"Merge Tagged PR 1012"

src/common/assert.h

@@ -28,22 +28,19 @@ __declspec(noinline, noreturn)
 }
 
 #define ASSERT(_a_)                                                                                \
-    do                                                                                             \
-        if (!(_a_)) {                                                                              \
-            assert_noinline_call([] { LOG_CRITICAL(Debug, "Assertion Failed!"); });                \
-        }                                                                                          \
-    while (0)
+    if (!(_a_)) {                                                                                  \
+        LOG_CRITICAL(Debug, "Assertion Failed!");                                                  \
+    }
 
 #define ASSERT_MSG(_a_, ...)                                                                       \
-    do                                                                                             \
-        if (!(_a_)) {                                                                              \
-            assert_noinline_call([&] { LOG_CRITICAL(Debug, "Assertion Failed!\n" __VA_ARGS__); }); \
-        }                                                                                          \
-    while (0)
+    if (!(_a_)) {                                                                                  \
+        LOG_CRITICAL(Debug, "Assertion Failed! " __VA_ARGS__);                                     \
+    }
 
-#define UNREACHABLE() assert_noinline_call([] { LOG_CRITICAL(Debug, "Unreachable code!"); })
+#define UNREACHABLE()                                                                              \
+    { LOG_CRITICAL(Debug, "Unreachable code!"); }
 #define UNREACHABLE_MSG(...)                                                                       \
-    assert_noinline_call([&] { LOG_CRITICAL(Debug, "Unreachable code!\n" __VA_ARGS__); })
+    { LOG_CRITICAL(Debug, "Unreachable code!\n" __VA_ARGS__); }
 
 #ifdef _DEBUG
 #define DEBUG_ASSERT(_a_) ASSERT(_a_)

src/common/threadsafe_queue.h

@@ -46,9 +46,8 @@ public:
         ElementPtr* new_ptr = new ElementPtr();
         write_ptr->next.store(new_ptr, std::memory_order_release);
         write_ptr = new_ptr;
-        cv.notify_one();
-
         ++size;
+        cv.notify_one();
     }
 
     void Pop() {

src/core/hle/kernel/mutex.cpp

@@ -7,6 +7,7 @@
 
 #include "common/assert.h"
 #include "core/core.h"
+#include "core/core_cpu.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/kernel.h"
@@ -23,7 +24,7 @@ namespace Kernel {
 /// Returns the number of threads that are waiting for a mutex, and the highest priority one among
 /// those.
 static std::pair<std::shared_ptr<Thread>, u32> GetHighestPriorityMutexWaitingThread(
-    const std::shared_ptr<Thread>& current_thread, VAddr mutex_addr) {
+    Thread* current_thread, VAddr mutex_addr) {
 
     std::shared_ptr<Thread> highest_priority_thread;
     u32 num_waiters = 0;
@@ -45,16 +46,15 @@ static std::pair<std::shared_ptr<Thread>, u32> GetHighestPriorityMutexWaitingThr
 }
 
 /// Update the mutex owner field of all threads waiting on the mutex to point to the new owner.
-static void TransferMutexOwnership(VAddr mutex_addr, std::shared_ptr<Thread> current_thread,
-                                   std::shared_ptr<Thread> new_owner) {
+static void TransferMutexOwnership(VAddr mutex_addr, Thread* current_thread, Thread* new_owner) {
     const auto threads = current_thread->GetMutexWaitingThreads();
     for (const auto& thread : threads) {
         if (thread->GetMutexWaitAddress() != mutex_addr)
             continue;
 
-        ASSERT(thread->GetLockOwner() == current_thread.get());
+        ASSERT(thread->GetLockOwner() == current_thread);
         current_thread->RemoveMutexWaiter(thread);
-        if (new_owner != thread)
+        if (new_owner != thread.get())
             new_owner->AddMutexWaiter(thread);
     }
 }
@@ -79,7 +79,7 @@ ResultCode Mutex::TryAcquire(VAddr address, Handle holding_thread_handle,
     // thread.
     ASSERT(requesting_thread == current_thread);
 
-    const u32 addr_value = system.Memory().Read32(address);
+    u32 addr_value = system.Memory().Read32(address);
 
     // If the mutex isn't being held, just return success.
     if (addr_value != (holding_thread_handle | Mutex::MutexHasWaitersFlag)) {
@@ -90,6 +90,20 @@ ResultCode Mutex::TryAcquire(VAddr address, Handle holding_thread_handle,
         return ERR_INVALID_HANDLE;
     }
 
