Kernel: Fix framebuffer bounds checks

running yes and pressing ctrl+c paniced the kernel :nekocatwoah:

BAN/include/BAN/Math.h

@@ -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 - clz(x) - 1;
-			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;
 	}
 
 // This is ugly but my clangd does not like including

README.md

@@ -9,7 +9,7 @@ This is my hobby operating system written in C++. Currently supports x86\_64 and
 
 You can find a live demo [here](https://bananymous.com/banan-os)
 
-If you want to try out DOOM, you should first enter the GUI environment using the `start-gui` command. Then you can run `doom` in the GUI terminal.
+![screenshot from qemu running banan-os](assets/banan-os.png)
 
 ### Features
 
@@ -18,6 +18,8 @@ If you want to try out DOOM, you should first enter the GUI environment using th
 - [x] SMP (multiprocessing)
 - [x] Linear framebuffer (VESA and GOP)
 - [x] Network stack
+- [x] USB stack
+- [x] Audio support
 - [x] ELF executable loading
 - [x] AML interpreter (partial)
 - [x] Basic graphical environment
@@ -27,16 +29,14 @@ If you want to try out DOOM, you should first enter the GUI environment using th
   - [ ] Some nice apps
 - [x] ELF dynamic linking
 - [x] copy-on-write memory
-  - [x] file mappings
-  - [ ] anonymous mappings
 
 #### Drivers
 - [x] NVMe disks
 - [x] ATA (IDE, SATA) disks
 - [x] E1000 and E1000E NICs
 - [x] RTL8111/8168/8211/8411 NICs
-- [x] PS2 keyboard (all scancode sets)
+- [x] AC97 and iHDA audio cards
-- [x] PS2 mouse
+- [x] PS2 keyboard and mouse
 - [x] USB
   - [x] xHCI
   - [ ] EHCI
@@ -46,7 +46,8 @@ If you want to try out DOOM, you should first enter the GUI environment using th
   - [x] Mouse
   - [x] Mass storage
   - [x] Hubs
-  - [ ] ...
+  - [ ] Network
+  - [ ] Audio
 - [ ] virtio devices (network, storage)
 
 #### Network
@@ -54,7 +55,7 @@ If you want to try out DOOM, you should first enter the GUI environment using th
 - [x] ICMP
 - [x] IPv4
 - [x] UDP
-- [x] TCP (partial and buggy)
+- [x] TCP
 - [x] Unix domain sockets
 - [ ] SSL
 
@@ -65,7 +66,6 @@ If you want to try out DOOM, you should first enter the GUI environment using th
 - [x] Dev
 - [x] Ram
 - [x] Proc
-- [ ] Sys
 - [ ] 9P
 
 #### Bootloader support
@@ -73,8 +73,6 @@ If you want to try out DOOM, you should first enter the GUI environment using th
 - [x] Custom BIOS bootloader
 - [ ] Custom UEFI bootloader
 
-![screenshot from qemu running banan-os](assets/banan-os.png)
-
 ## Code structure
 
