Scheduler refactor Pt. 1

* Simplifies scheduling logic, specifically regarding thread status. It should be much clearer which statuses are valid
for a thread at any given point in the system.
* Removes dead code from thread.cpp.
* Moves the implementation of resetting a ThreadContext to the corresponding core's implementation.

Other changes:
* Fixed comments in arm interfaces.
* Updated comments in thread.cpp
* Removed confusing, useless, functions like MakeReady() and ChangeStatus() from thread.cpp.
* Removed stack_size from Thread. In the CTR kernel, the thread's stack would be allocated before thread creation.

src/core/arm/arm_interface.h

@@ -85,6 +85,15 @@ public:
      */
     virtual void AddTicks(u64 ticks) = 0;
 
+    /**
+     * Initializes a CPU context for use on this CPU
+     * @param context Thread context to reset
+     * @param stack_top Pointer to the top of the stack
+     * @param entry_point Entry point for execution
+     * @param arg User argument for thread
+     */
+    virtual void ResetContext(Core::ThreadContext& context, u32 stack_top, u32 entry_point, u32 arg) = 0;
+
     /**
      * Saves the current CPU context
      * @param ctx Thread context to save

src/core/arm/dyncom/arm_dyncom.cpp

@@ -93,6 +93,16 @@ void ARM_DynCom::ExecuteInstructions(int num_instructions) {
     AddTicks(ticks_executed);
 }
 
+void ARM_DynCom::ResetContext(Core::ThreadContext& context, u32 stack_top, u32 entry_point, u32 arg) {
+    memset(&context, 0, sizeof(Core::ThreadContext));
+
+    context.cpu_registers[0] = arg;
+    context.pc = entry_point;
+    context.sp = stack_top;
+    context.cpsr = 0x1F; // Usermode
+    context.mode = 8;    // Instructs dyncom CPU core to start execution as if it's "resuming" a thread.
+}
+
 void ARM_DynCom::SaveContext(Core::ThreadContext& ctx) {
     memcpy(ctx.cpu_registers, state->Reg, sizeof(ctx.cpu_registers));
     memcpy(ctx.fpu_registers, state->ExtReg, sizeof(ctx.fpu_registers));

src/core/arm/dyncom/arm_dyncom.h

@@ -13,79 +13,24 @@
 
 class ARM_DynCom final : virtual public ARM_Interface {
 public:
-
     ARM_DynCom();
     ~ARM_DynCom();
 
-    /**
-     * Set the Program Counter to an address
-     * @param pc Address to set PC to
-     */
     void SetPC(u32 pc) override;
-
-    /*
-     * Get the current Program Counter
-     * @return Returns current PC
-     */
     u32 GetPC() const override;
-
-    /**
-     * Get an ARM register
-     * @param index Register index (0-15)
-     * @return Returns the value in the register
-     */
     u32 GetReg(int index) const override;
-
-    /**
-     * Set an ARM register
-     * @param index Register index (0-15)
-     * @param value Value to set register to
-     */
     void SetReg(int index, u32 value) override;
-
-    /**
-     * Get the current CPSR register
-     * @return Returns the value of the CPSR register
-     */
     u32 GetCPSR() const override;
-
-    /**
-     * Set the current CPSR register
-     * @param cpsr Value to set CPSR to
-     */
     void SetCPSR(u32 cpsr) override;
 
-    /**
-     * Returns the number of clock ticks since the last reset
-     * @return Returns number of clock ticks
-     */
     u64 GetTicks() const override;
-
-    /**
-    * Advance the CPU core by the specified number of ticks (e.g. to simulate CPU execution time)
-    * @param ticks Number of ticks to advance the CPU core
-    */
     void AddTicks(u64 ticks) override;
 
-    /**
-     * Saves the current CPU context
-     * @param ctx Thread context to save
-     */
+    void ResetContext(Core::ThreadContext& context, u32 stack_top, u32 entry_point, u32 arg);
     void SaveContext(Core::ThreadContext& ctx) override;
-
-    /**
-     * Loads a CPU context
-     * @param ctx Thread context to load
-     */
     void LoadContext(const Core::ThreadContext& ctx) override;
 
