Kernel: Fix AML PCIConfig OpRegion accesses

Apparently I'm not supposted to calculate device/function from the
offset, but parse them from the acpi namespace :)

This allows PCI PIN interrupt routing actually work

kernel/CMakeLists.txt

@@ -31,6 +31,7 @@ set(KERNEL_SOURCES
 	kernel/FS/ProcFS/Inode.cpp
 	kernel/FS/TmpFS/FileSystem.cpp
 	kernel/FS/TmpFS/Inode.cpp
+	kernel/FS/USTARModule.cpp
 	kernel/FS/VirtualFileSystem.cpp
 	kernel/GDT.cpp
 	kernel/IDT.cpp

kernel/arch/i686/PageTable.cpp

@@ -206,6 +206,7 @@ namespace Kernel
 	{
 		ASSERT(s_kernel);
 		ASSERT(paddr);
+		ASSERT(paddr % PAGE_SIZE == 0);
 
 		ASSERT(s_fast_page_lock.current_processor_has_lock());
 

kernel/arch/i686/boot.S

@@ -36,6 +36,12 @@ multiboot2_start:
 	.long 12
 	.long V2P(_start)
 
+	# page align modules
+	.align 8
+	.short 6
+	.short 0
+	.long 8
+
 	.align 8
 	.short 0
 	.short 0
@@ -53,10 +59,10 @@ bananboot_start:
 bananboot_end:
 
 .section .bss, "aw", @nobits
-	.align 4096
-	boot_stack_bottom:
+    .global g_boot_stack_top
+	g_boot_stack_bottom:
 		.skip 4096 * 4
-	boot_stack_top:
+	g_boot_stack_top:
 
 	.global g_kernel_cmdline
 	g_kernel_cmdline:
@@ -188,7 +194,7 @@ _start:
 	movl %ebx, V2P(bootloader_info)
 
 	# load boot stack
-	movl $V2P(boot_stack_top), %esp
+	movl $V2P(g_boot_stack_top), %esp
 
 	# load boot GDT
 	lgdt V2P(boot_gdtr)
@@ -206,7 +212,7 @@ gdt_flush:
 	call initialize_paging
 
 	# load higher half stack pointer
-	movl $boot_stack_top, %esp
+	movl $g_boot_stack_top, %esp
 
 	# jump to higher half
 	leal higher_half, %ecx

kernel/arch/x86_64/PageTable.cpp

@@ -503,6 +503,7 @@ namespace Kernel
 	{
 		ASSERT(s_kernel);
 		ASSERT(paddr);
+		ASSERT(paddr % PAGE_SIZE == 0);
 
 		ASSERT(s_fast_page_lock.current_processor_has_lock());
 

kernel/arch/x86_64/boot.S

@@ -36,6 +36,12 @@ multiboot2_start:
 	.long 12
 	.long V2P(_start)
 
+	# page align modules
+	.align 8
+	.short 6
+	.short 0
+	.long 8
+
 	.align 8
 	.short 0
 	.short 0
@@ -53,9 +59,10 @@ bananboot_start:
 bananboot_end:
 
 .section .bss, "aw", @nobits
-	boot_stack_bottom:
-		.skip 4096 * 64
-	boot_stack_top:
+    .global g_boot_stack_top
+	g_boot_stack_bottom:
+		.skip 4096 * 4
+	g_boot_stack_top:
 
 	.global g_kernel_cmdline
 	g_kernel_cmdline:
@@ -187,7 +194,7 @@ _start:
 	movl %eax, V2P(bootloader_magic)
 	movl %ebx, V2P(bootloader_info)
 
-	movl $V2P(boot_stack_top), %esp
+	movl $V2P(g_boot_stack_top), %esp
 
 	call check_requirements
 	call enable_sse

kernel/include/kernel/ACPI/AML/Node.h

@@ -87,6 +87,12 @@ namespace Kernel::ACPI::AML
 	struct OpRegion
 	{
 		GAS::AddressSpaceID address_space;
+
+		uint16_t seg;
+		uint8_t bus;
+		uint8_t dev;
+		uint8_t func;
+
 		uint64_t offset;
 		uint64_t length;
 	};

kernel/include/kernel/BootInfo.h

@@ -41,6 +41,12 @@ namespace Kernel
 		Type     type;
 	};
 
+	struct BootModule
+	{
+		paddr_t start;
+		size_t size;
+	};
+
 	struct BootInfo
 	{
 		BAN::String     command_line;
@@ -48,6 +54,7 @@ namespace Kernel
 		RSDP            rsdp         {};
 		paddr_t         kernel_paddr {};
 
+		BAN::Vector<BootModule> modules;
 		BAN::Vector<MemoryMapEntry>	memory_map_entries;
 	};
 

kernel/include/kernel/Device/FramebufferDevice.h

@@ -17,6 +17,7 @@ namespace Kernel
 
 		uint32_t get_pixel(uint32_t x, uint32_t y) const;
 		void set_pixel(uint32_t x, uint32_t y, uint32_t rgb);
+		void fill(uint32_t rgb);
 
 		// positive rows -> empty pixels on bottom
 		// negative rows -> empty pixels on top

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

@@ -62,8 +62,6 @@ namespace Kernel
 
 		virtual BAN::RefPtr<Inode> root_inode() override { return m_root_inode; }
 
-		virtual dev_t dev() const override { return m_block_device->rdev(); };
-
 	private:
 		Ext2FS(BAN::RefPtr<BlockDevice> block_device)
 			: m_block_device(block_device)

kernel/include/kernel/FS/Ext2/Inode.h

@@ -46,6 +46,7 @@ namespace Kernel
 		virtual BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override;
 		virtual BAN::ErrorOr<void> truncate_impl(size_t) override;
 		virtual BAN::ErrorOr<void> chmod_impl(mode_t) override;
+		virtual BAN::ErrorOr<void> chown_impl(uid_t, gid_t) override;
 		virtual BAN::ErrorOr<void> utimens_impl(const timespec[2]) override;
 		virtual BAN::ErrorOr<void> fsync_impl() override;
 

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

@@ -37,8 +37,6 @@ namespace Kernel
 
 		virtual BAN::RefPtr<Inode> root_inode() override { return m_root_inode; }
 
-		virtual dev_t dev() const override { return m_block_device->rdev(); };
-
 		BAN::ErrorOr<BAN::RefPtr<FATInode>> open_inode(BAN::RefPtr<FATInode> parent, const FAT::DirectoryEntry& entry, uint32_t cluster_index, uint32_t entry_index);
 		BAN::ErrorOr<void> inode_read_cluster(BAN::RefPtr<FATInode>, size_t index, BAN::ByteSpan buffer);
 		blksize_t inode_block_size(BAN::RefPtr<const FATInode>) const;

kernel/include/kernel/FS/FileSystem.h

@@ -26,8 +26,6 @@ namespace Kernel
 		static BAN::ErrorOr<BAN::RefPtr<FileSystem>> from_block_device(BAN::RefPtr<BlockDevice>);
 
 		virtual BAN::RefPtr<Inode> root_inode() = 0;
-
-		virtual dev_t dev() const = 0;
 	};
 
 }

kernel/include/kernel/FS/TmpFS/FileSystem.h

@@ -58,7 +58,7 @@ namespace Kernel
 
 		virtual BAN::RefPtr<Inode> root_inode() override { return m_root_inode; }
 
-		virtual dev_t dev() const override { return m_rdev; }
+		dev_t rdev() const { return m_rdev; }
 
 		BAN::ErrorOr<BAN::RefPtr<TmpInode>> open_inode(ino_t ino);
 
@@ -118,16 +118,8 @@ namespace Kernel
 
 	private:
 		InodeLocation find_inode(ino_t ino);
-
 		paddr_t find_block(size_t index);
 
-		template<TmpFuncs::for_each_indirect_paddr_allocating_callback F>
-		BAN::ErrorOr<void> for_each_indirect_paddr_allocating(PageInfo page_info, F callback, size_t depth);
-		template<TmpFuncs::for_each_indirect_paddr_allocating_callback F>
-		BAN::ErrorOr<BAN::Iteration> for_each_indirect_paddr_allocating_internal(PageInfo page_info, F callback, size_t depth);
-
-		paddr_t find_indirect(PageInfo root, size_t index, size_t depth);
-
 	private:
 		const dev_t m_rdev;
 
@@ -146,14 +138,14 @@ namespace Kernel
 		static constexpr size_t max_data_pages =
 			(PAGE_SIZE / sizeof(PageInfo)) *
 			(PAGE_SIZE / sizeof(PageInfo)) *
-			(PAGE_SIZE / sizeof(PageInfo));
+			(PAGE_SIZE / sizeof(PageInfo) - 1);
 
 		// We store inodes in pages with double indirection.
 		// With 64-bit pointers we can store 512^2 pages of inodes
 		// which should be enough for now.
 		// In future this should be dynamically calculated based on maximum
 		// number of pages for this file system.
-		PageInfo m_inode_pages;
+		PageInfo m_inode_pages {};
 		static constexpr size_t first_inode = 1;
 		static constexpr size_t max_inodes =
 			(PAGE_SIZE / sizeof(PageInfo)) *

kernel/include/kernel/FS/TmpFS/Inode.h

@@ -48,15 +48,21 @@ namespace Kernel
 		TmpInode(TmpFileSystem&, ino_t, const TmpInodeInfo&);
 
 		virtual BAN::ErrorOr<void> chmod_impl(mode_t) override;
+		virtual BAN::ErrorOr<void> chown_impl(uid_t, gid_t) override;
 		virtual BAN::ErrorOr<void> utimens_impl(const timespec[2]) override;
 		virtual BAN::ErrorOr<void> fsync_impl() override { return {}; }
 
 		void sync();
-		void free_all_blocks();
 		virtual BAN::ErrorOr<void> prepare_unlink() { return {}; };
 
+		void free_all_blocks();
+		void free_indirect_blocks(size_t block, uint32_t depth);
+
 		BAN::Optional<size_t> block_index(size_t data_block_index);
+		BAN::Optional<size_t> block_index_from_indirect(size_t block, size_t index, uint32_t depth);
+
 		BAN::ErrorOr<size_t> block_index_with_allocation(size_t data_block_index);
+		BAN::ErrorOr<size_t> block_index_from_indirect_with_allocation(size_t& block, size_t index, uint32_t depth);
 
 	protected:
 		TmpFileSystem& m_fs;
@@ -117,10 +123,9 @@ namespace Kernel
 		static BAN::ErrorOr<BAN::RefPtr<TmpSymlinkInode>> create_new(TmpFileSystem&, mode_t, uid_t, gid_t, BAN::StringView target);
 		~TmpSymlinkInode();
 
-		BAN::ErrorOr<void> set_link_target(BAN::StringView);
-
 	protected:
-		virtual BAN::ErrorOr<BAN::String> link_target_impl() override;
+		BAN::ErrorOr<BAN::String> link_target_impl() override;
+		BAN::ErrorOr<void> set_link_target_impl(BAN::StringView) override;
 
 		virtual bool can_read_impl() const override { return false; }
 		virtual bool can_write_impl() const override { return false; }

kernel/include/kernel/FS/USTARModule.h

@@ -0,0 +1,12 @@
+#pragma once
+
+#include <kernel/BootInfo.h>
+#include <kernel/FS/FileSystem.h>
+
+namespace Kernel
+{
+
+	bool is_ustar_boot_module(const BootModule&);
+	BAN::ErrorOr<void> unpack_boot_module_into_filesystem(BAN::RefPtr<FileSystem>, const BootModule&);
+
+}

kernel/include/kernel/FS/VirtualFileSystem.h

@@ -29,9 +29,6 @@ namespace Kernel
 
 		virtual BAN::RefPtr<Inode> root_inode() override { return m_root_fs->root_inode(); }
 
-		// FIXME:
-		virtual dev_t dev() const override { return 0; }
-
 		BAN::ErrorOr<void> mount(const Credentials&, BAN::StringView, BAN::StringView);
 		BAN::ErrorOr<void> mount(const Credentials&, BAN::RefPtr<FileSystem>, BAN::StringView);
 

kernel/include/kernel/IDT.h

@@ -75,7 +75,7 @@ namespace Kernel
 	private:
 		IDT() = default;
 
-		void register_interrupt_handler(uint8_t index, void (*handler)());
+		void register_interrupt_handler(uint8_t index, void (*handler)(), uint8_t ist = 0);
 		void register_syscall_handler(uint8_t index, void (*handler)());
 
 	private:

kernel/include/kernel/Input/PS2/Controller.h

@@ -27,7 +27,9 @@ namespace Kernel::Input
 	private:
 		PS2Controller() = default;
 		BAN::ErrorOr<void> initialize_impl(uint8_t scancode_set);
-		BAN::ErrorOr<void> initialize_device(uint8_t, uint8_t scancode_set);
+		BAN::ErrorOr<void> identify_device(uint8_t, uint8_t scancode_set);
+
+		void device_initialize_task(void*);
 
 		BAN::ErrorOr<uint8_t> read_byte();
 		BAN::ErrorOr<void> send_byte(uint16_t port, uint8_t byte);

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

@@ -22,6 +22,7 @@ namespace Kernel::Input
 
 	protected:
 		PS2Device(PS2Controller&, InputDevice::Type type);
+		virtual ~PS2Device();
 
 	protected:
 		PS2Controller& m_controller;

kernel/include/kernel/Memory/VirtualRange.h

@@ -15,21 +15,21 @@ namespace Kernel
 
 	public:
 		// Create virtual range to fixed virtual address
-		static BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> create_to_vaddr(PageTable&, vaddr_t, size_t, PageTable::flags_t flags, bool preallocate_pages);
+		static BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> create_to_vaddr(PageTable&, vaddr_t, size_t, PageTable::flags_t flags, bool preallocate_pages, bool add_guard_pages);
 		// Create virtual range to virtual address range
-		static BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> create_to_vaddr_range(PageTable&, vaddr_t vaddr_start, vaddr_t vaddr_end, size_t, PageTable::flags_t flags, bool preallocate_pages);
+		static BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> create_to_vaddr_range(PageTable&, vaddr_t vaddr_start, vaddr_t vaddr_end, size_t, PageTable::flags_t flags, bool preallocate_pages, bool add_guard_pages);
 		~VirtualRange();
 
 		BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> clone(PageTable&);
 
-		vaddr_t vaddr() const { return m_vaddr; }
-		size_t size() const { return m_size; }
+		vaddr_t vaddr() const { return m_vaddr + (m_has_guard_pages ? PAGE_SIZE : 0); }
+		size_t size() const { return m_size - (m_has_guard_pages ? 2 * PAGE_SIZE : 0); }
 		PageTable::flags_t flags() const { return m_flags; }
 
 		paddr_t paddr_of(vaddr_t vaddr) const
 		{
 			ASSERT(vaddr % PAGE_SIZE == 0);
-			const size_t index = (vaddr - m_vaddr) / PAGE_SIZE;
+			const size_t index = (vaddr - this->vaddr()) / PAGE_SIZE;
 			ASSERT(index < m_paddrs.size());
 			const paddr_t paddr = m_paddrs[index];
 			ASSERT(paddr);
@@ -41,12 +41,13 @@ namespace Kernel
 		BAN::ErrorOr<void> allocate_page_for_demand_paging(vaddr_t address);
 
 	private:
-		VirtualRange(PageTable&, bool preallocated, vaddr_t, size_t, PageTable::flags_t);
+		VirtualRange(PageTable&, bool preallocated, bool has_guard_pages, vaddr_t, size_t, PageTable::flags_t);
 		BAN::ErrorOr<void> initialize();
 
 	private:
 		PageTable&               m_page_table;
 		const bool               m_preallocated;
+		const bool               m_has_guard_pages;
 		const vaddr_t            m_vaddr;
 		const size_t             m_size;
 		const PageTable::flags_t m_flags;

kernel/include/kernel/Networking/ARPTable.h

@@ -4,7 +4,7 @@
 #include <BAN/HashMap.h>
 #include <BAN/UniqPtr.h>
 #include <kernel/Networking/NetworkInterface.h>
-#include <kernel/Process.h>
+#include <kernel/Thread.h>
 #include <kernel/ThreadBlocker.h>
 
 namespace Kernel
@@ -56,7 +56,7 @@ namespace Kernel
 
 		BAN::HashMap<BAN::IPv4Address, BAN::MACAddress> m_arp_table;
 
-		Process*									m_process = nullptr;
+		Thread*										m_thread { nullptr };
 		BAN::CircularQueue<PendingArpPacket, 128>	m_pending_packets;
 		ThreadBlocker								m_pending_thread_blocker;
 

kernel/include/kernel/Networking/IPv4Layer.h

@@ -11,7 +11,7 @@
 #include <kernel/Networking/NetworkInterface.h>
 #include <kernel/Networking/NetworkLayer.h>
 #include <kernel/Networking/NetworkSocket.h>
-#include <kernel/Process.h>
+#include <kernel/Thread.h>
 
 namespace Kernel
 {
@@ -72,7 +72,7 @@ namespace Kernel
 		RecursiveSpinLock		m_bound_socket_lock;
 
 		BAN::UniqPtr<ARPTable>	m_arp_table;
-		Process*				m_process { nullptr };
+		Thread*					m_thread { nullptr };
 
 		static constexpr size_t pending_packet_buffer_size = 128 * PAGE_SIZE;
 		BAN::UniqPtr<VirtualRange>					m_pending_packet_buffer;

kernel/include/kernel/Networking/TCPSocket.h

@@ -6,7 +6,7 @@
 #include <kernel/Memory/VirtualRange.h>
 #include <kernel/Networking/NetworkInterface.h>
 #include <kernel/Networking/NetworkSocket.h>
-#include <kernel/Process.h>
+#include <kernel/Thread.h>
 #include <kernel/ThreadBlocker.h>
 
 namespace Kernel
@@ -162,7 +162,7 @@ namespace Kernel
 		State m_next_state		{ State::Closed };
 		uint8_t m_next_flags	{ 0 };
 
-		Process* m_process { nullptr };
+		Thread* m_thread { nullptr };
 
 		uint64_t m_time_wait_start_ms { 0 };
 

kernel/include/kernel/Process.h

@@ -39,8 +39,6 @@ namespace Kernel
 		using entry_t = Thread::entry_t;
 
 	public:
-		static Process* create_kernel();
-		static Process* create_kernel(entry_t, void*);
 		static BAN::ErrorOr<Process*> create_userspace(const Credentials&, BAN::StringView path, BAN::Span<BAN::StringView> arguments);
 		~Process();
 		void cleanup_function(Thread*);
@@ -217,8 +215,6 @@ namespace Kernel
 		size_t proc_cmdline(off_t offset, BAN::ByteSpan) const;
 		size_t proc_environ(off_t offset, BAN::ByteSpan) const;
 
-		bool is_userspace() const { return m_is_userspace; }
-
 		// Returns error if page could not be allocated
 		// Returns true if the page was allocated successfully
 		// Return false if access was page violation (segfault)
@@ -331,8 +327,6 @@ namespace Kernel
 		BAN::Vector<BAN::String> m_cmdline;
 		BAN::Vector<BAN::String> m_environ;
 
