WindowServer: Start TaskBar and Terminal after starting up

This removes the race condition from `start-gui` not starting WindowServer fast enough and leading to the apps crashing as they timeout while connecting to the server
Kernel: Fix 32 bit target compilation
2026-08-22 16:05:54 +03:00 · 2026-08-22 15:33:49 +03:00 · 2026-08-22 14:45:23 +03:00 · 2026-08-22 14:39:42 +03:00 · 2026-08-22 14:39:42 +03:00 · 2026-08-22 14:39:42 +03:00
122 changed files with 1676 additions and 997 deletions
@@ -82,6 +82,37 @@ namespace BAN::Math
 		return x + 1;
 	}

+	template<unsigned_integral T>
+	requires is_same_v<T, unsigned int> || is_same_v<T, unsigned long> || is_same_v<T, unsigned long long>
+	inline constexpr int clz(T x)
+	{
+		if constexpr (is_same_v<T, unsigned int>)
+			return __builtin_clz(x);
+		if constexpr (is_same_v<T, unsigned long>)
+			return __builtin_clzl(x);
+		return __builtin_clzll(x);
+	}
+
+	template<unsigned_integral T> requires(sizeof(T) <= sizeof(unsigned long long))
+	inline constexpr int ctz(T x)
+	{
+		if constexpr (sizeof(T) <= sizeof(unsigned int))
+			return __builtin_ctz(x);
+		if constexpr (sizeof(T) <= sizeof(unsigned long))
+			return __builtin_ctzl(x);
+		return __builtin_ctzll(x);
+	}
+
+	template<unsigned_integral T> requires(sizeof(T) <= sizeof(unsigned long long))
+	inline constexpr int popcount(T x)
+	{
+		if constexpr (sizeof(T) <= sizeof(unsigned int))
+			return __builtin_popcount(x);
+		if constexpr (sizeof(T) <= sizeof(unsigned long))
+			return __builtin_popcountl(x);
+		return __builtin_popcountll(x);
+	}
+
 	template<integral T>
 	__attribute__((always_inline))
 	inline constexpr bool will_multiplication_overflow(T a, T b)
@@ -102,11 +133,7 @@ namespace BAN::Math
 	requires is_same_v<T, unsigned int> || is_same_v<T, unsigned long> || is_same_v<T, unsigned long long>
 	inline constexpr T ilog2(T x)
 	{
-		if constexpr(is_same_v<T, unsigned int>)
-			return sizeof(T) * 8 - __builtin_clz(x) - 1;
-		if constexpr(is_same_v<T, unsigned long>)
-			return sizeof(T) * 8 - __builtin_clzl(x) - 1;
-		return sizeof(T) * 8 - __builtin_clzll(x) - 1;
+		return sizeof(T) * 8 - clz(x) - 1;
 	}

 // This is ugly but my clangd does not like including
@@ -51,6 +51,8 @@ set(KERNEL_SOURCES
 	kernel/Interruptable.cpp
 	kernel/InterruptController.cpp
 	kernel/kernel.cpp
+	kernel/Lock/Mutex.cpp
+	kernel/Lock/RWLock.cpp
 	kernel/Lock/SpinLock.cpp
 	kernel/Memory/ByteRingBuffer.cpp
 	kernel/Memory/DMARegion.cpp
@@ -83,6 +85,7 @@ set(KERNEL_SOURCES
 	kernel/Processor.cpp
 	kernel/Random.cpp
 	kernel/Scheduler.cpp
+	kernel/SchedulerThreadNode.cpp
 	kernel/SSP.cpp
 	kernel/Storage/ATA/AHCI/Controller.cpp
 	kernel/Storage/ATA/AHCI/Device.cpp
@@ -63,7 +63,7 @@ sys_fork_trampoline:

 	call read_ip
 	testl %eax, %eax
-	jz .done
+	jz .Lsys_fork_trampoline_done

 	movl %esp, %ebx

@@ -73,7 +73,7 @@ sys_fork_trampoline:
 	call sys_fork
 	addl $16, %esp

- .done:
+ .Lsys_fork_trampoline_done:
 	popl %edi
 	popl %esi
 	popl %ebx
@@ -28,27 +28,26 @@ safe_user_strncpy:
 	testl %ecx, %ecx
 	jz safe_user_strncpy_fault

- .safe_user_strncpy_loop:
+ .Lsafe_user_strncpy_loop:
 	movb (%esi), %al
 	movb %al, (%edi)
 	testb %al, %al
-	jz .safe_user_strncpy_done
+	jz .Lsafe_user_strncpy_done

 	incl %edi
 	incl %esi
 	decl %ecx
-	jnz .safe_user_strncpy_loop
+	jnz .Lsafe_user_strncpy_loop

 safe_user_strncpy_fault:
 	xorl %eax, %eax
-	jmp .safe_user_strncpy_return
+	jmp .Lsafe_user_strncpy_return

- .safe_user_strncpy_done:
+ .Lsafe_user_strncpy_done:
 	movl $1, %eax

- .safe_user_strncpy_return:
+ .Lsafe_user_strncpy_return:
 	movl 4(%esp), %edi
 	movl 8(%esp), %esi
 	ret
-
 safe_user_strncpy_end:
@@ -168,14 +168,12 @@ has_sse:

 check_requirements:
 	call has_cpuid
-	jz .exit
+	jz system_halt
 	call has_pae
-	jz .exit
+	jz system_halt
 	call has_sse
-	jz .exit
+	jz system_halt
 	ret
-.exit:
-	jmp system_halt

 enable_sse:
 	movl %cr0, %eax
@@ -225,8 +223,8 @@ _start:

 	# load boot GDT
 	lgdt V2P(boot_gdtr)
-	ljmpl $0x08, $V2P(gdt_flush)
-gdt_flush:
+	ljmpl $0x08, $V2P(.Lgdt_flush)
+.Lgdt_flush:
 	# set correct segment registers
 	movw $0x10, %ax
 	movw %ax, %ds
@@ -243,10 +241,10 @@ gdt_flush:
 	movl $g_boot_stack_top, %esp

 	# jump to higher half
-	leal higher_half, %ecx
+	leal .Lhigher_half, %ecx
 	jmp *%ecx

-higher_half:
+.Lhigher_half:
 	# call global constuctors
 	call _init

@@ -298,18 +296,18 @@ ap_ready:
 	.skip 4

 1:	cli; cld
-	ljmpl $0x00, $AP_REL(ap_cs_clear)
+	ljmpl $0x00, $AP_REL(.Lap_cs_clear)

-ap_cs_clear:
+.Lap_cs_clear:
 	# load ap gdt and enter protected mode
 	lgdt AP_REL(ap_gdtr)
 	movl %cr0, %eax
 	orb $1, %al
 	movl %eax, %cr0
-	ljmpl $0x08, $AP_REL(ap_protected_mode)
+	ljmpl $0x08, $AP_REL(.Lap_protected_mode)

 .code32
-ap_protected_mode:
+.Lap_protected_mode:
 	movw $0x10, %ax
 	movw %ax, %ds
 	movw %ax, %ss
@@ -323,13 +321,13 @@ ap_protected_mode:

 	# load boot gdt and enter long mode
 	lgdt V2P(boot_gdtr)
-	ljmpl $0x08, $AP_REL(ap_flush_gdt)
+	ljmpl $0x08, $AP_REL(.Lap_flush_gdt)

-ap_flush_gdt:
-	movl $ap_higher_half, %ecx
+.Lap_flush_gdt:
+	movl $.Lap_higher_half, %ecx
 	jmp *%ecx

-ap_higher_half:
+.Lap_higher_half:
 	movl AP_REL(ap_prepare_paging), %eax
 	call *%eax

@@ -33,13 +33,13 @@ sys_fork_trampoline:

 	call read_ip
 	testq %rax, %rax
-	jz .done
+	jz .Lsys_fork_trampoline_done

 	movq %rax, %rsi
 	movq %rsp, %rdi
 	call sys_fork

- .done:
+ .Lsys_fork_trampoline_done:
 	popq %r15
 	popq %r14
 	popq %r13
@@ -20,28 +20,28 @@ safe_user_strncpy:
 	testq %rcx, %rcx
 	jz safe_user_strncpy_fault

- .safe_user_strncpy_align_loop:
+ .Lsafe_user_strncpy_align_loop:
 	testb $0x7, %sil
-	jz .safe_user_strncpy_align_done
+	jz .Lsafe_user_strncpy_align_done

 	movb (%rsi), %al
 	movb %al, (%rdi)
 	testb %al, %al
-	jz .safe_user_strncpy_done
+	jz .Lsafe_user_strncpy_done

 	incq %rdi
 	incq %rsi
 	decq %rcx
-	jnz .safe_user_strncpy_align_loop
+	jnz .Lsafe_user_strncpy_align_loop
 	jmp safe_user_strncpy_fault

- .safe_user_strncpy_align_done:
+ .Lsafe_user_strncpy_align_done:
 	movq $0x0101010101010101, %r8
 	movq $0x8080808080808080, %r9

- .safe_user_strncpy_qword_loop:
+ .Lsafe_user_strncpy_qword_loop:
 	cmpq $8, %rcx
-	jb .safe_user_strncpy_qword_done
+	jb .Lsafe_user_strncpy_qword_done

 	movq (%rsi), %rax
 	movq %rax, %r10
@@ -52,36 +52,36 @@ safe_user_strncpy:
 	notq %r11
 	andq %r11, %r10
 	andq %r9, %r10
-	jnz .safe_user_strncpy_byte_loop
+	jnz .Lsafe_user_strncpy_byte_loop

 	movq %rax, (%rdi)

 	addq $8, %rdi
 	addq $8, %rsi
 	subq $8, %rcx
-	jnz .safe_user_strncpy_qword_loop
+	jnz .Lsafe_user_strncpy_qword_loop
 	jmp safe_user_strncpy_fault

- .safe_user_strncpy_qword_done:
+ .Lsafe_user_strncpy_qword_done:
 	testq %rcx, %rcx
 	jz safe_user_strncpy_fault

- .safe_user_strncpy_byte_loop:
+ .Lsafe_user_strncpy_byte_loop:
 	movb (%rsi), %al
 	movb %al, (%rdi)
 	testb %al, %al
-	jz .safe_user_strncpy_done
+	jz .Lsafe_user_strncpy_done

 	incq %rdi
 	incq %rsi
 	decq %rcx
-	jnz .safe_user_strncpy_byte_loop
+	jnz .Lsafe_user_strncpy_byte_loop

 safe_user_strncpy_fault:
 	xorq %rax, %rax
 	ret

- .safe_user_strncpy_done:
+ .Lsafe_user_strncpy_done:
 	movb $1, %al
 	ret
 safe_user_strncpy_end:
@@ -161,12 +161,10 @@ is_64_bit:

 check_requirements:
 	call has_cpuid
-	jz .exit
+	jz system_halt
 	call is_64_bit
-	jz .exit
+	jz system_halt
 	ret
-.exit:
-	jmp system_halt

 enable_sse:
 	movl %cr0, %eax
@@ -226,10 +224,10 @@ _start:

 	# flush gdt and jump to 64 bit
 	lgdt V2P(boot_gdtr)
-	ljmpl $0x08, $V2P(long_mode)
+	ljmpl $0x08, $V2P(.Llong_mode)

 .code64
-long_mode:
+.Llong_mode:
 	movw $0x10, %ax
 	movw %ax, %ds
 	movw %ax, %ss
@@ -240,10 +238,10 @@ long_mode:
 	addq $KERNEL_OFFSET, %rsp

 	# jump to higher half
-	movabsq $higher_half, %rcx
+	movabsq $.Lhigher_half, %rcx
 	jmp *%rcx

-higher_half:
+.Lhigher_half:
 	# call global constuctors
 	call _init

@@ -293,18 +291,18 @@ ap_ready:
 	.skip 8

 1:	cli; cld
-	ljmpl $0x00, $AP_REL(ap_cs_clear)
+	ljmpl $0x00, $AP_REL(.Lap_cs_clear)

-ap_cs_clear:
+.Lap_cs_clear:
 	# load ap gdt and enter protected mode
 	lgdt AP_REL(ap_gdtr)
 	movl %cr0, %eax
 	orb $1, %al
 	movl %eax, %cr0
-	ljmpl $0x08, $AP_REL(ap_protected_mode)
+	ljmpl $0x08, $AP_REL(.Lap_protected_mode)

 .code32
-ap_protected_mode:
+.Lap_protected_mode:
 	movw $0x10, %ax
 	movw %ax, %ds
 	movw %ax, %ss
@@ -318,14 +316,14 @@ ap_protected_mode:

 	# load boot gdt and enter long mode
 	lgdt V2P(boot_gdtr)
-	ljmpl $0x08, $AP_REL(ap_long_mode)
+	ljmpl $0x08, $AP_REL(.Lap_long_mode)

 .code64
-ap_long_mode:
-	movq $ap_higher_half, %rax
+.Lap_long_mode:
+	movq $.Lap_higher_half, %rax
 	jmp *%rax

-ap_higher_half:
+.Lap_higher_half:
 	movq AP_REL(ap_prepare_paging), %rax
 	call *%rax

@@ -38,25 +38,39 @@ extern "C" void __cxa_finalize(void* f)
 	}
 };

-namespace __cxxabiv1
+extern "C" int __cxa_guard_acquire(uint64_t* g)
 {
-	using __guard = uint64_t;
+	uint8_t* guard = reinterpret_cast<uint8_t*>(g);

-	extern "C" int __cxa_guard_acquire (__guard* g)
+	for (;;)
 	{
-		uint8_t* byte = reinterpret_cast<uint8_t*>(g);
-		uint8_t zero = 0;
-		return __atomic_compare_exchange_n(byte, &zero, 1, false, __ATOMIC_ACQUIRE, __ATOMIC_ACQUIRE);
+		if (BAN::atomic_load(guard[0], BAN::memory_order_acquire))
+			return 0;
+
+		uint8_t expected = 0;
+		if (BAN::atomic_compare_exchange(guard[1], expected, 1, BAN::memory_order_acquire))
+		{
+			if (BAN::atomic_load(guard[0], BAN::memory_order_acquire))
+			{
+				BAN::atomic_store(guard[1], 0, BAN::memory_order_release);
+				return 0;
+			}
+			return 1;
 		}

-	extern "C" void __cxa_guard_release (__guard* g)
-	{
-		uint8_t* byte = reinterpret_cast<uint8_t*>(g);
-		__atomic_store_n(byte, 0, __ATOMIC_RELEASE);
-	}
-
-	extern "C" void __cxa_guard_abort (__guard*)
-	{
-		Kernel::panic("__cxa_guard_abort");
+		while (BAN::atomic_load(guard[1], BAN::memory_order_acquire))
+			Kernel::Processor::pause();
 	}
 }
+
+extern "C" void __cxa_guard_release(uint64_t* g)
+{
+	uint8_t* guard = reinterpret_cast<uint8_t*>(g);
+	BAN::atomic_store(guard[0], 1, BAN::memory_order_release);
+	BAN::atomic_store(guard[1], 0, BAN::memory_order_release);
+}
+
+extern "C" void __cxa_guard_abort(uint64_t*)
+{
+	Kernel::panic("__cxa_guard_abort");
+}
@@ -4,6 +4,7 @@
 #include <kernel/ACPI/AML/Namespace.h>
 #include <kernel/ACPI/EmbeddedController.h>
 #include <kernel/ACPI/Headers.h>
+#include <kernel/Interruptable.h>
 #include <kernel/Memory/Types.h>
 #include <kernel/ThreadBlocker.h>

@@ -1,11 +1,12 @@
 #pragma once

 #include <BAN/Atomic.h>
+#include <BAN/Optional.h>
 #include <BAN/UniqPtr.h>

 #include <kernel/ACPI/AML/Scope.h>
 #include <kernel/Lock/Mutex.h>
-#include <kernel/Thread.h>
+#include <kernel/ThreadBlocker.h>

 namespace Kernel::ACPI
 {
@@ -2,6 +2,7 @@

 #include <BAN/Vector.h>
 #include <kernel/InterruptController.h>
+#include <kernel/InterruptNumbers.h>
 #include <kernel/Lock/SpinLock.h>
 #include <kernel/Memory/Types.h>

@@ -67,6 +67,7 @@ namespace Kernel
 		BAN::Vector<BAN::Vector<const HDAudio::AFGWidget*>> m_output_paths;
 		BAN::Vector<const HDAudio::AFGWidget*> m_output_pins;
 		size_t m_output_path_index { SIZE_MAX };
+		size_t m_amplifier_idx { SIZE_MAX };

 		uint8_t m_stream_id    { 0xFF };
 		uint8_t m_stream_index { 0xFF };
@@ -1,5 +1,6 @@
 #pragma once

+#include <BAN/Optional.h>
 #include <BAN/Vector.h>

 namespace Kernel::HDAudio
@@ -64,6 +65,7 @@ namespace Kernel::HDAudio
 		};

 		BAN::Optional<Amplifier> output_amplifier;
+		BAN::Optional<Amplifier> input_amplifier;

 		BAN::Vector<uint16_t> connections;
 	};
@@ -3,6 +3,7 @@
 #include <BAN/HashMap.h>
 #include <kernel/FS/Inode.h>
 #include <kernel/Lock/Mutex.h>
+#include <kernel/ThreadBlocker.h>

 #include <sys/epoll.h>

@@ -1,7 +1,6 @@
 #pragma once

 #include <BAN/Vector.h>
-#include <kernel/Device/Device.h>
 #include <kernel/FS/TmpFS/FileSystem.h>
 #include <kernel/Lock/Mutex.h>
 #include <kernel/ThreadBlocker.h>
@@ -9,6 +8,8 @@
 namespace Kernel
 {

+	class Device;
+
 	class DevFileSystem final : public TmpFileSystem
 	{
 	public:
@@ -2,6 +2,7 @@

 #include <kernel/FS/Inode.h>
 #include <kernel/Lock/Mutex.h>
+#include <kernel/ThreadBlocker.h>

 namespace Kernel
 {
@@ -22,16 +23,16 @@ namespace Kernel
 		BAN::ErrorOr<size_t> read_impl(off_t, BAN::ByteSpan) override;
 		BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override;

-		bool can_read_impl() const override { return m_value > 0; }
-		bool can_write_impl() const override { return m_value < UINT64_MAX - 1; }
+		bool can_read_impl() const override;
+		bool can_write_impl() const override;
 		bool has_error_impl() const override { return false; }
 		bool has_hungup_impl() const override { return false; }

 	private:
 		const bool m_is_semaphore;
-		BAN::Atomic<uint64_t> m_value;
+		uint64_t m_value;

-		Mutex m_mutex;
+		mutable Mutex m_mutex;
 		ThreadBlocker m_thread_blocker;
 	};

@@ -2,8 +2,9 @@

 #include <BAN/HashMap.h>
 #include <kernel/Device/Device.h>
-#include <kernel/FS/FileSystem.h>
 #include <kernel/FS/Ext2/Inode.h>
+#include <kernel/FS/FileSystem.h>
+#include <kernel/Memory/VirtualRange.h>

 namespace Kernel
 {
@@ -5,6 +5,7 @@
 #include <kernel/FS/FAT/Definitions.h>
 #include <kernel/FS/FAT/Inode.h>
 #include <kernel/FS/FileSystem.h>
+#include <kernel/Lock/Mutex.h>

 namespace Kernel
 {
@@ -1,11 +1,12 @@
 #pragma once

-#include <kernel/Device/Device.h>
 #include <kernel/FS/Inode.h>

 namespace Kernel
 {

+	class BlockDevice;
+
 	class FileSystem : public BAN::RefCounted<FileSystem>
 	{
 	public:
@@ -2,11 +2,12 @@

 #include <kernel/FS/TmpFS/FileSystem.h>
 #include <kernel/FS/TmpFS/Inode.h>
-#include <kernel/Process.h>

 namespace Kernel
 {

+	class Process;
+
 	class ProcFileSystem final : public TmpFileSystem
 	{
 	public:
@@ -2,11 +2,12 @@