+    // This a workaround where an unknown bug writes the mutex value to give ownership to a cond var
+    // waiting thread.
+    if (holding_thread->GetStatus() == ThreadStatus::WaitCondVar) {
+        if (holding_thread->GetMutexWaitAddress() == address) {
+            Release(address, holding_thread.get());
+            addr_value = system.Memory().Read32(address);
+            if (addr_value == 0)
+                return RESULT_SUCCESS;
+            else {
+                holding_thread = handle_table.Get<Thread>(addr_value & Mutex::MutexOwnerMask);
+            }
+        }
+    }
+
     // Wait until the mutex is released
     current_thread->SetMutexWaitAddress(address);
     current_thread->SetWaitHandle(requesting_thread_handle);
@@ -105,15 +119,13 @@ ResultCode Mutex::TryAcquire(VAddr address, Handle holding_thread_handle,
     return RESULT_SUCCESS;
 }
 
-ResultCode Mutex::Release(VAddr address) {
+ResultCode Mutex::Release(VAddr address, Thread* holding_thread) {
     // The mutex address must be 4-byte aligned
     if ((address % sizeof(u32)) != 0) {
         return ERR_INVALID_ADDRESS;
     }
 
-    std::shared_ptr<Thread> current_thread =
-        SharedFrom(system.CurrentScheduler().GetCurrentThread());
-    auto [thread, num_waiters] = GetHighestPriorityMutexWaitingThread(current_thread, address);
+    auto [thread, num_waiters] = GetHighestPriorityMutexWaitingThread(holding_thread, address);
 
     // There are no more threads waiting for the mutex, release it completely.
     if (thread == nullptr) {
@@ -122,7 +134,7 @@ ResultCode Mutex::Release(VAddr address) {
     }
 
     // Transfer the ownership of the mutex from the previous owner to the new one.
-    TransferMutexOwnership(address, current_thread, thread);
+    TransferMutexOwnership(address, holding_thread, thread.get());
 
     u32 mutex_value = thread->GetWaitHandle();
 
@@ -143,7 +155,10 @@ ResultCode Mutex::Release(VAddr address) {
     thread->SetWaitHandle(0);
     thread->SetWaitSynchronizationResult(RESULT_SUCCESS);
 
-    system.PrepareReschedule();
+    if (thread->GetProcessorID() >= 0)
+        system.CpuCore(thread->GetProcessorID()).PrepareReschedule();
+    if (holding_thread->GetProcessorID() >= 0)
+        system.CpuCore(holding_thread->GetProcessorID()).PrepareReschedule();
 
     return RESULT_SUCCESS;
 }

src/core/hle/kernel/mutex.h

@@ -29,7 +29,7 @@ public:
                           Handle requesting_thread_handle);
 
     /// Releases the mutex at the specified address.
-    ResultCode Release(VAddr address);
+    ResultCode Release(VAddr address, Thread* holding_thread);
 
 private:
     Core::System& system;

src/core/hle/kernel/svc.cpp

@@ -588,7 +588,8 @@ static ResultCode ArbitrateUnlock(Core::System& system, VAddr mutex_addr) {
     }
 
     auto* const current_process = system.Kernel().CurrentProcess();
-    return current_process->GetMutex().Release(mutex_addr);
+    return current_process->GetMutex().Release(mutex_addr,
+                                               system.CurrentScheduler().GetCurrentThread());
 }
 
 enum class BreakType : u32 {
@@ -1628,12 +1629,13 @@ static ResultCode WaitProcessWideKeyAtomic(Core::System& system, VAddr mutex_add
     std::shared_ptr<Thread> thread = handle_table.Get<Thread>(thread_handle);
     ASSERT(thread);
 
-    const auto release_result = current_process->GetMutex().Release(mutex_addr);
+    Thread* current_thread = system.CurrentScheduler().GetCurrentThread();
+
+    const auto release_result = current_process->GetMutex().Release(mutex_addr, current_thread);
     if (release_result.IsError()) {
         return release_result;
     }
 
-    Thread* current_thread = system.CurrentScheduler().GetCurrentThread();
     current_thread->SetCondVarWaitAddress(condition_variable_addr);
     current_thread->SetMutexWaitAddress(mutex_addr);
     current_thread->SetWaitHandle(thread_handle);

src/core/hle/kernel/transfer_memory.cpp

@@ -23,6 +23,8 @@ std::shared_ptr<TransferMemory> TransferMemory::Create(KernelCore& kernel, VAddr
     transfer_memory->owner_permissions = permissions;
     transfer_memory->owner_process = kernel.CurrentProcess();
 
+    transfer_memory->MapMemory(base_address, size, permissions);
+
     return transfer_memory;
 }
 

src/core/hle/kernel/wait_object.cpp

@@ -51,17 +51,8 @@ std::shared_ptr<Thread> WaitObject::GetHighestPriorityReadyThread() const {
         if (ShouldWait(thread.get()))
             continue;
 