-		bool m_is_userspace { false };
-
 		BAN::Vector<ChildExitStatus> m_child_exit_statuses;
 		ThreadBlocker m_child_exit_blocker;
 

kernel/include/kernel/Scheduler.h

@@ -14,33 +14,12 @@ namespace Kernel
 	class BaseMutex;
 	class Thread;
 	class ThreadBlocker;
+	struct SchedulerQueueNode;
 
 	class SchedulerQueue
 	{
 	public:
-		struct Node
-		{
-			Node(Thread* thread)
-				: thread(thread)
-			{}
-
-			Thread* const thread;
-
-			Node* next { nullptr };
-			Node* prev { nullptr };
-
-			uint64_t wake_time_ns { static_cast<uint64_t>(-1) };
-
-			ThreadBlocker* blocker { nullptr };
-			Node* block_chain_next { nullptr };
-			Node* block_chain_prev { nullptr };
-
-			ProcessorID processor_id { PROCESSOR_NONE };
-			bool blocked { false };
-
-			uint64_t last_start_ns { 0 };
-			uint64_t time_used_ns  { 0 };
-		};
+		using Node = SchedulerQueueNode;
 
 	public:
 		void add_thread_to_back(Node*);

kernel/include/kernel/SchedulerQueueNode.h

@@ -0,0 +1,35 @@
+#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) };
+
+		SpinLock blocker_lock;
+		ThreadBlocker* blocker { nullptr };
+
+		ProcessorID processor_id { PROCESSOR_NONE };
+		bool blocked { false };
+
+		uint64_t last_start_ns { 0 };
+		uint64_t time_used_ns  { 0 };
+	};
+
+}

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

@@ -40,8 +40,6 @@ namespace Kernel
 		void select_device(bool is_secondary);
 		BAN::ErrorOr<DeviceType> identify(bool is_secondary, BAN::Span<uint16_t> buffer);
 
-		BAN::ErrorOr<void> block_until_irq();
-
 		uint8_t io_read(uint16_t);
 		void io_write(uint16_t, uint8_t);
 		void read_buffer(uint16_t, uint16_t*, size_t);
@@ -54,7 +52,7 @@ namespace Kernel
 		const uint16_t m_ctrl;
 		Mutex m_mutex;
 
-		BAN::Atomic<bool> m_has_got_irq { false };
+		ThreadBlocker m_thread_blocker;
 
 		// Non-owning pointers
 		BAN::Vector<ATADevice*> m_devices;

kernel/include/kernel/Thread.h

@@ -36,7 +36,7 @@ namespace Kernel
 		static constexpr size_t userspace_stack_size { PAGE_SIZE * 128 };
 
 	public:
-		static BAN::ErrorOr<Thread*> create_kernel(entry_t, void*, Process*);
+		static BAN::ErrorOr<Thread*> create_kernel(entry_t, void*);
 		static BAN::ErrorOr<Thread*> create_userspace(Process*, PageTable&);
 		~Thread();
 

kernel/include/kernel/ThreadBlocker.h

@@ -33,7 +33,9 @@ namespace Kernel
 
 	private:
 		SpinLock m_lock;
-		SchedulerQueue::Node* m_block_chain { nullptr };
+
+		SchedulerQueue::Node* m_block_chain[32] {};
+		size_t m_block_chain_length { 0 };
 
 		friend class Scheduler;
 	};

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

@@ -34,7 +34,7 @@ namespace Kernel
 
 		BAN::Atomic<uint32_t> m_changed_ports { 0 };
 		ThreadBlocker m_changed_port_blocker;
-		BAN::Atomic<Process*> m_port_updater { nullptr };
+		BAN::Atomic<Thread*> m_port_updater { nullptr };
 
 		struct PortInfo
 		{

kernel/include/kernel/USB/XHCI/Controller.h

@@ -73,7 +73,7 @@ namespace Kernel
 
 		Mutex m_mutex;
 
-		BAN::Atomic<Process*> m_port_updater { nullptr };
+		BAN::Atomic<Thread*> m_port_updater { nullptr };
 		ThreadBlocker m_port_thread_blocker;
 		BAN::Atomic<bool> m_port_changed { false };
 

kernel/include/kernel/multiboot2.h

@@ -6,6 +6,7 @@
 
 #define MULTIBOOT2_TAG_END			0
 #define MULTIBOOT2_TAG_CMDLINE		1
+#define MULTIBOOT2_TAG_MODULES		3
 #define MULTIBOOT2_TAG_MMAP			6
 #define MULTIBOOT2_TAG_FRAMEBUFFER	8
 #define MULTIBOOT2_TAG_OLD_RSDP		14
@@ -33,6 +34,13 @@ struct multiboot2_cmdline_tag_t : public multiboot2_tag_t
 	char cmdline[];
 } __attribute__((packed));
 
+struct multiboot2_modules_tag_t : public multiboot2_tag_t
+{
+	uint32_t mod_start;
+	uint32_t mod_end;
+	uint8_t string[];
+} __attribute__((packed));
+
 struct multiboot2_mmap_entry_t
 {
 	uint64_t base_addr;

kernel/kernel/ACPI/ACPI.cpp

@@ -885,7 +885,10 @@ acpi_release_global_lock:
 			set_irq(irq);
 			InterruptController::get().enable_irq(irq);
 
-			Process::create_kernel([](void*) { get().acpi_event_task(); }, nullptr);
+			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())
+				dwarnln("Failed to create ACPI thread, power button will not work: {}", ret.error());
 		}
 
 		dprintln("Initialized ACPI interrupts");

kernel/kernel/ACPI/AML/Node.cpp

@@ -1,3 +1,10 @@
+// FIXME: Rewrite aml interpreter to not be recursive.
+//        Not inlining TRYs drops our stack usage a ton...
+#pragma GCC push_options
+#pragma GCC optimize "no-inline"
+#include <BAN/Errors.h>
+#pragma GCC pop_options
+
 #include <BAN/Assert.h>
 #include <BAN/String.h>
 
@@ -75,7 +82,12 @@ namespace Kernel::ACPI::AML
 				);
 				return BAN::Error::from_errno(EINVAL);
 			}
-			name.parts[i] = aml_data.as<const uint32_t>();
+
+			name.parts[i] =
+				static_cast<uint32_t>(aml_data[0] <<  0) |
+				static_cast<uint32_t>(aml_data[1] <<  8) |
+				static_cast<uint32_t>(aml_data[2] << 16) |
+				static_cast<uint32_t>(aml_data[3] << 24);
 			aml_data = aml_data.slice(4);
 		}
 
@@ -377,9 +389,6 @@ namespace Kernel::ACPI::AML
 		return result;
 	}
 
-// FIXME: WHY TF IS THIS USING OVER 1 KiB of stack
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wstack-usage="
 	static BAN::ErrorOr<Node> parse_logical_op(ParseContext& context)
 	{
 		dprintln_if(AML_DUMP_FUNCTION_CALLS, "parse_logical_op");
@@ -470,7 +479,6 @@ namespace Kernel::ACPI::AML
 
 		return result;
 	}
-#pragma GCC diagnostic pop
 
 	static BAN::ErrorOr<Node> parse_index_op(ParseContext& context);
 
@@ -746,8 +754,6 @@ namespace Kernel::ACPI::AML
 		return {};
 	}
 
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wstack-usage="
 	static BAN::ErrorOr<void> perform_store(const Node& source, Reference* target, TargetType target_type)
 	{
 		dprintln_if(AML_DUMP_FUNCTION_CALLS, "perform_store");
@@ -828,7 +834,6 @@ namespace Kernel::ACPI::AML
 
 		return {};
 	}
-#pragma GCC diagnostic pop
 
 	static BAN::ErrorOr<void> store_into_target(ParseContext& context, const Node& node)
 	{
@@ -1235,7 +1240,7 @@ namespace Kernel::ACPI::AML
 		}
 
 		ASSERT(object);
-		return TRY(sizeof_impl(object->node));
+		return sizeof_impl(object->node);
 	}
 
 	static BAN::ErrorOr<Node> derefof_impl(const Node& source)
@@ -1243,7 +1248,7 @@ namespace Kernel::ACPI::AML
 		switch (source.type)
 		{
 			case Node::Type::Reference:
-				return TRY(source.as.reference->node.copy());
+				return source.as.reference->node.copy();
 			case Node::Type::Index:
 			{
 				switch (source.as.index.type)
@@ -1262,7 +1267,7 @@ namespace Kernel::ACPI::AML
 					{
 						ASSERT(source.as.index.index < source.as.index.as.package->num_elements);
 						TRY(resolve_package_element(source.as.index.as.package->elements[source.as.index.index], true));
-						return TRY(source.as.index.as.package->elements[source.as.index.index].value.node->copy());
+						return source.as.index.as.package->elements[source.as.index.index].value.node->copy();
 					}
 					default: ASSERT_NOT_REACHED();
 				}
@@ -1544,9 +1549,6 @@ namespace Kernel::ACPI::AML
 		return result;
 	}
 
-// FIXME: WHY TF IS THIS USING OVER 1 KiB of stack
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wstack-usage="
 	static BAN::ErrorOr<Node> parse_explicit_conversion(ParseContext& context)
 	{
 		dprintln_if(AML_DUMP_FUNCTION_CALLS, "parse_explicit_conversion");
@@ -1699,7 +1701,6 @@ namespace Kernel::ACPI::AML
 
 		return result;
 	}
-#pragma GCC diagnostic pop
 
 	static BAN::ErrorOr<Node> parse_to_string_op(ParseContext& context)
 	{
@@ -2001,7 +2002,6 @@ namespace Kernel::ACPI::AML
 		return {};
 	}
 
-
 	static BAN::ErrorOr<Node> parse_wait_op(ParseContext& context)
 	{
 		dprintln_if(AML_DUMP_FUNCTION_CALLS, "parse_wait_op");
@@ -2617,7 +2617,7 @@ namespace Kernel::ACPI::AML
 			case Node::Type::Buffer:
 			case Node::Type::Index:
 			case Node::Type::Reference:
-				return TRY(node.copy());
+				return node.copy();
 			case Node::Type::BufferField:
 				dwarnln("TODO: evaluate BufferField");
 				return BAN::Error::from_errno(ENOTSUP);
@@ -2626,7 +2626,7 @@ namespace Kernel::ACPI::AML
 			case Node::Type::Method:
 				if (node.as.method.arg_count != 0)
 					return BAN::Error::from_errno(EFAULT);
-				return TRY(method_call(node_path, node, BAN::Array<Reference*, 7>{}));
+				return method_call(node_path, node, BAN::Array<Reference*, 7>{});
 		}
 
 		dwarnln("evaluate {}", node);
@@ -2761,9 +2761,6 @@ namespace Kernel::ACPI::AML
 		return method_call(scope, method, BAN::move(args));
 	}
 
-// FIXME: WHY TF IS THIS USING OVER 2 KiB of stack
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wstack-usage="
 	BAN::ErrorOr<Node> parse_node(ParseContext& context, bool return_ref)
 	{
 		if (context.aml_data.empty())
@@ -2778,15 +2775,15 @@ namespace Kernel::ACPI::AML
 			switch (static_cast<AML::ExtOp>(opcode))
 			{
 				case AML::ExtOp::CondRefOfOp:
-					return TRY(parse_condrefof_op(context));
+					return parse_condrefof_op(context);
 				case AML::ExtOp::AcquireOp:
-					return TRY(parse_acquire_op(context));
+					return parse_acquire_op(context);
 				case AML::ExtOp::LoadOp:
-					return TRY(parse_load_op(context));
+					return parse_load_op(context);
 				case AML::ExtOp::TimerOp:
-					return TRY(parse_timer_op(context));
+					return parse_timer_op(context);
 				case AML::ExtOp::WaitOp:
-					return TRY(parse_wait_op(context));
+					return parse_wait_op(context);
 				case AML::ExtOp::DebugOp:
 				{
 					context.aml_data = context.aml_data.slice(2);
@@ -2812,46 +2809,46 @@ namespace Kernel::ACPI::AML
 			case AML::Byte::WordPrefix:
 			case AML::Byte::DWordPrefix:
 			case AML::Byte::QWordPrefix:
-				return TRY(parse_integer(context.aml_data));
+				return parse_integer(context.aml_data);
 			case AML::Byte::StringPrefix:
-				return TRY(parse_string(context.aml_data));
+				return parse_string(context.aml_data);
 			case AML::Byte::BufferOp:
-				return TRY(parse_buffer_op(context));
+				return parse_buffer_op(context);
 			case AML::Byte::PackageOp:
 			case AML::Byte::VarPackageOp:
-				return TRY(parse_package_op(context));
+				return parse_package_op(context);
 			case AML::Byte::SizeOfOp:
-				return TRY(parse_sizeof_op(context));
+				return parse_sizeof_op(context);
 			case AML::Byte::RefOfOp:
-				return TRY(parse_refof_op(context));
+				return parse_refof_op(context);
 			case AML::Byte::DerefOfOp:
-				return TRY(parse_derefof_op(context));
+				return parse_derefof_op(context);
 			case AML::Byte::StoreOp:
-				return TRY(parse_store_op(context));
+				return parse_store_op(context);
 			case AML::Byte::CopyObjectOp:
-				return TRY(parse_copy_object_op(context));
+				return parse_copy_object_op(context);
 			case AML::Byte::ConcatOp:
-				return TRY(parse_concat_op(context));
+				return parse_concat_op(context);
 			case AML::Byte::MidOp:
-				return TRY(parse_mid_op(context));
+				return parse_mid_op(context);
 			case AML::Byte::IndexOp:
-				return TRY(parse_index_op(context));
+				return parse_index_op(context);
 			case AML::Byte::ObjectTypeOp:
-				return TRY(parse_object_type_op(context));
+				return parse_object_type_op(context);
 			case AML::Byte::MatchOp:
-				return TRY(parse_match_op(context));
+				return parse_match_op(context);
 			case AML::Byte::ToBufferOp:
 			case AML::Byte::ToDecimalStringOp:
 			case AML::Byte::ToHexStringOp:
 			case AML::Byte::ToIntegerOp:
-				return TRY(parse_explicit_conversion(context));
+				return parse_explicit_conversion(context);
 			case AML::Byte::ToStringOp:
-				return TRY(parse_to_string_op(context));
+				return parse_to_string_op(context);
 			case AML::Byte::IncrementOp:
 			case AML::Byte::DecrementOp:
-				return TRY(parse_inc_dec_op(context));
+				return parse_inc_dec_op(context);
 			case AML::Byte::NotOp:
-				return TRY(parse_unary_integer_op(context));
+				return parse_unary_integer_op(context);
 			case AML::Byte::AddOp:
 			case AML::Byte::SubtractOp:
 			case AML::Byte::MultiplyOp:
@@ -2864,17 +2861,17 @@ namespace Kernel::ACPI::AML
 			case AML::Byte::NorOp:
 			case AML::Byte::XorOp:
 			case AML::Byte::ModOp:
-				return TRY(parse_binary_integer_op(context));
+				return parse_binary_integer_op(context);
 			case AML::Byte::LAndOp:
 			case AML::Byte::LEqualOp:
 			case AML::Byte::LGreaterOp:
 			case AML::Byte::LLessOp:
 			case AML::Byte::LNotOp:
 			case AML::Byte::LOrOp:
-				return TRY(parse_logical_op(context));
+				return parse_logical_op(context);
 			case AML::Byte::FindSetLeftBitOp:
 			case AML::Byte::FindSetRightBitOp:
-				return TRY(parse_find_set_bit_op(context));
+				return parse_find_set_bit_op(context);
 			case AML::Byte::Local0:
 			case AML::Byte::Local1:
 			case AML::Byte::Local2:
@@ -2892,7 +2889,7 @@ namespace Kernel::ACPI::AML
 					return BAN::Error::from_errno(EINVAL);
 				}
 				if (!return_ref)
-					return TRY(context.locals[local_index]->node.copy());
+					return context.locals[local_index]->node.copy();
 				Node reference;
 				reference.type = Node::Type::Reference;
 				reference.as.reference = context.locals[local_index];
@@ -2915,7 +2912,7 @@ namespace Kernel::ACPI::AML
 					return BAN::Error::from_errno(EINVAL);
 				}
 				if (!return_ref)
-					return TRY(context.args[arg_index]->node.copy());
+					return context.args[arg_index]->node.copy();
 				Node reference;
 				reference.type = Node::Type::Reference;
 				reference.as.reference = context.args[arg_index];
@@ -2962,11 +2959,11 @@ namespace Kernel::ACPI::AML
 				}
 			}
 
-			return TRY(method_call(BAN::move(object_scope), named_object->node, BAN::move(args), context.call_depth));
+			return method_call(BAN::move(object_scope), named_object->node, BAN::move(args), context.call_depth);
 		}
 
 		if (!return_ref)
-			return TRY(named_object->node.copy());
+			return named_object->node.copy();
 
 		Node reference;
 		reference.type = Node::Type::Reference;
@@ -2974,20 +2971,13 @@ namespace Kernel::ACPI::AML
 		reference.as.reference->ref_count++;
 		return reference;
 	}
-#pragma GCC diagnostic pop
 
-// FIXME: WHY TF IS THIS USING ALMOST 2 KiB of stack
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wstack-usage="
 	BAN::ErrorOr<ExecutionFlowResult> parse_node_or_execution_flow(ParseContext& context)
 	{
 		if (context.aml_data.empty())
 			return BAN::Error::from_errno(ENODATA);
 