 #include <kernel/FS/TmpFS/FileSystem.h>
 #include <kernel/FS/TmpFS/Inode.h>
-#include <kernel/Process.h>

 namespace Kernel
 {

+	class Process;
+
 	class ProcPidInode final : public TmpDirectoryInode
 	{
 		// FIXME: dynamically update ruid/rgid.
@@ -10,21 +10,6 @@
 namespace Kernel
 {

-	// IDT entries
-	//   0x00->0x1F (32):  ISR
-	//   0x20->0x7F (96):  PIC/IOAPIC
-	//   0x80->0xEF (112): MSI
-	//   0xF0->0xFE (15):  internal
-
-	constexpr uint8_t IRQ_VECTOR_BASE   = 0x20;
-	constexpr uint8_t IRQ_MSI_BASE      = 0x80;
-	constexpr uint8_t IRQ_MSI_END       = 0xF0;
-#if ARCH(i686)
-	constexpr uint8_t IRQ_SYSCALL       = 0xF0; // hard coded in kernel/API/Syscall.h
-#endif
-	constexpr uint8_t IRQ_IPI           = 0xF1;
-	constexpr uint8_t IRQ_TIMER         = 0xF2;
-
 #if ARCH(x86_64)
 	struct GateDescriptor
 	{
@@ -36,6 +36,7 @@ namespace Kernel::Input
 		uint8_t m_byte_index { 0 };
 		bool m_basic { false };

+		bool m_scancode_set_uncertain { false };
 		uint8_t m_scancode_set { 0xFF };
 		uint16_t m_modifiers { 0 };

@@ -0,0 +1,23 @@
+#pragma once
+
+#include <kernel/Arch.h>
+
+namespace Kernel
+{
+
+	// IDT entries
+	//   0x00->0x1F (32):  ISR
+	//   0x20->0x7F (96):  PIC/IOAPIC
+	//   0x80->0xEF (112): MSI
+	//   0xF0->0xFE (15):  internal
+
+	constexpr uint8_t IRQ_VECTOR_BASE   = 0x20;
+	constexpr uint8_t IRQ_MSI_BASE      = 0x80;
+	constexpr uint8_t IRQ_MSI_END       = 0xF0;
+#if ARCH(i686)
+	constexpr uint8_t IRQ_SYSCALL       = 0xF0; // hard coded in kernel/API/Syscall.h
+#endif
+	constexpr uint8_t IRQ_IPI           = 0xF1;
+	constexpr uint8_t IRQ_TIMER         = 0xF2;
+
+}
@@ -2,7 +2,6 @@

 #include <BAN/Atomic.h>
 #include <BAN/NoCopyMove.h>
-#include <kernel/Thread.h>

 #include <sys/types.h>

@@ -27,63 +26,14 @@ namespace Kernel
 	public:
 		Mutex() = default;

-		void lock() override
-		{
-			const auto tid = Thread::current_tid();
-			if (tid == m_locker)
-				ASSERT(m_lock_depth > 0);
-			else
-			{
-				ASSERT(!tid || !Thread::current().has_spinlock());
-				pid_t expected = -1;
-				while (!m_locker.compare_exchange(expected, tid))
-				{
-					ASSERT(Processor::get_interrupt_state() == InterruptState::Enabled);
-					Processor::yield();
-					expected = -1;
-				}
-				ASSERT(m_lock_depth == 0);
-				if (tid)
-					Thread::current().add_mutex();
-			}
-			m_lock_depth++;
-		}
-
-		bool try_lock()
-		{
-			const auto tid = Thread::current_tid();
-			if (tid == m_locker)
-				ASSERT(m_lock_depth > 0);
-			else
-			{
-				pid_t expected = -1;
-				if (!m_locker.compare_exchange(expected, tid))
-					return false;
-				ASSERT(m_lock_depth == 0);
-				if (tid)
-					Thread::current().add_mutex();
-			}
-			m_lock_depth++;
-			return true;
-		}
-
-		void unlock() override
-		{
-			const auto tid = Thread::current_tid();
-			ASSERT(m_locker == tid);
-			ASSERT(m_lock_depth > 0);
-			if (--m_lock_depth == 0)
-			{
-				m_locker = -1;
-				if (tid)
-					Thread::current().remove_mutex();
-			}
-		}
+		bool try_lock();
+		void lock() override;
+		void unlock() override;

 		pid_t locker() const { return m_locker; }
 		bool is_locked() const { return m_locker != -1; }
 		uint32_t lock_depth() const override { return m_lock_depth; }
-		bool is_locked_by_current_thread() const override { return m_locker == Thread::current_tid(); }
+		bool is_locked_by_current_thread() const override;

 	private:
 		BAN::Atomic<pid_t>	m_locker		{ -1 };
@@ -98,74 +48,14 @@ namespace Kernel
 	public:
 		PriorityMutex() = default;

-		void lock() override
-		{
-			const auto tid = Thread::current_tid();
-
-			if (tid == m_locker)
-				ASSERT(m_lock_depth > 0);
-			else
-			{
-				ASSERT(!tid || !Thread::current().has_spinlock());
-				bool has_priority = tid ? !Thread::current().is_userspace() : true;
-				if (has_priority)
-					m_queue_length++;
-				pid_t expected = -1;
-				while (!(has_priority || m_queue_length == 0) || !m_locker.compare_exchange(expected, tid))
-				{
-					ASSERT(Processor::get_interrupt_state() == InterruptState::Enabled);
-					Processor::yield();
-					expected = -1;
-				}
-				ASSERT(m_lock_depth == 0);
-				if (tid)
-					Thread::current().add_mutex();
-			}
-			m_lock_depth++;
-		}
-
-		bool try_lock()
-		{
-			const auto tid = Thread::current_tid();
-
-			if (tid == m_locker)
-				ASSERT(m_lock_depth > 0);
-			else
-			{
-				bool has_priority = tid ? !Thread::current().is_userspace() : true;
-				pid_t expected = -1;
-				if (!(has_priority || m_queue_length == 0) || !m_locker.compare_exchange(expected, tid))
-					return false;
-				if (has_priority)
-					m_queue_length++;
-				ASSERT(m_lock_depth == 0);
-				if (tid)
-					Thread::current().add_mutex();
-			}
-			m_lock_depth++;
-			return true;
-		}
-
-		void unlock() override
-		{
-			const auto tid = Thread::current_tid();
-			ASSERT(m_locker == tid);
-			ASSERT(m_lock_depth > 0);
-			if (--m_lock_depth == 0)
-			{
-				bool has_priority = tid ? !Thread::current().is_userspace() : true;
-				if (has_priority)
-					m_queue_length--;
-				m_locker = -1;
-				if (tid)
-					Thread::current().remove_mutex();
-			}
-		}
+		bool try_lock();
+		void lock() override;
+		void unlock() override;

 		pid_t locker() const { return m_locker; }
 		bool is_locked() const { return m_locker != -1; }
 		uint32_t lock_depth() const override { return m_lock_depth; }
-		bool is_locked_by_current_thread() const override { return m_locker == Thread::current_tid(); }
+		bool is_locked_by_current_thread() const override;

 	private:
 		BAN::Atomic<pid_t>		m_locker		{ -1 };
@@ -1,7 +1,7 @@
 #pragma once

-#include <kernel/Lock/BlockableSpinLock.h>
 #include <kernel/Lock/SpinLock.h>
+#include <kernel/ThreadBlocker.h>

 namespace Kernel
 {
@@ -13,54 +13,10 @@ namespace Kernel
 	public:
 		RWLock() = default;

-		void rd_lock()
-		{
-			SpinLockGuard _(m_lock);
-			while (m_writers_waiting > 0 || m_writer != -1)
-			{
-				BlockableSpinLock block(m_lock);
-				m_thread_blocker.block_indefinite(&block);
-			}
-			m_readers_active++;
-		}
-
-		void rd_unlock()
-		{
-			SpinLockGuard _(m_lock);
-			if (--m_readers_active == 0)
-				m_thread_blocker.unblock();
-		}
-
-		void wr_lock()
-		{
-			if (m_writer == Thread::current_tid())
-			{
-				m_writer_depth++;
-				return;
-			}
-
-			SpinLockGuard _(m_lock);
-
-			m_writers_waiting++;
-			while (m_readers_active > 0 || m_writer != -1)
-			{
-				BlockableSpinLock block(m_lock);
-				m_thread_blocker.block_indefinite(&block);
-			}
-			m_writers_waiting--;
-
-			m_writer = Thread::current_tid();
-			m_writer_depth = 1;
-		}
-
-		void wr_unlock()
-		{
-			if (--m_writer_depth != 0)
-				return;
-			SpinLockGuard _(m_lock);
-			m_writer = -1;
-			m_thread_blocker.unblock();
-		}
+		void rd_lock();
+		void rd_unlock();
+		void wr_lock();
+		void wr_unlock();

 	private:
 		SpinLock m_lock;
@@ -1,5 +1,6 @@
 #pragma once

+#include <BAN/Vector.h>
 #include <kernel/Memory/MemoryRegion.h>

 namespace Kernel
@@ -1,7 +1,9 @@
 #pragma once

 #include <BAN/HashMap.h>
+#include <BAN/RefPtr.h>
 #include <BAN/UniqPtr.h>
+#include <BAN/Vector.h>
 #include <kernel/Lock/Mutex.h>
 #include <kernel/Lock/SpinLock.h>
 #include <kernel/Memory/MemoryRegion.h>
@@ -27,8 +29,9 @@ namespace Kernel
 	private:
 		struct Object : public BAN::RefCounted<Object>
 		{
-			Object(key_t key, shmid_ds info)
+			Object(key_t key, int id, shmid_ds info)
 				: key(key)
+				, id(id)
 				, info(info)
 			{ }
 			~Object();
@@ -36,6 +39,7 @@ namespace Kernel
 			bool can_current_process_access(int flags) const;

 			const key_t key;
+			const int id;
 			shmid_ds info;

 			Mutex mutex;
@@ -4,8 +4,6 @@
 #include <BAN/HashMap.h>
 #include <BAN/UniqPtr.h>
 #include <kernel/Networking/NetworkInterface.h>
-#include <kernel/Thread.h>
-#include <kernel/ThreadBlocker.h>

 namespace Kernel
 {
@@ -11,7 +11,6 @@
 #include <kernel/Networking/NetworkInterface.h>
 #include <kernel/Networking/NetworkLayer.h>
 #include <kernel/Networking/NetworkSocket.h>
-#include <kernel/Thread.h>

 namespace Kernel
 {
@@ -1,6 +1,7 @@
 #pragma once

 #include <kernel/Networking/NetworkInterface.h>
+#include <kernel/Memory/VirtualRange.h>

 namespace Kernel
 {
@@ -7,7 +7,6 @@
 #include <kernel/Memory/ByteRingBuffer.h>
 #include <kernel/Networking/NetworkInterface.h>
 #include <kernel/Networking/NetworkSocket.h>
-#include <kernel/Thread.h>
 #include <kernel/ThreadBlocker.h>

 namespace Kernel
@@ -6,6 +6,7 @@
 #include <kernel/FS/Socket.h>
 #include <kernel/FS/TmpFS/Inode.h>
 #include <kernel/FS/VirtualFileSystem.h>
+#include <kernel/Memory/VirtualRange.h>
 #include <kernel/OpenFileDescriptorSet.h>

 namespace Kernel
@@ -3,6 +3,8 @@
 #include <BAN/UniqPtr.h>
 #include <BAN/Vector.h>
 #include <kernel/ACPI/AML/Node.h>
+#include <kernel/InterruptNumbers.h>
+#include <kernel/Interruptable.h>
 #include <kernel/Memory/Types.h>

 #include <sys/types.h>
@@ -19,6 +19,7 @@

 #include <poll.h>
 #include <sys/banan-os.h>
+#include <sys/epoll.h>
 #include <sys/mman.h>
 #include <sys/select.h>
 #include <sys/socket.h>
@@ -26,8 +27,6 @@
 #include <sys/time.h>
 #include <termios.h>

-struct epoll_event;
-
 namespace Kernel
 {

@@ -1,17 +1,14 @@
 #pragma once

+#include <BAN/Array.h>
 #include <BAN/Atomic.h>
-#include <BAN/ForwardList.h>
+#include <BAN/NoCopyMove.h>
 #include <BAN/Math.h>

 #include <kernel/API/SharedPage.h>
 #include <kernel/Arch.h>
-#include <kernel/GDT.h>
-#include <kernel/IDT.h>
-#include <kernel/InterruptStack.h>
 #include <kernel/Memory/Types.h>
 #include <kernel/ProcessorID.h>
-#include <kernel/Scheduler.h>

 namespace Kernel
 {
@@ -22,6 +19,12 @@ namespace Kernel
 		Enabled,
 	};

+	class GDT;
+	class IDT;
+	class Scheduler;
+	class SchedulerThreadNode;
+	class Thread;
+
 #if ARCH(x86_64) || ARCH(i686)
 	class Processor
 	{
@@ -51,8 +54,8 @@ namespace Kernel
 			union
 			{
 				TLBEntry flush_tlb;
-				SchedulerQueue::Node* new_thread;
-				SchedulerQueue::Node* unblock_thread;
+				SchedulerThreadNode* new_thread;
+				SchedulerThreadNode* unblock_thread;
 				bool dummy;
 			};
 		};
@@ -179,16 +182,22 @@ namespace Kernel
 		}

 		template<typename T>
-		static T read_gs_sized(uintptr_t offset) requires(sizeof(T) <= 8 && BAN::Math::is_power_of_two(sizeof(T)))
+		static T read_gs_sized(uintptr_t offset) requires(sizeof(T) <= sizeof(uintptr_t) && BAN::Math::is_power_of_two(sizeof(T)))
 		{
 			T value;
-			asm volatile("mov %%gs:%a[offset], %[value]" : [value]"=r"(value) : [offset]"ir"(offset));
+			if constexpr (sizeof(T) == 1 && ARCH(i686))
+				asm volatile("mov %%gs:%a[offset], %[value]" : [value]"=q"(value) : [offset]"ir"(offset) : "memory");
+			else
+				asm volatile("mov %%gs:%a[offset], %[value]" : [value]"=r"(value) : [offset]"ir"(offset) : "memory");
 			return value;
 		}

 		template<typename T>
-		static void write_gs_sized(uintptr_t offset, T value) requires(sizeof(T) <= 8 && BAN::Math::is_power_of_two(sizeof(T)))
+		static void write_gs_sized(uintptr_t offset, T value) requires(sizeof(T) <= sizeof(uintptr_t) && BAN::Math::is_power_of_two(sizeof(T)))
 		{
+			if constexpr (sizeof(T) == 1 && ARCH(i686))
+				asm volatile("mov %[value], %%gs:%a[offset]" :: [value]"q"(value), [offset]"ir"(offset) : "memory");
+			else
 				asm volatile("mov %[value], %%gs:%a[offset]" :: [value]"r"(value), [offset]"ir"(offset) : "memory");
 		}

@@ -5,6 +5,7 @@
 #include <BAN/NoCopyMove.h>
 #include <kernel/InterruptStack.h>
 #include <kernel/ProcessorID.h>
+#include <kernel/SchedulerThreadNode.h>

 #include <sys/types.h>

@@ -14,29 +15,6 @@ namespace Kernel
 	class BaseMutex;
 	class Thread;
 	class ThreadBlocker;
-	struct SchedulerQueueNode;
-
-	class SchedulerQueue
-	{
-	public:
-		using Node = SchedulerQueueNode;
-
-	public:
-		void add_thread_to_back(Node*);
-		bool add_thread_with_wake_time(Node*); // return true if node was inserted as the first element
-		template<typename F>
-		Node* remove_with_condition(F callback);
-		void remove_node(Node*);
-		Node* front();
-		Node* pop_front();
-
-		bool empty() const { return m_head == nullptr; }
-
-	private:
-		Node* m_head { nullptr };
-		Node* m_tail { nullptr };
-	};
-
 	class Scheduler
 	{
 		BAN_NON_COPYABLE(Scheduler);
@@ -45,12 +23,12 @@ namespace Kernel
 	public:
 		struct NewThreadRequest
 		{
-			SchedulerQueue::Node* node;
+			SchedulerThreadNode* node;
 		};

 		struct UnblockRequest
 		{
-			SchedulerQueue::Node* node;
+			SchedulerThreadNode* node;
 		};

 	public:
@@ -79,9 +57,9 @@ namespace Kernel
 	private:
 		Scheduler() = default;

-		void add_current_to_most_loaded(SchedulerQueue* target_queue);
-		void update_most_loaded_node_queue(SchedulerQueue::Node*, SchedulerQueue* target_queue);
-		void remove_node_from_most_loaded(SchedulerQueue::Node*);
+		void add_current_to_most_loaded(void* target_list);
+		void update_most_loaded_node_list(SchedulerThreadNode*, void* target_list);
+		void remove_node_from_most_loaded(SchedulerThreadNode*);

 		void update_wake_up_deadline();
 		void wake_up_sleeping_threads();
@@ -90,13 +68,13 @@ namespace Kernel

 		class ProcessorID find_least_loaded_processor() const;

-		void add_thread(SchedulerQueue::Node*);
-		void unblock_thread(SchedulerQueue::Node*);
+		void add_thread(SchedulerThreadNode*);
+		void unblock_thread(SchedulerThreadNode*);

 	private:
-		SchedulerQueue m_run_queue;
-		SchedulerQueue m_block_queue;
-		SchedulerQueue::Node* m_current { nullptr };
+		SchedulerQueue m_run_list;
+		SchedulerHeap m_block_list;
+		SchedulerThreadNode* m_current { nullptr };

 		uint32_t m_thread_count { 0 };

@@ -108,8 +86,8 @@ namespace Kernel

 		struct ThreadInfo
 		{
-			SchedulerQueue*       queue { nullptr };
-			SchedulerQueue::Node* node  { nullptr };
+			void*                list { nullptr };
+			SchedulerThreadNode* node { nullptr };
 		};
 		BAN::Array<ThreadInfo, 10> m_most_loaded_threads;