-        // A thread is ready to run if it's either in ThreadStatus::WaitSynch
-        // and the rest of the objects it is waiting on are ready.
-        bool ready_to_run = true;
-        if (thread_status == ThreadStatus::WaitSynch) {
-            ready_to_run = thread->AllWaitObjectsReady();
-        }
-
-        if (ready_to_run) {
-            candidate = thread.get();
-            candidate_priority = thread->GetPriority();
-        }
+        candidate = thread.get();
+        candidate_priority = thread->GetPriority();
     }
 
     return SharedFrom(candidate);

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

@@ -857,17 +857,16 @@ private:
     void PopInteractiveOutData(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_AM, "called");
 
-        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-
         const auto storage = applet->GetBroker().PopInteractiveDataToGame();
         if (storage == nullptr) {
             LOG_ERROR(Service_AM,
                       "storage is a nullptr. There is no data in the current interactive channel");
-
+            IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ERR_NO_DATA_IN_CHANNEL);
             return;
         }
 
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
         rb.PushIpcInterface<IStorage>(std::move(*storage));
     }

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

@@ -88,6 +88,7 @@ std::unique_ptr<IStorage> AppletDataBroker::PopInteractiveDataToApplet() {
 
 void AppletDataBroker::PushNormalDataFromGame(IStorage storage) {
     in_channel.push_back(std::make_unique<IStorage>(storage));
+    pop_out_data_event.writable->Clear();
 }
 
 void AppletDataBroker::PushNormalDataFromApplet(IStorage storage) {
@@ -97,6 +98,7 @@ void AppletDataBroker::PushNormalDataFromApplet(IStorage storage) {
 
 void AppletDataBroker::PushInteractiveDataFromGame(IStorage storage) {
     in_interactive_channel.push_back(std::make_unique<IStorage>(storage));
+    pop_interactive_out_data_event.writable->Clear();
 }
 
 void AppletDataBroker::PushInteractiveDataFromApplet(IStorage storage) {

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

@@ -91,6 +91,7 @@ void SoftwareKeyboard::ExecuteInteractive() {
 
     if (status == INTERACTIVE_STATUS_OK) {
         complete = true;
+        broker.SignalStateChanged();
     } else {
         std::array<char16_t, SWKBD_OUTPUT_INTERACTIVE_BUFFER_SIZE / 2 - 2> string;
         std::memcpy(string.data(), data.data() + 4, string.size() * 2);

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

@@ -40,7 +40,10 @@ static FileSys::VirtualDir GetDirectoryRelativeWrapped(FileSys::VirtualDir base,
     if (dir_name.empty() || dir_name == "." || dir_name == "/" || dir_name == "\\")
         return base;
 
-    return base->GetDirectoryRelative(dir_name);
+    const auto res = base->GetDirectoryRelative(dir_name);
+    if (res == nullptr)
+        return base->CreateDirectoryRelative(dir_name);
+    return res;
 }
 
 VfsDirectoryServiceWrapper::VfsDirectoryServiceWrapper(FileSys::VirtualDir backing_)

src/core/hle/service/nifm/nifm.cpp

@@ -9,6 +9,7 @@
 #include "core/hle/kernel/writable_event.h"
 #include "core/hle/service/nifm/nifm.h"
 #include "core/hle/service/service.h"
+#include "core/settings.h"
 
 namespace Service::NIFM {
 
@@ -86,7 +87,12 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(RESULT_SUCCESS);
-        rb.PushEnum(RequestState::Connected);
+
+        if (Settings::values.bcat_backend == "none") {
+            rb.PushEnum(RequestState::NotSubmitted);
+        } else {
+            rb.PushEnum(RequestState::Connected);
+        }
     }
 
     void GetResult(Kernel::HLERequestContext& ctx) {
@@ -194,14 +200,22 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(RESULT_SUCCESS);
-        rb.Push<u8>(1);
+        if (Settings::values.bcat_backend == "none") {
+            rb.Push<u8>(0);
+        } else {
+            rb.Push<u8>(1);
+        }
     }
     void IsAnyInternetRequestAccepted(Kernel::HLERequestContext& ctx) {
         LOG_WARNING(Service_NIFM, "(STUBBED) called");
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(RESULT_SUCCESS);
-        rb.Push<u8>(1);
+        if (Settings::values.bcat_backend == "none") {
+            rb.Push<u8>(0);
+        } else {
+            rb.Push<u8>(1);
+        }
     }
     Core::System& system;
 };

