driver-install: implemented a simple installer

Banos is a stable WIP C driver API that is supposed to provide a simple
interface to interact with the kernel and load the modules dynamically.
It is WIP and atm this just implements module loading with a custom
banos_install syscall. Banos will not try to substitute parts of the
kernel instead it will just expose kernel functionality via a stable
BINARY API. Meaning binaries (should) remain forward and backward
compatible on a binary level.

Banos modules work similarly to those in linux, you expose symbols via
BANOS_EXPORT which allows you to export a name + addr paired symbol.
It puts it in the .banos-export section. Drivers provide metadata about
themselves in the REQUIRED .banos-driver section. Symbols are resolved
at runtime. The kernel exposes the driver functionality via the same
.banos-export export mechanism.

Banos modules are elf RELOCATABLE files (object files) which have
partial linking (only banos symbols should remain). Modules will
eventually define dependencies, will export symbols and will allow you
to build a complex object hierarchy.

This patch adds the banos_install syscall which takes in the driver
image to install and may only be executed by super users. The API
doesn't validate already loaded modules, as thats something the
userspace MAY choose to keep track of. Multi-instance functionality
shall be implemented via driver specific behaviuor (exposed in the dev
filesystem or some other means).

Modules are supposed to allow you to alter kernel behavior and extend
it, allowing you to create filesystems, drivers, networking
modifications, schedulers, probers, and more (hopefully) whilst
remaining binary compatible with any version of the kernel (again,
hopefully).

BAN/include/BAN/HashMap.h

@@ -98,8 +98,7 @@ namespace BAN
 
 		struct EntryHash
 		{
-			template<detail::HashMapFindable<Key, HASH, COMP> U>
+			constexpr bool operator()(const Key& a)
-			constexpr bool operator()(const U& a)
 			{
 				return HASH()(a);
 			}
@@ -111,8 +110,7 @@ namespace BAN
 
 		struct EntryComp
 		{
-			template<detail::HashMapFindable<Key, HASH, COMP> U>
+			constexpr bool operator()(const Entry& a, const Key& b)
-			constexpr bool operator()(const Entry& a, const U& b)
 			{
 				return COMP()(a.key, b);
 			}

BAN/include/BAN/Sort.h

@@ -67,20 +67,11 @@ namespace BAN::sort
 	template<typename It, typename Comp = less<it_value_type_t<It>>>
 	void quick_sort(It begin, It end, Comp comp = {})
 	{
-		while (distance(begin, end) > 1)
+		if (distance(begin, end) <= 1)
-		{
+			return;
-			const auto [lt, gt] = detail::partition(begin, end, comp);
+		const auto [lt, gt] = detail::partition(begin, end, comp);
-			if (distance(begin, lt) < distance(gt, end))
+		quick_sort(begin, lt, comp);
-			{
+		quick_sort(gt, end, comp);
-				quick_sort(begin, lt);
-				begin = gt;
-			}
-			else
-			{
-				quick_sort(gt, end);
-				end = lt;
-			}
-		}
 	}
 
 	template<typename It, typename Comp = less<it_value_type_t<It>>>
@@ -111,28 +102,13 @@ namespace BAN::sort
 		template<typename It, typename Comp>
 		void intro_sort_impl(It begin, It end, size_t max_depth, Comp comp)
 		{
-			for (size_t i = 0; i < max_depth; i++)
+			if (distance(begin, end) <= 16)
-			{
+				return insertion_sort(begin, end, comp);
-				const size_t len = distance(begin, end);
+			if (max_depth == 0)
-				if (len <= 1)
+				return heap_sort(begin, end, comp);
-					return;
+			const auto [lt, gt] = detail::partition(begin, end, comp);
-				if (len <= 16)
+			intro_sort_impl(begin, lt, max_depth - 1, comp);
-					return insertion_sort(begin, end, comp);
+			intro_sort_impl(gt, end, max_depth - 1, comp);
-
-				const auto [lt, gt] = detail::partition(begin, end, comp);
-				if (distance(begin, lt) < distance(gt, end))
-				{
-					intro_sort_impl(begin, lt, max_depth - 1, comp);
-					begin = gt;
-				}
-				else
-				{
-					intro_sort_impl(gt, end, max_depth - 1, comp);
-					end = lt;
-				}
-			}
-
-			heap_sort(begin, end, comp);
 		}
 
 	}

BAN/include/BAN/Span.h

@@ -100,22 +100,3 @@ namespace BAN
 	};
 
 }
-
-#include <BAN/Formatter.h>
-
-namespace BAN::Formatter
-{
-
-	template<typename F, typename T>
-	void print_argument(F putc, const Span<T>& span, const ValueFormat& format)
-	{
-		putc('[');
-		for (typename Span<T>::size_type i = 0; i < span.size(); i++)
-		{
-			if (i != 0) putc(',');
-			print_argument(putc, span[i], format);
-		}
-		putc(']');
-	}
-
-}

README.md

@@ -118,7 +118,7 @@ To change the bootloader you can set environment variable BANAN\_BOOTLOADER; sup
 
 To run with UEFI set environment variable BANAN\_UEFI\_BOOT=1. You will also have to set OVMF\_PATH to the correct OVMF (default */usr/share/ovmf/x64/OVMF.fd*).
 
-To build an image with no physical root filesystem, but an initrd set environment variable BANAN\_INITRD=1. This can be used when testing on hardware with unsupported USB controller. If BANAN\_INITRD is set to a value larger than 1, initrd will be gzip compressed.
+To build an image with no physical root filesystem, but an initrd set environment variable BANAN\_INITRD=1. This can be used when testing on hardware with unsupported USB controller.
 
