Kernel: Cleanup multiprocessor startup by removing magic numbers

This allows easier access to processors fields

BAN/BAN/Assert.cpp

@@ -0,0 +1,22 @@
+#include <BAN/Assert.h>
+
+#if __is_kernel
+
+#include <kernel/Panic.h>
+
+[[noreturn]] void __ban_assertion_failed(const char* location, const char* msg)
+{
+	Kernel::panic_impl(location, msg);
+}
+
+#else
+
+#include <BAN/Debug.h>
+
+[[noreturn]] void __ban_assertion_failed(const char* location, const char* msg)
+{
+	derrorln("{}: {}", location, msg);
+	__builtin_trap();
+}
+
+#endif

BAN/CMakeLists.txt

@@ -3,6 +3,7 @@ cmake_minimum_required(VERSION 3.26)
 project(BAN CXX)
 
 set(BAN_SOURCES
+	BAN/Assert.cpp
 	BAN/New.cpp
 	BAN/String.cpp
 	BAN/StringView.cpp

BAN/include/BAN/Array.h

@@ -1,6 +1,5 @@
 #pragma once
 
-#include <BAN/Errors.h>
 #include <BAN/Iterators.h>
 #include <BAN/Span.h>
 
@@ -19,7 +18,7 @@ namespace BAN
 		using const_iterator = ConstIteratorSimple<T, Array>;
 
 	public:
-		Array();
+		Array() = default;
 		Array(const T&);
 
 		iterator begin() { return iterator(m_data); }
@@ -44,16 +43,9 @@ namespace BAN
 		T* data() { return m_data; }
 
 	private:
-		T m_data[S];
+		T m_data[S] {};
 	};
 
-	template<typename T, size_t S>
-	Array<T, S>::Array()
-	{
-		for (size_type i = 0; i < S; i++)
-			m_data[i] = T();
-	}
-
 	template<typename T, size_t S>
 	Array<T, S>::Array(const T& value)
 	{

BAN/include/BAN/Assert.h

@@ -1,33 +1,13 @@
 #pragma once
 
-#include <BAN/Traits.h>
+#define __ban_assert_stringify_helper(s) #s
+#define __ban_assert_stringify(s) __ban_assert_stringify_helper(s)
 
-#if defined(__is_kernel)
-	#include <kernel/Panic.h>
+#define ASSERT(cond)																						\
+	(__builtin_expect(!(cond), 0)																			\
+		? __ban_assertion_failed(__FILE__ ":" __ban_assert_stringify(__LINE__), "ASSERT(" #cond ") failed")	\
+		: (void)0)
 
-	#define ASSERT(cond)									\
-		do {												\
-			if (!(cond))									\
-				Kernel::panic("ASSERT(" #cond ") failed");	\
-		} while (false)
+#define ASSERT_NOT_REACHED() ASSERT(false)
 
-	#define __ASSERT_BIN_OP(lhs, rhs, name, op)														\
-		do {																						\
-			auto&& _lhs = (lhs);																	\
-			auto&& _rhs = (rhs);																	\
-			if (!(_lhs op _rhs))																	\
-					Kernel::panic(name "(" #lhs ", " #rhs ") ({} " #op " {}) failed", _lhs, _rhs);	\
-		} while (false)
-
-	#define ASSERT_LT(lhs, rhs)		__ASSERT_BIN_OP(lhs, rhs, "ASSERT_LT",	<)
-	#define ASSERT_LTE(lhs, rhs)	__ASSERT_BIN_OP(lhs, rhs, "ASSERT_LTE",	<=)
-	#define ASSERT_GT(lhs, rhs)		__ASSERT_BIN_OP(lhs, rhs, "ASSERT_GT",	>)
-	#define ASSERT_GTE(lhs, rhs)	__ASSERT_BIN_OP(lhs, rhs, "ASSERT_GTE",	>=)
-	#define ASSERT_EQ(lhs, rhs)		__ASSERT_BIN_OP(lhs, rhs, "ASSERT_EQ",	==)
-	#define ASSERT_NEQ(lhs, rhs)	__ASSERT_BIN_OP(lhs, rhs, "ASSERT_NEQ",	!=)
-	#define ASSERT_NOT_REACHED() Kernel::panic("ASSERT_NOT_REACHED() failed")
-#else
-	#include <assert.h>
-	#define ASSERT(cond) assert((cond) && "ASSERT("#cond") failed")
-	#define ASSERT_NOT_REACHED() do { assert(false && "ASSERT_NOT_REACHED() failed"); __builtin_unreachable(); } while (false)
-#endif
+[[noreturn]] void __ban_assertion_failed(const char* location, const char* msg);

BAN/include/BAN/CircularQueue.h

@@ -1,6 +1,7 @@
 #pragma once
 
 #include <BAN/Assert.h>
+#include <BAN/Move.h>
 #include <BAN/PlacementNew.h>
 
 #include <stdint.h>

BAN/include/BAN/Debug.h

@@ -1,8 +1,10 @@
 #pragma once
 
 #if __is_kernel
-#error "This is userspace only file"
-#endif
+
+#include <kernel/Debug.h>
+
+#else
 
 #include <BAN/Formatter.h>
 #include <stdio.h>
@@ -29,3 +31,5 @@
 		dprintln(__VA_ARGS__);								\
 		BAN::Formatter::print(__debug_putchar, "\e[m");		\
 	} while(false)
+
+#endif

BAN/include/BAN/PlacementNew.h

@@ -1,4 +1,6 @@
 #pragma once
 
+#include <stddef.h>
+
 inline void* operator new(size_t, void* addr)	{ return addr; }
 inline void* operator new[](size_t, void* addr)	{ return addr; }

kernel/CMakeLists.txt

@@ -40,9 +40,9 @@ set(KERNEL_SOURCES
 	kernel/Input/PS2/Keyboard.cpp
 	kernel/Input/PS2/Keymap.cpp
 	kernel/Input/PS2/Mouse.cpp
+	kernel/Interruptable.cpp
 	kernel/InterruptController.cpp
 	kernel/kernel.cpp
-	kernel/Lock/SpinLock.cpp
 	kernel/Memory/DMARegion.cpp
 	kernel/Memory/FileBackedRegion.cpp
 	kernel/Memory/Heap.cpp
@@ -136,6 +136,7 @@ set(LAI_SOURCES
 )
 
 set(BAN_SOURCES
+	../BAN/BAN/Assert.cpp
 	../BAN/BAN/New.cpp
 	../BAN/BAN/String.cpp
 	../BAN/BAN/StringView.cpp

kernel/arch/x86_64/GDT.cpp

@@ -1,68 +1,32 @@
-#include <BAN/Array.h>
 #include <kernel/GDT.h>
+#include <kernel/Debug.h>
 
 #include <string.h>
 
-namespace Kernel::GDT
+namespace Kernel
 {
 
-	struct TaskStateSegment
+	GDT* GDT::create()
 	{
-		uint32_t reserved1;
-		uint64_t rsp0;
-		uint64_t rsp1;
-		uint64_t rsp2;
-		uint64_t reserved2;
-		uint64_t ist1;
-		uint64_t ist2;
-		uint64_t ist3;
-		uint64_t ist4;
-		uint64_t ist5;
-		uint64_t ist6;
-		uint64_t ist7;
-		uint64_t reserved3;
-		uint16_t reserved4;
-		uint16_t iopb;
-	} __attribute__((packed));
+		auto* gdt = new GDT();
+		ASSERT(gdt);
 
-	union SegmentDescriptor
-	{
-		struct
-		{
-			uint16_t	limit1;
-			uint16_t	base1;
-			uint8_t		base2;
-			uint8_t		access;
-			uint8_t		limit2 : 4;
-			uint8_t		flags  : 4;
-			uint8_t		base3;
-		} __attribute__((packed));
+		gdt->write_entry(0x00, 0x00000000, 0x00000, 0x00, 0x0); // null
+		gdt->write_entry(0x08, 0x00000000, 0xFFFFF, 0x9A, 0xA); // kernel code
+		gdt->write_entry(0x10, 0x00000000, 0xFFFFF, 0x92, 0xC); // kernel data
+		gdt->write_entry(0x18, 0x00000000, 0xFFFFF, 0xFA, 0xA); // user code
+		gdt->write_entry(0x20, 0x00000000, 0xFFFFF, 0xF2, 0xC); // user data
+		gdt->write_tss();
 
-		struct
-		{
-			uint32_t low;
-			uint32_t high;
-		} __attribute__((packed));
+		return gdt;
+	}
 
-	} __attribute__((packed));
-
-	struct GDTR
-	{
-		uint16_t size;
-		uint64_t address;
-	} __attribute__((packed));
-
-	static constexpr uint16_t s_tss_offset = 0x28;
-
-	static TaskStateSegment s_tss;
-	static BAN::Array<SegmentDescriptor, 7> s_gdt; // null, kernel code, kernel data, user code, user data, tss low, tss high
-	static GDTR s_gdtr;
-
-	static void write_entry(uint8_t offset, uint32_t base, uint32_t limit, uint8_t access, uint8_t flags)
+	void GDT::write_entry(uint8_t offset, uint32_t base, uint32_t limit, uint8_t access, uint8_t flags)
 	{
 		ASSERT(offset % sizeof(SegmentDescriptor) == 0);
+		uint8_t idx = offset / sizeof(SegmentDescriptor);
 
