Cross platform thread implementation. (#41)

* Cross-platform thread implementation.

* Put set thread name calls behind a Win32 macro.

UnleashedRecomp/cpu/guest_thread.cpp

@@ -27,7 +27,7 @@ GuestThreadContext::GuestThreadContext(uint32_t cpuNumber)
     *(thread + 0x10C) = cpuNumber;
 
     *(uint32_t*)(thread + PCR_SIZE + 0x10) = 0xFFFFFFFF; // that one TLS entry that felt quirky
-    *(uint32_t*)(thread + PCR_SIZE + TLS_SIZE + 0x14C) = ByteSwap(GetCurrentThreadId()); // thread id
+    *(uint32_t*)(thread + PCR_SIZE + TLS_SIZE + 0x14C) = ByteSwap(GuestThread::GetCurrentThreadId()); // thread id
 
     ppcContext.fn = (uint8_t*)g_codeCache.bucket;
     ppcContext.r1.u64 = g_memory.MapVirtual(thread + PCR_SIZE + TLS_SIZE + TEB_SIZE + STACK_SIZE); // stack pointer
@@ -42,42 +42,82 @@ GuestThreadContext::~GuestThreadContext()
     g_userHeap.Free(thread);
 }
 
-DWORD GuestThread::Start(uint32_t function)
+static void GuestThreadFunc(GuestThreadHandle* hThread)
 {
-    const GuestThreadParameter parameter{ function };
+    hThread->suspended.wait(true);
-    return Start(parameter);
+    GuestThread::Start(hThread->params);
 }
 
-DWORD GuestThread::Start(const GuestThreadParameter& parameter)
+GuestThreadHandle::GuestThreadHandle(const GuestThreadParams& params)
+    : params(params), suspended((params.flags & 0x1) != 0), thread(GuestThreadFunc, this)
 {
-    const auto procMask = (uint8_t)(parameter.flags >> 24);
+}
+
+GuestThreadHandle::~GuestThreadHandle()
+{
+    if (thread.joinable())
+        thread.join();
+}
+
+void GuestThreadHandle::Wait(uint32_t timeout)
+{
+    assert(timeout == INFINITE);
+
+    if (thread.joinable())
+        thread.join();
+}
+
+uint32_t GuestThread::Start(const GuestThreadParams& params)
+{
+    const auto procMask = (uint8_t)(params.flags >> 24);
     const auto cpuNumber = procMask == 0 ? 0 : 7 - std::countl_zero(procMask);
 
     GuestThreadContext ctx(cpuNumber);
-    ctx.ppcContext.r3.u64 = parameter.value;
+    ctx.ppcContext.r3.u64 = params.value;
 
-    GuestCode::Run(g_codeCache.Find(parameter.function), &ctx.ppcContext, g_memory.Translate(0));
+    GuestCode::Run(g_codeCache.Find(params.function), &ctx.ppcContext, g_memory.Translate(0));
 
-    return (DWORD)ctx.ppcContext.r3.u64;
+    return ctx.ppcContext.r3.u32;
 }
 
-DWORD HostThreadStart(void* pParameter)
+static uint32_t GetThreadId(const std::thread::id& id)
 {
-    auto* parameter = static_cast<GuestThreadParameter*>(pParameter);
+    if constexpr (sizeof(id) == 4)
-    const auto result = GuestThread::Start(*parameter);
+        return *reinterpret_cast<const uint32_t*>(&id);
-
+    else
-    delete parameter;
+        return XXH32(&id, sizeof(id), 0);
-    return result;
 }
 
-HANDLE GuestThread::Start(uint32_t function, uint32_t parameter, uint32_t flags, LPDWORD threadId)
+GuestThreadHandle* GuestThread::Start(const GuestThreadParams& params, uint32_t* threadId)
 {
-    const auto hostCreationFlags = (flags & 1) != 0 ? CREATE_SUSPENDED : 0;
+    auto hThread = CreateKernelObject<GuestThreadHandle>(params);
-    //return CreateThread(nullptr, 0, Start, (void*)((uint64_t(parameter) << 32) | function), suspended ? CREATE_SUSPENDED : 0, threadId);
+
-    return CreateThread(nullptr, 0, HostThreadStart, new GuestThreadParameter{ function, parameter, flags }, hostCreationFlags, threadId);
+    if (threadId != nullptr)
+        *threadId = GetThreadId(hThread->thread.get_id());
+
+    return hThread;
 }
 