-    /// Prepare core for thread reschedule (if needed to correctly handle state)
     void PrepareReschedule() override;
-
-    /**
-     * Executes the given number of instructions
-     * @param num_instructions Number of instructions to executes
-     */
     void ExecuteInstructions(int num_instructions) override;
 
 private:

src/core/hle/kernel/kernel.cpp

@@ -153,12 +153,8 @@ void Shutdown() {
  * @return True on success, otherwise false
  */
 bool LoadExec(u32 entry_point) {
-    Core::g_app_core->SetPC(entry_point);
-
     // 0x30 is the typical main thread priority I've seen used so far
-    g_main_thread = Kernel::SetupMainThread(0x30, Kernel::DEFAULT_STACK_SIZE);
-    // Setup the idle thread
-    Kernel::SetupIdleThread();
+    g_main_thread = Kernel::SetupMainThread(Kernel::DEFAULT_STACK_SIZE, entry_point, 0x30);
 
     return true;
 }

src/core/hle/kernel/thread.h

@@ -31,13 +31,13 @@ enum ThreadProcessorId {
 };
 
 enum ThreadStatus {
-    THREADSTATUS_RUNNING        = 1,
-    THREADSTATUS_READY          = 2,
-    THREADSTATUS_WAIT           = 4,
-    THREADSTATUS_SUSPEND        = 8,
-    THREADSTATUS_DORMANT        = 16,
-    THREADSTATUS_DEAD           = 32,
-    THREADSTATUS_WAITSUSPEND    = THREADSTATUS_WAIT | THREADSTATUS_SUSPEND
+    THREADSTATUS_RUNNING,       ///< Currently running
+    THREADSTATUS_READY,         ///< Ready to run
+    THREADSTATUS_WAIT_ARB,      ///< Waiting on an address arbiter
+    THREADSTATUS_WAIT_SLEEP,    ///< Waiting due to a SleepThread SVC
+    THREADSTATUS_WAIT_SYNCH,    ///< Waiting due to a WaitSynchronization SVC
+    THREADSTATUS_DORMANT,       ///< Created but not yet made ready
+    THREADSTATUS_DEAD           ///< Run to completion, or forcefully terminated
 };
 
 namespace Kernel {
@@ -46,8 +46,19 @@ class Mutex;
 
 class Thread final : public WaitObject {
 public:
+    /**
+     * Creates and returns a new thread. The new thread is immediately scheduled
+     * @param name The friendly name desired for the thread
+     * @param entry_point The address at which the thread should start execution
+     * @param priority The thread's priority
+     * @param arg User data to pass to the thread
+     * @param processor_id The ID(s) of the processors on which the thread is desired to be run
+     * @param stack_top The address of the thread's stack top
+     * @param stack_size The size of the thread's stack
+     * @return A shared pointer to the newly created thread
+     */
     static ResultVal<SharedPtr<Thread>> Create(std::string name, VAddr entry_point, s32 priority,
-        u32 arg, s32 processor_id, VAddr stack_top, u32 stack_size);
+        u32 arg, s32 processor_id, VAddr stack_top);
 
     std::string GetName() const override { return name; }
     std::string GetTypeName() const override { return "Thread"; }
@@ -55,22 +66,32 @@ public:
     static const HandleType HANDLE_TYPE = HandleType::Thread;
     HandleType GetHandleType() const override { return HANDLE_TYPE; }
 
-    inline bool IsRunning() const { return (status & THREADSTATUS_RUNNING) != 0; }
-    inline bool IsStopped() const { return (status & THREADSTATUS_DORMANT) != 0; }
-    inline bool IsReady() const { return (status & THREADSTATUS_READY) != 0; }
-    inline bool IsWaiting() const { return (status & THREADSTATUS_WAIT) != 0; }
-    inline bool IsSuspended() const { return (status & THREADSTATUS_SUSPEND) != 0; }
-    inline bool IsIdle() const { return idle; }
-
     bool ShouldWait() override;
     void Acquire() override;
 