-		SegmentDescriptor& desc = s_gdt[offset / sizeof(SegmentDescriptor)];
+		auto& desc = m_gdt[idx];
 		desc.base1 = (base >>  0) & 0xFFFF;
 		desc.base2 = (base >> 16) & 0xFF;
 		desc.base3 = (base >> 24) & 0xFF;
@@ -75,49 +39,18 @@ namespace Kernel::GDT
 		desc.flags = flags & 0x0F;
 	}
 
-	static void write_tss()
+	void GDT::write_tss()
 	{
-		memset(&s_tss, 0x00, sizeof(TaskStateSegment));
-		s_tss.iopb = sizeof(TaskStateSegment);
+		memset(&m_tss, 0x00, sizeof(TaskStateSegment));
+		m_tss.iopb = sizeof(TaskStateSegment);
 
-		uint64_t base = (uint64_t)&s_tss;
+		uint64_t base = reinterpret_cast<uint64_t>(&m_tss);
 
-		write_entry(s_tss_offset, (uint32_t)base, sizeof(TaskStateSegment), 0x89, 0x0);
+		write_entry(0x28, (uint32_t)base, sizeof(TaskStateSegment), 0x89, 0x0);
 
-		SegmentDescriptor& desc = s_gdt[s_tss_offset / sizeof(SegmentDescriptor) + 1];
+		auto& desc = m_gdt[0x30 / sizeof(SegmentDescriptor)];
 		desc.low = base >> 32;
 		desc.high = 0;
 	}
 
-	void set_tss_stack(uintptr_t rsp)
-	{
-		s_tss.rsp0 = rsp;
-	}
-
-	static void flush_gdt()
-	{
-		asm volatile("lgdt %0" :: "m"(s_gdtr));
-	}
-
-	static void flush_tss()
-	{
-		asm volatile("ltr %0" :: "m"(s_tss_offset));
-	}
-
-	void initialize()
-	{
-		s_gdtr.address = (uint64_t)&s_gdt;
-		s_gdtr.size = s_gdt.size() * sizeof(SegmentDescriptor) - 1;
-
-		write_entry(0x00, 0x00000000, 0x00000, 0x00, 0x0); // null
-		write_entry(0x08, 0x00000000, 0xFFFFF, 0x9A, 0xA); // kernel code
-		write_entry(0x10, 0x00000000, 0xFFFFF, 0x92, 0xC); // kernel data
-		write_entry(0x18, 0x00000000, 0xFFFFF, 0xFA, 0xA); // user code
-		write_entry(0x20, 0x00000000, 0xFFFFF, 0xF2, 0xC); // user data
-		write_tss();
-
-		flush_gdt();
-		flush_tss();
-	}
-
 }

kernel/arch/x86_64/IDT.cpp

@@ -12,7 +12,7 @@
 #define ISR_LIST_X X(0) X(1) X(2) X(3) X(4) X(5) X(6) X(7) X(8) X(9) X(10) X(11) X(12) X(13) X(14) X(15) X(16) X(17) X(18) X(19) X(20) X(21) X(22) X(23) X(24) X(25) X(26) X(27) X(28) X(29) X(30) X(31)
 #define IRQ_LIST_X X(0) X(1) X(2) X(3) X(4) X(5) X(6) X(7) X(8) X(9) X(10) X(11) X(12) X(13) X(14) X(15) X(16) X(17) X(18) X(19) X(20) X(21) X(22) X(23) X(24) X(25) X(26) X(27) X(28) X(29) X(30) X(31)
 
-namespace Kernel::IDT
+namespace Kernel
 {
 
 	struct Registers
@@ -42,26 +42,6 @@ namespace Kernel::IDT
 		uint64_t rax;
 	};
 
-	struct GateDescriptor
-	{
-		uint16_t offset1;
-		uint16_t selector;
-		uint8_t IST;
-		uint8_t flags;
-		uint16_t offset2;
-		uint32_t offset3;
-		uint32_t reserved;
-	} __attribute__((packed));
-
-	struct IDTR
-	{
-		uint16_t size;
-		uint64_t offset;
-	} __attribute__((packed));
-
-	static IDTR				s_idtr;
-	static GateDescriptor*	s_idt = nullptr;
-
 #define X(num) 1 +
 	static BAN::Array<Interruptable*, IRQ_LIST_X 0> s_interruptables;
 #undef X
@@ -163,6 +143,13 @@ namespace Kernel::IDT
 
 	extern "C" void cpp_isr_handler(uint64_t isr, uint64_t error, InterruptStack& interrupt_stack, const Registers* regs)
 	{
+		if (g_paniced)
+		{
+			// FIXME: tell other processors kernel panic has occured
+			dprintln("Processor {} halted", Processor::current_id());
+			asm volatile("cli; 1: hlt; jmp 1b");
+		}
+
 #if __enable_sse
 		bool from_userspace = (interrupt_stack.cs & 0b11) == 0b11;
 		if (from_userspace)
@@ -334,14 +321,9 @@ done:
 		ASSERT(Thread::current().state() != Thread::State::Terminated);
 	}
 
-	static void flush_idt()
+	void IDT::register_interrupt_handler(uint8_t index, void (*handler)())
 	{
-		asm volatile("lidt %0"::"m"(s_idtr));
-	}
-
-	static void register_interrupt_handler(uint8_t index, void(*handler)())
-	{
-		GateDescriptor& descriptor = s_idt[index];
+		auto& descriptor = m_idt[index];
 		descriptor.offset1 = (uint16_t)((uint64_t)handler >> 0);
 		descriptor.offset2 = (uint16_t)((uint64_t)handler >> 16);
 		descriptor.offset3 = (uint32_t)((uint64_t)handler >> 32);
@@ -351,13 +333,13 @@ done:
 		descriptor.flags = 0x8E;
 	}
 
-	static void register_syscall_handler(uint8_t index, void(*handler)())
+	void IDT::register_syscall_handler(uint8_t index, void (*handler)())
 	{
 		register_interrupt_handler(index, handler);
-		s_idt[index].flags = 0xEE;
+		m_idt[index].flags = 0xEE;
 	}
 
-	void register_irq_handler(uint8_t irq, Interruptable* interruptable)
+	void IDT::register_irq_handler(uint8_t irq, Interruptable* interruptable)
 	{
 		if (irq > s_interruptables.size())
 			Kernel::panic("Trying to assign handler for irq {} while only {} are supported", irq, s_interruptables.size());
@@ -374,34 +356,36 @@ done:
 
 	extern "C" void syscall_asm();
 
-	void initialize()
+	IDT* IDT::create(bool is_bsb)
 	{
-		s_idt = (GateDescriptor*)kmalloc(0x100 * sizeof(GateDescriptor));
-		ASSERT(s_idt);
-		memset(s_idt, 0x00, 0x100 * sizeof(GateDescriptor));
+		auto* idt = new IDT();
+		ASSERT(idt);
 
-		s_idtr.offset = (uint64_t)s_idt;
-		s_idtr.size = 0x100 * sizeof(GateDescriptor) - 1;
+		memset(idt->m_idt.data(), 0x00, 0x100 * sizeof(GateDescriptor));
 
-#define X(num) register_interrupt_handler(num, isr ## num);
+#define X(num) idt->register_interrupt_handler(num, isr ## num);
 		ISR_LIST_X
 #undef X
 
-#define X(num) register_interrupt_handler(IRQ_VECTOR_BASE + num, irq ## num);
-		IRQ_LIST_X
+		// FIXME: distribute IRQs more evenly?
+#define X(num) idt->register_interrupt_handler(IRQ_VECTOR_BASE + num, irq ## num);
+		if (is_bsb)
+		{
+			IRQ_LIST_X
+		}
 #undef X
 
-		register_syscall_handler(0x80, syscall_asm);
+		idt->register_syscall_handler(0x80, syscall_asm);
 
-		flush_idt();
+		return idt;
 	}
 
-	[[noreturn]] void force_triple_fault()
+	[[noreturn]] void IDT::force_triple_fault()
 	{
 		// load 0 sized IDT and trigger an interrupt to force triple fault
-		asm volatile("cli");
-		s_idtr.size = 0;
-		flush_idt();
+		Processor::set_interrupt_state(InterruptState::Disabled);
+		Processor::idt().m_idtr.size = 0;
+		Processor::idt().load();
 		asm volatile("int $0x00");
 		ASSERT_NOT_REACHED();
 	}

kernel/arch/x86_64/PageTable.cpp

@@ -20,7 +20,6 @@ namespace Kernel
 	RecursiveSpinLock PageTable::s_fast_page_lock;
 
 	static PageTable* s_kernel = nullptr;
-	static PageTable* s_current = nullptr;
 	static bool s_has_nxe = false;
 	static bool s_has_pge = false;
 