@@ -1,36 +0,0 @@
-#pragma once
-
-#include <kernel/ProcessorID.h>
-#include <kernel/Lock/SpinLock.h>
-
-namespace Kernel
-{
-
-	class Thread;
-	class ThreadBlocker;
-
-	struct SchedulerQueueNode
-	{
-		SchedulerQueueNode(Thread* thread)
-			: thread(thread)
-		{}
-
-		Thread* const thread;
-
-		SchedulerQueueNode* next { nullptr };
-		SchedulerQueueNode* prev { nullptr };
-
-		uint64_t wake_time_ns { static_cast<uint64_t>(-1) };
-
-		BAN::Atomic<ThreadBlocker*> blocker { nullptr };
-		SchedulerQueueNode* block_chain_prev { nullptr };
-		SchedulerQueueNode* block_chain_next { nullptr };
-
-		ProcessorID processor_id { PROCESSOR_NONE };
-		bool blocked { false };
-
-		uint64_t last_start_ns { 0 };
-		uint64_t time_used_ns  { 0 };
-	};
-
-}
@@ -0,0 +1,96 @@
+#pragma once
+
+#include <BAN/Atomic.h>
+#include <BAN/NoCopyMove.h>
+#include <kernel/ProcessorID.h>
+
+namespace Kernel
+{
+
+	class Thread;
+	class ThreadBlocker;
+
+	struct SchedulerThreadNode
+	{
+		SchedulerThreadNode(Thread* thread)
+			: thread(thread)
+			, heap({ nullptr, nullptr, nullptr })
+		{}
+
+		Thread* const thread;
+
+		union
+		{
+			struct
+			{
+				SchedulerThreadNode* next;
+				SchedulerThreadNode* prev;
+			} queue;
+			struct
+			{
+				SchedulerThreadNode* parent;
+				SchedulerThreadNode* lchild;
+				SchedulerThreadNode* rchild;
+			} heap;
+		};
+
+		uint64_t wake_time_ns { static_cast<uint64_t>(-1) };
+
+		BAN::Atomic<ThreadBlocker*> blocker { nullptr };
+		SchedulerThreadNode* block_chain_prev { nullptr };
+		SchedulerThreadNode* block_chain_next { nullptr };
+
+		ProcessorID processor_id { PROCESSOR_NONE };
+		bool blocked { false };
+
+		uint64_t last_start_ns { 0 };
+		uint64_t time_used_ns  { 0 };
+	};
+
+	class SchedulerQueue
+	{
+		BAN_NON_COPYABLE(SchedulerQueue);
+		BAN_NON_MOVABLE(SchedulerQueue);
+	public:
+		SchedulerQueue() = default;
+
+		SchedulerThreadNode* front();
+		SchedulerThreadNode* pop_front();
+
+		void push(SchedulerThreadNode*);
+		void pop(SchedulerThreadNode*);
+
+		void walk(void (*)(const SchedulerThreadNode*, void*), void*) const;
+		bool empty() const { return m_head == nullptr; }
+
+	private:
+		SchedulerThreadNode* m_head { nullptr };
+		SchedulerThreadNode* m_tail { nullptr };
+	};
+
+	class SchedulerHeap
+	{
+		BAN_NON_COPYABLE(SchedulerHeap);
+		BAN_NON_MOVABLE(SchedulerHeap);
+	public:
+		SchedulerHeap() = default;
+
+		SchedulerThreadNode* front();
+		SchedulerThreadNode* pop_front();
+
+		void push(SchedulerThreadNode*);
+		void pop(SchedulerThreadNode*);
+
+		void walk(void (*)(const SchedulerThreadNode*, void*), void*) const;
+		bool empty() const { return m_root == nullptr; }
+
+	private:
+		void walk_impl(void (*)(const SchedulerThreadNode*, void*), void*, const SchedulerThreadNode*) const;
+		void swap_nodes(SchedulerThreadNode*, SchedulerThreadNode*);
+
+	private:
+		SchedulerThreadNode* m_root { nullptr };
+		SchedulerThreadNode* m_last { nullptr };
+	};
+
+}
@@ -1,6 +1,7 @@
 #pragma once

 #include <BAN/WeakPtr.h>
+#include <kernel/Memory/VirtualRange.h>
 #include <kernel/Terminal/TTY.h>

 namespace Kernel
@@ -96,8 +96,6 @@ namespace Kernel
 		uint32_t m_last_cursor_row { static_cast<uint32_t>(-1) };
 		uint32_t m_last_cursor_column { static_cast<uint32_t>(-1) };

-		const Palette& m_palette;
-
 		TerminalDriver::Color m_foreground;
 		TerminalDriver::Color m_background;
 		bool m_colors_inverted { false };
@@ -5,8 +5,8 @@
 #include <BAN/RefPtr.h>
 #include <BAN/UniqPtr.h>
 #include <kernel/InterruptStack.h>
+#include <kernel/Lock/Mutex.h>
 #include <kernel/Memory/VirtualRange.h>
-#include <kernel/ThreadBlocker.h>

 #include <LibELF/AuxiliaryVector.h>

@@ -18,6 +18,7 @@ namespace Kernel

 	class MemoryBackedRegion;
 	class Process;
+	class ThreadBlocker;

 	class Thread
 	{
@@ -124,10 +125,14 @@ namespace Kernel

 		bool is_userspace() const { return m_is_userspace; }

-		uint64_t cpu_time_ns() const;
+		uint64_t cpu_time_total_ns() const;
+		void cpu_time_ns(uint64_t& user_ns, uint64_t& system_ns) const;
 		void set_cpu_time_start();
 		void set_cpu_time_stop();

+		void set_is_in_syscall(bool is_in_syscall);
+		bool is_in_syscall() const { return m_is_in_syscall; }
+
 		void update_processor_index_address();

 		void set_fsbase(vaddr_t base) { m_fsbase = base; }
@@ -187,7 +192,7 @@ namespace Kernel
 		vaddr_t                    m_fsbase               { 0 };
 		vaddr_t                    m_gsbase               { 0 };

-		SchedulerQueue::Node*      m_scheduler_node       { nullptr };
+		SchedulerThreadNode*       m_scheduler_node       { nullptr };

 		YieldRegisters             m_yield_registers      { };

@@ -200,8 +205,10 @@ namespace Kernel
 		static_assert(_SIGMAX < 64);

 		mutable SpinLock           m_cpu_time_lock;
-		uint64_t                   m_cpu_time_ns          { 0 };
+		uint64_t                   m_cpu_time_user_ns     { 0 };
+		uint64_t                   m_cpu_time_system_ns   { 0 };
 		uint64_t                   m_cpu_time_start_ns    { UINT64_MAX };
+		BAN::Atomic<bool>          m_is_in_syscall        { false };

 		BAN::Atomic<uint32_t>      m_spinlock_count       { 0 };
 		BAN::Atomic<uint32_t>      m_mutex_count          { 0 };
@@ -1,12 +1,14 @@
 #pragma once

 #include <BAN/Math.h>
+#include <kernel/Lock/Mutex.h>
 #include <kernel/Lock/SpinLock.h>
-#include <kernel/Scheduler.h>

 namespace Kernel
 {

+	class SchedulerThreadNode;
+
 	class ThreadBlocker
 	{
 	public:
@@ -27,11 +29,11 @@ namespace Kernel
 		}

 	private:
-		void add_thread_to_block_queue(SchedulerQueue::Node*);
-		void remove_thread_from_block_queue(SchedulerQueue::Node*);
+		void add_thread_to_block_queue(SchedulerThreadNode*);
+		void remove_thread_from_block_queue(SchedulerThreadNode*);

 	private:
-		SchedulerQueue::Node* m_block_chain { nullptr };
+		SchedulerThreadNode* m_block_chain { nullptr };
 		SpinLock m_lock;

 		friend class Scheduler;
@@ -23,6 +23,8 @@ namespace Kernel

 		virtual void handle_stall(uint8_t endpoint_id) = 0;
 		virtual void handle_input_data(size_t byte_count, uint8_t endpoint_id) = 0;
+
+		virtual bool is_hid_driver() const { return false; }
 	};

 	class USBDevice
@@ -62,6 +64,12 @@ namespace Kernel
 			uint8_t tt_think_time;
 		};

+		struct HIDInfo
+		{
+			BAN::Atomic<uint32_t> led_mask { 0 };
+			BAN::Vector<USBClassDriver*> led_controls;
+		};
+
 	public:
 		USBDevice(USBController& controller, USB::SpeedClass speed_class, uint8_t depth)
 			: m_controller(controller)
@@ -92,6 +100,8 @@ namespace Kernel
 		void register_hub_to_init() { m_controller.register_hub_to_init(m_depth + 1); };
 		void mark_hub_init_done() { m_controller.mark_hub_init_done(m_depth + 1); };

+		void update_led_mask(uint32_t led_mask);
+
 	protected:
 		void handle_stall(uint8_t endpoint_id);
 		void handle_input_data(size_t byte_count, uint8_t endpoint_id);
@@ -112,6 +122,7 @@ namespace Kernel
 		BAN::UniqPtr<DMARegion> m_dma_buffer;

 		BAN::Vector<BAN::UniqPtr<USBClassDriver>> m_class_drivers;
+		HIDInfo m_hid_info;
 	};

 }
@@ -74,6 +74,7 @@ namespace Kernel
 		struct DeviceReport
 		{
 			BAN::Vector<USBHID::Report> inputs;
+			BAN::Vector<USBHID::Report> outputs;
 			BAN::RefPtr<USBHIDDevice> device;
 		};

@@ -81,27 +82,43 @@ namespace Kernel
 		void handle_stall(uint8_t endpoint_id) override;
 		void handle_input_data(size_t byte_count, uint8_t endpoint_id) override;

+		bool is_hid_driver() const override { return true; }
+
 		USBDevice& device() { return m_device; }
 		const USBDevice::InterfaceDescriptor& interface() const { return m_interface; }

+		bool has_led_control() const { return !m_led_controls.empty(); }
+		void set_leds(uint32_t led_mask);
+
 	private:
 		USBHIDDriver(USBDevice&, const USBDevice::InterfaceDescriptor&);
 		~USBHIDDriver();

 		BAN::ErrorOr<void> initialize() override;

-		BAN::ErrorOr<BAN::Vector<DeviceReport>> initializes_device_reports(const BAN::Vector<USBHID::Collection>&);
+		BAN::ErrorOr<void> initializes_device_reports(const BAN::Vector<USBHID::Collection>&);
+
+	private:
+		struct LEDControl
+		{
+			DeviceReport* report;
+			uint32_t report_id;
+			uint32_t report_bits;
+		};

 	private:
 		USBDevice& m_device;
 		USBDevice::InterfaceDescriptor m_interface;

 		bool m_uses_report_id { false };
-		BAN::Vector<DeviceReport> m_device_inputs;
+		BAN::Vector<DeviceReport> m_device_reports;

 		uint8_t m_data_endpoint_id = 0;
 		BAN::UniqPtr<DMARegion> m_data_buffer;

+		BAN::Vector<LEDControl> m_led_controls;
+		BAN::UniqPtr<DMARegion> m_led_region;
+
 		friend class BAN::UniqPtr<USBHIDDriver>;
 	};

@@ -11,8 +11,6 @@ namespace Kernel
 		BAN_NON_MOVABLE(USBKeyboard);

 	public:
-		BAN::ErrorOr<void> initialize() override;
-
 		void start_report() override;
 		void stop_report() override;

@@ -23,12 +21,9 @@ namespace Kernel
 		void update() override;

 	private:
-		USBKeyboard(USBHIDDriver& driver, BAN::Vector<USBHID::Report>&& outputs);
+		USBKeyboard(USBHIDDriver& driver);
 		~USBKeyboard() = default;

-		void set_leds(uint16_t mask);
-		void set_leds(uint8_t report_id, uint16_t mask);
-
 	private:
 		USBHIDDriver& m_driver;

@@ -38,11 +33,7 @@ namespace Kernel
 		BAN::Array<bool, 0x100> m_keyboard_state      { false };
 		BAN::Array<bool, 0x100> m_keyboard_state_temp { false };
 		uint16_t m_toggle_mask { 0 };
-
 		uint16_t m_led_mask { 0 };
-		BAN::UniqPtr<DMARegion> m_led_region;
-
-		BAN::Vector<USBHID::Report> m_outputs;

 		BAN::Optional<uint8_t> m_repeat_scancode;
 		uint8_t m_repeat_modifier { 0 };
@@ -1,6 +1,5 @@
 #pragma once

-#include <kernel/Process.h>
 #include <kernel/USB/Device.h>

 namespace Kernel
@@ -373,6 +373,33 @@ namespace Kernel::XHCI
 				uint32_t slot_id               : 8;
 			} configure_endpoint_command;

+			struct
+			{
+				uint32_t             : 32;
+				uint32_t             : 32;
+				uint32_t             : 32;
+				uint32_t cycle_bit   : 1;
+				uint32_t             : 8;
+				uint32_t tsp         : 1;
+				uint32_t trb_type    : 6;
+				uint32_t endpoint_id : 5;
+				uint32_t             : 3;
+				uint32_t slot_id     : 8;
+			} reset_endpoint_command;
+
+			struct
+			{
+				uint64_t new_tr_deque_pointer : 64;
+				uint32_t                      : 16;
+				uint32_t stream_id            : 16;
+				uint32_t cycle_bit            : 1;
+				uint32_t                      : 9;
+				uint32_t trb_type             : 6;
+				uint32_t endpoint_id          : 5;
+				uint32_t                      : 3;
+				uint32_t slot_id              : 8;
+			} set_tr_deque_pointer_command;
+
 			struct
 			{
 				uint64_t ring_segment_ponter     : 64;
@@ -8,6 +8,7 @@
 #include <kernel/IO.h>
 #include <kernel/Memory/PageTable.h>
 #include <kernel/Process.h>
+#include <kernel/Scheduler.h>
 #include <kernel/Timer/Timer.h>

 #define RSPD_SIZE	20
@@ -789,7 +790,7 @@ acpi_release_global_lock:
 			auto [gpe_path, gpe_obj] = TRY(m_namespace->find_named_object(embedded_controller, TRY(AML::NameString::from_string("_GPE"_sv)), true));
 			if (gpe_obj == nullptr)
 			{
-				dwarnln("EC {} does have _GPE", embedded_controller);
+				dwarnln("EC {} does not have _GPE", embedded_controller);
 				break;
 			}

@@ -806,7 +807,7 @@ acpi_release_global_lock:
 		auto [crs_path, crs_obj] = TRY(m_namespace->find_named_object(embedded_controller, TRY(AML::NameString::from_string("_CRS"_sv)), true));
 		if (crs_obj == nullptr)
 		{
-			dwarnln("EC {} does have _CRS", embedded_controller);
+			dwarnln("EC {} does not have _CRS", embedded_controller);
 			return BAN::Error::from_errno(ENOENT);
 		}

@@ -1198,6 +1199,8 @@ acpi_release_global_lock:
 			continue;

 handle_event:
+			IO::outw(sts_port, pending);
+
 			if (pending & PM1_EVN_PWRBTN)
 			{
 				dprintln("Power button pressed");
@@ -1208,8 +1211,6 @@ handle_event:
 			{
 				dwarnln("Unhandled ACPI fixed event {H}", pending);
 			}
-
-			IO::outw(sts_port, pending);
 		}
 	}

@@ -1,5 +1,6 @@
 #include <kernel/ACPI/BatterySystem.h>
-#include <kernel/FS/DevFS//FileSystem.h>
+#include <kernel/Device/Device.h>
+#include <kernel/FS/DevFS/FileSystem.h>
 #include <kernel/Timer/Timer.h>

 namespace Kernel::ACPI
@@ -35,34 +36,11 @@ namespace Kernel::ACPI
 			if (offset < 0)
 				return BAN::Error::from_errno(EINVAL);

-			if (SystemTimer::get().ms_since_boot() > m_last_read_ms + 1000)
-			{
-				auto [method_path, method_ref] = TRY(m_acpi_namespace.find_named_object(m_battery_path, m_method_name));
-				if (method_ref == nullptr)
-					return BAN::Error::from_errno(EFAULT);
-
-				auto result = TRY(AML::method_call(method_path, method_ref->node, BAN::Array<AML::Reference*, 7>{}));
-				if (result.type != AML::Node::Type::Package || result.as.package->num_elements < m_result_index)
-					return BAN::Error::from_errno(EFAULT);
-
-				auto& target_elem = result.as.package->elements[m_result_index];
-				if (!target_elem.resolved || !target_elem.value.node)
-					return BAN::Error::from_errno(EFAULT);
-
-				auto target_conv = AML::convert_node(TRY(target_elem.value.node->copy()), AML::ConvInteger, sizeof(uint64_t));
-				if (target_conv.is_error())
-					return BAN::Error::from_errno(EFAULT);
-
-				m_last_read_ms = SystemTimer::get().ms_since_boot();
-				m_last_value = target_conv.value().as.integer.value;
-			}
-
-			auto target_str = TRY(BAN::String::formatted("{}", m_last_value.load()));
-
+			auto target_str = TRY(BAN::String::formatted("{}", TRY(get_value())));
 			if (static_cast<size_t>(offset) >= target_str.size())
 				return 0;

-			const size_t ncopy = BAN::Math::min(buffer.size(), target_str.size() - offset);
+			const size_t ncopy = BAN::Math::min<size_t>(buffer.size(), target_str.size() - offset);
 			memcpy(buffer.data(), target_str.data() + offset, ncopy);
 			return ncopy;
 		}
@@ -84,14 +62,44 @@ namespace Kernel::ACPI
 			, m_result_index(index)
 		{ }

+		BAN::ErrorOr<uint64_t> get_value()
+		{
+			LockGuard _(m_mutex);
+
+			if (SystemTimer::get().ms_since_boot() < m_last_read_ms + 1000)
+				return m_last_value;
+
+			auto [method_path, method_ref] = TRY(m_acpi_namespace.find_named_object(m_battery_path, m_method_name));
+			if (method_ref == nullptr)
+				return BAN::Error::from_errno(EFAULT);
+
+			auto result = TRY(AML::method_call(method_path, method_ref->node, BAN::Array<AML::Reference*, 7>{}));
+			if (result.type != AML::Node::Type::Package || result.as.package->num_elements < m_result_index)
+				return BAN::Error::from_errno(EFAULT);
+
+			auto& target_elem = result.as.package->elements[m_result_index];
+			if (!target_elem.resolved || !target_elem.value.node)
+				return BAN::Error::from_errno(EFAULT);
+
+			auto target_conv = AML::convert_node(TRY(target_elem.value.node->copy()), AML::ConvInteger, sizeof(uint64_t));
+			if (target_conv.is_error())
+				return BAN::Error::from_errno(EFAULT);
+
+			m_last_read_ms = SystemTimer::get().ms_since_boot();
+			m_last_value = target_conv.value().as.integer.value;
+
+			return m_last_value;
+		}
+
 	private:
 		AML::Namespace& m_acpi_namespace;
 		AML::Scope m_battery_path;
 		AML::NameString m_method_name;
 		size_t m_result_index;

-		BAN::Atomic<uint64_t> m_last_read_ms = 0;
-		BAN::Atomic<uint64_t> m_last_value = 0;
+		Mutex m_mutex;
+		uint64_t m_last_read_ms = 0;
+		uint64_t m_last_value = 0;
 	};

 	BAN::ErrorOr<void> BatterySystem::initialize(AML::Namespace& acpi_namespace)
@@ -4,6 +4,8 @@
 #include <kernel/ACPI/EmbeddedController.h>
 #include <kernel/IO.h>
 #include <kernel/Lock/LockGuard.h>
+#include <kernel/Scheduler.h>
+#include <kernel/Thread.h>
 #include <kernel/Timer/Timer.h>

 namespace Kernel::ACPI
@@ -351,17 +351,17 @@ namespace Kernel
 			// give processor upto 100 * 100 us + 200 us to boot
 			PageTable::with_fast_page(ap_init_paddr, [&] {
 				for (int i = 0; i < 100; i++, udelay(100))
-					if (__atomic_load_n(&PageTable::fast_page_as<ap_init_info_t>(8).ready, __ATOMIC_SEQ_CST))
+					if (BAN::atomic_load(PageTable::fast_page_as<ap_init_info_t>(8).ready))
 						break;
 			});

 			initialized_aps++;
 		}

-		__atomic_store_n(&g_ap_startup_done[0], 1, __ATOMIC_SEQ_CST);
+		BAN::atomic_store(g_ap_startup_done[0], 1);

 		const size_t timeout_ms = SystemTimer::get().ms_since_boot() + 500;
-		while (__atomic_load_n(&g_ap_running_count[0], __ATOMIC_SEQ_CST) < initialized_aps)
+		while (BAN::atomic_load(g_ap_running_count[0]) < initialized_aps)
 		{
 			if (SystemTimer::get().ms_since_boot() >= timeout_ms)
 				Kernel::panic("Could not start all APs ({}/{} started)", g_ap_running_count[0], initialized_aps);
@@ -511,9 +511,12 @@ namespace Kernel
 		redir.lo_dword = ioapic->read(IOAPIC_REDIRS + pin * 2);
 		redir.hi_dword = ioapic->read(IOAPIC_REDIRS + pin * 2 + 1);

-		redir.trigger_mode = level_triggered;
 		redir.vector = IRQ_VECTOR_BASE + irq;
+		redir.delivery_mode = 0; // fixed
+		redir.pin_polarity  = 0; // active high
+		redir.trigger_mode = level_triggered;
 		redir.mask = 0;
+		redir.destination_mode = 0; // physical
 		// FIXME: distribute IRQs more evenly?
 		redir.destination = Kernel::Processor::bsp_id().as_u32();