 Each major component and library has its own subdirectory (kernel, userspace, libc, ...). Each directory contains directory *include*, which has **all** of the header files of the component. Every header is included by its absolute path.
@@ -125,8 +123,22 @@ If you have corrupted your disk image or want to create new one, you can either
 ./bos image-full
 ```
 
+To test on real hardware, I have a script to generate a compressed ISO. The ISO can be copied directly to an USB drive with for example `dd`. The file is created at build/banan-os.iso and can be generated with
+```sh
+./bos iso
+```
+
 I have also created shell completion script for zsh. You can either copy the file in _script/shell-completion/zsh/\_bos_ to _/usr/share/zsh/site-functions/_ or add the _script/shell-completion/zsh_ to your fpath in _.zshrc_.
 
+### Package Manager
+
+banan-os uses xbps as its package manager for ports. All upstream ports are packaged into xbps packages hosted on a [repository on my server](https://packages.bananymous.com/banan-os). You can manage sysroot's xbps packages with `./bos` wrapper. You can use `xbps-install`, `xbps-remove` and `xbps-query` with repository and sysroot set to the correct values. For example to install xbps that can be used inside banan-os you can use
+```sh
+./bos xbps-install -S xbps
+```
+For xbps usage see [their documentation](https://docs.voidlinux.org/xbps/index.html).
+
+
 ## Contributing
 
 As the upstream is hosted on my server https://git.bananymous.com/Bananymous/banan-os, merging contributions is not as trivial as it would be on GitHub. You can still send PRs in GitHub in which case I should be able to download the diff and apply it manually. If you want, I can also provide you an account to my git server. In this case please contact me ([email](mailto:[email protected]), [discord](https://discord.gg/ehjGySwYdK)).

kernel/CMakeLists.txt

@@ -41,6 +41,7 @@ set(KERNEL_SOURCES
 	kernel/FS/USTARModule.cpp
 	kernel/FS/VirtualFileSystem.cpp
 	kernel/GDT.cpp
+	kernel/Graphics/BGA.cpp
 	kernel/IDT.cpp
 	kernel/Input/InputDevice.cpp
 	kernel/Input/PS2/Controller.cpp
@@ -51,6 +52,8 @@ 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 +86,7 @@ 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

kernel/arch/i686/Syscall.S

@@ -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

kernel/arch/i686/User.S

@@ -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:

kernel/arch/i686/boot.S

@@ -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)
+	ljmpl $0x08, $V2P(.Lgdt_flush)
-gdt_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:
+.Lap_flush_gdt:
-	movl $ap_higher_half, %ecx
+	movl $.Lap_higher_half, %ecx
 	jmp *%ecx
 
-ap_higher_half:
+.Lap_higher_half:
 	movl AP_REL(ap_prepare_paging), %eax
 	call *%eax
 

kernel/arch/x86_64/Syscall.S

@@ -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

kernel/arch/x86_64/User.S

@@ -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:

kernel/arch/x86_64/boot.S

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

kernel/icxxabi.cpp

@@ -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);
+		if (BAN::atomic_load(guard[0], BAN::memory_order_acquire))
-		uint8_t zero = 0;
+			return 0;
-		return __atomic_compare_exchange_n(byte, &zero, 1, false, __ATOMIC_ACQUIRE, __ATOMIC_ACQUIRE);
-	}
 
-	extern "C" void __cxa_guard_release (__guard* g)
+		uint8_t expected = 0;
-	{
+		if (BAN::atomic_compare_exchange(guard[1], expected, 1, BAN::memory_order_acquire))
-		uint8_t* byte = reinterpret_cast<uint8_t*>(g);
+		{
-		__atomic_store_n(byte, 0, __ATOMIC_RELEASE);
+			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_abort (__guard*)
+		while (BAN::atomic_load(guard[1], BAN::memory_order_acquire))
-	{
+			Kernel::Processor::pause();
-		Kernel::panic("__cxa_guard_abort");
 	}
 }
+
+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");
+}

kernel/include/kernel/ACPI/ACPI.h

@@ -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>
 
@@ -53,6 +54,7 @@ namespace Kernel::ACPI
 
 		BAN::ErrorOr<void> initialize_embedded_controller(const AML::Scope& embedded_controller);
 		BAN::ErrorOr<void> initialize_embedded_controllers();
+		void initialize_embedded_controller_gpes();
 
 		BAN::Optional<GAS> find_gpe_block(size_t index);
 		bool enable_gpe(uint8_t gpe);

kernel/include/kernel/ACPI/EmbeddedController.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
 {
@@ -13,7 +14,7 @@ namespace Kernel::ACPI
 	class EmbeddedController
 	{
 	public:
-		static BAN::ErrorOr<BAN::UniqPtr<EmbeddedController>> create(AML::Scope&& scope, uint16_t command_port, uint16_t data_port, BAN::Optional<uint8_t> gpe);
+		static BAN::ErrorOr<BAN::UniqPtr<EmbeddedController>> create(AML::Scope&& scope, uint16_t command_port, uint16_t data_port);
 		~EmbeddedController();
 
 		BAN::ErrorOr<uint8_t> read_byte(uint8_t offset);
@@ -21,21 +22,21 @@ namespace Kernel::ACPI
 
 		const AML::Scope& scope() const { return m_scope; }
 
+		static void handle_gpe_trampoline(void*);
+
 	private:
-		EmbeddedController(AML::Scope&& scope, uint16_t command_port, uint16_t data_port, bool has_gpe)
+		EmbeddedController(AML::Scope&& scope, uint16_t command_port, uint16_t data_port)
 			: m_scope(BAN::move(scope))
 			, m_command_port(command_port)
 			, m_data_port(data_port)
-			, m_has_gpe(has_gpe)
 		{ }
 
 	private:
-		void wait_status_bit(uint8_t bit, uint8_t value);
+		void wait_status_bit(uint8_t mask, bool set);
 
 		uint8_t read_one(uint16_t port);
 		void write_one(uint16_t port, uint8_t value);
 
-		static void handle_gpe_wrapper(void*);
 		void handle_gpe();
 
 		BAN::ErrorOr<void> call_query_method(uint8_t notification);
@@ -56,7 +57,6 @@ namespace Kernel::ACPI
 		const AML::Scope m_scope;
 		const uint16_t m_command_port;
 		const uint16_t m_data_port;
-		const bool m_has_gpe;
 
 		Mutex m_mutex;
 		ThreadBlocker m_thread_blocker;

kernel/include/kernel/APIC.h

@@ -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>
 

kernel/include/kernel/Audio/HDAudio/AudioFunctionGroup.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 };

kernel/include/kernel/Audio/HDAudio/Definitions.h

@@ -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;
 	};

kernel/include/kernel/Device/FramebufferDevice.h

@@ -6,15 +6,22 @@
 namespace Kernel
 {
 
+	class BGAController;
+
 	class FramebufferDevice : public CharacterDevice
 	{
 	public:
+		static BAN::ErrorOr<BAN::RefPtr<FramebufferDevice>> create(paddr_t paddr, uint32_t width, uint32_t height, uint32_t pitch, uint8_t bpp);
+		static BAN::ErrorOr<BAN::RefPtr<FramebufferDevice>> create(BAN::RefPtr<BGAController>);
 		static BAN::ErrorOr<BAN::RefPtr<FramebufferDevice>> create_from_boot_framebuffer();
 		static BAN::RefPtr<FramebufferDevice> boot_framebuffer();
 		~FramebufferDevice();
 
 		uint32_t width() const { return m_width; }
 		uint32_t height() const { return m_height; }
+		uint8_t bpp() const { return m_bpp; }
+
+		BAN::ErrorOr<void> set_bga_controller(BAN::RefPtr<BGAController>);
 
 		uint32_t get_pixel(uint32_t x, uint32_t y) const;
 		void set_pixel(uint32_t x, uint32_t y, uint32_t rgb);
@@ -50,14 +57,16 @@ namespace Kernel
 	private:
 		const BAN::String m_name;
 
-		vaddr_t			m_video_memory_vaddr { 0 };
+		vaddr_t m_video_memory_vaddr { 0 };
-		const paddr_t	m_video_memory_paddr;
+		paddr_t m_video_memory_paddr { 0 };
-		const uint32_t	m_width;
+
-		const uint32_t	m_height;
+		uint32_t m_width  { 0 };
-		const uint32_t	m_pitch;
+		uint32_t m_height { 0 };
-		const uint8_t	m_bpp;
+		uint32_t m_pitch  { 0 };
+		uint8_t  m_bpp    { 0 };
 
 		BAN::UniqPtr<VirtualRange> m_video_buffer;
+		BAN::RefPtr<BGAController> m_bga_controller;
 
 		friend class FramebufferMemoryRegion;
 	};

kernel/include/kernel/Epoll.h

@@ -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>
 

kernel/include/kernel/FS/DevFS/FileSystem.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:

kernel/include/kernel/FS/EventFD.h

@@ -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_read_impl() const override;
-		bool can_write_impl() const override { return m_value < UINT64_MAX - 1; }
+		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;
 	};
 

kernel/include/kernel/FS/Ext2/FileSystem.h

@@ -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
 {

kernel/include/kernel/FS/FAT/FileSystem.h

@@ -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
 {

kernel/include/kernel/FS/FileSystem.h

@@ -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:

kernel/include/kernel/FS/ProcFS/FileSystem.h

@@ -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:

kernel/include/kernel/FS/ProcFS/Inode.h

@@ -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.

kernel/include/kernel/Graphics/BGA.h

@@ -0,0 +1,37 @@
+#pragma once
+
+#include <BAN/RefPtr.h>
+#include <kernel/Lock/SpinLock.h>
+#include <kernel/PCI.h>
+
+#include <sys/framebuffer.h>
+
+namespace Kernel
+{
+
+	class BGAController : public BAN::RefCounted<BGAController>
+	{
+	public:
+		static BAN::ErrorOr<BAN::RefPtr<BGAController>> create(PCI::Device&);
+
+		fb_fix_info get_fb_fix_info() const;
+		fb_var_info get_fb_var_info() const;
+		BAN::ErrorOr<void> set_fb_var_info(const fb_var_info&);
+
+	private:
+		BGAController(PCI::Device&);
+		BAN::ErrorOr<void> initialize();
+
+		void write_reg(uint16_t reg, uint16_t value);
+		uint16_t read_reg(uint16_t reg);
+
+	private:
+		mutable RecursiveSpinLock m_lock;
+		PCI::Device& m_pci_device;
+		BAN::UniqPtr<PCI::BarRegion> m_lfb_bar;
+
+		fb_fix_info m_fix_info {};
+		fb_var_info m_var_info {};
+	};
+
+}

kernel/include/kernel/IDT.h

@@ -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
 	{

kernel/include/kernel/Input/PS2/Keyboard.h

@@ -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 };
 

kernel/include/kernel/InterruptNumbers.h

@@ -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;
+
+}

kernel/include/kernel/Lock/Mutex.h

@@ -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
+		bool try_lock();
-		{
+		void lock() override;
-			const auto tid = Thread::current_tid();
+		void unlock() override;
-			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();
-			}
-		}
 
 		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
+		bool try_lock();
-		{
+		void lock() override;
-			const auto tid = Thread::current_tid();
+		void unlock() override;
-
-			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();
-			}
-		}
 
 		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 };

kernel/include/kernel/Lock/RWLock.h

@@ -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()
+		void rd_lock();
-		{
+		void rd_unlock();
-			SpinLockGuard _(m_lock);
+		void wr_lock();
-			while (m_writers_waiting > 0 || m_writer != -1)
+		void wr_unlock();
-			{
-				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();
-		}
 
 	private:
 		SpinLock m_lock;

kernel/include/kernel/Lock/SpinLock.h

@@ -61,9 +61,6 @@ namespace Kernel
 		uint32_t                             m_lock_depth { 0 };
 	};
 
-	template<typename Lock>
-	class SpinLockGuardAsMutex;
-
 	template<typename Lock>
 	class SpinLockGuard
 	{
@@ -85,7 +82,6 @@ namespace Kernel
 	private:
 		Lock& m_lock;
 		InterruptState m_state;
-		friend class SpinLockGuardAsMutex<Lock>;
 	};
 
 }

kernel/include/kernel/Memory/MemoryBackedRegion.h

@@ -1,5 +1,6 @@
 #pragma once
 
+#include <BAN/Vector.h>
 #include <kernel/Memory/MemoryRegion.h>
 
 namespace Kernel

kernel/include/kernel/Memory/SharedMemoryObject.h

@@ -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;

kernel/include/kernel/Networking/ARPTable.h

@@ -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
 {

kernel/include/kernel/Networking/IPv4Layer.h

@@ -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
 {

kernel/include/kernel/Networking/Loopback.h

@@ -1,6 +1,7 @@
 #pragma once
 
 #include <kernel/Networking/NetworkInterface.h>
+#include <kernel/Memory/VirtualRange.h>
 
 namespace Kernel
 {

kernel/include/kernel/Networking/TCPSocket.h

@@ -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

kernel/include/kernel/Networking/UNIX/Socket.h

@@ -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

kernel/include/kernel/PCI.h

@@ -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>

kernel/include/kernel/Process.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
 {
 
@@ -287,7 +286,7 @@ namespace Kernel
 		BAN::ErrorOr<FileParent> find_parent_file(int fd, const char* path, int flags) const;
 		BAN::ErrorOr<VirtualFileSystem::File> find_relative_parent(int fd, const char* path) const;
 
-		BAN::ErrorOr<MemoryRegion*> validate_and_pin_pointer_access(const void*, size_t, bool needs_write);
+		BAN::ErrorOr<void> validate_and_pin_pointer_access(const void*, size_t, bool needs_write, BAN::Vector<MemoryRegion*>&);
 
 		uint64_t signal_pending_mask() const
 		{

kernel/include/kernel/Processor.h

