | (size << 0x0A) | (unk_0x01 << 0x1F)
+            BitField<0, 8, u32_le> gpu_va_hi;
+            BitField<8, 2, u32_le> unk1;
+            BitField<10, 21, u32_le> sz;
+            BitField<31, 1, u32_le> unk2;
+        };
+
+        VAddr Address() const {
+            return (static_cast<VAddr>(gpu_va_hi) << 32) | entry0;
+        }
+    };
+    static_assert(sizeof(IoctlGpfifoEntry) == 8, "IoctlGpfifoEntry is incorrect size");
+
+    struct IoctlSubmitGpfifo {
+        u64_le gpfifo;      // (ignored) pointer to gpfifo fence structs
+        u32_le num_entries; // number of fence objects being submitted
+        u32_le flags;
+        IoctlFence fence_out; // returned new fence object for others to wait on
+    };
+    static_assert(sizeof(IoctlSubmitGpfifo) == 16 + sizeof(IoctlFence),
+                  "submit_gpfifo is incorrect size");
+
+    u32_le nvmap_fd{};
+    u64_le user_data{};
+    IoctlZCullBind zcull_params{};
+    u32_le channel_priority{};
+
+    u32 SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 SetClientData(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetClientData(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 ZCullBind(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 SetErrorNotifier(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 SetChannelPriority(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 AllocGPFIFOEx2(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 AllocateObjectContext(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& output);
+
+    std::shared_ptr<nvmap> nvmap_dev;
+};
+
+} // namespace Devices
+} // namespace Nvidia
+} // namespace Service

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

@@ -13,16 +13,14 @@ namespace Nvidia {
 namespace Devices {
 
 VAddr nvmap::GetObjectAddress(u32 handle) const {
-    auto itr = handles.find(handle);
-    ASSERT(itr != handles.end());
-
-    auto object = itr->second;
+    auto object = GetObject(handle);
+    ASSERT(object);
     ASSERT(object->status == Object::Status::Allocated);
     return object->addr;
 }
 
-u32 nvmap::ioctl(u32 command, const std::vector<u8>& input, std::vector<u8>& output) {
-    switch (static_cast<IoctlCommand>(command)) {
+u32 nvmap::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+    switch (static_cast<IoctlCommand>(command.raw)) {
     case IoctlCommand::Create:
         return IocCreate(input, output);
     case IoctlCommand::Alloc:
@@ -52,7 +50,7 @@ u32 nvmap::IocCreate(const std::vector<u8>& input, std::vector<u8>& output) {
     u32 handle = next_handle++;
     handles[handle] = std::move(object);
 
-    LOG_WARNING(Service_NVDRV, "(STUBBED) size 0x%08X", params.size);
+    LOG_DEBUG(Service_NVDRV, "size=0x%08X", params.size);
 
     params.handle = handle;
 
@@ -64,17 +62,16 @@ u32 nvmap::IocAlloc(const std::vector<u8>& input, std::vector<u8>& output) {
     IocAllocParams params;
     std::memcpy(&params, input.data(), sizeof(params));
 
-    auto itr = handles.find(params.handle);
-    ASSERT(itr != handles.end());
+    auto object = GetObject(params.handle);
+    ASSERT(object);
 
-    auto object = itr->second;
     object->flags = params.flags;
     object->align = params.align;
     object->kind = params.kind;
     object->addr = params.addr;
     object->status = Object::Status::Allocated;
 
-    LOG_WARNING(Service_NVDRV, "(STUBBED) Allocated address 0x%llx", params.addr);
+    LOG_DEBUG(Service_NVDRV, "called, addr=0x%llx", params.addr);
 
     std::memcpy(output.data(), &params, sizeof(params));
     return 0;
@@ -86,10 +83,10 @@ u32 nvmap::IocGetId(const std::vector<u8>& input, std::vector<u8>& output) {
 
     LOG_WARNING(Service_NVDRV, "called");
 
-    auto itr = handles.find(params.handle);
-    ASSERT(itr != handles.end());
+    auto object = GetObject(params.handle);
+    ASSERT(object);
 
-    params.id = itr->second->id;
+    params.id = object->id;
 
     std::memcpy(output.data(), &params, sizeof(params));
     return 0;
@@ -123,10 +120,8 @@ u32 nvmap::IocParam(const std::vector<u8>& input, std::vector<u8>& output) {
 
     LOG_WARNING(Service_NVDRV, "(STUBBED) called type=%u", params.type);
 
-    auto itr = handles.find(params.handle);
-    ASSERT(itr != handles.end());
-
-    auto object = itr->second;
+    auto object = GetObject(params.handle);
+    ASSERT(object);
     ASSERT(object->status == Object::Status::Allocated);
 
     switch (static_cast<ParamTypes>(params.type)) {

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

@@ -24,10 +24,9 @@ public:
     /// Returns the allocated address of an nvmap object given its handle.
     VAddr GetObjectAddress(u32 handle) const;
 
-    u32 ioctl(u32 command, const std::vector<u8>& input, std::vector<u8>& output) override;
+    u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
 
-private:
-    // Represents an nvmap object.
+    /// Represents an nvmap object.
     struct Object {
         enum class Status { Created, Allocated };
         u32 id;
@@ -39,10 +38,19 @@ private:
         Status status;
     };
 
+    std::shared_ptr<Object> GetObject(u32 handle) const {
+        auto itr = handles.find(handle);
+        if (itr != handles.end()) {
+            return itr->second;
+        }
+        return {};
+    }
+
+private:
     /// Id to use for the next handle that is created.
     u32 next_handle = 1;
 
-    // Id to use for the next object that is created.
+    /// Id to use for the next object that is created.
     u32 next_id = 1;
 
     /// Mapping of currently allocated handles to the objects they represent.

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

@@ -4,6 +4,7 @@
 
 #include "common/logging/log.h"
 #include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/event.h"
 #include "core/hle/service/nvdrv/interface.h"
 #include "core/hle/service/nvdrv/nvdrv.h"
 
@@ -11,7 +12,7 @@ namespace Service {
 namespace Nvidia {
 
 void NVDRV::Open(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    LOG_DEBUG(Service_NVDRV, "called");
 
     auto buffer = ctx.BufferDescriptorA()[0];
 
@@ -25,31 +26,35 @@ void NVDRV::Open(Kernel::HLERequestContext& ctx) {
 }
 
 void NVDRV::Ioctl(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    LOG_DEBUG(Service_NVDRV, "called");
 
     IPC::RequestParser rp{ctx};
     u32 fd = rp.Pop<u32>();
     u32 command = rp.Pop<u32>();
 