@@ -224,7 +224,12 @@ namespace Kernel
 	BAN::ErrorOr<void> AC97AudioController::set_volume_mdB(int32_t mdB)
 	{
 		m_volume_info.mdB = BAN::Math::clamp(mdB, m_volume_info.min_mdB, m_volume_info.max_mdB);
-		m_mixer->write16(AudioMixerRegister::MasterVolume, get_volume_data());
+
+		const uint32_t volume_data = get_volume_data();
+		m_mixer->write16(AudioMixerRegister::MasterVolume, volume_data);
+
+		m_volume_info.mdB = -(volume_data & 0xFF) * m_volume_info.step_mdB;
+
 		return {};
 	}

@@ -4,6 +4,7 @@
 #include <kernel/Device/DeviceNumbers.h>
 #include <kernel/FS/DevFS/FileSystem.h>
 #include <kernel/Lock/BlockableSpinLock.h>
+#include <kernel/Thread.h>

 #include <sys/ioctl.h>
 #include <sys/sysmacros.h>
@@ -139,31 +139,27 @@ namespace Kernel

 	BAN::ErrorOr<void> HDAudioFunctionGroup::set_volume_mdB(int32_t mdB)
 	{
+		if (m_amplifier_idx == SIZE_MAX)
+			return BAN::Error::from_errno(ENOTSUP);
+
 		mdB = BAN::Math::clamp(mdB, m_volume_info.min_mdB, m_volume_info.max_mdB);

-		const auto& path = m_output_paths[m_output_path_index];
-		for (size_t i = 0; i < path.size(); i++)
-		{
-			if (!path[i]->output_amplifier.has_value())
-				continue;
+		const auto* node = m_output_paths[m_output_path_index][m_amplifier_idx];

 		const int32_t step_round = (mdB >= 0)
 			? +m_volume_info.step_mdB / 2
 			: -m_volume_info.step_mdB / 2;
-			const uint32_t step = (mdB + step_round) / m_volume_info.step_mdB + path[i]->output_amplifier->offset;
+		const uint32_t step = (mdB + step_round) / m_volume_info.step_mdB + node->output_amplifier->offset;
 		const uint32_t volume = 0b1'0'1'1'0000'0'0000000 | step;

 		TRY(m_controller->send_command({
 			.data = static_cast<uint8_t>(volume & 0xFF),
 			.command = static_cast<uint16_t>(0x300 | (volume >> 8)),
-				.node_index = path[i]->id,
+			.node_index = node->id,
 			.codec_address = m_cid,
 		}));

-			break;
-		}
-
-		m_volume_info.mdB = mdB;
+		m_volume_info.mdB = (mdB + step_round) / m_volume_info.step_mdB * m_volume_info.step_mdB;

 		return {};
 	}
@@ -331,8 +327,8 @@ namespace Kernel
 	uint16_t HDAudioFunctionGroup::get_format_data() const
 	{
 		// TODO: don't hardcode this
-		// format: PCM, 48 kHz, 16 bit, 2 channels
-		return 0b0'0'000'000'0'001'0001;
+		// format: PCM, 48 kHz, 16 bit
+		return 0b0'0'000'000'0'001'0000 | (get_channels() - 1);
 	}

 	BAN::ErrorOr<void> HDAudioFunctionGroup::enable_output_path(uint8_t index)
@@ -346,6 +342,7 @@ namespace Kernel
 			{
 				using HDAudio::AFGWidget;
 				case AFGWidget::Type::OutputConverter:
+				case AFGWidget::Type::Mixer:
 				case AFGWidget::Type::PinComplex:
 					break;
 				default:
@@ -367,7 +364,7 @@ namespace Kernel
 			}));

 			// set connection index
-			if (i + 1 < path.size() && path[i]->connections.size() > 1)
+			if (i + 1 < path.size() && path[i]->connections.size() > 1 && path[i]->type != HDAudio::AFGWidget::Type::Mixer)
 			{
 				uint8_t index = 0;
 				for (; index < path[i]->connections.size(); index++)
@@ -407,6 +404,13 @@ namespace Kernel
 						.node_index = path[i]->id,
 						.codec_address = m_cid,
 					}));
+					// set channel count
+					TRY(m_controller->send_command({
+						.data = static_cast<uint8_t>(get_channels() - 1),
+						.command = 0x72D,
+						.node_index = path[i]->id,
+						.codec_address = m_cid,
+					}));
 					// set format
 					TRY(m_controller->send_command({
 						.data = static_cast<uint8_t>(format & 0xFF),
@@ -416,6 +420,25 @@ namespace Kernel
 					}));
 					break;

+				case AFGWidget::Type::Mixer:
+					if (path[i]->input_amplifier.has_value())
+					{
+						for (size_t idx = 0; idx < path[i]->connections.size(); idx++)
+						{
+							const uint8_t step = (path[i]->connections[idx] == path[i + 1]->id)
+								? path[i]->input_amplifier->offset
+								: 0b1'0000000;
+							const uint32_t volume = 0b0'1'1'1'0000'0'0000000 | (idx << 8) | step;
+							TRY(m_controller->send_command({
+								.data = static_cast<uint8_t>(volume & 0xFF),
+								.command = static_cast<uint16_t>(0x300 | (volume >> 8)),
+								.node_index = path[i]->id,
+								.codec_address = m_cid,
+							}));
+						}
+					}
+					break;
+
 				case AFGWidget::Type::PinComplex:
 					// enable output and H-Phn
 					TRY(m_controller->send_command({
@@ -438,15 +461,25 @@ namespace Kernel
 			}
 		}

-		// update volume info to this path
-		m_volume_info.min_mdB = 0;
-		m_volume_info.max_mdB = 0;
-		m_volume_info.step_mdB = 0;
+		int32_t max_steps { 0 };
+		m_amplifier_idx = SIZE_MAX;
+
 		for (size_t i = 0; i < path.size(); i++)
 		{
 			if (!path[i]->output_amplifier.has_value())
 				continue;
-			const auto& amp = path[i]->output_amplifier.value();
+			if (auto steps = path[i]->output_amplifier->num_steps; steps > max_steps)
+			{
+				max_steps = steps;
+				m_amplifier_idx = i;
+			}
+		}
+
+		if (m_amplifier_idx == SIZE_MAX)
+			m_volume_info = {};
+		else
+		{
+			const auto& amp = path[m_amplifier_idx]->output_amplifier.value();

 			const int32_t step_mdB = (amp.step_size + 1) * 250;
 			m_volume_info.step_mdB = step_mdB;
@@ -463,16 +496,11 @@ namespace Kernel
 			TRY(m_controller->send_command({
 				.data = static_cast<uint8_t>(volume & 0xFF),
 				.command = static_cast<uint16_t>(0x300 | (volume >> 8)),
-				.node_index = path[i]->id,
+				.node_index = path[m_amplifier_idx]->id,
 				.codec_address = m_cid,
 			}));
-
-			break;
 		}

-		if (m_volume_info.min_mdB == 0 && m_volume_info.max_mdB == 0)
-			m_volume_info.mdB = 0;
-
 		return {};
 	}

@@ -496,6 +524,7 @@ namespace Kernel
 				using HDAudio::AFGWidget;

 				case AFGWidget::Type::OutputConverter:
+				case AFGWidget::Type::Mixer:
 					break;

 				case AFGWidget::Type::PinComplex:
@@ -635,11 +664,8 @@ namespace Kernel
 			m_bdl_tail = (m_bdl_tail + 1) % m_bdl_entry_count;
 			if (m_bdl_tail == m_bdl_head)
 			{
-				if (auto ret = reset_stream(); ret.is_error())
-				{
-					dwarnln("failed to reset HDA stream: {}", ret.error());
-					return;
-				}
+				bar.write8(base + Regs::SDCTL, bar.read8(base + Regs::SDCTL) & 0xFD);
+				m_stream_running = false;
 			}

 			queue_bdl_data();
@@ -333,6 +333,20 @@ namespace Kernel
 			};
 		}

+		if (const uint32_t in_amp_cap = send_command_or_zero(0xF00, 0x0D))
+		{
+			const uint8_t offset    = (in_amp_cap >>  0) & 0x7F;
+			const uint8_t num_steps = (in_amp_cap >>  8) & 0x7F;
+			const uint8_t step_size = (in_amp_cap >> 16) & 0x7F;
+			const bool    mute      = (in_amp_cap >> 31);
+			result.input_amplifier = HDAudio::AFGWidget::Amplifier {
+				.offset = offset,
+				.num_steps = num_steps,
+				.step_size = step_size,
+				.mute = mute,
+			};
+		}
+
 		const uint8_t connection_info = send_command_or_zero(0xF00, 0x0E);
 		const uint8_t conn_width = (connection_info & 0x80) ? 2 : 1;
 		const uint8_t conn_count = connection_info & 0x3F;
@@ -14,6 +14,8 @@
 #include <kernel/Lock/SpinLock.h>
 #include <kernel/UserCopy.h>

+#if ARCH(x86_64)
+
 using namespace LibELF;
 using namespace Kernel;

@@ -204,7 +206,7 @@ BAN::ErrorOr<size_t> Banos::load_driver_from_image(const char* u_image) {
 // NOTE: should be more than plenty ;)
 extern char g_drv_builtin_begin[];
 extern char g_drv_builtin_end[];
-void Banos::initialize_initial_drivers(void) {
+void Banos::initialize_initial_drivers() {
 	import_symbols(g_banos_export, g_banos_export_end - g_banos_export);
 	char* head = g_drv_builtin_begin;
 	while(head < g_drv_builtin_end) {
@@ -213,3 +215,14 @@ void Banos::initialize_initial_drivers(void) {
 		head += drv->driver_size;
 	}
 }
+
+#else
+void Banos::initialize_initial_drivers()
+{
+}
+BAN::ErrorOr<size_t> Banos::load_driver_from_image(const char* u_image)
+{
+	(void)u_image;
+	return BAN::Error::from_errno(ENOTSUP);
+}
+#endif
@@ -1,6 +1,7 @@
 #include <kernel/Epoll.h>
 #include <kernel/Lock/BlockableSpinLock.h>
 #include <kernel/Lock/LockGuard.h>
+#include <kernel/Thread.h>
 #include <kernel/Timer/Timer.h>

 namespace Kernel
@@ -1,5 +1,6 @@
 #include <kernel/FS/EventFD.h>
 #include <kernel/Lock/LockGuard.h>
+#include <kernel/Thread.h>

 #include <sys/epoll.h>

@@ -43,7 +44,7 @@ namespace Kernel
 		while (m_value == 0)
 			TRY(Thread::current().block_or_eintr_indefinite(m_thread_blocker, &m_mutex));

-		const uint64_t read_value = m_is_semaphore ? 1 : m_value.load();
+		const uint64_t read_value = m_is_semaphore ? 1 : m_value;
 		m_value -= read_value;

 		buffer.as<uint64_t>() = read_value;
@@ -79,4 +80,16 @@ namespace Kernel
 		return sizeof(uint64_t);
 	}

+	bool EventFD::can_read_impl() const
+	{
+		LockGuard _(m_mutex);
+		return m_value > 0;
+	}
+
+	bool EventFD::can_write_impl() const
+	{
+		LockGuard _(m_mutex);
+		return m_value < UINT64_MAX - 1;
+	}
+
 }
@@ -2,6 +2,7 @@
 #include <BAN/Sort.h>
 #include <kernel/FS/Ext2/FileSystem.h>
 #include <kernel/Lock/LockGuard.h>
+#include <kernel/Thread.h>

 #define EXT2_DEBUG_PRINT 0
 #define EXT2_VERIFY_INODE 0
@@ -1,3 +1,4 @@
+#include <kernel/Device/Device.h>
 #include <kernel/FS/FAT/FileSystem.h>
 #include <kernel/Lock/LockGuard.h>

@@ -123,18 +124,6 @@ namespace Kernel
 	{
 		LockGuard _(m_mutex);

-		uint32_t block_count = 0;
-		{
-			uint32_t cluster = entry.first_cluster_lo;
-			if (m_type == Type::FAT32)
-				cluster |= static_cast<uint32_t>(entry.first_cluster_hi) << 16;
-			while (cluster >= 2 && cluster < cluster_count())
-			{
-				block_count++;
-				cluster = TRY(get_next_cluster(cluster));
-			}
-		}
-
 		uint32_t entry_cluster;
 		switch (m_type)
 		{
@@ -143,25 +132,34 @@ namespace Kernel
 				if (parent == m_root_inode)
 					entry_cluster = 1;
 				else
-				{
 					entry_cluster = parent->entry().first_cluster_lo;
-					for (uint32_t i = 0; i < cluster_index; i++)
-						entry_cluster = TRY(get_next_cluster(entry_cluster));
-				}
 				break;
 			case Type::FAT32:
 				if (parent == m_root_inode)
 					entry_cluster = m_bpb.ext_32.root_cluster;
 				else
 					entry_cluster = (static_cast<uint32_t>(parent->entry().first_cluster_hi) << 16) | parent->entry().first_cluster_lo;
-				for (uint32_t i = 0; i < cluster_index; i++)
-					entry_cluster = TRY(get_next_cluster(entry_cluster));
 				break;
 			default:
 				ASSERT_NOT_REACHED();
 		}

-		const ino_t ino = (static_cast<ino_t>(entry_cluster) << 32) | entry_index;
+		uint32_t block_count = 0;
+		for (uint32_t i = 0; i < cluster_index; i++)
+		{
+			block_count++;
+			entry_cluster = TRY(get_next_cluster(entry_cluster));
+		}
+
+		const uint32_t dirent_per_cluster = m_bpb.bytes_per_sector * m_bpb.sectors_per_cluster / sizeof(FAT::DirectoryEntry);
+		ASSERT(BAN::Math::is_power_of_two(dirent_per_cluster));
+		ASSERT(entry_index < dirent_per_cluster);
+
+		const uint32_t ino_cluster_shift = BAN::Math::ctz(dirent_per_cluster);
+		const ino_t ino = (static_cast<ino_t>(entry_cluster) << ino_cluster_shift) | entry_index;
+		if (ino >> ino_cluster_shift != entry_cluster)
+			dwarnln("FAT ino mapping not unique");
+
 		auto it = m_inode_cache.find(ino);
 		if (it != m_inode_cache.end())
 		{
@@ -1,5 +1,6 @@
 #include <BAN/Time.h>

+#include <kernel/Device/Device.h>
 #include <kernel/FS/FAT/FileSystem.h>
 #include <kernel/FS/FAT/Inode.h>

@@ -1,6 +1,7 @@
 #include <kernel/BootInfo.h>
 #include <kernel/FS/ProcFS/FileSystem.h>
 #include <kernel/FS/ProcFS/Inode.h>
+#include <kernel/Process.h>

 namespace Kernel
 {
@@ -1,4 +1,5 @@
 #include <kernel/FS/ProcFS/Inode.h>
+#include <kernel/Process.h>

 #include <ctype.h>

@@ -1,7 +1,9 @@
 #include <BAN/Array.h>
 #include <BAN/Errors.h>
+#include <kernel/GDT.h>
 #include <kernel/IDT.h>
 #include <kernel/InterruptController.h>
+#include <kernel/InterruptNumbers.h>
 #include <kernel/InterruptStack.h>
 #include <kernel/Memory/kmalloc.h>
 #include <kernel/Panic.h>
@@ -4,7 +4,9 @@
 #include <kernel/FS/DevFS/FileSystem.h>
 #include <kernel/Input/InputDevice.h>
 #include <kernel/Lock/BlockableSpinLock.h>
+#include <kernel/Scheduler.h>
 #include <kernel/Terminal/TTY.h>
+#include <kernel/Thread.h>

 #include <LibInput/Joystick.h>
 #include <LibInput/KeyEvent.h>
@@ -8,6 +8,8 @@
 #include <kernel/Input/PS2/Keyboard.h>
 #include <kernel/Input/PS2/Mouse.h>
 #include <kernel/IO.h>
+#include <kernel/Scheduler.h>
+#include <kernel/Thread.h>
 #include <kernel/Timer/Timer.h>

 namespace Kernel::Input
@@ -345,7 +347,7 @@ namespace Kernel::Input
 									result.command_port = data->as.fixed_io_port.range_base;
 								break;
 							case ACPI::ResourceData::Type::IRQ:
-								if (__builtin_popcount(data->as.irq.irq_mask) != 1)
+								if (BAN::Math::popcount(data->as.irq.irq_mask) != 1)
 									break;
 								for (int i = 0; i < 16; i++)
 									if (data->as.irq.irq_mask & (1 << i))
@@ -386,14 +388,11 @@ namespace Kernel::Input
 			}

 			dprintln("Found {} PS/2 devices from ACPI namespace", acpi_devices.size());
-			if (acpi_devices.empty())
-				return {};

 			if (acpi_devices.size() > 2)
 			{
 				dwarnln("TODO: over 2 PS/2 devices");
-				while (acpi_devices.size() > 2)
-					acpi_devices.pop_back();
+				MUST(acpi_devices.resize(2));
 			}

 			if (acpi_devices.size() == 2)
@@ -419,6 +418,8 @@ namespace Kernel::Input
 				}
 			}

+			if (!acpi_devices.empty())
+			{
 				BAN::Optional<uint16_t> command_port;
 				command_port = acpi_devices[0].command_port;
 				if (!command_port.has_value() && acpi_devices.size() >= 2)
@@ -432,21 +433,20 @@ namespace Kernel::Input
 					data_port = acpi_devices[1].data_port;
 				if (data_port.has_value())
 					m_data_port = data_port.value();
+			}

-			devices[0] = {
-				.type      = acpi_devices[0].type,
-				.interrupt = acpi_devices[0].irq.value_or(PS2::IRQ::DEVICE0)
-			};
-
-			if (acpi_devices.size() > 1)
+			for (size_t i = 0; i < acpi_devices.size(); i++)
 			{
-				devices[1] = {
-					.type      = acpi_devices[1].type,
-					.interrupt = acpi_devices[1].irq.value_or(PS2::IRQ::DEVICE1)
+				devices[i] = {
+					.type      = acpi_devices[i].type,
+					.interrupt = acpi_devices[i].irq.value_or(i == 0 ? PS2::IRQ::DEVICE0 : PS2::IRQ::DEVICE1),
 				};
 			}
 		}
-		else if (has_legacy_8042())
+
+		if (devices[0].type != DeviceType::None)
+			;
+		else if (scancode_set || has_legacy_8042())
 		{
 			devices[0] = {
 				.type = DeviceType::Unknown,
@@ -53,8 +53,9 @@ namespace Kernel::Input

 		if (command_data[0] == Command::CONFIG_SCANCODE_SET && m_scancode_set >= 0xFE)
 		{
-			dwarnln("Could not detect scancode set, assuming 2");
-			m_scancode_set = 2;
+			dwarnln("Could not detect scancode set, assuming 1");
+			m_scancode_set_uncertain = true;
+			m_scancode_set = 1;
 			m_keymap.initialize(m_scancode_set);
 			append_command_queue(PS2::DeviceCommand::ENABLE_SCANNING, 0);
 		}
@@ -89,6 +90,7 @@ namespace Kernel::Input
 			else
 			{
 				dwarnln("Could not detect scancode set, assuming 1");
+				m_scancode_set_uncertain = true;
 				m_scancode_set = 1;
 			}
 			m_keymap.initialize(m_scancode_set);
@@ -103,6 +105,17 @@ namespace Kernel::Input
 			return;
 		}