-		auto dummy_return = ExecutionFlowResult {
-			.elem1 = ExecutionFlow::Normal,
-			.elem2 = BAN::Optional<Node>(),
-		};
+		BAN::ErrorOr<void> (*function)(ParseContext&) = nullptr;
 
 		if (context.aml_data[0] == static_cast<uint8_t>(AML::Byte::ExtOpPrefix))
 		{
@@ -2996,91 +2986,95 @@ namespace Kernel::ACPI::AML
 			switch (static_cast<AML::ExtOp>(context.aml_data[1]))
 			{
 				case AML::ExtOp::MutexOp:
-					TRY(parse_mutex_op(context));
-					return dummy_return;
+					function = parse_mutex_op;
+					break;
 				case AML::ExtOp::FatalOp:
-					TRY(parse_fatal_op(context));
-					return dummy_return;
+					function = parse_fatal_op;
+					break;
 				case AML::ExtOp::EventOp:
-					TRY(parse_event_op(context));
-					return dummy_return;
+					function = parse_event_op;
+					break;
 				case AML::ExtOp::ResetOp:
 				case AML::ExtOp::SignalOp:
-					TRY(parse_reset_signal_op(context));
-					return dummy_return;
+					function = parse_reset_signal_op;
+					break;
 				case AML::ExtOp::CreateFieldOp:
-					TRY(parse_createfield_op(context));
-					return dummy_return;
+					function = parse_createfield_op;
+					break;
 				case AML::ExtOp::SleepOp:
-					TRY(parse_sleep_op(context));
-					return dummy_return;
+					function = parse_sleep_op;
+					break;
 				case AML::ExtOp::StallOp:
-					TRY(parse_stall_op(context));
-					return dummy_return;
+					function = parse_stall_op;
+					break;
 				case AML::ExtOp::ReleaseOp:
-					TRY(parse_release_op(context));
-					return dummy_return;
+					function = parse_release_op;
+					break;
 				case AML::ExtOp::OpRegionOp:
-					TRY(parse_opregion_op(context));
-					return dummy_return;
+					function = parse_opregion_op;
+					break;
 				case AML::ExtOp::FieldOp:
-					TRY(parse_field_op(context));
-					return dummy_return;
+					function = parse_field_op;
+					break;
 				case AML::ExtOp::IndexFieldOp:
-					TRY(parse_index_field_op(context));
-					return dummy_return;
+					function = parse_index_field_op;
+					break;
 				case AML::ExtOp::BankFieldOp:
-					TRY(parse_bank_field_op(context));
-					return dummy_return;
+					function = parse_bank_field_op;
+					break;
 				case AML::ExtOp::DeviceOp:
-					TRY(parse_device_op(context));
-					return dummy_return;
+					function = parse_device_op;
+					break;
 				case AML::ExtOp::ProcessorOp:
-					TRY(parse_processor_op(context));
-					return dummy_return;
+					function = parse_processor_op;
+					break;
 				case AML::ExtOp::PowerResOp:
-					TRY(parse_power_resource_op(context));
-					return dummy_return;
+					function = parse_power_resource_op;
+					break;
 				case AML::ExtOp::ThermalZoneOp:
-					TRY(parse_thermal_zone_op(context));
-					return dummy_return;
+					function = parse_thermal_zone_op;
+					break;
 				default:
 					break;
 			}
 		}
-
+		else
+		{
 			switch (static_cast<AML::Byte>(context.aml_data[0]))
 			{
 				case AML::Byte::AliasOp:
-				TRY(parse_alias_op(context));
-				return dummy_return;
+					function = parse_alias_op;
+					break;
 				case AML::Byte::NameOp:
-				TRY(parse_name_op(context));
-				return dummy_return;
+					function = parse_name_op;
+					break;
 				case AML::Byte::MethodOp:
-				TRY(parse_method_op(context));
-				return dummy_return;
-			case AML::Byte::NoopOp:
-			case AML::Byte::BreakPointOp:
-				context.aml_data = context.aml_data.slice(1);
-				return dummy_return;
+					function = parse_method_op;
+					break;
 				case AML::Byte::ScopeOp:
-				TRY(parse_scope_op(context));
-				return dummy_return;
+					function = parse_scope_op;
+					break;
 				case AML::Byte::NotifyOp:
-				TRY(parse_notify_op(context));
-				return dummy_return;
+					function = parse_notify_op;
+					break;
 				case AML::Byte::CreateBitFieldOp:
 				case AML::Byte::CreateByteFieldOp:
 				case AML::Byte::CreateWordFieldOp:
 				case AML::Byte::CreateDWordFieldOp:
 				case AML::Byte::CreateQWordFieldOp:
-				TRY(parse_createfield_op(context));
-				return dummy_return;
+					function = parse_createfield_op;
+					break;
 				case AML::Byte::IfOp:
 					return parse_if_op(context);
 				case AML::Byte::WhileOp:
 					return parse_while_op(context);
+				case AML::Byte::NoopOp:
+				case AML::Byte::BreakPointOp:
+					context.aml_data = context.aml_data.slice(1);
+					return ExecutionFlowResult {
+						.elem1 = ExecutionFlow::Normal,
+						.elem2 = BAN::Optional<Node>(),
+					};;
 				case AML::Byte::BreakOp:
 					dprintln_if(AML_DUMP_FUNCTION_CALLS, "parse_break_op");
 					context.aml_data = context.aml_data.slice(1);
@@ -3107,6 +3101,16 @@ namespace Kernel::ACPI::AML
 				default:
 					break;
 			}
+		}
+
+		if (function)
+		{
+			TRY(function(context));
+			return ExecutionFlowResult {
+				.elem1 = ExecutionFlow::Normal,
+				.elem2 = BAN::Optional<Node>(),
+			};;
+		}
 
 		auto node = TRY(parse_node(context));
 		return ExecutionFlowResult {
@@ -3114,7 +3118,6 @@ namespace Kernel::ACPI::AML
 			.elem2 = BAN::move(node)
 		};
 	}
-#pragma GCC diagnostic pop
 
 	BAN::ErrorOr<NameString> NameString::from_string(BAN::StringView name)
 	{

kernel/kernel/ACPI/AML/OpRegion.cpp

@@ -1,3 +1,9 @@
+// FIXME: Find better ways to manage stack usage
+#pragma GCC push_options
+#pragma GCC optimize "no-inline"
+#include <BAN/Errors.h>
+#pragma GCC pop_options
+
 #include <kernel/ACPI/AML/Bytes.h>
 #include <kernel/ACPI/AML/Namespace.h>
 #include <kernel/ACPI/AML/OpRegion.h>
@@ -92,6 +98,39 @@ namespace Kernel::ACPI::AML
 		opregion.as.opregion.offset = region_offset.as.integer.value;
 		opregion.as.opregion.length = region_length.as.integer.value;
 
+		opregion.as.opregion.seg = 0;
+		opregion.as.opregion.bus = 0;
+		opregion.as.opregion.dev = 0;
+		opregion.as.opregion.func = 0;
+
+		if (opregion.as.opregion.address_space == GAS::AddressSpaceID::PCIConfig)
+		{
+			// FIXME: Am I actually allowed to read these here or should I determine
+			//        them on every read/write access
+
+			if (auto seg_res = TRY(Namespace::root_namespace().find_named_object(context.scope, TRY(AML::NameString::from_string("_SEG"_sv)))); seg_res.node != nullptr)
+			{
+				auto seg_node = TRY(convert_node(TRY(evaluate_node(seg_res.path, seg_res.node->node)), ConvInteger, -1));
+				opregion.as.opregion.seg = seg_node.as.integer.value;
+			}
+
+			if (auto bbn_res = TRY(Namespace::root_namespace().find_named_object(context.scope, TRY(AML::NameString::from_string("_BBN"_sv)))); bbn_res.node != nullptr)
+			{
+				auto bbn_node = TRY(convert_node(TRY(evaluate_node(bbn_res.path, bbn_res.node->node)), ConvInteger, -1));
+				opregion.as.opregion.bus = bbn_node.as.integer.value;
+			}
+
+			auto adr_res = TRY(Namespace::root_namespace().find_named_object(context.scope, TRY(AML::NameString::from_string("_ADR"_sv))));
+			if (adr_res.node == nullptr)
+			{
+				dwarnln("No _ADR for PCIConfig OpRegion");
+				return BAN::Error::from_errno(EFAULT);
+			}
+			auto adr_node = TRY(convert_node(TRY(evaluate_node(adr_res.path, adr_res.node->node)), ConvInteger, -1));
+			opregion.as.opregion.dev  = adr_node.as.integer.value >> 16;
+			opregion.as.opregion.func = adr_node.as.integer.value & 0xFF;
+		}
+
 		TRY(Namespace::root_namespace().add_named_object(context, region_name, BAN::move(opregion)));
 
 		return {};
@@ -416,19 +455,19 @@ namespace Kernel::ACPI::AML
 				ASSERT_NOT_REACHED();
 			case GAS::AddressSpaceID::PCIConfig:
 			{
-				// https://uefi.org/htmlspecs/ACPI_Spec_6_4_html/05_ACPI_Software_Programming_Model/ACPI_Software_Programming_Model.html#address-space-format
-				// PCI configuration space is confined to segment 0, bus 0
+				if (opregion.seg != 0)
+				{
+					dwarnln("PCIConfig OpRegion with segment");
+					return BAN::Error::from_errno(ENOTSUP);
+				}
 
-				const uint16_t device   = (byte_offset >> 32) & 0xFFFF;
-				const uint16_t function = (byte_offset >> 16) & 0xFFFF;
-				const uint16_t offset   =  byte_offset        & 0xFFFF;
 				switch (access_size)
 				{
-					case 1: return PCI::PCIManager::get().read_config_byte (0, device, function, offset);
-					case 2: return PCI::PCIManager::get().read_config_word (0, device, function, offset);
-					case 4: return PCI::PCIManager::get().read_config_dword(0, device, function, offset);
+					case 1: return PCI::PCIManager::get().read_config_byte (opregion.bus, opregion.dev, opregion.func, byte_offset);
+					case 2: return PCI::PCIManager::get().read_config_word (opregion.bus, opregion.dev, opregion.func, byte_offset);
+					case 4: return PCI::PCIManager::get().read_config_dword(opregion.bus, opregion.dev, opregion.func, byte_offset);
 					default:
-						dwarnln("{} byte read from PCI {2H}:{2H}:{2H}", device, function, offset);
+						dwarnln("{} byte read from PCI {2H}:{2H}:{2H} offset {2H}", access_size, opregion.bus, opregion.dev, opregion.func, byte_offset);
 						return BAN::Error::from_errno(EINVAL);
 				}
 				ASSERT_NOT_REACHED();
@@ -486,19 +525,19 @@ namespace Kernel::ACPI::AML
 				return {};
 			case GAS::AddressSpaceID::PCIConfig:
 			{
-				// https://uefi.org/htmlspecs/ACPI_Spec_6_4_html/05_ACPI_Software_Programming_Model/ACPI_Software_Programming_Model.html#address-space-format
-				// PCI configuration space is confined to segment 0, bus 0
+				if (opregion.seg != 0)
+				{
+					dwarnln("PCIConfig OpRegion with segment");
+					return BAN::Error::from_errno(ENOTSUP);
+				}
 
-				const uint16_t device   = (byte_offset >> 32) & 0xFFFF;
-				const uint16_t function = (byte_offset >> 16) & 0xFFFF;
-				const uint16_t offset   =  byte_offset        & 0xFFFF;
 				switch (access_size)
 				{
-					case 1: PCI::PCIManager::get().write_config_byte (0, device, function, offset, value); break;
-					case 2: PCI::PCIManager::get().write_config_word (0, device, function, offset, value); break;
-					case 4: PCI::PCIManager::get().write_config_dword(0, device, function, offset, value); break;
+					case 1: PCI::PCIManager::get().write_config_byte (opregion.bus, opregion.dev, opregion.func, byte_offset, value); break;
+					case 2: PCI::PCIManager::get().write_config_word (opregion.bus, opregion.dev, opregion.func, byte_offset, value); break;
+					case 4: PCI::PCIManager::get().write_config_dword(opregion.bus, opregion.dev, opregion.func, byte_offset, value); break;
 					default:
-						dwarnln("{} byte write to PCI {2H}:{2H}:{2H}", device, function, offset);
+						dwarnln("{} byte write to PCI {2H}:{2H}:{2H} offset {2H}", access_size, opregion.bus, opregion.dev, opregion.func, byte_offset);
 						return BAN::Error::from_errno(EINVAL);
 				}
 				return {};

kernel/kernel/APIC.cpp

@@ -540,8 +540,6 @@ namespace Kernel
 	//        this is a hack to allow direct GSI reservation
 	BAN::ErrorOr<uint8_t> APIC::reserve_gsi(uint32_t gsi)
 	{
-		dwarnln("TRYING TO RESERVE GSI {}", gsi);
-
 		size_t irq = 0;
 		for (; irq < 0x100; irq++)
 			if (m_irq_overrides[irq] == gsi)
@@ -553,8 +551,6 @@ namespace Kernel
 			return BAN::Error::from_errno(ENOTSUP);
 		}
 
-		dwarnln("  matches IRQ {}", irq);
-
 		TRY(reserve_irq(irq));
 
 		return irq;

kernel/kernel/BootInfo.cpp

@@ -26,12 +26,21 @@ namespace Kernel
 
 		for (const auto* tag = multiboot2_info.tags; tag->type != MULTIBOOT2_TAG_END; tag = tag->next())
 		{
-			if (tag->type == MULTIBOOT2_TAG_CMDLINE)
+			switch (tag->type)
+			{
+				case MULTIBOOT2_TAG_CMDLINE:
 				{
 					const auto& command_line_tag = *static_cast<const multiboot2_cmdline_tag_t*>(tag);
 					MUST(g_boot_info.command_line.append(command_line_tag.cmdline));
+					break;
 				}
-			else if (tag->type == MULTIBOOT2_TAG_FRAMEBUFFER)
+				case MULTIBOOT2_TAG_MODULES:
+				{
+					const auto& modules_tag = *static_cast<const multiboot2_modules_tag_t*>(tag);
+					MUST(g_boot_info.modules.emplace_back(modules_tag.mod_start, modules_tag.mod_end - modules_tag.mod_start));
+					break;
+				}
+				case MULTIBOOT2_TAG_FRAMEBUFFER:
 				{
 					const auto& framebuffer_tag = *static_cast<const multiboot2_framebuffer_tag_t*>(tag);
 					g_boot_info.framebuffer.address	= framebuffer_tag.framebuffer_addr;
@@ -45,8 +54,9 @@ namespace Kernel
 						g_boot_info.framebuffer.type = FramebufferInfo::Type::Text;
 					else
 						g_boot_info.framebuffer.type = FramebufferInfo::Type::Unknown;
+					break;
 				}
-			else if (tag->type == MULTIBOOT2_TAG_MMAP)
+				case MULTIBOOT2_TAG_MMAP:
 				{
 					const auto& mmap_tag = *static_cast<const multiboot2_mmap_tag_t*>(tag);
 
@@ -66,19 +76,23 @@ namespace Kernel
 						g_boot_info.memory_map_entries[i].length  = mmap_entry.length;
 						g_boot_info.memory_map_entries[i].type    = bios_number_to_memory_type(mmap_entry.type);
 					}
+					break;
 				}
-			else if (tag->type == MULTIBOOT2_TAG_OLD_RSDP)
+				case MULTIBOOT2_TAG_OLD_RSDP:
 				{
 					if (g_boot_info.rsdp.length == 0)
 					{
 						memcpy(&g_boot_info.rsdp, static_cast<const multiboot2_rsdp_tag_t*>(tag)->data, 20);
 						g_boot_info.rsdp.length = 20;
 					}
+					break;
 				}
-			else if (tag->type == MULTIBOOT2_TAG_NEW_RSDP)
+				case MULTIBOOT2_TAG_NEW_RSDP:
 				{
 					const auto& rsdp = *reinterpret_cast<const RSDP*>(static_cast<const multiboot2_rsdp_tag_t*>(tag)->data);
 					memcpy(&g_boot_info.rsdp, &rsdp, BAN::Math::min<uint32_t>(rsdp.length, sizeof(g_boot_info.rsdp)));
+					break;
+				}
 			}
 		}
 

kernel/kernel/Device/FramebufferDevice.cpp

@@ -78,7 +78,7 @@ namespace Kernel
 			KERNEL_OFFSET, UINTPTR_MAX,
 			BAN::Math::div_round_up<size_t>(m_width * m_height * (BANAN_FB_BPP / 8), PAGE_SIZE) * PAGE_SIZE,
 			PageTable::Flags::ReadWrite | PageTable::Flags::Present,
-			true
+			true, false
 		));
 
 		return {};
@@ -128,22 +128,24 @@ namespace Kernel
 	uint32_t FramebufferDevice::get_pixel(uint32_t x, uint32_t y) const
 	{
 		ASSERT(x < m_width && y < m_height);
-		const auto* video_buffer_u8 = reinterpret_cast<const uint8_t*>(m_video_buffer->vaddr());
-		return (video_buffer_u8[(y * m_width + x) * (BANAN_FB_BPP / 8) + 0] << 0)
-		     | (video_buffer_u8[(y * m_width + x) * (BANAN_FB_BPP / 8) + 1] << 8)
-		     | (video_buffer_u8[(y * m_width + x) * (BANAN_FB_BPP / 8) + 2] << 16);
+		static_assert(BANAN_FB_BPP == 32);
+		return reinterpret_cast<uint32_t*>(m_video_buffer->vaddr())[y * m_width + x];
 	}
 
 	void FramebufferDevice::set_pixel(uint32_t x, uint32_t y, uint32_t rgb)
 	{
 		if (x >= m_width || y >= m_height)
 			return;
-		auto* video_buffer_u8 = reinterpret_cast<uint8_t*>(m_video_buffer->vaddr());
-		video_buffer_u8[(y * m_width + x) * (BANAN_FB_BPP / 8) + 0] = rgb >>  0;
-		video_buffer_u8[(y * m_width + x) * (BANAN_FB_BPP / 8) + 1] = rgb >>  8;
-		video_buffer_u8[(y * m_width + x) * (BANAN_FB_BPP / 8) + 2] = rgb >> 16;
-		if constexpr(BANAN_FB_BPP == 32)
-			video_buffer_u8[(y * m_width + x) * (BANAN_FB_BPP / 8) + 3] = rgb >> 24;
+		static_assert(BANAN_FB_BPP == 32);
+		reinterpret_cast<uint32_t*>(m_video_buffer->vaddr())[y * m_width + x] = rgb;
+	}
+
+	void FramebufferDevice::fill(uint32_t rgb)
+	{
+		static_assert(BANAN_FB_BPP == 32);
+		auto* video_buffer_u32 = reinterpret_cast<uint32_t*>(m_video_buffer->vaddr());
+		for (uint32_t i = 0; i < m_width * m_height; i++)
+			video_buffer_u32[i] = rgb;
 	}
 
 	void FramebufferDevice::scroll(int32_t rows, uint32_t rgb)

kernel/kernel/FS/DevFS/FileSystem.cpp

@@ -45,7 +45,7 @@ namespace Kernel
 
 	void DevFileSystem::initialize_device_updater()
 	{
-		Process::create_kernel(
+		auto* updater_thread = MUST(Thread::create_kernel(
 			[](void* _devfs)
 			{
 				auto* devfs = static_cast<DevFileSystem*>(_devfs);
@@ -59,44 +59,39 @@ namespace Kernel
 					SystemTimer::get().sleep_ms(10);
 				}
 			}, s_instance
-		);
+		));
+		MUST(Processor::scheduler().add_thread(updater_thread));
 
-		auto* sync_process = Process::create_kernel();
-
-		sync_process->add_thread(MUST(Thread::create_kernel(
+		auto* disk_sync_thread = MUST(Thread::create_kernel(
 			[](void* _devfs)
 			{
 				auto* devfs = static_cast<DevFileSystem*>(_devfs);
+
+				constexpr uint64_t sync_interval_ms = 10'000;
+				uint64_t next_sync_ms { sync_interval_ms };
 				while (true)
 				{
 					LockGuard _(devfs->m_device_lock);
 					while (!devfs->m_should_sync)
-						devfs->m_sync_thread_blocker.block_indefinite(&devfs->m_device_lock);
+					{
+						const uint64_t current_ms = SystemTimer::get().ms_since_boot();
+						if (devfs->m_should_sync || current_ms >= next_sync_ms)
+							break;
+						devfs->m_sync_thread_blocker.block_with_timeout_ms(next_sync_ms - current_ms, &devfs->m_device_lock);
+					}
 
 					for (auto& device : devfs->m_devices)
 						if (device->is_storage_device())
 							if (auto ret = static_cast<StorageDevice*>(device.ptr())->sync_disk_cache(); ret.is_error())
 								dwarnln("disk sync: {}", ret.error());
 