-    auto input_buffer = ctx.BufferDescriptorA()[0];
-    auto output_buffer = ctx.BufferDescriptorB()[0];
-
-    std::vector<u8> input(input_buffer.Size());
-    std::vector<u8> output(output_buffer.Size());
-
-    Memory::ReadBlock(input_buffer.Address(), input.data(), input_buffer.Size());
-
-    u32 nv_result = nvdrv->Ioctl(fd, command, input, output);
-
-    Memory::WriteBlock(output_buffer.Address(), output.data(), output_buffer.Size());
-
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(RESULT_SUCCESS);
-    rb.Push(nv_result);
+    if (ctx.BufferDescriptorA()[0].Size() != 0) {
+        auto input_buffer = ctx.BufferDescriptorA()[0];
+        auto output_buffer = ctx.BufferDescriptorB()[0];
+        std::vector<u8> input(input_buffer.Size());
+        std::vector<u8> output(output_buffer.Size());
+        Memory::ReadBlock(input_buffer.Address(), input.data(), input_buffer.Size());
+        rb.Push(nvdrv->Ioctl(fd, command, input, output));
+        Memory::WriteBlock(output_buffer.Address(), output.data(), output_buffer.Size());
+    } else {
+        auto input_buffer = ctx.BufferDescriptorX()[0];
+        auto output_buffer = ctx.BufferDescriptorC()[0];
+        std::vector<u8> input(input_buffer.size);
+        std::vector<u8> output(output_buffer.size);
+        Memory::ReadBlock(input_buffer.Address(), input.data(), input_buffer.size);
+        rb.Push(nvdrv->Ioctl(fd, command, input, output));
+        Memory::WriteBlock(output_buffer.Address(), output.data(), output_buffer.size);
+    }
 }
 
 void NVDRV::Close(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    LOG_DEBUG(Service_NVDRV, "called");
 
     IPC::RequestParser rp{ctx};
     u32 fd = rp.Pop<u32>();
@@ -67,16 +72,34 @@ void NVDRV::Initialize(Kernel::HLERequestContext& ctx) {
     rb.Push<u32>(0);
 }
 
+void NVDRV::QueryEvent(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    u32 fd = rp.Pop<u32>();
+    u32 event_id = rp.Pop<u32>();
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called, fd=%x, event_id=%x", fd, event_id);
+
+    IPC::ResponseBuilder rb{ctx, 3, 1};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushCopyObjects(query_event);
+    rb.Push<u32>(0);
+}
+
 void NVDRV::SetClientPID(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     pid = rp.Pop<u64>();
 
-    LOG_INFO(Service_NVDRV, "called, pid=0x%lx", pid);
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called, pid=0x%lx", pid);
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(RESULT_SUCCESS);
     rb.Push<u32>(0);
 }
 
+void NVDRV::FinishInitialize(Kernel::HLERequestContext& ctx) {
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
 NVDRV::NVDRV(std::shared_ptr<Module> nvdrv, const char* name)
     : ServiceFramework(name), nvdrv(std::move(nvdrv)) {
     static const FunctionInfo functions[] = {
@@ -84,9 +107,13 @@ NVDRV::NVDRV(std::shared_ptr<Module> nvdrv, const char* name)
         {1, &NVDRV::Ioctl, "Ioctl"},
         {2, &NVDRV::Close, "Close"},
         {3, &NVDRV::Initialize, "Initialize"},
+        {4, &NVDRV::QueryEvent, "QueryEvent"},
         {8, &NVDRV::SetClientPID, "SetClientPID"},
+        {13, &NVDRV::FinishInitialize, "FinishInitialize"},
     };
     RegisterHandlers(functions);
+
+    query_event = Kernel::Event::Create(Kernel::ResetType::OneShot, "NVDRV::query_event");
 }
 
 } // namespace Nvidia

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

@@ -6,6 +6,7 @@
 
 #include <memory>
 #include <string>
+#include "core/hle/kernel/event.h"
 #include "core/hle/service/nvdrv/nvdrv.h"
 #include "core/hle/service/service.h"
 
@@ -22,11 +23,15 @@ private:
     void Ioctl(Kernel::HLERequestContext& ctx);
     void Close(Kernel::HLERequestContext& ctx);
     void Initialize(Kernel::HLERequestContext& ctx);
+    void QueryEvent(Kernel::HLERequestContext& ctx);
     void SetClientPID(Kernel::HLERequestContext& ctx);
+    void FinishInitialize(Kernel::HLERequestContext& ctx);
 
     std::shared_ptr<Module> nvdrv;
 
     u64 pid{};
+
+    Kernel::SharedPtr<Kernel::Event> query_event;
 };
 
 } // namespace Nvidia

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

@@ -7,6 +7,8 @@
 #include "core/hle/service/nvdrv/devices/nvdisp_disp0.h"
 #include "core/hle/service/nvdrv/devices/nvhost_as_gpu.h"
 #include "core/hle/service/nvdrv/devices/nvhost_ctrl.h"
+#include "core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h"
+#include "core/hle/service/nvdrv/devices/nvhost_gpu.h"
 #include "core/hle/service/nvdrv/devices/nvmap.h"
 #include "core/hle/service/nvdrv/interface.h"
 #include "core/hle/service/nvdrv/nvdrv.h"
@@ -21,13 +23,17 @@ void InstallInterfaces(SM::ServiceManager& service_manager) {
     auto module_ = std::make_shared<Module>();
     std::make_shared<NVDRV>(module_, "nvdrv")->InstallAsService(service_manager);
     std::make_shared<NVDRV>(module_, "nvdrv:a")->InstallAsService(service_manager);
+    std::make_shared<NVDRV>(module_, "nvdrv:s")->InstallAsService(service_manager);
+    std::make_shared<NVDRV>(module_, "nvdrv:t")->InstallAsService(service_manager);
     std::make_shared<NVMEMP>()->InstallAsService(service_manager);
     nvdrv = module_;
 }
 
 Module::Module() {
     auto nvmap_dev = std::make_shared<Devices::nvmap>();
-    devices["/dev/nvhost-as-gpu"] = std::make_shared<Devices::nvhost_as_gpu>();
+    devices["/dev/nvhost-as-gpu"] = std::make_shared<Devices::nvhost_as_gpu>(nvmap_dev);
+    devices["/dev/nvhost-gpu"] = std::make_shared<Devices::nvhost_gpu>(nvmap_dev);
+    devices["/dev/nvhost-ctrl-gpu"] = std::make_shared<Devices::nvhost_ctrl_gpu>();
     devices["/dev/nvmap"] = nvmap_dev;
     devices["/dev/nvdisp_disp0"] = std::make_shared<Devices::nvdisp_disp0>(nvmap_dev);
     devices["/dev/nvhost-ctrl"] = std::make_shared<Devices::nvhost_ctrl>();
@@ -45,12 +51,12 @@ u32 Module::Open(std::string device_name) {
     return fd;
 }
 
-u32 Module::Ioctl(u32 fd, u32 command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 Module::Ioctl(u32 fd, u32_le command, const std::vector<u8>& input, std::vector<u8>& output) {
     auto itr = open_files.find(fd);
     ASSERT_MSG(itr != open_files.end(), "Tried to talk to an invalid device");
 
     auto device = itr->second;
-    return device->ioctl(command, input, output);
+    return device->ioctl({command}, input, output);
 }
 