@@ -71,40 +70,51 @@ namespace Kernel
 	void PageTable::initialize()
 	{
 		if (CPUID::has_nxe())
-		{
-			asm volatile(
-				"movl $0xC0000080, %ecx;"
-				"rdmsr;"
-				"orl $0x800, %eax;"
-				"wrmsr"
-			);
 			s_has_nxe = true;
-		}
 
-		uint32_t ecx, edx;
-		CPUID::get_features(ecx, edx);
-		if (edx & CPUID::EDX_PGE)
-		{
-			asm volatile(
-				"movq %cr4, %rax;"
-				"orq $0x80, %rax;"
-				"movq %rax, %cr4;"
-			);
+		if (CPUID::has_pge())
 			s_has_pge = true;
-		}
-
-		// enable write protect to kernel
-		asm volatile(
-			"movq %cr0, %rax;"
-			"orq $0x10000, %rax;"
-			"movq %rax, %cr0;"
-		);
 
 		ASSERT(s_kernel == nullptr);
 		s_kernel = new PageTable();
 		ASSERT(s_kernel);
+
 		s_kernel->initialize_kernel();
-		s_kernel->load();
+		s_kernel->initial_load();
+	}
+
+	void PageTable::initial_load()
+	{
+		if (s_has_nxe)
+		{
+			asm volatile(
+				"movl $0xC0000080, %%ecx;"
+				"rdmsr;"
+				"orl $0x800, %%eax;"
+				"wrmsr"
+				::: "eax", "ecx", "edx", "memory"
+			);
+		}
+
+		if (s_has_pge)
+		{
+			asm volatile(
+				"movq %%cr4, %%rax;"
+				"orq $0x80, %%rax;"
+				"movq %%rax, %%cr4;"
+				::: "rax"
+			);
+		}
+
+		// enable write protect
+		asm volatile(
+			"movq %%cr0, %%rax;"
+			"orq $0x10000, %%rax;"
+			"movq %%rax, %%cr0;"
+			::: "rax"
+		);
+
+		load();
 	}
 
 	PageTable& PageTable::kernel()
@@ -113,12 +123,6 @@ namespace Kernel
 		return *s_kernel;
 	}
 
-	PageTable& PageTable::current()
-	{
-		ASSERT(s_current);
-		return *s_current;
-	}
-
 	bool PageTable::is_valid_pointer(uintptr_t pointer)
 	{
 		if (!is_canonical(pointer))
@@ -208,7 +212,7 @@ namespace Kernel
 	void PageTable::map_fast_page(paddr_t paddr)
 	{
 		ASSERT(s_kernel);
-		ASSERT_NEQ(paddr, 0);
+		ASSERT(paddr);
 
 		SpinLockGuard _(s_fast_page_lock);
 
@@ -308,7 +312,7 @@ namespace Kernel
 	{
 		SpinLockGuard _(m_lock);
 		asm volatile("movq %0, %%cr3" :: "r"(m_highest_paging_struct));
-		s_current = this;
+		Processor::set_current_page_table(this);
 	}
 
 	void PageTable::invalidate(vaddr_t vaddr)
@@ -322,7 +326,7 @@ namespace Kernel
 		ASSERT(vaddr);
 		ASSERT(vaddr != fast_page());
 		if (vaddr >= KERNEL_OFFSET)
-			ASSERT_GTE(vaddr, (vaddr_t)g_kernel_start);
+			ASSERT(vaddr >= (vaddr_t)g_kernel_start);
 		if ((vaddr >= KERNEL_OFFSET) != (this == s_kernel))
 			Kernel::panic("unmapping {8H}, kernel: {}", vaddr, this == s_kernel);
 
@@ -367,8 +371,6 @@ namespace Kernel
 	{
 		ASSERT(vaddr);
 		ASSERT(vaddr != fast_page());
-		if (vaddr >= KERNEL_OFFSET && s_current)
-			ASSERT_GTE(vaddr, (vaddr_t)g_kernel_start);
 		if ((vaddr >= KERNEL_OFFSET) != (this == s_kernel))
 			Kernel::panic("mapping {8H} to {8H}, kernel: {}", paddr, vaddr, this == s_kernel);
 

kernel/arch/x86_64/boot.S

@@ -158,13 +158,6 @@ enable_sse:
 	movl %eax, %cr4
 	ret
 
-initialize_lapic_id:
-	movl $1, %eax
-	cpuid
-	shrl $24, %ebx
-	movw %bx, %gs
-	ret
-
 initialize_paging:
 	# enable PAE
 	movl %cr4, %ecx
@@ -198,7 +191,6 @@ _start:
 	movl %ebx, V2P(bootloader_info)
 
 	movl $V2P(boot_stack_top), %esp
-	call initialize_lapic_id
 
 	call check_requirements
 	call enable_sse
@@ -278,7 +270,6 @@ ap_protected_mode:
 
 	movl ap_stack_ptr, %esp
 	movb $1, V2P(g_ap_stack_loaded)
-	call V2P(initialize_lapic_id)
 
 	call V2P(enable_sse)
 

kernel/include/kernel/CPUID.h

@@ -78,5 +78,6 @@ namespace CPUID
 	void get_features(uint32_t& ecx, uint32_t& edx);
 	bool is_64_bit();
 	bool has_nxe();
+	bool has_pge();
 
 }

kernel/include/kernel/GDT.h

@@ -1,16 +1,101 @@
 #pragma once
 
+#include <BAN/Array.h>
+#include <BAN/NoCopyMove.h>
+
 #include <stdint.h>
 
-namespace Kernel::GDT
+namespace Kernel
 {
 
-	static constexpr inline bool is_user_segment(uint8_t segment)
+	struct TaskStateSegment
 	{
-		return (segment & 3) == 3;
-	}
+		uint32_t reserved1;
+		uint64_t rsp0;
+		uint64_t rsp1;
+		uint64_t rsp2;
+		uint64_t reserved2;
+		uint64_t ist1;
+		uint64_t ist2;
+		uint64_t ist3;
+		uint64_t ist4;
+		uint64_t ist5;
+		uint64_t ist6;
+		uint64_t ist7;
+		uint64_t reserved3;
+		uint16_t reserved4;
+		uint16_t iopb;
+	} __attribute__((packed));
 
-	void initialize();
-	void set_tss_stack(uintptr_t);
+	union SegmentDescriptor
+	{
+		struct
+		{
+			uint16_t	limit1;
+			uint16_t	base1;
+			uint8_t		base2;
+			uint8_t		access;
+			uint8_t		limit2 : 4;
+			uint8_t		flags  : 4;
+			uint8_t		base3;
+		} __attribute__((packed));
+
+		struct
+		{
+			uint32_t low;
+			uint32_t high;
+		} __attribute__((packed));
+
+	} __attribute__((packed));
+
+	struct GDTR
+	{
+		uint16_t size;
+		uint64_t address;
+	} __attribute__((packed));
+
+	class GDT
+	{
+		BAN_NON_COPYABLE(GDT);
+		BAN_NON_MOVABLE(GDT);
+
+	public:
+		static GDT* create();
+		void load() { flush_gdt(); flush_tss(); }
+
+		static constexpr inline bool is_user_segment(uint8_t segment)
+		{
+			return (segment & 3) == 3;
+		}
+
+		void set_tss_stack(uintptr_t rsp)
+		{
+			m_tss.rsp0 = rsp;
+		}
+
+	private:
+		GDT() = default;
+
+		void write_entry(uint8_t offset, uint32_t base, uint32_t limit, uint8_t access, uint8_t flags);
+		void write_tss();
+
+		void flush_gdt()
+		{
+			asm volatile("lgdt %0" :: "m"(m_gdtr) : "memory");
+		}
+
+		void flush_tss()
+		{
+			asm volatile("ltr %0" :: "rm"((uint16_t)0x28) : "memory");
+		}
+
+	private:
+		BAN::Array<SegmentDescriptor, 7> m_gdt; // null, kernel code, kernel data, user code, user data, tss low, tss high
+		TaskStateSegment m_tss;
+		const GDTR m_gdtr {
+			.size = m_gdt.size() * sizeof(SegmentDescriptor) - 1,
+			.address = reinterpret_cast<uint64_t>(m_gdt.data())
+		};
+	};
 
 }

kernel/include/kernel/IDT.h

@@ -1,13 +1,62 @@
 #pragma once
 
+#include <BAN/Array.h>
+#include <BAN/NoCopyMove.h>
+#include <kernel/Interruptable.h>
+
 #include <stdint.h>
 
 constexpr uint8_t IRQ_VECTOR_BASE = 0x20;
 