+					next_sync_ms = SystemTimer::get().ms_since_boot() + sync_interval_ms;
 					devfs->m_should_sync = false;
 					devfs->m_sync_done.unblock();
 				}
-			}, s_instance, sync_process
-		)));
-
-		sync_process->add_thread(MUST(Kernel::Thread::create_kernel(
-			[](void* _devfs)
-			{
-				auto* devfs = static_cast<DevFileSystem*>(_devfs);
-				while (true)
-				{
-					SystemTimer::get().sleep_ms(10'000);
-					devfs->initiate_sync(false);
-				}
-			}, s_instance, sync_process
-		)));
-
-		sync_process->register_to_scheduler();
+			}, s_instance
+		));
+		MUST(Processor::scheduler().add_thread(disk_sync_thread));
 	}
 
 	void DevFileSystem::initiate_sync(bool should_block)

kernel/kernel/FS/Ext2/Inode.cpp

@@ -289,6 +289,26 @@ namespace Kernel
 		return {};
 	}
 
+	BAN::ErrorOr<void> Ext2Inode::chown_impl(uid_t uid, gid_t gid)
+	{
+		if (m_inode.uid == uid && m_inode.gid == gid)
+			return {};
+
+		const auto old_uid = m_inode.uid;
+		const auto old_gid = m_inode.gid;
+
+		m_inode.uid = uid;
+		m_inode.gid = gid;
+		if (auto ret = sync(); ret.is_error())
+		{
+			m_inode.uid = old_uid;
+			m_inode.gid = old_gid;
+			return ret.release_error();
+		}
+
+		return {};
+	}
+
 	BAN::ErrorOr<void> Ext2Inode::utimens_impl(const timespec times[2])
 	{
 		const uint32_t old_times[2] {

kernel/kernel/FS/FAT/FileSystem.cpp

@@ -2,6 +2,7 @@
 #include <kernel/Lock/LockGuard.h>
 
 #include <ctype.h>
+#include <sys/statvfs.h>
 
 namespace Kernel
 {
@@ -73,7 +74,7 @@ namespace Kernel
 	fsfilcnt_t    FATFS::ffree()   const { return 0; } // FIXME
 	fsfilcnt_t    FATFS::favail()  const { return 0; } // FIXME
 	unsigned long FATFS::fsid()    const { return m_type == Type::FAT32 ? m_bpb.ext_32.volume_id : m_bpb.ext_12_16.volume_id; }
-	unsigned long FATFS::flag()    const { return 0; }
+	unsigned long FATFS::flag()    const { return ST_RDONLY; }
 	unsigned long FATFS::namemax() const { return 255; }
 
 	BAN::ErrorOr<BAN::RefPtr<FATFS>> FATFS::create(BAN::RefPtr<BlockDevice> block_device)

kernel/kernel/FS/Inode.cpp

@@ -1,9 +1,11 @@
 #include <kernel/Epoll.h>
+#include <kernel/FS/FileSystem.h>
 #include <kernel/FS/Inode.h>
 #include <kernel/Lock/LockGuard.h>
 #include <kernel/Memory/FileBackedRegion.h>
 
 #include <fcntl.h>
+#include <sys/statvfs.h>
 
 namespace Kernel
 {
@@ -81,6 +83,8 @@ namespace Kernel
 			return BAN::Error::from_errno(ENOTDIR);
 		if (Mode(mode).ifdir())
 			return BAN::Error::from_errno(EINVAL);
+		if (auto* fs = filesystem(); fs && (fs->flag() & ST_RDONLY))
+			return BAN::Error::from_errno(EROFS);
 		return create_file_impl(name, mode, uid, gid);
 	}
 
@@ -91,6 +95,8 @@ namespace Kernel
 			return BAN::Error::from_errno(ENOTDIR);
 		if (!Mode(mode).ifdir())
 			return BAN::Error::from_errno(EINVAL);
+		if (auto* fs = filesystem(); fs && (fs->flag() & ST_RDONLY))
+			return BAN::Error::from_errno(EROFS);
 		return create_directory_impl(name, mode, uid, gid);
 	}
 
@@ -101,6 +107,8 @@ namespace Kernel
 			return BAN::Error::from_errno(ENOTDIR);
 		if (inode->mode().ifdir())
 			return BAN::Error::from_errno(EINVAL);
+		if (auto* fs = filesystem(); fs && (fs->flag() & ST_RDONLY))
+			return BAN::Error::from_errno(EROFS);
 		return link_inode_impl(name, inode);
 	}
 
@@ -111,6 +119,8 @@ namespace Kernel
 			return BAN::Error::from_errno(ENOTDIR);
 		if (name == "."_sv || name == ".."_sv)
 			return BAN::Error::from_errno(EINVAL);
+		if (auto* fs = filesystem(); fs && (fs->flag() & ST_RDONLY))
+			return BAN::Error::from_errno(EROFS);
 		return unlink_impl(name);
 	}
 
@@ -127,6 +137,8 @@ namespace Kernel
 		LockGuard _(m_mutex);
 		if (!mode().iflnk())
 			return BAN::Error::from_errno(EINVAL);
+		if (auto* fs = filesystem(); fs && (fs->flag() & ST_RDONLY))
+			return BAN::Error::from_errno(EROFS);
 		return set_link_target_impl(target);
 	}
 
@@ -207,6 +219,8 @@ namespace Kernel
 		LockGuard _(m_mutex);
 		if (mode().ifdir())
 			return BAN::Error::from_errno(EISDIR);
+		if (auto* fs = filesystem(); fs && (fs->flag() & ST_RDONLY))
+			return BAN::Error::from_errno(EROFS);
 		return write_impl(offset, buffer);
 	}
 
@@ -215,6 +229,8 @@ namespace Kernel
 		LockGuard _(m_mutex);
 		if (mode().ifdir())
 			return BAN::Error::from_errno(EISDIR);
+		if (auto* fs = filesystem(); fs && (fs->flag() & ST_RDONLY))
+			return BAN::Error::from_errno(EROFS);
 		return truncate_impl(size);
 	}
 
@@ -222,18 +238,24 @@ namespace Kernel
 	{
 		ASSERT((mode & Inode::Mode::TYPE_MASK) == 0);
 		LockGuard _(m_mutex);
+		if (auto* fs = filesystem(); fs && (fs->flag() & ST_RDONLY))
+			return BAN::Error::from_errno(EROFS);
 		return chmod_impl(mode);
 	}
 
 	BAN::ErrorOr<void> Inode::chown(uid_t uid, gid_t gid)
 	{
 		LockGuard _(m_mutex);
+		if (auto* fs = filesystem(); fs && (fs->flag() & ST_RDONLY))
+			return BAN::Error::from_errno(EROFS);
 		return chown_impl(uid, gid);
 	}
 
 	BAN::ErrorOr<void> Inode::utimens(const timespec times[2])
 	{
 		LockGuard _(m_mutex);
+		if (auto* fs = filesystem(); fs && (fs->flag() & ST_RDONLY))
+			return BAN::Error::from_errno(EROFS);
 		return utimens_impl(times);
 	}
 

kernel/kernel/FS/TmpFS/FileSystem.cpp

@@ -1,3 +1,4 @@
+#include <BAN/ScopeGuard.h>
 #include <kernel/Device/DeviceNumbers.h>
 #include <kernel/FS/TmpFS/FileSystem.h>
 #include <kernel/Memory/Heap.h>
@@ -105,7 +106,7 @@ namespace Kernel
 	{
 		LockGuard _(m_mutex);
 
-		auto inode_location = find_inode(ino);
+		const auto inode_location = find_inode(ino);
 		PageTable::with_fast_page(inode_location.paddr, [&] {
 			out = PageTable::fast_page_as_sized<TmpInodeInfo>(inode_location.index);
 		});
@@ -115,7 +116,7 @@ namespace Kernel
 	{
 		LockGuard _(m_mutex);
 
-		auto inode_location = find_inode(ino);
+		const auto inode_location = find_inode(ino);
 		PageTable::with_fast_page(inode_location.paddr, [&] {
 			auto& inode_info = PageTable::fast_page_as_sized<TmpInodeInfo>(inode_location.index);
 			inode_info = info;
@@ -126,7 +127,7 @@ namespace Kernel
 	{
 		LockGuard _(m_mutex);
 
-		auto inode_location = find_inode(ino);
+		const auto inode_location = find_inode(ino);
 		PageTable::with_fast_page(inode_location.paddr, [&] {
 			auto& inode_info = PageTable::fast_page_as_sized<TmpInodeInfo>(inode_location.index);
 			ASSERT(inode_info.nlink == 0);
@@ -134,6 +135,7 @@ namespace Kernel
 				ASSERT(paddr == 0);
 			inode_info = {};
 		});
+
 		ASSERT(!m_inode_cache.contains(ino));
 	}
 
@@ -141,26 +143,86 @@ namespace Kernel
 	{
 		LockGuard _(m_mutex);
 
-		constexpr size_t inodes_per_page = PAGE_SIZE / sizeof(TmpInodeInfo);
+		constexpr size_t inode_infos_per_page = PAGE_SIZE / sizeof(TmpInodeInfo);
+		constexpr size_t page_infos_per_page = PAGE_SIZE / sizeof(PageInfo);
 
-		ino_t ino = first_inode;
-		TRY(for_each_indirect_paddr_allocating(m_inode_pages, [&](paddr_t paddr, bool) {
-			BAN::Iteration result = BAN::Iteration::Continue;
+		for (size_t layer0_index = 0; layer0_index < page_infos_per_page; layer0_index++)
+		{
+			PageInfo layer0_page;
+			PageTable::with_fast_page(m_inode_pages.paddr(), [&] {
+				layer0_page = PageTable::fast_page_as_sized<PageInfo>(layer0_index);
+			});
+
+			if (!(layer0_page.flags() & PageInfo::Flags::Present))
+			{
+				if (m_used_pages >= m_max_pages)
+					return BAN::Error::from_errno(ENOSPC);
+				const paddr_t paddr = Heap::get().take_free_page();
+				if (paddr == 0)
+					return BAN::Error::from_errno(ENOMEM);
 				PageTable::with_fast_page(paddr, [&] {
-				for (size_t i = 0; i < inodes_per_page; i++, ino++)
+					memset(PageTable::fast_page_as_ptr(), 0, PAGE_SIZE);
+				});
+				PageTable::with_fast_page(m_inode_pages.paddr(), [&] {
+					auto& page_info = PageTable::fast_page_as_sized<PageInfo>(layer0_index);
+					page_info.set_paddr(paddr);
+					page_info.set_flags(PageInfo::Flags::Present);
+					layer0_page = page_info;
+				});
+				m_used_pages++;
+			}
+
+			for (size_t layer1_index = 0; layer1_index < page_infos_per_page; layer1_index++)
+			{
+				PageInfo layer1_page;
+				PageTable::with_fast_page(layer0_page.paddr(), [&] {
+					layer1_page = PageTable::fast_page_as_sized<PageInfo>(layer1_index);
+				});
+
+				if (!(layer1_page.flags() & PageInfo::Flags::Present))
+				{
+					if (m_used_pages >= m_max_pages)
+						return BAN::Error::from_errno(ENOSPC);
+					const paddr_t paddr = Heap::get().take_free_page();
+					if (paddr == 0)
+						return BAN::Error::from_errno(ENOMEM);
+					PageTable::with_fast_page(paddr, [&] {
+						memset(PageTable::fast_page_as_ptr(), 0, PAGE_SIZE);
+					});
+					PageTable::with_fast_page(layer0_page.paddr(), [&] {
+						auto& page_info = PageTable::fast_page_as_sized<PageInfo>(layer1_index);
+						page_info.set_paddr(paddr);
+						page_info.set_flags(PageInfo::Flags::Present);
+						layer1_page = page_info;
+					});
+					m_used_pages++;
+				}
+
+				size_t layer2_index = SIZE_MAX;
+
+				PageTable::with_fast_page(layer1_page.paddr(), [&] {
+					for (size_t i = 0; i < PAGE_SIZE / sizeof(TmpInodeInfo); i++)
 					{
 						auto& inode_info = PageTable::fast_page_as_sized<TmpInodeInfo>(i);
 						if (inode_info.mode != 0)
 							continue;
 						inode_info = info;
-					result = BAN::Iteration::Break;
+						layer2_index = i;
 						return;
 					}
 				});
-			return result;
-		}, 2));
 
-		return ino;
+				if (layer2_index != SIZE_MAX)
+				{
+					const size_t layer0_offset = layer0_index * inode_infos_per_page * page_infos_per_page;
+					const size_t layer1_offset = layer1_index * inode_infos_per_page;
+					const size_t layer2_offset = layer2_index;
+					return layer0_offset + layer1_offset + layer2_offset + first_inode;
+				}
+			}
+		}
+
+		ASSERT_NOT_REACHED();
 	}
 
 	TmpFileSystem::InodeLocation TmpFileSystem::find_inode(ino_t ino)
@@ -168,16 +230,30 @@ namespace Kernel
 		LockGuard _(m_mutex);
 
 		ASSERT(ino >= first_inode);
-		ASSERT(ino < max_inodes);
+		ASSERT(ino - first_inode < max_inodes);
 
-		constexpr size_t inodes_per_page = PAGE_SIZE / sizeof(TmpInodeInfo);
+		constexpr size_t inode_infos_per_page = PAGE_SIZE / sizeof(TmpInodeInfo);
+		constexpr size_t page_infos_per_page = PAGE_SIZE / sizeof(PageInfo);
+		const size_t layer0_index = (ino - first_inode) / inode_infos_per_page / page_infos_per_page;
+		const size_t layer1_index = (ino - first_inode) / inode_infos_per_page % page_infos_per_page;
+		const size_t layer2_index = (ino - first_inode) % inode_infos_per_page;
+		ASSERT(layer0_index < page_infos_per_page);
 
-		size_t index_of_page = (ino - first_inode) / inodes_per_page;
-		size_t index_in_page = (ino - first_inode) % inodes_per_page;
+		PageInfo layer0_page;
+		PageTable::with_fast_page(m_inode_pages.paddr(), [&] {
+			layer0_page = PageTable::fast_page_as_sized<PageInfo>(layer0_index);
+		});
+		ASSERT(layer0_page.flags() & PageInfo::Flags::Present);
+
+		PageInfo layer1_page;
+		PageTable::with_fast_page(layer0_page.paddr(), [&] {
+			layer1_page = PageTable::fast_page_as_sized<PageInfo>(layer1_index);
+		});
+		ASSERT(layer1_page.flags() & PageInfo::Flags::Present);
 
 		return {
-			.paddr = find_indirect(m_inode_pages, index_of_page, 2),
-			.index = index_in_page
+			.paddr = layer1_page.paddr(),
+			.index = layer2_index,
 		};
 	}
 
@@ -185,145 +261,185 @@ namespace Kernel
 	{
 		LockGuard _(m_mutex);
 
-		constexpr size_t addresses_per_page = PAGE_SIZE / sizeof(PageInfo);
+		ASSERT(index >= first_data_page);
+		ASSERT(index - first_data_page < max_data_pages);
 
-		const size_t index_of_page = (index - first_data_page) / addresses_per_page;
-		const size_t index_in_page = (index - first_data_page) % addresses_per_page;
+		constexpr size_t page_infos_per_page = PAGE_SIZE / sizeof(PageInfo);
+		const size_t layer0_index = (index - first_data_page) / (page_infos_per_page - 1) / page_infos_per_page;
+		const size_t layer1_index = (index - first_data_page) / (page_infos_per_page - 1) % page_infos_per_page;
+		const size_t layer2_index = (index - first_data_page) % (page_infos_per_page - 1);
+		ASSERT(layer0_index < page_infos_per_page);
 
-		paddr_t page_containing = find_indirect(m_data_pages, index_of_page, 2);
+		PageInfo layer0_page;
+		PageTable::with_fast_page(m_data_pages.paddr(), [&] {
+			layer0_page = PageTable::fast_page_as_sized<PageInfo>(layer0_index);
+		});
+		ASSERT(layer0_page.flags() & PageInfo::Flags::Present);
 
-		paddr_t paddr_to_free = 0;
-		PageTable::with_fast_page(page_containing, [&] {
-			auto& page_info = PageTable::fast_page_as_sized<PageInfo>(index_in_page);
+		PageInfo layer1_page;
+		PageTable::with_fast_page(layer0_page.paddr(), [&] {
+			layer1_page = PageTable::fast_page_as_sized<PageInfo>(layer1_index);
+		});
+		ASSERT(layer1_page.flags() & PageInfo::Flags::Present);
+
+		paddr_t page_to_free;
+		PageTable::with_fast_page(layer1_page.paddr(), [&] {
+			auto& allocated_pages = PageTable::fast_page_as_sized<size_t>(page_infos_per_page - 1);
+			ASSERT(allocated_pages > 0);
+			allocated_pages--;
+
+			auto& page_info = PageTable::fast_page_as_sized<PageInfo>(layer2_index);
 			ASSERT(page_info.flags() & PageInfo::Flags::Present);
-			paddr_to_free = page_info.paddr();
-			m_used_pages--;
-
+			page_to_free = page_info.paddr();
 			page_info.set_paddr(0);
 			page_info.set_flags(0);
 		});
-		Heap::get().release_page(paddr_to_free);
-	}
 
-	BAN::ErrorOr<size_t> TmpFileSystem::allocate_block()
-	{
-		LockGuard _(m_mutex);
-
-		size_t result = first_data_page;
-		TRY(for_each_indirect_paddr_allocating(m_data_pages, [&] (paddr_t, bool allocated) {
-			if (allocated)
-				return BAN::Iteration::Break;
-			result++;
-			return BAN::Iteration::Continue;
-		}, 3));
-		return result;
+		Heap::get().release_page(page_to_free);
 	}
 
 	paddr_t TmpFileSystem::find_block(size_t index)
 	{
 		LockGuard _(m_mutex);
 
-		ASSERT(index > 0);
-		return find_indirect(m_data_pages, index - first_data_page, 3);
+		ASSERT(index >= first_data_page);
+		ASSERT(index - first_data_page < max_data_pages);
+
+		constexpr size_t page_infos_per_page = PAGE_SIZE / sizeof(PageInfo);
+		const size_t layer0_index = (index - first_data_page) / (page_infos_per_page - 1) / page_infos_per_page;
+		const size_t layer1_index = (index - first_data_page) / (page_infos_per_page - 1) % page_infos_per_page;
+		const size_t layer2_index = (index - first_data_page) % (page_infos_per_page - 1);
+		ASSERT(layer0_index < page_infos_per_page);
+
+		PageInfo layer0_page;
+		PageTable::with_fast_page(m_data_pages.paddr(), [&] {
+			layer0_page = PageTable::fast_page_as_sized<PageInfo>(layer0_index);
+		});
+		ASSERT(layer0_page.flags() & PageInfo::Flags::Present);
+
+		PageInfo layer1_page;
+		PageTable::with_fast_page(layer0_page.paddr(), [&] {
+			layer1_page = PageTable::fast_page_as_sized<PageInfo>(layer1_index);
+		});
+		ASSERT(layer1_page.flags() & PageInfo::Flags::Present);
+
+		PageInfo layer2_page;
+		PageTable::with_fast_page(layer1_page.paddr(), [&] {
+			layer2_page = PageTable::fast_page_as_sized<PageInfo>(layer2_index);
+		});
+		ASSERT(layer2_page.flags() & PageInfo::Flags::Present);
+
+		return layer2_page.paddr();
 	}
 