@@ -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 +19,11 @@ namespace Kernel
 		Enabled,
 	};
 
+	class GDT;
+	class IDT;
+	class Scheduler;
+	class Thread;
+
 #if ARCH(x86_64) || ARCH(i686)
 	class Processor
 	{
@@ -51,8 +53,8 @@ namespace Kernel
 			union
 			{
 				TLBEntry flush_tlb;
-				SchedulerQueue::Node* new_thread;
+				Thread* new_thread;
-				SchedulerQueue::Node* unblock_thread;
+				Thread* unblock_thread;
 				bool dummy;
 			};
 		};
@@ -179,17 +181,23 @@ 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)))
 		{
-			asm volatile("mov %[value], %%gs:%a[offset]" :: [value]"r"(value), [offset]"ir"(offset) : "memory");
+			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");
 		}
 
 		void lock_tlb_lock();

kernel/include/kernel/Scheduler.h

@@ -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:
@@ -63,9 +41,8 @@ namespace Kernel
 		void on_timer_interrupt();
 		void on_yield(YieldRegisters*);
 
-		static BAN::ErrorOr<void> bind_thread_to_processor(Thread*, ProcessorID);
+		static void bind_thread_to_processor(Thread*, ProcessorID);
-		// if thread is already bound, this will never fail
+		void add_thread(Thread*);
-		BAN::ErrorOr<void> add_thread(Thread*);
 
 		void block_current_thread(ThreadBlocker* thread_blocker, uint64_t wake_time_ns, BaseMutex* mutex);
 		void unblock_thread(Thread*);
@@ -79,9 +56,9 @@ namespace Kernel
 	private:
 		Scheduler() = default;
 
-		void add_current_to_most_loaded(SchedulerQueue* target_queue);
+		void add_current_to_most_loaded(void* target_list);
-		void update_most_loaded_node_queue(SchedulerQueue::Node*, SchedulerQueue* target_queue);
+		void update_most_loaded_node_list(SchedulerThreadNode*, void* target_list);
-		void remove_node_from_most_loaded(SchedulerQueue::Node*);
+		void remove_node_from_most_loaded(SchedulerThreadNode*);
 
 		void update_wake_up_deadline();
 		void wake_up_sleeping_threads();
@@ -90,13 +67,10 @@ namespace Kernel
 
 		class ProcessorID find_least_loaded_processor() const;
 
-		void add_thread(SchedulerQueue::Node*);
-		void unblock_thread(SchedulerQueue::Node*);
-
 	private:
-		SchedulerQueue m_run_queue;
+		SchedulerQueue m_run_list;
-		SchedulerQueue m_block_queue;
+		SchedulerHeap m_block_list;
-		SchedulerQueue::Node* m_current { nullptr };
+		SchedulerThreadNode* m_current { nullptr };
 
 		uint32_t m_thread_count { 0 };
 
@@ -108,8 +82,8 @@ namespace Kernel
 
 		struct ThreadInfo
 		{
-			SchedulerQueue*       queue { nullptr };
+			void*                list { nullptr };
-			SchedulerQueue::Node* node  { nullptr };
+			SchedulerThreadNode* node { nullptr };
 		};
 		BAN::Array<ThreadInfo, 10> m_most_loaded_threads;
 

kernel/include/kernel/SchedulerQueueNode.h

@@ -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 };
-	};
-
-}

kernel/include/kernel/SchedulerThreadNode.h

@@ -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 };
+	};
+
+}

kernel/include/kernel/Terminal/PseudoTerminal.h

@@ -1,6 +1,7 @@
 #pragma once
 
 #include <BAN/WeakPtr.h>
+#include <kernel/Memory/VirtualRange.h>
 #include <kernel/Terminal/TTY.h>
 
 namespace Kernel

kernel/include/kernel/Terminal/TTY.h

@@ -53,6 +53,8 @@ namespace Kernel
 
 		virtual bool master_has_closed() const { return false; }
 
+		virtual bool is_vtty() const { return false; }
+
 	protected:
 		TTY(termios termios, mode_t mode, uid_t uid, gid_t gid);
 
@@ -100,9 +102,9 @@ namespace Kernel
 		termios m_termios;
 
 	protected:
-		Mutex m_mutex;
+		Mutex m_input_mutex;
 
-		Mutex m_write_lock;
+		RecursiveSpinLock m_write_lock;
 		ThreadBlocker m_write_blocker;
 	};
 

kernel/include/kernel/Terminal/VirtualTTY.h

@@ -21,6 +21,9 @@ namespace Kernel
 
 		void clear() override;
 
+		bool is_vtty() const override { return true; }
+		BAN::ErrorOr<void> set_terminal_driver(BAN::RefPtr<TerminalDriver>);
+
 	protected:
 		BAN::StringView name() const override { return m_name; }
 		bool putchar_impl(uint8_t ch) override;
@@ -96,8 +99,6 @@ 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 };

kernel/include/kernel/Thread.h

@@ -5,8 +5,9 @@
 #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 <kernel/SchedulerThreadNode.h>
 
 #include <LibELF/AuxiliaryVector.h>
 
@@ -18,6 +19,7 @@ namespace Kernel
 
 	class MemoryBackedRegion;
 	class Process;
+	class ThreadBlocker;
 
 	class Thread
 	{
@@ -124,10 +126,14 @@ 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 +193,7 @@ namespace Kernel
 		vaddr_t                    m_fsbase               { 0 };
 		vaddr_t                    m_gsbase               { 0 };
 
-		SchedulerQueue::Node*      m_scheduler_node       { nullptr };
+		SchedulerThreadNode        m_scheduler_node;
 
 		YieldRegisters             m_yield_registers      { };
 
@@ -200,8 +206,10 @@ 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 };

kernel/include/kernel/ThreadBlocker.h

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

kernel/include/kernel/USB/Definitions.h

@@ -30,6 +30,7 @@ namespace Kernel
 			Hub                        = 0x09,
 			BillboardDeviceClass       = 0x11,
 			DiagnosticDevice           = 0xDC,
+			WirelessController         = 0xE0,
 			Miscellaneous              = 0xEF,
 			VendorSpecific             = 0xFF,
 		};

kernel/include/kernel/USB/Device.h

@@ -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;
 	};
 
 }

kernel/include/kernel/USB/HID/HIDDriver.h

@@ -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>;
 	};
 

kernel/include/kernel/USB/HID/Keyboard.h

@@ -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 };

kernel/include/kernel/USB/Hub/HubDriver.h

@@ -1,6 +1,5 @@
 #pragma once
 
-#include <kernel/Process.h>
 #include <kernel/USB/Device.h>
 
 namespace Kernel

kernel/include/kernel/USB/XHCI/Definitions.h

@@ -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;

kernel/kernel/ACPI/ACPI.cpp

@@ -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
@@ -783,30 +784,10 @@ acpi_release_global_lock:
 
 	BAN::ErrorOr<void> ACPI::initialize_embedded_controller(const AML::Scope& embedded_controller)
 	{
-		BAN::Optional<uint8_t> gpe_int;
-
-		do {
-			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);
-				break;
-			}
-
-			auto gpe = TRY(AML::evaluate_node(gpe_path, gpe_obj->node));
-			if (gpe.type == AML::Node::Type::Package)
-			{
-				dwarnln("TODO: EC {} has package _GPE");
-				break;
-			}
-
-			gpe_int = TRY(AML::convert_node(BAN::move(gpe), AML::ConvInteger, -1)).as.integer.value;
-		} while (false);
-
 		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);
 		}
 
@@ -856,7 +837,7 @@ acpi_release_global_lock:
 		const auto data_port = TRY(extract_io_port(crs_buffer));
 		const auto command_port = TRY(extract_io_port(crs_buffer));
 
-		TRY(m_embedded_controllers.push_back(TRY(EmbeddedController::create(TRY(embedded_controller.copy()), command_port, data_port, gpe_int))));
+		TRY(m_embedded_controllers.push_back(TRY(EmbeddedController::create(TRY(embedded_controller.copy()), command_port, data_port))));
 		return {};
 	}
 
@@ -876,6 +857,33 @@ acpi_release_global_lock:
 		return {};
 	}
 
+	void ACPI::initialize_embedded_controller_gpes()
+	{
+		const auto initialize_gpe = [this](EmbeddedController& embedded_controller) -> BAN::ErrorOr<void> {
+			auto [gpe_path, gpe_obj] = TRY(m_namespace->find_named_object(embedded_controller.scope(), TRY(AML::NameString::from_string("_GPE"_sv)), true));
+			if (gpe_obj == nullptr)
+			{
+				dprintln("EC {} does not have _GPE", embedded_controller.scope());
+				return {};
+			}
+
+			auto gpe = TRY(AML::evaluate_node(gpe_path, gpe_obj->node));
+			if (gpe.type == AML::Node::Type::Package)
+			{
+				dwarnln("TODO: EC {} has package _GPE");
+				return {};
+			}
+
+			const auto gpe_int = TRY(AML::convert_node(BAN::move(gpe), AML::ConvInteger, -1)).as.integer.value;
+			TRY(register_gpe_handler(gpe_int, &EmbeddedController::handle_gpe_trampoline, &embedded_controller));
+
+			return {};
+		};
+
+		for (auto& embedded_controller : m_embedded_controllers)
+			(void)initialize_gpe(*embedded_controller);
+	}
+
 	BAN::ErrorOr<void> ACPI::register_gpe_handler(uint8_t gpe, void (*callback)(void*), void* argument)
 	{
 		if (m_gpe_methods[gpe].method)
@@ -995,57 +1003,27 @@ acpi_release_global_lock:
 			// FIXME: add support for GPE blocks inside the ACPI namespace
 		}
 
-		if (auto ret = initialize_embedded_controllers(); ret.is_error())
-			dwarnln("Failed to initialize Embedded Controllers: {}", ret.error());
-
-		if (auto ret = m_namespace->post_load_initialize(); ret.is_error())
-			dwarnln("Failed to initialize ACPI namespace: {}", ret.error());
-
-		auto [pic_path, pic_obj] = TRY(m_namespace->find_named_object({}, TRY(AML::NameString::from_string("\\_PIC"_sv))));
-		if (pic_obj && pic_obj->node.type == AML::Node::Type::Method)
-		{
-			auto& pic_node = pic_obj->node;
-			if (pic_node.as.method.arg_count != 1)
-			{
-				dwarnln("Method \\_PIC has {} arguments, expected 1", pic_node.as.method.arg_count);
-				return BAN::Error::from_errno(EINVAL);
-			}
-
-			AML::Reference arg_ref;
-			arg_ref.node.type = AML::Node::Type::Integer;
-			arg_ref.node.as.integer.value = InterruptController::get().is_using_apic() ? 1 : 0;
-			arg_ref.ref_count = 2;
-
-			BAN::Array<AML::Reference*, 7> arguments(nullptr);
-			arguments[0] = &arg_ref; // method call should not delete argument
-			TRY(AML::method_call(pic_path, pic_node, BAN::move(arguments)));
-		}
-
-		dprintln("Evaluated \\_PIC({})", InterruptController::get().is_using_apic() ? 1 : 0);
-
 		uint8_t irq = fadt().sci_int;
 		if (auto ret = InterruptController::get().reserve_irq(irq); ret.is_error())
 			dwarnln("Could not enable ACPI interrupt: {}", ret.error());
 		else
 		{
-			auto hex_sv_to_int =
+			const auto hex_sv_to_int = [](BAN::StringView sv) -> BAN::Optional<uint32_t> {
-				[](BAN::StringView sv) -> BAN::Optional<uint32_t>
+				uint32_t ret = 0;
+				for (char c : sv)
 				{
-					uint32_t ret = 0;
+					ret <<= 4;
-					for (char c : sv)
+					if (c >= '0' && c <= '9')
-					{
+						ret += c - '0';
-						ret <<= 4;
+					else if (c >= 'A' && c <= 'F')
-						if (c >= '0' && c <= '9')
+						ret += c - 'A' + 10;
-							ret += c - '0';
+					else if (c >= 'a' && c <= 'f')
-						else if (c >= 'A' && c <= 'F')
+						ret += c - 'a' + 10;
-							ret += c - 'A' + 10;
+					else
-						else if (c >= 'a' && c <= 'f')
+						return {};
-							ret += c - 'a' + 10;
+				}
-						else
+				return ret;
-							return {};
+			};
-					}
-					return ret;
-				};
 
 			auto [gpe_scope, gpe_obj] = TRY(m_namespace->find_named_object({}, TRY(AML::NameString::from_string("\\_GPE"))));
 			if (gpe_obj && gpe_obj->node.is_scope())
@@ -1091,11 +1069,43 @@ acpi_release_global_lock:
 
 			if (auto thread_or_error = Thread::create_kernel([](void*) { get().acpi_event_task(); }, nullptr); thread_or_error.is_error())
 				dwarnln("Failed to create ACPI thread, power button will not work: {}", thread_or_error.error());
-			else if (auto ret = Processor::scheduler().add_thread(thread_or_error.value()); ret.is_error())
+			else
-				dwarnln("Failed to create ACPI thread, power button will not work: {}", ret.error());
+			{
+				Processor::scheduler().add_thread(thread_or_error.value());
+				dprintln("Initialized ACPI interrupts");
+			}
 		}
 
-		dprintln("Initialized ACPI interrupts");
+		if (auto ret = initialize_embedded_controllers(); ret.is_error())
+			dwarnln("Failed to initialize Embedded Controllers: {}", ret.error());
+
+		if (auto ret = m_namespace->post_load_initialize(); ret.is_error())
+			dwarnln("Failed to initialize ACPI namespace: {}", ret.error());
+
+		// NOTE: We cannot initialize EC GPEs before the post init is done, but post init does need ECs initialized
+		initialize_embedded_controller_gpes();
+
+		auto [pic_path, pic_obj] = TRY(m_namespace->find_named_object({}, TRY(AML::NameString::from_string("\\_PIC"_sv))));
+		if (pic_obj && pic_obj->node.type == AML::Node::Type::Method)
+		{
+			auto& pic_node = pic_obj->node;
+			if (pic_node.as.method.arg_count != 1)
+			{
+				dwarnln("Method \\_PIC has {} arguments, expected 1", pic_node.as.method.arg_count);
+				return BAN::Error::from_errno(EINVAL);
+			}
+
+			AML::Reference arg_ref;
+			arg_ref.node.type = AML::Node::Type::Integer;
+			arg_ref.node.as.integer.value = InterruptController::get().is_using_apic() ? 1 : 0;
+			arg_ref.ref_count = 2;
+
+			BAN::Array<AML::Reference*, 7> arguments(nullptr);
+			arguments[0] = &arg_ref; // method call should not delete argument
+			TRY(AML::method_call(pic_path, pic_node, BAN::move(arguments)));
+
+			dprintln("Evaluated \\_PIC({})", InterruptController::get().is_using_apic() ? 1 : 0);
+		}
 
 		if (InterruptController::get().is_using_apic())
 		{
@@ -1198,6 +1208,8 @@ acpi_release_global_lock:
 			continue;
 
 handle_event:
+			IO::outw(sts_port, pending);
+
 			if (pending & PM1_EVN_PWRBTN)