-namespace Kernel::IDT
+namespace Kernel
 {
 
-	void initialize();
-	[[noreturn]] void force_triple_fault();
+	struct GateDescriptor
+	{
+		uint16_t offset1;
+		uint16_t selector;
+		uint8_t IST;
+		uint8_t flags;
+		uint16_t offset2;
+		uint32_t offset3;
+		uint32_t reserved;
+	} __attribute__((packed));
+
+	struct IDTR
+	{
+		uint16_t size;
+		uint64_t offset;
+	} __attribute__((packed));
+
+	class IDT
+	{
+		BAN_NON_COPYABLE(IDT);
+		BAN_NON_MOVABLE(IDT);
+
+	public:
+		static IDT* create(bool is_bsb);
+
+		[[noreturn]] static void force_triple_fault();
+
+		void register_irq_handler(uint8_t irq, Interruptable* interruptable);
+
+		void load()
+		{
+			asm volatile("lidt %0" :: "m"(m_idtr) : "memory");
+		}
+
+	private:
+		IDT() = default;
+
+		void register_interrupt_handler(uint8_t index, void (*handler)());
+		void register_syscall_handler(uint8_t index, void (*handler)());
+
+	private:
+		BAN::Array<GateDescriptor, 0x100> m_idt;
+		IDTR m_idtr {
+			.size = m_idt.size() * sizeof(GateDescriptor) - 1,
+			.offset = reinterpret_cast<uint64_t>(m_idt.data())
+		};
+	};
 
 }

kernel/include/kernel/Input/PS2/Device.h

@@ -1,7 +1,7 @@
 #pragma once
 
 #include <kernel/Input/PS2/Controller.h>
-#include <kernel/InterruptController.h>
+#include <kernel/Interruptable.h>
 
 namespace Kernel::Input
 {

kernel/include/kernel/InterruptController.h

@@ -8,23 +8,6 @@
 namespace Kernel
 {
 
-	class Interruptable
-	{
-	public:
-		void set_irq(int irq);
-		void enable_interrupt();
-		void disable_interrupt();
-
-		virtual void handle_irq() = 0;
-
-	protected:
-		Interruptable() = default;
-		~Interruptable() {}
-
-	private:
-		int m_irq { -1 };
-	};
-
 	class InterruptController
 	{
 	public:

kernel/include/kernel/Interruptable.h

@@ -0,0 +1,23 @@
+#pragma once
+
+namespace Kernel
+{
+
+	class Interruptable
+	{
+	public:
+		void set_irq(int irq);
+		void enable_interrupt();
+		void disable_interrupt();
+
+		virtual void handle_irq() = 0;
+
+	protected:
+		Interruptable() = default;
+		~Interruptable() {}
+
+	private:
+		int m_irq { -1 };
+	};
+
+}

kernel/include/kernel/Lock/Mutex.h

@@ -21,12 +21,12 @@ namespace Kernel
 		{
 			auto tid = Scheduler::current_tid();
 			if (tid == m_locker)
-				ASSERT_GT(m_lock_depth, 0);
+				ASSERT(m_lock_depth > 0);
 			else
 			{
 				while (!m_locker.compare_exchange(-1, tid))
 					Scheduler::get().reschedule();
-				ASSERT_EQ(m_lock_depth, 0);
+				ASSERT(m_lock_depth == 0);
 			}
 			m_lock_depth++;
 		}
@@ -35,20 +35,20 @@ namespace Kernel
 		{
 			auto tid = Scheduler::current_tid();
 			if (tid == m_locker)
-				ASSERT_GT(m_lock_depth, 0);
+				ASSERT(m_lock_depth > 0);
 			else
 			{
 				if (!m_locker.compare_exchange(-1, tid))
 					return false;
-				ASSERT_EQ(m_lock_depth, 0);
+				ASSERT(m_lock_depth == 0);
 			}
 			m_lock_depth++;
 		}
 
 		void unlock()
 		{
-			ASSERT_EQ(m_locker.load(), Scheduler::current_tid());
-			ASSERT_GT(m_lock_depth, 0);
+			ASSERT(m_locker == Scheduler::current_tid());
+			ASSERT(m_lock_depth > 0);
 			if (--m_lock_depth == 0)
 				m_locker = -1;
 		}
@@ -74,7 +74,7 @@ namespace Kernel
 		{
 			auto tid = Scheduler::current_tid();
 			if (tid == m_locker)
-				ASSERT_GT(m_lock_depth, 0);
+				ASSERT(m_lock_depth > 0);
 			else
 			{
 				bool has_priority = tid ? !Thread::current().is_userspace() : true;
@@ -82,7 +82,7 @@ namespace Kernel
 					m_queue_length++;
 				while (!(has_priority || m_queue_length == 0) || !m_locker.compare_exchange(-1, tid))
 					Scheduler::get().reschedule();
-				ASSERT_EQ(m_lock_depth, 0);
+				ASSERT(m_lock_depth == 0);
 			}
 			m_lock_depth++;
 		}
@@ -91,7 +91,7 @@ namespace Kernel
 		{
 			auto tid = Scheduler::current_tid();
 			if (tid == m_locker)
-				ASSERT_GT(m_lock_depth, 0);
+				ASSERT(m_lock_depth > 0);
 			else
 			{
 				bool has_priority = tid ? !Thread::current().is_userspace() : true;
@@ -99,7 +99,7 @@ namespace Kernel
 					return false;
 				if (has_priority)
 					m_queue_length++;
-				ASSERT_EQ(m_lock_depth, 0);
+				ASSERT(m_lock_depth == 0);
 			}
 			m_lock_depth++;
 		}
@@ -107,8 +107,8 @@ namespace Kernel
 		void unlock()
 		{
 			auto tid = Scheduler::current_tid();
-			ASSERT_EQ(m_locker.load(), tid);
-			ASSERT_GT(m_lock_depth, 0);
+			ASSERT(m_locker == tid);
+			ASSERT(m_lock_depth > 0);
 			if (--m_lock_depth == 0)
 			{
 				bool has_priority = tid ? !Thread::current().is_userspace() : true;

kernel/include/kernel/Lock/SpinLock.h

@@ -1,5 +1,6 @@
 #pragma once
 
+#include <BAN/Assert.h>
 #include <BAN/Atomic.h>
 #include <BAN/NoCopyMove.h>
 #include <kernel/Processor.h>
@@ -17,11 +18,35 @@ namespace Kernel
 	public:
 		SpinLock() = default;
 
-		InterruptState lock();
-		void unlock(InterruptState state);
+		InterruptState lock()
+		{
+			auto state = Processor::get_interrupt_state();
+			Processor::set_interrupt_state(InterruptState::Disabled);
+
+			auto id = Processor::current_id();
+			ASSERT(m_locker != id);
+
+			while (!m_locker.compare_exchange(PROCESSOR_NONE, id, BAN::MemoryOrder::memory_order_acquire))
+				__builtin_ia32_pause();
+
+			return state;
+		}
+
+		void unlock(InterruptState state)
+		{
+			ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
+			ASSERT(m_locker == Processor::current_id());
+			m_locker.store(PROCESSOR_NONE, BAN::MemoryOrder::memory_order_release);
+			Processor::set_interrupt_state(state);
+		}
+
+		bool current_processor_has_lock() const
+		{
+			return m_locker == Processor::current_id();
+		}
 
 	private:
-		BAN::Atomic<ProcessorID>	m_locker { PROCESSOR_NONE };
+		BAN::Atomic<ProcessorID> m_locker { PROCESSOR_NONE };
 	};
 
 	class RecursiveSpinLock
@@ -32,45 +57,44 @@ namespace Kernel
 	public:
 		RecursiveSpinLock() = default;
 
-		InterruptState lock();
-		void unlock(InterruptState state);
-
-	private:
-		BAN::Atomic<ProcessorID>	m_locker { PROCESSOR_NONE };
-		uint32_t					m_lock_depth { 0 };
-	};
-
-	class SpinLockUnsafe
-	{
-		BAN_NON_COPYABLE(SpinLockUnsafe);
-		BAN_NON_MOVABLE(SpinLockUnsafe);
-
-	public:
-		SpinLockUnsafe() = default;
-
 		InterruptState lock()
 		{
-			auto id = Processor::current_id();
-
 			auto state = Processor::get_interrupt_state();
 			Processor::set_interrupt_state(InterruptState::Disabled);
 
-			while (!m_locker.compare_exchange(PROCESSOR_NONE, id, BAN::MemoryOrder::memory_order_acquire))
-				__builtin_ia32_pause();
+			auto id = Processor::current_id();
+			if (m_locker == id)
+				ASSERT(m_lock_depth > 0);
+			else
+			{
+				while (!m_locker.compare_exchange(PROCESSOR_NONE, id, BAN::MemoryOrder::memory_order_acquire))
+					__builtin_ia32_pause();
+				ASSERT(m_lock_depth == 0);
+			}
+
+			m_lock_depth++;
 
 			return state;
 		}
 
 		void unlock(InterruptState state)
 		{
-			m_locker.store(PROCESSOR_NONE, BAN::MemoryOrder::memory_order_release);
+			ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
+			ASSERT(m_locker == Processor::current_id());
+			ASSERT(m_lock_depth > 0);
+			if (--m_lock_depth == 0)
+				m_locker.store(PROCESSOR_NONE, BAN::MemoryOrder::memory_order_release);
 			Processor::set_interrupt_state(state);
 		}
 