 If you have corrupted your disk image or want to create new one, you can either manually delete *build/banan-os.img* and build system will automatically create you a new one or you can run the following command.
 ```sh

kernel/arch/i686/Signal.S

@@ -59,7 +59,7 @@ signal_trampoline:
 	addl $24, %esp
 
 	// restore sigmask
-	movl $78, %eax      // SYS_SIGPROCMASK
+	movl $83, %eax      // SYS_SIGPROCMASK
 	movl $3,  %ebx      // SIG_SETMASK
 	leal 72(%esp), %ecx // set
 	xorl %edx, %edx     // oset

kernel/arch/i686/Yield.S

@@ -12,7 +12,7 @@ asm_yield_trampoline:
 	pushl %ebp
 
 	pushl %esp
-	call scheduler_on_yield_trampoline
+	call scheduler_on_yield
 	addl $4, %esp
 
 	popl %ebp

kernel/arch/i686/interrupts.S

@@ -27,10 +27,14 @@
 
 isr_stub:
 	intr_header 12
+	movl %cr0, %eax; pushl %eax
+	movl %cr2, %eax; pushl %eax
+	movl %cr3, %eax; pushl %eax
+	movl %cr4, %eax; pushl %eax
 
-	movl 32(%esp), %edi // isr number
+	movl 48(%esp), %edi // isr number
-	movl 36(%esp), %esi // error code
+	movl 52(%esp), %esi // error code
-	leal 40(%esp), %edx // interrupt stack ptr
+	leal 56(%esp), %edx // interrupt stack ptr
 	movl %esp,     %ecx // register ptr
 
 	# stack frame for stack trace
@@ -48,7 +52,7 @@ isr_stub:
 	call cpp_isr_handler
 
 	movl %ebp, %esp
-	addl $8, %esp
+	addl $24, %esp
 
 	intr_footer 12
 	addl $8, %esp

kernel/arch/x86_64/Signal.S

@@ -61,7 +61,7 @@ signal_trampoline:
 	addq $40, %rsp
 
 	// restore sigmask
-	movq $78, %rdi       // SYS_SIGPROCMASK
+	movq $83, %rdi       // SYS_SIGPROCMASK
 	movq $3,  %rsi       // SIG_SETMASK
 	leaq 192(%rsp), %rdx // set
 	xorq %r10, %r10      // oset

kernel/arch/x86_64/Yield.S

@@ -15,7 +15,7 @@ asm_yield_trampoline:
 	pushq %r15
 
 	movq %rsp, %rdi
-	call scheduler_on_yield_trampoline
+	call scheduler_on_yield
 
 	popq %r15
 	popq %r14

kernel/arch/x86_64/interrupts.S

@@ -44,12 +44,17 @@
 
 isr_stub:
 	intr_header 24
+	movq %cr0, %rax; pushq %rax
+	movq %cr2, %rax; pushq %rax
+	movq %cr3, %rax; pushq %rax
+	movq %cr4, %rax; pushq %rax
 
-	movq 120(%rsp), %rdi // isr number
+	movq 152(%rsp), %rdi	// isr number
-	movq 128(%rsp), %rsi // error code
+	movq 160(%rsp), %rsi	// error code
-	leaq 136(%rsp), %rdx // interrupt stack ptr
+	leaq 168(%rsp), %rdx	// interrupt stack ptr
-	movq %rsp, %rcx      // register ptr
+	movq %rsp, %rcx			// register ptr
 	call cpp_isr_handler
+	addq $32, %rsp
 
 	intr_footer 24
 	addq $16, %rsp

kernel/include/kernel/API/SharedPage.h

@@ -8,7 +8,6 @@ namespace Kernel::API
 	enum SharedPageFeature : uint32_t
 	{
 		SPF_GETTIME = 1 << 0,
-		SPF_RDTSCP  = 1 << 1,
 	};
 
 	struct SharedPage
@@ -19,8 +18,9 @@ namespace Kernel::API
 
 		struct
 		{
-			uint64_t realtime_s;
+			uint8_t shift;
-			uint32_t realtime_ns;
+			uint64_t mult;
+			uint64_t realtime_seconds;
 		} gettime_shared;
 
 		struct
@@ -28,8 +28,6 @@ namespace Kernel::API
 			struct
 			{
 				uint32_t seq;
-				uint32_t mult;
-				int8_t shift;
 				uint64_t last_ns;
 				uint64_t last_tsc;
 			} gettime_local;

kernel/include/kernel/APIC.h

@@ -26,7 +26,6 @@ namespace Kernel
 		BAN::ErrorOr<uint8_t> reserve_gsi(uint32_t gsi);
 
 		void initialize_timer();
-		void set_timer_dealine(uint64_t ns);
 
 	private:
 		uint32_t read_from_local_apic(ptrdiff_t);

kernel/include/kernel/Audio/Controller.h

@@ -38,7 +38,7 @@ namespace Kernel
 
 		BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override;
 
-		BAN::ErrorOr<long> ioctl_impl(unsigned long request, void* arg) override;
+		BAN::ErrorOr<long> ioctl_impl(int cmd, void* arg) override;
 
 	protected:
 		ThreadBlocker m_sample_data_blocker;

kernel/include/kernel/CPUID.h

@@ -82,8 +82,5 @@ namespace CPUID
 	bool has_pat();
 	bool has_1gib_pages();
 	bool has_invariant_tsc();
-	bool has_rdtscp();
-	uint64_t get_tsc_frequency();
-	bool has_kvm_pvclock();
 
 }

kernel/include/kernel/Device/FramebufferDevice.h

@@ -36,7 +36,7 @@ namespace Kernel
 		virtual BAN::ErrorOr<size_t> read_impl(off_t, BAN::ByteSpan) override;
 		virtual BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override;
 
-		BAN::ErrorOr<long> ioctl_impl(unsigned long request, void* arg) override;
+		BAN::ErrorOr<long> ioctl_impl(int cmd, void* arg) override;
 
 		virtual bool can_read_impl() const override { return true; }
 		virtual bool can_write_impl() const override { return true; }

kernel/include/kernel/ELF.h

@@ -14,8 +14,8 @@ namespace Kernel::ELF
 			size_t size;
 		};
 
+		bool open_execfd;
 		vaddr_t entry_point;
-		BAN::Optional<vaddr_t> interp_base;
 		BAN::Optional<TLS> master_tls;
 		BAN::Vector<BAN::UniqPtr<MemoryRegion>> regions;
 	};

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

@@ -99,8 +99,7 @@ namespace Kernel
 		BAN::ErrorOr<uint32_t> reserve_free_block(uint32_t primary_bgd);
 		BAN::ErrorOr<void> release_block(uint32_t block);
 
-		BAN::ErrorOr<BAN::RefPtr<Ext2Inode>> open_inode(ino_t);
+		BAN::HashMap<ino_t, BAN::RefPtr<Ext2Inode>>& inode_cache() { return m_inode_cache; }
-		void remove_from_cache(ino_t);
 
 		const Ext2::Superblock& superblock() const { return m_superblock; }
 
@@ -156,7 +155,6 @@ namespace Kernel
 		BAN::RefPtr<Inode> m_root_inode;
 		BAN::Vector<uint32_t> m_superblock_backups;
 
-		Mutex m_inode_cache_lock;
 		BAN::HashMap<ino_t, BAN::RefPtr<Ext2Inode>> m_inode_cache;
 
 		BlockBufferManager m_buffer_manager;

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

@@ -55,7 +55,7 @@ namespace Kernel
 		BAN::ErrorOr<BAN::RefPtr<Inode>> find_inode_no_lock(BAN::StringView);
 
 		/* needs write end of the lock when allocate is true*/
-		BAN::ErrorOr<BAN::Optional<uint32_t>> block_from_indirect_block_no_lock(uint32_t block, uint32_t index, uint32_t depth, bool allocate);
+		BAN::ErrorOr<BAN::Optional<uint32_t>> block_from_indirect_block_no_lock(uint32_t& block, uint32_t index, uint32_t depth, bool allocate);
 		BAN::ErrorOr<BAN::Optional<uint32_t>> fs_block_of_data_block_index_no_lock(uint32_t data_block_index, bool allocate);
 
 		/* needs write end of the lock */
@@ -71,13 +71,7 @@ namespace Kernel
 	private:
 		Ext2Inode(Ext2FS& fs, Ext2::Inode inode, uint32_t ino);
 
-		BAN::Optional<uint32_t> block_cache_find(uint32_t block, uint32_t index) const;
+		static BAN::ErrorOr<BAN::RefPtr<Ext2Inode>> create(Ext2FS&, uint32_t);
-		void block_cache_remove(uint32_t block, uint32_t index);
-		void block_cache_add(uint32_t block, uint32_t index, uint32_t target);
-
-		BAN::RefPtr<Inode> dir_cache_find(BAN::StringView) const;
-		void dir_cache_remove(BAN::StringView);
-		void dir_cache_add(BAN::StringView, BAN::RefPtr<Inode>);
 
 	private:
 		struct ScopedSync
@@ -113,27 +107,6 @@ namespace Kernel
 		const uint32_t m_og_faddr;
 		const Ext2::Osd2 m_og_osd2;
 
-		struct BlockCacheEntry
-		{
-			mutable uint32_t freq;
-			uint32_t block;
-			uint32_t index;
-			uint32_t target;
-		};
-		mutable SpinLock m_block_cache_lock;
-		BAN::Array<BlockCacheEntry, 8> m_block_cache;
-
-		struct DirCacheEntry
-		{
-			mutable size_t freq { 0 };
-			BAN::RefPtr<Inode> inode;
-			size_t name_len { 0 };
-			char   name[256];
-		};
-		static constexpr size_t dir_cache_size = 32;
-		mutable RWLock m_dir_cache_lock;
-		BAN::Vector<DirCacheEntry> m_dir_cache;
-
 		friend class Ext2FS;
 		friend class BAN::RefPtr<Ext2Inode>;
 	};

kernel/include/kernel/FS/Inode.h

@@ -150,7 +150,7 @@ namespace Kernel
 		bool has_error() const  { return has_error_impl(); }
 		bool has_hungup() const { return has_hungup_impl(); }
 
-		BAN::ErrorOr<long> ioctl(unsigned long request, void* arg);
+		BAN::ErrorOr<long> ioctl(int request, void* arg);
 
 		BAN::ErrorOr<void> add_epoll(class Epoll*);
 		void del_epoll(class Epoll*);
@@ -199,7 +199,7 @@ namespace Kernel
 		virtual bool has_error_impl() const = 0;
 		virtual bool has_hungup_impl() const = 0;
 
-		virtual BAN::ErrorOr<long> ioctl_impl(unsigned long, void*) { return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<long> ioctl_impl(int, void*) { return BAN::Error::from_errno(ENOTSUP); }
 
 	protected:
 		// TODO: this is supposed to be const I guess?

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

@@ -13,8 +13,6 @@ namespace Kernel
 		static void initialize();
 		static ProcFileSystem& get();
 
-		void post_scheduler_initialize();
-
 		BAN::ErrorOr<void> on_process_create(Process&);
 		void on_process_delete(Process&);
 

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

@@ -76,7 +76,7 @@ namespace Kernel
 	class ProcROInode final : public TmpInode
 	{
 	public:
-		static BAN::ErrorOr<BAN::RefPtr<ProcROInode>> create_new(BAN::ErrorOr<size_t> (*callback)(off_t, BAN::ByteSpan, void*), TmpFileSystem&, void*, mode_t, uid_t, gid_t);
+		static BAN::ErrorOr<BAN::RefPtr<ProcROInode>> create_new(size_t (*callback)(off_t, BAN::ByteSpan), TmpFileSystem&, mode_t, uid_t, gid_t);
 		~ProcROInode() = default;
 
 	protected:
@@ -92,11 +92,10 @@ namespace Kernel
 		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
-		ProcROInode(BAN::ErrorOr<size_t> (*callback)(off_t, BAN::ByteSpan, void*), TmpFileSystem&, void*, const TmpInodeInfo&);
+		ProcROInode(size_t (*callback)(off_t, BAN::ByteSpan), TmpFileSystem&, const TmpInodeInfo&);
 
 	private:
-		BAN::ErrorOr<size_t> (*m_callback)(off_t, BAN::ByteSpan, void*);
+		size_t (*m_callback)(off_t, BAN::ByteSpan);
-		void* m_argument;
 	};
 
 	class ProcSymlinkInode final : public TmpInode

kernel/include/kernel/FS/VirtualFileSystem.h

@@ -92,9 +92,8 @@ namespace Kernel
 		MountPoint* mount_from_root_inode(BAN::RefPtr<Inode>);
 
 	private:
+		Mutex					m_mutex;
 		BAN::RefPtr<FileSystem>	m_root_fs;
-
-		Mutex					m_mount_point_lock;
 		BAN::Vector<MountPoint>	m_mount_points;
 
 		friend class BAN::RefPtr<VirtualFileSystem>;

kernel/include/kernel/Lock/RWLock.h

@@ -1,7 +1,7 @@
 #pragma once
 
-#include <kernel/Lock/SpinLock.h>
+#include <kernel/Lock/Mutex.h>
-#include <kernel/Lock/SpinLockAsMutex.h>
+#include <kernel/Lock/LockGuard.h>
 
 namespace Kernel
 {
@@ -15,18 +15,15 @@ namespace Kernel
 
 		void rd_lock()
 		{
-			SpinLockGuard _(m_lock);
+			LockGuard _(m_mutex);
 			while (m_writers_waiting > 0 || m_writer != -1)
-			{
+				m_thread_blocker.block_indefinite(&m_mutex);
-				SpinLockGuardAsMutex smutex(_);
-				m_thread_blocker.block_indefinite(&smutex);
-			}
 			m_readers_active++;
 		}
 
 		void rd_unlock()
 		{
-			SpinLockGuard _(m_lock);
+			LockGuard _(m_mutex);
 			if (--m_readers_active == 0)
 				m_thread_blocker.unblock();
 		}
@@ -39,14 +36,11 @@ namespace Kernel
 				return;
 			}
 
-			SpinLockGuard _(m_lock);
+			LockGuard _(m_mutex);
 
 			m_writers_waiting++;
 			while (m_readers_active > 0 || m_writer != -1)
-			{
+				m_thread_blocker.block_indefinite(&m_mutex);
-				SpinLockGuardAsMutex smutex(_);
-				m_thread_blocker.block_indefinite(&smutex);
-			}
 			m_writers_waiting--;
 
 			m_writer = Thread::current_tid();
@@ -57,13 +51,13 @@ namespace Kernel
 		{
 			if (--m_writer_depth != 0)
 				return;
-			SpinLockGuard _(m_lock);
+			LockGuard _(m_mutex);
 			m_writer = -1;
 			m_thread_blocker.unblock();
 		}
 
 	private:
-		SpinLock m_lock;
+		Mutex m_mutex;
 		ThreadBlocker m_thread_blocker;
 		uint32_t m_readers_active  { 0 };
 		uint32_t m_writers_waiting { 0 };

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

@@ -29,7 +29,6 @@ namespace Kernel
 		REG_TDLEN	= 0x3808,
 		REG_TDH		= 0x3810,
 		REG_TDT		= 0x3818,
-		REG_RXCSUM	= 0x5000,
 		REG_MTA		= 0x5200,