 			{
 				dprintln("Power button pressed");
@@ -1208,8 +1220,6 @@ handle_event:
 			{
 				dwarnln("Unhandled ACPI fixed event {H}", pending);
 			}
-
-			IO::outw(sts_port, pending);
 		}
 	}
 

kernel/kernel/ACPI/AML/OpRegion.cpp

@@ -494,8 +494,23 @@ namespace Kernel::ACPI::AML
 
 				return TRY(embedded_controller->read_byte(offset));
 			}
-			case GAS::AddressSpaceID::SMBus:
 			case GAS::AddressSpaceID::SystemCMOS:
+				if (byte_offset >= 64)
+				{
+					dwarnln("CMOS read from offset 0x{H}", byte_offset);
+					return BAN::Error::from_errno(ENOTSUP);
+				}
+				switch (access_size)
+				{
+					case 1:
+						IO::outb(0x70, offset);
+						return IO::inb(0x71);
+					default:
+						dwarnln("{} byte read from CMOS offset {2H}", access_size, byte_offset);
+						return BAN::Error::from_errno(EINVAL);
+				}
+				ASSERT_NOT_REACHED();
+			case GAS::AddressSpaceID::SMBus:
 			case GAS::AddressSpaceID::PCIBarTarget:
 			case GAS::AddressSpaceID::IPMI:
 			case GAS::AddressSpaceID::GeneralPurposeIO:
@@ -581,8 +596,24 @@ namespace Kernel::ACPI::AML
 				TRY(embedded_controller->write_byte(offset, value));
 				return {};
 			}
-			case GAS::AddressSpaceID::SMBus:
 			case GAS::AddressSpaceID::SystemCMOS:
+				if (byte_offset >= 64)
+				{
+					dwarnln("CMOS write to offset 0x{H}", byte_offset);
+					return BAN::Error::from_errno(ENOTSUP);
+				}
+				switch (access_size)
+				{
+					case 1:
+						IO::outb(0x70, offset);
+						IO::outb(0x71, value);
+						break;
+					default:
+						dwarnln("{} byte write to CMOS offset {2H}", access_size, byte_offset);
+						return BAN::Error::from_errno(EINVAL);
+				}
+				return {};
+			case GAS::AddressSpaceID::SMBus:
 			case GAS::AddressSpaceID::PCIBarTarget:
 			case GAS::AddressSpaceID::IPMI:
 			case GAS::AddressSpaceID::GeneralPurposeIO:

kernel/kernel/ACPI/BatterySystem.cpp

@@ -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 target_str = TRY(BAN::String::formatted("{}", TRY(get_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;
-			}
-
-			auto target_str = TRY(BAN::String::formatted("{}", m_last_value.load()));
-
 			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;
+		Mutex m_mutex;
-		BAN::Atomic<uint64_t> m_last_value = 0;
+		uint64_t m_last_read_ms = 0;
+		uint64_t m_last_value = 0;
 	};
 
 	BAN::ErrorOr<void> BatterySystem::initialize(AML::Namespace& acpi_namespace)

kernel/kernel/ACPI/EmbeddedController.cpp

@@ -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
@@ -28,28 +30,18 @@ namespace Kernel::ACPI
 		CMD_QUERY     = 0x84,
 	};
 
-	BAN::ErrorOr<BAN::UniqPtr<EmbeddedController>> EmbeddedController::create(AML::Scope&& scope, uint16_t command_port, uint16_t data_port, BAN::Optional<uint8_t> gpe)
+	BAN::ErrorOr<BAN::UniqPtr<EmbeddedController>> EmbeddedController::create(AML::Scope&& scope, uint16_t command_port, uint16_t data_port)
 	{
-		auto* embedded_controller_ptr = new EmbeddedController(BAN::move(scope), command_port, data_port, gpe.has_value());
+		auto* embedded_controller_ptr = new EmbeddedController(BAN::move(scope), command_port, data_port);
 		if (embedded_controller_ptr == nullptr)
 			return BAN::Error::from_errno(ENOMEM);
 
 		auto* thread = TRY(Thread::create_kernel([](void* ec) { static_cast<EmbeddedController*>(ec)->thread_task(); }, embedded_controller_ptr));
-		TRY(Processor::scheduler().add_thread(thread));
+		Processor::scheduler().add_thread(thread);
 
 		auto embedded_controller = BAN::UniqPtr<EmbeddedController>::adopt(embedded_controller_ptr);
 		embedded_controller->m_thread = thread;
 
-		if (gpe.has_value())
-			TRY(ACPI::get().register_gpe_handler(gpe.value(), &handle_gpe_wrapper, embedded_controller.ptr()));
-		else
-		{
-			// FIXME: Restructure EC such that SCI_EVT can be polled.
-			//        Simple solution would be spawning another thread,
-			//        but that feels too hacky.
-			dwarnln("TODO: SCI_EVT polling without GPE");
-		}
-
 		return embedded_controller;
 	}
 
@@ -93,26 +85,25 @@ namespace Kernel::ACPI
 
 	uint8_t EmbeddedController::read_one(uint16_t port)
 	{
-		wait_status_bit(STS_OBF, 1);
+		wait_status_bit(STS_OBF, true);
 		return IO::inb(port);
 	}
 
 	void EmbeddedController::write_one(uint16_t port, uint8_t value)
 	{
-		wait_status_bit(STS_IBF, 0);
+		wait_status_bit(STS_IBF, false);
 		IO::outb(port, value);
 	}
 
-	void EmbeddedController::wait_status_bit(uint8_t bit, uint8_t value)
+	void EmbeddedController::wait_status_bit(uint8_t mask, bool set)
 	{
 		// FIXME: timeouts
-		const uint8_t mask = 1 << bit;
+		const uint8_t comp = set ? mask : 0;
-		const uint8_t comp = value ? mask : 0;
 		while ((IO::inb(m_command_port) & mask) != comp)
 			continue;
 	}
 
-	void EmbeddedController::handle_gpe_wrapper(void* embedded_controller)
+	void EmbeddedController::handle_gpe_trampoline(void* embedded_controller)
 	{
 		static_cast<EmbeddedController*>(embedded_controller)->handle_gpe();
 	}
@@ -215,7 +206,7 @@ namespace Kernel::ACPI
 
 		for (;;)
 		{
-			Command* const command = m_queued_command.has_value() ? m_queued_command.value() : nullptr;
+			auto* const command = m_queued_command.value_or(nullptr);
 			m_queued_command.clear();
 
 			if (command == nullptr)
@@ -224,27 +215,22 @@ namespace Kernel::ACPI
 				continue;
 			}
 
+			// TODO: use burst mode
+
 			m_mutex.unlock();
 
-			if (command)
+			write_one(m_command_port, command->command);
-			{
+			if (command->data1.has_value())
-				// TODO: use burst mode
+				write_one(m_data_port, command->data1.value());
-
+			if (command->data2.has_value())
-				write_one(m_command_port, command->command);
+				write_one(m_data_port, command->data2.value());
-				if (command->data1.has_value())
+			if (command->response)
-					write_one(m_data_port, command->data1.value());
+				*command->response = read_one(m_data_port);
-				if (command->data2.has_value())
-					write_one(m_data_port, command->data2.value());
-				if (command->response)
-					*command->response = read_one(m_data_port);
-
-				m_mutex.lock();
-				command->done = true;
-				m_thread_blocker.unblock();
-				m_mutex.unlock();
-			}
 
 			m_mutex.lock();
+
+			command->done = true;
+			m_thread_blocker.unblock();
 		}
 	}
 

kernel/kernel/APIC.cpp

@@ -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();
 

kernel/kernel/Audio/AC97/Controller.cpp

@@ -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 {};
 	}
 

kernel/kernel/Audio/Controller.cpp

@@ -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>

kernel/kernel/Audio/HDAudio/AudioFunctionGroup.cpp

@@ -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];
+		const auto* node = m_output_paths[m_output_path_index][m_amplifier_idx];
-		for (size_t i = 0; i < path.size(); i++)
-		{
-			if (!path[i]->output_amplifier.has_value())
-				continue;
 
-			const int32_t step_round = (mdB >= 0)
+		const int32_t step_round = (mdB >= 0)
-				? +m_volume_info.step_mdB / 2
+			? +m_volume_info.step_mdB / 2
-				: -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;
+		const uint32_t volume = 0b1'0'1'1'0000'0'0000000 | step;
 
-			TRY(m_controller->send_command({
+		TRY(m_controller->send_command({
-				.data = static_cast<uint8_t>(volume & 0xFF),
+			.data = static_cast<uint8_t>(volume & 0xFF),
-				.command = static_cast<uint16_t>(0x300 | (volume >> 8)),
+			.command = static_cast<uint16_t>(0x300 | (volume >> 8)),
-				.node_index = path[i]->id,
+			.node_index = node->id,
-				.codec_address = m_cid,
+			.codec_address = m_cid,
-			}));
+		}));
 
-			break;
+		m_volume_info.mdB = (mdB + step_round) / m_volume_info.step_mdB * m_volume_info.step_mdB;
-		}
-
-		m_volume_info.mdB = 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
+		// format: PCM, 48 kHz, 16 bit
-		return 0b0'0'000'000'0'001'0001;
+		return 0b0'0'000'000'0'001'0000 | (get_channels() - 1);
 	}
 
 	BAN::ErrorOr<void> HDAudioFunctionGroup::enable_output_path(uint8_t index)
@@ -346,6 +342,8 @@ namespace Kernel
 			{
 				using HDAudio::AFGWidget;
 				case AFGWidget::Type::OutputConverter:
+				case AFGWidget::Type::Mixer:
+				case AFGWidget::Type::Selector:
 				case AFGWidget::Type::PinComplex:
 					break;
 				default:
@@ -367,7 +365,7 @@ 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 +405,13 @@ 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 +421,28 @@ 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::Selector:
+					break;
+
 				case AFGWidget::Type::PinComplex:
 					// enable output and H-Phn
 					TRY(m_controller->send_command({
@@ -438,15 +465,25 @@ namespace Kernel
 			}
 		}
 
-		// update volume info to this path
+		int32_t max_steps { 0 };
-		m_volume_info.min_mdB = 0;
+		m_amplifier_idx = SIZE_MAX;
-		m_volume_info.max_mdB = 0;
+
-		m_volume_info.step_mdB = 0;
 		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 +500,11 @@ 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 +528,8 @@ namespace Kernel
 				using HDAudio::AFGWidget;
 
 				case AFGWidget::Type::OutputConverter:
+				case AFGWidget::Type::Mixer:
+				case AFGWidget::Type::Selector:
 					break;
 
 				case AFGWidget::Type::PinComplex:
@@ -635,11 +669,8 @@ 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())
+				bar.write8(base + Regs::SDCTL, bar.read8(base + Regs::SDCTL) & 0xFD);
-				{
+				m_stream_running = false;
-					dwarnln("failed to reset HDA stream: {}", ret.error());
-					return;
-				}
 			}
 
 			queue_bdl_data();

kernel/kernel/Audio/HDAudio/Controller.cpp

@@ -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;

kernel/kernel/Banos.cpp

@@ -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

kernel/kernel/Device/FramebufferDevice.cpp

@@ -2,6 +2,7 @@
 #include <kernel/Device/DeviceNumbers.h>
 #include <kernel/Device/FramebufferDevice.h>
 #include <kernel/FS/DevFS/FileSystem.h>
+#include <kernel/Graphics/BGA.h>
 #include <kernel/Memory/Heap.h>
 #include <kernel/Terminal/FramebufferTerminal.h>
 
@@ -26,19 +27,14 @@ namespace Kernel
 		return s_boot_framebuffer;
 	}
 
-	BAN::ErrorOr<BAN::RefPtr<FramebufferDevice>> FramebufferDevice::create_from_boot_framebuffer()
+	BAN::ErrorOr<BAN::RefPtr<FramebufferDevice>> FramebufferDevice::create(paddr_t paddr, uint32_t width, uint32_t height, uint32_t pitch, uint8_t bpp)
 	{
-		ASSERT(g_boot_info.framebuffer.type == FramebufferInfo::Type::RGB);
+		if (bpp != 24 && bpp != 32)
-		if (g_boot_info.framebuffer.bpp != 24 && g_boot_info.framebuffer.bpp != 32)
 			return BAN::Error::from_errno(ENOTSUP);
 		auto* device_ptr = new FramebufferDevice(
 			0660, 0, 900,
 			makedev(DeviceNumber::Framebuffer, get_framebuffer_device_index()),
-			g_boot_info.framebuffer.address,
+			paddr, width, height, pitch, bpp
-			g_boot_info.framebuffer.width,
-			g_boot_info.framebuffer.height,
-			g_boot_info.framebuffer.pitch,
-			g_boot_info.framebuffer.bpp
 		);
 		if (device_ptr == nullptr)
 			return BAN::Error::from_errno(ENOMEM);
@@ -49,6 +45,33 @@ namespace Kernel
 		return device;
 	}
 
+	BAN::ErrorOr<BAN::RefPtr<FramebufferDevice>> FramebufferDevice::create(BAN::RefPtr<BGAController> bga_controller)
+	{
+		const auto fix_info = bga_controller->get_fb_fix_info();
+		const auto var_info = bga_controller->get_fb_var_info();
+		auto controller = TRY(create(
+			fix_info.mem_start,
+			var_info.xres,
+			var_info.yres,
+			fix_info.pitch,
+			var_info.bpp
+		));
+		controller->m_bga_controller = bga_controller;
+		return controller;
+	}
+
+	BAN::ErrorOr<BAN::RefPtr<FramebufferDevice>> FramebufferDevice::create_from_boot_framebuffer()
+	{
+		s_boot_framebuffer = TRY(create(
+			g_boot_info.framebuffer.address,
+			g_boot_info.framebuffer.width,
+			g_boot_info.framebuffer.height,
+			g_boot_info.framebuffer.pitch,
+			g_boot_info.framebuffer.bpp
+		));
+		return s_boot_framebuffer;
+	}
+
 	FramebufferDevice::FramebufferDevice(mode_t mode, uid_t uid, gid_t gid, dev_t rdev, paddr_t paddr, uint32_t width, uint32_t height, uint32_t pitch, uint8_t bpp)
 		: CharacterDevice(mode, uid, gid)
 		, m_name(MUST(BAN::String::formatted("fb{}", minor(rdev))))
@@ -65,16 +88,20 @@ namespace Kernel
 	{
 		if (m_video_memory_vaddr == 0)
 			return;
-		size_t video_memory_pages = range_page_count(m_video_memory_paddr, m_height * m_pitch);
+		const size_t video_memory_bytes = m_bga_controller ? m_bga_controller->get_fb_fix_info().mem_size : m_height * m_pitch;
+		const size_t video_memory_pages = range_page_count(m_video_memory_paddr, video_memory_bytes);
 		PageTable::kernel().unmap_range(m_video_memory_vaddr, video_memory_pages * PAGE_SIZE);
 	}
 
 	BAN::ErrorOr<void> FramebufferDevice::initialize()
 	{
-		size_t video_memory_pages = range_page_count(m_video_memory_paddr, m_height * m_pitch);
+		const size_t video_memory_bytes = m_bga_controller ? m_bga_controller->get_fb_fix_info().mem_size : m_height * m_pitch;
+		const size_t video_memory_pages = range_page_count(m_video_memory_paddr, video_memory_bytes);