-void GuestThread::SetThreadName(uint32_t id, const char* name)
+uint32_t GuestThread::GetCurrentThreadId()
+{
+    return GetThreadId(std::this_thread::get_id());
+}
+
+void GuestThread::SetLastError(uint32_t error)
+{
+    auto* thread = (char*)g_memory.Translate(GetPPCContext()->r13.u32);
+    if (*(uint32_t*)(thread + 0x150))
+    {
+        // Program doesn't want errors
+        return;
+    }
+
+    // TEB + 0x160 : Win32LastError
+    *(uint32_t*)(thread + TEB_OFFSET + 0x160) = ByteSwap(error);
+}
+
+#ifdef _WIN32
+void GuestThread::SetThreadName(uint32_t threadId, const char* name)
 {
 #pragma pack(push,8)
     const DWORD MS_VC_EXCEPTION = 0x406D1388;
@@ -94,7 +134,7 @@ void GuestThread::SetThreadName(uint32_t id, const char* name)
     THREADNAME_INFO info;
     info.dwType = 0x1000;
     info.szName = name;
-    info.dwThreadID = id;
+    info.dwThreadID = threadId;
     info.dwFlags = 0;
 
     __try
@@ -105,41 +145,31 @@ void GuestThread::SetThreadName(uint32_t id, const char* name)
     {
     }
 }
-
+#endif
-void GuestThread::SetLastError(DWORD error)
-{
-    auto* thread = (char*)g_memory.Translate(GetPPCContext()->r13.u32);
-    if (*(DWORD*)(thread + 0x150))
-    {
-        // Program doesn't want errors
-        return;
-    }
-
-    // TEB + 0x160 : Win32LastError
-    *(DWORD*)(thread + TEB_OFFSET + 0x160) = ByteSwap(error);
-}
-
-PPCContext* GuestThread::Invoke(uint32_t address)
-{
-    auto* ctx = GetPPCContext();
-    GuestCode::Run(g_codeCache.Find(address), ctx);
-
-    return ctx;
-}
 
 void SetThreadNameImpl(uint32_t a1, uint32_t threadId, uint32_t* name)
 {
+#ifdef _WIN32
     GuestThread::SetThreadName(threadId, (const char*)g_memory.Translate(ByteSwap(*name)));
+#endif
 }
 
-int GetThreadPriorityImpl(uint32_t hThread)
+int GetThreadPriorityImpl(GuestThreadHandle* hThread)
 {
-    return GetThreadPriority((HANDLE)hThread);
+#ifdef _WIN32
+    return GetThreadPriority(hThread == GetKernelObject(CURRENT_THREAD_HANDLE) ? GetCurrentThread() : hThread->thread.native_handle());
+#else 
+    return 0;
+#endif
 }
 
-DWORD SetThreadIdealProcessorImpl(uint32_t hThread, DWORD dwIdealProcessor)
+uint32_t SetThreadIdealProcessorImpl(GuestThreadHandle* hThread, uint32_t dwIdealProcessor)
 {
-    return SetThreadIdealProcessor((HANDLE)hThread, dwIdealProcessor);
+#ifdef _WIN32
+    return SetThreadIdealProcessor(hThread == GetKernelObject(CURRENT_THREAD_HANDLE) ? GetCurrentThread() : hThread->thread.native_handle(), dwIdealProcessor);
+#else
+    return 0;
+#endif
 }
 
 GUEST_FUNCTION_HOOK(sub_82DFA2E8, SetThreadNameImpl);

UnleashedRecomp/cpu/guest_thread.h

@@ -1,12 +1,8 @@
 #pragma once
 
-struct PPCContext;
+#include <kernel/xdm.h>
-struct GuestThreadParameter
+
-{
+#define CURRENT_THREAD_HANDLE uint32_t(-2)
-    uint32_t function;
-    uint32_t value;
-    uint32_t flags;
-};
 
 struct GuestThreadContext
 {
@@ -17,13 +13,34 @@ struct GuestThreadContext
     ~GuestThreadContext();
 };
 
+struct GuestThreadParams
+{
+    uint32_t function;
+    uint32_t value;
+    uint32_t flags;
+};
+
+struct GuestThreadHandle : KernelObject
+{
+    GuestThreadParams params;
+    std::atomic<bool> suspended;
+    std::thread thread;
+
+    GuestThreadHandle(const GuestThreadParams& params);
+    ~GuestThreadHandle() override;
+
+    void Wait(uint32_t timeout) override;
+};
+
 struct GuestThread
 {
-    static DWORD Start(uint32_t function);
+    static uint32_t Start(const GuestThreadParams& params);
-    static DWORD Start(const GuestThreadParameter& parameter);
+    static GuestThreadHandle* Start(const GuestThreadParams& params, uint32_t* threadId);
-    static HANDLE Start(uint32_t function, uint32_t parameter, uint32_t flags, LPDWORD threadId);
 