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

@@ -22,6 +22,18 @@ u32 nvhost_nvdec::ioctl(Ioctl command, const std::vector<u8>& input, const std::
     switch (static_cast<IoctlCommand>(command.raw)) {
     case IoctlCommand::IocSetNVMAPfdCommand:
         return SetNVMAPfd(input, output);
+    case IoctlCommand::IocSubmit:
+        return Submit(input, output);
+    case IoctlCommand::IocGetSyncpoint:
+        return GetSyncpoint(input, output);
+    case IoctlCommand::IocGetWaitbase:
+        return GetWaitbase(input, output);
+    case IoctlCommand::IocMapBuffer:
+        return MapBuffer(input, output);
+    case IoctlCommand::IocMapBufferEx:
+        return MapBufferEx(input, output);
+    case IoctlCommand::IocUnmapBufferEx:
+        return UnmapBufferEx(input, output);
     }
 
     UNIMPLEMENTED_MSG("Unimplemented ioctl");
@@ -30,11 +42,67 @@ u32 nvhost_nvdec::ioctl(Ioctl command, const std::vector<u8>& input, const std::
 
 u32 nvhost_nvdec::SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output) {
     IoctlSetNvmapFD params{};
-    std::memcpy(&params, input.data(), input.size());
+    std::memcpy(&params, input.data(), sizeof(IoctlSetNvmapFD));
     LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
 
     nvmap_fd = params.nvmap_fd;
     return 0;
 }
 
+u32 nvhost_nvdec::Submit(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlSubmit params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSubmit));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
+    return 0;
+}
+
+u32 nvhost_nvdec::GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlGetSyncpoint params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlGetSyncpoint));
+    LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
+    params.value = 0; // Seems to be hard coded at 0
+    std::memcpy(output.data(), &params, sizeof(IoctlGetSyncpoint));
+    return 0;
+}
+
+u32 nvhost_nvdec::GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlGetWaitbase params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlGetWaitbase));
+    LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
+    params.value = 0; // Seems to be hard coded at 0
+    std::memcpy(output.data(), &params, sizeof(IoctlGetWaitbase));
+    return 0;
+}
+
+u32 nvhost_nvdec::MapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlMapBuffer params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
+                params.address_1);
+    params.address_1 = 0;
+    params.address_2 = 0;
+    std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
+    return 0;
+}
+
+u32 nvhost_nvdec::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlMapBufferEx params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlMapBufferEx));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
+                params.address_1);
+    params.address_1 = 0;
+    params.address_2 = 0;
+    std::memcpy(output.data(), &params, sizeof(IoctlMapBufferEx));
+    return 0;
+}
+
+u32 nvhost_nvdec::UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlUnmapBufferEx params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlUnmapBufferEx));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    std::memcpy(output.data(), &params, sizeof(IoctlUnmapBufferEx));
+    return 0;
+}
+
 } // namespace Service::Nvidia::Devices

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

@@ -23,16 +23,66 @@ public:
 private:
     enum class IoctlCommand : u32_le {
         IocSetNVMAPfdCommand = 0x40044801,
+        IocSubmit = 0xC0400001,
+        IocGetSyncpoint = 0xC0080002,
+        IocGetWaitbase = 0xC0080003,
+        IocMapBuffer = 0xC01C0009,
+        IocMapBufferEx = 0xC0A40009,
+        IocUnmapBufferEx = 0xC0A4000A,
     };
 
     struct IoctlSetNvmapFD {
         u32_le nvmap_fd;
     };
-    static_assert(sizeof(IoctlSetNvmapFD) == 4, "IoctlSetNvmapFD is incorrect size");
+    static_assert(sizeof(IoctlSetNvmapFD) == 0x4, "IoctlSetNvmapFD is incorrect size");
+
+    struct IoctlSubmit {
+        INSERT_PADDING_BYTES(0x40); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlSubmit) == 0x40, "IoctlSubmit has incorrect size");
+
+    struct IoctlGetSyncpoint {
+        u32 unknown; // seems to be ignored? Nintendo added this
+        u32 value;
+    };
+    static_assert(sizeof(IoctlGetSyncpoint) == 0x08, "IoctlGetSyncpoint has incorrect size");
+
+    struct IoctlGetWaitbase {
+        u32 unknown; // seems to be ignored? Nintendo added this
+        u32 value;
+    };
+    static_assert(sizeof(IoctlGetWaitbase) == 0x08, "IoctlGetWaitbase has incorrect size");
+
+    struct IoctlMapBuffer {
+        u32 unknown;
+        u32 address_1;
+        u32 address_2;
+        INSERT_PADDING_BYTES(0x10); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlMapBuffer) == 0x1C, "IoctlMapBuffer is incorrect size");
+
+    struct IoctlMapBufferEx {
+        u32 unknown;
+        u32 address_1;
+        u32 address_2;
+        INSERT_PADDING_BYTES(0x98); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlMapBufferEx) == 0xA4, "IoctlMapBufferEx has incorrect size");
+
+    struct IoctlUnmapBufferEx {
+        INSERT_PADDING_BYTES(0xA4); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlUnmapBufferEx) == 0xA4, "IoctlUnmapBufferEx has incorrect size");
 
     u32_le nvmap_fd{};
 
     u32 SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 Submit(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 MapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
 };
 