+		// If we could not detect scancode set, we assume it to be 1
+		// If we get byte 0xF0 which is indicates release in scancode set 2
+		// and nothing in scancode set 1, switch to scancode set 2
+		if (m_scancode_set_uncertain && byte == 0xF0)
+		{
+			dprintln("Switching to scancode set 2");
+			m_scancode_set_uncertain = false;
+			m_scancode_set = 2;
+			m_keymap.initialize(m_scancode_set);
+		}
+
 		m_byte_buffer[m_byte_index++] = byte;
 		if (byte == 0xE0)
 			return;
@@ -0,0 +1,134 @@
+#include <kernel/Lock/Mutex.h>
+#include <kernel/Thread.h>
+
+namespace Kernel
+{
+
+	bool Mutex::try_lock()
+	{
+		const auto tid = Thread::current_tid();
+		if (tid == m_locker)
+			ASSERT(m_lock_depth > 0);
+		else
+		{
+			pid_t expected = -1;
+			if (!m_locker.compare_exchange(expected, tid))
+				return false;
+			ASSERT(m_lock_depth == 0);
+			if (tid)
+				Thread::current().add_mutex();
+		}
+		m_lock_depth++;
+		return true;
+	}
+
+	void Mutex::lock()
+	{
+		const auto tid = Thread::current_tid();
+		if (tid == m_locker)
+			ASSERT(m_lock_depth > 0);
+		else
+		{
+			ASSERT(!tid || !Thread::current().has_spinlock());
+			pid_t expected = -1;
+			while (!m_locker.compare_exchange(expected, tid))
+			{
+				ASSERT(Processor::get_interrupt_state() == InterruptState::Enabled);
+				Processor::yield();
+				expected = -1;
+			}
+			ASSERT(m_lock_depth == 0);
+			if (tid)
+				Thread::current().add_mutex();
+		}
+		m_lock_depth++;
+	}
+
+	void Mutex::unlock()
+	{
+		const auto tid = Thread::current_tid();
+		ASSERT(m_locker == tid);
+		ASSERT(m_lock_depth > 0);
+		if (--m_lock_depth == 0)
+		{
+			m_locker = -1;
+			if (tid)
+				Thread::current().remove_mutex();
+		}
+	}
+
+	bool Mutex::is_locked_by_current_thread() const
+	{
+		return m_locker == Thread::current_tid();
+	}
+
+	bool PriorityMutex::try_lock()
+	{
+		const auto tid = Thread::current_tid();
+
+		if (tid == m_locker)
+			ASSERT(m_lock_depth > 0);
+		else
+		{
+			bool has_priority = tid ? !Thread::current().is_userspace() : true;
+			pid_t expected = -1;
+			if (!(has_priority || m_queue_length == 0) || !m_locker.compare_exchange(expected, tid))
+				return false;
+			if (has_priority)
+				m_queue_length++;
+			ASSERT(m_lock_depth == 0);
+			if (tid)
+				Thread::current().add_mutex();
+		}
+		m_lock_depth++;
+		return true;
+	}
+
+	void PriorityMutex::lock()
+	{
+		const auto tid = Thread::current_tid();
+
+		if (tid == m_locker)
+			ASSERT(m_lock_depth > 0);
+		else
+		{
+			ASSERT(!tid || !Thread::current().has_spinlock());
+			bool has_priority = tid ? !Thread::current().is_userspace() : true;
+			if (has_priority)
+				m_queue_length++;
+			pid_t expected = -1;
+			while (!(has_priority || m_queue_length == 0) || !m_locker.compare_exchange(expected, tid))
+			{
+				ASSERT(Processor::get_interrupt_state() == InterruptState::Enabled);
+				Processor::yield();
+				expected = -1;
+			}
+			ASSERT(m_lock_depth == 0);
+			if (tid)
+				Thread::current().add_mutex();
+		}
+		m_lock_depth++;
+	}
+
+	void PriorityMutex::unlock()
+	{
+		const auto tid = Thread::current_tid();
+		ASSERT(m_locker == tid);
+		ASSERT(m_lock_depth > 0);
+		if (--m_lock_depth == 0)
+		{
+			bool has_priority = tid ? !Thread::current().is_userspace() : true;
+			if (has_priority)
+				m_queue_length--;
+			m_locker = -1;
+			if (tid)
+				Thread::current().remove_mutex();
+		}
+	}
+
+	bool PriorityMutex::is_locked_by_current_thread() const
+	{
+		return m_locker == Thread::current_tid();
+	}
+
+}
@@ -0,0 +1,58 @@
+
+#include <kernel/Lock/BlockableSpinLock.h>
+#include <kernel/Lock/RWLock.h>
+#include <kernel/Thread.h>
+
+namespace Kernel
+{
+
+	void RWLock::rd_lock()
+	{
+		SpinLockGuard _(m_lock);
+		while (m_writers_waiting > 0 || m_writer != -1)
+		{
+			BlockableSpinLock block(m_lock);
+			m_thread_blocker.block_indefinite(&block);
+		}
+		m_readers_active++;
+	}
+
+	void RWLock::rd_unlock()
+	{
+		SpinLockGuard _(m_lock);
+		if (--m_readers_active == 0)
+			m_thread_blocker.unblock();
+	}
+
+	void RWLock::wr_lock()
+	{
+		if (m_writer == Thread::current_tid())
+		{
+			m_writer_depth++;
+			return;
+		}
+
+		SpinLockGuard _(m_lock);
+
+		m_writers_waiting++;
+		while (m_readers_active > 0 || m_writer != -1)
+		{
+			BlockableSpinLock block(m_lock);
+			m_thread_blocker.block_indefinite(&block);
+		}
+		m_writers_waiting--;
+
+		m_writer = Thread::current_tid();
+		m_writer_depth = 1;
+	}
+
+	void RWLock::wr_unlock()
+	{
+		if (--m_writer_depth != 0)
+			return;
+		SpinLockGuard _(m_lock);
+		m_writer = -1;
+		m_thread_blocker.unblock();
+	}
+
+}
@@ -115,7 +115,7 @@ namespace Kernel
 			return {};

 		const vaddr_t first_page = BAN::Math::max(m_vaddr, address) & PAGE_ADDR_MASK;
-		const vaddr_t last_page = BAN::Math::div_round_up(BAN::Math::min(m_vaddr + m_size, address + size), PAGE_SIZE) * PAGE_SIZE;
+		const vaddr_t last_page = BAN::Math::div_round_up<vaddr_t>(BAN::Math::min(m_vaddr + m_size, address + size), PAGE_SIZE) * PAGE_SIZE;

 		RWLockRDGuard _(m_shared_data->rw_lock);
 		for (vaddr_t page_addr = first_page; page_addr < last_page; page_addr += PAGE_SIZE)
@@ -43,11 +43,10 @@ namespace Kernel
 			PageTable::with_per_cpu_fast_page(current_paddr, [&page_matched_bit](void* addr) {
 				for (size_t j = 0; j < PAGE_SIZE / sizeof(size_t); j++)
 				{
-					static_assert(sizeof(size_t) == sizeof(long));
 					auto& current = static_cast<size_t*>(addr)[j];
 					if (current == BAN::numeric_limits<size_t>::max())
 						continue;
-					const int ctz = __builtin_ctzl(~current);
+					const int ctz = BAN::Math::ctz(~current);
 					current |= static_cast<size_t>(1) << ctz;
 					page_matched_bit = j * sizeof(size_t) * 8 + ctz;
 					return;
@@ -93,7 +93,15 @@ namespace Kernel
 		const pid_t pid = process.pid();
 		const mode_t mode = shmflg & 0777;

-		auto object = TRY(BAN::RefPtr<Object>::create(key, shmid_ds {
+		const int shmid = ({
+			int id;
+			do {
+				id = Random::get<unsigned>() & BAN::numeric_limits<int>::max();
+			} while (m_ids.contains(shmid));
+			id;
+		});
+
+		auto object = TRY(BAN::RefPtr<Object>::create(key, shmid, shmid_ds {
 				.shm_perm = {
 					.uid = uid,
 					.gid = gid,
@@ -109,13 +117,7 @@ namespace Kernel
 				.shm_dtime = 0,
 				.shm_ctime = SystemTimer::get().real_time().tv_sec,
 		}));
-		TRY(object->paddrs.resize(BAN::Math::div_round_up(size, PAGE_SIZE), 0));
-
-		auto generate_id = []() { return Random::get<unsigned>() & BAN::numeric_limits<int>::max(); };
-
-		int shmid = generate_id();
-		while (m_ids.contains(shmid))
-			shmid = generate_id();
+		TRY(object->paddrs.resize(BAN::Math::div_round_up<size_t>(size, PAGE_SIZE), 0));

 		if (key != IPC_PRIVATE)
 			TRY(m_ids.insert(key, shmid));
@@ -222,14 +224,16 @@ namespace Kernel
 		LockGuard _(m_object->mutex);
 		m_object->info.shm_nattch++;
 		m_object->info.shm_atime = SystemTimer::get().real_time().tv_sec;
+		m_object->info.shm_lpid = Process::current().pid();
 	}

 	SharedMemoryObject::~SharedMemoryObject()
 	{
 		LockGuard _(m_object->mutex);
 		if (--m_object->info.shm_nattch == 0 && m_object->marked_for_deletion)
-			SharedMemoryObjectManager::get().m_objects.remove(m_object->key);
+			SharedMemoryObjectManager::get().m_objects.remove(m_object->id);
 		m_object->info.shm_dtime = SystemTimer::get().real_time().tv_sec;
+		m_object->info.shm_lpid = Process::current().pid();
 	}

 	BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> SharedMemoryObject::clone(PageTable& new_page_table)
@@ -125,8 +125,6 @@ struct BitmapAllocator
 		// NOTE: We could optimize other bitmap functions than this
 		//       but this one is the bottle neck so it doesn't matter

-		static_assert(sizeof(unsigned long long) == sizeof(uint64_t));
-
 		if (index >= total_chunks)
 			return index;

@@ -134,7 +132,7 @@ struct BitmapAllocator
 		{
 			const uint64_t qword = *reinterpret_cast<const uint64_t*>(base + (index - rem) / 8) >> rem;
 			if (qword != (1ull << (64 - rem)) - 1)
-				return index + __builtin_ctzll(~qword);
+				return index + BAN::Math::ctz(~qword);
 			index += 64 - rem;
 		}

@@ -142,7 +140,7 @@ struct BitmapAllocator
 		{
 			const uint64_t qword = *reinterpret_cast<const uint64_t*>(base + index / 8);
 			if (qword != UINT64_MAX)
-				return index + __builtin_ctzll(~qword);
+				return index + BAN::Math::ctz(~qword);
 			index += 64;
 		}

@@ -6,6 +6,8 @@
 #include <kernel/MMIO.h>
 #include <kernel/Networking/E1000/E1000.h>
 #include <kernel/Networking/NetworkManager.h>
+#include <kernel/Scheduler.h>
+#include <kernel/Thread.h>

 namespace Kernel
 {
@@ -1,6 +1,8 @@
 #include <kernel/Lock/BlockableSpinLock.h>
 #include <kernel/Networking/Loopback.h>
 #include <kernel/Networking/NetworkManager.h>
+#include <kernel/Scheduler.h>
+#include <kernel/Thread.h>

 namespace Kernel
 {
@@ -2,6 +2,8 @@
 #include <kernel/Networking/NetworkManager.h>
 #include <kernel/Networking/RTL8169/Definitions.h>
 #include <kernel/Networking/RTL8169/RTL8169.h>
+#include <kernel/Scheduler.h>
+#include <kernel/Thread.h>
 #include <kernel/Timer/Timer.h>

 namespace Kernel
@@ -3,6 +3,7 @@
 #include <kernel/Networking/TCPSocket.h>
 #include <kernel/Process.h>
 #include <kernel/Random.h>
+#include <kernel/Scheduler.h>
 #include <kernel/Timer/Timer.h>

 #include <fcntl.h>
@@ -75,7 +75,7 @@ namespace Kernel
 		if (flags & ~(O_ACCMODE | O_NOFOLLOW | O_APPEND | O_TRUNC | O_CLOEXEC | O_TTY_INIT | O_NOCTTY | O_DIRECTORY | O_CREAT | O_EXCL | O_NONBLOCK))
 			return BAN::Error::from_errno(ENOTSUP);

-		if ((flags & O_ACCMODE) != O_RDWR && __builtin_popcount(flags & O_ACCMODE) != 1)
+		if ((flags & O_ACCMODE) != O_RDWR && BAN::Math::popcount<unsigned>(flags & O_ACCMODE) != 1)
 			return BAN::Error::from_errno(EINVAL);

 		if ((flags & O_DIRECTORY) && !file.inode->mode().ifdir())
@@ -1,4 +1,5 @@
 #include <kernel/IDT.h>
+#include <kernel/InterruptNumbers.h>
 #include <kernel/IO.h>
 #include <kernel/PIC.h>

@@ -696,19 +696,24 @@ namespace Kernel

 	size_t Process::proc_cputime(off_t offset, BAN::ByteSpan buffer) const
 	{
-		const uint64_t cpu_time_ns = [this] {
-			uint64_t cpu_time_ns { 0 };
+		uint64_t user_ns { 0 }, system_ns { 0 };
+
+		{
 			LockGuard _(m_process_lock);
 			for (auto* thread : m_threads)
-				cpu_time_ns += thread->cpu_time_ns();
-			return cpu_time_ns;
-		}();
+			{
+				uint64_t u, s;
+				thread->cpu_time_ns(u, s);
+				user_ns   += u;
+				system_ns += s;
+			}
+		}

-		auto data = MUST(BAN::String::formatted("{}", cpu_time_ns));
-		if (static_cast<size_t>(offset) >= data.size() + 1)
+		auto data = MUST(BAN::String::formatted("{} {} {}", user_ns + system_ns, user_ns, system_ns));
+		if (static_cast<size_t>(offset) >= data.size())
 			return 0;

-		const size_t to_copy = BAN::Math::min<size_t>(data.size() - offset + 1, buffer.size());
+		const size_t to_copy = BAN::Math::min<size_t>(data.size() - offset, buffer.size());
 		memcpy(buffer.data(), data.data(), to_copy);
 		return to_copy;
 	}
@@ -3054,7 +3059,7 @@ namespace Kernel
 				LockGuard _(m_process_lock);
 				uint64_t cpu_time_ns { 0 };
 				for (auto* thread : m_threads)
-					cpu_time_ns += thread->cpu_time_ns();
+					cpu_time_ns += thread->cpu_time_total_ns();
 				tp = {
 					.tv_sec = static_cast<time_t>(cpu_time_ns / 1'000'000'000),
 					.tv_nsec = static_cast<long>(cpu_time_ns % 1'000'000'000),
@@ -3063,7 +3068,7 @@ namespace Kernel
 			}
 			case CLOCK_THREAD_CPUTIME_ID:
 			{
-				const auto cpu_time_ns = Thread::current().cpu_time_ns();
+				const auto cpu_time_ns = Thread::current().cpu_time_total_ns();
 				tp = {
 					.tv_sec = static_cast<time_t>(cpu_time_ns / 1'000'000'000),
 					.tv_nsec = static_cast<long>(cpu_time_ns % 1'000'000'000),
@@ -1,8 +1,12 @@
 #include <kernel/CPUID.h>
+#include <kernel/GDT.h>
+#include <kernel/IDT.h>
 #include <kernel/InterruptController.h>
+#include <kernel/IO.h>
 #include <kernel/Memory/Heap.h>
 #include <kernel/Memory/kmalloc.h>
 #include <kernel/Processor.h>
+#include <kernel/Scheduler.h>
 #include <kernel/Terminal/TerminalDriver.h>
 #include <kernel/Thread.h>
 #include <kernel/Timer/Timer.h>
@@ -1,11 +1,11 @@
 #include <BAN/Optional.h>
 #include <BAN/Sort.h>
 #include <kernel/APIC.h>
+#include <kernel/GDT.h>
 #include <kernel/InterruptController.h>
 #include <kernel/Lock/Mutex.h>
 #include <kernel/Process.h>
 #include <kernel/Scheduler.h>
-#include <kernel/SchedulerQueueNode.h>
 #include <kernel/Thread.h>
 #include <kernel/Timer/Timer.h>

@@ -36,85 +36,6 @@ namespace Kernel

 	static BAN::Atomic<size_t> s_next_processor_index { 0 };

-
-	void SchedulerQueue::add_thread_to_back(Node* node)
-	{
-		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
-		node->next = nullptr;
-		node->prev = m_tail;
-		(m_tail ? m_tail->next : m_head) = node;
-		m_tail = node;
-	}
-
-	bool SchedulerQueue::add_thread_with_wake_time(Node* node)
-	{
-		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
-
-		if (m_tail == nullptr || node->wake_time_ns >= m_tail->wake_time_ns)
-		{
-			add_thread_to_back(node);
-			return node == m_head;
-		}
-
-		Node* next = m_head;
-		Node* prev = nullptr;
-		while (next && node->wake_time_ns > next->wake_time_ns)
-		{
-			prev = next;
-			next = next->next;
-		}
-
-		node->next = next;
-		node->prev = prev;
-		(next ? next->prev : m_tail) = node;
-		(prev ? prev->next : m_head) = node;
-
-		return node == m_head;
-	}
-
-	template<typename F>
-	SchedulerQueue::Node* SchedulerQueue::remove_with_condition(F callback)
-	{
-		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
-
-		for (Node* node = m_head; node; node = node->next)
-		{
-			if (!callback(node))
-				continue;
-			remove_node(node);
-			return node;
-		}
-
-		return nullptr;
-	}
-
-	void SchedulerQueue::remove_node(Node* node)
-	{
-		(node->prev ? node->prev->next : m_head) = node->next;
-		(node->next ? node->next->prev : m_tail) = node->prev;
-		node->prev = nullptr;
-		node->next = nullptr;
-	}
-
-	SchedulerQueue::Node* SchedulerQueue::front()
-	{
-		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
-		ASSERT(!empty());
-		return m_head;
-	}
-
-	SchedulerQueue::Node* SchedulerQueue::pop_front()
-	{
-		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
-		if (empty())
-			return nullptr;
-		Node* result = m_head;
-		m_head = m_head->next;
-		(m_head ? m_head->prev : m_tail) = nullptr;
-		result->next = nullptr;
-		return result;
-	}
-
 	BAN::ErrorOr<Scheduler*> Scheduler::create()
 	{
 		auto* scheduler = new Scheduler();
@@ -142,7 +63,7 @@ namespace Kernel
 		return {};
 	}

-	void Scheduler::add_current_to_most_loaded(SchedulerQueue* target_queue)
+	void Scheduler::add_current_to_most_loaded(void* target_list)
 	{
 		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);

@@ -151,7 +72,7 @@ namespace Kernel
 		{
 			if (info.node == m_current)
 			{
-				info.queue = target_queue;
+				info.list = target_list;
 				has_current = true;
 				break;
 			}
@@ -163,8 +84,10 @@ namespace Kernel
 			for (; index < m_most_loaded_threads.size() - 1; index++)
 				if (m_most_loaded_threads[index].node == nullptr)
 					break;
-			m_most_loaded_threads[index].queue = target_queue;
-			m_most_loaded_threads[index].node  = m_current;
+			m_most_loaded_threads[index] = {
+				.list = target_list,
+				.node = m_current,
+			};
 		}

 		BAN::sort::sort(m_most_loaded_threads.begin(), m_most_loaded_threads.end(),
@@ -177,7 +100,7 @@ namespace Kernel
 		);
 	}

-	void Scheduler::update_most_loaded_node_queue(SchedulerQueue::Node* node, SchedulerQueue* target_queue)
+	void Scheduler::update_most_loaded_node_list(SchedulerThreadNode* node, void* target_list)
 	{
 		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);

@@ -185,13 +108,13 @@ namespace Kernel
 		{
 			if (info.node == node)
 			{
-				info.queue = target_queue;
+				info.list = target_list;
 				break;
 			}
 		}
 	}

-	void Scheduler::remove_node_from_most_loaded(SchedulerQueue::Node* node)
+	void Scheduler::remove_node_from_most_loaded(SchedulerThreadNode* node)
 	{
 		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);

@@ -203,8 +126,10 @@ namespace Kernel
 		for (; i < m_most_loaded_threads.size() - 1; i++)
 			m_most_loaded_threads[i] = m_most_loaded_threads[i + 1];