 	};
 
@@ -168,15 +167,6 @@ namespace Kernel
 		RCTL_BSIZE_16384	= (0b01 << 16) | RCTL_BSEX,
 	};
 
-	enum E1000_RXCSUM : uint32_t
-	{
-		RXSUM_IPOFLD  = 1 << 8,
-		RXSUM_TUOFLD  = 1 << 9,
-		RXSUM_CRCOFLD = 1 << 11,
-		RXSUM_IPPCE   = 1 << 12,
-		RXSUM_PCSD    = 1 << 13,
-	};
-
 	enum E1000_CMD : uint8_t
 	{
 		CMD_EOP		= 1 << 0,
@@ -188,26 +178,6 @@ namespace Kernel
 		CMD_IDE		= 1 << 7,
 	};
 
-	enum E1000_RX_STS : uint8_t
-	{
-		RX_STS_IPCS  = 1 << 6,
-		RX_STS_TCPCS = 1 << 5,
-		RX_STS_UDPCS = 1 << 5,
-		RX_STS_EOP   = 1 << 1,
-		RX_STS_DD    = 1 << 0,
-	};
-
-	enum E1000_RX_ERR : uint8_t
-	{
-		RX_ERR_RXE  = 1 << 7,
-		RX_ERR_IPE  = 1 << 6,
-		RX_ERR_TCPE = 1 << 5,
-		RX_ERR_CXE  = 1 << 4,
-		RX_ERR_SEQ  = 1 << 2,
-		RX_ERR_SE   = 1 << 1,
-		RX_ERR_CE   = 1 << 0,
-	};
-
 	struct e1000_rx_desc
 	{
 		uint64_t addr;

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

@@ -45,7 +45,7 @@ namespace Kernel
 		uint32_t read32(uint16_t reg);
 		void write32(uint16_t reg, uint32_t value);
 
-		BAN::ErrorOr<void> send_raw_bytes(BAN::Span<const BAN::ConstByteSpan> buffers) override;
+		BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::Span<const BAN::ConstByteSpan> payload) override;
 
 		bool can_read_impl() const override { return false; }
 		bool can_write_impl() const override { return false; }
@@ -74,8 +74,8 @@ namespace Kernel
 		BAN::UniqPtr<DMARegion>	m_rx_descriptor_region;
 		BAN::UniqPtr<DMARegion>	m_tx_descriptor_region;
 
-		BAN::Atomic<uint32_t> m_tx_head   { 0 };
+		BAN::Atomic<uint32_t> m_tx_head1 { 0 };
-		BAN::Atomic<uint32_t> m_tx_commit { 0 };
+		BAN::Atomic<uint32_t> m_tx_head2 { 0 };
 
 		SpinLock m_rx_lock;
 		ThreadBlocker m_rx_blocker;

kernel/include/kernel/Networking/IPv4Layer.h

@@ -41,14 +41,14 @@ namespace Kernel
 
 		ARPTable& arp_table() { return *m_arp_table; }
 
-		BAN::ErrorOr<void> handle_ipv4_packet(NetworkInterface&, BAN::ConstByteSpan, uint32_t validated_cksums);
+		BAN::ErrorOr<void> handle_ipv4_packet(NetworkInterface&, BAN::ConstByteSpan);
 
 		virtual void unbind_socket(uint16_t port) override;
 		virtual BAN::ErrorOr<void> bind_socket_with_target(BAN::RefPtr<NetworkSocket>, const sockaddr* target_address, socklen_t target_address_len) override;
 		virtual BAN::ErrorOr<void> bind_socket_to_address(BAN::RefPtr<NetworkSocket>, const sockaddr* address, socklen_t address_len) override;
 		virtual BAN::ErrorOr<void> get_socket_address(BAN::RefPtr<NetworkSocket>, sockaddr* address, socklen_t* address_len) override;
 
-		virtual BAN::ErrorOr<void> sendto(NetworkSocket&, BAN::ConstByteSpan, const sockaddr*, socklen_t) override;
+		virtual BAN::ErrorOr<size_t> sendto(NetworkSocket&, BAN::ConstByteSpan, const sockaddr*, socklen_t) override;
 
 		virtual Socket::Domain domain() const override { return Socket::Domain::INET ;}
 		virtual size_t header_size() const override { return sizeof(IPv4Header); }

kernel/include/kernel/Networking/Loopback.h

@@ -27,7 +27,7 @@ namespace Kernel
 		{}
 		~LoopbackInterface();
 
-		BAN::ErrorOr<void> send_raw_bytes(BAN::Span<const BAN::ConstByteSpan> buffers) override;
+		BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::Span<const BAN::ConstByteSpan> payload) override;
 
 		bool can_read_impl() const override { return false; }
 		bool can_write_impl() const override { return false; }
@@ -46,7 +46,7 @@ namespace Kernel
 		};
 
 	private:
-		SpinLock m_buffer_lock;
+		Mutex m_buffer_lock;
 		BAN::UniqPtr<VirtualRange> m_buffer;
 
 		uint32_t m_buffer_tail { 0 };

kernel/include/kernel/Networking/NetworkInterface.h

@@ -23,13 +23,6 @@ namespace Kernel
 		ARP = 0x0806,
 	};
 
-	enum NetworkChecksum : uint32_t
-	{
-		CKSUM_IPV4 = 1 << 0,
-		CKSUM_TCP  = 1 << 1,
-		CKSUM_UDP  = 1 << 2,
-	};
-
 	class NetworkInterface : public CharacterDevice
 	{
 		BAN_NON_COPYABLE(NetworkInterface);
@@ -66,33 +59,11 @@ namespace Kernel
 
 		virtual BAN::StringView name() const override { return m_name; }
 
-		BAN::ErrorOr<void> send_with_ethernet_header(BAN::MACAddress dst_mac, EtherType ether_type, BAN::ConstByteSpan buffer)
+		BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::ConstByteSpan payload)
 		{
-			BAN::ConstByteSpan buffer_array[1] { buffer };
+			return send_bytes(destination, protocol, { &payload, 1 });
-			return send_with_ethernet_header(dst_mac, ether_type, buffer_array);
 		}
-
+		virtual BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::Span<const BAN::ConstByteSpan> payload) = 0;
-		template<size_t SIZE>
-		BAN::ErrorOr<void> send_with_ethernet_header(BAN::MACAddress dst_mac, EtherType ether_type, const BAN::ConstByteSpan (&buffer_array)[SIZE])
-		{
-			const auto ethernet_header = EthernetHeader {
-				.dst_mac = dst_mac,
-				.src_mac = get_mac_address(),
-				.ether_type = ether_type,
-			};
-
-			BAN::ConstByteSpan new_buffer_array[SIZE + 1];
-			new_buffer_array[0] = BAN::ConstByteSpan::from(ethernet_header);
-			for (size_t i = 0; i < SIZE; i++)
-				new_buffer_array[i + 1] = buffer_array[i];
-
-			return send_raw_bytes({ new_buffer_array, SIZE + 1 });
-		}
-
-		virtual BAN::ErrorOr<void> send_raw_bytes(BAN::Span<const BAN::ConstByteSpan> buffers) = 0;
-
-	private:
-		BAN::ErrorOr<long> ioctl_impl(unsigned long, void*) override;
 
 	private:
 		const Type m_type;

kernel/include/kernel/Networking/NetworkLayer.h

@@ -27,7 +27,7 @@ namespace Kernel
 		virtual BAN::ErrorOr<void> bind_socket_to_address(BAN::RefPtr<NetworkSocket>, const sockaddr* address, socklen_t address_len) = 0;
 		virtual BAN::ErrorOr<void> get_socket_address(BAN::RefPtr<NetworkSocket>, sockaddr* address, socklen_t* address_len) = 0;
 
-		virtual BAN::ErrorOr<void> sendto(NetworkSocket&, BAN::ConstByteSpan, const sockaddr*, socklen_t) = 0;
+		virtual BAN::ErrorOr<size_t> sendto(NetworkSocket&, BAN::ConstByteSpan, const sockaddr*, socklen_t) = 0;
 
 		virtual Socket::Domain domain() const = 0;
 		virtual size_t header_size() const = 0;

kernel/include/kernel/Networking/NetworkManager.h

@@ -27,7 +27,7 @@ namespace Kernel
 		BAN::ErrorOr<BAN::RefPtr<Socket>> create_socket(Socket::Domain, Socket::Type, mode_t, uid_t, gid_t);
 		BAN::ErrorOr<void> connect_sockets(Socket::Domain, BAN::RefPtr<Socket>, BAN::RefPtr<Socket>);
 
-		void on_receive(NetworkInterface&, BAN::ConstByteSpan, uint32_t validated_cksums);
+		void on_receive(NetworkInterface&, BAN::ConstByteSpan);
 
 	private:
 		NetworkManager() {}

kernel/include/kernel/Networking/NetworkSocket.h

@@ -35,7 +35,7 @@ namespace Kernel
 		virtual void get_protocol_header(BAN::ByteSpan header, BAN::ConstByteSpan payload, uint16_t dst_port, PseudoHeader) = 0;
 		virtual NetworkProtocol protocol() const = 0;
 
-		virtual void receive_packet(BAN::ConstByteSpan, const sockaddr* sender, socklen_t sender_len, uint32_t validated_cksums) = 0;
+		virtual void receive_packet(BAN::ConstByteSpan, const sockaddr* sender, socklen_t sender_len) = 0;
 
 		bool is_bound() const { return m_address_len >= static_cast<socklen_t>(sizeof(sa_family_t)) && m_address.ss_family != AF_UNSPEC; }
 		in_port_t bound_port() const
@@ -54,7 +54,7 @@ namespace Kernel
 	protected:
 		NetworkSocket(NetworkLayer&, const Socket::Info&);
 
-		virtual BAN::ErrorOr<long> ioctl_impl(unsigned long request, void* arg) override;
+		virtual BAN::ErrorOr<long> ioctl_impl(int request, void* arg) override;
 		virtual BAN::ErrorOr<void> getsockname_impl(sockaddr*, socklen_t*) override;
 		virtual BAN::ErrorOr<void> getpeername_impl(sockaddr*, socklen_t*) override = 0;
 

kernel/include/kernel/Networking/RTL8169/Definitions.h

@@ -7,25 +7,25 @@ namespace Kernel
 
 	enum RTL8169_IO_REGS : uint16_t
 	{
-		RTL8169_IO_IDR0    = 0x00,
+		RTL8169_IO_IDR0   = 0x00,
-		RTL8169_IO_IDR1    = 0x01,
+		RTL8169_IO_IDR1   = 0x01,
-		RTL8169_IO_IDR2    = 0x02,
+		RTL8169_IO_IDR2   = 0x02,
-		RTL8169_IO_IDR3    = 0x03,
+		RTL8169_IO_IDR3   = 0x03,
-		RTL8169_IO_IDR4    = 0x04,
+		RTL8169_IO_IDR4   = 0x04,
-		RTL8169_IO_IDR5    = 0x05,
+		RTL8169_IO_IDR5   = 0x05,
 
-		RTL8169_IO_TNPDS   = 0x20,
+		RTL8169_IO_TNPDS  = 0x20,
-		RTL8169_IO_CR      = 0x37,
+		RTL8169_IO_CR     = 0x37,
-		RTL8169_IO_TPPoll  = 0x38,
+		RTL8169_IO_TPPoll = 0x38,
-		RTL8169_IO_IMR     = 0x3C,
+		RTL8169_IO_IMR    = 0x3C,
-		RTL8169_IO_ISR     = 0x3E,
+		RTL8169_IO_ISR    = 0x3E,
-		RTL8169_IO_TCR     = 0x40,
+		RTL8169_IO_TCR    = 0x40,
-		RTL8169_IO_RCR     = 0x44,
+		RTL8169_IO_RCR    = 0x44,
-		RTL8169_IO_PHYSts  = 0x6C,
+		RTL8169_IO_9346CR = 0x50,
-		RTL8169_IO_RMS     = 0xDA,
+		RTL8169_IO_PHYSts = 0x6C,
-		RTL8169_IO_CPlusCR = 0xE0,
+		RTL8169_IO_RMS    = 0xDA,
-		RTL8169_IO_RDSAR   = 0xE4,
+		RTL8169_IO_RDSAR  = 0xE4,
-		RTL8169_IO_MTPS    = 0xEC,
+		RTL8169_IO_MTPS   = 0xEC,
 	};
 
 	enum RTL8169_CR : uint8_t

kernel/include/kernel/Networking/RTL8169/RTL8169.h

@@ -20,7 +20,7 @@ namespace Kernel
 		virtual bool link_up() override { return m_link_up; }
 		virtual int link_speed() override;
 
-		virtual size_t payload_mtu() const override { return 0x1FF8 - sizeof(EthernetHeader); }
+		virtual size_t payload_mtu() const override { return 7436 - sizeof(EthernetHeader); }
 
 		virtual void handle_irq() override;
 
@@ -33,7 +33,7 @@ namespace Kernel
 
 		BAN::ErrorOr<void> initialize();
 
-		virtual BAN::ErrorOr<void> send_raw_bytes(BAN::Span<const BAN::ConstByteSpan> buffers) override;
+		virtual BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::Span<const BAN::ConstByteSpan>) override;
 
 		virtual bool can_read_impl() const override { return false; }
 		virtual bool can_write_impl() const override { return false; }
@@ -64,17 +64,14 @@ namespace Kernel
 		BAN::UniqPtr<DMARegion>	m_rx_descriptor_region;
 		BAN::UniqPtr<DMARegion>	m_tx_descriptor_region;
 
-		BAN::Atomic<bool> m_rx_thread_should_die { false };
+		SpinLock m_lock;
-		BAN::Atomic<bool> m_rx_thread_is_dead { true };
 
-		uint32_t m_rx_head { 0 };
+		bool m_thread_should_die { false };
-		SpinLock m_rx_lock;
+		BAN::Atomic<bool> m_thread_is_dead { true };
-		ThreadBlocker m_rx_blocker;
+		ThreadBlocker m_thread_blocker;
 
-		BAN::Atomic<uint32_t> m_tx_head { 0 };
+		uint32_t m_rx_current { 0 };
-		BAN::Atomic<uint32_t> m_tx_commit { 0 };
+		size_t m_tx_current { 0 };
-		SpinLock m_tx_lock;
-		ThreadBlocker m_tx_blocker;
 
 		BAN::MACAddress	m_mac_address {};
 		BAN::Atomic<bool> m_link_up { false };

kernel/include/kernel/Networking/TCPSocket.h

@@ -66,9 +66,9 @@ namespace Kernel
 		BAN::ErrorOr<void> getsockopt_impl(int, int, void*, socklen_t*) override;
 		BAN::ErrorOr<void> setsockopt_impl(int, int, const void*, socklen_t) override;
 
-		BAN::ErrorOr<long> ioctl_impl(unsigned long, void*) override;
+		BAN::ErrorOr<long> ioctl_impl(int, void*) override;
 
-		void receive_packet(BAN::ConstByteSpan, const sockaddr* sender, socklen_t sender_len, uint32_t validated_cksums) override;
+		void receive_packet(BAN::ConstByteSpan, const sockaddr* sender, socklen_t sender_len) override;
 
 		bool can_read_impl() const override;
 		bool can_write_impl() const override;

kernel/include/kernel/Networking/UDPSocket.h

@@ -31,7 +31,7 @@ namespace Kernel
 		void get_protocol_header(BAN::ByteSpan header, BAN::ConstByteSpan payload, uint16_t dst_port, PseudoHeader) override;
 
 	protected:
-		void receive_packet(BAN::ConstByteSpan, const sockaddr* sender, socklen_t sender_len, uint32_t validated_cksums) override;
+		void receive_packet(BAN::ConstByteSpan, const sockaddr* sender, socklen_t sender_len) override;
 