 } // namespace Service::Nvidia::Devices

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

@@ -22,6 +22,18 @@ u32 nvhost_vic::ioctl(Ioctl command, const std::vector<u8>& input, const std::ve
     switch (static_cast<IoctlCommand>(command.raw)) {
     case IoctlCommand::IocSetNVMAPfdCommand:
         return SetNVMAPfd(input, output);
+    case IoctlCommand::IocSubmit:
+        return Submit(input, output);
+    case IoctlCommand::IocGetSyncpoint:
+        return GetSyncpoint(input, output);
+    case IoctlCommand::IocGetWaitbase:
+        return GetWaitbase(input, output);
+    case IoctlCommand::IocMapBuffer:
+        return MapBuffer(input, output);
+    case IoctlCommand::IocMapBufferEx:
+        return MapBuffer(input, output);
+    case IoctlCommand::IocUnmapBufferEx:
+        return UnmapBufferEx(input, output);
     }
 
     UNIMPLEMENTED_MSG("Unimplemented ioctl");
@@ -30,11 +42,67 @@ u32 nvhost_vic::ioctl(Ioctl command, const std::vector<u8>& input, const std::ve
 
 u32 nvhost_vic::SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output) {
     IoctlSetNvmapFD params{};
-    std::memcpy(&params, input.data(), input.size());
+    std::memcpy(&params, input.data(), sizeof(IoctlSetNvmapFD));
     LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
 
     nvmap_fd = params.nvmap_fd;
     return 0;
 }
 
+u32 nvhost_vic::Submit(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlSubmit params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSubmit));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
+    return 0;
+}
+
+u32 nvhost_vic::GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlGetSyncpoint params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlGetSyncpoint));
+    LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
+    params.value = 0; // Seems to be hard coded at 0
+    std::memcpy(output.data(), &params, sizeof(IoctlGetSyncpoint));
+    return 0;
+}
+
+u32 nvhost_vic::GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlGetWaitbase params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlGetWaitbase));
+    LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
+    params.value = 0; // Seems to be hard coded at 0
+    std::memcpy(output.data(), &params, sizeof(IoctlGetWaitbase));
+    return 0;
+}
+
+u32 nvhost_vic::MapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlMapBuffer params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
+                params.address_1);
+    params.address_1 = 0;
+    params.address_2 = 0;
+    std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
+    return 0;
+}
+
+u32 nvhost_vic::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlMapBufferEx params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlMapBufferEx));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
+                params.address_1);
+    params.address_1 = 0;
+    params.address_2 = 0;
+    std::memcpy(output.data(), &params, sizeof(IoctlMapBufferEx));
+    return 0;
+}
+
+u32 nvhost_vic::UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlUnmapBufferEx params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlUnmapBufferEx));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    std::memcpy(output.data(), &params, sizeof(IoctlUnmapBufferEx));
+    return 0;
+}
+
 } // namespace Service::Nvidia::Devices

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

@@ -23,6 +23,12 @@ public:
 private:
     enum class IoctlCommand : u32_le {
         IocSetNVMAPfdCommand = 0x40044801,
+        IocSubmit = 0xC0400001,
+        IocGetSyncpoint = 0xC0080002,
+        IocGetWaitbase = 0xC0080003,
+        IocMapBuffer = 0xC01C0009,
+        IocMapBufferEx = 0xC03C0009,
+        IocUnmapBufferEx = 0xC03C000A,
     };
 
     struct IoctlSetNvmapFD {
@@ -30,9 +36,53 @@ private:
     };
     static_assert(sizeof(IoctlSetNvmapFD) == 4, "IoctlSetNvmapFD is incorrect size");
 