-		bool is_locked() const { return m_locker != PROCESSOR_NONE; }
+		bool current_processor_has_lock() const
+		{
+			return m_locker == Processor::current_id();
+		}
 
 	private:
-		BAN::Atomic<ProcessorID> m_locker { PROCESSOR_NONE };
+		BAN::Atomic<ProcessorID>	m_locker { PROCESSOR_NONE };
+		uint32_t					m_lock_depth { 0 };
 	};
 
 	template<typename Lock>

kernel/include/kernel/Memory/PageTable.h

@@ -40,7 +40,7 @@ namespace Kernel
 		static void initialize();
 
 		static PageTable& kernel();
-		static PageTable& current();
+		static PageTable& current() { return *reinterpret_cast<PageTable*>(Processor::get_current_page_table()); }
 
 		static constexpr vaddr_t fast_page() { return KERNEL_OFFSET; }
 
@@ -109,6 +109,7 @@ namespace Kernel
 		vaddr_t reserve_free_contiguous_pages(size_t page_count, vaddr_t first_address, vaddr_t last_address = UINTPTR_MAX);
 
 		void load();
+		void initial_load();
 
 		InterruptState lock() const { return m_lock.lock(); }
 		void unlock(InterruptState state) const { m_lock.unlock(state); }

kernel/include/kernel/Networking/E1000/E1000.h

@@ -1,7 +1,7 @@
 #pragma once
 
 #include <BAN/UniqPtr.h>
-#include <kernel/InterruptController.h>
+#include <kernel/Interruptable.h>
 #include <kernel/Memory/DMARegion.h>
 #include <kernel/Networking/E1000/Definitions.h>
 #include <kernel/Networking/NetworkInterface.h>

kernel/include/kernel/Panic.h

@@ -2,28 +2,29 @@
 
 #include <kernel/Debug.h>
 
-#define panic(...) detail::panic_impl(__FILE__, __LINE__, __VA_ARGS__)
+#define __panic_stringify_helper(s) #s
+#define __panic_stringify(s) __panic_stringify_helper(s)
 
-namespace Kernel::detail
+#define panic(...) panic_impl(__FILE__ ":" __panic_stringify(__LINE__), __VA_ARGS__)
+
+namespace Kernel
 {
 
-	extern bool g_paniced;
+	extern volatile bool g_paniced;
 
 	template<typename... Args>
 	__attribute__((__noreturn__))
-	static void panic_impl(const char* file, int line, const char* message, Args&&... args)
+	static void panic_impl(const char* location, const char* message, Args&&... args)
 	{
 		asm volatile("cli");
-		derrorln("Kernel panic at {}:{}", file, line);
+		derrorln("Kernel panic at {}", location);
 		derrorln(message, BAN::forward<Args>(args)...);
 		if (!g_paniced)
 		{
 			g_paniced = true;
 			Debug::dump_stack_trace();
 		}
-		for (;;)
-			asm volatile("hlt");
-		__builtin_unreachable();
+		asm volatile("ud2");
 	}
 
 }

kernel/include/kernel/Processor.h

@@ -1,7 +1,10 @@
 #pragma once
 
 #include <BAN/ForwardList.h>
+
 #include <kernel/Arch.h>
+#include <kernel/GDT.h>
+#include <kernel/IDT.h>
 
 namespace Kernel
 {
@@ -19,20 +22,16 @@ namespace Kernel
 	class Processor
 	{
 		BAN_NON_COPYABLE(Processor);
+		BAN_NON_MOVABLE(Processor);
 
 	public:
 		static Processor& create(ProcessorID id);
+		static Processor& initialize();
 
-		static ProcessorID current_id()
-		{
-			uint16_t id;
-			asm volatile("movw %%gs, %0" : "=rm"(id));
-			return id;
-		}
+		static ProcessorID current_id() { return read_gs_sized<ProcessorID>(offsetof(Processor, m_id)); }
 
-		static Processor& get(ProcessorID);
-
-		static Processor& current() { return get(current_id()); }
+		static ProcessorID bsb_id() { return s_bsb_id; }
+		static bool current_is_bsb() { return current_id() == bsb_id(); }
 
 		static void set_interrupt_state(InterruptState state)
 		{
@@ -51,23 +50,71 @@ namespace Kernel
 			return InterruptState::Disabled;
 		};
 
-		uintptr_t stack_bottom() const { return reinterpret_cast<uintptr_t>(m_stack); }
-		uintptr_t stack_top() const { return stack_bottom() + m_stack_size; }
+		static uintptr_t current_stack_bottom() { return reinterpret_cast<uintptr_t>(read_gs_ptr(offsetof(Processor, m_stack))); }
+		static uintptr_t current_stack_top()	{ return current_stack_bottom() + s_stack_size; }
+
+		uintptr_t stack_bottom() const	{ return reinterpret_cast<uintptr_t>(m_stack); }
+		uintptr_t stack_top() const		{ return stack_bottom() + s_stack_size; }
+
+		static GDT& gdt() { return *reinterpret_cast<GDT*>(read_gs_ptr(offsetof(Processor, m_gdt))); }
+		static IDT& idt() { return *reinterpret_cast<IDT*>(read_gs_ptr(offsetof(Processor, m_idt))); }
+
+		static void* get_current_page_table()					{ return read_gs_ptr(offsetof(Processor, m_current_page_table)); }
+		static void set_current_page_table(void* page_table)	{ write_gs_ptr(offsetof(Processor, m_current_page_table), page_table); }
 
 	private:
 		Processor() = default;
-		Processor(Processor&& other)
+		~Processor() { ASSERT_NOT_REACHED(); }
+
+		template<typename T>
+		static T read_gs_sized(uintptr_t offset) requires(sizeof(T) <= 8)
 		{
-			m_stack = other.m_stack;
-			other.m_stack = nullptr;
+#define __ASM_INPUT(operation) operation " %%gs:(%[offset]), %[result]" : [result]"=rm"(result) : [offset]"rm"(offset)
+			T result;
+			if constexpr(sizeof(T) == 8)
+				asm volatile(__ASM_INPUT("movq"));
+			if constexpr(sizeof(T) == 4)
+				asm volatile(__ASM_INPUT("movl"));
+			if constexpr(sizeof(T) == 2)
+				asm volatile(__ASM_INPUT("movw"));
+			if constexpr(sizeof(T) == 1)
+				asm volatile(__ASM_INPUT("movb"));
+			return result;
+#undef __ASM_INPUT
 		}
-		~Processor();
+
+		template<typename T>
+		static void write_gs_sized(uintptr_t offset, T value) requires(sizeof(T) <= 8)
+		{
+#define __ASM_INPUT(operation) operation " %[value], %%gs:(%[offset])" :: [value]"rm"(value), [offset]"rm"(offset) : "memory"
+			if constexpr(sizeof(T) == 8)
+				asm volatile(__ASM_INPUT("movq"));
+			if constexpr(sizeof(T) == 4)
+				asm volatile(__ASM_INPUT("movl"));
+			if constexpr(sizeof(T) == 2)
+				asm volatile(__ASM_INPUT("movw"));
+			if constexpr(sizeof(T) == 1)
+				asm volatile(__ASM_INPUT("movb"));
+#undef __ASM_INPUT
+		}
+
+		static void* read_gs_ptr(uintptr_t offset) { return read_gs_sized<void*>(offset); }
+		static void write_gs_ptr(uintptr_t offset, void* value) { write_gs_sized<void*>(offset, value); }
 
 	private:
-		void* m_stack { nullptr };
-		static constexpr size_t m_stack_size { 4096 };
+		static ProcessorID s_bsb_id;
 
-		friend class BAN::Vector<Processor>;
+		ProcessorID m_id { PROCESSOR_NONE };
+
+		static constexpr size_t s_stack_size { 4096 };
+		void* m_stack { nullptr };
+
+		GDT* m_gdt { nullptr };
+		IDT* m_idt { nullptr };
+
+		void* m_current_page_table { nullptr };
+
+		friend class BAN::Array<Processor, 0xFF>;
 	};
 #else
 	#error

kernel/include/kernel/Scheduler.h

@@ -62,7 +62,7 @@ namespace Kernel
 			Semaphore*	semaphore;
 		};
 
-		SpinLockUnsafe m_lock;
+		SpinLock m_lock;
 