 		BAN::ErrorOr<void> connect_impl(const sockaddr*, socklen_t) override;
 		BAN::ErrorOr<void> bind_impl(const sockaddr* address, socklen_t address_len) override;
@@ -41,7 +41,7 @@ namespace Kernel
 		BAN::ErrorOr<void> getsockopt_impl(int, int, void*, socklen_t*) override;
 		BAN::ErrorOr<void> setsockopt_impl(int, int, const void*, socklen_t) override;
 
-		BAN::ErrorOr<long> ioctl_impl(unsigned long, void*) override;
+		BAN::ErrorOr<long> ioctl_impl(int, void*) override;
 
 		bool can_read_impl() const override { return !m_packets.empty(); }
 		bool can_write_impl() const override { return true; }

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

@@ -30,7 +30,6 @@ namespace Kernel
 		virtual BAN::ErrorOr<void> bind_impl(const sockaddr*, socklen_t) override;
 		virtual BAN::ErrorOr<size_t> recvmsg_impl(msghdr& message, int flags) override;
 		virtual BAN::ErrorOr<size_t> sendmsg_impl(const msghdr& message, int flags) override;
-		virtual BAN::ErrorOr<void> getsockname_impl(sockaddr*, socklen_t*) override;
 		virtual BAN::ErrorOr<void> getpeername_impl(sockaddr*, socklen_t*) override;
 		virtual BAN::ErrorOr<void> getsockopt_impl(int, int, void*, socklen_t*) override;
 		virtual BAN::ErrorOr<void> setsockopt_impl(int, int, const void*, socklen_t) override;

kernel/include/kernel/OpenFileDescriptorSet.h

@@ -58,8 +58,6 @@ namespace Kernel
 		BAN::ErrorOr<size_t> recvmsg(int socket, msghdr& message, int flags);
 		BAN::ErrorOr<size_t> sendmsg(int socket, const msghdr& message, int flags);
 
-		int get_max_open_fd() const;
-
 		BAN::ErrorOr<VirtualFileSystem::File> file_of(int) const;
 		BAN::ErrorOr<BAN::String> path_of(int) const;
 		BAN::ErrorOr<BAN::RefPtr<Inode>> inode_of(int);

kernel/include/kernel/PCI.h

@@ -4,6 +4,7 @@
 #include <BAN/Vector.h>
 #include <kernel/ACPI/AML/Node.h>
 #include <kernel/Memory/Types.h>
+#include <kernel/Storage/StorageController.h>
 
 #include <sys/types.h>
 
@@ -178,7 +179,7 @@ namespace Kernel::PCI
 	private:
 		struct PCIeInfo
 		{
-			paddr_t bus_paddr[256] {};
+			paddr_t bus_paddr[256];
 		};
 
 		PCIManager() = default;

kernel/include/kernel/Process.h

@@ -44,10 +44,10 @@ namespace Kernel
 		~Process();
 		void cleanup_function(Thread*);
 
-		BAN::ErrorOr<vaddr_t> setup_initial_process_stack(MemoryBackedRegion&, BAN::Span<BAN::String> argv, BAN::Span<BAN::String> envp, BAN::Span<LibELF::AuxiliaryVector> auxv);
+		void register_to_scheduler();
-
 		void exit(int status, int signal);
 
+		void add_thread(Thread*);
 		// returns true if thread was the last one
 		bool on_thread_exit(Thread&);
 
@@ -68,6 +68,7 @@ namespace Kernel
 		BAN::ErrorOr<long> sys_exec(const char* path, const char* const* argv, const char* const* envp);
 
 		BAN::ErrorOr<long> sys_wait(pid_t pid, int* stat_loc, int options);
+		BAN::ErrorOr<long> sys_sleep(int seconds);
 		BAN::ErrorOr<long> sys_nanosleep(const timespec* rqtp, timespec* rmtp);
 		BAN::ErrorOr<long> sys_setitimer(int which, const itimerval* value, itimerval* ovalue);
 
@@ -136,7 +137,7 @@ namespace Kernel
 		BAN::ErrorOr<long> sys_recvmsg(int socket, msghdr* message, int flags);
 		BAN::ErrorOr<long> sys_sendmsg(int socket, const msghdr* message, int flags);
 
-		BAN::ErrorOr<long> sys_ioctl(int fildes, unsigned long request, void* arg);
+		BAN::ErrorOr<long> sys_ioctl(int fildes, int request, void* arg);
 
 		BAN::ErrorOr<long> sys_pselect(sys_pselect_t* arguments);
 		BAN::ErrorOr<long> sys_ppoll(pollfd* fds, nfds_t nfds, const timespec* tmp_p, const sigset_t* sigmask);
@@ -209,12 +210,12 @@ namespace Kernel
 		BAN::ErrorOr<long> sys_set_gsbase(void*);
 		BAN::ErrorOr<long> sys_get_gsbase();
 
-		BAN::ErrorOr<long> sys_thread_create(void (*entry)(void*), void* arg, void* stack_base, size_t stack_size);
+		BAN::ErrorOr<long> sys_pthread_create(const pthread_attr_t* attr, void (*entry)(void*), void* arg);
-		BAN::ErrorOr<long> sys_thread_exit(void* value);
+		BAN::ErrorOr<long> sys_pthread_exit(void* value);
-		BAN::ErrorOr<long> sys_thread_join(pid_t tid, void** value);
+		BAN::ErrorOr<long> sys_pthread_join(pthread_t thread, void** value);
-		BAN::ErrorOr<long> sys_thread_getid();
+		BAN::ErrorOr<long> sys_pthread_self();
-		BAN::ErrorOr<long> sys_thread_kill(pid_t tid, int signal);
+		BAN::ErrorOr<long> sys_pthread_kill(pthread_t thread, int signal);
-		BAN::ErrorOr<long> sys_thread_detach(pid_t tid);
+		BAN::ErrorOr<long> sys_pthread_detach(pthread_t thread);
 
 		BAN::ErrorOr<long> sys_clock_gettime(clockid_t, timespec*);
 
@@ -228,12 +229,11 @@ namespace Kernel
 
 		vaddr_t shared_page_vaddr() const { return m_shared_page_vaddr; }
 
-		PageTable& page_table() { return *m_page_table; }
+		PageTable& page_table() { return m_page_table ? *m_page_table : PageTable::kernel(); }
 
 		size_t proc_meminfo(off_t offset, BAN::ByteSpan) const;
 		size_t proc_cmdline(off_t offset, BAN::ByteSpan) const;
 		size_t proc_environ(off_t offset, BAN::ByteSpan) const;
-		size_t proc_cputime(off_t offset, BAN::ByteSpan) const;
 		BAN::ErrorOr<BAN::String> proc_cwd() const;
 		BAN::ErrorOr<BAN::String> proc_exe() const;
 
@@ -364,13 +364,13 @@ namespace Kernel
 
 		BAN::Vector<Thread*> m_threads;
 
-		struct exited_thread_info_t
+		struct pthread_info_t
 		{
-			pid_t tid;
+			pthread_t thread;
 			void* value;
 		};
-		BAN::Vector<exited_thread_info_t> m_exited_threads;
+		BAN::Vector<pthread_info_t> m_exited_pthreads;
-		ThreadBlocker m_thread_exit_blocker;
+		ThreadBlocker m_pthread_exit_blocker;
 
 		uint64_t m_alarm_interval_ns { 0 };
 		uint64_t m_alarm_wake_time_ns { 0 };

kernel/include/kernel/Processor.h

@@ -57,12 +57,6 @@ namespace Kernel
 			};
 		};
 
-		struct LoadStats
-		{
-			uint64_t ns_idle;
-			uint64_t ns_total;
-		};
-
 	public:
 		static Processor& create(ProcessorID id);
 		static Processor& initialize();
@@ -78,6 +72,9 @@ namespace Kernel
 		static void set_smp_enabled() { s_is_smp_enabled = true; }
 		static void wait_until_processors_ready();
 
+		static void toggle_should_print_cpu_load() { s_should_print_cpu_load = !s_should_print_cpu_load; }
+		static bool get_should_print_cpu_load() { return s_should_print_cpu_load; }
+
 		static ProcessorID bsp_id() { return s_bsp_id; }
 		static bool current_is_bsp() { return current_id() == bsp_id(); }
 
@@ -119,12 +116,10 @@ namespace Kernel
 		static void* get_current_page_table()					{ return read_gs_sized<void*>(offsetof(Processor, m_current_page_table)); }
 		static void set_current_page_table(void* page_table)	{ write_gs_sized<void*>(offsetof(Processor, m_current_page_table), page_table); }
 
-		static LoadStats get_load_stats(size_t index);
-
 		static void yield();
 		static Scheduler& scheduler() { return *read_gs_sized<Scheduler*>(offsetof(Processor, m_scheduler)); }
 
-		static void initialize_tsc(uint64_t realtime_seconds);
+		static void initialize_tsc(uint8_t shift, uint64_t mult, uint64_t realtime_seconds);
 		static void update_tsc();
 		static uint64_t ns_since_boot_tsc();
 
@@ -192,6 +187,7 @@ namespace Kernel
 		static ProcessorID s_bsp_id;
 		static BAN::Atomic<uint8_t> s_processor_count;
 		static BAN::Atomic<bool>    s_is_smp_enabled;
+		static BAN::Atomic<bool>    s_should_print_cpu_load;
 		static paddr_t              s_shared_page_paddr;
 		static vaddr_t              s_shared_page_vaddr;
 
@@ -211,9 +207,10 @@ namespace Kernel
 
 		Scheduler* m_scheduler { nullptr };
 
-		BAN::Atomic<bool> m_load_stat_lock;
+		uint64_t m_start_ns { 0 };
-		uint64_t m_load_start_ns { 0 };
+		uint64_t m_idle_ns { 0 };
-		LoadStats m_load_stats {};
+		uint64_t m_last_update_ns { 0 };
+		uint64_t m_next_update_ns { 0 };
 
 		BAN::Atomic<SMPMessage*> m_smp_pending { nullptr };
 		BAN::Atomic<SMPMessage*> m_smp_free    { nullptr };

kernel/include/kernel/Scheduler.h

@@ -23,7 +23,7 @@ namespace Kernel
 
 	public:
 		void add_thread_to_back(Node*);
-		bool add_thread_with_wake_time(Node*); // return true if node was inserted as the first element
+		void add_thread_with_wake_time(Node*);
 		template<typename F>
 		Node* remove_with_condition(F callback);
 		void remove_node(Node*);
@@ -60,8 +60,7 @@ namespace Kernel
 		void reschedule(YieldRegisters*);
 		void reschedule_if_idle();
 
-		void on_timer_interrupt();
+		void timer_interrupt();
-		void on_yield(YieldRegisters*);
 
 		static BAN::ErrorOr<void> bind_thread_to_processor(Thread*, ProcessorID);
 		// if thread is already bound, this will never fail
@@ -83,7 +82,6 @@ namespace Kernel
 		void update_most_loaded_node_queue(SchedulerQueue::Node*, SchedulerQueue* target_queue);
 		void remove_node_from_most_loaded(SchedulerQueue::Node*);
 
-		void update_wake_up_deadline();
 		void wake_up_sleeping_threads();
 
 		void do_load_balancing();

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

@@ -10,14 +10,12 @@
 namespace Kernel
 {
 
-	class AHCIController final : public BAN::RefCounted<AHCIController>, public Interruptable
+	class AHCIController final : public StorageController, public Interruptable
 	{
 		BAN_NON_COPYABLE(AHCIController);
 		BAN_NON_MOVABLE(AHCIController);
 
 	public:
-		static BAN::ErrorOr<BAN::RefPtr<AHCIController>> create(PCI::Device&);
-
 		~AHCIController();
 
 		virtual void handle_irq() override;
@@ -29,7 +27,6 @@ namespace Kernel
 			: m_pci_device(pci_device)
 		{ }
 		BAN::ErrorOr<void> initialize();
-
 		BAN::Optional<AHCIPortType> check_port_type(volatile HBAPortMemorySpace&);
 
 	private:

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

@@ -56,6 +56,8 @@ namespace Kernel
 
 		// Non-owning pointers
 		BAN::Vector<ATADevice*> m_devices;
+
+		friend class ATAController;
 	};
 
 }

kernel/include/kernel/Storage/ATA/ATAController.h

@@ -2,22 +2,23 @@
 
 #include <BAN/UniqPtr.h>
 #include <kernel/PCI.h>
+#include <kernel/Storage/StorageController.h>
 #include <kernel/Storage/ATA/ATABus.h>
 #include <kernel/Storage/ATA/ATADevice.h>
 
 namespace Kernel
 {
 
-	class ATAController : public BAN::RefCounted<ATAController>
+	class ATAController : public StorageController
 	{
 	public:
-		static BAN::ErrorOr<BAN::RefPtr<ATAController>> create(PCI::Device&);
+		static BAN::ErrorOr<BAN::RefPtr<StorageController>> create(PCI::Device&);
+		virtual BAN::ErrorOr<void> initialize() override;
 
 	private:
 		ATAController(PCI::Device& pci_device)
 			: m_pci_device(pci_device)
 		{ }
-		BAN::ErrorOr<void> initialize();
 
 	private:
 		PCI::Device& m_pci_device;

kernel/include/kernel/Storage/NVMe/Controller.h

@@ -10,13 +10,13 @@
 namespace Kernel
 {
 
-	class NVMeController final : public CharacterDevice
+	class NVMeController final : public StorageController, public CharacterDevice
 	{
 		BAN_NON_COPYABLE(NVMeController);
 		BAN_NON_MOVABLE(NVMeController);
 
 	public:
-		static BAN::ErrorOr<BAN::RefPtr<NVMeController>> create(PCI::Device&);
+		static BAN::ErrorOr<BAN::RefPtr<StorageController>> create(PCI::Device&);
 
 		NVMeQueue& io_queue() { return *m_io_queue; }
 
@@ -30,7 +30,7 @@ namespace Kernel
 
 	private:
 		NVMeController(PCI::Device& pci_device);
-		BAN::ErrorOr<void> initialize();
+		virtual BAN::ErrorOr<void> initialize() override;
 
 		BAN::ErrorOr<void> identify_controller();
 		BAN::ErrorOr<void> identify_namespaces();

kernel/include/kernel/Storage/StorageController.h

@@ -0,0 +1,15 @@
+#pragma once
+
+#include <BAN/RefPtr.h>
+
+namespace Kernel
+{
+
+	class StorageController : public BAN::RefCounted<StorageController>