+
 		m_video_memory_vaddr = PageTable::kernel().reserve_free_contiguous_pages(video_memory_pages, KERNEL_OFFSET);
 		if (m_video_memory_vaddr == 0)
 			return BAN::Error::from_errno(ENOMEM);
+
 		PageTable::kernel().map_range_at(
 			m_video_memory_paddr & PAGE_ADDR_MASK,
 			m_video_memory_vaddr,
@@ -86,7 +113,7 @@ namespace Kernel
 		m_video_buffer = TRY(VirtualRange::create_to_vaddr_range(
 			PageTable::kernel(),
 			{ KERNEL_OFFSET, UINTPTR_MAX },
-			BAN::Math::div_round_up<size_t>(m_width * m_height * (BANAN_FB_BPP / 8), PAGE_SIZE) * PAGE_SIZE,
+			video_memory_pages * PAGE_SIZE,
 			PageTable::Flags::ReadWrite | PageTable::Flags::Present,
 			false
 		));
@@ -94,6 +121,37 @@ namespace Kernel
 		return {};
 	}
 
+	BAN::ErrorOr<void> FramebufferDevice::set_bga_controller(BAN::RefPtr<BGAController> bga_controller)
+	{
+		ASSERT(!m_bga_controller);
+
+		if (auto var_info = bga_controller->get_fb_var_info(); var_info.bpp != 32)
+		{
+			var_info.bpp = 32;
+			TRY(bga_controller->set_fb_var_info(var_info));
+		}
+
+		const auto fix_info = bga_controller->get_fb_fix_info();
+		const auto var_info = bga_controller->get_fb_var_info();
+
+		ASSERT(m_video_memory_vaddr);
+		PageTable::kernel().unmap_range(m_video_memory_vaddr, m_height * m_pitch);
+
+		m_video_memory_vaddr = 0;
+		m_video_memory_paddr = fix_info.mem_start;
+
+		m_width  = var_info.xres;
+		m_height = var_info.yres;
+		m_pitch  = fix_info.pitch;
+		m_bpp    = var_info.bpp;
+
+		m_bga_controller = bga_controller;
+
+		TRY(initialize());
+
+		return {};
+	}
+
 	BAN::ErrorOr<size_t> FramebufferDevice::read_impl(off_t offset, BAN::ByteSpan buffer)
 	{
 		// Reading from negative offset will fill buffer with framebuffer info

kernel/kernel/Epoll.cpp

@@ -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

kernel/kernel/FS/DevFS/FileSystem.cpp

@@ -63,7 +63,7 @@ namespace Kernel
 				}
 			}, s_instance
 		));
-		MUST(Processor::scheduler().add_thread(updater_thread));
+		Processor::scheduler().add_thread(updater_thread);
 
 		auto* disk_cache_drop_thread = MUST(Thread::create_kernel(
 			[](void* _devfs)
@@ -93,7 +93,7 @@ namespace Kernel
 				}
 			}, s_instance
 		));
-		MUST(Processor::scheduler().add_thread(disk_cache_drop_thread));
+		Processor::scheduler().add_thread(disk_cache_drop_thread);
 
 		auto* disk_sync_thread = MUST(Thread::create_kernel(
 			[](void* _devfs)
@@ -129,7 +129,7 @@ namespace Kernel
 				}
 			}, s_instance
 		));
-		MUST(Processor::scheduler().add_thread(disk_sync_thread));
+		Processor::scheduler().add_thread(disk_sync_thread);
 	}
 
 	void DevFileSystem::initiate_disk_cache_drop()

kernel/kernel/FS/EventFD.cpp

@@ -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;
+	}
+
 }

kernel/kernel/FS/Ext2/FileSystem.cpp

@@ -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

kernel/kernel/FS/FAT/FileSystem.cpp

@@ -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())
 		{

kernel/kernel/FS/FAT/Inode.cpp

@@ -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>
 

kernel/kernel/FS/ProcFS/FileSystem.cpp

@@ -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
 {

kernel/kernel/FS/ProcFS/Inode.cpp

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

kernel/kernel/Graphics/BGA.cpp

@@ -0,0 +1,242 @@
+#include <kernel/Device/FramebufferDevice.h>
+#include <kernel/Graphics/BGA.h>
+#include <kernel/IO.h>
+#include <kernel/Terminal/FramebufferTerminal.h>
+#include <kernel/Terminal/VirtualTTY.h>
+
+namespace Kernel
+{
+
+	enum BGA_IO_PORT : uint16_t
+	{
+		BGA_IO_PORT_INDEX = 0x1CE,
+		BGA_IO_PORT_DATA  = 0x1CF,
+	};
+
+	enum BGA_REG : uint16_t
+	{
+		BGA_REG_ID          = 0,
+		BGA_REG_XRES        = 1,
+		BGA_REG_YRES        = 2,
+		BGA_REG_BPP         = 3,
+		BGA_REG_ENABLE      = 4,
+		BGA_REG_BANK        = 5,
+		BGA_REG_XRES_VIRT  = 6,
+		BGA_REG_YRES_VIRT = 7,
+		BGA_REG_XOFF    = 8,
+		BGA_REG_YOFF    = 9,
+	};
+
+	enum BGA_ID : uint16_t
+	{
+		BGA_ID_0 = 0xB0C0,
+		BGA_ID_1 = 0xB0C1,
+		BGA_ID_2 = 0xB0C2,
+		BGA_ID_3 = 0xB0C3,
+		BGA_ID_4 = 0xB0C4,
+		BGA_ID_5 = 0xB0C5,
+	};
+
+	enum BGA_ENABLE : uint16_t
+	{
+		BGA_ENABLE_ENABLE     = 0x01,
+		BGA_ENABLE_LFB_ENABLE = 0x40,
+		BGA_ENABLE_NOCLEARMEM = 0x80,
+	};
+
+	BAN::ErrorOr<BAN::RefPtr<BGAController>> BGAController::create(PCI::Device& pci_device)
+	{
+		auto* bga_controller_ptr = new BGAController(pci_device);
+		if (bga_controller_ptr == nullptr)
+			return BAN::Error::from_errno(ENOMEM);
+		auto bga_controller = BAN::RefPtr<BGAController>::adopt(bga_controller_ptr);
+		TRY(bga_controller->initialize());
+		return bga_controller;
+	}
+
+	BGAController::BGAController(PCI::Device& pci_device)
+		: m_pci_device(pci_device)
+	{ }
+
+	BAN::ErrorOr<void> BGAController::initialize()
+	{
+		auto boot_framebuffer = FramebufferDevice::boot_framebuffer();
+
+		m_lfb_bar = TRY(m_pci_device.allocate_bar_region(0));
+		if (m_lfb_bar->type() != PCI::BarType::MEM)
+		{
+			dwarnln("BGA LFB is not memory bar");
+			return BAN::Error::from_errno(EINVAL);
+		}
+
+		m_fix_info = {
+			.xpanstep = 1,
+			.ypanstep = 1,
+			.pitch = 0,
+			.mem_start = static_cast<uint64_t>(m_lfb_bar->paddr()),
+			.mem_size  = static_cast<uint32_t>(m_lfb_bar->size()),
+		};
+
+		const auto ready_to_use_flags = BGA_ENABLE_LFB_ENABLE | BGA_ENABLE_ENABLE;
+		if ((read_reg(BGA_REG_ENABLE) & ready_to_use_flags) == ready_to_use_flags)
+		{
+			m_var_info = {
+				.xres      = read_reg(BGA_REG_XRES),
+				.yres      = read_reg(BGA_REG_YRES),
+				.xres_virt = read_reg(BGA_REG_XRES_VIRT),
+				.yres_virt = read_reg(BGA_REG_XRES_VIRT),
+				.xoff      = read_reg(BGA_REG_XOFF),
+				.yoff      = read_reg(BGA_REG_YOFF),
+				.bpp       = read_reg(BGA_REG_BPP),
+			};
+
+			m_fix_info.pitch = m_var_info.xres_virt * m_var_info.bpp / 8;
+		}
+		else
+		{
+			const uint32_t xres = boot_framebuffer ? boot_framebuffer->width()  : 1280;
+			const uint32_t yres = boot_framebuffer ? boot_framebuffer->height() :  800;
+			const uint32_t bpp  = boot_framebuffer ? boot_framebuffer->bpp()    :   32;
+
+			TRY(set_fb_var_info({
+				.xres = xres,
+				.yres = yres,
+				.xres_virt = xres,
+				.yres_virt = yres * 2,
+				.xoff = 0,
+				.yoff = 0,
+				.bpp = bpp,
+			}));
+		}
+
+		if (boot_framebuffer)
+		{
+			// FIXME: check that this is actually the boot framebuffer
+			TRY(boot_framebuffer->set_bga_controller(this));
+		}
+		else
+		{
+			auto framebuffer_device = TRY(FramebufferDevice::create(this));
+			auto fb_terminal_driver = TRY(FramebufferTerminalDriver::create(framebuffer_device));
+
+			// FIXME: query vtty instead of checking the current tty
+			if (auto tty = TTY::current(); tty && tty->is_vtty())
+				TRY(static_cast<VirtualTTY*>(tty.ptr())->set_terminal_driver(fb_terminal_driver));
+			else
+				TRY(VirtualTTY::create(fb_terminal_driver));
+		}
+
+		return {};
+	}
+
+	BAN::ErrorOr<void> BGAController::set_fb_var_info(const fb_var_info& var_info)
+	{
+		const auto validate_u16 = [](auto value) -> BAN::ErrorOr<void> {
+			if (value > BAN::numeric_limits<uint16_t>::max())
+				return BAN::Error::from_errno(EINVAL);
+			return {};
+		};
+		TRY(validate_u16(var_info.xres));
+		TRY(validate_u16(var_info.xres));
+		TRY(validate_u16(var_info.xres_virt));
+		TRY(validate_u16(var_info.yres_virt));
+		TRY(validate_u16(var_info.xoff));
+		TRY(validate_u16(var_info.yoff));
+		TRY(validate_u16(var_info.bpp));
+
+		if (var_info.xres + var_info.xoff > var_info.xres_virt)
+			return BAN::Error::from_errno(EINVAL);
+		if (var_info.yres + var_info.yoff > var_info.yres_virt)
+			return BAN::Error::from_errno(EINVAL);
+		if (static_cast<uint64_t>(var_info.xres_virt) * var_info.yres_virt * var_info.bpp / 8 > m_lfb_bar->size())
+			return BAN::Error::from_errno(EINVAL);
+
+		SpinLockGuard _(m_lock);
+
+		const bool needs_reconf =
+			(m_var_info.xres      != var_info.xres) ||
+			(m_var_info.yres      != var_info.yres) ||
+			(m_var_info.xres_virt != var_info.xres_virt) ||
+			(m_var_info.yres_virt != var_info.yres_virt) ||
+			(m_var_info.bpp       != var_info.bpp);
+
+		if (!needs_reconf)
+		{
+			write_reg(BGA_REG_XOFF, var_info.xoff);
+			write_reg(BGA_REG_YOFF, var_info.yoff);
+			m_var_info.xoff = var_info.xoff;
+			m_var_info.yoff = var_info.yoff;
+			return {};
+		}
+
+		const auto old_enable = read_reg(BGA_REG_ENABLE);
+		write_reg(BGA_REG_ENABLE, 0);
+
+		write_reg(BGA_REG_BPP,       var_info.bpp);
+		write_reg(BGA_REG_XRES,      var_info.xres);
+		write_reg(BGA_REG_YRES,      var_info.yres);
+		write_reg(BGA_REG_XRES_VIRT, var_info.xres_virt);
+		write_reg(BGA_REG_YRES_VIRT, var_info.yres_virt);
+
+		const bool valid_config =
+			read_reg(BGA_REG_BPP)       == var_info.bpp &&
+			read_reg(BGA_REG_XRES)      == var_info.xres &&
+			read_reg(BGA_REG_YRES)      == var_info.yres &&
+			read_reg(BGA_REG_XRES_VIRT) == var_info.xres_virt;
+
+		if (!valid_config)
+		{
+			write_reg(BGA_REG_BPP,       m_var_info.bpp);
+			write_reg(BGA_REG_XRES,      m_var_info.xres);
+			write_reg(BGA_REG_YRES,      m_var_info.yres);
+			write_reg(BGA_REG_XRES_VIRT, m_var_info.xres_virt);
+			write_reg(BGA_REG_YRES_VIRT, m_var_info.yres_virt);
+
+			write_reg(BGA_REG_ENABLE, BGA_ENABLE_NOCLEARMEM | old_enable);
+
+			return BAN::Error::from_errno(EINVAL);
+		}
+
+		write_reg(BGA_REG_ENABLE, BGA_ENABLE_LFB_ENABLE | BGA_ENABLE_ENABLE);
+
+		// NOTE: At least qemu only sets virtual yres on enable and it will be
+		//       maximum that fits within vram. Our initial bounds check should
+		//       make sure this always succeeds
+		ASSERT(read_reg(BGA_REG_YRES_VIRT) >= var_info.yres_virt);
+
+		write_reg(BGA_REG_XOFF, var_info.xoff);
+		write_reg(BGA_REG_YOFF, var_info.yoff);
+
+		m_var_info = var_info;
+		m_fix_info.pitch = m_var_info.xres_virt * m_var_info.bpp / 8;
+
+		return {};
+	}
+
+	void BGAController::write_reg(uint16_t reg, uint16_t value)
+	{
+		SpinLockGuard _(m_lock);
+		IO::outw(BGA_IO_PORT_INDEX, reg);
+		IO::outw(BGA_IO_PORT_DATA, value);
+	}
+
+	uint16_t BGAController::read_reg(uint16_t reg)
+	{
+		SpinLockGuard _(m_lock);
+		IO::outw(BGA_IO_PORT_INDEX, reg);
+		return IO::inw(BGA_IO_PORT_DATA);
+	}
+
+	fb_fix_info BGAController::get_fb_fix_info() const
+	{
+		SpinLockGuard _(m_lock);
+		return m_fix_info;
+	}
+
+	fb_var_info BGAController::get_fb_var_info() const
+	{
+		SpinLockGuard _(m_lock);
+		return m_var_info;
+	}
+
+}

kernel/kernel/IDT.cpp

@@ -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>

kernel/kernel/Input/InputDevice.cpp

@@ -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>
@@ -305,8 +307,7 @@ namespace Kernel
 	BAN::ErrorOr<void> KeyboardDevice::initialize_tty_thread()
 	{
 		auto* thread = TRY(Thread::create_kernel(tty_keyboard_thread, nullptr));
-		ASSERT(thread);
+		Processor::scheduler().add_thread(thread);
-		TRY(Processor::scheduler().add_thread(thread));
 		return {};
 	}
 

kernel/kernel/Input/PS2/Controller.cpp

@@ -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
@@ -317,11 +319,18 @@ namespace Kernel::Input
 					if (crs_obj == nullptr)
 						return PS2ResourceSetting {};
 
+					const auto crs = TRY(ACPI::AML::evaluate_node(crs_path, crs_obj->node));
+					if (crs.type != ACPI::AML::Node::Type::Buffer)
+					{
+						dwarnln("PS/2 _CRS is not a buffer, but {}", crs);
+						return BAN::Error::from_errno(EINVAL);
+					}
+
 					PS2ResourceSetting result;
 					result.type = type;
 
 					BAN::Optional<ACPI::ResourceData> data;
-					ACPI::ResourceParser parser({ crs_obj->node.as.str_buf->bytes, static_cast<size_t>(crs_obj->node.as.str_buf->size) });
+					ACPI::ResourceParser parser({ crs.as.str_buf->bytes, static_cast<size_t>(crs.as.str_buf->size) });
 					while ((data = parser.get_next()).has_value())
 					{
 						switch (data->type)
@@ -345,7 +354,7 @@ 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 +395,11 @@ 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)
+				MUST(acpi_devices.resize(2));
-					acpi_devices.pop_back();
 			}
 
 			if (acpi_devices.size() == 2)
@@ -419,34 +425,35 @@ namespace Kernel::Input
 				}
 			}
 
-			BAN::Optional<uint16_t> command_port;
+			if (!acpi_devices.empty())
-			command_port = acpi_devices[0].command_port;
-			if (!command_port.has_value() && acpi_devices.size() >= 2)
-				command_port = acpi_devices[1].command_port;
-			if (command_port.has_value())
-				m_command_port = command_port.value();
-
-			BAN::Optional<uint16_t> data_port;
-			data_port = acpi_devices[0].data_port;
-			if (!data_port.has_value() && acpi_devices.size() >= 2)
-				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)
 			{
-				devices[1] = {
+				BAN::Optional<uint16_t> command_port;
-					.type      = acpi_devices[1].type,
+				command_port = acpi_devices[0].command_port;
-					.interrupt = acpi_devices[1].irq.value_or(PS2::IRQ::DEVICE1)
+				if (!command_port.has_value() && acpi_devices.size() >= 2)