-    static void SetThreadName(uint32_t id, const char* name);
+    static uint32_t GetCurrentThreadId();
-    static void SetLastError(DWORD error);
+    static void SetLastError(uint32_t error);
-    static PPCContext* Invoke(uint32_t address);
+
+#ifdef _WIN32
+    static void SetThreadName(uint32_t threadId, const char* name);
+#endif
 };

UnleashedRecomp/gpu/video.cpp

@@ -4047,8 +4047,9 @@ static void ProcSetPixelShader(const RenderCommand& cmd)
 
 static std::thread g_renderThread([]
     {
+#ifdef _WIN32
         GuestThread::SetThreadName(GetCurrentThreadId(), "Render Thread");
-
+#endif
         RenderCommand commands[32];
 
         while (true)
@@ -4857,7 +4858,9 @@ static moodycamel::BlockingConcurrentQueue<PipelineStateQueueItem> g_pipelineSta
 
 static void PipelineCompilerThread()
 {
+#ifdef _WIN32
     GuestThread::SetThreadName(GetCurrentThreadId(), "Pipeline Compiler Thread");
+#endif
     std::unique_ptr<GuestThreadContext> ctx;
 
     while (true)
@@ -5508,8 +5511,9 @@ static bool CheckMadeAll(const T& modelData)
 
 static void ModelConsumerThread()
 {
+#ifdef _WIN32
     GuestThread::SetThreadName(GetCurrentThreadId(), "Model Consumer Thread");
-
+#endif
     std::vector<boost::shared_ptr<Hedgehog::Database::CDatabaseData>> localPendingDataQueue;
     std::unique_ptr<GuestThreadContext> ctx;
 

UnleashedRecomp/kernel/imports.cpp

@@ -256,7 +256,16 @@ uint32_t FscSetCacheElementCount()
 
 DWORD NtWaitForSingleObjectEx(DWORD Handle, DWORD WaitMode, DWORD Alertable, XLPQWORD Timeout)
 {
-    const auto status = WaitForSingleObjectEx((HANDLE)Handle, GuestTimeoutToMilliseconds(Timeout), Alertable);
+    uint32_t timeout = GuestTimeoutToMilliseconds(Timeout);
+    assert(timeout == 0 || timeout == INFINITE);
+
+    if (IsKernelObject(Handle))
+    {
+        GetKernelObject(Handle)->Wait(timeout);
+    }
+    else
+    {
+        const auto status = WaitForSingleObjectEx((HANDLE)Handle, timeout, Alertable);
 
         if (status == WAIT_IO_COMPLETION)
         {
@@ -266,6 +275,7 @@ DWORD NtWaitForSingleObjectEx(DWORD Handle, DWORD WaitMode, DWORD Alertable, XLP
         {
             return STATUS_ALERTED;
         }
+    }
 
     return STATUS_SUCCESS;
 }
@@ -474,8 +484,9 @@ void ObDereferenceObject()
     LOG_UTILITY("!!! STUB !!!");
 }
 
-void KeSetBasePriorityThread(uint32_t thread, int priority)
+void KeSetBasePriorityThread(GuestThreadHandle* hThread, int priority)
 {
+#ifdef _WIN32
     if (priority == 16)
     {
         priority = 15;
@@ -485,7 +496,8 @@ void KeSetBasePriorityThread(uint32_t thread, int priority)
         priority = -15;
     }
 
-    SetThreadPriority((HANDLE)thread, priority);
+    SetThreadPriority(hThread == GetKernelObject(CURRENT_THREAD_HANDLE) ? GetCurrentThread() : hThread->thread.native_handle(), priority);
+#endif
 }
 
 uint32_t ObReferenceObjectByHandle(uint32_t handle, uint32_t objectType, XLPDWORD object)
@@ -499,15 +511,12 @@ void KeQueryBasePriorityThread()
     LOG_UTILITY("!!! STUB !!!");
 }
 
-uint32_t NtSuspendThread(uint32_t hThread, uint32_t* suspendCount)
+uint32_t NtSuspendThread(GuestThreadHandle* hThread, uint32_t* suspendCount)
 {
-    DWORD count = SuspendThread((HANDLE)hThread);
+    assert(hThread != GetKernelObject(CURRENT_THREAD_HANDLE) && hThread->thread.get_id() == std::this_thread::get_id());
 
-    if (count == (DWORD)-1)
+    hThread->suspended = true;
-        return E_FAIL;
+    hThread->suspended.wait(true);
-
-    if (suspendCount != nullptr)
-        *suspendCount = ByteSwap(count);
 
     return S_OK;
 }
@@ -890,29 +899,32 @@ DWORD KeWaitForSingleObject(XDISPATCHER_HEADER* Object, DWORD WaitReason, DWORD
     return WaitForSingleObjectEx(handle, timeout, Alertable);
 }
 