 		Thread* m_idle_thread { nullptr };
 		BAN::LinkedList<SchedulerThread> m_active_threads;

kernel/include/kernel/Storage/ATA/AHCI/Controller.h

@@ -2,7 +2,7 @@
 
 #include <BAN/Array.h>
 #include <BAN/RefPtr.h>
-#include <kernel/InterruptController.h>
+#include <kernel/Interruptable.h>
 #include <kernel/Memory/DMARegion.h>
 #include <kernel/PCI.h>
 #include <kernel/Storage/ATA/AHCI/Definitions.h>

kernel/include/kernel/Storage/ATA/ATABus.h

@@ -3,7 +3,7 @@
 #include <BAN/ByteSpan.h>
 #include <BAN/RefPtr.h>
 #include <BAN/Vector.h>
-#include <kernel/InterruptController.h>
+#include <kernel/Interruptable.h>
 #include <kernel/Lock/Mutex.h>
 
 namespace Kernel

kernel/include/kernel/Storage/NVMe/Queue.h

@@ -2,7 +2,7 @@
 
 #include <BAN/UniqPtr.h>
 #include <BAN/Vector.h>
-#include <kernel/InterruptController.h>
+#include <kernel/Interruptable.h>
 #include <kernel/Lock/Mutex.h>
 #include <kernel/Memory/DMARegion.h>
 #include <kernel/Semaphore.h>

kernel/include/kernel/Terminal/Serial.h

@@ -2,7 +2,7 @@
 
 #include <BAN/CircularQueue.h>
 #include <BAN/Errors.h>
-#include <kernel/InterruptController.h>
+#include <kernel/Interruptable.h>
 #include <kernel/Terminal/TTY.h>
 
 namespace Kernel

kernel/include/kernel/Timer/HPET.h

@@ -1,6 +1,6 @@
 #pragma once
 
-#include <kernel/InterruptController.h>
+#include <kernel/Interruptable.h>
 #include <kernel/Lock/SpinLock.h>
 #include <kernel/Timer/Timer.h>
 

kernel/include/kernel/Timer/PIT.h

@@ -1,6 +1,6 @@
 #pragma once
 
-#include <kernel/InterruptController.h>
+#include <kernel/Interruptable.h>
 #include <kernel/Timer/Timer.h>
 
 namespace Kernel

kernel/kernel/APIC.cpp

@@ -31,6 +31,35 @@ extern volatile uint8_t g_ap_stack_loaded[];
 namespace Kernel
 {
 
+	enum ICR_LO : uint32_t
+	{
+		ICR_LO_reserved_mask							= 0xFFF32000,
+
+		ICR_LO_delivery_mode_fixed						= 0b000 << 8,
+		ICR_LO_delivery_mode_lowest_priority			= 0b001 << 8,
+		ICR_LO_delivery_mode_smi						= 0b010 << 8,
+		ICR_LO_delivery_mode_nmi						= 0b100 << 8,
+		ICR_LO_delivery_mode_init						= 0b101 << 8,
+		ICR_LO_delivery_mode_start_up					= 0b110 << 8,
+
+		ICR_LO_destination_mode_physical				= 0 << 11,
+		ICR_LO_destination_mode_logical					= 1 << 11,
+
+		ICR_LO_delivery_status_idle						= 0 << 12,
+		ICR_LO_delivery_status_send_pending				= 1 << 12,
+
+		ICR_LO_level_deassert							= 0 << 14,
+		ICR_LO_level_assert								= 1 << 14,
+
+		ICR_LO_trigger_mode_edge						= 0 << 15,
+		ICR_LO_trigger_mode_level						= 1 << 15,
+
+		ICR_LO_destination_shorthand_none				= 0b00 << 18,
+		ICR_LO_destination_shorthand_self				= 0b01 << 18,
+		ICR_LO_destination_shorthand_all_including_self	= 0b10 << 18,
+		ICR_LO_destination_shorthand_all_excluding_self	= 0b11 << 18,
+	};
+
 	struct MADT : public Kernel::ACPI::SDTHeader
 	{
 		uint32_t local_apic;
@@ -208,7 +237,7 @@ namespace Kernel
 				write_to_local_apic(LAPIC_ICR_HI_REG, (read_from_local_apic(LAPIC_ICR_HI_REG) & 0x00FFFFFF) | (processor << 24));
 				write_to_local_apic(LAPIC_ICR_LO_REG, data);
 				udelay(ud);
-				while (read_from_local_apic(LAPIC_ICR_LO_REG) & (1 << 12))
+				while ((read_from_local_apic(LAPIC_ICR_LO_REG) & ICR_LO_delivery_status_send_pending) == ICR_LO_delivery_status_send_pending)
 					__builtin_ia32_pause();
 			};
 
@@ -232,31 +261,55 @@ namespace Kernel
 
 			dprintln("Trying to enable processor (lapic id {})", processor.apic_id);
 
-			Kernel::Processor::create(processor.processor_id);
-
+			auto& proc = Kernel::Processor::create(processor.apic_id);
 			PageTable::with_fast_page((paddr_t)g_ap_init_addr, [&] {
-				PageTable::fast_page_as_sized<uint32_t>(2) = V2P(Kernel::Processor::get(processor.processor_id).stack_top());
+				PageTable::fast_page_as_sized<uint32_t>(2) = V2P(proc.stack_top());
 			});
 			*g_ap_stack_loaded = 0;
 
 			write_to_local_apic(LAPIC_ERROR_REG, 0x00);
-			send_ipi(processor.processor_id, (read_from_local_apic(LAPIC_ICR_LO_REG) & 0xFFF00000) | 0x0000C500, 0);
-			send_ipi(processor.processor_id, (read_from_local_apic(LAPIC_ICR_LO_REG) & 0xFFF0F800) | 0x00008500, 0);
+
+			// send INIT IPI
+			send_ipi(processor.apic_id,
+				(read_from_local_apic(LAPIC_ICR_LO_REG) & ICR_LO_reserved_mask)
+				| ICR_LO_delivery_mode_init
+				| ICR_LO_destination_mode_physical
+				| ICR_LO_level_assert
+				| ICR_LO_trigger_mode_edge
+				| ICR_LO_destination_shorthand_none
+				, 0
+			);
+
+			// TODO: If we are on processor predating Pentium, we need to send deassert
 
 			udelay(10 * 1000);
 
 			for (int i = 0; i < 2; i++)
 			{
 				write_to_local_apic(LAPIC_ERROR_REG, 0x00);
-				send_ipi(processor.processor_id, (read_from_local_apic(LAPIC_ICR_LO_REG) & 0xFFF0F800) | 0x00000600 | ap_init_page, 200);
+
+				// send 2 SETUP IPIs with 200 us delay
+				send_ipi(processor.apic_id,
+					(read_from_local_apic(LAPIC_ICR_LO_REG) & ICR_LO_reserved_mask)
+					| ICR_LO_delivery_mode_start_up
+					| ICR_LO_destination_mode_physical
+					| ICR_LO_level_assert
+					| ICR_LO_trigger_mode_edge
+					| ICR_LO_destination_shorthand_none
+					| ap_init_page
+					, 200
+				);
 			}
 
-			// give processor upto 100 * 100 us (10 ms to boot)
+			// give processor upto 100 * 100 us + 200 us to boot
 			for (int i = 0; *g_ap_stack_loaded == 0 && i < 100; i++)
 				udelay(100);
 		}
 
 		*g_ap_startup_done = 1;
+
+		// give processors 100 us time to increment running count
+		udelay(100);
 		dprintln("{} processors started", *g_ap_running_count);
 	}
 

kernel/kernel/CPUID.cpp

@@ -50,6 +50,13 @@ namespace CPUID
 		return buffer[3] & (1 << 20);
 	}
 
+	bool has_pge()
+	{
+		uint32_t ecx, edx;
+		get_features(ecx, edx);
+		return edx & CPUID::EDX_PGE;
+	}
+
 	const char* feature_string_ecx(uint32_t feat)
 	{
 		switch (feat)

kernel/kernel/FS/TmpFS/FileSystem.cpp

@@ -128,9 +128,9 @@ namespace Kernel
 		auto inode_location = find_inode(ino);
 		PageTable::with_fast_page(inode_location.paddr, [&] {
 			auto& inode_info = PageTable::fast_page_as_sized<TmpInodeInfo>(inode_location.index);
-			ASSERT_EQ(inode_info.nlink, 0);
+			ASSERT(inode_info.nlink == 0);
 			for (auto paddr : inode_info.block)
-				ASSERT_EQ(paddr, 0);
+				ASSERT(paddr == 0);
 			inode_info = {};
 		});
 		ASSERT(!m_inode_cache.contains(ino));
@@ -166,8 +166,8 @@ namespace Kernel
 	{
 		LockGuard _(m_mutex);
 
-		ASSERT_GTE(ino, first_inode);
-		ASSERT_LT(ino, max_inodes);
+		ASSERT(ino >= first_inode);
+		ASSERT(ino < max_inodes);
 
 		constexpr size_t inodes_per_page = PAGE_SIZE / sizeof(TmpInodeInfo);
 
@@ -220,7 +220,7 @@ namespace Kernel
 	{
 		LockGuard _(m_mutex);
 
-		ASSERT_GT(index, 0);
+		ASSERT(index > 0);
 		return find_indirect(m_data_pages, index - first_data_page, 3);
 	}
 

kernel/kernel/Input/PS2/Keymap.cpp

@@ -1,6 +1,9 @@
+#include <kernel/Debug.h>
 #include <kernel/Input/KeyEvent.h>
 #include <kernel/Input/PS2/Keymap.h>
 
+#include <string.h>
+
 namespace Kernel::Input
 {
 

kernel/kernel/InterruptController.cpp

@@ -8,29 +8,8 @@
 namespace Kernel
 {
 
-	namespace IDT { void register_irq_handler(uint8_t irq, Interruptable*); }
-
 	static InterruptController* s_instance = nullptr;
 