+	{
+	public:
+		virtual ~StorageController() {}
+		virtual BAN::ErrorOr<void> initialize() = 0;
+	};
+
+}

kernel/include/kernel/Terminal/PseudoTerminal.h

@@ -32,7 +32,7 @@ namespace Kernel
 
 		void on_close(int) override;
 
-		BAN::ErrorOr<long> ioctl_impl(unsigned long, void*) override;
+		BAN::ErrorOr<long> ioctl_impl(int, void*) override;
 
 	private:
 		PseudoTerminalMaster(BAN::UniqPtr<VirtualRange>, mode_t, uid_t, gid_t);

kernel/include/kernel/Terminal/TTY.h

@@ -68,7 +68,7 @@ namespace Kernel
 		virtual bool has_error_impl() const override { return false; }
 		virtual bool has_hungup_impl() const override { return false; }
 
-		virtual BAN::ErrorOr<long> ioctl_impl(unsigned long, void*) override;
+		virtual BAN::ErrorOr<long> ioctl_impl(int, void*) override;
 
 	private:
 		bool putchar(uint8_t ch);

kernel/include/kernel/Thread.h

@@ -41,20 +41,22 @@ namespace Kernel
 		// TODO: userspace stack size is hard limited, maybe make this dynamic?
 #if ARCH(x86_64)
 		static constexpr size_t userspace_stack_size { 32 << 20 };
-		static constexpr vaddr_t userspace_stack_base { 0x0000700000000000 };
 #elif ARCH(i686)
 		static constexpr size_t userspace_stack_size { 4 << 20 };
-		static constexpr vaddr_t userspace_stack_base { 0xB0000000 };
 #endif
 
 	public:
 		static BAN::ErrorOr<Thread*> create_kernel(entry_t, void*);
-		static BAN::ErrorOr<Thread*> create_userspace(Process*, PageTable&, vaddr_t userspace_stack_vaddr, size_t userspace_stack_size, vaddr_t entry_point, vaddr_t stack_pointer);
+		static BAN::ErrorOr<Thread*> create_userspace(Process*, PageTable&);
 		~Thread();
 
+		BAN::ErrorOr<Thread*> pthread_create(entry_t, void*);
+
 		BAN::ErrorOr<Thread*> clone(Process*, uintptr_t sp, uintptr_t ip);
 		void setup_process_cleanup();
 
+		BAN::ErrorOr<void> initialize_userspace(vaddr_t entry, BAN::Span<BAN::String> argv, BAN::Span<BAN::String> envp, BAN::Span<LibELF::AuxiliaryVector> auxv);
+
 		// Returns true, if thread is going to trigger signal
 		bool is_interrupted_by_signal(bool skip_stop_and_cont = false) const;
 
@@ -75,21 +77,15 @@ namespace Kernel
 
 		// blocks current thread and returns either on unblock, eintr, spuriously or after timeout
 		// if mutex is not nullptr, it will be atomically freed before blocking and automatically locked on wake
-		BAN::ErrorOr<void> sleep_or_eintr_for_ns(uint64_t timeout_ns);
+		BAN::ErrorOr<void> sleep_or_eintr_ns(uint64_t ns);
-		BAN::ErrorOr<void> sleep_or_eintr_until_ns(uint64_t waketime_ns);
 		BAN::ErrorOr<void> block_or_eintr_indefinite(ThreadBlocker& thread_blocker, BaseMutex* mutex);
 		BAN::ErrorOr<void> block_or_eintr_or_timeout_ns(ThreadBlocker& thread_blocker, uint64_t timeout_ns, bool etimedout, BaseMutex* mutex);
 		BAN::ErrorOr<void> block_or_eintr_or_waketime_ns(ThreadBlocker& thread_blocker, uint64_t wake_time_ns, bool etimedout, BaseMutex* mutex);
 
-		BAN::ErrorOr<void> sleep_or_eintr_for_ms(uint64_t timeout_ms)
+		BAN::ErrorOr<void> sleep_or_eintr_ms(uint64_t ms)
 		{
-			ASSERT(!BAN::Math::will_multiplication_overflow<uint64_t>(timeout_ms, 1'000'000));
+			ASSERT(!BAN::Math::will_multiplication_overflow<uint64_t>(ms, 1'000'000));
-			return sleep_or_eintr_for_ns(timeout_ms * 1'000'000);
+			return sleep_or_eintr_ns(ms * 1'000'000);
-		}
-		BAN::ErrorOr<void> sleep_or_eintr_until_ms(uint64_t waketime_ms)
-		{
-			ASSERT(!BAN::Math::will_multiplication_overflow<uint64_t>(waketime_ms, 1'000'000));
-			return sleep_or_eintr_until_ns(waketime_ms * 1'000'000);
 		}
 		BAN::ErrorOr<void> block_or_eintr_or_timeout_ms(ThreadBlocker& thread_blocker, uint64_t timeout_ms, bool etimedout, BaseMutex* mutex)
 		{
@@ -110,6 +106,8 @@ namespace Kernel
 		vaddr_t kernel_stack_top() const	{ return m_kernel_stack->vaddr() + m_kernel_stack->size(); }
 		VirtualRange& kernel_stack() { return *m_kernel_stack; }
 
+		MemoryBackedRegion& userspace_stack() { ASSERT(is_userspace() && m_userspace_stack); return *m_userspace_stack; }
+
 		static Thread& current();
 		static pid_t current_tid();
 
@@ -153,6 +151,8 @@ namespace Kernel
 	private:
 		Thread(pid_t tid, Process*);
 
+		void setup_exec(vaddr_t ip, vaddr_t sp);
+
 		static void on_exit_trampoline(Thread*);
 		void on_exit();
 
@@ -174,6 +174,7 @@ namespace Kernel
 		BAN::UniqPtr<PageTable>    m_keep_alive_page_table;
 
 		BAN::UniqPtr<VirtualRange> m_kernel_stack;
+		MemoryBackedRegion*        m_userspace_stack      { nullptr };
 		const pid_t                m_tid                  { 0 };
 		State                      m_state                { State::NotStarted };
 		Process*                   m_process              { nullptr };
@@ -181,9 +182,6 @@ namespace Kernel
 		BAN::Atomic<bool>          m_is_detached          { false };
 		bool                       m_delete_process       { false };
 
-		vaddr_t                    m_userspace_stack_vaddr { 0 };
-		size_t                     m_userspace_stack_size  { 0 };
-
 		vaddr_t                    m_fsbase               { 0 };
 		vaddr_t                    m_gsbase               { 0 };
 

kernel/include/kernel/Timer/HPET.h

@@ -12,7 +12,7 @@ namespace Kernel
 	class HPET final : public Timer, public Interruptable
 	{
 	public:
-		static BAN::ErrorOr<BAN::UniqPtr<HPET>> create();
+		static BAN::ErrorOr<BAN::UniqPtr<HPET>> create(bool force_pic);
 		~HPET();
 
 		virtual uint64_t ms_since_boot() const override;
@@ -26,7 +26,7 @@ namespace Kernel
 
 	private:
 		HPET() = default;
-		BAN::ErrorOr<void> initialize();
+		BAN::ErrorOr<void> initialize(bool force_pic);
 
 		volatile HPETRegisters& registers();
 		const volatile HPETRegisters& registers() const;

kernel/include/kernel/Timer/Timer.h

@@ -2,7 +2,6 @@
 
 #include <BAN/UniqPtr.h>
 #include <BAN/Vector.h>
-#include <kernel/Memory/DMARegion.h>
 #include <kernel/Timer/RTC.h>
 
 #include <time.h>
@@ -32,14 +31,7 @@ namespace Kernel
 	class SystemTimer : public Timer
 	{
 	public:
-		struct TSCInfo
+		static void initialize(bool force_pic);
-		{
-			int8_t shift;
-			uint32_t mult;
-		};
-
-	public:
-		static void initialize();
 		static SystemTimer& get();
 		static bool is_initialized();
 
@@ -52,60 +44,29 @@ namespace Kernel
 		virtual bool pre_scheduler_sleep_needs_lock() const override;
 		virtual void pre_scheduler_sleep_ns(uint64_t) override;
 
-		void sleep_for_ns(uint64_t timeout_ns) const;
+		void sleep_ms(uint64_t ms) const { ASSERT(!BAN::Math::will_multiplication_overflow<uint64_t>(ms, 1'000'000)); return sleep_ns(ms * 1'000'000); }
-		void sleep_until_ns(uint64_t waketime_ns) const;
+		void sleep_ns(uint64_t ns) const;
-
-		void sleep_for_ms(uint64_t timeout_ms) const
-		{
-			ASSERT(!BAN::Math::will_multiplication_overflow<uint64_t>(timeout_ms, 1'000'000));
-			return sleep_for_ns(timeout_ms * 1'000'000);
-		}
-		void sleep_until_ms(uint64_t waketime_ms) const
-		{
-			ASSERT(!BAN::Math::will_multiplication_overflow<uint64_t>(waketime_ms, 1'000'000));
-			return sleep_until_ns(waketime_ms * 1'000'000);
-		}
 
 		void dont_invoke_scheduler() { m_timer->m_should_invoke_scheduler = false; }
 
-		void update_tsc();
+		void update_tsc() const;
-		TSCInfo tsc_info() const;
+		uint64_t ns_since_boot_no_tsc() const;
-
-		uint64_t ns_since_boot_no_tsc() const { return m_timer->ns_since_boot(); }
 
 		timespec real_time() const;
 
 	private:
 		SystemTimer() = default;
 
-		void initialize_timers();
+		void initialize_timers(bool force_pic);
 
-		void initialize_invariant_tsc();
+		uint64_t get_tsc_frequency() const;
-		void initialize_pvclock();
 
 	private:
-		enum class TSCType
-		{
-			None,
-			Invariant,
-			PVClock,
-		};
-
 		uint64_t m_boot_time { 0 };
 		BAN::UniqPtr<RTC> m_rtc;
 		BAN::UniqPtr<Timer> m_timer;
-
+		bool m_has_invariant_tsc { false };
-		TSCType m_tsc_type = TSCType::None;
+		mutable uint32_t m_timer_ticks { 0 };
-
-		union {
-			struct {
-				int8_t shift;
-				uint32_t mult;
-			} invariant;
-		} m_tsc_info;
-
-		BAN::UniqPtr<DMARegion> m_tsc_page;
-		uint64_t m_tsc_update_ns { 0 };
 	};
 
 }

kernel/include/kernel/USB/HID/Joystick.h

@@ -34,7 +34,7 @@ namespace Kernel
 		BAN::ErrorOr<size_t> read_impl(off_t, BAN::ByteSpan) override;
 		bool can_read_impl() const override { return true; }
 
-		BAN::ErrorOr<long> ioctl_impl(unsigned long request, void* arg) override;
+		BAN::ErrorOr<long> ioctl_impl(int request, void* arg) override;
 
 	private:
 		USBJoystick(USBHIDDriver&);

kernel/include/kernel/USB/MassStorage/SCSIDevice.h

@@ -33,7 +33,7 @@ namespace Kernel
 		const uint64_t m_block_count;
 		const uint32_t m_block_size;
 
-		char m_name[4];
+		const char m_name[4];
 
 		friend class BAN::RefPtr<USBSCSIDevice>;
 	};

kernel/include/kernel/UserCopy.h

@@ -1,12 +1,7 @@
 #pragma once
-
 #include <BAN/Errors.h>
-
+namespace Kernel {
-namespace Kernel
-{
-
 	BAN::ErrorOr<void> read_from_user(const void* user_addr, void* out, size_t size);
 	BAN::ErrorOr<void> read_string_from_user(const char* user_addr, char* out, size_t max_size);
 	BAN::ErrorOr<void> write_to_user(void* user_addr, const void* in, size_t size);
-
 };

kernel/kernel/ACPI/ACPI.cpp

@@ -952,7 +952,7 @@ acpi_release_global_lock:
 			{
 				if (IO::inw(fadt().pm1a_cnt_blk) & PM1_CNT_SCI_EN)
 					break;
-				SystemTimer::get().sleep_for_ms(10);
+				SystemTimer::get().sleep_ms(10);
 			}
 
 			if (!(IO::inw(fadt().pm1a_cnt_blk) & PM1_CNT_SCI_EN))

kernel/kernel/ACPI/AML/Node.cpp

@@ -2146,9 +2146,9 @@ namespace Kernel::ACPI::AML
 
 		auto milliseconds = TRY(convert_node(TRY(parse_node(context)), ConvInteger, sizeof(uint64_t)));
 
-		const uint64_t waketime_ms = SystemTimer::get().ms_since_boot() + milliseconds.as.integer.value;
+		const uint64_t wakeup_ms = SystemTimer::get().ms_since_boot() + milliseconds.as.integer.value;
-		while (SystemTimer::get().ms_since_boot() < waketime_ms)
+		for (uint64_t curr_ms = SystemTimer::get().ms_since_boot(); curr_ms < wakeup_ms; curr_ms = SystemTimer::get().ms_since_boot())
-			SystemTimer::get().sleep_until_ms(waketime_ms);
+			SystemTimer::get().sleep_ms(wakeup_ms - curr_ms);
 
 		return {};
 	}
@@ -2392,7 +2392,7 @@ namespace Kernel::ACPI::AML
 		result.type = Node::Type::Integer;
 		result.as.integer.value = BAN::numeric_limits<uint64_t>::max();
 
-		const uint64_t waketime_ms = (timeout_ms >= 0xFFFF)
+		const uint64_t wake_time_ms = (timeout_ms >= 0xFFFF)
 			? BAN::numeric_limits<uint64_t>::max()
 			: SystemTimer::get().ms_since_boot() + timeout_ms;
 
@@ -2401,9 +2401,9 @@ namespace Kernel::ACPI::AML
 		{
 			if (sync_object->node.as.mutex->mutex.try_lock())
 				break;
-			if (SystemTimer::get().ms_since_boot() >= waketime_ms)
+			if (SystemTimer::get().ms_since_boot() >= wake_time_ms)
 				return result;
-			Processor::yield();
+			SystemTimer::get().sleep_ms(1);
 		}
 
 		if (sync_object->node.as.mutex->global_lock)

kernel/kernel/APIC.cpp

@@ -437,27 +437,8 @@ namespace Kernel
 
 		dprintln("CPU {}: lapic timer frequency: {} Hz", Kernel::Processor::current_id(), m_lapic_timer_frequency_hz);
 
-		write_to_local_apic(LAPIC_TIMER_LVT,         TimerModeOneShot | IRQ_TIMER);
+		write_to_local_apic(LAPIC_TIMER_LVT,         TimerModePeriodic | IRQ_TIMER);
-		write_to_local_apic(LAPIC_TIMER_INITIAL_REG, 0);
+		write_to_local_apic(LAPIC_TIMER_INITIAL_REG, m_lapic_timer_frequency_hz / 2 / 100);
-	}
-
-	void APIC::set_timer_dealine(uint64_t target_ns)
-	{
-		// TODO: don't hardcode divide by 2