+    /**
+     * Checks if the thread is an idle (stub) thread
+     * @return True if the thread is an idle (stub) thread, false otherwise
+     */
+    inline bool IsIdle() const { return idle; }
+
+    /**
+     * Gets the thread's current priority
+     * @return The current thread's priority
+     */
     s32 GetPriority() const { return current_priority; }
+
+    /**
+     * Sets the thread's current priority
+     * @param priority The new priority
+     */
     void SetPriority(s32 priority);
 
+    /**
+     * Gets the thread's thread ID
+     * @return The thread's ID
+     */
     u32 GetThreadId() const { return thread_id; }
-
-    void Stop(const char* reason);
     
     /**
      * Release an acquired wait object
@@ -78,12 +99,14 @@ public:
      */
     void ReleaseWaitObject(WaitObject* wait_object);
 
-    /// Resumes a thread from waiting by marking it as "ready"
+    /**
+     * Resumes a thread from waiting
+     */
     void ResumeFromWait();
 
     /**
-    * Schedules an event to wake up the specified thread after the specified delay.
-    * @param nanoseconds The time this thread will be allowed to sleep for.
+    * Schedules an event to wake up the specified thread after the specified delay
+    * @param nanoseconds The time this thread will be allowed to sleep for
     */
     void WakeAfterDelay(s64 nanoseconds);
 
@@ -99,6 +122,11 @@ public:
      */
     void SetWaitSynchronizationOutput(s32 output);
 
+    /**
+     * Stops a thread, invalidating it from further use
+     */
+    void Stop();
+
     Core::ThreadContext context;
 
     u32 thread_id;
@@ -106,7 +134,6 @@ public:
     u32 status;
     u32 entry_point;
     u32 stack_top;
-    u32 stack_size;
 
     s32 initial_priority;
     s32 current_priority;
@@ -136,31 +163,49 @@ private:
 
 extern SharedPtr<Thread> g_main_thread;
 
-/// Sets up the primary application thread
-SharedPtr<Thread> SetupMainThread(s32 priority, u32 stack_size);
+/**
+ * Sets up the primary application thread
+ * @param stack_size The size of the thread's stack
+ * @param entry_point The address at which the thread should start execution
+ * @param priority The priority to give the main thread
+ * @return A shared pointer to the main thread
+ */
+SharedPtr<Thread> SetupMainThread(u32 stack_size, u32 entry_point, s32 priority);
 
-/// Reschedules to the next available thread (call after current thread is suspended)
+/**
+ * Reschedules to the next available thread (call after current thread is suspended)
+ */
 void Reschedule();
 
-/// Arbitrate the highest priority thread that is waiting
+/**
+ * Arbitrate the highest priority thread that is waiting
+ * @param address The address for which waiting threads should be arbitrated
+ */
 Thread* ArbitrateHighestPriorityThread(u32 address);
 
-/// Arbitrate all threads currently waiting...
+/**
+ * Arbitrate all threads currently waiting.
+ * @param address The address for which waiting threads should be arbitrated
+ */
 void ArbitrateAllThreads(u32 address);
 
-/// Gets the current thread
+/**
+ * Gets the current thread
+ */
 Thread* GetCurrentThread();
 
-/// Waits the current thread on a sleep
+/**
+ * Waits the current thread on a sleep
+ */
 void WaitCurrentThread_Sleep();
 
 /**
  * Waits the current thread from a WaitSynchronization call
- * @param wait_object Kernel object that we are waiting on
+ * @param wait_objects Kernel objects that we are waiting on
  * @param wait_set_output If true, set the output parameter on thread wakeup (for WaitSynchronizationN only)
  * @param wait_all If true, wait on all objects before resuming (for WaitSynchronizationN only)
  */
-void WaitCurrentThread_WaitSynchronization(SharedPtr<WaitObject> wait_object, bool wait_set_output, bool wait_all);
+void WaitCurrentThread_WaitSynchronization(std::vector<SharedPtr<WaitObject>> wait_objects, bool wait_set_output, bool wait_all);
 