-	void Interruptable::set_irq(int irq)
-	{
-		if (m_irq != -1)
-			IDT::register_irq_handler(m_irq, nullptr);
-		m_irq = irq;
-		IDT::register_irq_handler(irq, this);
-	}
-
-	void Interruptable::enable_interrupt()
-	{
-		ASSERT(m_irq != -1);
-		InterruptController::get().enable_irq(m_irq);
-	}
-
-	void Interruptable::disable_interrupt()
-	{
-		ASSERT_NOT_REACHED();
-	}
-
 	InterruptController& InterruptController::get()
 	{
 		ASSERT(s_instance);

kernel/kernel/Interruptable.cpp

@@ -0,0 +1,28 @@
+#include <kernel/IDT.h>
+#include <kernel/Interruptable.h>
+#include <kernel/InterruptController.h>
+#include <kernel/Processor.h>
+
+namespace Kernel
+{
+
+	void Interruptable::set_irq(int irq)
+	{
+		if (m_irq != -1)
+			Processor::idt().register_irq_handler(m_irq, nullptr);
+		m_irq = irq;
+		Processor::idt().register_irq_handler(irq, this);
+	}
+
+	void Interruptable::enable_interrupt()
+	{
+		ASSERT(m_irq != -1);
+		InterruptController::get().enable_irq(m_irq);
+	}
+
+	void Interruptable::disable_interrupt()
+	{
+		ASSERT_NOT_REACHED();
+	}
+
+}

kernel/kernel/Lock/SpinLock.cpp

@@ -1,61 +0,0 @@
-#include <kernel/InterruptController.h>
-#include <kernel/Lock/SpinLock.h>
-#include <kernel/Scheduler.h>
-
-// FIXME: try to move these to header
-
-namespace Kernel
-{
-
-	InterruptState SpinLock::lock()
-	{
-		auto id = Processor::current_id();
-		ASSERT_NEQ(m_locker.load(), id);
-
-		auto state = Processor::get_interrupt_state();
-		Processor::set_interrupt_state(InterruptState::Disabled);
-
-		while (!m_locker.compare_exchange(PROCESSOR_NONE, id, BAN::MemoryOrder::memory_order_acquire))
-			__builtin_ia32_pause();
-
-		return state;
-	}
-
-	void SpinLock::unlock(InterruptState state)
-	{
-		ASSERT_EQ(m_locker.load(), Processor::current_id());
-		m_locker.store(PROCESSOR_NONE, BAN::MemoryOrder::memory_order_release);
-		Processor::set_interrupt_state(state);
-	}
-
-	InterruptState RecursiveSpinLock::lock()
-	{
-		auto id = Processor::current_id();
-
-		auto state = Processor::get_interrupt_state();
-		Processor::set_interrupt_state(InterruptState::Disabled);
-
-		if (id == m_locker)
-			ASSERT_GT(m_lock_depth, 0);
-		else
-		{
-			while (!m_locker.compare_exchange(PROCESSOR_NONE, id, BAN::MemoryOrder::memory_order_acquire))
-				__builtin_ia32_pause();
-			ASSERT_EQ(m_lock_depth, 0);
-		}
-
-		m_lock_depth++;
-
-		return state;
-	}
-
-	void RecursiveSpinLock::unlock(InterruptState state)
-	{
-		ASSERT_EQ(m_locker.load(), Processor::current_id());
-		ASSERT_GT(m_lock_depth, 0);
-		if (--m_lock_depth == 0)
-			m_locker.store(PROCESSOR_NONE, BAN::MemoryOrder::memory_order_release);
-		Processor::set_interrupt_state(state);
-	}
-
-}

kernel/kernel/Memory/FileBackedRegion.cpp

@@ -129,7 +129,7 @@ namespace Kernel
 			size_t file_offset = m_offset + (vaddr - m_vaddr);
 			size_t bytes = BAN::Math::min<size_t>(m_size - file_offset, PAGE_SIZE);
 
-			ASSERT_EQ(&PageTable::current(), &m_page_table);
+			ASSERT(&PageTable::current() == &m_page_table);
 			auto read_ret = m_inode->read(file_offset, BAN::ByteSpan((uint8_t*)vaddr, bytes));
 
 			if (read_ret.is_error())

kernel/kernel/Memory/VirtualRange.cpp

@@ -158,9 +158,9 @@ namespace Kernel
 			return;
 
 		// Verify no overflow
-		ASSERT_LTE(bytes, size());
-		ASSERT_LTE(offset, size());
-		ASSERT_LTE(offset, size() - bytes);
+		ASSERT(bytes <= size());
+		ASSERT(offset <= size());
+		ASSERT(offset <= size() - bytes);
 
 		if (&PageTable::current() == &m_page_table || &PageTable::kernel() == &m_page_table)
 		{

kernel/kernel/Networking/E1000/E1000.cpp

@@ -259,7 +259,7 @@ namespace Kernel
 
 	BAN::ErrorOr<void> E1000::send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::ConstByteSpan buffer)
 	{
-		ASSERT_LTE(buffer.size() + sizeof(EthernetHeader), E1000_TX_BUFFER_SIZE);
+		ASSERT(buffer.size() + sizeof(EthernetHeader) <= E1000_TX_BUFFER_SIZE);
 
 		SpinLockGuard _(m_lock);
 
@@ -299,7 +299,7 @@ namespace Kernel
 			auto& descriptor = reinterpret_cast<volatile e1000_rx_desc*>(m_rx_descriptor_region->vaddr())[rx_current];
 			if (!(descriptor.status & 1))
 				break;
-			ASSERT_LTE((uint16_t)descriptor.length, E1000_RX_BUFFER_SIZE);
+			ASSERT(descriptor.length <= E1000_RX_BUFFER_SIZE);
 
 			NetworkManager::get().on_receive(*this, BAN::ConstByteSpan {
 				reinterpret_cast<const uint8_t*>(m_rx_buffer_region->vaddr() + rx_current * E1000_RX_BUFFER_SIZE),

kernel/kernel/Networking/TCPSocket.cpp

@@ -455,7 +455,7 @@ namespace Kernel
 				if (m_send_window.data_size > 0 && m_send_window.current_ack - m_send_window.has_ghost_byte > m_send_window.start_seq)
 				{
 					uint32_t acknowledged_bytes = m_send_window.current_ack - m_send_window.start_seq - m_send_window.has_ghost_byte;
-					ASSERT_LTE(acknowledged_bytes, m_send_window.data_size);
+					ASSERT(acknowledged_bytes <= m_send_window.data_size);
 
 					m_send_window.data_size -= acknowledged_bytes;
 					m_send_window.start_seq += acknowledged_bytes;

kernel/kernel/Panic.cpp

@@ -1,6 +1,6 @@
 #include <kernel/Panic.h>
 
-namespace Kernel::detail
+namespace Kernel
 {
-	bool g_paniced = false;
+	volatile bool g_paniced = false;
 }

kernel/kernel/Processor.cpp

@@ -1,34 +1,70 @@
-#include <BAN/Vector.h>
+#include <kernel/Memory/kmalloc.h>
 #include <kernel/Processor.h>
 
+#include <kernel/Debug.h>
+
 namespace Kernel
 {
 
-	static BAN::Vector<Processor> s_processors;
+	static constexpr uint32_t MSR_IA32_GS_BASE = 0xC0000101;
+
+	ProcessorID Processor::s_bsb_id { PROCESSOR_NONE };
+
+	static BAN::Array<Processor, 0xFF> s_processors;
+
+	static ProcessorID read_processor_id()
+	{
+		uint8_t id;
+		asm volatile(
+			"movl $1, %%eax;"
+			"cpuid;"
+			"shrl $24, %%ebx;"
+			"movb %%bl, %0;"
+			: "=rm"(id)
+			:: "eax", "ebx", "ecx", "edx"
+		);
+		return id;
+	}
 
 	Processor& Processor::create(ProcessorID id)
 	{
-		while (s_processors.size() <= id)
-			MUST(s_processors.emplace_back());
+		// bsb is the first processor
+		if (s_bsb_id == PROCESSOR_NONE)
+			s_bsb_id = id = read_processor_id();
+
 		auto& processor = s_processors[id];
-		if (processor.m_stack == nullptr)
-		{
-			processor.m_stack = kmalloc(m_stack_size, 4096, true);
-			ASSERT(processor.m_stack);
-		}
+
+		ASSERT(processor.m_id == PROCESSOR_NONE);
+		processor.m_id = id;
+
+		processor.m_stack = kmalloc(s_stack_size, 4096, true);
+		ASSERT(processor.m_stack);
+
+		processor.m_gdt = GDT::create();
+		ASSERT(processor.m_gdt);
+
+		processor.m_idt = IDT::create(id == s_bsb_id);
+		ASSERT(processor.m_idt);
+
 		return processor;
 	}
 
-	Processor::~Processor()
+	Processor& Processor::initialize()
 	{
-		if (m_stack)
-			kfree(m_stack);
-		m_stack = nullptr;
-	}
+		auto id = read_processor_id();
+		auto& processor = s_processors[id];
 