 ResultCode Module::Close(u32 fd) {

src/core/hle/service/nvflinger/buffer_queue.cpp

@@ -40,7 +40,11 @@ u32 BufferQueue::DequeueBuffer(u32 pixel_format, u32 width, u32 height) {
         return igbp_buffer.format == pixel_format && igbp_buffer.width == width &&
                igbp_buffer.height == height;
     });
-    ASSERT(itr != queue.end());
+    if (itr == queue.end()) {
+        LOG_CRITICAL(Service_NVDRV, "no free buffers for pixel_format=%d, width=%d, height=%d",
+                     pixel_format, width, height);
+        itr = queue.begin();
+    }
 
     itr->status = Buffer::Status::Dequeued;
     return itr->slot;
@@ -54,12 +58,13 @@ const IGBPBuffer& BufferQueue::RequestBuffer(u32 slot) const {
     return itr->igbp_buffer;
 }
 
-void BufferQueue::QueueBuffer(u32 slot) {
+void BufferQueue::QueueBuffer(u32 slot, BufferTransformFlags transform) {
     auto itr = std::find_if(queue.begin(), queue.end(),
                             [&](const Buffer& buffer) { return buffer.slot == slot; });
     ASSERT(itr != queue.end());
     ASSERT(itr->status == Buffer::Status::Dequeued);
     itr->status = Buffer::Status::Queued;
+    itr->transform = transform;
 }
 
 boost::optional<const BufferQueue::Buffer&> BufferQueue::AcquireBuffer() {

src/core/hle/service/nvflinger/buffer_queue.h

@@ -46,18 +46,32 @@ public:
     BufferQueue(u32 id, u64 layer_id);
     ~BufferQueue() = default;
 
+    enum class BufferTransformFlags : u32 {
+        /// Flip source image horizontally (around the vertical axis)
+        FlipH = 0x01,
+        /// Flip source image vertically (around the horizontal axis)
+        FlipV = 0x02,
+        /// Rotate source image 90 degrees clockwise
+        Rotate90 = 0x04,
+        /// Rotate source image 180 degrees
+        Roate180 = 0x03,
+        /// Rotate source image 270 degrees clockwise
+        Roate270 = 0x07,
+    };
+
     struct Buffer {
         enum class Status { Free = 0, Queued = 1, Dequeued = 2, Acquired = 3 };
 
         u32 slot;
         Status status = Status::Free;
         IGBPBuffer igbp_buffer;
+        BufferTransformFlags transform;
     };
 
     void SetPreallocatedBuffer(u32 slot, IGBPBuffer& buffer);
     u32 DequeueBuffer(u32 pixel_format, u32 width, u32 height);
     const IGBPBuffer& RequestBuffer(u32 slot) const;
-    void QueueBuffer(u32 slot);
+    void QueueBuffer(u32 slot, BufferTransformFlags transform);
     boost::optional<const Buffer&> AcquireBuffer();
     void ReleaseBuffer(u32 slot);
     u32 Query(QueryType type);

src/core/hle/service/nvflinger/nvflinger.cpp

@@ -145,7 +145,7 @@ void NVFlinger::Compose() {
         ASSERT(nvdisp);
 
         nvdisp->flip(igbp_buffer.gpu_buffer_id, igbp_buffer.offset, igbp_buffer.format,
-                     igbp_buffer.width, igbp_buffer.height, igbp_buffer.stride);
+                     igbp_buffer.width, igbp_buffer.height, igbp_buffer.stride, buffer->transform);
 
         buffer_queue->ReleaseBuffer(buffer->slot);
     }

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

@@ -103,6 +103,8 @@ void SM::GetService(Kernel::HLERequestContext& ctx) {
         rb.Push(client_port.Code());
         LOG_ERROR(Service_SM, "called service=%s -> error 0x%08X", name.c_str(),
                   client_port.Code().raw);
+        if (name.length() == 0)
+            return; // LibNX Fix
         UNIMPLEMENTED();
         return;
     }

src/core/hle/service/time/time.cpp

@@ -68,8 +68,15 @@ public:
     ITimeZoneService() : ServiceFramework("ITimeZoneService") {
         static const FunctionInfo functions[] = {
             {0, &ITimeZoneService::GetDeviceLocationName, "GetDeviceLocationName"},
+            {1, nullptr, "SetDeviceLocationName"},
             {2, &ITimeZoneService::GetTotalLocationNameCount, "GetTotalLocationNameCount"},
+            {3, nullptr, "LoadLocationNameList"},
+            {4, &ITimeZoneService::LoadTimeZoneRule, "LoadTimeZoneRule"},
+            {5, nullptr, "GetTimeZoneRuleVersion"},
+            {100, nullptr, "ToCalendarTime"},
             {101, &ITimeZoneService::ToCalendarTimeWithMyRule, "ToCalendarTimeWithMyRule"},
+            {200, nullptr, "ToPosixTime"},
+            {201, nullptr, "ToPosixTimeWithMyRule"},
         };
         RegisterHandlers(functions);
     }
@@ -90,6 +97,12 @@ private:
         rb.Push<u32>(0);
     }
 
+    void LoadTimeZoneRule(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_Time, "(STUBBED) called");
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
     void ToCalendarTimeWithMyRule(Kernel::HLERequestContext& ctx) {
         IPC::RequestParser rp{ctx};
         u64 posix_time = rp.Pop<u64>();

src/core/hle/service/vi/vi.cpp

@@ -3,7 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <algorithm>
-
+#include <array>
 #include "common/alignment.h"
 #include "common/scope_exit.h"
 #include "core/core_timing.h"
@@ -19,6 +19,15 @@
 namespace Service {
 namespace VI {
 
+struct DisplayInfo {
+    char display_name[0x40]{"Default"};
+    u64 unknown_1{1};
+    u64 unknown_2{1};
+    u64 width{1920};
+    u64 height{1080};
+};
+static_assert(sizeof(DisplayInfo) == 0x60, "DisplayInfo has wrong size");
+
 class Parcel {
 public:
     // This default size was chosen arbitrarily.
@@ -92,8 +101,10 @@ public:
         SerializeData();
 
         Header header{};
-        header.data_offset = sizeof(Header);
         header.data_size = static_cast<u32_le>(write_index - sizeof(Header));
+        header.data_offset = sizeof(Header);
+        header.objects_size = 4;
+        header.objects_offset = sizeof(Header) + header.data_size;
         std::memcpy(buffer.data(), &header, sizeof(Header));
 
         return buffer;
@@ -133,11 +144,11 @@ protected:
 private:
     struct Data {
         u32_le magic = 2;
-        u32_le process_id;
+        u32_le process_id = 1;
         u32_le id;
-        INSERT_PADDING_BYTES(0xC);
+        INSERT_PADDING_WORDS(3);
         std::array<u8, 8> dispdrv = {'d', 'i', 's', 'p', 'd', 'r', 'v', '\0'};
-        INSERT_PADDING_BYTES(8);
+        INSERT_PADDING_WORDS(2);
     };
     static_assert(sizeof(Data) == 0x28, "ParcelData has wrong size");
 
@@ -202,7 +213,6 @@ public:
     void DeserializeData() override {
         std::u16string token = ReadInterfaceToken();
         data = Read<Data>();
-        ASSERT(data.graphic_buffer_length == sizeof(NVFlinger::IGBPBuffer));
         buffer = Read<NVFlinger::IGBPBuffer>();
     }
 