+    struct IoctlSubmit {
+        INSERT_PADDING_BYTES(0x40); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlSubmit) == 0x40, "IoctlSubmit is incorrect size");
+
+    struct IoctlGetSyncpoint {
+        u32 unknown; // seems to be ignored? Nintendo added this
+        u32 value;
+    };
+    static_assert(sizeof(IoctlGetSyncpoint) == 0x8, "IoctlGetSyncpoint is incorrect size");
+
+    struct IoctlGetWaitbase {
+        u32 unknown; // seems to be ignored? Nintendo added this
+        u32 value;
+    };
+    static_assert(sizeof(IoctlGetWaitbase) == 0x8, "IoctlGetWaitbase is incorrect size");
+
+    struct IoctlMapBuffer {
+        u32 unknown;
+        u32 address_1;
+        u32 address_2;
+        INSERT_PADDING_BYTES(0x10); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlMapBuffer) == 0x1C, "IoctlMapBuffer is incorrect size");
+
+    struct IoctlMapBufferEx {
+        u32 unknown;
+        u32 address_1;
+        u32 address_2;
+        INSERT_PADDING_BYTES(0x30); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlMapBufferEx) == 0x3C, "IoctlMapBufferEx is incorrect size");
+
+    struct IoctlUnmapBufferEx {
+        INSERT_PADDING_BYTES(0x3C); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlUnmapBufferEx) == 0x3C, "IoctlUnmapBufferEx is incorrect size");
+
     u32_le nvmap_fd{};
 
     u32 SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 Submit(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 MapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
 };
 
 } // namespace Service::Nvidia::Devices

src/video_core/gpu_thread.cpp

@@ -19,8 +19,7 @@ static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_p
     MicroProfileOnThreadCreate("GpuThread");
 
     // Wait for first GPU command before acquiring the window context
-    while (state.queue.Empty())
-        ;
+    state.sync_event.Wait();
 
     // If emulation was stopped during disk shader loading, abort before trying to acquire context
     if (!state.is_running) {
@@ -31,22 +30,25 @@ static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_p
 
     CommandDataContainer next;
     while (state.is_running) {
-        next = state.queue.PopWait();
-        if (const auto submit_list = std::get_if<SubmitListCommand>(&next.data)) {
-            dma_pusher.Push(std::move(submit_list->entries));
-            dma_pusher.DispatchCalls();
-        } else if (const auto data = std::get_if<SwapBuffersCommand>(&next.data)) {
-            renderer.SwapBuffers(data->framebuffer ? &*data->framebuffer : nullptr);
-        } else if (const auto data = std::get_if<FlushRegionCommand>(&next.data)) {
-            renderer.Rasterizer().FlushRegion(data->addr, data->size);
-        } else if (const auto data = std::get_if<InvalidateRegionCommand>(&next.data)) {
-            renderer.Rasterizer().InvalidateRegion(data->addr, data->size);
-        } else if (std::holds_alternative<EndProcessingCommand>(next.data)) {
-            return;
-        } else {
-            UNREACHABLE();
+        while (!state.queue.Empty()) {
+            state.queue.Pop(next);
+            if (const auto submit_list = std::get_if<SubmitListCommand>(&next.data)) {
+                dma_pusher.Push(std::move(submit_list->entries));
+                dma_pusher.DispatchCalls();
+            } else if (const auto data = std::get_if<SwapBuffersCommand>(&next.data)) {
+                renderer.SwapBuffers(data->framebuffer ? &*data->framebuffer : nullptr);
+            } else if (const auto data = std::get_if<FlushRegionCommand>(&next.data)) {
+                renderer.Rasterizer().FlushRegion(data->addr, data->size);
+            } else if (const auto data = std::get_if<InvalidateRegionCommand>(&next.data)) {
+                renderer.Rasterizer().InvalidateRegion(data->addr, data->size);
+            } else if (std::holds_alternative<EndProcessingCommand>(next.data)) {
+                return;
+            } else {
+                UNREACHABLE();
+            }
+            state.signaled_fence.store(next.fence);
         }
-        state.signaled_fence.store(next.fence);
+        state.sync_event.Wait();
     }
 }
 
@@ -95,6 +97,7 @@ void ThreadManager::WaitIdle() const {
 u64 ThreadManager::PushCommand(CommandData&& command_data) {
     const u64 fence{++state.last_fence};
     state.queue.Push(CommandDataContainer(std::move(command_data), fence));
+    state.sync_event.Set();
     return fence;
 }
 

src/video_core/gpu_thread.h

@@ -11,6 +11,7 @@
 #include <thread>
 #include <variant>
 
+#include "common/thread.h"
 #include "common/threadsafe_queue.h"
 #include "video_core/gpu.h"
 
@@ -90,6 +91,7 @@ struct SynchState final {
     CommandQueue queue;
     u64 last_fence{};
     std::atomic<u64> signaled_fence{};
+    Common::Event sync_event{};
 };
 
 /// Class used to manage the GPU thread

src/video_core/texture_cache/surface_params.cpp

@@ -246,6 +246,16 @@ SurfaceParams SurfaceParams::CreateForFermiCopySurface(
     return params;
 }
 