+					command_port = acpi_devices[1].command_port;
+				if (command_port.has_value())
+					m_command_port = command_port.value();
+
+				BAN::Optional<uint16_t> data_port;
+				data_port = acpi_devices[0].data_port;
+				if (!data_port.has_value() && acpi_devices.size() >= 2)
+					data_port = acpi_devices[1].data_port;
+				if (data_port.has_value())
+					m_data_port = data_port.value();
+			}
+
+			for (size_t i = 0; i < acpi_devices.size(); i++)
+			{
+				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,
@@ -551,7 +558,7 @@ namespace Kernel::Input
 				static_cast<PS2DeviceInitInfo*>(info)->controller->device_initialize_task(info);
 			}, &info
 		));
-		TRY(Processor::scheduler().add_thread(init_thread));
+		Processor::scheduler().add_thread(init_thread);
 
 		while (!info.thread_started)
 			Processor::pause();

kernel/kernel/Input/PS2/Keyboard.cpp

@@ -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");
+			dwarnln("Could not detect scancode set, assuming 1");
-			m_scancode_set = 2;
+			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;

kernel/kernel/Lock/Mutex.cpp

@@ -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();
+	}
+
+}

kernel/kernel/Lock/RWLock.cpp

@@ -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();
+	}
+
+}

kernel/kernel/Memory/FileBackedRegion.cpp

@@ -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)

kernel/kernel/Memory/PhysicalRange.cpp

@@ -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;

kernel/kernel/Memory/SharedMemoryObject.cpp

@@ -87,13 +87,24 @@ namespace Kernel
 				return BAN::Error::from_errno(ENOENT);
 		}
 
+		if (size == 0)
+			return BAN::Error::from_errno(EINVAL);
+
 		const auto& process = Process::current();
 		const uid_t uid = process.credentials().euid();
 		const gid_t gid = process.credentials().egid();
 		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 +120,7 @@ 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));
+		TRY(object->paddrs.resize(BAN::Math::div_round_up<size_t>(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();
 
 		if (key != IPC_PRIVATE)
 			TRY(m_ids.insert(key, shmid));
@@ -222,14 +227,16 @@ 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)

kernel/kernel/Memory/kmalloc.cpp

@@ -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;
 		}
 

kernel/kernel/Networking/E1000/E1000.cpp

@@ -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
 {
@@ -93,11 +95,8 @@ namespace Kernel
 		auto* thread = TRY(Thread::create_kernel([](void* e1000_ptr) {
 			static_cast<E1000*>(e1000_ptr)->receive_thread();
 		}, this));
-		if (auto ret = Processor::scheduler().add_thread(thread); ret.is_error())
+
-		{
+		Processor::scheduler().add_thread(thread);
-			delete thread;
-			return ret.release_error();
-		}
 		m_thread_is_dead = false;
 
 		return {};

kernel/kernel/Networking/Loopback.cpp

@@ -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
 {
@@ -23,11 +25,8 @@ namespace Kernel
 		auto* thread = TRY(Thread::create_kernel([](void* loopback_ptr) {
 			static_cast<LoopbackInterface*>(loopback_ptr)->receive_thread();
 		}, loopback_ptr));
-		if (auto ret = Processor::scheduler().add_thread(thread); ret.is_error())
+
-		{
+		Processor::scheduler().add_thread(thread);
-			delete thread;
-			return ret.release_error();
-		}
 		loopback->m_thread_is_dead = false;
 
 		loopback->set_ipv4_address({ 127, 0, 0, 1 });

kernel/kernel/Networking/RTL8169/RTL8169.cpp

@@ -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
@@ -73,11 +75,8 @@ namespace Kernel
 		auto* thread = TRY(Thread::create_kernel([](void* rtl8169_ptr) {
 			static_cast<RTL8169*>(rtl8169_ptr)->receive_thread();
 		}, this));
-		if (auto ret = Processor::scheduler().add_thread(thread); ret.is_error())
+
-		{
+		Processor::scheduler().add_thread(thread);
-			delete thread;
-			return ret.release_error();
-		}
 		m_rx_thread_is_dead = false;
 
 		return {};

kernel/kernel/Networking/TCPSocket.cpp

@@ -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>
@@ -44,7 +45,7 @@ namespace Kernel
 				reinterpret_cast<TCPSocket*>(socket_ptr)->process_task();
 			}, socket.ptr()
 		));
-		TRY(Processor::scheduler().add_thread(socket->m_thread));
+		Processor::scheduler().add_thread(socket->m_thread);
 		// hack to keep socket alive until its process starts
 		socket->ref();
 		return socket;

kernel/kernel/Networking/UNIX/Socket.cpp

@@ -21,7 +21,7 @@ namespace Kernel
 
 	static BAN::ErrorOr<BAN::StringView> validate_sockaddr_un(const sockaddr* address, socklen_t address_len)
 	{
-		if (address_len < static_cast<socklen_t>(sizeof(sa_family_t)))
+		if (address_len <= static_cast<socklen_t>(sizeof(sa_family_t)))
 			return BAN::Error::from_errno(EINVAL);
 		if (address_len > static_cast<socklen_t>(sizeof(sockaddr_un)))
 			address_len = sizeof(sockaddr_un);
@@ -30,13 +30,12 @@ namespace Kernel
 		if (sockaddr_un.sun_family != AF_UNIX)
 			return BAN::Error::from_errno(EINVAL);
 
-		size_t length = 0;
+		auto sun_path = BAN::StringView { sockaddr_un.sun_path, address_len - sizeof(sa_family_t) };
-		while (length < sizeof(sockaddr_un::sun_path) && sockaddr_un.sun_path[length])
+		if (const auto null_idx = sun_path.find('\0'); null_idx.has_value())
-			length++;
+			sun_path = sun_path.substring(0, null_idx.value());
-		if (length >= sizeof(sockaddr_un::sun_path))
+		if (sun_path.empty())
-			return BAN::Error::from_errno(ENAMETOOLONG);
+			return BAN::Error::from_errno(EINVAL);
-
+		return sun_path;
-		return BAN::StringView { sockaddr_un.sun_path, length };
 	}
 
 	BAN::ErrorOr<BAN::RefPtr<UnixDomainSocket>> UnixDomainSocket::create(Socket::Type socket_type, const Socket::Info& info)
@@ -264,14 +263,15 @@ namespace Kernel
 	BAN::ErrorOr<void> UnixDomainSocket::bind_impl(const sockaddr* address, socklen_t address_len)
 	{
 		const auto sun_path = TRY(validate_sockaddr_un(address, address_len));
-		if (sun_path.empty())
+
-			return BAN::Error::from_errno(EINVAL);
+		BAN::String sun_path_nul;
+		TRY(sun_path_nul.append(sun_path));
 
 		// FIXME: This feels sketchy
 		auto parent_file = sun_path.front() == '/'
 			? TRY(Process::current().root_file().clone())
 			: TRY(Process::current().working_directory().clone());
-		if (auto ret = Process::current().create_file(AT_FDCWD, sun_path.data(), 0755 | Inode::Mode::IFSOCK); ret.is_error())
+		if (auto ret = Process::current().create_file(AT_FDCWD, sun_path_nul.data(), 0755 | Inode::Mode::IFSOCK); ret.is_error())
 		{
 			if (ret.error().get_error_code() == EEXIST)
 				return BAN::Error::from_errno(EADDRINUSE);

kernel/kernel/OpenFileDescriptorSet.cpp

@@ -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())

kernel/kernel/PCI.cpp

@@ -3,6 +3,7 @@
 #include <kernel/ACPI/ACPI.h>
 #include <kernel/APIC.h>
 #include <kernel/Audio/Controller.h>
+#include <kernel/Graphics/BGA.h>
 #include <kernel/IDT.h>
 #include <kernel/IO.h>
 #include <kernel/Memory/PageTable.h>
@@ -258,6 +259,13 @@ namespace Kernel::PCI
 		for_each_device(
 			[&](PCI::Device& pci_device)
 			{
+				if (pci_device.vendor_id() == 0x1234 && pci_device.device_id() == 0x1111)
+				{
+					if (auto ret = BGAController::create(pci_device); ret.is_error())
+						dprintln("BGA: {}", ret.error());
+					return;
+				}
+
 				switch (pci_device.class_code())
 				{
 					case 0x01:
@@ -912,10 +920,10 @@ namespace Kernel::PCI
 		}
 
 		// disable io/mem space while reading bar
-		uint16_t command = device.read_word(PCI_REG_COMMAND);
+		const uint16_t command = device.read_word(PCI_REG_COMMAND);
 		device.write_word(PCI_REG_COMMAND, command & ~(PCI_CMD_IO_SPACE | PCI_CMD_MEM_SPACE));
 
-		uint8_t offset = 0x10 + bar_num * 4;
+		const uint8_t offset = 0x10 + bar_num * 4;
 
 		uint64_t addr = device.read_dword(offset);
 
@@ -924,6 +932,13 @@ namespace Kernel::PCI
 		size = ~size + 1;
 		device.write_dword(offset, addr);
 
+		if (size == 0)
+		{
+			device.write_word(PCI_REG_COMMAND, command);
+			dwarnln("BAR{} has size 0", bar_num);
+			return BAN::Error::from_errno(EINVAL);
+		}
+
 		// determine bar type
 		BarType type = BarType::INVALID;
 		if (addr & 1)
@@ -956,8 +971,7 @@ namespace Kernel::PCI
 		TRY(region->initialize());
 
 		// restore old command register and enable correct IO/MEM space
-		command |= (type == BarType::IO) ? PCI_CMD_IO_SPACE : PCI_CMD_MEM_SPACE;
+		device.write_word(PCI_REG_COMMAND, command | ((type == BarType::IO) ? PCI_CMD_IO_SPACE : PCI_CMD_MEM_SPACE));
-		device.write_word(PCI_REG_COMMAND, command);
 
 #if DEBUG_PCI
 		dprintln("created BAR region for PCI {2H}:{2H}.{2H}",
@@ -997,7 +1011,7 @@ namespace Kernel::PCI
 		if (m_type == BarType::IO)
 			return {};
 
-		size_t needed_pages = BAN::Math::div_round_up<size_t>(m_size, PAGE_SIZE);
+		const size_t needed_pages = BAN::Math::div_round_up<size_t>(m_size, PAGE_SIZE);
 		m_vaddr = PageTable::kernel().reserve_free_contiguous_pages(needed_pages, KERNEL_OFFSET);
 		if (m_vaddr == 0)
 			return BAN::Error::from_errno(ENOMEM);

kernel/kernel/PIC.cpp

@@ -1,4 +1,5 @@
 #include <kernel/IDT.h>
+#include <kernel/InterruptNumbers.h>
 #include <kernel/IO.h>
 #include <kernel/PIC.h>
 

kernel/kernel/Process.cpp

@@ -237,9 +237,8 @@ namespace Kernel
 #endif
 		}
 
-		// NOTE: make sure the last two `MUST`s don't fail
+		// NOTE: make sure the last `MUST` doesn't fail
 		TRY(process->m_threads.reserve(1));
-		TRY(Processor::scheduler().bind_thread_to_processor(thread, Processor::current_id()));
 
 		{
 			SpinLockGuard _(s_process_lock);
@@ -247,7 +246,7 @@ namespace Kernel
 		}
 
 		MUST(process->m_threads.push_back(thread));
-		MUST(Processor::scheduler().add_thread(thread));
+		Processor::scheduler().add_thread(thread);
 
 		process_deleter.disable();
 		thread_deleter.disable();
@@ -696,19 +695,24 @@ namespace Kernel
 
 	size_t Process::proc_cputime(off_t offset, BAN::ByteSpan buffer) const
 	{
-		const uint64_t cpu_time_ns = [this] {
+		uint64_t user_ns { 0 }, system_ns { 0 };
-			uint64_t cpu_time_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));
+		auto data = MUST(BAN::String::formatted("{} {} {}", user_ns + system_ns, user_ns, system_ns));
-		if (static_cast<size_t>(offset) >= data.size() + 1)
+		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;
 	}
@@ -872,9 +876,8 @@ namespace Kernel
 		Thread* thread = TRY(Thread::current().clone(forked, sp, ip));
 		BAN::ScopeGuard thread_deleter([thread] { delete thread; });
 
-		// NOTE: make sure the last two `MUST`s don't fail
+		// NOTE: make sure the last `MUST` doesn't fail
 		TRY(forked->m_threads.reserve(1));
-		TRY(Processor::scheduler().bind_thread_to_processor(thread, Processor::current_id()));
 
 		{
 			SpinLockGuard _(s_process_lock);
@@ -900,7 +903,7 @@ namespace Kernel
 		ASSERT(this == &Process::current());
 
 		MUST(forked->m_threads.push_back(thread));
-		MUST(Processor::scheduler().add_thread(thread));
+		Processor::scheduler().add_thread(thread);
 
 		process_deleter.disable();
 		thread_deleter.disable();
@@ -1067,14 +1070,6 @@ namespace Kernel
 #endif
 			}
 
-			// NOTE: bind new thread to this processor so it wont be rescheduled before end of this function
-			//       and so that adding the thread to the scheduler cannot fail
-			if (auto ret = Scheduler::bind_thread_to_processor(new_thread, Processor::current_id()); ret.is_error())
-			{
-				Processor::set_interrupt_state(InterruptState::Enabled);
-				delete new_thread;
-				return ret.release_error();
-			}
 
 			RWLockWRGuard wr_guard(m_memory_region_lock);
 
@@ -1098,7 +1093,9 @@ namespace Kernel
 			m_threads.front()->m_process = nullptr;
 			m_threads.front()->give_keep_alive_page_table(BAN::move(m_page_table));
 
-			MUST(Processor::scheduler().add_thread(new_thread));
+			// NOTE: bind new thread to this processor so it wont be rescheduled before end of this function
+			Scheduler::bind_thread_to_processor(new_thread, Processor::current_id());
+			Processor::scheduler().add_thread(new_thread);
 			m_threads.front() = new_thread;
 
 			for (size_t i = 0; i < sizeof(m_signal_handlers) / sizeof(*m_signal_handlers); i++)
@@ -1265,18 +1262,12 @@ namespace Kernel
 				return BAN::Error::from_errno(EINVAL);
 		}
 
-		MemoryRegion* value_region = nullptr;
+		BAN::Vector<MemoryRegion*> regions;
-		MemoryRegion* ovalue_region = nullptr;
+		BAN::ScopeGuard _([&] { for (auto* region : regions) region->unpin(); });
-		BAN::ScopeGuard _([&] {
-			if (value_region)
-				value_region->unpin();
-			if (ovalue_region)
-				ovalue_region->unpin();
-		});
 
-		value_region = TRY(validate_and_pin_pointer_access(value, sizeof(itimerval), false));
+		TRY(validate_and_pin_pointer_access(value, sizeof(itimerval), false, regions));
 		if (ovalue != nullptr)
-			ovalue_region = TRY(validate_and_pin_pointer_access(ovalue, sizeof(itimerval), true));
+			TRY(validate_and_pin_pointer_access(ovalue, sizeof(itimerval), true, regions));
 