-		const uint64_t effective_freq = m_lapic_timer_frequency_hz / 2;
-
-		const uint64_t delta_ns = [&]() -> uint64_t {
-			const uint64_t current_ns = SystemTimer::get().ns_since_boot();
-			if (current_ns >= target_ns)
-				return 1;
-			const uint64_t delta_ns = target_ns - current_ns;
-			if (BAN::Math::will_multiplication_overflow(delta_ns, effective_freq))
-				return BAN::numeric_limits<uint64_t>::max() / effective_freq;
-			return delta_ns;
-		}();
-
-		const uint64_t ticks = BAN::Math::div_round_up<uint64_t>(delta_ns * effective_freq, 1'000'000'000);
-		write_to_local_apic(LAPIC_TIMER_INITIAL_REG, BAN::Math::min<uint64_t>(ticks, BAN::numeric_limits<uint32_t>::max()));
 	}
 
 	uint32_t APIC::read_from_local_apic(ptrdiff_t offset)

kernel/kernel/Audio/Controller.cpp

@@ -77,9 +77,9 @@ namespace Kernel
 		return to_copy;
 	}
 
-	BAN::ErrorOr<long> AudioController::ioctl_impl(unsigned long request, void* arg)
+	BAN::ErrorOr<long> AudioController::ioctl_impl(int cmd, void* arg)
 	{
-		switch (request)
+		switch (cmd)
 		{
 			case SND_GET_CHANNELS:
 				*static_cast<uint32_t*>(arg) = get_channels();
@@ -92,7 +92,7 @@ namespace Kernel
 			{
 				SpinLockGuard _(m_spinlock);
 				*static_cast<uint32_t*>(arg) = m_sample_data->size();
-				if (request == SND_RESET_BUFFER)
+				if (cmd == SND_RESET_BUFFER)
 					m_sample_data->pop(m_sample_data->size());
 				return 0;
 			}
@@ -113,7 +113,7 @@ namespace Kernel
 				return 0;
 		}
 
-		return CharacterDevice::ioctl_impl(request, arg);
+		return CharacterDevice::ioctl_impl(cmd, arg);
 	}
 
 }

kernel/kernel/Audio/HDAudio/AudioFunctionGroup.cpp

@@ -448,7 +448,7 @@ namespace Kernel
 				continue;
 			const auto& amp = path[i]->output_amplifier.value();
 
-			const int32_t step_mdB = (amp.step_size + 1) * 250;
+			const int32_t step_mdB = amp.step_size * 250;
 			m_volume_info.step_mdB = step_mdB;
 			m_volume_info.min_mdB = -amp.offset * step_mdB;
 			m_volume_info.max_mdB = (amp.num_steps - amp.offset) * step_mdB;

kernel/kernel/Audio/HDAudio/Controller.cpp

@@ -113,7 +113,7 @@ namespace Kernel
 		// 4.3 The software must wait at least 521 us (25 frames) after reading CRST as a 1
 		// before assuming that codecs have all made status change requests and have been
 		// registered by the controller
-		SystemTimer::get().sleep_for_ns(521'000);
+		SystemTimer::get().sleep_ms(1);
 
 		return {};
 	}

kernel/kernel/CPUID.cpp

@@ -85,38 +85,6 @@ namespace CPUID
 		return buffer[3] & (1 << 8);
 	}
 
-	bool has_rdtscp()
-	{
-		uint32_t buffer[4] {};
-		get_cpuid(0x80000000, buffer);
-		if (buffer[0] < 0x80000001)
-			return false;
-		get_cpuid(0x80000001, buffer);
-		return buffer[3] & (1 << 27);
-	}
-
-	uint64_t get_tsc_frequency()
-	{
-		uint32_t buffer[4];
-		get_cpuid(0x00, buffer);
-		if (buffer[0] < 0x15)
-			return 0;
-		get_cpuid(0x15, buffer);
-		if (buffer[0] == 0 || buffer[1] == 0 || buffer[2] == 0)
-			return 0;
-		return static_cast<uint64_t>(buffer[2]) * buffer[1] / buffer[0];
-	}
-
-	bool has_kvm_pvclock()
-	{
-		uint32_t buffer[4] {};
-		get_cpuid(0x40000000, buffer);
-		if (buffer[1] != 0x4B4D564B || buffer[2] != 0x564B4D56 || buffer[3] != 0x4D)
-			return false;
-		get_cpuid(0x40000001, buffer);
-		return buffer[0] & (1 << 3);
-	}
-
 	const char* feature_string_ecx(uint32_t feat)
 	{
 		switch (feat)

kernel/kernel/Device/FramebufferDevice.cpp

@@ -135,9 +135,9 @@ namespace Kernel
 		return bytes_to_copy;
 	}
 
-	BAN::ErrorOr<long> FramebufferDevice::ioctl_impl(unsigned long request, void* arg)
+	BAN::ErrorOr<long> FramebufferDevice::ioctl_impl(int cmd, void* arg)
 	{
-		switch (request)
+		switch (cmd)
 		{
 			case FB_MSYNC_RECTANGLE:
 			{
@@ -152,7 +152,7 @@ namespace Kernel
 			}
 		}
 
-		return CharacterDevice::ioctl_impl(request, arg);
+		return CharacterDevice::ioctl(cmd, arg);
 	}
 
 	uint32_t FramebufferDevice::get_pixel(uint32_t x, uint32_t y) const
@@ -356,7 +356,51 @@ namespace Kernel
 
 		void do_msync(uint32_t first_pixel, uint32_t pixel_count)
 		{
-			m_framebuffer->sync_pixels_linear(first_pixel, pixel_count);
+			if (!Processor::get_should_print_cpu_load())
+				return m_framebuffer->sync_pixels_linear(first_pixel, pixel_count);
+
+			const uint32_t fb_width = m_framebuffer->width();
+
+			// If we are here (in FramebufferMemoryRegion), our terminal driver is FramebufferTerminalDriver
+			ASSERT(g_terminal_driver->has_font());
+			const auto& font = g_terminal_driver->font();
+
+			const uint32_t x = first_pixel % fb_width;
+			const uint32_t y = first_pixel / fb_width;
+
+			const uint32_t load_w = 16                 * font.width();
+			const uint32_t load_h = Processor::count() * font.height();
+
+			if (y >= load_h || x + pixel_count <= fb_width - load_w)
+				return m_framebuffer->sync_pixels_linear(first_pixel, pixel_count);
+
+			if (x >= fb_width - load_w && x + pixel_count <= fb_width)
+				return;
+
+			if (x < fb_width - load_w)
+				m_framebuffer->sync_pixels_linear(first_pixel, fb_width - load_w - x);
+
+			if (x + pixel_count > fb_width)
+			{
+				const uint32_t past_last_pixel = first_pixel + pixel_count;
+
+				first_pixel = (y + 1) * fb_width;
+				pixel_count = past_last_pixel - first_pixel;
+
+				const uint32_t cpu_load_end = load_h * fb_width;
+
+				while (pixel_count && first_pixel < cpu_load_end)
+				{
+					m_framebuffer->sync_pixels_linear(first_pixel, BAN::Math::min(pixel_count, fb_width - load_w));
+
+					const uint32_t advance = BAN::Math::min(pixel_count, fb_width);
+					pixel_count -= advance;
+					first_pixel += advance;
+				}
+
+				if (pixel_count)
+					m_framebuffer->sync_pixels_linear(first_pixel, pixel_count);
+			}
 		}
 
 	private:

kernel/kernel/ELF.cpp

@@ -8,6 +8,7 @@
 
 #include <ctype.h>
 #include <fcntl.h>
+#include <pthread.h>
 
 namespace Kernel::ELF
 {
@@ -290,9 +291,7 @@ namespace Kernel::ELF
 			TRY(memory_regions.emplace_back(BAN::move(region)));
 		}
 
-		if (!interpreter.empty())
+		result.open_execfd = !interpreter.empty();
-			result.interp_base = load_base_vaddr;
-
 		result.entry_point = load_base_vaddr + file_header.e_entry;
 		result.regions = BAN::move(memory_regions);
 		return BAN::move(result);

kernel/kernel/FS/DevFS/FileSystem.cpp

@@ -50,7 +50,6 @@ namespace Kernel
 			[](void* _devfs)
 			{
 				auto* devfs = static_cast<DevFileSystem*>(_devfs);
-				uint64_t next_update_ms = SystemTimer::get().ms_since_boot();
 				while (true)
 				{
 					{
@@ -58,8 +57,7 @@ namespace Kernel
 						for (auto& device : devfs->m_devices)
 							device->update();
 					}
-					SystemTimer::get().sleep_until_ms(next_update_ms);
+					SystemTimer::get().sleep_ms(10);
-					next_update_ms += 10;
 				}
 			}, s_instance
 		));

kernel/kernel/FS/Ext2/FileSystem.cpp

@@ -167,36 +167,10 @@ namespace Kernel
 
 	BAN::ErrorOr<void> Ext2FS::initialize_root_inode()
 	{
-		m_root_inode = TRY(open_inode(Ext2::Enum::ROOT_INO));
+		m_root_inode = TRY(Ext2Inode::create(*this, Ext2::Enum::ROOT_INO));
 		return {};
 	}
 
-	BAN::ErrorOr<BAN::RefPtr<Ext2Inode>> Ext2FS::open_inode(ino_t ino)
-	{
-		LockGuard _(m_inode_cache_lock);
-
-		auto it = m_inode_cache.find(ino);
-		if (it != m_inode_cache.end())
-			return it->value;
-
-		auto inode_location = TRY(locate_inode(ino));
-
-		auto block_buffer = TRY(get_block_buffer());
-		TRY(read_block(inode_location.block, block_buffer));
-
-		auto& inode = block_buffer.span().slice(inode_location.offset).as<Ext2::Inode>();
-
-		auto result = TRY(BAN::RefPtr<Ext2Inode>::create(*this, inode, ino));
-		TRY(m_inode_cache.insert(ino, result));
-		return result;
-	}
-
-	void Ext2FS::remove_from_cache(ino_t ino)
-	{
-		LockGuard _(m_inode_cache_lock);
-		m_inode_cache.remove(ino);
-	}
-
 	BAN::ErrorOr<uint32_t> Ext2FS::create_inode(const Ext2::Inode& ext2_inode)
 	{
 		auto bgd_buffer = TRY(m_buffer_manager.get_buffer());

kernel/kernel/FS/Ext2/Inode.cpp

@@ -42,6 +42,24 @@ namespace Kernel
 		return (m_ino - 1) / m_fs.superblock().blocks_per_group;
 	}
 
+	BAN::ErrorOr<BAN::RefPtr<Ext2Inode>> Ext2Inode::create(Ext2FS& fs, uint32_t inode_ino)
+	{
+		auto it = fs.inode_cache().find(inode_ino);
+		if (it != fs.inode_cache().end())
+			return it->value;
+
+		auto inode_location = TRY(fs.locate_inode(inode_ino));
+
+		auto block_buffer = TRY(fs.get_block_buffer());
+		TRY(fs.read_block(inode_location.block, block_buffer));
+
+		auto& inode = block_buffer.span().slice(inode_location.offset).as<Ext2::Inode>();
+
+		auto result = TRY(BAN::RefPtr<Ext2Inode>::create(fs, inode, inode_ino));
+		TRY(fs.inode_cache().insert(inode_ino, result));
+		return result;
+	}
+
 	Ext2Inode::~Ext2Inode()
 	{
 		if (m_nlink > 0)
@@ -71,53 +89,57 @@ namespace Kernel
 		return {};
 	}
 
-	BAN::ErrorOr<BAN::Optional<uint32_t>> Ext2Inode::block_from_indirect_block_no_lock(uint32_t block, uint32_t index, uint32_t depth, bool allocate)
+	BAN::ErrorOr<BAN::Optional<uint32_t>> Ext2Inode::block_from_indirect_block_no_lock(uint32_t& block, uint32_t index, uint32_t depth, bool allocate)
 	{
 		const uint32_t indices_per_fs_block = blksize() / sizeof(uint32_t);
 
-		ASSERT(block != 0);
+		if (block == 0 && !allocate)
+			return BAN::Optional<uint32_t>();
 
 		if (depth == 0)
-			return BAN::Optional<uint32_t>(block);
-
-		uint32_t local_index = index;
-		for (uint32_t i = 1; i < depth; i++)
-			local_index /= indices_per_fs_block;
-		local_index %= indices_per_fs_block;
-
-		uint32_t next_block = 0;
-
-		if (auto cached = block_cache_find(block, local_index); cached.has_value())
-			next_block = cached.value();
-
-		if (next_block == 0)
 		{
-			auto block_buffer = TRY(m_fs.get_block_buffer());
+			if (block == 0)
-			TRY(m_fs.read_block(block, block_buffer));
-			auto block_span = block_buffer.span().as_span<uint32_t>();
-
-			next_block = block_span[local_index];
-
-			if (next_block == 0)
 			{
-				if (!allocate)
+				block = TRY(m_fs.reserve_free_block(block_group()));
-					return BAN::Optional<uint32_t>();
-
-				auto zero_buffer = TRY(m_fs.get_block_buffer());
-				memset(zero_buffer.data(), 0, zero_buffer.size());
-
-				next_block = TRY(m_fs.reserve_free_block(block_group()));
-				TRY(m_fs.write_block(next_block, zero_buffer));
 				m_blocks++;
 
-				block_span[local_index] = next_block;
+				auto block_buffer = TRY(m_fs.get_block_buffer());
+				memset(block_buffer.data(), 0x00, block_buffer.size());
 				TRY(m_fs.write_block(block, block_buffer));
 			}
 
-			block_cache_add(block, local_index, next_block);
+			return BAN::Optional<uint32_t>(block);
 		}
 
-		return block_from_indirect_block_no_lock(next_block, index, depth - 1, allocate);
+		auto block_buffer = TRY(m_fs.get_block_buffer());
+
+		bool needs_write = false;
+
+		if (block != 0)
+			TRY(m_fs.read_block(block, block_buffer));
+		else
+		{
+			block = TRY(m_fs.reserve_free_block(block_group()));
+			m_blocks++;
+
+			memset(block_buffer.data(), 0, block_buffer.size());
+
+			needs_write = true;
+		}
+
+		uint32_t divisor = 1;
+		for (uint32_t i = 1; i < depth; i++)
+			divisor *= indices_per_fs_block;
+
+		uint32_t& new_block = block_buffer.span().as_span<uint32_t>()[(index / divisor) % indices_per_fs_block];
+		const auto old_block = new_block;
+
+		const auto result = TRY(block_from_indirect_block_no_lock(new_block, index, depth - 1, allocate));
+
+		if (needs_write || old_block != new_block)
+			TRY(m_fs.write_block(block, block_buffer));
+
+		return result;