-	Processor& Processor::get(ProcessorID id)
-	{
-		return s_processors[id];
+		// set gs base to pointer to this processor
+		uint64_t ptr = reinterpret_cast<uint64_t>(&processor);
+		asm volatile("wrmsr" :: "d"(ptr >> 32), "a"(ptr), "c"(MSR_IA32_GS_BASE));
+
+		ASSERT(processor.m_gdt);
+		processor.gdt().load();
+
+		ASSERT(processor.m_idt);
+		processor.idt().load();
+
+		return processor;
 	}
 
 }

kernel/kernel/Scheduler.cpp

@@ -18,7 +18,7 @@ namespace Kernel
 
 	ALWAYS_INLINE static void load_temp_stack()
 	{
-		asm volatile("movq %0, %%rsp" :: "rm"(Processor::current().stack_top()));
+		asm volatile("movq %0, %%rsp" :: "rm"(Processor::current_stack_top()));
 	}
 
 	BAN::ErrorOr<void> Scheduler::initialize()
@@ -64,7 +64,7 @@ namespace Kernel
 		auto state = m_lock.lock();
 		wake_threads();
 		if (save_current_thread())
-			return m_lock.unlock(state);
+			return Processor::set_interrupt_state(state);
 		advance_current_thread();
 		execute_current_thread_locked();
 		ASSERT_NOT_REACHED();
@@ -86,7 +86,7 @@ namespace Kernel
 		if (m_active_threads.empty() || &current_thread() != m_idle_thread)
 			return m_lock.unlock(state);
 		if (save_current_thread())
-			return m_lock.unlock(state);
+			return Processor::set_interrupt_state(state);
 		m_current_thread = m_active_threads.begin();
 		execute_current_thread_locked();
 		ASSERT_NOT_REACHED();
@@ -94,7 +94,7 @@ namespace Kernel
 
 	void Scheduler::wake_threads()
 	{
-		ASSERT(m_lock.is_locked());
+		ASSERT(m_lock.current_processor_has_lock());
 
 		uint64_t current_time = SystemTimer::get().ms_since_boot();
 		while (!m_sleeping_threads.empty() && m_sleeping_threads.front().wake_time <= current_time)
@@ -124,7 +124,7 @@ namespace Kernel
 
 	void Scheduler::advance_current_thread()
 	{
-		ASSERT(m_lock.is_locked());
+		ASSERT(m_lock.current_processor_has_lock());
 
 		if (m_active_threads.empty())
 		{
@@ -137,7 +137,7 @@ namespace Kernel
 
 	void Scheduler::remove_and_advance_current_thread()
 	{
-		ASSERT(m_lock.is_locked());
+		ASSERT(m_lock.current_processor_has_lock());
 
 		ASSERT(m_current_thread);
 
@@ -158,7 +158,7 @@ namespace Kernel
 	//       after getting the rsp
 	ALWAYS_INLINE bool Scheduler::save_current_thread()
 	{
-		ASSERT(m_lock.is_locked());
+		ASSERT(m_lock.current_processor_has_lock());
 
 		uintptr_t rsp, rip;
 		push_callee_saved();
@@ -209,7 +209,7 @@ namespace Kernel
 
 	void Scheduler::execute_current_thread_locked()
 	{
-		ASSERT(m_lock.is_locked());
+		ASSERT(m_lock.current_processor_has_lock());
 		load_temp_stack();
 		PageTable::kernel().load();
 		execute_current_thread_stack_loaded();
@@ -218,12 +218,12 @@ namespace Kernel
 
 	NEVER_INLINE void Scheduler::execute_current_thread_stack_loaded()
 	{
-		ASSERT(m_lock.is_locked());
+		ASSERT(m_lock.current_processor_has_lock());
 
 #if SCHEDULER_VERIFY_STACK
 		vaddr_t rsp;
 		read_rsp(rsp);
-		ASSERT(Processor::current().stack_bottom() <= rsp && rsp <= Processor::current().stack_top());
+		ASSERT(Processor::current_stack_bottom() <= rsp && rsp <= Processor::current_stack_top());
 		ASSERT(&PageTable::current() == &PageTable::kernel());
 #endif
 
@@ -256,7 +256,7 @@ namespace Kernel
 		if (current->has_process())
 		{
 			current->process().page_table().load();
-			GDT::set_tss_stack(current->interrupt_stack_base() + current->interrupt_stack_size());
+			Processor::gdt().set_tss_stack(current->interrupt_stack_base() + current->interrupt_stack_size());
 		}
 		else
 			PageTable::kernel().load();
@@ -281,7 +281,7 @@ namespace Kernel
 
 	void Scheduler::set_current_thread_sleeping_impl(uint64_t wake_time)
 	{
-		ASSERT(m_lock.is_locked());
+		ASSERT(m_lock.current_processor_has_lock());
 
 		if (save_current_thread())
 			return;
@@ -307,16 +307,18 @@ namespace Kernel
 
 	void Scheduler::set_current_thread_sleeping(uint64_t wake_time)
 	{
-		SpinLockGuard _(m_lock);
+		auto state = m_lock.lock();
 		m_current_thread->semaphore = nullptr;
 		set_current_thread_sleeping_impl(wake_time);
+		Processor::set_interrupt_state(state);
 	}
 
 	void Scheduler::block_current_thread(Semaphore* semaphore, uint64_t wake_time)
 	{
-		SpinLockGuard _(m_lock);
+		auto state = m_lock.lock();
 		m_current_thread->semaphore = semaphore;
 		set_current_thread_sleeping_impl(wake_time);
+		Processor::set_interrupt_state(state);
 	}
 
 	void Scheduler::unblock_threads(Semaphore* semaphore)

kernel/kernel/Storage/ATA/ATAController.cpp

@@ -1,4 +1,5 @@
 #include <kernel/FS/DevFS/FileSystem.h>
+#include <kernel/InterruptController.h>
 #include <kernel/Storage/ATA/AHCI/Controller.h>
 #include <kernel/Storage/ATA/ATABus.h>
 #include <kernel/Storage/ATA/ATAController.h>

kernel/kernel/Thread.cpp

@@ -170,7 +170,7 @@ namespace Kernel
 		// Signal mask is inherited
 
 		// Setup stack for returning
-		ASSERT_EQ(m_rsp % PAGE_SIZE, 0u);
+		ASSERT(m_rsp % PAGE_SIZE == 0);
 		PageTable::with_fast_page(process().page_table().physical_address_of(m_rsp - PAGE_SIZE), [&] {
 			uintptr_t rsp = PageTable::fast_page() + PAGE_SIZE;
 			write_to_stack(rsp, nullptr); // alignment
@@ -199,7 +199,7 @@ namespace Kernel
 		m_signal_pending_mask = 0;
 		m_signal_block_mask = ~0ull;
 
-		ASSERT_EQ(m_rsp % PAGE_SIZE, 0u);
+		ASSERT(m_rsp % PAGE_SIZE == 0);
 		PageTable::with_fast_page(process().page_table().physical_address_of(m_rsp - PAGE_SIZE), [&] {
 			uintptr_t rsp = PageTable::fast_page() + PAGE_SIZE;
 			write_to_stack(rsp, nullptr); // alignment

kernel/kernel/kernel.cpp

@@ -102,16 +102,12 @@ extern "C" void kernel_main(uint32_t boot_magic, uint32_t boot_info)
 	parse_boot_info(boot_magic, boot_info);
 	dprintln("boot info parsed");
 
-	Processor::create(Processor::current_id());
+	Processor::create(0);
+	Processor::initialize();
 	dprintln("BSP initialized");
 
-	GDT::initialize();
-	dprintln("GDT initialized");
-
-	IDT::initialize();
-	dprintln("IDT initialized");
-
 	PageTable::initialize();
+	PageTable::kernel().initial_load();
 	dprintln("PageTable initialized");
 
 	Heap::initialize();
@@ -129,8 +125,6 @@ extern "C" void kernel_main(uint32_t boot_magic, uint32_t boot_info)
 	SystemTimer::initialize(cmdline.force_pic);
 	dprintln("Timers initialized");
 
-	InterruptController::get().initialize_multiprocessor();
-
 	DevFileSystem::initialize();
 	dprintln("devfs initialized");
 
@@ -143,6 +137,8 @@ extern "C" void kernel_main(uint32_t boot_magic, uint32_t boot_info)
 	if (g_terminal_driver)
 		dprintln("Framebuffer terminal initialized");
 
+	InterruptController::get().initialize_multiprocessor();
+
 	ProcFileSystem::initialize();
 	dprintln("procfs initialized");
 
@@ -215,8 +211,11 @@ extern "C" void ap_main()
 {
 	using namespace Kernel;
 
-	dprintln("hello from processor {}", Processor::current_id());
+	Processor::initialize();
+	PageTable::kernel().initial_load();
+
+	dprintln("ap{} initialized", Processor::current_id());
 
 	for (;;)
-		asm volatile("");
+		asm volatile("hlt");
 }

libc/CMakeLists.txt

@@ -30,6 +30,8 @@ set(LIBC_SOURCES
 	unistd.cpp
 	math.S
 	icxxabi.cpp
+
+	../BAN/BAN/Assert.cpp
 )
 
 add_custom_target(libc-headers