-static thread_local std::vector<uint32_t> g_tlsValues;
 static std::vector<size_t> g_tlsFreeIndices;
 static size_t g_tlsNextIndex = 0;
 static Mutex g_tlsAllocationMutex;
 
-static void KeTlsEnsureTlsCapacity(size_t index)
+static uint32_t& KeTlsGetValueRef(size_t index)
 {
-    if (g_tlsValues.size() <= index)
+    // Having this a global thread_local variable
+    // for some reason crashes on boot in debug builds.
+    thread_local std::vector<uint32_t> s_tlsValues;
+
+    if (s_tlsValues.size() <= index)
     {
-        g_tlsValues.resize(index + 1, 0);
+        s_tlsValues.resize(index + 1, 0);
     }
+
+    return s_tlsValues[index];
 }
 
 uint32_t KeTlsGetValue(DWORD dwTlsIndex)
 {
-    KeTlsEnsureTlsCapacity(dwTlsIndex);
+    return KeTlsGetValueRef(dwTlsIndex);
-    return g_tlsValues[dwTlsIndex];
 }
 
 BOOL KeTlsSetValue(DWORD dwTlsIndex, DWORD lpTlsValue)
 {
-    KeTlsEnsureTlsCapacity(dwTlsIndex);
+    KeTlsGetValueRef(dwTlsIndex) = lpTlsValue;
-    g_tlsValues[dwTlsIndex] = lpTlsValue;
     return TRUE;
 }
 
@@ -1169,15 +1181,12 @@ uint32_t NtClearEvent(uint32_t handle, uint32_t* previousState)
     return ResetEvent((HANDLE)handle) ? 0 : 0xFFFFFFFF;
 }
 
-uint32_t NtResumeThread(uint32_t hThread, uint32_t* suspendCount)
+uint32_t NtResumeThread(GuestThreadHandle* hThread, uint32_t* suspendCount)
 {
-    DWORD count = ResumeThread((HANDLE)hThread);
+    assert(hThread != GetKernelObject(CURRENT_THREAD_HANDLE));
 
-    if (count == (DWORD)-1)
+    hThread->suspended = false;
-        return E_FAIL;
+    hThread->suspended.notify_all();
-
-    if (suspendCount != nullptr)
-        *suspendCount = ByteSwap(count);
 
     return S_OK;
 }
@@ -1270,9 +1279,9 @@ uint32_t ExCreateThread(XLPDWORD handle, uint32_t stackSize, XLPDWORD threadId,
     LOGF_UTILITY("0x{:X}, 0x{:X}, 0x{:X}, 0x{:X}, 0x{:X}, 0x{:X}, 0x{:X}",
         (intptr_t)handle, stackSize, (intptr_t)threadId, xApiThreadStartup, startAddress, startContext, creationFlags);
 
-    DWORD hostThreadId;
+    uint32_t hostThreadId;
 
-    *handle = (uint32_t)GuestThread::Start(startAddress, startContext, creationFlags, &hostThreadId);
+    *handle = GetKernelHandle(GuestThread::Start({ startAddress, startContext, creationFlags }, &hostThreadId));
 
     if (threadId != nullptr)
         *threadId = hostThreadId;
@@ -1391,10 +1400,13 @@ DWORD XAudioGetVoiceCategoryVolumeChangeMask(DWORD Driver, XLPDWORD Mask)
     return 0;
 }
 
-uint32_t KeResumeThread(uint32_t object)
+uint32_t KeResumeThread(GuestThreadHandle* object)
 {
-    LOGF_UTILITY("0x{:x}", object);
+    assert(object != GetKernelObject(CURRENT_THREAD_HANDLE));
-    return ResumeThread((HANDLE)object);
+
+    object->suspended = false;
+    object->suspended.notify_all();
+    return 0;
 }
 
 void KeInitializeSemaphore(XKSEMAPHORE* semaphore, uint32_t count, uint32_t limit)

UnleashedRecomp/kernel/xdm.h

@@ -10,7 +10,11 @@ struct KernelObject
 {
     virtual ~KernelObject() 
     {
-        ;
+    }
+
+    virtual void Wait(uint32_t timeout) 
+    {
+        assert(false && "Wait not implemented for this kernel object.");
     }
 };
 

UnleashedRecomp/main.cpp

@@ -185,7 +185,7 @@ int main(int argc, char *argv[])
 
     Video::StartPipelinePrecompilation();
 
-    GuestThread::Start(entry);
+    GuestThread::Start({ entry, 0, 0 });
 
     return 0;
 }