-		m_most_loaded_threads.back().node = nullptr;
-		m_most_loaded_threads.back().queue = nullptr;
+		m_most_loaded_threads.back() = {
+			.list = nullptr,
+			.node = nullptr,
+		};
 	}

 	void Scheduler::reschedule(YieldRegisters* yield_registers)
@@ -212,7 +137,7 @@ namespace Kernel
 		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);

 		// If there are no other threads in run queue, reschedule can be no-op :)
-		if (m_run_queue.empty() && (!m_current || !m_current->blocked) && current_thread().state() == Thread::State::Executing)
+		if (m_run_list.empty() && (!m_current || !m_current->blocked) && current_thread().state() == Thread::State::Executing)
 			return;

 		if (m_current == nullptr)
@@ -232,12 +157,13 @@ namespace Kernel
 				case Thread::State::Executing:
 					m_current->thread->yield_registers() = *yield_registers;
 					m_current->time_used_ns += SystemTimer::get().ns_since_boot() - m_current->last_start_ns;
-					add_current_to_most_loaded(m_current->blocked ? &m_block_queue : &m_run_queue);
+					add_current_to_most_loaded(m_current->blocked ? static_cast<void*>(&m_block_list) : &m_run_list);
 					if (!m_current->blocked)
-						m_run_queue.add_thread_to_back(m_current);
+						m_run_list.push(m_current);
 					else
 					{
-						if (m_block_queue.add_thread_with_wake_time(m_current))
+						m_block_list.push(m_current);
+						if (m_block_list.front() == m_current)
 							update_wake_up_deadline();
 						Processor::set_disable_smp_messages(false);
 					}
@@ -246,12 +172,12 @@ namespace Kernel
 					ASSERT(!m_current->blocked);
 					m_current->time_used_ns = 0;
 					remove_node_from_most_loaded(m_current);
-					m_run_queue.add_thread_to_back(m_current);
+					m_run_list.push(m_current);
 					break;
 			}
 		}

-		while ((m_current = m_run_queue.pop_front()))
+		while ((m_current = m_run_list.pop_front()))
 		{
 			if (m_current->thread->state() != Thread::State::Terminated)
 				break;
@@ -273,7 +199,7 @@ namespace Kernel
 			return;
 		}

-		update_most_loaded_node_queue(m_current, nullptr);
+		update_most_loaded_node_list(m_current, nullptr);

 		auto* thread = m_current->thread;

@@ -309,8 +235,8 @@ namespace Kernel
 		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);

 		const uint64_t current_ns = SystemTimer::get().ns_since_boot();
-		while (!m_block_queue.empty() && current_ns >= m_block_queue.front()->wake_time_ns)
-			unblock_thread(m_block_queue.front());
+		while (!m_block_list.empty() && current_ns >= m_block_list.front()->wake_time_ns)
+			unblock_thread(m_block_list.front());
 	}

 	void Scheduler::update_wake_up_deadline()
@@ -324,8 +250,8 @@ namespace Kernel
 			return;

 		uint64_t deadline_ns = m_next_reschedule_ns;
-		if (!m_block_queue.empty())
-			deadline_ns = BAN::Math::min(deadline_ns, m_block_queue.front()->wake_time_ns);
+		if (!m_block_list.empty())
+			deadline_ns = BAN::Math::min(deadline_ns, m_block_list.front()->wake_time_ns);
 		if (Processor::is_smp_enabled())
 			deadline_ns = BAN::Math::min(deadline_ns, m_last_load_balance_ns + s_load_balance_interval_ns);

@@ -336,7 +262,7 @@ namespace Kernel
 	{
 		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);

-		if ((is_idle() && !m_run_queue.empty()) || m_has_pending_reschedule)
+		if ((is_idle() && !m_run_list.empty()) || m_has_pending_reschedule)
 		{
 			m_has_pending_reschedule = false;
 			Processor::yield();
@@ -374,7 +300,7 @@ namespace Kernel
 			update_wake_up_deadline();
 	}

-	void Scheduler::unblock_thread(SchedulerQueue::Node* node)
+	void Scheduler::unblock_thread(SchedulerThreadNode* node)
 	{
 		auto state = Processor::get_interrupt_state();
 		Processor::set_interrupt_state(InterruptState::Disabled);
@@ -385,12 +311,12 @@ namespace Kernel
 				return;
 			ASSERT(node != m_current);
 			Processor::set_disable_smp_messages(true);
-			m_block_queue.remove_node(node);
+			m_block_list.pop(node);
 			if (auto* blocker = node->blocker.load())
 				blocker->remove_thread_from_block_queue(node);
 			node->blocked = false;
-			m_run_queue.add_thread_to_back(node);
-			update_most_loaded_node_queue(node, &m_run_queue);
+			m_run_list.push(node);
+			update_most_loaded_node_list(node, &m_run_list);
 			Processor::set_disable_smp_messages(false);
 		}
 		else
@@ -404,7 +330,7 @@ namespace Kernel
 		Processor::set_interrupt_state(state);
 	}

-	void Scheduler::add_thread(SchedulerQueue::Node* node)
+	void Scheduler::add_thread(SchedulerThreadNode* node)
 	{
 		auto state = Processor::get_interrupt_state();
 		Processor::set_interrupt_state(InterruptState::Disabled);
@@ -412,9 +338,13 @@ namespace Kernel
 		ASSERT(node->processor_id == Processor::current_id());

 		if (!node->blocked)
-			m_run_queue.add_thread_to_back(node);
-		else if (m_block_queue.add_thread_with_wake_time(node))
+			m_run_list.push(node);
+		else
+		{
+			m_block_list.push(node);
+			if (m_block_list.front() == node)
 				update_wake_up_deadline();
+		}

 		if (auto* thread = node->thread; thread->is_userspace() && thread->has_process())
 			thread->update_processor_index_address();
@@ -482,21 +412,21 @@ namespace Kernel
 				const uint64_t load_percent_x1000 = BAN::Math::div_round_up<uint64_t>(m_current->time_used_ns * 100'000, processing_ns);
 				dprintln("  tid { 2}: { 3}.{3}% <{}> current", m_current->thread->tid(), load_percent_x1000 / 1000, load_percent_x1000 % 1000, name);
 			}
-			m_run_queue.remove_with_condition(
-				[&](SchedulerQueue::Node* node)
+			m_run_list.walk(
+				[](const SchedulerThreadNode* node, void* arg)
 				{
+					const uint64_t processing_ns = *static_cast<const uint64_t*>(arg);
 					const uint64_t load_percent_x1000 = BAN::Math::div_round_up<uint64_t>(node->time_used_ns * 100'000, processing_ns);
 					dprintln("  tid { 2}: { 3}.{3}% active", node->thread->tid(), load_percent_x1000 / 1000, load_percent_x1000 % 1000);
-					return false;
-				}
+				}, const_cast<uint64_t*>(&processing_ns)
 			);
-			m_block_queue.remove_with_condition(
-				[&](SchedulerQueue::Node* node)
+			m_block_list.walk(
+				[](const SchedulerThreadNode* node, void* arg)
 				{
+					const uint64_t processing_ns = *static_cast<const uint64_t*>(arg);
 					const uint64_t load_percent_x1000 = BAN::Math::div_round_up<uint64_t>(node->time_used_ns * 100'000, processing_ns);
 					dprintln("  tid { 2}: { 3}.{3}% blocked", node->thread->tid(), load_percent_x1000 / 1000, load_percent_x1000 % 1000);
-					return false;
-				}
+				}, const_cast<uint64_t*>(&processing_ns)
 			);
 		}

@@ -532,7 +462,7 @@ namespace Kernel
 			auto& thread_info = m_most_loaded_threads[i];
 			if (thread_info.node == nullptr)
 				break;
-			if (thread_info.node == m_current || thread_info.queue == nullptr)
+			if (thread_info.node == m_current || thread_info.list == nullptr)
 				continue;

 			auto least_loaded_id = find_least_loaded_processor();
@@ -596,11 +526,11 @@ namespace Kernel

 			thread_info.node->time_used_ns = 0;

-			{
-				auto& my_queue = (thread_info.queue == &m_run_queue) ? m_run_queue : m_block_queue;
-				my_queue.remove_node(thread_info.node);
+			if (thread_info.list == &m_run_list)
+				m_run_list.pop(thread_info.node);
+			else
+				m_block_list.pop(thread_info.node);
 			m_thread_count--;
-			}

 			thread_info.node->processor_id = least_loaded_id;
 			Processor::send_smp_message(least_loaded_id, {
@@ -608,8 +538,10 @@ namespace Kernel
 				.new_thread = thread_info.node
 			});

-			thread_info.node = nullptr;
-			thread_info.queue = nullptr;
+			thread_info = {
+				.list = nullptr,
+				.node = nullptr,
+			};

 			if (m_idle_ns == 0)
 				break;
@@ -622,8 +554,8 @@ namespace Kernel
 			m_current->time_used_ns = 0;
 		for (auto& thread_info : m_most_loaded_threads)
 			thread_info = {};
-		m_run_queue  .remove_with_condition([&](SchedulerQueue::Node* node) { node->time_used_ns = 0; return false; });
-		m_block_queue.remove_with_condition([&](SchedulerQueue::Node* node) { node->time_used_ns = 0; return false; });
+		m_run_list  .walk([](const SchedulerThreadNode* node, void*) { const_cast<SchedulerThreadNode*>(node)->time_used_ns = 0; }, nullptr);
+		m_block_list.walk([](const SchedulerThreadNode* node, void*) { const_cast<SchedulerThreadNode*>(node)->time_used_ns = 0; }, nullptr);
 		m_idle_ns = 0;

 		m_should_calculate_max_load_threads = true;
@@ -634,7 +566,7 @@ namespace Kernel
 	BAN::ErrorOr<void> Scheduler::bind_thread_to_processor(Thread* thread, ProcessorID processor_id)
 	{
 		ASSERT(thread->m_scheduler_node == nullptr);
-		auto* new_node = new SchedulerQueue::Node(thread);
+		auto* new_node = new SchedulerThreadNode(thread);
 		if (new_node == nullptr)
 			return BAN::Error::from_errno(ENOMEM);

@@ -0,0 +1,266 @@
+#include <BAN/Assert.h>
+#include <BAN/Swap.h>
+#include <kernel/SchedulerThreadNode.h>
+
+namespace Kernel
+{
+
+	SchedulerThreadNode* SchedulerQueue::front()
+	{
+		return m_head;
+	}
+
+	SchedulerThreadNode* SchedulerQueue::pop_front()
+	{
+		if (empty())
+			return nullptr;
+		auto* const result = m_head;
+		m_head = m_head->queue.next;
+		(m_head ? m_head->queue.prev : m_tail) = nullptr;
+		result->queue.prev = nullptr;
+		result->queue.next = nullptr;
+		return result;
+	}
+
+	void SchedulerQueue::push(SchedulerThreadNode* node)
+	{
+		ASSERT(node->queue.prev == nullptr);
+		ASSERT(node->queue.next == nullptr);
+
+		node->queue.prev = m_tail;
+		node->queue.next = nullptr;
+		(m_tail ? m_tail->queue.next : m_head) = node;
+		m_tail = node;
+	}
+
+	void SchedulerQueue::pop(SchedulerThreadNode* node)
+	{
+		(node->queue.prev ? node->queue.prev->queue.next : m_head) = node->queue.next;
+		(node->queue.next ? node->queue.next->queue.prev : m_tail) = node->queue.prev;
+		node->queue.prev = nullptr;
+		node->queue.next = nullptr;
+	}
+
+	void SchedulerQueue::walk(void (*callback)(const SchedulerThreadNode*, void*), void* arg) const
+	{
+		for (auto* node = m_head; node; node = node->queue.next)
+			callback(node, arg);
+	}
+
+
+
+	SchedulerThreadNode* SchedulerHeap::front()
+	{
+		return m_root;
+	}
+
+	SchedulerThreadNode* SchedulerHeap::pop_front()
+	{
+		if (empty())
+			return nullptr;
+		auto* const result = m_root;
+		pop(result);
+		return result;
+	}
+
+	void SchedulerHeap::push(SchedulerThreadNode* node)
+	{
+		ASSERT(node->heap.parent == nullptr);
+		ASSERT(node->heap.lchild == nullptr);
+		ASSERT(node->heap.rchild == nullptr);
+
+		if (m_root == nullptr)
+		{
+			// push to empty heap
+			node->heap.parent = nullptr;
+			node->heap.lchild = nullptr;
+			node->heap.rchild = nullptr;
+			m_root = node;
+			m_last = node;
+			return;
+		}
+
+		auto* parent = m_last;
+
+		{
+			// find parent of the new node
+			SchedulerThreadNode* temp;
+			while ((temp = parent->heap.parent) && parent == temp->heap.rchild)
+				parent = temp;
+			if (temp && temp->heap.rchild == nullptr)
+				parent = temp;
+			else
+			{
+				if (temp != nullptr)
+					parent = temp->heap.rchild;
+				while ((temp = parent->heap.lchild))
+					parent = temp;
+			}
+		}
+
+		// insert node as the last node
+		(parent->heap.lchild ? parent->heap.rchild : parent->heap.lchild) = node;
+		node->heap.parent = parent;
+		node->heap.lchild = nullptr;
+		node->heap.rchild = nullptr;
+		m_last = node;
+
+		// fix heap properties
+		while ((parent = node->heap.parent) && node->wake_time_ns < parent->wake_time_ns)
+			swap_nodes(node, parent);
+	}
+
+	void SchedulerHeap::pop(SchedulerThreadNode* old_node)
+	{
+		if (m_root == m_last)
+		{
+			// remove the only node
+			old_node->heap.parent = nullptr;
+			old_node->heap.lchild = nullptr;
+			old_node->heap.rchild = nullptr;
+			m_root = nullptr;
+			m_last = nullptr;
+			return;
+		}
+
+		auto* fix_node = m_last;
+		swap_nodes(old_node, m_last);
+
+		{
+			// update last to point to the previous node
+			SchedulerThreadNode* temp;
+			while ((temp = m_last->heap.parent) && m_last == temp->heap.lchild)
+				m_last = temp;
+			if (temp != nullptr)
+				m_last = temp->heap.lchild;
+			ASSERT(m_last);
+			while ((temp = m_last->heap.rchild))
+				m_last = temp;
+		}
+
+		{
+			// delete links to/from the deleted node
+			if (auto* parent = old_node->heap.parent)
+				(old_node == parent->heap.rchild ? parent->heap.rchild : parent->heap.lchild) = nullptr;
+			old_node->heap.parent = nullptr;
+			old_node->heap.lchild = nullptr;
+			old_node->heap.rchild = nullptr;
+		}
+
+		// fix heap properties
+		if (fix_node->wake_time_ns == old_node->wake_time_ns)
+			;
+		else if (fix_node->wake_time_ns < old_node->wake_time_ns)
+		{
+			SchedulerThreadNode* parent;
+			while ((parent = fix_node->heap.parent) && fix_node->wake_time_ns < parent->wake_time_ns)
+				swap_nodes(fix_node, parent);
+		}
+		else for (;;)
+		{
+			const bool l_ok = !fix_node->heap.lchild || fix_node->wake_time_ns <= fix_node->heap.lchild->wake_time_ns;
+			const bool r_ok = !fix_node->heap.rchild || fix_node->wake_time_ns <= fix_node->heap.rchild->wake_time_ns;
+			if (l_ok && r_ok)
+				break;
+			auto* child = (!l_ok && !r_ok)
+				? (fix_node->heap.lchild->wake_time_ns < fix_node->heap.rchild->wake_time_ns ? fix_node->heap.lchild : fix_node->heap.rchild)
+				: (r_ok ? fix_node->heap.lchild : fix_node->heap.rchild);
+			swap_nodes(fix_node, child);
+		}
+	}
+
+	void SchedulerHeap::walk(void (*callback)(const SchedulerThreadNode*, void*), void* arg) const
+	{
+		walk_impl(callback, arg, m_root);
+	}
+
+	void SchedulerHeap::walk_impl(void (*callback)(const SchedulerThreadNode*, void*), void* arg, const SchedulerThreadNode* node) const
+	{
+		if (node == nullptr)
+			return;
+		callback(node, arg);
+		walk_impl(callback, arg, node->heap.lchild);
+		walk_impl(callback, arg, node->heap.rchild);
+	}
+
+	void SchedulerHeap::swap_nodes(SchedulerThreadNode* node1, SchedulerThreadNode* node2)
+	{
+		if (node1 == node2)
+			return;
+
+		if (node2 == node1->heap.parent)
+			BAN::swap(node1, node2);
+
+		auto* const p1 = node1->heap.parent;
+		auto* const l1 = node1->heap.lchild;
+		auto* const r1 = node1->heap.rchild;
+
+		auto* const p2 = node2->heap.parent;
+		auto* const l2 = node2->heap.lchild;
+		auto* const r2 = node2->heap.rchild;
+
+		if (node1 == node2->heap.parent)
+		{
+			node1->heap.parent = node2;
+			node1->heap.lchild = l2;
+			node1->heap.rchild = r2;
+
+			node2->heap.parent = p1;
+
+			if (l1 == node2)
+			{
+				node2->heap.lchild = node1;
+				node2->heap.rchild = r1;
+				if (r1) r1->heap.parent = node2;
+			}
+			else
+			{
+				node2->heap.lchild = l1;
+				node2->heap.rchild = node1;
+				if (l1) l1->heap.parent = node2;
+			}
+
+			if (p1) (node1 == p1->heap.lchild ? p1->heap.lchild : p1->heap.rchild) = node2;
+
+			if (l2) l2->heap.parent = node1;
+			if (r2) r2->heap.parent = node1;
+		}
+		else
+		{
+			node1->heap.parent = p2;
+			node1->heap.lchild = l2;
+			node1->heap.rchild = r2;
+
+			node2->heap.parent = p1;
+			node2->heap.lchild = l1;
+			node2->heap.rchild = r1;
+
+			if (l1) l1->heap.parent = node2;
+			if (r1) r1->heap.parent = node2;
+
+			if (l2) l2->heap.parent = node1;
+			if (r2) r2->heap.parent = node1;
+
+			if (p1 || p2)
+			{
+				if (p1 == p2)
+					BAN::swap(p1->heap.lchild, p1->heap.rchild);
+				else
+				{
+					if (p1) (p1->heap.lchild == node1 ? p1->heap.lchild : p1->heap.rchild) = node2;
+					if (p2) (p2->heap.lchild == node2 ? p2->heap.lchild : p2->heap.rchild) = node1;
+				}
+			}
+		}
+
+		auto* const root = m_root;
+		auto* const last = m_last;
+
+		if      (node1 == root) m_root = node2;
+		else if (node1 == last) m_last = node2;
+
+		if      (node2 == root) m_root = node1;
+		else if (node2 == last) m_last = node1;
+	}
+
+}
@@ -94,7 +94,7 @@ namespace Kernel

 			while (is != 0)
 			{
-				const size_t idx = __builtin_ctz(is);
+				const size_t idx = BAN::Math::ctz(is);
 				if (auto& device = m_devices[idx])
 					device->handle_irq();
 				else
@@ -1,3 +1,4 @@
+#include <kernel/Lock/BlockableSpinLock.h>
 #include <kernel/Lock/LockGuard.h>
 #include <kernel/Scheduler.h>
 #include <kernel/Storage/ATA/AHCI/Controller.h>
@@ -301,7 +302,7 @@ namespace Kernel
 			const vaddr_t buffer_vaddr = reinterpret_cast<vaddr_t>(buffer.data());
 			const paddr_t buffer_paddr = to_paddr(buffer_vaddr);