-	paddr_t TmpFileSystem::find_indirect(PageInfo root, size_t index, size_t depth)
+	BAN::ErrorOr<size_t> TmpFileSystem::allocate_block()
 	{
 		LockGuard _(m_mutex);
 
-		ASSERT(root.flags() & PageInfo::Flags::Present);
-		if (depth == 0)
-		{
-			ASSERT(index == 0);
-			return root.paddr();
-		}
-
-		constexpr size_t addresses_per_page = PAGE_SIZE / sizeof(PageInfo);
-
-		size_t divisor = 1;
-		for (size_t i = 1; i < depth; i++)
-			divisor *= addresses_per_page;
-
-		size_t index_of_page = index / divisor;
-		size_t index_in_page = index % divisor;
-
-		ASSERT(index_of_page < addresses_per_page);
-
-		PageInfo next;
-		PageTable::with_fast_page(root.paddr(), [&] {
-			next = PageTable::fast_page_as_sized<PageInfo>(index_of_page);
-		});
-
-		return find_indirect(next, index_in_page, depth - 1);
-	}
-
-	template<TmpFuncs::for_each_indirect_paddr_allocating_callback F>
-	BAN::ErrorOr<BAN::Iteration> TmpFileSystem::for_each_indirect_paddr_allocating_internal(PageInfo page_info, F callback, size_t depth)
-	{
-		LockGuard _(m_mutex);
-
-		ASSERT(page_info.flags() & PageInfo::Flags::Present);
-		if (depth == 0)
-		{
-			bool is_new_block = page_info.flags() & PageInfo::Flags::Internal;
-			return callback(page_info.paddr(), is_new_block);
-		}
-
-		for (size_t i = 0; i < PAGE_SIZE / sizeof(PageInfo); i++)
-		{
-			PageInfo next_info;
-			PageTable::with_fast_page(page_info.paddr(), [&] {
-				next_info = PageTable::fast_page_as_sized<PageInfo>(i);
-			});
-
-			if (!(next_info.flags() & PageInfo::Flags::Present))
-			{
 		if (m_used_pages >= m_max_pages)
 			return BAN::Error::from_errno(ENOSPC);
-				paddr_t new_paddr = Heap::get().take_free_page();
-				if (new_paddr == 0)
+
+		const paddr_t new_block = Heap::get().take_free_page();
+		if (new_block == 0)
 			return BAN::Error::from_errno(ENOMEM);
+		PageTable::with_fast_page(new_block, [] {
+			memset(PageTable::fast_page_as_ptr(), 0, PAGE_SIZE);
+		});
+		BAN::ScopeGuard block_deleter([new_block] { Heap::get().release_page(new_block); });
+
+		constexpr size_t page_infos_per_page = PAGE_SIZE / sizeof(PageInfo);
+
+		for (size_t layer0_index = 0; layer0_index < PAGE_SIZE / sizeof(PageInfo); layer0_index++)
+		{
+			PageInfo layer0_page;
+			PageTable::with_fast_page(m_data_pages.paddr(), [&] {
+				layer0_page = PageTable::fast_page_as_sized<PageInfo>(layer0_index);
+			});
+
+			if (!(layer0_page.flags() & PageInfo::Flags::Present))
+			{
+				if (m_used_pages + 1 >= m_max_pages)
+					return BAN::Error::from_errno(ENOSPC);
+				const paddr_t paddr = Heap::get().take_free_page();
+				if (paddr == 0)
+					return BAN::Error::from_errno(ENOMEM);
+				PageTable::with_fast_page(paddr, [&] {
+					memset(PageTable::fast_page_as_ptr(), 0, PAGE_SIZE);
+				});
+				PageTable::with_fast_page(m_data_pages.paddr(), [&] {
+					auto& page_info = PageTable::fast_page_as_sized<PageInfo>(layer0_index);
+					page_info.set_paddr(paddr);
+					page_info.set_flags(PageInfo::Flags::Present);
+					layer0_page = page_info;
+				});
+				m_used_pages++;
+			}
+
+			for (size_t layer1_index = 0; layer1_index < PAGE_SIZE / sizeof(PageInfo); layer1_index++)
+			{
+				PageInfo layer1_page;
+				PageTable::with_fast_page(layer0_page.paddr(), [&] {
+					layer1_page = PageTable::fast_page_as_sized<PageInfo>(layer1_index);
+				});
+
+				if (!(layer1_page.flags() & PageInfo::Flags::Present))
+				{
+					if (m_used_pages + 1 >= m_max_pages)
+						return BAN::Error::from_errno(ENOSPC);
+					const paddr_t paddr = Heap::get().take_free_page();
+					if (paddr == 0)
+						return BAN::Error::from_errno(ENOMEM);
+					PageTable::with_fast_page(paddr, [&] {
+						memset(PageTable::fast_page_as_ptr(), 0, PAGE_SIZE);
+					});
+					PageTable::with_fast_page(layer0_page.paddr(), [&] {
+						auto& page_info = PageTable::fast_page_as_sized<PageInfo>(layer1_index);
+						page_info.set_paddr(paddr);
+						page_info.set_flags(PageInfo::Flags::Present);
+						layer1_page = page_info;
+					});
+					m_used_pages++;
+				}
+
+				size_t layer2_index = SIZE_MAX;
+
+				PageTable::with_fast_page(layer1_page.paddr(), [&] {
+					constexpr size_t pages_per_block = page_infos_per_page - 1;
+
+					auto& allocated_pages = PageTable::fast_page_as_sized<size_t>(pages_per_block);
+					if (allocated_pages == pages_per_block)
+						return;
+
+					for (size_t i = 0; i < pages_per_block; i++)
+					{
+						auto& page_info = PageTable::fast_page_as_sized<PageInfo>(i);
+						if (page_info.flags() & PageInfo::Flags::Present)
+							continue;
+						page_info.set_paddr(new_block);
+						page_info.set_flags(PageInfo::Flags::Present);
+						allocated_pages++;
+						layer2_index = i;
+						return;
+					}
+
+					ASSERT_NOT_REACHED();
+				});
+
+				if (layer2_index != SIZE_MAX)
+				{
+					block_deleter.disable();
+
 					m_used_pages++;
 
-				PageTable::with_fast_page(new_paddr, [&] {
-					memset(PageTable::fast_page_as_ptr(), 0x00, PAGE_SIZE);
-				});
-
-				next_info.set_paddr(new_paddr);
-				next_info.set_flags(PageInfo::Flags::Present);
-
-				PageTable::with_fast_page(page_info.paddr(), [&] {
-					auto& to_update_info = PageTable::fast_page_as_sized<PageInfo>(i);
-					to_update_info = next_info;
-				});
-
-				// Don't sync the internal bit to actual memory
-				next_info.set_flags(PageInfo::Flags::Internal | PageInfo::Flags::Present);
+					const size_t layer0_offset = layer0_index * (page_infos_per_page - 1) * page_infos_per_page;
+					const size_t layer1_offset = layer1_index * (page_infos_per_page - 1);
+					const size_t layer2_offset = layer2_index;
+					return layer0_offset + layer1_offset + layer2_offset + first_data_page;
+				}
+			}
 		}
 
-			auto result = TRY(for_each_indirect_paddr_allocating_internal(next_info, callback, depth - 1));
-			switch (result)
-			{
-				case BAN::Iteration::Continue:
-					break;
-				case BAN::Iteration::Break:
-					return BAN::Iteration::Break;
-				default:
 		ASSERT_NOT_REACHED();
 	}
-		}
-
-		return BAN::Iteration::Continue;
-	}
-
-	template<TmpFuncs::for_each_indirect_paddr_allocating_callback F>
-	BAN::ErrorOr<void> TmpFileSystem::for_each_indirect_paddr_allocating(PageInfo page_info, F callback, size_t depth)
-	{
-		LockGuard _(m_mutex);
-
-		BAN::Iteration result = TRY(for_each_indirect_paddr_allocating_internal(page_info, callback, depth));
-		ASSERT(result == BAN::Iteration::Break);
-		return {};
-	}
 
 }

kernel/kernel/FS/TmpFS/Inode.cpp

@@ -50,7 +50,7 @@ namespace Kernel
 
 	dev_t TmpInode::dev() const
 	{
-		return m_fs.dev();
+		return m_fs.rdev();
 	}
 
 	BAN::ErrorOr<BAN::RefPtr<TmpInode>> TmpInode::create_from_existing(TmpFileSystem& fs, ino_t ino, const TmpInodeInfo& info)
@@ -94,12 +94,19 @@ namespace Kernel
 
 	BAN::ErrorOr<void> TmpInode::chmod_impl(mode_t new_mode)
 	{
-		ASSERT(!(new_mode & Mode::TYPE_MASK));
-		m_inode_info.mode &= ~Mode::TYPE_MASK;
+		ASSERT(!(new_mode & Inode::Mode::TYPE_MASK));
+		m_inode_info.mode &= Inode::Mode::TYPE_MASK;
 		m_inode_info.mode |= new_mode;
 		return {};
 	}
 
+	BAN::ErrorOr<void> TmpInode::chown_impl(uid_t new_uid, gid_t new_gid)
+	{
+		m_inode_info.uid = new_uid;
+		m_inode_info.gid = new_gid;
+		return {};
+	}
+
 	BAN::ErrorOr<void> TmpInode::utimens_impl(const timespec times[2])
 	{
 		if (times[0].tv_nsec != UTIME_OMIT)
@@ -117,30 +124,101 @@ namespace Kernel
 	void TmpInode::free_all_blocks()
 	{
 		for (size_t i = 0; i < TmpInodeInfo::direct_block_count; i++)
-		{
 			if (m_inode_info.block[i])
 				m_fs.free_block(m_inode_info.block[i]);
-			m_inode_info.block[i] = 0;
+		if (size_t block = m_inode_info.block[TmpInodeInfo::direct_block_count + 0])
+			free_indirect_blocks(block, 1);
+		if (size_t block = m_inode_info.block[TmpInodeInfo::direct_block_count + 1])
+			free_indirect_blocks(block, 2);
+		if (size_t block = m_inode_info.block[TmpInodeInfo::direct_block_count + 2])
+			free_indirect_blocks(block, 3);
+		for (auto& block : m_inode_info.block)
+			block = 0;
 	}
-		for (auto block : m_inode_info.block)
-			ASSERT(block == 0);
+
+	void TmpInode::free_indirect_blocks(size_t block, uint32_t depth)
+	{
+		ASSERT(block != 0);
+
+		if (depth == 0)
+		{
+			m_fs.free_block(block);
+			return;
+		}
+
+		const size_t indices_per_block = blksize() / sizeof(size_t);
+		for (size_t index = 0; index < indices_per_block; index++)
+		{
+			size_t next_block;
+			m_fs.with_block_buffer(block, [&](BAN::ByteSpan block_buffer) {
+				next_block = block_buffer.as_span<size_t>()[index];
+			});
+
+			if (next_block == 0)
+				continue;
+
+			free_indirect_blocks(next_block, depth - 1);
+		}
+
+		m_fs.free_block(block);
 	}
 
 	BAN::Optional<size_t> TmpInode::block_index(size_t data_block_index)
 	{
-		ASSERT(data_block_index < TmpInodeInfo::direct_block_count);
-		if (m_inode_info.block[data_block_index])
-			return m_inode_info.block[data_block_index];
+		if (data_block_index < TmpInodeInfo::direct_block_count)
+		{
+			if (m_inode_info.block[data_block_index] == 0)
 				return {};
+			return m_inode_info.block[data_block_index];
+		}
+		data_block_index -= TmpInodeInfo::direct_block_count;
+
+		const size_t indices_per_block = blksize() / sizeof(size_t);
+
+		if (data_block_index < indices_per_block)
+			return block_index_from_indirect(m_inode_info.block[TmpInodeInfo::direct_block_count + 0], data_block_index, 1);
+		data_block_index -= indices_per_block;
+
+		if (data_block_index < indices_per_block * indices_per_block)
+			return block_index_from_indirect(m_inode_info.block[TmpInodeInfo::direct_block_count + 1], data_block_index, 2);
+		data_block_index -= indices_per_block * indices_per_block;
+
+		if (data_block_index < indices_per_block * indices_per_block * indices_per_block)
+			return block_index_from_indirect(m_inode_info.block[TmpInodeInfo::direct_block_count + 2], data_block_index, 3);
+
+		ASSERT_NOT_REACHED();
+	}
+
+	BAN::Optional<size_t> TmpInode::block_index_from_indirect(size_t block, size_t index, uint32_t depth)
+	{
+		if (block == 0)
+			return {};
+		ASSERT(depth >= 1);
+
+		const size_t indices_per_block = blksize() / sizeof(size_t);
+
+		size_t divisor = 1;
+		for (size_t i = 1; i < depth; i++)
+			divisor *= indices_per_block;
+
+		size_t next_block;
+		m_fs.with_block_buffer(block, [&](BAN::ByteSpan block_buffer) {
+			next_block = block_buffer.as_span<size_t>()[(index / divisor) % indices_per_block];
+		});
+
+		if (next_block == 0)
+			return {};
+
+		if (depth == 1)
+			return next_block;
+
+		return block_index_from_indirect(next_block, index, depth - 1);
 	}
 
 	BAN::ErrorOr<size_t> TmpInode::block_index_with_allocation(size_t data_block_index)
 	{
-		if (data_block_index >= TmpInodeInfo::direct_block_count)
+		if (data_block_index < TmpInodeInfo::direct_block_count)
 		{
-			dprintln("only {} blocks supported :D", TmpInodeInfo::direct_block_count);
-			return BAN::Error::from_errno(ENOSPC);
-		}
 			if (m_inode_info.block[data_block_index] == 0)
 			{
 				m_inode_info.block[data_block_index] = TRY(m_fs.allocate_block());
@@ -148,6 +226,59 @@ namespace Kernel
 			}
 			return m_inode_info.block[data_block_index];
 		}
+		data_block_index -= TmpInodeInfo::direct_block_count;
+
+		const size_t indices_per_block = blksize() / sizeof(size_t);
+
+		if (data_block_index < indices_per_block)
+			return block_index_from_indirect_with_allocation(m_inode_info.block[TmpInodeInfo::direct_block_count + 0], data_block_index, 1);
+		data_block_index -= indices_per_block;
+
+		if (data_block_index < indices_per_block * indices_per_block)
+			return block_index_from_indirect_with_allocation(m_inode_info.block[TmpInodeInfo::direct_block_count + 1], data_block_index, 2);
+		data_block_index -= indices_per_block * indices_per_block;
+
+		if (data_block_index < indices_per_block * indices_per_block * indices_per_block)
+			return block_index_from_indirect_with_allocation(m_inode_info.block[TmpInodeInfo::direct_block_count + 2], data_block_index, 3);
+
+		ASSERT_NOT_REACHED();
+	}
+
+	BAN::ErrorOr<size_t> TmpInode::block_index_from_indirect_with_allocation(size_t& block, size_t index, uint32_t depth)
+	{
+		if (block == 0)
+		{
+			block = TRY(m_fs.allocate_block());
+			m_inode_info.blocks++;
+		}
+		ASSERT(depth >= 1);
+
+		const size_t indices_per_block = blksize() / sizeof(size_t);
+
+		size_t divisor = 1;
+		for (size_t i = 1; i < depth; i++)
+			divisor *= indices_per_block;
+
+		size_t next_block;
+		m_fs.with_block_buffer(block, [&](BAN::ByteSpan block_buffer) {
+			next_block = block_buffer.as_span<size_t>()[(index / divisor) % indices_per_block];
+		});
+
+		if (next_block == 0)
+		{
+			next_block = TRY(m_fs.allocate_block());
+			m_inode_info.blocks++;
+
+			m_fs.with_block_buffer(block, [&](BAN::ByteSpan block_buffer) {
+				block_buffer.as_span<size_t>()[(index / divisor) % indices_per_block] = next_block;
+			});
+		}
+
+		if (depth == 1)
+			return next_block;
+
+		return block_index_from_indirect_with_allocation(next_block, index, depth - 1);
+	}
 
 	/* FILE INODE */
 
@@ -234,6 +365,9 @@ namespace Kernel
 
 	BAN::ErrorOr<void> TmpFileInode::truncate_impl(size_t new_size)
 	{
+		// FIXME: if size is decreased, we should probably free
+		//        unused blocks
+
 		m_inode_info.size = new_size;
 		return {};
 	}
@@ -288,7 +422,7 @@ namespace Kernel
 	{
 	}
 