 	}
 
 	BAN::ErrorOr<BAN::Optional<uint32_t>> Ext2Inode::fs_block_of_data_block_index_no_lock(uint32_t data_block_index, bool allocate)
@@ -144,29 +166,16 @@ namespace Kernel
 		}
 		data_block_index -= 12;
 
-		uint32_t depth_block_count = indices_per_block;
+		if (data_block_index < indices_per_block)
-		for (size_t i = 0; i < 3; i++)
+			return block_from_indirect_block_no_lock(m_ext2_blocks.block[12], data_block_index, 1, allocate);
-		{
+		data_block_index -= indices_per_block;
-			if (data_block_index < depth_block_count)
-			{
-				auto& block = m_ext2_blocks.block[12 + i];
-				if (block == 0)
-				{
-					if (!allocate)
-						return BAN::Optional<uint32_t>();
 
-					auto zero_buffer = TRY(m_fs.get_block_buffer());
+		if (data_block_index < indices_per_block * indices_per_block)
-					memset(zero_buffer.data(), 0, zero_buffer.size());
+			return block_from_indirect_block_no_lock(m_ext2_blocks.block[13], data_block_index, 2, allocate);
+		data_block_index -= indices_per_block * indices_per_block;
 
-					block = TRY(m_fs.reserve_free_block(block_group()));
+		if (data_block_index < indices_per_block * indices_per_block * indices_per_block)
-					TRY(m_fs.write_block(block, zero_buffer));
+			return block_from_indirect_block_no_lock(m_ext2_blocks.block[14], data_block_index, 3, allocate);
-					m_blocks++;
-				}
-				return block_from_indirect_block_no_lock(block, data_block_index, i + 1, allocate);
-			}
-			data_block_index -= depth_block_count;
-			depth_block_count *= indices_per_block;
-		}
 
 		ASSERT_NOT_REACHED();
 	}
@@ -533,7 +542,7 @@ done:
 
 		const uint32_t new_ino = TRY(m_fs.create_inode(initialize_new_inode_info(mode, uid, gid)));
 
-		auto inode_or_error = m_fs.open_inode(new_ino);
+		auto inode_or_error = Ext2Inode::create(m_fs, new_ino);
 		if (inode_or_error.is_error())
 		{
 			TRY(m_fs.delete_inode(new_ino));
@@ -559,7 +568,7 @@ done:
 
 		const uint32_t new_ino = TRY(m_fs.create_inode(initialize_new_inode_info(mode, uid, gid)));
 
-		auto inode_or_error = m_fs.open_inode(new_ino);
+		auto inode_or_error = Ext2Inode::create(m_fs, new_ino);
 		if (inode_or_error.is_error())
 		{
 			TRY(m_fs.delete_inode(new_ino));
@@ -805,7 +814,7 @@ needs_new_block:
 					}
 					else if (entry_name == ".."_sv)
 					{
-						auto parent = TRY(m_fs.open_inode(entry.inode));
+						auto parent = TRY(Ext2Inode::create(m_fs, entry.inode));
 						parent->m_nlink--;
 						TRY(parent->sync_inode_no_lock());
 					}
@@ -851,7 +860,7 @@ needs_new_block:
 				auto& entry = block_buffer.span().slice(offset).as<Ext2::LinkedDirectoryEntry>();
 				if (entry.inode && name == BAN::StringView(entry.name, entry.name_len))
 				{
-					auto inode = TRY(m_fs.open_inode(entry.inode));
+					auto inode = TRY(Ext2Inode::create(m_fs, entry.inode));
 					if (cleanup_directory && inode->mode().ifdir())
 					{
 						if (!TRY(inode->is_directory_empty_no_lock()))
@@ -869,7 +878,11 @@ needs_new_block:
 					// NOTE: If this was the last link to inode we must
 					//       remove it from inode cache to trigger cleanup
 					if (inode->nlink() == 0)
-						m_fs.remove_from_cache(inode->ino());
+					{
+						auto& cache = m_fs.inode_cache();
+						if (cache.contains(inode->ino()))
+							cache.remove(inode->ino());
+					}
 
 					// FIXME: This should expand the last inode if exists
 					entry.inode = 0;
@@ -879,8 +892,6 @@ needs_new_block:
 			}
 		}
 
-		dir_cache_remove(name);
-
 		return {};
 	}
 
@@ -917,8 +928,6 @@ needs_new_block:
 		auto inode_location = TRY(m_fs.locate_inode(ino()));
 		auto block_buffer = TRY(m_fs.get_block_buffer());
 
-		// FIXME: race condition when syncing multiple inodes :)
-
 		TRY(m_fs.read_block(inode_location.block, block_buffer));
 		if (memcmp(block_buffer.data() + inode_location.offset, &inode, sizeof(Ext2::Inode)))
 		{
@@ -939,9 +948,6 @@ needs_new_block:
 	{
 		ASSERT(mode().ifdir());
 
-		if (auto cached = dir_cache_find(file_name))
-			return cached;
-
 		auto block_buffer = TRY(m_fs.get_block_buffer());
 
 		for (uint32_t i = 0; i < max_used_data_block_count(); i++)
@@ -957,129 +963,11 @@ needs_new_block:
 				auto& entry = entry_span.as<const Ext2::LinkedDirectoryEntry>();
 				BAN::StringView entry_name(entry.name, entry.name_len);
 				if (entry.inode && entry_name == file_name)
-				{
+					return BAN::RefPtr<Inode>(TRY(Ext2Inode::create(m_fs, entry.inode)));
-					auto inode = BAN::RefPtr<Inode>(TRY(m_fs.open_inode(entry.inode)));
-					dir_cache_add(file_name, inode);
-					return inode;
-				}
 				entry_span = entry_span.slice(entry.rec_len);
 			}
 		}
 
 		return BAN::Error::from_errno(ENOENT);
 	}
-
-	BAN::Optional<uint32_t> Ext2Inode::block_cache_find(uint32_t block, uint32_t index) const
-	{
-		SpinLockGuard _(m_block_cache_lock);
-		for (const auto& cache : m_block_cache)
-		{
-			if (cache.block != block || cache.index != index)
-				continue;
-			cache.freq++;
-			return cache.target;
-		}
-		return {};
-	}
-
-	void Ext2Inode::block_cache_remove(uint32_t block, uint32_t index)
-	{
-		SpinLockGuard _(m_block_cache_lock);
-		for (auto& cache : m_block_cache)
-		{
-			if (cache.block != block || cache.index != index)
-				continue;
-			cache = {};
-			return;
-		}
-	}
-
-	void Ext2Inode::block_cache_add(uint32_t block, uint32_t index, uint32_t target)
-	{
-		SpinLockGuard _(m_block_cache_lock);
-
-		size_t min_freq = BAN::numeric_limits<size_t>::max();
-		size_t min_index = 0;
-		for (size_t i = 0; i < m_block_cache.size(); i++)
-		{
-			const auto& cache = m_block_cache[i];
-			if (cache.block == block && cache.index == index)
-			{
-				ASSERT(cache.target == target);
-				cache.freq++;
-				return;
-			}
-			if (cache.freq < min_freq)
-			{
-				min_freq = cache.freq;
-				min_index = i;
-			}
-		}
-
-		m_block_cache[min_index] = {
-			.freq = 1,
-			.block = block,
-			.index = index,
-			.target = target,
-		};
-	}
-
-	BAN::RefPtr<Inode> Ext2Inode::dir_cache_find(BAN::StringView name) const
-	{
-		RWLockRDGuard _(m_dir_cache_lock);
-		for (const auto& cache : m_dir_cache)
-		{
-			if (!cache.inode || name != BAN::StringView(cache.name, cache.name_len))
-				continue;
-			BAN::atomic_add_fetch(cache.freq, 1);
-			return cache.inode;
-		}
-		return {};
-	}
-
-	void Ext2Inode::dir_cache_remove(BAN::StringView name)
-	{
-		RWLockWRGuard _(m_dir_cache_lock);
-		for (auto& cache : m_dir_cache)
-		{
-			if (!cache.inode || name != BAN::StringView(cache.name, cache.name_len))
-				continue;
-			cache.freq = 0;
-			cache.inode = {};
-			return;
-		}
-	}
-
-	void Ext2Inode::dir_cache_add(BAN::StringView name, BAN::RefPtr<Inode> inode)
-	{
-		RWLockWRGuard _(m_dir_cache_lock);
-
-		if (m_dir_cache.empty() && m_dir_cache.resize(dir_cache_size).is_error())
-			return;
-
-		size_t min_freq = BAN::numeric_limits<size_t>::max();
-		size_t min_index = 0;
-		for (size_t i = 0; i < m_dir_cache.size(); i++)
-		{
-			const auto& cache = m_dir_cache[i];
-			if (cache.inode && name == BAN::StringView(cache.name, cache.name_len))
-			{
-				ASSERT(cache.inode == inode);
-				cache.freq++;
-				return;
-			}
-			if (cache.freq < min_freq)
-			{
-				min_freq  = cache.freq;
-				min_index = i;
-			}
-		}
-
-		auto& cache = m_dir_cache[min_index];
-		cache.freq     = 1;
-		cache.inode    = inode;
-		cache.name_len = name.size();
-		memcpy(cache.name, name.data(), name.size());
-	}
-
 }

kernel/kernel/FS/Inode.cpp

@@ -316,7 +316,7 @@ namespace Kernel
 		return {};
 	}
 
-	BAN::ErrorOr<long> Inode::ioctl(unsigned long request, void* arg)
+	BAN::ErrorOr<long> Inode::ioctl(int request, void* arg)
 	{
 		auto ret = ioctl_impl(request, arg);
 		if (!ret.is_error() || ret.error().get_error_code() != ENOTSUP)

kernel/kernel/FS/ProcFS/FileSystem.cpp

@@ -16,23 +16,22 @@ namespace Kernel
 		MUST(s_instance->TmpFileSystem::initialize(0555, 0, 0));
 
 		auto meminfo_inode = MUST(ProcROInode::create_new(
-			[](off_t offset, BAN::ByteSpan buffer, void*) -> BAN::ErrorOr<size_t>
+			[](off_t offset, BAN::ByteSpan buffer) -> size_t
 			{
 				ASSERT(offset >= 0);
 				if ((size_t)offset >= sizeof(full_meminfo_t))
 					return 0;
 
-				const full_meminfo_t meminfo {
+				full_meminfo_t meminfo;
-					.page_size = PAGE_SIZE,
+				meminfo.page_size = PAGE_SIZE;
-					.free_pages = Heap::get().free_pages(),
+				meminfo.free_pages = Heap::get().free_pages();
-					.used_pages = Heap::get().used_pages(),
+				meminfo.used_pages = Heap::get().used_pages();
-				};
 
 				size_t bytes = BAN::Math::min<size_t>(sizeof(full_meminfo_t) - offset, buffer.size());
 				memcpy(buffer.data(), (uint8_t*)&meminfo + offset, bytes);
 				return bytes;
 			},
-			*s_instance, nullptr, 0444, 0, 0
+			*s_instance, 0444, 0, 0
 		));
 		MUST(static_cast<TmpDirectoryInode*>(s_instance->root_inode().ptr())->link_inode(*meminfo_inode, "meminfo"_sv));
 
@@ -49,35 +48,6 @@ namespace Kernel
 		MUST(static_cast<TmpDirectoryInode*>(s_instance->root_inode().ptr())->link_inode(*self_inode, "self"_sv));
 	}
 
-	void ProcFileSystem::post_scheduler_initialize()
-	{
-		MUST(s_instance->root_inode()->create_directory("cpu"_sv, Inode::Mode::IFDIR | 0555, 0, 0));
-		auto cpu_directory = MUST(s_instance->root_inode()->find_inode("cpu"_sv));
-
-		for (size_t i = 0; i < Processor::count(); i++)
-		{
-			auto cpu_inode = MUST(ProcROInode::create_new(
-				[](off_t offset, BAN::ByteSpan buffer, void* index_ptr) -> BAN::ErrorOr<size_t>
-				{
-					ASSERT(offset >= 0);
-
-					const size_t index = reinterpret_cast<uintptr_t>(index_ptr);
-					const auto load_stats = Processor::get_load_stats(index);
-
-					auto string = TRY(BAN::String::formatted("{} {}", load_stats.ns_idle, load_stats.ns_total));
-					if (static_cast<size_t>(offset) >= string.size())
-						return 0;
-
-					const size_t bytes = BAN::Math::min<size_t>(string.size() - offset, buffer.size());
-					memcpy(buffer.data(), string.data() + offset, bytes);
-					return bytes;
-				},
-				*s_instance, reinterpret_cast<void*>(i), 0444, 0, 0
-			));
-			MUST(cpu_directory->link_inode(MUST(BAN::String::formatted("{}", i)), cpu_inode));
-		}
-	}
-
 	ProcFileSystem& ProcFileSystem::get()
 	{
 		ASSERT(s_instance);

kernel/kernel/FS/ProcFS/Inode.cpp

@@ -16,13 +16,12 @@ namespace Kernel
 
 		TRY(inode->link_inode(*inode, "."_sv));
 		TRY(inode->link_inode(static_cast<TmpInode&>(*fs.root_inode()), ".."_sv));
-		TRY(inode->link_inode(*MUST(ProcROProcessInode::create_new(process, &Process::proc_meminfo, fs, 0444)), "meminfo"_sv));
+		TRY(inode->link_inode(*MUST(ProcROProcessInode::create_new(process, &Process::proc_meminfo, fs, 0400)), "meminfo"_sv));
 		TRY(inode->link_inode(*MUST(ProcROProcessInode::create_new(process, &Process::proc_cmdline, fs, 0444)), "cmdline"_sv));
 		TRY(inode->link_inode(*MUST(ProcROProcessInode::create_new(process, &Process::proc_environ, fs, 0400)), "environ"_sv));
-		TRY(inode->link_inode(*MUST(ProcROProcessInode::create_new(process, &Process::proc_cputime, fs, 0444)), "cputime"_sv));
 		TRY(inode->link_inode(*MUST(ProcSymlinkProcessInode::create_new(process, &Process::proc_cwd, fs, 0777)), "cwd"_sv));
 		TRY(inode->link_inode(*MUST(ProcSymlinkProcessInode::create_new(process, &Process::proc_exe, fs, 0777)), "exe"_sv));
-		TRY(inode->link_inode(*MUST(ProcFDDirectoryInode::create_new(process, fs, 0500)), "fd"_sv));
+		TRY(inode->link_inode(*MUST(ProcFDDirectoryInode::create_new(process, fs, 0777)), "fd"_sv));
 