+VideoCore::Surface::SurfaceTarget SurfaceParams::ExpectedTarget(
+    const VideoCommon::Shader::Sampler& entry) {
+    return TextureTypeToSurfaceTarget(entry.GetType(), entry.IsArray());
+}
+
+VideoCore::Surface::SurfaceTarget SurfaceParams::ExpectedTarget(
+    const VideoCommon::Shader::Image& entry) {
+    return ImageTypeToSurfaceTarget(entry.GetType());
+}
+
 bool SurfaceParams::IsLayered() const {
     switch (target) {
     case SurfaceTarget::Texture1DArray:

src/video_core/texture_cache/surface_params.h

@@ -45,6 +45,14 @@ public:
     static SurfaceParams CreateForFermiCopySurface(
         const Tegra::Engines::Fermi2D::Regs::Surface& config);
 
+    /// Obtains the texture target from a shader's sampler entry.
+    static VideoCore::Surface::SurfaceTarget ExpectedTarget(
+        const VideoCommon::Shader::Sampler& entry);
+
+    /// Obtains the texture target from a shader's sampler entry.
+    static VideoCore::Surface::SurfaceTarget ExpectedTarget(
+        const VideoCommon::Shader::Image& entry);
+
     std::size_t Hash() const {
         return static_cast<std::size_t>(
             Common::CityHash64(reinterpret_cast<const char*>(this), sizeof(*this)));

src/video_core/texture_cache/texture_cache.h

@@ -95,10 +95,16 @@ public:
         std::lock_guard lock{mutex};
         const auto gpu_addr{tic.Address()};
         if (!gpu_addr) {
-            return {};
+            return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
+        }
+
+        const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)};
+        const auto cache_addr{ToCacheAddr(host_ptr)};
+        if (!cache_addr) {
+            return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
         }
         const auto params{SurfaceParams::CreateForTexture(format_lookup_table, tic, entry)};
-        const auto [surface, view] = GetSurface(gpu_addr, params, true, false);
+        const auto [surface, view] = GetSurface(gpu_addr, cache_addr, params, true, false);
         if (guard_samplers) {
             sampled_textures.push_back(surface);
         }
@@ -110,10 +116,15 @@ public:
         std::lock_guard lock{mutex};
         const auto gpu_addr{tic.Address()};
         if (!gpu_addr) {
-            return {};
+            return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
+        }
+        const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)};
+        const auto cache_addr{ToCacheAddr(host_ptr)};
+        if (!cache_addr) {
+            return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
         }
         const auto params{SurfaceParams::CreateForImage(format_lookup_table, tic, entry)};
-        const auto [surface, view] = GetSurface(gpu_addr, params, true, false);
+        const auto [surface, view] = GetSurface(gpu_addr, cache_addr, params, true, false);
         if (guard_samplers) {
             sampled_textures.push_back(surface);
         }
@@ -143,11 +154,17 @@ public:
             SetEmptyDepthBuffer();
             return {};
         }
+        const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)};
+        const auto cache_addr{ToCacheAddr(host_ptr)};
+        if (!cache_addr) {
+            SetEmptyDepthBuffer();
+            return {};
+        }
         const auto depth_params{SurfaceParams::CreateForDepthBuffer(
             system, regs.zeta_width, regs.zeta_height, regs.zeta.format,
             regs.zeta.memory_layout.block_width, regs.zeta.memory_layout.block_height,
             regs.zeta.memory_layout.block_depth, regs.zeta.memory_layout.type)};
-        auto surface_view = GetSurface(gpu_addr, depth_params, preserve_contents, true);
+        auto surface_view = GetSurface(gpu_addr, cache_addr, depth_params, preserve_contents, true);
         if (depth_buffer.target)
             depth_buffer.target->MarkAsRenderTarget(false, NO_RT);
         depth_buffer.target = surface_view.first;
@@ -180,8 +197,16 @@ public:
             return {};
         }
 