-	BAN::ErrorOr<void> TmpSymlinkInode::set_link_target(BAN::StringView new_target)
+	BAN::ErrorOr<void> TmpSymlinkInode::set_link_target_impl(BAN::StringView new_target)
 	{
 		free_all_blocks();
 		m_inode_info.size = 0;
@@ -500,6 +634,9 @@ namespace Kernel
 			case Mode::IFREG:
 				new_inode = TRY(TmpFileInode::create_new(m_fs, mode, uid, gid));
 				break;
+			case Mode::IFLNK:
+				new_inode = TRY(TmpSymlinkInode::create_new(m_fs, mode, uid, gid, ""_sv));
+				break;
 			case Mode::IFSOCK:
 				new_inode = TRY(TmpSocketInode::create_new(m_fs, mode, uid, gid));
 				break;

kernel/kernel/FS/USTARModule.cpp

@@ -0,0 +1,166 @@
+#include <BAN/ScopeGuard.h>
+#include <kernel/FS/USTARModule.h>
+
+#include <tar.h>
+
+namespace Kernel
+{
+
+	bool is_ustar_boot_module(const BootModule& module)
+	{
+		if (module.start % PAGE_SIZE)
+		{
+			dprintln("ignoring non-page-aligned module");
+			return false;
+		}
+
+		if (module.size < 512)
+			return false;
+
+		bool has_ustar_signature;
+		PageTable::with_fast_page(module.start, [&] {
+			has_ustar_signature = memcmp(PageTable::fast_page_as_ptr(257), "ustar", 5) == 0;
+		});
+
+		return has_ustar_signature;
+	}
+
+	BAN::ErrorOr<void> unpack_boot_module_into_filesystem(BAN::RefPtr<FileSystem> filesystem, const BootModule& module)
+	{
+		ASSERT(is_ustar_boot_module(module));
+
+		auto root_inode = filesystem->root_inode();
+
+		uint8_t* temp_page = static_cast<uint8_t*>(kmalloc(PAGE_SIZE));
+		if (temp_page == nullptr)
+			return BAN::Error::from_errno(ENOMEM);
+		BAN::ScopeGuard _([temp_page] { kfree(temp_page); });
+
+		size_t offset = 0;
+		while (offset + 512 <= module.size)
+		{
+			size_t file_size = 0;
+			mode_t file_mode = 0;
+			uid_t file_uid = 0;
+			gid_t file_gid = 0;
+			uint8_t file_type = 0;
+			char file_path[100 + 1 + 155 + 1] {};
+
+			PageTable::with_fast_page((module.start + offset) & PAGE_ADDR_MASK, [&] {
+				const size_t page_off = offset % PAGE_SIZE;
+
+				const auto parse_octal =
+					[page_off](size_t offset, size_t length) -> size_t
+					{
+						size_t result = 0;
+						for (size_t i = 0; i < length; i++)
+						{
+							const char ch = PageTable::fast_page_as<char>(page_off + offset + i);
+							if (ch == '\0')
+								break;
+							result = (result * 8) + (ch - '0');
+						}
+						return result;
+					};
+
+				if (memcmp(PageTable::fast_page_as_ptr(page_off + 257), "ustar", 5)) {
+					file_size = SIZE_MAX;
+					return;
+				}
+
+				memcpy(file_path, PageTable::fast_page_as_ptr(page_off + 345), 155);
+				const size_t prefix_len = strlen(file_path);
+				file_path[prefix_len] = '/';
+				memcpy(file_path + prefix_len + 1, PageTable::fast_page_as_ptr(page_off), 100);
+
+				file_mode = parse_octal(100, 8);
+				file_uid  = parse_octal(108, 8);
+				file_gid  = parse_octal(116, 8);
+				file_size = parse_octal(124, 12);
+				file_type = PageTable::fast_page_as<char>(page_off + 156);
+			});
+
+			if (file_size == SIZE_MAX)
+				break;
+			if (offset + 512 + file_size > module.size)
+				break;
+
+			auto parent_inode = filesystem->root_inode();
+
+			auto file_path_parts = TRY(BAN::StringView(file_path).split('/'));
+			for (size_t i = 0; i < file_path_parts.size() - 1; i++)
+				parent_inode = TRY(parent_inode->find_inode(file_path_parts[i]));
+
+			switch (file_type)
+			{
+				case REGTYPE:
+				case AREGTYPE: file_mode |= Inode::Mode::IFREG; break;
+				case LNKTYPE:                                   break;
+				case SYMTYPE:  file_mode |= Inode::Mode::IFLNK; break;
+				case CHRTYPE:  file_mode |= Inode::Mode::IFCHR; break;
+				case BLKTYPE:  file_mode |= Inode::Mode::IFBLK; break;
+				case DIRTYPE:  file_mode |= Inode::Mode::IFDIR; break;
+				case FIFOTYPE: file_mode |= Inode::Mode::IFIFO; break;
+				default:
+					ASSERT_NOT_REACHED();
+			}
+
+			auto file_name_sv = file_path_parts.back();
+
+			if (file_type == DIRTYPE)
+			{
+				TRY(parent_inode->create_directory(file_name_sv, file_mode, file_uid, file_gid));
+			}
+			else if (file_type == LNKTYPE)
+			{
+				dwarnln("TODO: hardlink");
+			}
+			else if (file_type == SYMTYPE)
+			{
+				TRY(parent_inode->create_file(file_name_sv, file_mode, file_uid, file_gid));
+
+				char link_target[101] {};
+				const paddr_t paddr = module.start + offset;
+				PageTable::with_fast_page(paddr & PAGE_ADDR_MASK, [&] {
+					memcpy(link_target, PageTable::fast_page_as_ptr((paddr % PAGE_SIZE) + 157), 100);
+				});
+
+				if (link_target[0])
+				{
+					auto inode = TRY(parent_inode->find_inode(file_name_sv));
+					TRY(inode->set_link_target(link_target));
+				}
+			}
+			else
+			{
+				TRY(parent_inode->create_file(file_name_sv, file_mode, file_uid, file_gid));
+
+				if (file_size)
+				{
+					auto inode = TRY(parent_inode->find_inode(file_name_sv));
+
+					size_t nwritten = 0;
+					while (nwritten < file_size)
+					{
+						const paddr_t paddr = module.start + offset + 512 + nwritten;
+						PageTable::with_fast_page(paddr & PAGE_ADDR_MASK, [&] {
+							memcpy(temp_page, PageTable::fast_page_as_ptr(), PAGE_SIZE);
+						});
+
+						const size_t page_off = paddr % PAGE_SIZE;
+						const size_t to_write = BAN::Math::min(file_size - nwritten, PAGE_SIZE - page_off);
+						TRY(inode->write(nwritten, { temp_page + page_off, to_write }));
+						nwritten += to_write;
+					}
+				}
+			}
+
+			offset += 512 + file_size;
+			if (auto rem = offset % 512)
+				offset += 512 - rem;
+		}
+
+		return {};
+	}
+
+}

kernel/kernel/FS/VirtualFileSystem.cpp

@@ -3,6 +3,7 @@
 #include <kernel/FS/DevFS/FileSystem.h>
 #include <kernel/FS/ProcFS/FileSystem.h>
 #include <kernel/FS/TmpFS/FileSystem.h>
+#include <kernel/FS/USTARModule.h>
 #include <kernel/FS/VirtualFileSystem.h>
 #include <kernel/Lock/LockGuard.h>
 #include <kernel/Storage/Partition.h>
@@ -51,7 +52,31 @@ namespace Kernel
 		return BAN::RefPtr<BlockDevice>(static_cast<BlockDevice*>(device_inode.ptr()));
 	}
 
-	static BAN::RefPtr<BlockDevice> find_root_device(BAN::StringView root_path)
+	static BAN::RefPtr<FileSystem> load_fallback_root_filesystem()
+	{
+		if (g_boot_info.modules.empty())
+			panic("No fallback boot modules given");
+
+		auto filesystem_or_error = TmpFileSystem::create(-1, 0755, 0, 0);
+		if (filesystem_or_error.is_error())
+			panic("Failed to create fallback filesystem: {}", filesystem_or_error.error());
+
+		dwarnln("Attempting to load fallback filesystem from {} modules", g_boot_info.modules.size());
+
+		auto filesystem = BAN::RefPtr<FileSystem>::adopt(filesystem_or_error.release_value());
+
+		for (const auto& module : g_boot_info.modules)
+		{
+			if (!is_ustar_boot_module(module))
+				continue;
+			if (auto ret = unpack_boot_module_into_filesystem(filesystem, module); ret.is_error())
+				dwarnln("Failed to unpack boot module: {}", ret.error());
+		}
+
+		return filesystem;
+	}
+
+	static BAN::RefPtr<FileSystem> load_root_filesystem(BAN::StringView root_path)
 	{
 		enum class RootType
 		{
@@ -66,19 +91,26 @@ namespace Kernel
 		{
 			entry = root_path.substring(9);
 			if (entry.size() != 36)
-				panic("Invalid UUID '{}'", entry);
+			{
+				derrorln("Invalid UUID '{}'", entry);
+				return load_fallback_root_filesystem();
+			}
 			type = RootType::PartitionUUID;
 		}
 		else if (root_path.starts_with("/dev/"_sv))
 		{
 			entry = root_path.substring(5);
 			if (entry.empty() || entry.contains('/'))
-				panic("Invalid root path '{}'", root_path);
+			{
+				derrorln("Invalid root path '{}'", root_path);
+				return load_fallback_root_filesystem();
+			}
 			type = RootType::BlockDeviceName;
 		}
 		else
 		{
-			panic("Unsupported root path format '{}'", root_path);
+			derrorln("Unsupported root path format '{}'", root_path);
+			return load_fallback_root_filesystem();
 		}
 
 		constexpr size_t timeout_ms = 10'000;
@@ -99,15 +131,30 @@ namespace Kernel
 			}
 
 			if (!ret.is_error())
-				return ret.release_value();
+			{
+				auto filesystem_or_error = FileSystem::from_block_device(ret.release_value());
+				if (filesystem_or_error.is_error())
+				{
+					derrorln("Could not create filesystem from '{}': {}", root_path, filesystem_or_error.error());
+					return load_fallback_root_filesystem();
+				}
+				return filesystem_or_error.release_value();;
+			}
 
 			if (ret.error().get_error_code() != ENOENT)
-				panic("could not open root device '{}': {}", root_path, ret.error());
+			{
+				derrorln("Could not open root device '{}': {}", root_path, ret.error());
+				return load_fallback_root_filesystem();
+			}
+
+			if (i == 4)
+				dwarnln("Could not find specified root device, waiting for it to get loaded...");
 
 			SystemTimer::get().sleep_ms(sleep_ms);
 		}
 
-		panic("could not find root device '{}' after {} ms", root_path, timeout_ms);
+		derrorln("Could not find root device '{}' after {} ms", root_path, timeout_ms);
+		return load_fallback_root_filesystem();
 	}
 
 	void VirtualFileSystem::initialize(BAN::StringView root_path)
@@ -115,13 +162,9 @@ namespace Kernel
 		ASSERT(!s_instance);
 		s_instance = MUST(BAN::RefPtr<VirtualFileSystem>::create());
 
-		auto root_device = find_root_device(root_path);
-		ASSERT(root_device);
-
-		auto filesystem_result = FileSystem::from_block_device(root_device);
-		if (filesystem_result.is_error())
-			panic("Could not create filesystem from '{}': {}", root_path, filesystem_result.error());
-		s_instance->m_root_fs = filesystem_result.release_value();
+		s_instance->m_root_fs = load_root_filesystem(root_path);
+		if (!s_instance->m_root_fs)
+			panic("Could not load root filesystem");
 
 		Credentials root_creds { 0, 0, 0, 0 };
 		MUST(s_instance->mount(root_creds, &DevFileSystem::get(), "/dev"_sv));

kernel/kernel/GDT.cpp

@@ -1,8 +1,11 @@
 #include <kernel/GDT.h>
+#include <kernel/Memory/Types.h>
 #include <kernel/Processor.h>
 
 #include <string.h>
 
+extern "C" uint8_t g_boot_stack_top[];
+
 namespace Kernel
 {
 
@@ -62,6 +65,7 @@ namespace Kernel
 	{
 		memset(&m_tss, 0x00, sizeof(TaskStateSegment));
 		m_tss.iopb = sizeof(TaskStateSegment);
+		m_tss.ist1 = reinterpret_cast<vaddr_t>(g_boot_stack_top);
 
 		uintptr_t base = reinterpret_cast<uintptr_t>(&m_tss);
 

kernel/kernel/IDT.cpp

@@ -247,7 +247,7 @@ namespace Kernel
 			);
 		}
 
-		if (Thread::current().has_process() && Process::current().is_userspace())
+		if (Thread::current().has_process())
 			process_name = Process::current().name();
 
 #if ARCH(x86_64)
@@ -401,7 +401,7 @@ namespace Kernel
 		Thread::current().load_sse();
 	}
 
-	void IDT::register_interrupt_handler(uint8_t index, void (*handler)())
+	void IDT::register_interrupt_handler(uint8_t index, void (*handler)(), uint8_t ist)
 	{
 		auto& desc = m_idt[index];
 		memset(&desc, 0, sizeof(GateDescriptor));
@@ -412,6 +412,7 @@ namespace Kernel
 		desc.offset2 = (uint32_t)((uintptr_t)handler >> 32);
 #endif
 
+		desc.IST = ist;
 		desc.selector = 0x08;
 		desc.flags = 0x8E;
 	}
@@ -453,6 +454,9 @@ namespace Kernel
 		ISR_LIST_X
 #undef X
 
+		idt->register_interrupt_handler(DoubleFault, isr8, 1);
+		static_assert(DoubleFault == 8);
+
 #define X(num) idt->register_interrupt_handler(IRQ_VECTOR_BASE + num, irq ## num);
 		IRQ_LIST_X
 #undef X

kernel/kernel/Input/PS2/Controller.cpp

@@ -12,7 +12,7 @@
 namespace Kernel::Input
 {
 
-	static constexpr uint64_t s_ps2_timeout_ms = 100;
+	static constexpr uint64_t s_ps2_timeout_ms = 300;
 
 	static PS2Controller* s_instance = nullptr;
 
@@ -238,6 +238,15 @@ namespace Kernel::Input
 		return *s_instance;
 	}
 
+	struct PS2DeviceInitInfo
+	{
+		PS2Controller* controller;
+		bool valid_ports[2];
+		uint8_t scancode_set;
+		uint8_t config;
+		BAN::Atomic<bool> thread_started;
+	};
+
 	BAN::ErrorOr<void> PS2Controller::initialize_impl(uint8_t scancode_set)
 	{
 		constexpr size_t iapc_flag_off = offsetof(ACPI::FADT, iapc_boot_arch);
@@ -315,6 +324,54 @@ namespace Kernel::Input
 		if (!valid_ports[0] && !valid_ports[1])
 			return {};
 
+		// Reserve IRQs
+		if (valid_ports[0] && InterruptController::get().reserve_irq(PS2::IRQ::DEVICE0).is_error())
+		{
+			dwarnln("Could not reserve irq for PS/2 port 1");
+			valid_ports[0] = false;
+		}
+		if (valid_ports[1] && InterruptController::get().reserve_irq(PS2::IRQ::DEVICE1).is_error())
+		{
+			dwarnln("Could not reserve irq for PS/2 port 2");
+			valid_ports[1] = false;
+		}
+
+		PS2DeviceInitInfo info {
+			.controller = this,
+			.valid_ports = { valid_ports[0], valid_ports[1] },
+			.scancode_set = scancode_set,
+			.config = config,
+			.thread_started { false },
+		};
+
+		auto* init_thread = TRY(Thread::create_kernel(
+			[](void* info) {
+				static_cast<PS2DeviceInitInfo*>(info)->controller->device_initialize_task(info);
+			}, &info
+		));
+		TRY(Processor::scheduler().add_thread(init_thread));
+
+		while (!info.thread_started)
+			Processor::pause();
+
+		return {};
+	}
+
+	void PS2Controller::device_initialize_task(void* _info)
+	{
+		bool valid_ports[2];
+		uint8_t scancode_set;
+		uint8_t config;
+
+		{
+			auto& info = *static_cast<PS2DeviceInitInfo*>(_info);
+			valid_ports[0] = info.valid_ports[0];
+			valid_ports[1] = info.valid_ports[1];
+			scancode_set = info.scancode_set;
+			config = info.config;
+			info.thread_started = true;
+		}
+
 		// Initialize devices
 		for (uint8_t device = 0; device < 2; device++)
 		{
@@ -325,7 +382,7 @@ namespace Kernel::Input
 				dwarnln_if(DEBUG_PS2, "PS/2 device enable failed: {}", ret.error());
 				continue;
 			}
-			if (auto res = initialize_device(device, scancode_set); res.is_error())
+			if (auto res = identify_device(device, scancode_set); res.is_error())
 			{
 				dwarnln_if(DEBUG_PS2, "PS/2 device initialization failed: {}", res.error());
 				(void)send_command(device == 0 ? PS2::Command::DISABLE_FIRST_PORT : PS2::Command::DISABLE_SECOND_PORT);
@@ -333,20 +390,8 @@ namespace Kernel::Input
 			}
 		}
 
-		// Reserve IRQs
-		if (m_devices[0] && InterruptController::get().reserve_irq(PS2::IRQ::DEVICE0).is_error())
-		{
-			dwarnln("Could not reserve irq for PS/2 port 1");
-			m_devices[0].clear();
-		}
-		if (m_devices[1] && InterruptController::get().reserve_irq(PS2::IRQ::DEVICE1).is_error())
-		{
-			dwarnln("Could not reserve irq for PS/2 port 2");
-			m_devices[1].clear();
-		}
-
 		if (!m_devices[0] && !m_devices[1])
-			return {};
+			return;
 
 		// Enable irqs on valid devices
 		if (m_devices[0])
@@ -362,21 +407,21 @@ namespace Kernel::Input
 			config |= PS2::Config::INTERRUPT_SECOND_PORT;
 		}
 
-		TRY(send_command(PS2::Command::WRITE_CONFIG, config));
+		if (auto ret = send_command(PS2::Command::WRITE_CONFIG, config); ret.is_error())
+		{
+			dwarnln("PS2 failed to enable interrupts: {}", ret.error());
+			m_devices[0].clear();
+			m_devices[1].clear();
+			return;
+		}
 
 		// Send device initialization sequence after interrupts are enabled
 		for (uint8_t i = 0; i < 2; i++)
-		{
-			if (!m_devices[i])
-				continue;
+			if (m_devices[i])
 				m_devices[i]->send_initialize();
-			DevFileSystem::get().add_device(m_devices[i]);
 	}
 
-		return {};
-	}
-
-	BAN::ErrorOr<void> PS2Controller::initialize_device(uint8_t device, uint8_t scancode_set)
+	BAN::ErrorOr<void> PS2Controller::identify_device(uint8_t device, uint8_t scancode_set)
 	{
 		// Reset device
 		TRY(device_send_byte_and_wait_ack(device, PS2::DeviceCommand::RESET));

kernel/kernel/Input/PS2/Device.cpp

@@ -9,7 +9,14 @@ namespace Kernel::Input
 	PS2Device::PS2Device(PS2Controller& controller, InputDevice::Type type)
 		: InputDevice(type)
 		, m_controller(controller)
-	{ }
+	{
+		DevFileSystem::get().add_device(this);
+	}
+
+	PS2Device::~PS2Device()
+	{
+		DevFileSystem::get().remove_device(this);
+	}
 
 	bool PS2Device::append_command_queue(uint8_t command, uint8_t response_size)
 	{

kernel/kernel/Memory/Heap.cpp

@@ -58,9 +58,13 @@ namespace Kernel
 			if (entry.type != MemoryMapEntry::Type::Available)
 				continue;
 
+			// FIXME: only reserve kernel area and modules, not everything from 0 -> kernel end
 			paddr_t start = entry.address;
 			if (start < (vaddr_t)g_kernel_end - KERNEL_OFFSET + g_boot_info.kernel_paddr)
 				start = (vaddr_t)g_kernel_end - KERNEL_OFFSET + g_boot_info.kernel_paddr;
+			for (const auto& module : g_boot_info.modules)
+				if (start < module.start + module.size)
+					start = module.start + module.size;
 			if (auto rem = start % PAGE_SIZE)
 				start += PAGE_SIZE - rem;
 

kernel/kernel/Memory/VirtualRange.cpp

@@ -4,21 +4,30 @@
 namespace Kernel
 {
 
-	BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> VirtualRange::create_to_vaddr(PageTable& page_table, vaddr_t vaddr, size_t size, PageTable::flags_t flags, bool preallocate_pages)
+	BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> VirtualRange::create_to_vaddr(PageTable& page_table, vaddr_t vaddr, size_t size, PageTable::flags_t flags, bool preallocate_pages, bool add_guard_pages)
 	{
 		ASSERT(size % PAGE_SIZE == 0);
 		ASSERT(vaddr % PAGE_SIZE == 0);
 		ASSERT(vaddr > 0);
 