 		{
 			SpinLockGuard _(s_process_lock);
@@ -1470,8 +1461,10 @@ namespace Kernel
 			return 0;
 		}
 
-		auto* buffer_region = TRY(validate_and_pin_pointer_access(buffer, count, true));
+		BAN::Vector<MemoryRegion*> regions;
-		BAN::ScopeGuard _([buffer_region] { buffer_region->unpin(); });
+		BAN::ScopeGuard _([&] { for (auto* region : regions) region->unpin(); });
+		TRY(validate_and_pin_pointer_access(buffer, count, true, regions));
+
 		return TRY(m_open_file_descriptors.read(fd, BAN::ByteSpan(static_cast<uint8_t*>(buffer), count)));
 	}
 
@@ -1483,8 +1476,10 @@ namespace Kernel
 			return 0;
 		}
 
-		auto* buffer_region = TRY(validate_and_pin_pointer_access(buffer, count, false));
+		BAN::Vector<MemoryRegion*> regions;
-		BAN::ScopeGuard _([buffer_region] { buffer_region->unpin(); });
+		BAN::ScopeGuard _([&] { for (auto* region : regions) region->unpin(); });
+		TRY(validate_and_pin_pointer_access(buffer, count, false, regions));
+
 		return TRY(m_open_file_descriptors.write(fd, BAN::ConstByteSpan(static_cast<const uint8_t*>(buffer), count)));
 	}
 
@@ -1674,8 +1669,9 @@ namespace Kernel
 	{
 		auto inode = TRY(m_open_file_descriptors.inode_of(fd));
 
-		auto* buffer_region = TRY(validate_and_pin_pointer_access(buffer, count, true));
+		BAN::Vector<MemoryRegion*> regions;
-		BAN::ScopeGuard _([buffer_region] { buffer_region->unpin(); });
+		BAN::ScopeGuard _([&] { for (auto* region : regions) region->unpin(); });
+		TRY(validate_and_pin_pointer_access(buffer, count, true, regions));
 
 		return TRY(inode->read(offset, { reinterpret_cast<uint8_t*>(buffer), count }));
 	}
@@ -1684,8 +1680,9 @@ namespace Kernel
 	{
 		auto inode = TRY(m_open_file_descriptors.inode_of(fd));
 
-		auto* buffer_region = TRY(validate_and_pin_pointer_access(buffer, count, false));
+		BAN::Vector<MemoryRegion*> regions;
-		BAN::ScopeGuard _([buffer_region] { buffer_region->unpin(); });
+		BAN::ScopeGuard _([&] { for (auto* region : regions) region->unpin(); });
+		TRY(validate_and_pin_pointer_access(buffer, count, false, regions));
 
 		return TRY(inode->write(offset, { reinterpret_cast<const uint8_t*>(buffer), count }));	}
 
@@ -1705,10 +1702,7 @@ namespace Kernel
 		{
 			LockGuard _(m_process_lock);
 			if (!m_credentials.is_superuser() && inode->uid() != m_credentials.euid())
-			{
-				dwarnln("cannot chmod uid {} vs {}", inode->uid(), m_credentials.euid());
 				return BAN::Error::from_errno(EPERM);
-			}
 		}
 
 		TRY(inode->chmod(mode & ~S_IFMASK));
@@ -1880,8 +1874,9 @@ namespace Kernel
 		if (!inode->mode().ifsock())
 			return BAN::Error::from_errno(ENOTSOCK);
 
-		auto* buffer = TRY(validate_and_pin_pointer_access(user_option_value, option_len, true));
+		BAN::Vector<MemoryRegion*> regions;
-		BAN::ScopeGuard _([buffer] { buffer->unpin(); });
+		BAN::ScopeGuard _([&] { for (auto* region : regions) region->unpin(); });
+		TRY(validate_and_pin_pointer_access(user_option_value, option_len, true, regions));
 
 		TRY(inode->getsockopt(level, option_name, user_option_value, &option_len));
 		TRY(write_to_user(user_option_len, &option_len, sizeof(socklen_t)));
@@ -1898,8 +1893,9 @@ namespace Kernel
 		if (!inode->mode().ifsock())
 			return BAN::Error::from_errno(ENOTSOCK);
 
-		auto* buffer = TRY(validate_and_pin_pointer_access(user_option_value, option_len, false));
+		BAN::Vector<MemoryRegion*> regions;
-		BAN::ScopeGuard _([buffer] { buffer->unpin(); });
+		BAN::ScopeGuard _([&] { for (auto* region : regions) region->unpin(); });
+		TRY(validate_and_pin_pointer_access(user_option_value, option_len, false, regions));
 
 		TRY(inode->setsockopt(level, option_name, user_option_value, option_len));
 
@@ -1913,20 +1909,13 @@ namespace Kernel
 		if (flags & ~(SOCK_NONBLOCK | SOCK_CLOEXEC))
 			return BAN::Error::from_errno(EINVAL);
 
-		MemoryRegion* address_region1 = nullptr;
+		BAN::Vector<MemoryRegion*> regions;
-		MemoryRegion* address_region2 = nullptr;
+		BAN::ScopeGuard _([&] { for (auto* region : regions) region->unpin(); });
-
-		BAN::ScopeGuard _([&] {
-			if (address_region1)
-				address_region1->unpin();
-			if (address_region2)
-				address_region2->unpin();
-		});
 
 		if (address_len)
 		{
-			address_region1 = TRY(validate_and_pin_pointer_access(address_len, sizeof(address_len), true));
+			TRY(validate_and_pin_pointer_access(address_len, sizeof(address_len), true, regions));
-			address_region2 = TRY(validate_and_pin_pointer_access(address, *address_len, true));
+			TRY(validate_and_pin_pointer_access(address,           *address_len,  true, regions));
 		}
 
 		auto inode = TRY(m_open_file_descriptors.inode_of(socket));
@@ -1993,23 +1982,17 @@ namespace Kernel
 		TRY(read_from_user(user_message, &message, sizeof(msghdr)));
 
 		BAN::Vector<MemoryRegion*> regions;
-		TRY(regions.reserve(!!message.msg_name + !!message.msg_control + !!message.msg_iov));
+		BAN::ScopeGuard _([&] { for (auto* region : regions) region->unpin(); });
-
-		BAN::ScopeGuard _([&regions] {
-			for (auto* region : regions)
-				region->unpin();
-		});
 
 		if (message.msg_name)
-			TRY(regions.push_back(TRY(validate_and_pin_pointer_access(message.msg_name, message.msg_namelen, true))));
+			TRY(validate_and_pin_pointer_access(message.msg_name, message.msg_namelen, true, regions));
 		if (message.msg_control)
-			TRY(regions.push_back(TRY(validate_and_pin_pointer_access(message.msg_control, message.msg_controllen, true))));
+			TRY(validate_and_pin_pointer_access(message.msg_control, message.msg_controllen, true, regions));
 		if (message.msg_iov)
 		{
-			TRY(regions.push_back(TRY(validate_and_pin_pointer_access(message.msg_iov, message.msg_iovlen * sizeof(iovec), true))));
+			TRY(validate_and_pin_pointer_access(message.msg_iov, message.msg_iovlen * sizeof(iovec), true, regions));
-			TRY(regions.reserve(regions.size() + message.msg_iovlen));
 			for (int i = 0; i < message.msg_iovlen; i++)
-				TRY(regions.push_back(TRY(validate_and_pin_pointer_access(message.msg_iov[i].iov_base, message.msg_iov[i].iov_len, true))));
+				TRY(validate_and_pin_pointer_access(message.msg_iov[i].iov_base, message.msg_iov[i].iov_len, true, regions));
 		}
 
 		const auto ret = TRY(m_open_file_descriptors.recvmsg(socket, message, flags));
@@ -2025,23 +2008,18 @@ namespace Kernel
 		TRY(read_from_user(user_message, &message, sizeof(msghdr)));
 
 		BAN::Vector<MemoryRegion*> regions;
-		TRY(regions.reserve(!!message.msg_name + !!message.msg_control + !!message.msg_iov));
+		BAN::ScopeGuard _([&regions] { for (auto* region : regions) region->unpin(); });
-
-		BAN::ScopeGuard _([&regions] {
-			for (auto* region : regions)
-				region->unpin();
-		});
 
 		if (message.msg_name)
-			TRY(regions.push_back(TRY(validate_and_pin_pointer_access(message.msg_name, message.msg_namelen, false))));
+			TRY(validate_and_pin_pointer_access(message.msg_name, message.msg_namelen, false, regions));
 		if (message.msg_control)
-			TRY(regions.push_back(TRY(validate_and_pin_pointer_access(message.msg_control, message.msg_controllen, false))));
+			TRY(validate_and_pin_pointer_access(message.msg_control, message.msg_controllen, false, regions));
 		if (message.msg_iov)
 		{
-			TRY(regions.push_back(TRY(validate_and_pin_pointer_access(message.msg_iov, message.msg_iovlen * sizeof(iovec), false))));
+			TRY(validate_and_pin_pointer_access(message.msg_iov, message.msg_iovlen * sizeof(iovec), false, regions));
 			TRY(regions.reserve(regions.size() + message.msg_iovlen));
 			for (int i = 0; i < message.msg_iovlen; i++)
-				TRY(regions.push_back(TRY(validate_and_pin_pointer_access(message.msg_iov[i].iov_base, message.msg_iov[i].iov_len, false))));
+				TRY(validate_and_pin_pointer_access(message.msg_iov[i].iov_base, message.msg_iov[i].iov_len, false, regions));
 		}
 
 		return TRY(m_open_file_descriptors.sendmsg(socket, message, flags));
@@ -2176,8 +2154,9 @@ namespace Kernel
 
 	BAN::ErrorOr<long> Process::sys_ppoll(pollfd* fds, nfds_t nfds, const timespec* user_timeout, const sigset_t* user_sigmask)
 	{
-		auto* fds_region = TRY(validate_and_pin_pointer_access(fds, nfds * sizeof(pollfd), true));
+		BAN::Vector<MemoryRegion*> regions;
-		BAN::ScopeGuard _([fds_region] { fds_region->unpin(); });
+		BAN::ScopeGuard _([&] { for (auto* region : regions) region->unpin(); });
+		TRY(validate_and_pin_pointer_access(fds, nfds * sizeof(pollfd), true, regions));
 
 		const auto old_sigmask = Thread::current().m_signal_block_mask;
 		if (user_sigmask != nullptr)
@@ -2359,8 +2338,9 @@ namespace Kernel
 				timeout.tv_nsec;
 		}
 
-		auto* events_region = TRY(validate_and_pin_pointer_access(events, maxevents * sizeof(epoll_event), true));
+		BAN::Vector<MemoryRegion*> regions;
-		BAN::ScopeGuard _([events_region] { events_region->unpin(); });
+		BAN::ScopeGuard _([&] { for (auto* region : regions) region->unpin(); });
+		TRY(validate_and_pin_pointer_access(events, maxevents * sizeof(epoll_event), true, regions));
 
 		const auto old_sigmask = Thread::current().m_signal_block_mask;
 		if (user_sigmask)
@@ -2573,8 +2553,10 @@ namespace Kernel
 		if (BAN::Math::will_multiplication_overflow(list_len, sizeof(struct dirent)))
 			return BAN::Error::from_errno(EOVERFLOW);
 
-		auto* list_region = TRY(validate_and_pin_pointer_access(list, list_len * sizeof(struct dirent), true));
+		BAN::Vector<MemoryRegion*> regions;
-		BAN::ScopeGuard _([list_region] { list_region->unpin(); });
+		BAN::ScopeGuard _([&] { for (auto* region : regions) region->unpin(); });
+		TRY(validate_and_pin_pointer_access(list, list_len * sizeof(struct dirent), true, regions));
+
 		return TRY(m_open_file_descriptors.read_dir_entries(fd, list, list_len));
 	}
 
@@ -3054,7 +3036,7 @@ 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 +3045,7 @@ 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),
@@ -3411,8 +3393,9 @@ namespace Kernel
 		const bool is_private = (op & FUTEX_PRIVATE);
 		op &= ~(FUTEX_PRIVATE | FUTEX_REALTIME);
 
-		auto* buffer_region = TRY(validate_and_pin_pointer_access(addr, sizeof(uint32_t), false));
+		BAN::Vector<MemoryRegion*> regions;
-		BAN::ScopeGuard pin_guard([buffer_region] { buffer_region->unpin(); });
+		BAN::ScopeGuard _0([&] { for (auto* region : regions) region->unpin(); });
+		TRY(validate_and_pin_pointer_access(addr, sizeof(uint32_t), false, regions));
 
 		const paddr_t paddr = m_page_table->physical_address_of(vaddr & PAGE_ADDR_MASK) | (vaddr & ~PAGE_ADDR_MASK);
 		ASSERT(paddr != 0);
@@ -3473,7 +3456,7 @@ namespace Kernel
 			}
 		}
 
-		LockGuard _(futex->mutex);
+		LockGuard _1(futex->mutex);
 
 		switch (op)
 		{
@@ -3556,8 +3539,9 @@ namespace Kernel
 		if (stack_vaddr % PAGE_SIZE || stack_size % PAGE_SIZE)
 			return BAN::Error::from_errno(EINVAL);
 
-		auto* memory_region = TRY(validate_and_pin_pointer_access(stack_base, stack_size, true));
+		BAN::Vector<MemoryRegion*> regions;
-		BAN::ScopeGuard _0([memory_region] { if (memory_region) memory_region->unpin(); });
+		BAN::ScopeGuard _0([&] { for (auto* region : regions) region->unpin(); });
+		TRY(validate_and_pin_pointer_access(stack_base, stack_size, true, regions));
 
 		const vaddr_t initial_stack_pointer = stack_vaddr + stack_size - sizeof(void*);
 		*reinterpret_cast<void**>(initial_stack_pointer) = arg;
@@ -3575,12 +3559,7 @@ namespace Kernel
 		LockGuard _1(m_process_lock);
 
 		TRY(m_threads.push_back(thread));
-		if (auto ret = Processor::scheduler().add_thread(thread); ret.is_error())
+		Processor::scheduler().add_thread(thread);
-		{
-			m_threads.pop_back();
-			delete thread;
-			return ret.release_error();
-		}
 
 		return thread->tid();
 	}
@@ -4030,13 +4009,14 @@ namespace Kernel
 		if (BAN::Math::will_multiplication_overflow(count, sizeof(gid_t)))
 			return BAN::Error::from_errno(EOVERFLOW);
 
-		LockGuard _(m_process_lock);
+		LockGuard _0(m_process_lock);
 
 		if (!m_credentials.is_superuser())
 			return BAN::Error::from_errno(EPERM);
 
-		auto* region = TRY(validate_and_pin_pointer_access(groups, count * sizeof(gid_t), false));
+		BAN::Vector<MemoryRegion*> regions;
-		BAN::ScopeGuard pin_guard([region] { region->unpin(); });
+		BAN::ScopeGuard _1([&] { for (auto* region : regions) region->unpin(); });
+		TRY(validate_and_pin_pointer_access(groups, count * sizeof(gid_t), false, regions));
 
 		TRY(m_credentials.set_groups({ groups, count }));
 
@@ -4101,26 +4081,23 @@ namespace Kernel
 		return region->allocate_page_containing(address, wants_write);
 	}
 
-	BAN::ErrorOr<MemoryRegion*> Process::validate_and_pin_pointer_access(const void* ptr, size_t size, bool needs_write)
+	BAN::ErrorOr<void> Process::validate_and_pin_pointer_access(const void* ptr, size_t size, bool needs_write, BAN::Vector<MemoryRegion*>& regions)
 	{
-		// TODO: allow pinning multiple regions?
+		vaddr_t vaddr = reinterpret_cast<vaddr_t>(ptr);
-
-		const vaddr_t user_vaddr = reinterpret_cast<vaddr_t>(ptr);
 