-			const size_t bytes = BAN::Math::min(buffer.size(), PAGE_SIZE - buffer_vaddr % PAGE_SIZE);
+			const size_t bytes = BAN::Math::min<size_t>(buffer.size(), PAGE_SIZE - buffer_vaddr % PAGE_SIZE);

 			bool can_extend = true;
 			if (prdt_count == 0)
@@ -424,7 +425,7 @@ namespace Kernel
 		{
 			if (const uint32_t usable_slots = ~(m_port->sact | m_port->ci) & m_free_slots)
 			{
-				const uint32_t slot = __builtin_ctz(usable_slots);
+				const uint32_t slot = BAN::Math::ctz(usable_slots);
 				m_free_slots &= ~(1u << slot);
 				return slot;
 			}
@@ -43,6 +43,8 @@ namespace Kernel

 	extern "C" long cpp_syscall_handler(int syscall, uintptr_t arg1, uintptr_t arg2, uintptr_t arg3, uintptr_t arg4, uintptr_t arg5)
 	{
+		Thread::current().set_is_in_syscall(true);
+
 		Processor::set_interrupt_state(InterruptState::Enabled);

 		Process::current().wait_while_stopped();
@@ -101,6 +103,8 @@ namespace Kernel

 		Processor::set_interrupt_state(InterruptState::Disabled);

+		Thread::current().set_is_in_syscall(false);
+
 		ASSERT(Kernel::Thread::current().state() == Kernel::Thread::State::Executing);

 		if (ret.is_error())
@@ -2,6 +2,7 @@
 #include <kernel/FS/DevFS/FileSystem.h>
 #include <kernel/Lock/BlockableSpinLock.h>
 #include <kernel/Terminal/PseudoTerminal.h>
+#include <kernel/Thread.h>

 #include <BAN/ScopeGuard.h>

@@ -237,8 +237,6 @@ namespace Kernel
 		if (ch == _POSIX_VDISABLE)
 			return;

-		LockGuard _(m_mutex);
-
 		const auto termios = get_termios();

 		if ((termios.c_iflag & ISTRIP))
@@ -271,6 +269,8 @@ namespace Kernel
 		bool should_flush = false;
 		bool force_echo = false;

+		LockGuard _(m_mutex);
+
 		if (!(termios.c_lflag & ICANON))
 			should_flush = true;
 		else
@@ -47,9 +47,8 @@ namespace Kernel
 		  }, 0600, 0, 0)
 		, m_name(MUST(BAN::String::formatted("tty{}", s_next_tty_number++)))
 		, m_terminal_driver(driver)
-		, m_palette(driver->palette())
-		, m_foreground(m_palette[15])
-		, m_background(m_palette[0])
+		, m_foreground(driver->palette()[15])
+		, m_background(driver->palette()[0])
 	{
 		m_width = m_terminal_driver->width();
 		m_height = m_terminal_driver->height();
@@ -122,11 +121,14 @@ namespace Kernel
 	void VirtualTTY::handle_ansi_csi_color(uint8_t value)
 	{
 		ASSERT(m_write_lock.is_locked_by_current_thread());
+
+		auto& palette = m_terminal_driver->palette();
+
 		switch (value)
 		{
 			case 0:
-				m_foreground = m_palette[15];
-				m_background = m_palette[0];
+				m_foreground = palette[15];
+				m_background = palette[0];
 				m_colors_inverted = false;
 				break;

@@ -138,25 +140,25 @@ namespace Kernel
 			case 27: m_colors_inverted = false; break;

 			case 30 ... 37:
-				m_foreground = m_palette[value - 30];
+				m_foreground = palette[value - 30];
 				break;
 			case 39:
-				m_foreground = m_palette[15];
+				m_foreground = palette[15];
 				break;

 			case 40 ... 47:
-				m_background = m_palette[value - 40];
+				m_background = palette[value - 40];
 				break;
 			case 49:
-				m_background = m_palette[0];
+				m_background = palette[0];
 				break;

 			case 90 ... 97:
-				m_foreground = m_palette[value - 90 + 8];
+				m_foreground = palette[value - 90 + 8];
 				break;

 			case 100 ... 107:
-				m_background = m_palette[value - 100 + 8];
+				m_background = palette[value - 100 + 8];
 				break;

 			default:
@@ -173,7 +175,7 @@ namespace Kernel

 		const uint8_t code = BAN::Math::min(m_ansi_state.nums[2], 255);
 		if (code < 16)
-			return m_palette[code];
+			return m_terminal_driver->palette()[code];

 		if (code < 232)
 		{
@@ -285,12 +285,22 @@ namespace Kernel
 		}
 	}

-	uint64_t Thread::cpu_time_ns() const
+	uint64_t Thread::cpu_time_total_ns() const
 	{
 		SpinLockGuard _(m_cpu_time_lock);
-		if (m_cpu_time_start_ns == UINT64_MAX)
-			return m_cpu_time_ns;
-		return m_cpu_time_ns + (SystemTimer::get().ns_since_boot() - m_cpu_time_start_ns);
+		uint64_t offset = 0;
+		if (m_cpu_time_start_ns != UINT64_MAX)
+			offset += SystemTimer::get().ns_since_boot() - m_cpu_time_start_ns;
+		return m_cpu_time_user_ns + m_cpu_time_system_ns + offset;
+	}
+
+	void Thread::cpu_time_ns(uint64_t& user_ns, uint64_t& system_ns) const
+	{
+		SpinLockGuard _(m_cpu_time_lock);
+		user_ns = m_cpu_time_user_ns;
+		system_ns = m_cpu_time_system_ns;
+		if (m_cpu_time_start_ns != UINT64_MAX)
+			system_ns += SystemTimer::get().ns_since_boot() - m_cpu_time_start_ns;
 	}

 	void Thread::set_cpu_time_start()
@@ -304,10 +314,21 @@ namespace Kernel
 	{
 		SpinLockGuard _(m_cpu_time_lock);
 		ASSERT(m_cpu_time_start_ns != UINT64_MAX);
-		m_cpu_time_ns += SystemTimer::get().ns_since_boot() - m_cpu_time_start_ns;
+		uint64_t& value = m_is_in_syscall ? m_cpu_time_system_ns : m_cpu_time_user_ns;
+		value += SystemTimer::get().ns_since_boot() - m_cpu_time_start_ns;
 		m_cpu_time_start_ns = UINT64_MAX;
 	}

+	void Thread::set_is_in_syscall(bool is_in_syscall)
+	{
+		SpinLockGuard _(m_cpu_time_lock);
+		const uint64_t current_ns = SystemTimer::get().ns_since_boot();
+		uint64_t& value = m_is_in_syscall ? m_cpu_time_system_ns : m_cpu_time_user_ns;
+		value += current_ns - m_cpu_time_start_ns;
+		m_is_in_syscall = is_in_syscall;
+		m_cpu_time_start_ns = current_ns;
+	}
+
 	void Thread::update_processor_index_address()
 	{
 		if (!is_userspace() || !has_process())
@@ -1,5 +1,6 @@
 #include <kernel/Processor.h>
-#include <kernel/SchedulerQueueNode.h>
+#include <kernel/Scheduler.h>
+#include <kernel/SchedulerThreadNode.h>
 #include <kernel/ThreadBlocker.h>
 #include <kernel/Timer/Timer.h>

@@ -44,7 +45,7 @@ namespace Kernel
 		m_block_chain = nullptr;
 	}

-	void ThreadBlocker::add_thread_to_block_queue(SchedulerQueue::Node* node)
+	void ThreadBlocker::add_thread_to_block_queue(SchedulerThreadNode* node)
 	{
 		SpinLockGuard _(m_lock);

@@ -60,7 +61,7 @@ namespace Kernel
 		m_block_chain = node;
 	}

-	void ThreadBlocker::remove_thread_from_block_queue(SchedulerQueue::Node* node)
+	void ThreadBlocker::remove_thread_from_block_queue(SchedulerThreadNode* node)
 	{
 		SpinLockGuard _(m_lock);

@@ -6,6 +6,7 @@
 #include <kernel/Memory/PageTable.h>
 #include <kernel/MMIO.h>
 #include <kernel/Processor.h>
+#include <kernel/Scheduler.h>
 #include <kernel/Timer/HPET.h>

 #define HPET_PERIOD_MAX 0x05F5E100
@@ -135,8 +136,7 @@ namespace Kernel
 		auto* header = static_cast<const ACPI::HPET*>(ACPI::ACPI::get().get_header("HPET"_sv, 0));
 		if (header == nullptr)
 			return BAN::Error::from_errno(ENODEV);
-
-		if (header->hardware_rev_id == 0)
+		if (header->length < sizeof(ACPI::HPET))
 			return BAN::Error::from_errno(EINVAL);

 		if (!InterruptController::get().is_using_apic() && !header->legacy_replacement_irq_routing_cable)
@@ -152,6 +152,12 @@ namespace Kernel

 		auto& regs = registers();

+		if ((regs.capabilities & 0xFF) == 0)
+		{
+			dwarnln("HPET: Invalid revision");
+			return BAN::Error::from_errno(EINVAL);
+		}
+
 #if ARCH(x86_64)
 		m_is_64bit = regs.capabilities & COUNT_SIZE_CAP;
 #else
@@ -191,17 +197,36 @@ namespace Kernel
 			return BAN::Error::from_errno(ENOTSUP);
 		}

+		// set timer period to 1000 Hz
+		const uint64_t ticks_per_ms = m_ticks_per_s / 1000;
+
+#if ARCH(x86_64)
+		const bool use_64bit_timer = !!(timer0.configuration & Tn_SIZE_CAP);
+#else
+		// Spec recommends using 32 bit timers when the cpu can't do atomic 64 bit accesses
+		const bool use_64bit_timer = !!(timer0.configuration & Tn_SIZE_CAP) && (ticks_per_ms > 0xFFFFFFFF);
+#endif
+
+		if (!use_64bit_timer && ticks_per_ms >= 0x100000000)
+		{
+			dprintln("HPET: cannot create 1 kHz timer");
+			return BAN::Error::from_errno(ENOTSUP);
+		}
+
 		// enable interrupts
-		timer0.configuration = timer0.configuration |  Tn_INT_ENB_CNF;
+		timer0.configuration |=  Tn_INT_ENB_CNF;
 		// edge triggered interrupts
-		timer0.configuration = timer0.configuration & ~Tn_INT_TYPE_CNF;
+		timer0.configuration &= ~Tn_INT_TYPE_CNF;
 		// periodic timer
-		timer0.configuration = timer0.configuration |  Tn_TYPE_CNF;
-		// disable 32 bit mode
-		timer0.configuration = timer0.configuration & ~Tn_32MODE_CNF;
+		timer0.configuration |=  Tn_TYPE_CNF;
+		// set 32 bit mode
+		if (use_64bit_timer)
+			timer0.configuration &= ~Tn_32MODE_CNF;
+		else
+			timer0.configuration |=  Tn_32MODE_CNF;
 		// disable FSB interrupts
 		if (timer0.configuration & Tn_FSB_INT_DEL_CAP)
-			timer0.configuration = timer0.configuration & ~Tn_FSB_EN_CNF;
+			timer0.configuration &= ~Tn_FSB_EN_CNF;

 		uint16_t irq;
 		if (header->legacy_replacement_irq_routing_cable)
@@ -238,27 +263,29 @@ namespace Kernel
 			timer0.configuration = (timer0.configuration & ~Tn_INT_ROUTE_CNF_MASK) | (gsi << Tn_INT_ROUTE_CNF_SHIFT);
 		}

-		// set timer period to 1000 Hz
-		const uint64_t ticks_per_ms = m_ticks_per_s / 1000;
+		if (use_64bit_timer)
+		{
+#if ARCH(x86_64)
+			timer0.configuration = timer0.configuration | Tn_VAL_SET_CNF;
+			timer0.comparator.full = ticks_per_ms;
+#else
 			timer0.configuration   = timer0.configuration | Tn_VAL_SET_CNF;
 			timer0.comparator.low  = ticks_per_ms;
-		if (timer0.configuration & Tn_SIZE_CAP)
-		{
 			timer0.configuration   = timer0.configuration | Tn_VAL_SET_CNF;
 			timer0.comparator.high = ticks_per_ms >> 32;
+#endif
 		}
-		else if (ticks_per_ms > 0xFFFFFFFF)
+		else
 		{
-			dprintln("HPET: cannot create 1 kHz timer");
-			return BAN::Error::from_errno(ENOTSUP);
+			timer0.configuration = timer0.configuration | Tn_VAL_SET_CNF;
+			timer0.comparator.low = ticks_per_ms;
 		}

-		// enable main counter
-		regs.configuration.low = regs.configuration.low | ENABLE_CNF;
-
 		set_irq(irq);
 		InterruptController::get().enable_irq(irq);

+		regs.configuration.low = regs.configuration.low | ENABLE_CNF;
+
 		return {};
 	}

@@ -2,6 +2,7 @@
 #include <kernel/InterruptController.h>
 #include <kernel/IO.h>
 #include <kernel/Processor.h>
+#include <kernel/Scheduler.h>
 #include <kernel/Timer/PIT.h>

 #define PIT_IRQ 0
@@ -252,14 +252,19 @@ namespace Kernel
 					dwarnln("Could not initialize USB interface {}", ret.error());
 					m_class_drivers.remove(i--);
 				}
+				else if (m_class_drivers[i]->is_hid_driver() && static_cast<USBHIDDriver*>(m_class_drivers[i].ptr())->has_led_control())
+					TRY(m_hid_info.led_controls.push_back(m_class_drivers[i].ptr()));
 			}

 			if (!m_class_drivers.empty())
 			{
+				update_led_mask(m_hid_info.led_mask);
+
 				dprintln("Successfully initialized USB device with {}/{} interfaces",
 					m_class_drivers.size(),
 					configuration.interfaces.size()
 				);
+
 				return {};
 			}
 		}
@@ -363,4 +368,13 @@ namespace Kernel
 			driver->handle_input_data(byte_count, endpoint_id);
 	}

+	void USBDevice::update_led_mask(uint32_t led_mask)
+	{
+		const uint32_t old_mask = m_hid_info.led_mask.exchange(led_mask);
+		if (old_mask == led_mask)
+			return;
+		for (auto* led_control : m_hid_info.led_controls)
+			static_cast<USBHIDDriver*>(led_control)->set_leds(led_mask);
+	}
+
 }
@@ -14,6 +14,7 @@ namespace Kernel

 	enum HIDRequest : uint8_t
 	{
+		GET_PROTOCOL = 0x03,
 		SET_PROTOCOL = 0x0B,
 	};

@@ -77,14 +78,14 @@ namespace Kernel

 	USBHIDDriver::~USBHIDDriver()
 	{
-		for (auto& device_input : m_device_inputs)
-			if (device_input.device)
-				DevFileSystem::get().remove_device(device_input.device);
+		for (auto& device_report : m_device_reports)
+			if (device_report.device)
+				DevFileSystem::get().remove_device(device_report.device);
 	}

 	BAN::ErrorOr<void> USBHIDDriver::initialize()
 	{
-		auto dma_buffer = TRY(DMARegion::create(1024));
+		auto dma_buffer = TRY(DMARegion::create(1024, PageTable::MemoryType::Normal));

 		ASSERT(static_cast<USB::InterfaceBaseClass>(m_interface.descriptor.bInterfaceClass) == USB::InterfaceBaseClass::HID);

@@ -119,13 +120,14 @@ namespace Kernel
 		// If this device supports boot protocol, make sure it is not used
 		if (m_interface.descriptor.bInterfaceSubClass == 0x01)
 		{
-			USBDeviceRequest request;
-			request.bmRequestType = USB::RequestType::HostToDevice | USB::RequestType::Class | USB::RequestType::Interface;
-			request.bRequest      = HIDRequest::SET_PROTOCOL;
-			request.wValue        = 1; // report protocol
-			request.wIndex        = m_interface.descriptor.bInterfaceNumber;
-			request.wLength       = 0;
-			TRY(m_device.send_request(request, 0));
+			const auto request = USBDeviceRequest {
+				.bmRequestType = USB::RequestType::HostToDevice | USB::RequestType::Class | USB::RequestType::Interface,
+				.bRequest      = HIDRequest::SET_PROTOCOL,
+				.wValue        = 1, // report protocol
+				.wIndex        = m_interface.descriptor.bInterfaceNumber,
+				.wLength       = 0,
+			};
+			(void)m_device.send_request(request, 0);
 		}

 		const auto& hid_descriptor = *reinterpret_cast<const HIDDescriptor*>(m_interface.misc_descriptors[hid_descriptor_index].data());
@@ -184,7 +186,10 @@ namespace Kernel
 			return BAN::Error::from_errno(EFAULT);
 		}

-		m_device_inputs = TRY(initializes_device_reports(collections));
+		TRY(initializes_device_reports(collections));
+
+		if (has_led_control())
+			m_led_region = TRY(DMARegion::create(PAGE_SIZE, PageTable::MemoryType::Normal));

 		for (const auto& endpoint : m_interface.endpoints)
 		{
@@ -196,7 +201,7 @@ namespace Kernel
 				continue;

 			TRY(m_device.configure_endpoint(desc));
-			m_data_buffer = TRY(DMARegion::create(desc.wMaxPacketSize & 0x07FF));
+			m_data_buffer = TRY(DMARegion::create(desc.wMaxPacketSize & 0x07FF, PageTable::MemoryType::Normal));

 			m_data_endpoint_id = (desc.bEndpointAddress & 0x0F) * 2 + !!(desc.bEndpointAddress & 0x80);

@@ -209,7 +214,7 @@ namespace Kernel
 			return BAN::Error::from_errno(EINVAL);
 		}

-		for (auto& report : m_device_inputs)
+		for (auto& report : m_device_reports)
 			if (report.device && report.device->initialize().is_error())
 				report.device.clear();

@@ -238,10 +243,9 @@ namespace Kernel
 		return {};
 	}

-	BAN::ErrorOr<BAN::Vector<USBHIDDriver::DeviceReport>> USBHIDDriver::initializes_device_reports(const BAN::Vector<USBHID::Collection>& collection_list)
+	BAN::ErrorOr<void> USBHIDDriver::initializes_device_reports(const BAN::Vector<USBHID::Collection>& collection_list)
 	{
-		BAN::Vector<USBHIDDriver::DeviceReport> result;
-		TRY(result.reserve(collection_list.size()));
+		TRY(m_device_reports.reserve(collection_list.size()));

 		for (size_t i = 0; i < collection_list.size(); i++)
 		{
@@ -249,9 +253,7 @@ namespace Kernel

 			USBHIDDriver::DeviceReport report;
 			TRY(gather_collection_reports(collection, report.inputs,  USBHID::Report::Type::Input));
-
-			BAN::Vector<USBHID::Report> outputs;
-			TRY(gather_collection_reports(collection, outputs, USBHID::Report::Type::Output));
+			TRY(gather_collection_reports(collection, report.outputs, USBHID::Report::Type::Output));

 			switch (collection.usage_page)
 			{
@@ -267,7 +269,7 @@ namespace Kernel
 						dprintln("Initialized an USB Joystick");
 						break;
 					case 0x06:
-						report.device = TRY(BAN::RefPtr<USBKeyboard>::create(*this, BAN::move(outputs)));
+						report.device = TRY(BAN::RefPtr<USBKeyboard>::create(*this));
 						dprintln("Initialized an USB Keyboard");
 						break;
 					default:
@@ -279,19 +281,47 @@ namespace Kernel
 				switch (collection.usage_id)
 				{
 					case 0x01:
-						report.device = TRY(BAN::RefPtr<USBKeyboard>::create(*this, BAN::move(outputs)));
+						report.device = TRY(BAN::RefPtr<USBKeyboard>::create(*this));
 						dprintln("Initialized an USB Consumer Control");
 						break;
 				}
 			}

-			TRY(result.push_back(BAN::move(report)));
+			TRY(m_device_reports.push_back(BAN::move(report)));
 		}