 /**
  * Waits the current thread from an ArbitrateAddress call
@@ -172,14 +217,18 @@ void WaitCurrentThread_ArbitrateAddress(VAddr wait_address);
  * Sets up the idle thread, this is a thread that is intended to never execute instructions,
  * only to advance the timing. It is scheduled when there are no other ready threads in the thread queue
  * and will try to yield on every call.
- * @returns The handle of the idle thread
+ * @return The handle of the idle thread
  */
 SharedPtr<Thread> SetupIdleThread();
 
-/// Initialize threading
+/**
+ * Initialize threading
+ */
 void ThreadingInit();
 
-/// Shutdown threading
+/**
+ * Shutdown threading
+ */
 void ThreadingShutdown();
 
 } // namespace

src/core/hle/svc.cpp

@@ -150,7 +150,7 @@ static ResultCode WaitSynchronization1(Handle handle, s64 nano_seconds) {
     if (object->ShouldWait()) {
 
         object->AddWaitingThread(Kernel::GetCurrentThread());
-        Kernel::WaitCurrentThread_WaitSynchronization(object, false, false);
+        Kernel::WaitCurrentThread_WaitSynchronization({ object }, false, false);
 
         // Create an event to wake the thread up after the specified nanosecond delay has passed
         Kernel::GetCurrentThread()->WakeAfterDelay(nano_seconds);
@@ -212,7 +212,6 @@ static ResultCode WaitSynchronizationN(s32* out, Handle* handles, s32 handle_cou
         // NOTE: This should deadlock the current thread if no timeout was specified
         if (!wait_all) {
             wait_thread = true;
-            Kernel::WaitCurrentThread_WaitSynchronization(nullptr, true, wait_all);
         }
     }
 
@@ -222,12 +221,17 @@ static ResultCode WaitSynchronizationN(s32* out, Handle* handles, s32 handle_cou
     if (wait_thread) {
 
         // Actually wait the current thread on each object if we decided to wait...
+        std::vector<SharedPtr<Kernel::WaitObject>> wait_objects;
+        wait_objects.reserve(handle_count);
+        
         for (int i = 0; i < handle_count; ++i) {
             auto object = Kernel::g_handle_table.GetWaitObject(handles[i]);
             object->AddWaitingThread(Kernel::GetCurrentThread());
-            Kernel::WaitCurrentThread_WaitSynchronization(object, true, wait_all);
+            wait_objects.push_back(object);
         }
 
+        Kernel::WaitCurrentThread_WaitSynchronization(std::move(wait_objects), true, wait_all);
+
         // Create an event to wake the thread up after the specified nanosecond delay has passed
         Kernel::GetCurrentThread()->WakeAfterDelay(nano_seconds);
 
@@ -319,7 +323,7 @@ static ResultCode CreateThread(u32* out_handle, u32 priority, u32 entry_point, u
     }
 
     CASCADE_RESULT(SharedPtr<Thread> thread, Kernel::Thread::Create(
-            name, entry_point, priority, arg, processor_id, stack_top, Kernel::DEFAULT_STACK_SIZE));
+            name, entry_point, priority, arg, processor_id, stack_top));
     CASCADE_RESULT(*out_handle, Kernel::g_handle_table.Create(std::move(thread)));
 
     LOG_TRACE(Kernel_SVC, "called entrypoint=0x%08X (%s), arg=0x%08X, stacktop=0x%08X, "
@@ -338,7 +342,7 @@ static ResultCode CreateThread(u32* out_handle, u32 priority, u32 entry_point, u
 static void ExitThread() {
     LOG_TRACE(Kernel_SVC, "called, pc=0x%08X", Core::g_app_core->GetPC());
 
-    Kernel::GetCurrentThread()->Stop(__func__);
+    Kernel::GetCurrentThread()->Stop();
     HLE::Reschedule(__func__);
 }