-		auto result = TRY(BAN::UniqPtr<VirtualRange>::create(page_table, preallocate_pages, vaddr, size, flags));
+		if (add_guard_pages)
+		{
+			vaddr -= PAGE_SIZE;
+			size += 2 * PAGE_SIZE;
+		}
+
+		auto result = TRY(BAN::UniqPtr<VirtualRange>::create(page_table, preallocate_pages, add_guard_pages, vaddr, size, flags));
 		ASSERT(page_table.reserve_range(vaddr, size));
 		TRY(result->initialize());
 
 		return result;
 	}
 
-	BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> VirtualRange::create_to_vaddr_range(PageTable& page_table, vaddr_t vaddr_start, vaddr_t vaddr_end, size_t size, PageTable::flags_t flags, bool preallocate_pages)
+	BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> VirtualRange::create_to_vaddr_range(PageTable& page_table, vaddr_t vaddr_start, vaddr_t vaddr_end, size_t size, PageTable::flags_t flags, bool preallocate_pages, bool add_guard_pages)
 	{
+		if (add_guard_pages)
+			size += 2 * PAGE_SIZE;
+
 		ASSERT(size % PAGE_SIZE == 0);
 		ASSERT(vaddr_start > 0);
 		ASSERT(vaddr_start + size <= vaddr_end);
@@ -31,13 +40,13 @@ namespace Kernel
 		ASSERT(vaddr_start < vaddr_end);
 		ASSERT(vaddr_end - vaddr_start + 1 >= size / PAGE_SIZE);
 
-		vaddr_t vaddr = page_table.reserve_free_contiguous_pages(size / PAGE_SIZE, vaddr_start, vaddr_end);
+		const vaddr_t vaddr = page_table.reserve_free_contiguous_pages(size / PAGE_SIZE, vaddr_start, vaddr_end);
 		if (vaddr == 0)
 			return BAN::Error::from_errno(ENOMEM);
 		ASSERT(vaddr >= vaddr_start);
 		ASSERT(vaddr + size <= vaddr_end);
 
-		auto result_or_error = BAN::UniqPtr<VirtualRange>::create(page_table, preallocate_pages, vaddr, size, flags);
+		auto result_or_error = BAN::UniqPtr<VirtualRange>::create(page_table, preallocate_pages, add_guard_pages, vaddr, size, flags);
 		if (result_or_error.is_error())
 		{
 			page_table.unmap_range(vaddr, size);
@@ -50,9 +59,10 @@ namespace Kernel
 		return result;
 	}
 
-	VirtualRange::VirtualRange(PageTable& page_table, bool preallocated, vaddr_t vaddr, size_t size, PageTable::flags_t flags)
+	VirtualRange::VirtualRange(PageTable& page_table, bool preallocated, bool has_guard_pages, vaddr_t vaddr, size_t size, PageTable::flags_t flags)
 		: m_page_table(page_table)
 		, m_preallocated(preallocated)
+		, m_has_guard_pages(has_guard_pages)
 		, m_vaddr(vaddr)
 		, m_size(size)
 		, m_flags(flags)
@@ -70,26 +80,26 @@ namespace Kernel
 
 	BAN::ErrorOr<void> VirtualRange::initialize()
 	{
-		TRY(m_paddrs.resize(m_size / PAGE_SIZE, 0));
+		TRY(m_paddrs.resize(size() / PAGE_SIZE, 0));
 
 		if (!m_preallocated)
 			return {};
 
-		const size_t page_count = m_size / PAGE_SIZE;
+		const size_t page_count = size() / PAGE_SIZE;
 		for (size_t i = 0; i < page_count; i++)
 		{
 			m_paddrs[i] = Heap::get().take_free_page();
 			if (m_paddrs[i] == 0)
 				return BAN::Error::from_errno(ENOMEM);
-			m_page_table.map_page_at(m_paddrs[i], m_vaddr + i * PAGE_SIZE, m_flags);
+			m_page_table.map_page_at(m_paddrs[i], vaddr() + i * PAGE_SIZE, m_flags);
 		}
 
 		if (&PageTable::current() == &m_page_table || &PageTable::kernel() == &m_page_table)
-			memset(reinterpret_cast<void*>(m_vaddr), 0, m_size);
+			memset(reinterpret_cast<void*>(vaddr()), 0, size());
 		else
 		{
-			const size_t page_count = m_size / PAGE_SIZE;
-			for (size_t i = 0; i < page_count; i++)
+			const size_t page_count = size() / PAGE_SIZE;
+			for (size_t i = m_has_guard_pages; i < page_count; i++)
 			{
 				PageTable::with_fast_page(m_paddrs[i], [&] {
 					memset(PageTable::fast_page_as_ptr(), 0, PAGE_SIZE);
@@ -107,10 +117,10 @@ namespace Kernel
 
 		SpinLockGuard _(m_lock);
 
-		auto result = TRY(create_to_vaddr(page_table, m_vaddr, m_size, m_flags, m_preallocated));
+		auto result = TRY(create_to_vaddr(page_table, vaddr(), size(), m_flags, m_preallocated, m_has_guard_pages));
 
-		const size_t page_count = m_size / PAGE_SIZE;
-		for (size_t i = 0; i < page_count; i++)
+		const size_t page_count = size() / PAGE_SIZE;
+		for (size_t i = m_has_guard_pages; i < page_count; i++)
 		{
 			if (m_paddrs[i] == 0)
 				continue;
@@ -119,11 +129,11 @@ namespace Kernel
 				result->m_paddrs[i] = Heap::get().take_free_page();
 				if (result->m_paddrs[i] == 0)
 					return BAN::Error::from_errno(ENOMEM);
-				result->m_page_table.map_page_at(result->m_paddrs[i], m_vaddr + i * PAGE_SIZE, m_flags);
+				result->m_page_table.map_page_at(result->m_paddrs[i], vaddr() + i * PAGE_SIZE, m_flags);
 			}
 
 			PageTable::with_fast_page(result->m_paddrs[i], [&] {
-				memcpy(PageTable::fast_page_as_ptr(), reinterpret_cast<void*>(m_vaddr + i * PAGE_SIZE), PAGE_SIZE);
+				memcpy(PageTable::fast_page_as_ptr(), reinterpret_cast<void*>(vaddr() + i * PAGE_SIZE), PAGE_SIZE);
 			});
 		}
 
@@ -137,7 +147,7 @@ namespace Kernel
 		ASSERT(contains(vaddr));
 		ASSERT(&PageTable::current() == &m_page_table);
 
-		const size_t index = (vaddr - m_vaddr) / PAGE_SIZE;
+		const size_t index = (vaddr - this->vaddr()) / PAGE_SIZE;
 		ASSERT(m_paddrs[index] == 0);
 
 		SpinLockGuard _(m_lock);

kernel/kernel/Networking/ARPTable.cpp

@@ -18,14 +18,14 @@ namespace Kernel
 	BAN::ErrorOr<BAN::UniqPtr<ARPTable>> ARPTable::create()
 	{
 		auto arp_table = TRY(BAN::UniqPtr<ARPTable>::create());
-		arp_table->m_process = Process::create_kernel(
+		arp_table->m_thread = TRY(Thread::create_kernel(
 			[](void* arp_table_ptr)
 			{
 				auto& arp_table = *reinterpret_cast<ARPTable*>(arp_table_ptr);
 				arp_table.packet_handle_task();
 			}, arp_table.ptr()
-		);
-		ASSERT(arp_table->m_process);
+		));
+		TRY(Processor::scheduler().add_thread(arp_table->m_thread));
 		return arp_table;
 	}
 
@@ -35,9 +35,9 @@ namespace Kernel
 
 	ARPTable::~ARPTable()
 	{
-		if (m_process)
-			m_process->exit(0, SIGKILL);
-		m_process = nullptr;
+		if (m_thread)
+			m_thread->add_signal(SIGKILL);
+		m_thread = nullptr;
 	}
 
 	BAN::ErrorOr<BAN::MACAddress> ARPTable::get_mac_from_ipv4(NetworkInterface& interface, BAN::IPv4Address ipv4_address)

kernel/kernel/Networking/IPv4Layer.cpp

@@ -21,21 +21,21 @@ namespace Kernel
 	BAN::ErrorOr<BAN::UniqPtr<IPv4Layer>> IPv4Layer::create()
 	{
 		auto ipv4_manager = TRY(BAN::UniqPtr<IPv4Layer>::create());
-		ipv4_manager->m_process = Process::create_kernel(
+		ipv4_manager->m_thread = TRY(Thread::create_kernel(
 			[](void* ipv4_manager_ptr)
 			{
 				auto& ipv4_manager = *reinterpret_cast<IPv4Layer*>(ipv4_manager_ptr);
 				ipv4_manager.packet_handle_task();
 			}, ipv4_manager.ptr()
-		);
-		ASSERT(ipv4_manager->m_process);
+		));
+		TRY(Processor::scheduler().add_thread(ipv4_manager->m_thread));
 		ipv4_manager->m_pending_packet_buffer = TRY(VirtualRange::create_to_vaddr_range(
 			PageTable::kernel(),
 			KERNEL_OFFSET,
 			~(uintptr_t)0,
 			pending_packet_buffer_size,
 			PageTable::Flags::ReadWrite | PageTable::Flags::Present,
-			true
+			true, false
 		));
 		ipv4_manager->m_arp_table = TRY(ARPTable::create());
 		return ipv4_manager;
@@ -46,9 +46,9 @@ namespace Kernel
 
 	IPv4Layer::~IPv4Layer()
 	{
-		if (m_process)
-			m_process->exit(0, SIGKILL);
-		m_process = nullptr;
+		if (m_thread)
+			m_thread->add_signal(SIGKILL);
+		m_thread = nullptr;
 	}
 
 	void IPv4Layer::add_ipv4_header(BAN::ByteSpan packet, BAN::IPv4Address src_ipv4, BAN::IPv4Address dst_ipv4, uint8_t protocol) const

kernel/kernel/Networking/Loopback.cpp

@@ -16,7 +16,7 @@ namespace Kernel
 			BAN::numeric_limits<vaddr_t>::max(),
 			buffer_size,
 			PageTable::Flags::ReadWrite | PageTable::Flags::Present,
-			true
+			true, false
 		));
 		loopback->set_ipv4_address({ 127, 0, 0, 1 });
 		loopback->set_netmask({ 255, 0, 0, 0 });

kernel/kernel/Networking/TCPSocket.cpp

@@ -1,6 +1,7 @@
 #include <kernel/Lock/LockGuard.h>
 #include <kernel/Networking/NetworkManager.h>
 #include <kernel/Networking/TCPSocket.h>
+#include <kernel/Process.h>
 #include <kernel/Random.h>
 #include <kernel/Timer/Timer.h>
 
@@ -31,7 +32,7 @@ namespace Kernel
 			~(vaddr_t)0,
 			s_recv_window_buffer_size,
 			PageTable::Flags::ReadWrite | PageTable::Flags::Present,
-			true
+			true, false
 		));
 		socket->m_recv_window.scale_shift = PAGE_SIZE_SHIFT; // use PAGE_SIZE windows
 		socket->m_send_window.buffer = TRY(VirtualRange::create_to_vaddr_range(
@@ -40,14 +41,15 @@ namespace Kernel
 			~(vaddr_t)0,
 			s_send_window_buffer_size,
 			PageTable::Flags::ReadWrite | PageTable::Flags::Present,
-			true
+			true, false
 		));
-		socket->m_process = Process::create_kernel(
+		socket->m_thread = TRY(Thread::create_kernel(
 			[](void* socket_ptr)
 			{
 				reinterpret_cast<TCPSocket*>(socket_ptr)->process_task();
 			}, socket.ptr()
-		);
+		));
+		TRY(Processor::scheduler().add_thread(socket->m_thread));
 		// hack to keep socket alive until its process starts
 		socket->ref();
 		return socket;
@@ -63,7 +65,7 @@ namespace Kernel
 	TCPSocket::~TCPSocket()
 	{
 		ASSERT(!is_bound());
-		ASSERT(m_process == nullptr);
+		ASSERT(m_thread == nullptr);
 		dprintln_if(DEBUG_TCP, "Socket destroyed");
 	}
 
@@ -620,7 +622,7 @@ namespace Kernel
 			dprintln_if(DEBUG_TCP, "Socket unbound");
 		}
 
-		m_process = nullptr;
+		m_thread = nullptr;
 	}
 
 	void TCPSocket::remove_listen_child(BAN::RefPtr<TCPSocket> socket)
@@ -652,7 +654,7 @@ namespace Kernel
 
 		LockGuard _(m_mutex);
 
-		while (m_process)
+		while (m_thread)
 		{
 			const uint64_t current_ms = SystemTimer::get().ms_since_boot();
 

kernel/kernel/Networking/UDPSocket.cpp

@@ -17,7 +17,7 @@ namespace Kernel
 			~(uintptr_t)0,
 			packet_buffer_size,
 			PageTable::Flags::ReadWrite | PageTable::Flags::Present,
-			true
+			true, false
 		));
 		return socket;
 	}

kernel/kernel/Networking/UNIX/Socket.cpp

@@ -3,6 +3,7 @@
 #include <kernel/Lock/SpinLockAsMutex.h>
 #include <kernel/Networking/NetworkManager.h>
 #include <kernel/Networking/UNIX/Socket.h>
+#include <kernel/Process.h>
 #include <kernel/Scheduler.h>
 
 #include <fcntl.h>
@@ -28,7 +29,7 @@ namespace Kernel
 			~(uintptr_t)0,
 			s_packet_buffer_size,
 			PageTable::Flags::ReadWrite | PageTable::Flags::Present,
-			true
+			true, false
 		));
 		return socket;
 	}

kernel/kernel/OpenFileDescriptorSet.cpp

@@ -3,6 +3,7 @@
 #include <kernel/Lock/LockGuard.h>
 #include <kernel/Networking/NetworkManager.h>
 #include <kernel/OpenFileDescriptorSet.h>
+#include <kernel/Process.h>
 
 #include <fcntl.h>
 #include <sys/file.h>

kernel/kernel/Process.cpp

@@ -97,21 +97,6 @@ namespace Kernel
 			MUST(Processor::scheduler().add_thread(thread));
 	}
 
-	Process* Process::create_kernel()
-	{
-		auto* process = create_process({ 0, 0, 0, 0 }, 0);
-		return process;
-	}
-
-	Process* Process::create_kernel(entry_t entry, void* data)
-	{
-		auto* process = create_process({ 0, 0, 0, 0 }, 0);
-		auto* thread = MUST(Thread::create_kernel(entry, data, process));
-		process->add_thread(thread);
-		process->register_to_scheduler();
-		return process;
-	}
-
 	BAN::ErrorOr<Process*> Process::create_userspace(const Credentials& credentials, BAN::StringView path, BAN::Span<BAN::StringView> arguments)
 	{
 		auto* process = create_process(credentials, 0);
@@ -609,7 +594,6 @@ namespace Kernel
 		forked->m_page_table = BAN::move(page_table);
 		forked->m_open_file_descriptors = BAN::move(*open_file_descriptors);
 		forked->m_mapped_regions = BAN::move(mapped_regions);
-		forked->m_is_userspace = m_is_userspace;
 		forked->m_has_called_exec = false;
 		memcpy(forked->m_signal_handlers, m_signal_handlers, sizeof(m_signal_handlers));
 

kernel/kernel/Scheduler.cpp

@@ -4,6 +4,7 @@
 #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>
 
@@ -118,7 +119,7 @@ namespace Kernel
 
 	BAN::ErrorOr<void> Scheduler::initialize()
 	{
-		m_idle_thread = TRY(Thread::create_kernel([](void*) { asm volatile("1: hlt; jmp 1b"); }, nullptr, nullptr));
+		m_idle_thread = TRY(Thread::create_kernel([](void*) { asm volatile("1: hlt; jmp 1b"); }, nullptr));
 		ASSERT(m_idle_thread);
 
 		size_t processor_index = 0;
@@ -307,8 +308,11 @@ namespace Kernel
 			while (!m_block_queue.empty() && current_ns >= m_block_queue.front()->wake_time_ns)
 			{
 				auto* node = m_block_queue.pop_front();
+				{
+					SpinLockGuard _(node->blocker_lock);
 					if (node->blocker)
 						node->blocker->remove_blocked_thread(node);
+				}
 				node->blocked = false;
 				update_most_loaded_node_queue(node, &m_run_queue);
 				m_run_queue.add_thread_to_back(node);
@@ -336,8 +340,11 @@ namespace Kernel
 				return;
 			if (node != m_current)
 				m_block_queue.remove_node(node);
+			{
+				SpinLockGuard _(node->blocker_lock);
 				if (node->blocker)
 					node->blocker->remove_blocked_thread(node);
+			}
 			node->blocked = false;
 			if (node != m_current)
 				m_run_queue.add_thread_to_back(node);
@@ -618,8 +625,13 @@ namespace Kernel
 
 		m_current->blocked = true;
 		m_current->wake_time_ns = wake_time_ns;
+
+		{
+			SpinLockGuard _(m_current->blocker_lock);
 			if (blocker)
 				blocker->add_thread_to_block_queue(m_current);
+		}
+
 		update_most_loaded_node_queue(m_current, &m_block_queue);
 
 		uint32_t lock_depth = 0;
@@ -642,10 +654,7 @@ namespace Kernel
 
 	void Scheduler::unblock_thread(Thread* thread)
 	{
-		auto state = Processor::get_interrupt_state();
-		Processor::set_interrupt_state(InterruptState::Disabled);
 		unblock_thread(thread->m_scheduler_node);
-		Processor::set_interrupt_state(state);
 	}
 
 	Thread& Scheduler::current_thread()

kernel/kernel/Storage/ATA/ATABus.cpp

@@ -126,26 +126,7 @@ namespace Kernel
 	{
 		if (io_read(ATA_PORT_STATUS) & ATA_STATUS_ERR)
 			dprintln("ATA Error: {}", error());
-
-		bool expected { false };
-		[[maybe_unused]] bool success = m_has_got_irq.compare_exchange(expected, true);
-		ASSERT(success);
-	}
-
-	BAN::ErrorOr<void> ATABus::block_until_irq()
-	{
-		const uint64_t timeout_ms = SystemTimer::get().ms_since_boot() + s_ata_timeout_ms;
-
-		bool expected { true };
-		while (!m_has_got_irq.compare_exchange(expected, false))
-		{
-			if (SystemTimer::get().ms_since_boot() >= timeout_ms)
-				return BAN::Error::from_errno(ETIMEDOUT);
-			Processor::pause();
-			expected = true;
-		}
-
-		return {};
+		m_thread_blocker.unblock();
 	}
 
 	uint8_t ATABus::io_read(uint16_t port)
@@ -192,22 +173,30 @@ namespace Kernel
 		for (uint32_t i = 0; i < 4; i++)
 			io_read(ATA_PORT_ALT_STATUS);
 