-		for (auto& report : result)
-			if (report.device)
+		for (auto& report : m_device_reports)
+		{
+			if (!report.device)
+				continue;
 			DevFileSystem::get().add_device(report.device);
-		return BAN::move(result);
+
+			uint8_t led_report_ids[0x100 / 8] {};
+			for (const auto& output : report.outputs)
+			{
+				if (output.usage_page != 0x08)
+					continue;
+
+				const auto byte = output.report_id / 8;
+				const auto bit  = output.report_id % 8;
+				if (led_report_ids[byte] & (1u << bit))
+					continue;
+				led_report_ids[byte] |= (1u << bit);
+
+				uint32_t report_bits = 0;
+				for (const auto& temp : report.outputs)
+					if (temp.report_id == output.report_id)
+						report_bits += temp.report_size * temp.report_count;
+
+				TRY(m_led_controls.push_back({
+					.report = &m_device_reports.back(),
+					.report_id = output.report_id,
+					.report_bits = report_bits,
+				}));
+			}
+		}
+
+		return {};
 	}

 	void USBHIDDriver::handle_stall(uint8_t endpoint_id)
@@ -368,19 +398,19 @@ namespace Kernel
 		}

 		size_t bit_offset = 0;
-		for (auto& device_input : m_device_inputs)
+		for (auto& device_report : m_device_reports)
 		{
-			if (device_input.device)
-				device_input.device->start_report();
+			if (device_report.device)
+				device_report.device->start_report();

-			for (const auto& input : device_input.inputs)
+			for (const auto& input : device_report.inputs)
 			{
 				if (report_id.value_or(input.report_id) != input.report_id)
 					continue;

 				ASSERT(input.report_size <= 32);

-				if (!device_input.device || (input.usage_id == 0 && input.usage_minimum == 0 && input.usage_maximum == 0))
+				if (!device_report.device || (input.usage_id == 0 && input.usage_minimum == 0 && input.usage_maximum == 0))
 				{
 					bit_offset += input.report_size * input.report_count;
 					continue;
@@ -401,7 +431,7 @@ namespace Kernel
 					const auto usage = input.usage_id ? input.usage_id : (input.usage_minimum + (variable ? i : logical));

 					if (!variable)
-						device_input.device->handle_array(input.usage_page, usage);
+						device_report.device->handle_array(input.usage_page, usage);
 					else
 					{
 						const int64_t physical =
@@ -411,17 +441,51 @@ namespace Kernel
 							input.physical_minimum;

 						if (relative)
-							device_input.device->handle_variable(input.usage_page, usage, physical);
+							device_report.device->handle_variable(input.usage_page, usage, physical);
 						else
-							device_input.device->handle_variable_absolute(input.usage_page, usage, physical, input.physical_minimum, input.physical_maximum);
+							device_report.device->handle_variable_absolute(input.usage_page, usage, physical, input.physical_minimum, input.physical_maximum);
 					}

 					bit_offset += input.report_size;
 				}
 			}

-			if (device_input.device)
-				device_input.device->stop_report();
+			if (device_report.device)
+				device_report.device->stop_report();
+		}
+	}
+
+	void USBHIDDriver::set_leds(uint32_t led_mask)
+	{
+		for (const auto& led_control : m_led_controls)
+		{
+			const size_t report_bytes = BAN::Math::div_round_up<uint32_t>(led_control.report_bits, 8);
+
+			auto led_data = BAN::ByteSpan(reinterpret_cast<uint8_t*>(m_led_region->vaddr()), report_bytes);
+			memset(led_data.data(), 0, report_bytes);
+
+			size_t bit_offset = 0;
+			for (const auto& output : led_control.report->outputs)
+			{
+				if (output.report_id != led_control.report_id)
+					continue;
+
+				const size_t usage_base = output.usage_id ? output.usage_id : output.usage_minimum;
+				for (size_t i = 0; output.report_size == 1 && i < output.report_count; i++, bit_offset++)
+					if (led_mask & (1u << (usage_base + bit_offset)))
+						led_data[bit_offset / 8] |= 1u << (bit_offset % 8);
+
+				bit_offset += output.report_size * output.report_count;
+			}
+
+			USBDeviceRequest request;
+			request.bmRequestType = USB::RequestType::HostToDevice | USB::RequestType::Class | USB::RequestType::Interface;
+			request.bRequest = 0x09;
+			request.wValue = 0x0200 | led_control.report_id;
+			request.wIndex = m_interface.descriptor.bInterfaceNumber;
+			request.wLength = report_bytes;
+			if (auto ret = m_device.send_request(request, m_led_region->paddr()); ret.is_error())
+				dprintln_if(DEBUG_USB_HID, "Failed to update LEDs: {}", ret.error());
 		}
 	}

@@ -14,19 +14,10 @@ namespace Kernel
 	static void initialize_scancode_to_keycode();
 	static constexpr bool is_repeatable_scancode(uint8_t scancode);

-	USBKeyboard::USBKeyboard(USBHIDDriver& driver, BAN::Vector<USBHID::Report>&& outputs)
+	USBKeyboard::USBKeyboard(USBHIDDriver& driver)
 		: USBHIDDevice(InputDevice::Type::Keyboard)
 		, m_driver(driver)
-		, m_outputs(BAN::move(outputs))
-	{
-		set_leds(0);
-	}
-
-	BAN::ErrorOr<void> USBKeyboard::initialize()
-	{
-		m_led_region = TRY(DMARegion::create(PAGE_SIZE, PageTable::MemoryType::Normal));
-		return {};
-	}
+	{ }

 	void USBKeyboard::start_report()
 	{
@@ -114,32 +105,6 @@ namespace Kernel
 	{
 		ASSERT(m_keyboard_lock.current_processor_has_lock());

-		if (usage_page != 0x07)
-		{
-			dprintln_if(DEBUG_USB_KEYBOARD, "Unsupported keyboard usage page {2H}", usage_page);
-			return;
-		}
-		if (usage >= 4 && usage < m_keyboard_state_temp.size())
-			m_keyboard_state_temp[usage] = true;
-	}
-
-	void USBKeyboard::handle_variable(uint16_t usage_page, uint16_t usage, int64_t state)
-	{
-		(void)usage_page;
-		(void)usage;
-		(void)state;
-		dprintln_if(DEBUG_USB_KEYBOARD, "Unsupported keyboard relative usage page {2H}", usage_page);
-	}
-
-	void USBKeyboard::handle_variable_absolute(uint16_t usage_page, uint16_t usage, int64_t state, int64_t min, int64_t max)
-	{
-		(void)min; (void)max;
-
-		ASSERT(m_keyboard_lock.current_processor_has_lock());
-
-		if (state == 0)
-			return;
-
 		switch (usage_page)
 		{
 			case 0x07:
@@ -149,6 +114,11 @@ namespace Kernel
 			case 0x0C:
 				switch (usage)
 				{
+					case 0x00:
+						break;
+					case 0xE2:
+						m_keyboard_state_temp[0x7F] = true;
+						break;
 					case 0xE9:
 						m_keyboard_state_temp[0x80] = true;
 						break;
@@ -166,6 +136,24 @@ namespace Kernel
 		}
 	}

+	void USBKeyboard::handle_variable(uint16_t usage_page, uint16_t usage, int64_t state)
+	{
+		(void)usage_page;
+		(void)usage;
+		(void)state;
+		dprintln_if(DEBUG_USB_KEYBOARD, "Unsupported keyboard relative usage page {2H}", usage_page);
+	}
+
+	void USBKeyboard::handle_variable_absolute(uint16_t usage_page, uint16_t usage, int64_t state, int64_t min, int64_t max)
+	{
+		(void)min; (void)max;
+
+		if (state == 0)
+			return;
+
+		handle_array(usage_page, usage);
+	}
+
 	void USBKeyboard::update()
 	{
 		using KeyModifier = LibInput::KeyEvent::Modifier;
@@ -173,8 +161,19 @@ namespace Kernel
 		const auto toggle_mask = ({ SpinLockGuard _(m_keyboard_lock); m_toggle_mask; });

 		if (m_led_mask != toggle_mask)
-			set_leds(toggle_mask);
+		{
+			uint32_t usb_led_mask { 0 };
+			if (toggle_mask & KeyModifier::NumLock)
+				usb_led_mask |= 1 << 1;
+			if (toggle_mask & KeyModifier::CapsLock)
+				usb_led_mask |= 1 << 2;
+			if (toggle_mask & KeyModifier::ScrollLock)
+				usb_led_mask |= 1 << 3;
+
+			m_driver.device().update_led_mask(usb_led_mask);
+
 			m_led_mask = toggle_mask;
+		}

 		SpinLockGuard _(m_keyboard_lock);

@@ -193,88 +192,6 @@ namespace Kernel
 		m_next_repeat_event_ms += s_repeat_interval_ms;
 	}

-	void USBKeyboard::set_leds(uint16_t mask)
-	{
-		uint8_t report_ids_done[0x100 / 8] {};
-
-		for (const auto& report : m_outputs)
-		{
-			if (report.usage_page != 0x08)
-				continue;
-
-			const auto byte = report.report_id / 8;
-			const auto bit  = report.report_id % 8;
-			if (report_ids_done[byte] & (1u << bit))
-				continue;
-
-			set_leds(report.report_id, mask);
-			report_ids_done[byte] |= (1u << bit);
-		}
-	}
-
-	void USBKeyboard::set_leds(uint8_t report_id, uint16_t mask)
-	{
-		using KeyModifier = LibInput::KeyEvent::Modifier;
-
-		if (!m_led_region)
-			return;
-
-		size_t report_bits = 0;
-		for (const auto& report : m_outputs)
-		{
-			if (report.report_id != report_id)
-				continue;
-			report_bits += report.report_size * report.report_count;
-		}
-
-		const size_t report_bytes = (report_bits + 7) / 8;
-		ASSERT(report_bytes <= PAGE_SIZE);
-
-		PageTable::with_fast_page(m_led_region->paddr(), [&] {
-			uint8_t* data = &PageTable::fast_page_as<uint8_t>();
-
-			memset(data, 0, report_bytes);
-
-			size_t bit_offset = 0;
-			for (const auto& report : m_outputs)
-			{
-				if (report.report_id != report_id)
-					continue;
-
-				for (size_t i = 0; report.report_size == 1 && i < report.report_count; i++, bit_offset++)
-				{
-					const size_t usage = (report.usage_id ? report.usage_id : report.usage_minimum) + bit_offset;
-					switch (usage)
-					{
-						case 0x01:
-							if (mask & KeyModifier::NumLock)
-								data[bit_offset / 8] |= 1u << (bit_offset % 8);
-							break;
-						case 0x02:
-							if (mask & KeyModifier::CapsLock)
-								data[bit_offset / 8] |= 1u << (bit_offset % 8);
-							break;
-						case 0x03:
-							if (mask & KeyModifier::ScrollLock)
-								data[bit_offset / 8] |= 1u << (bit_offset % 8);
-							break;
-					}
-				}
-
-				bit_offset += report.report_size * report.report_count;
-			}
-		});
-
-		USBDeviceRequest request;
-		request.bmRequestType = USB::RequestType::HostToDevice | USB::RequestType::Class | USB::RequestType::Interface;
-		request.bRequest = 0x09;
-		request.wValue = 0x0200 | report_id;
-		request.wIndex = m_driver.interface().descriptor.bInterfaceNumber;
-		request.wLength = report_bytes;
-		if (auto ret = m_driver.device().send_request(request, m_led_region->paddr()); ret.is_error())
-			dprintln_if(DEBUG_USB_KEYBOARD, "Failed to update LEDs: {}", ret.error());
-	}
-
 	void initialize_scancode_to_keycode()
 	{
 		using LibInput::keycode_function;
@@ -324,6 +241,7 @@ namespace Kernel
 		s_scancode_to_keycode[0x33] = keycode_normal(2, 10);
 		s_scancode_to_keycode[0x34] = keycode_normal(2, 11);
 		s_scancode_to_keycode[0x31] = keycode_normal(2, 12);
+		s_scancode_to_keycode[0x32] = keycode_normal(2, 12);
 		s_scancode_to_keycode[0x28] = keycode_normal(2, 13);
 		s_scancode_to_keycode[0xE1] = keycode_normal(3,  0);
 		s_scancode_to_keycode[0x64] = keycode_normal(3,  1);
@@ -1,7 +1,9 @@
 #include <BAN/ScopeGuard.h>
+#include <kernel/Scheduler.h>
+#include <kernel/Thread.h>
+#include <kernel/Timer/Timer.h>
 #include <kernel/USB/Hub/Definitions.h>
 #include <kernel/USB/Hub/HubDriver.h>
-#include <kernel/Timer/Timer.h>

 namespace Kernel
 {
@@ -192,7 +194,7 @@ namespace Kernel
 			auto result = m_device.send_request(request, m_data_region->paddr());
 			if (result.is_error() || result.value() != sizeof(USBHub::PortStatus))
 			{
-				dwarnln("Failed to get port {} status");
+				dwarnln("Failed to get port {} status", port_id);
 				continue;
 			}

--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Bananymous	315d8895a7	WindowServer: Start TaskBar and Terminal after starting up This removes the race condition from `start-gui` not starting WindowServer fast enough and leading to the apps crashing as they timeout while connecting to the server	2026-08-22 16:05:54 +03:00
Bananymous	4e9cfdde97	Kernel: Fix 32 bit target compilation banos is written with only 64 bit in mind so it has to be disabled	2026-08-22 15:33:49 +03:00
Bananymous	645b6d772d	Userspace: Use BAN::Math instead of builtins	2026-08-22 14:45:23 +03:00
Bananymous	ae24fb1dd0	Kernel: Use BAN::Math::{ctz,popcount} instead of builtins	2026-08-22 14:39:42 +03:00
Bananymous	fe64639842	Kernel/LibC: Make ino_t 32 bit integer 32 bit target could not use atomic ino_t when it was 32 bit. There isn't really any reason to have 64 bit ino_t as there aren't really filesystems with that large numbers :D	2026-08-22 14:39:42 +03:00
Bananymous	db9beac740	BAN: Add math functions for clz, ctz, popcount	2026-08-22 14:39:42 +03:00
Bananymous	471b6a6c40	Kernel: Replace __atomic builtins with BAN::atomic wrappers	2026-08-22 13:28:56 +03:00
Bananymous	fe2b130915	Kernel: Fix __cxa_guard_* implementations	2026-08-22 13:27:25 +03:00
Bananymous	67fdd24a19	Kernel: Prefer 32 bit HPET timers on 32 bit target	2026-08-22 13:23:43 +03:00
Bananymous	0d6b32dc3c	Kernel: Fix HPET validation	2026-08-22 13:23:43 +03:00
Bananymous	ffceb300db	Kernel: Cleanup xHCI debug printing	2026-08-22 11:57:19 +03:00
Bananymous	b818e2764c	Kernel: Only update USB max packet size for FS devices	2026-08-22 11:57:19 +03:00
Bananymous	4b7d90f939	Kernel: Ignore SET_PROTOCOL(1) errors on HID devices One of my USB mice is a boot protocol device but stalls on SET_PROTOCOL request. USB HID is supposed to be in report protocol on initialization so an error should not be an issue	2026-08-22 11:57:19 +03:00
Bananymous	6a7c1c1281	Kernel: Reset xHCI control endpoint on command stall	2026-08-22 11:57:19 +03:00
Bananymous	fdbfda189f	Kernel: Map non-us hashtag to backspace on HID keyboards Apparently one of my laptops advertise this key as non-us hashtag instead of backslash while the key layout is the exact same as my other keyboards :D	2026-08-22 11:57:19 +03:00
Bananymous	e496d889ba	Kernel: Deduplicate USB keyboard input code	2026-08-22 11:57:19 +03:00
Bananymous	9967878886	Kernel: Handle array usages for consumer page Also map consumer page mute button in variable case	2026-08-22 11:57:19 +03:00
Bananymous	51badd75f0	Kernel: Move HID led handling from keyboard to usb device Apparently some keyboards can have led control on another interface from the one where input is received. Now led mask is global to the usb devices instead of per keyboard instance.	2026-08-22 11:57:19 +03:00
Bananymous	f6e841e623	Kernel: Don't reset iHDA stream after running out of samples We can just stop the stream :D This also broke restarting the stream on some hardware when ig you need to reprogram the DAC after resetting stream(?)	2026-08-22 11:57:19 +03:00
Bananymous	28bd9cf353	Kernel: Fix PS/2 scancode set detection logic If we cannot determine current scancode set because either the keyboard does not respond or responds with bogus value, assume scancode set 1 and mark the scancode set as uncertain. If we receive the byte 0xF0 while having the uncertain flag set, swap to scancode set 2. 0xF0 is unused in scancode set 1 but indicates key release in scancode set 2, so it will be sent after every key press.	2026-08-22 11:57:19 +03:00
Bananymous	892829adeb	Kernel: Fix PS/2 device detection If we cannot find PS/2 devices in the acpi namespace check 8042 bit in FADT. Also if user explicitly specified PS/2 scancode set, assume that we do have a PS/2 controller even when other detection mechanisms fail	2026-08-22 11:57:19 +03:00
Bananymous	82fd57af9e	Kernel: Use local labels in assembly	2026-08-22 11:57:19 +03:00
Bananymous	2f25bfb2db	Kernel: Use min heap for scheduler block list This makes insertion to block list O(log n) instead of O(n) when blocking with wake time. This is more heavy and currently slower than the previous implementation, but that is just because we don't really have too many threads running at any given time. This will be more scalable in the future and even allows implementing priorities if I ever wish so	2026-08-19 21:32:36 +03:00
Bananymous	7f12e5aaf3	Kernel: Report actual hw volume from AC97	2026-08-19 21:32:36 +03:00
Bananymous	78e40ca353	AudioServer: Add software volume control	2026-08-19 21:32:36 +03:00
Bananymous	1231b16cc3	Kernel: Add mixer support to IHDA This has not been tested as I don't have any hardware with mixers on output paths :P	2026-08-19 21:32:36 +03:00
Bananymous	6cd3638a85	Kernel: Send set channel count on IHDA	2026-08-19 21:32:36 +03:00
Bananymous	4aead25970	Kernel: Cleanup IHDA volume control and report the actual volume	2026-08-19 21:32:36 +03:00
Bananymous	36b34b0df0	Kernel: Don't store copy of palette in virtual TTY It can be just queried from the terminal driver	2026-08-19 21:32:36 +03:00
Bananymous	bf0bd19289	Kernel: Refactor includes to reduce include dependencies Now modifying Scheduler.h or Thread.h doesnt trigger practically a full kernel rebuild. This required moving Mutex and RWLock of of line but that should be fine :^)	2026-08-19 21:32:36 +03:00
Bananymous	8cd9e6dfa8	Kernel: Keep track of thread user and system time separately	2026-08-19 21:32:36 +03:00
Bananymous	f0a114f4e5	Kernel: Fix shm object freeing and track lpid I was deleting based on the key instead of the id which of course does not work :D	2026-08-19 21:32:36 +03:00
Bananymous	ac99bf128b	LibAudio: Rename AudioLoader -> AudioStream	2026-08-19 21:32:36 +03:00
Bananymous	2c66de888f	BuildSystem: Add option to generate an ISO file This creates ISO file with compressed initrd and grub files that can be live booted easily	2026-08-19 21:32:36 +03:00
Bananymous	8f4db57ed8	WindowServer: Mark old window area as damaged on resize	2026-08-17 04:58:35 +03:00
Bananymous	b6f922895f	AudioServer: Queue more than 4k bytes per write Before this I was writing audio data to kernel in ~20 ms chuncks. This feels like way too many syscalls, so now we allow sending upto 1 second per syscall :D	2026-08-17 04:44:15 +03:00
Bananymous	7aa8eb1778	Kernel: Use 64 bit write on HPET on x86_64	2026-08-17 04:33:20 +03:00
Bananymous	1128e440d2	Kernel: Fix typos and fix power button status clear	2026-08-17 04:32:26 +03:00
Bananymous	ac052fe1c1	Kernel: Program missing entries from redirection entries This wasnt an issue but could be in the future	2026-08-17 04:23:04 +03:00