@@ -252,6 +262,11 @@ public:
     Data data;
 };
 
+// TODO(bunnei): Remove this. When set to 1, games will think a fence is valid and boot further.
+// This will break libnx and potentially other apps that more stringently check this. This is here
+// purely as a convenience, and should go away once we implement fences.
+static constexpr u32 FENCE_HACK = 0;
+
 class IGBPDequeueBufferResponseParcel : public Parcel {
 public:
     explicit IGBPDequeueBufferResponseParcel(u32 slot) : Parcel(), slot(slot) {}
@@ -259,11 +274,20 @@ public:
 
 protected:
     void SerializeData() override {
-        Write(slot);
-        // TODO(Subv): Find out how this Fence is used.
-        std::array<u32_le, 11> fence = {};
-        Write(fence);
-        Write<u32_le>(0);
+        // TODO(bunnei): Find out what this all means. Writing anything non-zero here breaks libnx.
+        Write<u32>(0);
+        Write<u32>(FENCE_HACK);
+        Write<u32>(0);
+        Write<u32>(0);
+        Write<u32>(0);
+        Write<u32>(0);
+        Write<u32>(0);
+        Write<u32>(0);
+        Write<u32>(0);
+        Write<u32>(0);
+        Write<u32>(0);
+        Write<u32>(0);
+        Write<u32>(0);
     }
 
     u32_le slot;
@@ -292,14 +316,11 @@ public:
 
 protected:
     void SerializeData() override {
-        // TODO(Subv): Find out what this all means
-        Write<u32_le>(1);
-
-        Write<u32_le>(sizeof(NVFlinger::IGBPBuffer));
-        Write<u32_le>(0); // Unknown
-
+        // TODO(bunnei): Find out what this all means. Writing anything non-zero here breaks libnx.
+        Write<u32_le>(0);
+        Write<u32_le>(FENCE_HACK);
+        Write<u32_le>(0);
         Write(buffer);
-
         Write<u32_le>(0);
     }
 
@@ -318,13 +339,29 @@ public:
         data = Read<Data>();
     }
 
+    struct Fence {
+        u32_le id;
+        u32_le value;
+    };
+    static_assert(sizeof(Fence) == 8, "Fence has wrong size");
+
     struct Data {
         u32_le slot;
-        INSERT_PADDING_WORDS(2);
+        INSERT_PADDING_WORDS(3);
         u32_le timestamp;
-        INSERT_PADDING_WORDS(20);
+        s32_le is_auto_timestamp;
+        s32_le crop_left;
+        s32_le crop_top;
+        s32_le crop_right;
+        s32_le crop_bottom;
+        s32_le scaling_mode;
+        NVFlinger::BufferQueue::BufferTransformFlags transform;
+        u32_le sticky_transform;
+        INSERT_PADDING_WORDS(2);
+        u32_le fence_is_valid;
+        std::array<Fence, 2> fences;
     };
-    static_assert(sizeof(Data) == 96, "ParcelData has wrong size");
+    static_assert(sizeof(Data) == 80, "ParcelData has wrong size");
 
     Data data;
 };
@@ -392,7 +429,7 @@ public:
             {0, &IHOSBinderDriver::TransactParcel, "TransactParcel"},
             {1, &IHOSBinderDriver::AdjustRefcount, "AdjustRefcount"},
             {2, &IHOSBinderDriver::GetNativeHandle, "GetNativeHandle"},
-            {3, nullptr, "TransactParcelAuto"},
+            {3, &IHOSBinderDriver::TransactParcelAuto, "TransactParcelAuto"},
         };
         RegisterHandlers(functions);
     }
@@ -416,35 +453,23 @@ private:
         SetPreallocatedBuffer = 14
     };
 
-    void TransactParcel(Kernel::HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        u32 id = rp.Pop<u32>();
-        auto transaction = static_cast<TransactionId>(rp.Pop<u32>());
-        u32 flags = rp.Pop<u32>();
-
-        auto& input_buffer = ctx.BufferDescriptorA()[0];
-        std::vector<u8> input_data(input_buffer.Size());
-        Memory::ReadBlock(input_buffer.Address(), input_data.data(), input_buffer.Size());
-
-        auto& output_buffer = ctx.BufferDescriptorB()[0];
-
+    void TransactParcel(u32 id, TransactionId transaction, const std::vector<u8>& input_data,
+                        VAddr output_addr, u64 output_size) {
         auto buffer_queue = nv_flinger->GetBufferQueue(id);
-        LOG_WARNING(Service_VI, "(STUBBED) called, transaction=%x", transaction);
+
         if (transaction == TransactionId::Connect) {
             IGBPConnectRequestParcel request{input_data};
             IGBPConnectResponseParcel response{1280, 720};
-            auto response_buffer = response.Serialize();
-            Memory::WriteBlock(output_buffer.Address(), response_buffer.data(),
-                               output_buffer.Size());
+            std::vector<u8> response_buffer = response.Serialize();
+            Memory::WriteBlock(output_addr, response_buffer.data(), response_buffer.size());
         } else if (transaction == TransactionId::SetPreallocatedBuffer) {
             IGBPSetPreallocatedBufferRequestParcel request{input_data};
 
             buffer_queue->SetPreallocatedBuffer(request.data.slot, request.buffer);
 
             IGBPSetPreallocatedBufferResponseParcel response{};
-            auto response_buffer = response.Serialize();
-            Memory::WriteBlock(output_buffer.Address(), response_buffer.data(),
-                               output_buffer.Size());
+            std::vector<u8> response_buffer = response.Serialize();
+            Memory::WriteBlock(output_addr, response_buffer.data(), response_buffer.size());
         } else if (transaction == TransactionId::DequeueBuffer) {
             IGBPDequeueBufferRequestParcel request{input_data};
 
@@ -452,27 +477,24 @@ private:
                                                    request.data.height);
 
             IGBPDequeueBufferResponseParcel response{slot};
-            auto response_buffer = response.Serialize();
-            Memory::WriteBlock(output_buffer.Address(), response_buffer.data(),
-                               output_buffer.Size());
+            std::vector<u8> response_buffer = response.Serialize();
+            Memory::WriteBlock(output_addr, response_buffer.data(), response_buffer.size());
         } else if (transaction == TransactionId::RequestBuffer) {
             IGBPRequestBufferRequestParcel request{input_data};
 
             auto& buffer = buffer_queue->RequestBuffer(request.slot);
 
             IGBPRequestBufferResponseParcel response{buffer};
-            auto response_buffer = response.Serialize();
-            Memory::WriteBlock(output_buffer.Address(), response_buffer.data(),
-                               output_buffer.Size());
+            std::vector<u8> response_buffer = response.Serialize();
+            Memory::WriteBlock(output_addr, response_buffer.data(), response_buffer.size());
         } else if (transaction == TransactionId::QueueBuffer) {
             IGBPQueueBufferRequestParcel request{input_data};
 
