[Kernel] Allocate TLS for other threads, fix CreateThread svc

include/kernel/kernel.hpp

@@ -64,7 +64,7 @@ class Kernel {
 	Handle makeProcess();
 	Handle makePort(const char* name);
 	Handle makeSession(Handle port);
-	Handle makeThread(u32 entrypoint, u32 initialSP, u32 priority, u32 id, ThreadStatus status = ThreadStatus::Dormant);
+	Handle makeThread(u32 entrypoint, u32 initialSP, u32 priority, s32 id, u32 arg,ThreadStatus status = ThreadStatus::Dormant);
 
 	void sleepThreadOnArbiter(u32 waitingAddress);
 	void switchThread(int newThreadIndex);

include/kernel/kernel_types.hpp

@@ -101,7 +101,7 @@ struct Thread {
     u32 initialSP;  // Initial r13 value
     u32 entrypoint; // Initial r15 value
     u32 priority;
-    u32 processorID;
+    s32 processorID;
     ThreadStatus status;
     Handle handle;  // OS handle for this thread
  

src/core/kernel/kernel.cpp

@@ -84,9 +84,11 @@ void Kernel::reset() {
 	// Allocate handle #0 to a dummy object and make a main process object
 	makeObject(KernelObjectType::Dummy);
 	currentProcess = makeProcess();
+
 	// Make main thread object. We do not have to set the entrypoint and SP for it as the ROM loader does.
-	// Main thread seems to have a priority of 0x30
+	// Main thread seems to have a priority of 0x30. TODO: This creates a dummy context for thread 0,
-	mainThread = makeThread(0, 0, 0x30, 0, ThreadStatus::Running);
+	// which is thankfully not used. Maybe we should prevent this
+	mainThread = makeThread(0, 0, 0x30, -2, 0, ThreadStatus::Running);
 	currentThreadIndex = 0;
 
 	// Create global service manager port

src/core/kernel/threads.cpp

@@ -32,7 +32,7 @@ void Kernel::switchThread(int newThreadIndex) {
 }
 
 // Internal OS function to spawn a thread
-Handle Kernel::makeThread(u32 entrypoint, u32 initialSP, u32 priority, u32 id, ThreadStatus status) {
+Handle Kernel::makeThread(u32 entrypoint, u32 initialSP, u32 priority, s32 id, u32 arg, ThreadStatus status) {
 	if (threadCount >= appResourceLimits.maxThreads) {
 		Helpers::panic("Overflowed the number of threads");
 	}
@@ -47,17 +47,21 @@ Handle Kernel::makeThread(u32 entrypoint, u32 initialSP, u32 priority, u32 id, T
 	t.fprs.fill(0);
 
 	t.initialSP = initialSP;
-	t.gprs[13] = initialSP;
 	t.entrypoint = entrypoint;
+
+	t.gprs[0] = arg;
+	t.gprs[13] = initialSP;
 	t.gprs[15] = entrypoint;
 	t.priority = priority;
 	t.processorID = id;
 	t.status = status;
 	t.handle = ret;
+	t.waitingAddress = 0;
 
 	t.cpsr = CPSR::UserMode | (isThumb ? CPSR::Thumb : 0);
 	t.fpscr = FPSCR::ThreadDefault;
 	// Initial TLS base has already been set in Kernel::Kernel()
+	// TODO: Does svcCreateThread zero-set the TLS of the new thread?
 
 	return ret;
 }
@@ -66,8 +70,9 @@ Handle Kernel::makeThread(u32 entrypoint, u32 initialSP, u32 priority, u32 id, T
 void Kernel::createThread() {
 	u32 priority = regs[0];
 	u32 entrypoint = regs[1];
+	u32 arg = regs[2]; // An argument value stored in r0 of the new thread
 	u32 initialSP = regs[3] & ~7; // SP is force-aligned to 8 bytes
-	u32 id = regs[4];
+	s32 id = static_cast<s32>(regs[4]);
 
 	printf("CreateThread(entry = %08X, stacktop = %08X, priority = %X, processor ID = %d)\n", entrypoint,
 		initialSP, priority, id);
@@ -79,7 +84,7 @@ void Kernel::createThread() {
 	}
 
 	regs[0] = SVCResult::Success;
-	regs[1] = makeThread(entrypoint, initialSP, priority, id);
+	regs[1] = makeThread(entrypoint, initialSP, priority, id, arg, ThreadStatus::Ready);
 }
 
 void Kernel::sleepThreadOnArbiter(u32 waitingAddress) {
@@ -87,5 +92,5 @@ void Kernel::sleepThreadOnArbiter(u32 waitingAddress) {
 
 	t.status = ThreadStatus::WaitArbiter;
 	t.waitingAddress = waitingAddress;
-	Helpers::panic("AOWPWOOOPDQ");
+	switchThread(1);
 }

src/core/memory.cpp

@@ -1,5 +1,6 @@
 #include "memory.hpp"
 #include "config_mem.hpp"
+#include "resource_limits.hpp"
 #include <cassert>
 
 using namespace KernelMemoryTypes;
@@ -27,9 +28,13 @@ void Memory::reset() {
 	u32 basePaddrForStack = FCRAM_APPLICATION_SIZE - VirtualAddrs::StackSize;
 	allocateMemory(VirtualAddrs::StackBottom, basePaddrForStack, VirtualAddrs::StackSize, true);
 
-	// And map 4KB of FCRAM before the stack for TLS
+	// And map (4 * 32)KB of FCRAM before the stack for the TLS of each thread
-	u32 basePaddrForTLS = basePaddrForStack - VirtualAddrs::TLSSize;
+	u32 basePaddrForTLS = basePaddrForStack;
-	allocateMemory(VirtualAddrs::TLSBase, basePaddrForTLS, VirtualAddrs::TLSSize, true);
+	for (int i = 0; i < appResourceLimits.maxThreads; i++) {
+		u32 vaddr = VirtualAddrs::TLSBase + i * VirtualAddrs::TLSSize;
+		basePaddrForTLS -= VirtualAddrs::TLSSize;
+		allocateMemory(vaddr, basePaddrForTLS, VirtualAddrs::TLSSize, true);
+	}
 
 	// Reserve 4KB of memory for GSP shared memory
 	constexpr u32 gspMemSize = 0x1000;