-        auto surface_view = GetSurface(gpu_addr, SurfaceParams::CreateForFramebuffer(system, index),
-                                       preserve_contents, true);
+        const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)};
+        const auto cache_addr{ToCacheAddr(host_ptr)};
+        if (!cache_addr) {
+            SetEmptyColorBuffer(index);
+            return {};
+        }
+
+        auto surface_view =
+            GetSurface(gpu_addr, cache_addr, SurfaceParams::CreateForFramebuffer(system, index),
+                       preserve_contents, true);
         if (render_targets[index].target)
             render_targets[index].target->MarkAsRenderTarget(false, NO_RT);
         render_targets[index].target = surface_view.first;
@@ -230,8 +255,14 @@ public:
         const GPUVAddr src_gpu_addr = src_config.Address();
         const GPUVAddr dst_gpu_addr = dst_config.Address();
         DeduceBestBlit(src_params, dst_params, src_gpu_addr, dst_gpu_addr);
-        std::pair<TSurface, TView> dst_surface = GetSurface(dst_gpu_addr, dst_params, true, false);
-        std::pair<TSurface, TView> src_surface = GetSurface(src_gpu_addr, src_params, true, false);
+        const auto dst_host_ptr{system.GPU().MemoryManager().GetPointer(dst_gpu_addr)};
+        const auto dst_cache_addr{ToCacheAddr(dst_host_ptr)};
+        const auto src_host_ptr{system.GPU().MemoryManager().GetPointer(src_gpu_addr)};
+        const auto src_cache_addr{ToCacheAddr(src_host_ptr)};
+        std::pair<TSurface, TView> dst_surface =
+            GetSurface(dst_gpu_addr, dst_cache_addr, dst_params, true, false);
+        std::pair<TSurface, TView> src_surface =
+            GetSurface(src_gpu_addr, src_cache_addr, src_params, true, false);
         ImageBlit(src_surface.second, dst_surface.second, copy_config);
         dst_surface.first->MarkAsModified(true, Tick());
     }
@@ -347,13 +378,6 @@ protected:
         return new_surface;
     }
 
-    std::pair<TSurface, TView> GetFermiSurface(
-        const Tegra::Engines::Fermi2D::Regs::Surface& config) {
-        SurfaceParams params = SurfaceParams::CreateForFermiCopySurface(config);
-        const GPUVAddr gpu_addr = config.Address();
-        return GetSurface(gpu_addr, params, true, false);
-    }
-
     Core::System& system;
 
 private:
@@ -614,22 +638,9 @@ private:
      *                          left blank.
      * @param is_render         Whether or not the surface is a render target.
      **/
-    std::pair<TSurface, TView> GetSurface(const GPUVAddr gpu_addr, const SurfaceParams& params,
-                                          bool preserve_contents, bool is_render) {
-        const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)};
-        const auto cache_addr{ToCacheAddr(host_ptr)};
-
-        // Step 0: guarantee a valid surface
-        if (!cache_addr) {
-            // Return a null surface if it's invalid
-            SurfaceParams new_params = params;
-            new_params.width = 1;
-            new_params.height = 1;
-            new_params.depth = 1;
-            new_params.block_height = 0;
-            new_params.block_depth = 0;
-            return InitializeSurface(gpu_addr, new_params, false);
-        }
+    std::pair<TSurface, TView> GetSurface(const GPUVAddr gpu_addr, const CacheAddr cache_addr,
+                                          const SurfaceParams& params, bool preserve_contents,
+                                          bool is_render) {
 
         // Step 1
         // Check Level 1 Cache for a fast structural match. If candidate surface
@@ -793,6 +804,41 @@ private:
         }
     }
 
+    /**
+     * Gets a null surface based on a target texture.
+     * @param target The target of the null surface.
+     */
+    TView GetNullSurface(SurfaceTarget target) {
+        const u32 i_target = static_cast<u32>(target);
+        if (const auto it = invalid_cache.find(i_target); it != invalid_cache.end()) {
+            return it->second->GetMainView();
+        }
+        SurfaceParams params{};
+        params.target = target;
+        params.is_tiled = false;
+        params.srgb_conversion = false;
+        params.is_layered = false;
+        params.block_width = 0;
+        params.block_height = 0;
+        params.block_depth = 0;
+        params.tile_width_spacing = 1;
+        params.width = 1;
+        params.height = 1;
+        params.depth = 1;
+        params.pitch = 4;
+        params.num_levels = 1;
+        params.emulated_levels = 1;
+        params.pixel_format = VideoCore::Surface::PixelFormat::RGBA16F;
+        params.type = VideoCore::Surface::SurfaceType::ColorTexture;
+        auto surface = CreateSurface(0ULL, params);
+        invalid_memory.clear();
+        invalid_memory.resize(surface->GetHostSizeInBytes(), 0U);
+        surface->UploadTexture(invalid_memory);
+        surface->MarkAsModified(false, Tick());
+        invalid_cache.emplace(i_target, surface);
+        return surface->GetMainView();
+    }
+
     /**
      * Gets the a source and destination starting address and parameters,
      * and tries to deduce if they are supposed to be depth textures. If so, their
@@ -991,6 +1037,11 @@ private:
 
     std::vector<TSurface> sampled_textures;
 
+    /// This cache stores null surfaces in order to be used as a placeholder
+    /// for invalid texture calls.
+    std::unordered_map<u32, TSurface> invalid_cache;
+    std::vector<u8> invalid_memory;
+
     StagingCache staging_cache;
     std::recursive_mutex mutex;
 };