-            buffer_queue->QueueBuffer(request.data.slot);
+            buffer_queue->QueueBuffer(request.data.slot, request.data.transform);
 
             IGBPQueueBufferResponseParcel response{1280, 720};
-            auto response_buffer = response.Serialize();
-            Memory::WriteBlock(output_buffer.Address(), response_buffer.data(),
-                               output_buffer.Size());
+            std::vector<u8> response_buffer = response.Serialize();
+            Memory::WriteBlock(output_addr, response_buffer.data(), response_buffer.size());
         } else if (transaction == TransactionId::Query) {
             IGBPQueryRequestParcel request{input_data};
 
@@ -480,12 +502,46 @@ private:
                 buffer_queue->Query(static_cast<NVFlinger::BufferQueue::QueryType>(request.type));
 
             IGBPQueryResponseParcel response{value};
-            auto response_buffer = response.Serialize();
-            Memory::WriteBlock(output_buffer.Address(), response_buffer.data(),
-                               output_buffer.Size());
+            std::vector<u8> response_buffer = response.Serialize();
+            Memory::WriteBlock(output_addr, response_buffer.data(), response_buffer.size());
+        } else if (transaction == TransactionId::CancelBuffer) {
+            LOG_WARNING(Service_VI, "(STUBBED) called, transaction=CancelBuffer");
         } else {
             ASSERT_MSG(false, "Unimplemented");
         }
+    }
+
+    void TransactParcel(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        u32 id = rp.Pop<u32>();
+        auto transaction = static_cast<TransactionId>(rp.Pop<u32>());
+        u32 flags = rp.Pop<u32>();
+        LOG_DEBUG(Service_VI, "called, transaction=%x", transaction);
+
+        auto& input_buffer = ctx.BufferDescriptorA()[0];
+        auto& output_buffer = ctx.BufferDescriptorB()[0];
+        std::vector<u8> input_data(input_buffer.Size());
+        Memory::ReadBlock(input_buffer.Address(), input_data.data(), input_buffer.Size());
+
+        TransactParcel(id, transaction, input_data, output_buffer.Address(), output_buffer.Size());
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void TransactParcelAuto(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        u32 id = rp.Pop<u32>();
+        auto transaction = static_cast<TransactionId>(rp.Pop<u32>());
+        u32 flags = rp.Pop<u32>();
+        LOG_DEBUG(Service_VI, "called, transaction=%x", transaction);
+
+        auto& input_buffer = ctx.BufferDescriptorX()[0];
+        auto& output_buffer = ctx.BufferDescriptorC()[0];
+        std::vector<u8> input_data(input_buffer.size);
+        Memory::ReadBlock(input_buffer.Address(), input_data.data(), input_buffer.size);
+
+        TransactParcel(id, transaction, input_data, output_buffer.Address(), output_buffer.Size());
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);
@@ -518,7 +574,7 @@ private:
     }
 
     std::shared_ptr<NVFlinger::NVFlinger> nv_flinger;
-};
+}; // namespace VI
 
 class ISystemDisplayService final : public ServiceFramework<ISystemDisplayService> {
 public:
@@ -648,12 +704,12 @@ void IApplicationDisplayService::CloseDisplay(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     u64 display_id = rp.Pop<u64>();
 
-    IPC::ResponseBuilder rb = rp.MakeBuilder(4, 0, 0);
+    IPC::ResponseBuilder rb = rp.MakeBuilder(2, 0, 0);
     rb.Push(RESULT_SUCCESS);
 }
 
 void IApplicationDisplayService::OpenLayer(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_VI, "(STUBBED) called");
+    LOG_DEBUG(Service_VI, "called");
     IPC::RequestParser rp{ctx};
     auto name_buf = rp.PopRaw<std::array<u8, 0x40>>();
     auto end = std::find(name_buf.begin(), name_buf.end(), '\0');
@@ -678,7 +734,7 @@ void IApplicationDisplayService::OpenLayer(Kernel::HLERequestContext& ctx) {
 }
 
 void IApplicationDisplayService::CreateStrayLayer(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service, "(STUBBED) called");
+    LOG_DEBUG(Service_VI, "called");
 
     IPC::RequestParser rp{ctx};
     u32 flags = rp.Pop<u32>();
@@ -722,6 +778,17 @@ void IApplicationDisplayService::SetLayerScalingMode(Kernel::HLERequestContext&
     rb.Push(RESULT_SUCCESS);
 }
 
+void IApplicationDisplayService::ListDisplays(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    DisplayInfo display_info;
+    auto& buffer = ctx.BufferDescriptorB()[0];
+    Memory::WriteBlock(buffer.Address(), &display_info, sizeof(DisplayInfo));
+    IPC::ResponseBuilder rb = rp.MakeBuilder(4, 0, 0);
+    rb.Push(RESULT_SUCCESS);
+    rb.Push<u64>(1);
+    LOG_WARNING(Service_VI, "(STUBBED) called");
+}
+
 void IApplicationDisplayService::GetDisplayVsyncEvent(Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service_VI, "(STUBBED) called");
     IPC::RequestParser rp{ctx};
@@ -743,7 +810,7 @@ IApplicationDisplayService::IApplicationDisplayService(
         {102, &IApplicationDisplayService::GetManagerDisplayService, "GetManagerDisplayService"},
         {103, &IApplicationDisplayService::GetIndirectDisplayTransactionService,
          "GetIndirectDisplayTransactionService"},
-        {1000, nullptr, "ListDisplays"},
+        {1000, &IApplicationDisplayService::ListDisplays, "ListDisplays"},
         {1010, &IApplicationDisplayService::OpenDisplay, "OpenDisplay"},
         {1020, &IApplicationDisplayService::CloseDisplay, "CloseDisplay"},
         {2101, &IApplicationDisplayService::SetLayerScalingMode, "SetLayerScalingMode"},

src/core/hle/service/vi/vi.h

@@ -30,6 +30,7 @@ private:
     void OpenDisplay(Kernel::HLERequestContext& ctx);
     void CloseDisplay(Kernel::HLERequestContext& ctx);
     void SetLayerScalingMode(Kernel::HLERequestContext& ctx);
+    void ListDisplays(Kernel::HLERequestContext& ctx);
     void OpenLayer(Kernel::HLERequestContext& ctx);
     void CreateStrayLayer(Kernel::HLERequestContext& ctx);
     void DestroyStrayLayer(Kernel::HLERequestContext& ctx);

src/core/loader/deconstructed_rom_directory.cpp

@@ -29,7 +29,7 @@ static std::string FindRomFS(const std::string& directory) {
 
         // Verify extension
         const std::string extension = physical_name.substr(physical_name.find_last_of(".") + 1);
-        if (Common::ToLower(extension) != "istorage") {
+        if (Common::ToLower(extension) != "romfs") {
             return true;
         }
 
@@ -38,7 +38,7 @@ static std::string FindRomFS(const std::string& directory) {
         return false;
     };
 
-    // Search the specified directory recursively, looking for the first .istorage file, which will
+    // Search the specified directory recursively, looking for the first .romfs file, which will
     // be used for the RomFS
     FileUtil::ForeachDirectoryEntry(nullptr, directory, callback);
 