-		uint64_t timeout = SystemTimer::get().ms_since_boot() + s_ata_timeout_ms;
+		const uint64_t start_ms = SystemTimer::get().ms_since_boot();
+		const uint64_t timeout_ms = start_ms + s_ata_timeout_ms;
 
-		uint8_t status;
-		while ((status = io_read(ATA_PORT_STATUS)) & ATA_STATUS_BSY)
-			if (SystemTimer::get().ms_since_boot() >= timeout)
-				return BAN::Error::from_errno(ETIMEDOUT);
-
-		while (wait_drq && !(status & ATA_STATUS_DRQ))
+		for (;;)
 		{
-			if (SystemTimer::get().ms_since_boot() >= timeout)
-				return BAN::Error::from_errno(ETIMEDOUT);
+			const uint8_t status = io_read(ATA_PORT_ALT_STATUS);
+			if (status & ATA_STATUS_BSY)
+				goto drive_not_ready;
+			if (!wait_drq || (status & ATA_STATUS_DRQ))
+				break;
 			if (status & ATA_STATUS_ERR)
 				return error();
 			if (status & ATA_STATUS_DF)
 				return BAN::Error::from_errno(EIO);
-			status = io_read(ATA_PORT_STATUS);
+
+		drive_not_ready:
+			const uint64_t current_ms = SystemTimer::get().ms_since_boot();
+			if (current_ms >= timeout_ms)
+				return BAN::Error::from_errno(ETIMEDOUT);
+			// NODE: poll for 5 milliseconds, then just block
+			//       until timeout or irq
+			if (current_ms < start_ms + 5)
+				continue;
+			m_thread_blocker.block_with_timeout_ms(timeout_ms - current_ms, nullptr);
 		}
 
 		return {};
@@ -249,7 +238,7 @@ namespace Kernel
 
 		for (uint32_t sector = 0; sector < sector_count; sector++)
 		{
-			TRY(block_until_irq());
+			TRY(wait(true));
 			read_buffer(ATA_PORT_DATA, (uint16_t*)buffer.data() + sector * device.words_per_sector(), device.words_per_sector());
 		}
 
@@ -269,12 +258,12 @@ namespace Kernel
 
 		for (uint32_t sector = 0; sector < sector_count; sector++)
 		{
+			TRY(wait(true));
 			write_buffer(ATA_PORT_DATA, (uint16_t*)buffer.data() + sector * device.words_per_sector(), device.words_per_sector());
-			TRY(block_until_irq());
 		}
 
+		TRY(wait(false));
 		io_write(ATA_PORT_COMMAND, ATA_COMMAND_CACHE_FLUSH);
-		TRY(block_until_irq());
 
 		return {};
 	}
@@ -310,9 +299,10 @@ namespace Kernel
 			io_lba2   = (cylinder >> 8) & 0xFF;
 		}
 
+		TRY(wait(false));
+
 		io_write(ATA_PORT_DRIVE_SELECT, io_select);
 		select_delay();
-		io_write(ATA_PORT_CONTROL, 0);
 
 		io_write(ATA_PORT_SECTOR_COUNT, sector_count);
 		io_write(ATA_PORT_LBA0, io_lba0);

kernel/kernel/Terminal/FramebufferTerminal.cpp

@@ -79,9 +79,7 @@ namespace Kernel
 		for (auto& pixel : m_cursor_data)
 			pixel = color.rgb;
 
-		for (uint32_t y = 0; y < m_framebuffer_device->height(); y++)
-			for (uint32_t x = 0; x < m_framebuffer_device->width(); x++)
-				m_framebuffer_device->set_pixel(x, y, color.rgb);
+		m_framebuffer_device->fill(color.rgb);
 		m_framebuffer_device->sync_pixels_full();
 
 		if (m_cursor_shown)

kernel/kernel/Terminal/PseudoTerminal.cpp

@@ -44,7 +44,8 @@ namespace Kernel
 			PageTable::kernel(),
 			KERNEL_OFFSET, static_cast<vaddr_t>(-1),
 			16 * PAGE_SIZE,
-			PageTable::Flags::ReadWrite | PageTable::Flags::Present, true
+			PageTable::Flags::ReadWrite | PageTable::Flags::Present,
+			true, false
 		));
 		auto pts_master = TRY(BAN::RefPtr<PseudoTerminalMaster>::create(BAN::move(pts_master_buffer), mode, uid, gid));
 		DevFileSystem::get().remove_from_cache(pts_master);

kernel/kernel/Terminal/TTY.cpp

@@ -161,7 +161,7 @@ namespace Kernel
 		static bool initialized = false;
 		ASSERT(!initialized);
 
-		auto* thread = MUST(Thread::create_kernel(&TTY::keyboard_task, nullptr, nullptr));
+		auto* thread = MUST(Thread::create_kernel(&TTY::keyboard_task, nullptr));
 		MUST(Processor::scheduler().add_thread(thread));
 
 		DevFileSystem::get().add_inode("tty", MUST(DevTTY::create(0666, 0, 0)));

kernel/kernel/Thread.cpp

@@ -69,10 +69,10 @@ namespace Kernel
 		s_default_sse_storage_initialized = true;
 	}
 
-	BAN::ErrorOr<Thread*> Thread::create_kernel(entry_t entry, void* data, Process* process)
+	BAN::ErrorOr<Thread*> Thread::create_kernel(entry_t entry, void* data)
 	{
 		// Create the thread object
-		Thread* thread = new Thread(s_next_tid++, process);
+		Thread* thread = new Thread(s_next_tid++, nullptr);
 		if (thread == nullptr)
 			return BAN::Error::from_errno(ENOMEM);
 		BAN::ScopeGuard thread_deleter([thread] { delete thread; });
@@ -84,7 +84,7 @@ namespace Kernel
 			~(uintptr_t)0,
 			kernel_stack_size,
 			PageTable::Flags::ReadWrite | PageTable::Flags::Present,
-			true
+			true, true
 		));
 
 		// Initialize stack for returning
@@ -124,7 +124,7 @@ namespace Kernel
 			0x200000, USERSPACE_END,
 			kernel_stack_size,
 			PageTable::Flags::ReadWrite | PageTable::Flags::Present,
-			true
+			true, true
 		));
 
 		thread->m_userspace_stack = TRY(VirtualRange::create_to_vaddr_range(
@@ -132,7 +132,7 @@ namespace Kernel
 			0x200000, USERSPACE_END,
 			userspace_stack_size,
 			PageTable::Flags::UserSupervisor | PageTable::Flags::ReadWrite | PageTable::Flags::Present,
-			true
+			true, true
 		));
 
 		thread_deleter.disable();
@@ -285,6 +285,9 @@ namespace Kernel
 		// auxv
 		needed_size += auxv.size() * sizeof(LibELF::AuxiliaryVector);
 
+		if (auto rem = needed_size % alignof(char*))
+			needed_size += alignof(char*) - rem;
+
 		if (needed_size > m_userspace_stack->size())
 			return BAN::Error::from_errno(ENOBUFS);
 
@@ -358,8 +361,6 @@ namespace Kernel
 			stack_push_str(envp[i]);
 		}
 
-		ASSERT(vaddr == userspace_stack_top());
-
 		setup_exec(entry, userspace_stack_top() - needed_size);
 
 		return {};

kernel/kernel/ThreadBlocker.cpp

@@ -1,4 +1,5 @@
 #include <kernel/Processor.h>
+#include <kernel/SchedulerQueueNode.h>
 #include <kernel/ThreadBlocker.h>
 #include <kernel/Timer/Timer.h>
 
@@ -22,71 +23,60 @@ namespace Kernel
 
 	void ThreadBlocker::unblock()
 	{
-		SchedulerQueue::Node* block_chain;
+		decltype(m_block_chain) temp_block_chain;
+		size_t temp_block_chain_length { 0 };
 
 		{
 			SpinLockGuard _(m_lock);
-			block_chain = m_block_chain;
-			m_block_chain = nullptr;
+			for (size_t i = 0; i < m_block_chain_length; i++)
+				temp_block_chain[i] = m_block_chain[i];
+			temp_block_chain_length = m_block_chain_length;
+			m_block_chain_length = 0;
 		}
 
-		for (auto* node = block_chain; node;)
-		{
-			ASSERT(node->blocked);
-
-			auto* next = node->block_chain_next;
-			node->blocker = nullptr;
-			node->block_chain_next = nullptr;
-			node->block_chain_prev = nullptr;
-			Processor::scheduler().unblock_thread(node);
-			node = next;
-		}
+		for (size_t i = 0; i < temp_block_chain_length; i++)
+			Processor::scheduler().unblock_thread(temp_block_chain[i]);
 	}
 
 	void ThreadBlocker::add_thread_to_block_queue(SchedulerQueue::Node* node)
 	{
+		ASSERT(node->blocker_lock.current_processor_has_lock());
+
+		SpinLockGuard _(m_lock);
+
+		ASSERT(m_block_chain_length < sizeof(m_block_chain) / sizeof(m_block_chain[0]));
+
 		ASSERT(node);
 		ASSERT(node->blocked);
 		ASSERT(node->blocker == nullptr);
-		ASSERT(node->block_chain_prev == nullptr);
-		ASSERT(node->block_chain_next == nullptr);
 
-		SpinLockGuard _(m_lock);
+		for (size_t i = 0 ; i < m_block_chain_length; i++)
+			ASSERT(m_block_chain[i] != node);
+		m_block_chain[m_block_chain_length++] = node;
+
 		node->blocker = this;
-		node->block_chain_prev = nullptr;
-		node->block_chain_next = m_block_chain;
-		if (m_block_chain)
-			m_block_chain->block_chain_prev = node;
-		m_block_chain = node;
 	}
 
 	void ThreadBlocker::remove_blocked_thread(SchedulerQueue::Node* node)
 	{
+		ASSERT(node->blocker_lock.current_processor_has_lock());
+
 		SpinLockGuard _(m_lock);
 
 		ASSERT(node);
 		ASSERT(node->blocked);
 		ASSERT(node->blocker == this);
 
-		if (node == m_block_chain)
+		for (size_t i = 0 ; i < m_block_chain_length; i++)
 		{
-			ASSERT(node->block_chain_prev == nullptr);
-			m_block_chain = node->block_chain_next;
-			if (m_block_chain)
-				m_block_chain->block_chain_prev = nullptr;
-		}
-		else
-		{
-			ASSERT(node->block_chain_prev);
-			node->block_chain_prev->block_chain_next = node->block_chain_next;
-			if (node->block_chain_next)
-				node->block_chain_next->block_chain_prev = node->block_chain_prev;
+			if (m_block_chain[i] != node)
+				continue;
+			for (size_t j = i + 1; j < m_block_chain_length; j++)
+				m_block_chain[j - 1] = m_block_chain[j];
+			m_block_chain_length--;
 		}
 
 		node->blocker = nullptr;
-		node->block_chain_next = nullptr;
-		node->block_chain_prev = nullptr;
 	}
 
-
 }

kernel/kernel/USB/Hub/HubDriver.cpp

@@ -215,9 +215,8 @@ namespace Kernel
 			m_changed_ports |= 1u << port_id;
 		}
 
-		m_port_updater = Process::create_kernel([](void* data) { reinterpret_cast<USBHubDriver*>(data)->port_updater_task(); }, this);
-		if (m_port_updater == nullptr)
-			return BAN::Error::from_errno(ENOMEM);
+		m_port_updater = TRY(Thread::create_kernel([](void* data) { reinterpret_cast<USBHubDriver*>(data)->port_updater_task(); }, this));
+		TRY(Processor::scheduler().add_thread(m_port_updater));
 
 		return {};
 	}

kernel/kernel/USB/XHCI/Controller.cpp

@@ -135,9 +135,8 @@ namespace Kernel
 		while (operational.usbsts & XHCI::USBSTS::HCHalted)
 			continue;
 
-		m_port_updater = Process::create_kernel([](void* data) { reinterpret_cast<XHCIController*>(data)->port_updater_task(); }, this);
-		if (m_port_updater == nullptr)
-			return BAN::Error::from_errno(ENOMEM);
+		m_port_updater = TRY(Thread::create_kernel([](void* data) { reinterpret_cast<XHCIController*>(data)->port_updater_task(); }, this));
+		TRY(Processor::scheduler().add_thread(m_port_updater));
 
 		return {};
 	}

kernel/kernel/kernel.cpp

@@ -199,7 +199,8 @@ extern "C" void kernel_main(uint32_t boot_magic, uint32_t boot_info)
 	Processor::wait_until_processors_ready();
 	MUST(Processor::scheduler().initialize());
 
-	Process::create_kernel(init2, nullptr);
+	auto* init_thread = MUST(Thread::create_kernel(init2, nullptr));
+	MUST(Processor::scheduler().add_thread(init_thread));
 	Processor::yield();
 
 	ASSERT_NOT_REACHED();
@@ -258,6 +259,9 @@ static void init2(void*)
 	VirtualFileSystem::initialize(cmdline.root);
 	dprintln("VFS initialized");
 
+	// FIXME: release memory used by modules. If modules are used
+	//        they are already loaded in here
+
 	TTY::initialize_devices();
 
 	auto console_path = MUST(BAN::String::formatted("/dev/{}", cmdline.console));

ports/make/build.sh

@@ -8,4 +8,5 @@ CONFIGURE_OPTIONS=(
 	'--disable-nls'
 	'--disable-posix-spawn'
 	'--enable-year2038'
+	'CFLAGS=-std=c17'
 )

toolchain/Toolchain.txt

@@ -22,6 +22,7 @@ set(CMAKE_SYSTEM_NAME banan-os)
 set(CMAKE_SYSTEM_PROCESSOR ${BANAN_ARCH})
 
 set(CMAKE_SYSROOT ${BANAN_SYSROOT})
+set(CMAKE_STAGING_PREFIX ${BANAN_SYSROOT}/usr)
 
 set(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)
 set(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)

userspace/libraries/LibC/stdio.cpp

@@ -908,12 +908,16 @@ void setbuf(FILE* file, char* buffer)
 
 int setvbuf(FILE* file, char* buffer, int type, size_t size)
 {
+	ScopeLock _(file);
+
 	if (file->fd == -1)
 	{
 		errno = EBADF;
 		return -1;
 	}
 
+	(void)fflush(file);
+
 	if (size == 0)
 		type = _IONBF;
 

userspace/programs/CMakeLists.txt

@@ -4,6 +4,7 @@ set(USERSPACE_PROGRAMS
 	cat
 	cat-mmap
 	chmod
+	chown
 	cp
 	dd
 	dhcp-client
@@ -11,6 +12,7 @@ set(USERSPACE_PROGRAMS
 	DynamicLoader
 	echo
 	env
+	false
 	getopt
 	http-server
 	id
@@ -36,6 +38,8 @@ set(USERSPACE_PROGRAMS
 	sudo
 	sync
 	tee
+	test
+	true
 	TaskBar
 	Terminal
 	touch

userspace/programs/chown/CMakeLists.txt

@@ -0,0 +1,9 @@
+set(SOURCES
+	main.cpp
+)
+
+add_executable(chown ${SOURCES})
+banan_link_library(chown ban)
+banan_link_library(chown libc)
+
+install(TARGETS chown OPTIONAL)

userspace/programs/chown/main.cpp

@@ -0,0 +1,109 @@
+#include <ctype.h>
+#include <errno.h>
+#include <grp.h>
+#include <pwd.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+void usage(const char* argv0, int ret)
+{
+	FILE* out = (ret == 0) ? stdout : stderr;
+	fprintf(out, "usage: %s [OWNER][:[GROUP]] FILE...\n", argv0);
+	fprintf(out, "  Change the owner and/or group of each FILE.\n");
+	exit(ret);
+}
+
+[[noreturn]] void print_error_and_exit(const char* format, ...)
+{
+	va_list args;
+	va_start(args, format);
+	vfprintf(stderr, format, args);
+	va_end(args);
+	exit(1);
+	__builtin_unreachable();
+}
+
+const passwd* get_user(const char* string)
+{
+	bool is_numeric = true;
+	for (size_t i = 0; string[i] && is_numeric; i++)
+		if (!isdigit(string[i]))
+			is_numeric = false;
+	if (is_numeric)
+		return getpwuid(atoll(string));
+	return getpwnam(string);
+}
+
+const group* get_group(const char* string)
+{
+	bool is_numeric = true;
+	for (size_t i = 0; string[i] && is_numeric; i++)
+		if (!isdigit(string[i]))
+			is_numeric = false;
+	if (is_numeric)
+		return getgrgid(atoll(string));
+	return getgrnam(string);
+}
+
+int main(int argc, char** argv)
+{
+	if (argc <= 2)
+		usage(argv[0], 1);
+
+	uid_t uid = -1;
+	gid_t gid = -1;
+
+	const char* owner_string = argv[1];
+
+	const char* colon = strchr(owner_string, ':');
+	if (colon == owner_string)
+	{
+		const auto* group = get_group(owner_string + 1);
+		if (group == nullptr)
+			print_error_and_exit("could not find group %s\n", owner_string + 1);
+		gid = group->gr_gid;
+	}
+	else if (colon == nullptr)
+	{
+		const auto* user = get_user(owner_string);
+		if (user == nullptr)
+			print_error_and_exit("could not find user %s\n", owner_string);
+		uid = user->pw_uid;
+	}
+	else
+	{
+		char* user_name = strndup(owner_string, colon - owner_string);
+		if (user_name == nullptr)
+			print_error_and_exit("strndup: %s\n", strerror(errno));
+		const auto* user = get_user(user_name);
+		if (user == nullptr)
+			print_error_and_exit("could not find user %s\n", user_name);
+		free(user_name);
+		uid = user->pw_uid;
+		if (colon[1] == '\0')
+			gid = user->pw_gid;
+		else
+		{
+			const auto* group = get_group(colon + 1);
+			if (group == nullptr)
+				print_error_and_exit("could not find group %s\n", colon + 1);
+			gid = group->gr_gid;
+		}
+	}
+
+	int ret = 0;
+	for (int i = 2; i < argc; i++)
+	{
+		if (chown(argv[i], uid, gid) == -1)
+		{
+			perror("chown");
+			ret = 1;
+		}
+	}
+
+	return ret;
+}

userspace/programs/false/CMakeLists.txt

@@ -0,0 +1,9 @@
+set(SOURCES
+	main.cpp
+)
+
+add_executable(false ${SOURCES})
+banan_link_library(false ban)
+banan_link_library(false libc)
+
+install(TARGETS false OPTIONAL)

userspace/programs/false/main.cpp

@@ -0,0 +1,6 @@
+#include <stdlib.h>
+
+int main()
+{
+	return EXIT_FAILURE;
+}

userspace/programs/true/CMakeLists.txt

@@ -0,0 +1,9 @@
+set(SOURCES
+	main.cpp
+)
+
+add_executable(true ${SOURCES})
+banan_link_library(true ban)
+banan_link_library(true libc)
+
+install(TARGETS true OPTIONAL)

userspace/programs/true/main.cpp

@@ -0,0 +1,6 @@
+#include <stdlib.h>
+
+int main()
+{
+	return EXIT_SUCCESS;
+}