 		{
 			RWLockRDGuard _(m_memory_region_lock);
 
-			const size_t first_index = find_mapped_region(user_vaddr);
+			const size_t first_index = find_mapped_region(vaddr);
 			for (size_t i = first_index; i < m_mapped_regions.size(); i++)
 			{
-				auto& region = m_mapped_regions[i];
+				const auto& region = m_mapped_regions[i];
-				if (user_vaddr >= region->vaddr() + region->size())
+				if (vaddr < region->vaddr())
 					break;
-				if (!region->contains_fully(user_vaddr, size))
-					continue;
 
-				const size_t page_count = range_page_count(user_vaddr, size);
+				const vaddr_t min_vaddr_end = BAN::Math::min(vaddr + size, region->vaddr() + region->size());
-				const vaddr_t page_base = user_vaddr & PAGE_ADDR_MASK;
+				const size_t page_count = range_page_count(vaddr, min_vaddr_end - vaddr);
+				const vaddr_t page_base = vaddr & PAGE_ADDR_MASK;
 				for (size_t p = 0; p < page_count; p++)
 				{
 					const auto flags = PageTable::UserSupervisor | (needs_write ? PageTable::ReadWrite : 0) | PageTable::Present;
@@ -4128,8 +4105,15 @@ namespace Kernel
 						goto validate_and_pin_pointer_access_with_allocation;
 				}
 
+				TRY(regions.push_back(region.ptr()));
 				region->pin();
-				return region.ptr();
+
+				if (region->contains(vaddr + size - 1))
+					return {};
+
+				const vaddr_t next_vaddr = region->vaddr() + region->size();
+				size -= next_vaddr - vaddr;
+				vaddr = next_vaddr;
 			}
 
 			return BAN::Error::from_errno(EFAULT);
@@ -4138,17 +4122,16 @@ namespace Kernel
 	validate_and_pin_pointer_access_with_allocation:
 		RWLockWRGuard _(m_memory_region_lock);
 
-		const size_t first_index = find_mapped_region(user_vaddr);
+		const size_t first_index = find_mapped_region(vaddr);
 		for (size_t i = first_index; i < m_mapped_regions.size(); i++)
 		{
 			auto& region = m_mapped_regions[i];
-			if (user_vaddr >= region->vaddr() + region->size())
+			if (vaddr < region->vaddr())
 				break;
-			if (!region->contains_fully(user_vaddr, size))
-				continue;
 
-			const size_t page_count = range_page_count(user_vaddr, size);
+			const vaddr_t min_vaddr_end = BAN::Math::min(vaddr + size, region->vaddr() + region->size());
-			const vaddr_t page_base = user_vaddr & PAGE_ADDR_MASK;
+			const size_t page_count = range_page_count(vaddr, min_vaddr_end - vaddr);
+			const vaddr_t page_base = vaddr & PAGE_ADDR_MASK;
 			for (size_t p = 0; p < page_count; p++)
 			{
 				const auto flags = PageTable::UserSupervisor | (needs_write ? PageTable::ReadWrite : 0) | PageTable::Present;
@@ -4158,8 +4141,15 @@ namespace Kernel
 					return BAN::Error::from_errno(EFAULT);
 			}
 
+			TRY(regions.push_back(region.ptr()));
 			region->pin();
-			return region.ptr();
+
+			if (region->contains(vaddr + size - 1))
+				return {};
+
+			const vaddr_t next_vaddr = region->vaddr() + region->size();
+			size -= next_vaddr - vaddr;
+			vaddr = next_vaddr;
 		}
 
 		return BAN::Error::from_errno(EFAULT);

kernel/kernel/Processor.cpp

@@ -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>

kernel/kernel/Scheduler.cpp

@@ -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>
 
@@ -33,88 +33,6 @@ namespace Kernel
 	static SpinLock                        s_processor_info_time_lock;
 	static BAN::Array<ProcessorInfo, 0xFF> s_processor_infos;
 
-
-	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 +60,7 @@ 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 +69,7 @@ namespace Kernel
 		{
 			if (info.node == m_current)
 			{
-				info.queue = target_queue;
+				info.list = target_list;
 				has_current = true;
 				break;
 			}
@@ -163,8 +81,10 @@ 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] = {
-			m_most_loaded_threads[index].node  = m_current;
+				.list = target_list,
+				.node = m_current,
+			};
 		}
 
 		BAN::sort::sort(m_most_loaded_threads.begin(), m_most_loaded_threads.end(),
@@ -177,7 +97,7 @@ 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 +105,13 @@ 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 +123,10 @@ 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() = {
-		m_most_loaded_threads.back().queue = nullptr;
+			.list = nullptr,
+			.node = nullptr,
+		};
 	}
 
 	void Scheduler::reschedule(YieldRegisters* yield_registers)
@@ -212,7 +134,7 @@ 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)
@@ -226,18 +148,18 @@ namespace Kernel
 					if (&PageTable::current() != &PageTable::kernel())
 						PageTable::kernel().load();
 					delete m_current->thread;
-					delete m_current;
 					m_thread_count--;
 					break;
 				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 +168,12 @@ 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;
@@ -259,7 +181,6 @@ namespace Kernel
 			if (&PageTable::current() != &PageTable::kernel())
 				PageTable::kernel().load();
 			delete m_current->thread;
-			delete m_current;
 			m_thread_count--;
 		}
 
@@ -273,7 +194,7 @@ 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 +230,8 @@ 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)
+		while (!m_block_list.empty() && current_ns >= m_block_list.front()->wake_time_ns)
-			unblock_thread(m_block_queue.front());
+			unblock_thread(m_block_list.front()->thread);
 	}
 
 	void Scheduler::update_wake_up_deadline()
@@ -324,8 +245,8 @@ namespace Kernel
 			return;
 
 		uint64_t deadline_ns = m_next_reschedule_ns;
-		if (!m_block_queue.empty())
+		if (!m_block_list.empty())
-			deadline_ns = BAN::Math::min(deadline_ns, m_block_queue.front()->wake_time_ns);
+			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 +257,7 @@ 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();
@@ -374,56 +295,6 @@ namespace Kernel
 			update_wake_up_deadline();
 	}
 
-	void Scheduler::unblock_thread(SchedulerQueue::Node* node)
-	{
-		auto state = Processor::get_interrupt_state();
-		Processor::set_interrupt_state(InterruptState::Disabled);
-
-		if (node->processor_id == Processor::current_id())
-		{
-			if (!node->blocked)
-				return;
-			ASSERT(node != m_current);
-			Processor::set_disable_smp_messages(true);
-			m_block_queue.remove_node(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);
-			Processor::set_disable_smp_messages(false);
-		}
-		else
-		{
-			Processor::send_smp_message(node->processor_id, {
-				.type = Processor::SMPMessage::Type::UnblockThread,
-				.unblock_thread = node
-			});
-		}
-
-		Processor::set_interrupt_state(state);
-	}
-
-	void Scheduler::add_thread(SchedulerQueue::Node* node)
-	{
-		auto state = Processor::get_interrupt_state();
-		Processor::set_interrupt_state(InterruptState::Disabled);
-
-		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))
-			update_wake_up_deadline();
-
-		if (auto* thread = node->thread; thread->is_userspace() && thread->has_process())
-			thread->update_processor_index_address();
-
-		m_thread_count++;
-
-		Processor::set_interrupt_state(state);
-	}
-
 	ProcessorID Scheduler::find_least_loaded_processor() const
 	{
 		ProcessorID least_loaded_id        = Processor::current_id();
@@ -482,21 +353,21 @@ 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(
+			m_run_list.walk(
-				[&](SchedulerQueue::Node* node)
+				[](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(
+			m_block_list.walk(
-				[&](SchedulerQueue::Node* node)
+				[](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 +403,7 @@ 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();
@@ -596,20 +467,22 @@ namespace Kernel
 
 			thread_info.node->time_used_ns = 0;
 
-			{
+			if (thread_info.list == &m_run_list)
-				auto& my_queue = (thread_info.queue == &m_run_queue) ? m_run_queue : m_block_queue;
+				m_run_list.pop(thread_info.node);
-				my_queue.remove_node(thread_info.node);
+			else
-				m_thread_count--;
+				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, {
 				.type = Processor::SMPMessage::Type::NewThread,
-				.new_thread = thread_info.node
+				.new_thread = thread_info.node->thread
 			});
 
-			thread_info.node = nullptr;
+			thread_info = {
-			thread_info.queue = nullptr;
+				.list = nullptr,
+				.node = nullptr,
+			};
 
 			if (m_idle_ns == 0)
 				break;
@@ -622,8 +495,8 @@ 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_run_list  .walk([](const SchedulerThreadNode* node, void*) { const_cast<SchedulerThreadNode*>(node)->time_used_ns = 0; }, nullptr);
-		m_block_queue.remove_with_condition([&](SchedulerQueue::Node* node) { node->time_used_ns = 0; return false; });
+		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;
@@ -631,41 +504,52 @@ namespace Kernel
 		m_last_load_balance_ns += s_load_balance_interval_ns;
 	}
 
-	BAN::ErrorOr<void> Scheduler::bind_thread_to_processor(Thread* thread, ProcessorID processor_id)
+	void Scheduler::bind_thread_to_processor(Thread* thread, ProcessorID processor_id)
 	{
-		ASSERT(thread->m_scheduler_node == nullptr);
-		auto* new_node = new SchedulerQueue::Node(thread);
-		if (new_node == nullptr)
-			return BAN::Error::from_errno(ENOMEM);
-
 		ASSERT(processor_id != PROCESSOR_NONE);
-		new_node->processor_id = processor_id;
+		ASSERT(thread->m_scheduler_node.processor_id == PROCESSOR_NONE);
-		thread->m_scheduler_node = new_node;
+		thread->m_scheduler_node.processor_id = processor_id;
-
-		return {};
 	}
 
-	BAN::ErrorOr<void> Scheduler::add_thread(Thread* thread)
+	void Scheduler::add_thread(Thread* thread)
 	{
-		if (thread->m_scheduler_node == nullptr)
+		ASSERT(thread);
+
+		if (thread->m_scheduler_node.processor_id == PROCESSOR_NONE)
 		{
+			static BAN::Atomic<size_t> s_next_processor_index { 0 };
 			const size_t processor_index = s_next_processor_index++ % Processor::count();
 			const auto processor_id = Processor::id_from_index(processor_index);
-			TRY(bind_thread_to_processor(thread, processor_id));
+			bind_thread_to_processor(thread, processor_id);
 		}
 
-		auto* node = thread->m_scheduler_node;
+		if (const auto proc_id = thread->m_scheduler_node.processor_id; proc_id != Processor::current_id())
-		if (node->processor_id == Processor::current_id())
+		{
-			add_thread(node);
+			Processor::send_smp_message(proc_id, {
+				.type = Processor::SMPMessage::Type::NewThread,
+				.new_thread = thread
+			});
+			return;
+		}
+
+		const auto state = Processor::get_interrupt_state();
+		Processor::set_interrupt_state(InterruptState::Disabled);
+
+		if (!thread->m_scheduler_node.blocked)
+			m_run_list.push(&thread->m_scheduler_node);
 		else
 		{
-			Processor::send_smp_message(node->processor_id, {
+			m_block_list.push(&thread->m_scheduler_node);
-				.type = Processor::SMPMessage::Type::NewThread,
+			if (m_block_list.front() == &thread->m_scheduler_node)
-				.new_thread = node
+				update_wake_up_deadline();
-			});
 		}
 
-		return {};
+		if (thread->is_userspace() && thread->has_process())
+			thread->update_processor_index_address();
+
+		m_thread_count++;
+
+		Processor::set_interrupt_state(state);
 	}
 
 	void Scheduler::block_current_thread(ThreadBlocker* blocker, uint64_t wake_time_ns, BaseMutex* mutex)
@@ -673,7 +557,7 @@ namespace Kernel
 		if (SystemTimer::get().ns_since_boot() >= wake_time_ns)
 			return;
 
-		auto state = Processor::get_interrupt_state();
+		const auto state = Processor::get_interrupt_state();
 		Processor::set_interrupt_state(InterruptState::Disabled);
 
 		ASSERT(m_current->processor_id == Processor::current_id());
@@ -707,7 +591,39 @@ namespace Kernel
 
 	void Scheduler::unblock_thread(Thread* thread)
 	{
-		unblock_thread(thread->m_scheduler_node);
+		if (const auto proc_id = thread->m_scheduler_node.processor_id; proc_id != Processor::current_id())
+		{
+			Processor::send_smp_message(proc_id, {
+				.type = Processor::SMPMessage::Type::UnblockThread,
+				.unblock_thread = thread
+			});
+			return;
+		}
+
+		const auto state = Processor::get_interrupt_state();
+		Processor::set_interrupt_state(InterruptState::Disabled);
+
+		if (!thread->m_scheduler_node.blocked)
+		{
+			Processor::set_interrupt_state(state);
+			return;
+		}
+
+		ASSERT(&thread->m_scheduler_node != m_current);
+
+		Processor::set_disable_smp_messages(true);
+
+		m_block_list.pop(&thread->m_scheduler_node);
+		if (auto* blocker = thread->m_scheduler_node.blocker.load())
+			blocker->remove_thread_from_block_queue(&thread->m_scheduler_node);
+		thread->m_scheduler_node.blocked = false;
+
+		m_run_list.push(&thread->m_scheduler_node);
+		update_most_loaded_node_list(&thread->m_scheduler_node, &m_run_list);
+
+		Processor::set_disable_smp_messages(false);
+
+		Processor::set_interrupt_state(state);
 	}
 
 	Thread& Scheduler::current_thread()

kernel/kernel/SchedulerThreadNode.cpp

@@ -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;
+	}
+
+}

kernel/kernel/Storage/ATA/AHCI/Controller.cpp

@@ -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

kernel/kernel/Storage/ATA/AHCI/Device.cpp

@@ -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;
 			}