@@ -128,10 +128,10 @@ ResultStatus AppLoader_DeconstructedRomDirectory::Load(
         Kernel::ResourceLimit::GetForCategory(Kernel::ResourceLimitCategory::APPLICATION);
     process->Run(Memory::PROCESS_IMAGE_VADDR, 48, Kernel::DEFAULT_STACK_SIZE);
 
-    // Find the RomFS by searching for a ".istorage" file in this directory
+    // Find the RomFS by searching for a ".romfs" file in this directory
     filepath_romfs = FindRomFS(directory);
 
-    // Register the RomFS if a ".istorage" file was found
+    // Register the RomFS if a ".romfs" file was found
     if (!filepath_romfs.empty()) {
         Service::FileSystem::RegisterFileSystem(std::make_unique<FileSys::RomFS_Factory>(*this),
                                                 Service::FileSystem::Type::RomFS);

src/core/loader/deconstructed_rom_directory.h

@@ -15,7 +15,7 @@ namespace Loader {
  * This class loads a "deconstructed ROM directory", which are the typical format we see for Switch
  * game dumps. The path should be a "main" NSO, which must be in a directory that contains the other
  * standard ExeFS NSOs (e.g. rtld, sdk, etc.). It will automatically find and load these.
- * Furthermore, it will look for the first .istorage file (optionally) and use this for the RomFS.
+ * Furthermore, it will look for the first .romfs file (optionally) and use this for the RomFS.
  */
 class AppLoader_DeconstructedRomDirectory final : public AppLoader {
 public:

src/core/memory.h

@@ -25,10 +25,11 @@ namespace Memory {
  * Page size used by the ARM architecture. This is the smallest granularity with which memory can
  * be mapped.
  */
-const int PAGE_BITS = 12;
-const u64 PAGE_SIZE = 1 << PAGE_BITS;
-const u64 PAGE_MASK = PAGE_SIZE - 1;
-const size_t PAGE_TABLE_NUM_ENTRIES = 1ULL << (36 - PAGE_BITS);
+constexpr size_t PAGE_BITS = 12;
+constexpr u64 PAGE_SIZE = 1 << PAGE_BITS;
+constexpr u64 PAGE_MASK = PAGE_SIZE - 1;
+constexpr size_t ADDRESS_SPACE_BITS = 36;
+constexpr size_t PAGE_TABLE_NUM_ENTRIES = 1ULL << (ADDRESS_SPACE_BITS - PAGE_BITS);
 
 enum class PageType : u8 {
     /// Page is unmapped and should cause an access error.

src/video_core/CMakeLists.txt

@@ -1,4 +1,15 @@
 add_library(video_core STATIC
+    command_processor.cpp
+    command_processor.h
+    engines/fermi_2d.cpp
+    engines/fermi_2d.h
+    engines/maxwell_3d.cpp
+    engines/maxwell_3d.h
+    engines/maxwell_compute.cpp
+    engines/maxwell_compute.h
+    gpu.h
+    memory_manager.cpp
+    memory_manager.h
     renderer_base.cpp
     renderer_base.h
     renderer_opengl/gl_resource_manager.h

src/video_core/command_processor.cpp

@@ -0,0 +1,119 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <array>
+#include <cstddef>
+#include <memory>
+#include <utility>
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "common/microprofile.h"
+#include "common/vector_math.h"
+#include "core/memory.h"
+#include "core/tracer/recorder.h"
+#include "video_core/command_processor.h"
+#include "video_core/engines/fermi_2d.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/engines/maxwell_compute.h"
+#include "video_core/gpu.h"
+#include "video_core/renderer_base.h"
+#include "video_core/video_core.h"
+
+namespace Tegra {
+
+enum class BufferMethods {
+    BindObject = 0,
+    CountBufferMethods = 0x100,
+};
+
+void GPU::WriteReg(u32 method, u32 subchannel, u32 value) {
+    LOG_WARNING(HW_GPU, "Processing method %08X on subchannel %u value %08X", method, subchannel,
+                value);
+
+    if (method == static_cast<u32>(BufferMethods::BindObject)) {
+        // Bind the current subchannel to the desired engine id.
+        LOG_DEBUG(HW_GPU, "Binding subchannel %u to engine %u", subchannel, value);
+        ASSERT(bound_engines.find(subchannel) == bound_engines.end());
+        bound_engines[subchannel] = static_cast<EngineID>(value);
+        return;
+    }
+
+    if (method < static_cast<u32>(BufferMethods::CountBufferMethods)) {
+        // TODO(Subv): Research and implement these methods.
+        LOG_ERROR(HW_GPU, "Special buffer methods other than Bind are not implemented");
+        return;
+    }
+
+    ASSERT(bound_engines.find(subchannel) != bound_engines.end());
+
+    const EngineID engine = bound_engines[subchannel];
+
+    switch (engine) {
+    case EngineID::FERMI_TWOD_A:
+        fermi_2d->WriteReg(method, value);
+        break;
+    case EngineID::MAXWELL_B:
+        maxwell_3d->WriteReg(method, value);
+        break;
+    case EngineID::MAXWELL_COMPUTE_B:
+        maxwell_compute->WriteReg(method, value);
+        break;
+    default:
+        UNIMPLEMENTED();
+    }
+}
+
+void GPU::ProcessCommandList(GPUVAddr address, u32 size) {
+    // TODO(Subv): PhysicalToVirtualAddress is a misnomer, it converts a GPU VAddr into an
+    // application VAddr.
+    const VAddr head_address = memory_manager->PhysicalToVirtualAddress(address);
+    VAddr current_addr = head_address;
+    while (current_addr < head_address + size * sizeof(CommandHeader)) {
+        const CommandHeader header = {Memory::Read32(current_addr)};
+        current_addr += sizeof(u32);
+
+        switch (header.mode.Value()) {
+        case SubmissionMode::IncreasingOld:
+        case SubmissionMode::Increasing: {
+            // Increase the method value with each argument.
+            for (unsigned i = 0; i < header.arg_count; ++i) {
+                WriteReg(header.method + i, header.subchannel, Memory::Read32(current_addr));
+                current_addr += sizeof(u32);
+            }
+            break;
+        }
+        case SubmissionMode::NonIncreasingOld:
+        case SubmissionMode::NonIncreasing: {
+            // Use the same method value for all arguments.
+            for (unsigned i = 0; i < header.arg_count; ++i) {
+                WriteReg(header.method, header.subchannel, Memory::Read32(current_addr));
+                current_addr += sizeof(u32);
+            }
+            break;
+        }
+        case SubmissionMode::IncreaseOnce: {
+            ASSERT(header.arg_count.Value() >= 1);
+            // Use the original method for the first argument and then the next method for all other
+            // arguments.
+            WriteReg(header.method, header.subchannel, Memory::Read32(current_addr));
+            current_addr += sizeof(u32);
+            // Use the same method value for all arguments.
+            for (unsigned i = 1; i < header.arg_count; ++i) {
+                WriteReg(header.method + 1, header.subchannel, Memory::Read32(current_addr));
+                current_addr += sizeof(u32);
+            }
+            break;
+        }
+        case SubmissionMode::Inline: {
+            // The register value is stored in the bits 16-28 as an immediate
+            WriteReg(header.method, header.subchannel, header.inline_data);
+            break;
+        }
+        default:
+            UNIMPLEMENTED();
+        }
+    }
+}
+
+} // namespace Tegra

src/video_core/command_processor.h

@@ -0,0 +1,39 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <type_traits>
+#include "common/bit_field.h"
+#include "common/common_types.h"
+
+namespace Tegra {
+
+enum class SubmissionMode : u32 {
+    IncreasingOld = 0,
+    Increasing = 1,
+    NonIncreasingOld = 2,
+    NonIncreasing = 3,
+    Inline = 4,
+    IncreaseOnce = 5
+};
+
+union CommandHeader {
+    u32 hex;
+
+    BitField<0, 13, u32> method;
+    BitField<13, 3, u32> subchannel;
+
+    BitField<16, 13, u32> arg_count;
+    BitField<16, 13, u32> inline_data;
+
+    BitField<29, 3, SubmissionMode> mode;
+};
+static_assert(std::is_standard_layout<CommandHeader>::value == true,
+              "CommandHeader does not use standard layout");
+static_assert(sizeof(CommandHeader) == sizeof(u32), "CommandHeader has incorrect size!");
+
+void ProcessCommandList(VAddr address, u32 size);
+
+} // namespace Tegra

src/video_core/engines/fermi_2d.cpp

@@ -0,0 +1,13 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/engines/fermi_2d.h"
+
+namespace Tegra {
+namespace Engines {
+
+void Fermi2D::WriteReg(u32 method, u32 value) {}
+
+} // namespace Engines
+} // namespace Tegra

src/video_core/engines/fermi_2d.h

@@ -0,0 +1,22 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Tegra {
+namespace Engines {
+
+class Fermi2D final {
+public:
+    Fermi2D() = default;
+    ~Fermi2D() = default;
+
+    /// Write the value to the register identified by method.
+    void WriteReg(u32 method, u32 value);
+};
+
+} // namespace Engines
+} // namespace Tegra

src/video_core/engines/maxwell_3d.cpp

@@ -0,0 +1,13 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/engines/maxwell_3d.h"
+
+namespace Tegra {
+namespace Engines {
+
+void Maxwell3D::WriteReg(u32 method, u32 value) {}
+
+} // namespace Engines
+} // namespace Tegra

src/video_core/engines/maxwell_3d.h

@@ -0,0 +1,22 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Tegra {
+namespace Engines {
+
+class Maxwell3D final {
+public:
+    Maxwell3D() = default;
+    ~Maxwell3D() = default;
+
+    /// Write the value to the register identified by method.
+    void WriteReg(u32 method, u32 value);
+};
+
+} // namespace Engines
+} // namespace Tegra

src/video_core/engines/maxwell_compute.cpp

@@ -0,0 +1,13 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/engines/maxwell_compute.h"
+
+namespace Tegra {
+namespace Engines {
+
+void MaxwellCompute::WriteReg(u32 method, u32 value) {}
+
+} // namespace Engines
+} // namespace Tegra

src/video_core/engines/maxwell_compute.h

@@ -0,0 +1,22 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Tegra {
+namespace Engines {
+
+class MaxwellCompute final {
+public:
+    MaxwellCompute() = default;
+    ~MaxwellCompute() = default;
+
+    /// Write the value to the register identified by method.
+    void WriteReg(u32 method, u32 value);
+};
+
+} // namespace Engines
+} // namespace Tegra

src/video_core/gpu.h

@@ -0,0 +1,55 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <unordered_map>
+#include "common/common_types.h"
+#include "video_core/engines/fermi_2d.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/engines/maxwell_compute.h"
+#include "video_core/memory_manager.h"
+
+namespace Tegra {
+
+enum class EngineID {
+    FERMI_TWOD_A = 0x902D, // 2D Engine
+    MAXWELL_B = 0xB197,    // 3D Engine
+    MAXWELL_COMPUTE_B = 0xB1C0,
+    KEPLER_INLINE_TO_MEMORY_B = 0xA140,
+    MAXWELL_DMA_COPY_A = 0xB0B5,
+};
+
+class GPU final {
+public:
+    GPU() {
+        memory_manager = std::make_unique<MemoryManager>();
+        maxwell_3d = std::make_unique<Engines::Maxwell3D>();
+        fermi_2d = std::make_unique<Engines::Fermi2D>();
+        maxwell_compute = std::make_unique<Engines::MaxwellCompute>();
+    }
+    ~GPU() = default;
+
+    /// Processes a command list stored at the specified address in GPU memory.
+    void ProcessCommandList(GPUVAddr address, u32 size);
+
+    std::unique_ptr<MemoryManager> memory_manager;
+
+private:
+    /// Writes a single register in the engine bound to the specified subchannel
+    void WriteReg(u32 method, u32 subchannel, u32 value);
+
+    /// Mapping of command subchannels to their bound engine ids.
+    std::unordered_map<u32, EngineID> bound_engines;
+
+    /// 3D engine
+    std::unique_ptr<Engines::Maxwell3D> maxwell_3d;
+    /// 2D engine
+    std::unique_ptr<Engines::Fermi2D> fermi_2d;
+    /// Compute engine
+    std::unique_ptr<Engines::MaxwellCompute> maxwell_compute;
+};
+
+} // namespace Tegra

src/video_core/memory_manager.cpp

@@ -0,0 +1,110 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "video_core/memory_manager.h"
+
+namespace Tegra {
+
+PAddr MemoryManager::AllocateSpace(u64 size, u64 align) {
+    boost::optional<PAddr> paddr = FindFreeBlock(size, align);
+    ASSERT(paddr);
+
+    for (u64 offset = 0; offset < size; offset += Memory::PAGE_SIZE) {
+        PageSlot(*paddr + offset) = static_cast<u64>(PageStatus::Allocated);
+    }
+
+    return *paddr;
+}
+
+PAddr MemoryManager::AllocateSpace(PAddr paddr, u64 size, u64 align) {
+    for (u64 offset = 0; offset < size; offset += Memory::PAGE_SIZE) {
+        if (IsPageMapped(paddr + offset)) {
+            return AllocateSpace(size, align);
+        }
+    }
+
+    for (u64 offset = 0; offset < size; offset += Memory::PAGE_SIZE) {
+        PageSlot(paddr + offset) = static_cast<u64>(PageStatus::Allocated);
+    }
+
+    return paddr;
+}
+
+PAddr MemoryManager::MapBufferEx(VAddr vaddr, u64 size) {
+    vaddr &= ~Memory::PAGE_MASK;
+
+    boost::optional<PAddr> paddr = FindFreeBlock(size);
+    ASSERT(paddr);
+
+    for (u64 offset = 0; offset < size; offset += Memory::PAGE_SIZE) {
+        PageSlot(*paddr + offset) = vaddr + offset;
+    }
+
+    return *paddr;
+}
+
+PAddr MemoryManager::MapBufferEx(VAddr vaddr, PAddr paddr, u64 size) {
+    vaddr &= ~Memory::PAGE_MASK;
+    paddr &= ~Memory::PAGE_MASK;
+
+    for (u64 offset = 0; offset < size; offset += Memory::PAGE_SIZE) {
+        if (PageSlot(paddr + offset) != static_cast<u64>(PageStatus::Allocated)) {
+            return MapBufferEx(vaddr, size);
+        }
+    }
+
+    for (u64 offset = 0; offset < size; offset += Memory::PAGE_SIZE) {
+        PageSlot(paddr + offset) = vaddr + offset;
+    }
+
+    return paddr;
+}
+
+boost::optional<PAddr> MemoryManager::FindFreeBlock(u64 size, u64 align) {
+    PAddr paddr{};
+    u64 free_space{};
+    align = (align + Memory::PAGE_MASK) & ~Memory::PAGE_MASK;
+
+    while (paddr + free_space < MAX_ADDRESS) {
+        if (!IsPageMapped(paddr + free_space)) {
+            free_space += Memory::PAGE_SIZE;
+            if (free_space >= size) {
+                return paddr;
+            }
+        } else {
+            paddr += free_space + Memory::PAGE_SIZE;
+            free_space = 0;
+            const u64 remainder{paddr % align};
+            if (!remainder) {
+                paddr = (paddr - remainder) + align;
+            }
+        }
+    }
+
+    return {};
+}
+
+VAddr MemoryManager::PhysicalToVirtualAddress(PAddr paddr) {
+    VAddr base_addr = PageSlot(paddr);
+    ASSERT(base_addr != static_cast<u64>(PageStatus::Unmapped));
+    return base_addr + (paddr & Memory::PAGE_MASK);
+}
+
+bool MemoryManager::IsPageMapped(PAddr paddr) {
+    return PageSlot(paddr) != static_cast<u64>(PageStatus::Unmapped);
+}
+
+VAddr& MemoryManager::PageSlot(PAddr paddr) {
+    auto& block = page_table[(paddr >> (Memory::PAGE_BITS + PAGE_TABLE_BITS)) & PAGE_TABLE_MASK];
+    if (!block) {
+        block = std::make_unique<PageBlock>();
+        for (unsigned index = 0; index < PAGE_BLOCK_SIZE; index++) {
+            (*block)[index] = static_cast<u64>(PageStatus::Unmapped);
+        }
+    }
+    return (*block)[(paddr >> Memory::PAGE_BITS) & PAGE_BLOCK_MASK];
+}
+
+} // namespace Tegra

src/video_core/memory_manager.h

@@ -0,0 +1,49 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <memory>
+#include "common/common_types.h"
+#include "core/memory.h"
+
+namespace Tegra {
+
+/// Virtual addresses in the GPU's memory map are 64 bit.
+using GPUVAddr = u64;
+
+class MemoryManager final {
+public:
+    MemoryManager() = default;
+
+    PAddr AllocateSpace(u64 size, u64 align);
+    PAddr AllocateSpace(PAddr paddr, u64 size, u64 align);
+    PAddr MapBufferEx(VAddr vaddr, u64 size);
+    PAddr MapBufferEx(VAddr vaddr, PAddr paddr, u64 size);
+    VAddr PhysicalToVirtualAddress(PAddr paddr);
+
+private:
+    boost::optional<PAddr> FindFreeBlock(u64 size, u64 align = 1);
+    bool IsPageMapped(PAddr paddr);
+    VAddr& PageSlot(PAddr paddr);
+
+    enum class PageStatus : u64 {
+        Unmapped = 0xFFFFFFFFFFFFFFFFULL,
+        Allocated = 0xFFFFFFFFFFFFFFFEULL,
+    };
+
+    static constexpr u64 MAX_ADDRESS{0x10000000000ULL};
+    static constexpr u64 PAGE_TABLE_BITS{14};
+    static constexpr u64 PAGE_TABLE_SIZE{1 << PAGE_TABLE_BITS};
+    static constexpr u64 PAGE_TABLE_MASK{PAGE_TABLE_SIZE - 1};
+    static constexpr u64 PAGE_BLOCK_BITS{14};
+    static constexpr u64 PAGE_BLOCK_SIZE{1 << PAGE_BLOCK_BITS};
+    static constexpr u64 PAGE_BLOCK_MASK{PAGE_BLOCK_SIZE - 1};
+
+    using PageBlock = std::array<VAddr, PAGE_BLOCK_SIZE>;
+    std::array<std::unique_ptr<PageBlock>, PAGE_TABLE_SIZE> page_table{};
+};
+
+} // namespace Tegra

src/video_core/renderer_base.h

@@ -43,6 +43,7 @@ public:
         u32 height;
         u32 stride;
         PixelFormat pixel_format;
+        bool flip_vertical;
     };
 
     virtual ~RendererBase() {}

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -262,6 +262,8 @@ void RendererOpenGL::LoadFBToScreenInfo(const FramebufferInfo& framebuffer_info,
     // only allows rows to have a memory alignement of 4.
     ASSERT(framebuffer_info.stride % 4 == 0);
 
+    framebuffer_flip_vertical = framebuffer_info.flip_vertical;
+
     // Reset the screen info's display texture to its own permanent texture
     screen_info.display_texture = screen_info.texture.resource.handle;
     screen_info.display_texcoords = MathUtil::Rectangle<float>(0.f, 0.f, 1.f, 1.f);
@@ -401,13 +403,15 @@ void RendererOpenGL::ConfigureFramebufferTexture(TextureInfo& texture,
 
 void RendererOpenGL::DrawSingleScreen(const ScreenInfo& screen_info, float x, float y, float w,
                                       float h) {
-    auto& texcoords = screen_info.display_texcoords;
+    const auto& texcoords = screen_info.display_texcoords;
+    const auto& left = framebuffer_flip_vertical ? texcoords.right : texcoords.left;
+    const auto& right = framebuffer_flip_vertical ? texcoords.left : texcoords.right;
 
     std::array<ScreenRectVertex, 4> vertices = {{
-        ScreenRectVertex(x, y, texcoords.top, texcoords.right),
-        ScreenRectVertex(x + w, y, texcoords.bottom, texcoords.right),
-        ScreenRectVertex(x, y + h, texcoords.top, texcoords.left),
-        ScreenRectVertex(x + w, y + h, texcoords.bottom, texcoords.left),
+        ScreenRectVertex(x, y, texcoords.top, right),
+        ScreenRectVertex(x + w, y, texcoords.bottom, right),
+        ScreenRectVertex(x, y + h, texcoords.top, left),
+        ScreenRectVertex(x + w, y + h, texcoords.bottom, left),
     }};
 
     state.texture_units[0].texture_2d = screen_info.display_texture;

src/video_core/renderer_opengl/renderer_opengl.h

@@ -86,4 +86,7 @@ private:
     // Shader attribute input indices
     GLuint attrib_position;
     GLuint attrib_tex_coord;
+
+    /// Flips the framebuffer vertically when true
+    bool framebuffer_flip_vertical;
 };