 		return inode;
 	}
@@ -39,7 +38,6 @@ namespace Kernel
 		(void)TmpDirectoryInode::unlink_impl("meminfo"_sv);
 		(void)TmpDirectoryInode::unlink_impl("cmdline"_sv);
 		(void)TmpDirectoryInode::unlink_impl("environ"_sv);
-		(void)TmpDirectoryInode::unlink_impl("cputime"_sv);
 		(void)TmpDirectoryInode::unlink_impl("cwd"_sv);
 		(void)TmpDirectoryInode::unlink_impl("exe"_sv);
 		(void)TmpDirectoryInode::unlink_impl("fd"_sv);
@@ -97,20 +95,19 @@ namespace Kernel
 		return (m_process.*m_callback)();
 	}
 
-	BAN::ErrorOr<BAN::RefPtr<ProcROInode>> ProcROInode::create_new(BAN::ErrorOr<size_t> (*callback)(off_t, BAN::ByteSpan, void*), TmpFileSystem& fs, void* argument, mode_t mode, uid_t uid, gid_t gid)
+	BAN::ErrorOr<BAN::RefPtr<ProcROInode>> ProcROInode::create_new(size_t (*callback)(off_t, BAN::ByteSpan), TmpFileSystem& fs, mode_t mode, uid_t uid, gid_t gid)
 	{
 		auto inode_info = create_inode_info(Mode::IFREG | mode, uid, gid);
 
-		auto* inode_ptr = new ProcROInode(callback, fs, argument, inode_info);
+		auto* inode_ptr = new ProcROInode(callback, fs, inode_info);
 		if (inode_ptr == nullptr)
 			return BAN::Error::from_errno(ENOMEM);
 		return BAN::RefPtr<ProcROInode>::adopt(inode_ptr);
 	}
 
-	ProcROInode::ProcROInode(BAN::ErrorOr<size_t> (*callback)(off_t, BAN::ByteSpan, void*), TmpFileSystem& fs, void* argument, const TmpInodeInfo& inode_info)
+	ProcROInode::ProcROInode(size_t (*callback)(off_t, BAN::ByteSpan), TmpFileSystem& fs, const TmpInodeInfo& inode_info)
 		: TmpInode(fs, MUST(fs.allocate_inode(inode_info)), inode_info)
 		, m_callback(callback)
-		, m_argument(argument)
 	{
 		m_mode |= Inode::Mode::IFREG;
 	}
@@ -119,7 +116,7 @@ namespace Kernel
 	{
 		if (offset < 0)
 			return BAN::Error::from_errno(EINVAL);
-		return TRY(m_callback(offset, buffer, m_argument));
+		return m_callback(offset, buffer);
 	}
 
 	BAN::ErrorOr<BAN::RefPtr<ProcSymlinkInode>> ProcSymlinkInode::create_new(BAN::ErrorOr<BAN::String> (*callback)(void*), void (*destructor)(void*), void* data, TmpFileSystem& fs, mode_t mode, uid_t uid, gid_t gid)

kernel/kernel/FS/VirtualFileSystem.cpp

@@ -157,7 +157,7 @@ namespace Kernel
 			if (i == 4)
 				dwarnln("Could not find specified root device, waiting for it to get loaded...");
 
-			SystemTimer::get().sleep_for_ms(sleep_ms);
+			SystemTimer::get().sleep_ms(sleep_ms);
 		}
 
 		derrorln("Could not find root device '{}' after {} ms", root_path, timeout_ms);
@@ -210,14 +210,14 @@ namespace Kernel
 		if (!file.inode->mode().ifdir())
 			return BAN::Error::from_errno(ENOTDIR);
 
-		LockGuard _(m_mount_point_lock);
+		LockGuard _(m_mutex);
 		TRY(m_mount_points.emplace_back(file_system, BAN::move(file)));
 		return {};
 	}
 
 	VirtualFileSystem::MountPoint* VirtualFileSystem::mount_from_host_inode(BAN::RefPtr<Inode> inode)
 	{
-		LockGuard _(m_mount_point_lock);
+		LockGuard _(m_mutex);
 		for (MountPoint& mount : m_mount_points)
 			if (*mount.host.inode == *inode)
 				return &mount;
@@ -226,52 +226,53 @@ namespace Kernel
 
 	VirtualFileSystem::MountPoint* VirtualFileSystem::mount_from_root_inode(BAN::RefPtr<Inode> inode)
 	{
-		LockGuard _(m_mount_point_lock);
+		LockGuard _(m_mutex);
 		for (MountPoint& mount : m_mount_points)
 			if (*mount.target->root_inode() == *inode)
 				return &mount;
 		return nullptr;
 	}
 
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wstack-usage="
 	BAN::ErrorOr<VirtualFileSystem::File> VirtualFileSystem::file_from_relative_path(BAN::RefPtr<Inode> root_inode, const File& parent, const Credentials& credentials, BAN::StringView path, int flags)
 	{
+		LockGuard _(m_mutex);
+
 		auto inode = parent.inode;
 		ASSERT(inode);
 
-		char canonical_buf[PATH_MAX];
+		BAN::String canonical_path;
-		size_t canonical_len = parent.canonical_path.size();
+		TRY(canonical_path.append(parent.canonical_path));
-		if (!parent.canonical_path.empty() && parent.canonical_path.back() == '/')
+		if (!canonical_path.empty() && canonical_path.back() == '/')
-			canonical_len--;
+			canonical_path.pop_back();
-		memcpy(canonical_buf, parent.canonical_path.data(), canonical_len);
+		ASSERT(canonical_path.empty() || canonical_path.back() != '/');
-		ASSERT(canonical_len == 0 || canonical_buf[canonical_len - 1] != '/');
 
-		char path_storage[PATH_MAX];
+		BAN::Vector<BAN::String> path_parts;
+
+		const auto append_string_view_in_reverse =
+			[&path_parts](BAN::StringView path) -> BAN::ErrorOr<void>
+			{
+				auto split_path = TRY(path.split('/'));
+				TRY(path_parts.reserve(path_parts.size() + split_path.size()));
+				for (size_t i = split_path.size(); i > 0; i--)
+				{
+					TRY(path_parts.emplace_back());
+					TRY(path_parts.back().append(split_path[i - 1]));
+				}
+				return {};
+			};
+		TRY(append_string_view_in_reverse(path));
 
 		size_t link_depth = 0;
-		for (;;)
+
+		while (!path_parts.empty())
 		{
-			const auto path_part = [&path] {
+			BAN::String path_part = BAN::move(path_parts.back());
-				size_t off = 0;
+			path_parts.pop_back();
-				while (off < path.size() && path[off] == '/')
-					off++;
 
-				size_t len = 0;
+			if (path_part.empty() || path_part == "."_sv)
-				while (off + len < path.size() && path[off + len] != '/')
-					len++;
-
-				const auto result = path.substring(off, len);
-				path = path.substring(off + len);
-				return result;
-			}();
-
-			if (path_part.empty())
-				break;
-			if (path_part == "."_sv)
 				continue;
 
-			auto orig_inode = inode;
+			auto orig = inode;
 
 			// resolve file name
 			{
@@ -288,26 +289,26 @@ namespace Kernel
 
 				if (path_part == ".."_sv)
 				{
-					while (canonical_len && canonical_buf[canonical_len - 1] != '/')
+					if (!canonical_path.empty())
-						canonical_len--;
+					{
-					if (canonical_len)
+						ASSERT(canonical_path.front() == '/');
-						canonical_len--;
+						while (canonical_path.back() != '/')
+							canonical_path.pop_back();
+						canonical_path.pop_back();
+					}
 				}
 				else
 				{
 					if (auto* mount_point = mount_from_host_inode(inode))
 						inode = mount_point->target->root_inode();
-					if (canonical_len + 1 + path_part.size() > sizeof(canonical_buf))
+					TRY(canonical_path.push_back('/'));
-						return BAN::Error::from_errno(ENAMETOOLONG);
+					TRY(canonical_path.append(path_part));
-					canonical_buf[canonical_len++] = '/';
-					memcpy(canonical_buf + canonical_len, path_part.data(), path_part.size());
-					canonical_len += path_part.size();
 				}
 			}
 
 			if (!inode->mode().iflnk())
 				continue;
-			if ((flags & O_NOFOLLOW) && !path.find([](char ch) { return ch != '/'; }).has_value())
+			if ((flags & O_NOFOLLOW) && path_parts.empty())
 				continue;
 
 			// resolve symbolic links
@@ -319,26 +320,21 @@ namespace Kernel
 				if (link_target.front() == '/')
 				{
 					inode = root_inode;
-					canonical_len = 0;
+					canonical_path.clear();
 				}
 				else
 				{
-					inode = orig_inode;
+					inode = orig;
 
-					while (canonical_len && canonical_buf[canonical_len - 1] != '/')
+					while (canonical_path.back() != '/')
-						canonical_len--;
+						canonical_path.pop_back();
-					if (canonical_len)
+					canonical_path.pop_back();
-						canonical_len--;
 				}
 
-				if (link_target.size() + path.size() > sizeof(path_storage))
+				TRY(append_string_view_in_reverse(link_target.sv()));
-					return BAN::Error::from_errno(ENAMETOOLONG);
-				memcpy(path_storage, link_target.data(), link_target.size());
-				memcpy(path_storage + link_target.size(), path.data(), path.size());
-				path = BAN::StringView(path_storage, link_target.size() + path.size());
 
 				link_depth++;
-				if (link_depth > SYMLOOP_MAX)
+				if (link_depth > 100)
 					return BAN::Error::from_errno(ELOOP);
 			}
 		}
@@ -346,14 +342,17 @@ namespace Kernel
 		if (!inode->can_access(credentials, flags))
 			return BAN::Error::from_errno(EACCES);
 
-		if (canonical_len == 0)
+		if (canonical_path.empty())
-			canonical_buf[canonical_len++] = '/';
+			TRY(canonical_path.push_back('/'));
 
-		BAN::String canonical_path;
+		File file;
-		TRY(canonical_path.append(BAN::StringView(canonical_buf, canonical_len)));
+		file.inode = inode;
+		file.canonical_path = BAN::move(canonical_path);
 
-		return File(BAN::move(inode), BAN::move(canonical_path));
+		if (file.canonical_path.empty())
+			TRY(file.canonical_path.push_back('/'));
+
+		return file;
 	}
-#pragma GCC diagnostic pop
 
 }

kernel/kernel/IDT.cpp

@@ -24,6 +24,11 @@ namespace Kernel
 #if ARCH(x86_64)
 	struct Registers
 	{
+		uint64_t cr4;
+		uint64_t cr3;
+		uint64_t cr2;
+		uint64_t cr0;
+
 		uint64_t r15;
 		uint64_t r14;
 		uint64_t r13;
@@ -44,6 +49,11 @@ namespace Kernel
 #elif ARCH(i686)
 	struct Registers
 	{
+		uint32_t cr4;
+		uint32_t cr3;
+		uint32_t cr2;
+		uint32_t cr0;
+
 		uint32_t edi;
 		uint32_t esi;
 		uint32_t ebp;
@@ -216,11 +226,8 @@ namespace Kernel
 					return;
 				}
 
-				uintptr_t cr2;
-				asm volatile("mov %%cr2, %0" : "=r"(cr2));
-
 				Processor::set_interrupt_state(InterruptState::Enabled);
-				auto result = Process::current().allocate_page_for_demand_paging(cr2, page_fault_error.write, page_fault_error.instruction);
+				auto result = Process::current().allocate_page_for_demand_paging(regs->cr2, page_fault_error.write, page_fault_error.instruction);
 				Processor::set_interrupt_state(InterruptState::Disabled);
 
 				if (result.is_error())
@@ -249,18 +256,6 @@ namespace Kernel
 
 				break;
 			}
-			case ISR::GeneralProtectionFault:
-			{
-				const uint8_t* ip = reinterpret_cast<const uint8_t*>(interrupt_stack->ip);
-				for (const auto& safe_user : s_safe_user_page_faults)
-				{
-					if (ip < safe_user.ip_start || ip >= safe_user.ip_end)
-						continue;
-					interrupt_stack->ip = reinterpret_cast<vaddr_t>(safe_user.ip_fault);
-					return;
-				}
-				break;
-			}
 			case ISR::DeviceNotAvailable:
 			{
 				if (pid == 0 || !Thread::current().is_userspace())
@@ -279,15 +274,6 @@ namespace Kernel
 			}
 		}
 
-		uintptr_t cr0, cr2, cr3, cr4;
-		asm volatile(
-			"mov %%cr0, %0;"
-			"mov %%cr2, %1;"
-			"mov %%cr3, %2;"
-			"mov %%cr4, %3;"
-			: "=r"(cr0), "=r"(cr2), "=r"(cr3), "=r"(cr4)
-		);
-
 		Debug::s_debug_lock.lock();
 
 		if (PageTable::current().get_page_flags(interrupt_stack->ip & PAGE_ADDR_MASK) & PageTable::Flags::Present)
@@ -325,7 +311,7 @@ namespace Kernel
 			regs->r8, regs->r9, regs->r10, regs->r11,
 			regs->r12, regs->r13, regs->r14, regs->r15,
 			interrupt_stack->ip, interrupt_stack->flags,
-			cr0, cr2, cr3, cr4
+			regs->cr0, regs->cr2, regs->cr3, regs->cr4
 		);
 #elif ARCH(i686)
 		dwarnln(
@@ -339,7 +325,7 @@ namespace Kernel
 			regs->eax, regs->ebx, regs->ecx, regs->edx,
 			interrupt_stack->sp, regs->ebp, regs->edi, regs->esi,
 			interrupt_stack->ip, interrupt_stack->flags,
-			cr0, cr2, cr3, cr4
+			regs->cr0, regs->cr2, regs->cr3, regs->cr4
 		);
 #endif
 		if (isr == ISR::PageFault)
@@ -387,11 +373,11 @@ namespace Kernel
 				break;
 			case ISR::PageFault:
 				signal_info.si_signo = SIGSEGV;
-				if (PageTable::current().get_page_flags(cr2 & PAGE_ADDR_MASK) & PageTable::Flags::Present)
+				if (PageTable::current().get_page_flags(regs->cr2 & PAGE_ADDR_MASK) & PageTable::Flags::Present)
 					signal_info.si_code = SEGV_ACCERR;
 				else
 					signal_info.si_code = SEGV_MAPERR;
-				signal_info.si_addr = reinterpret_cast<void*>(cr2);
+				signal_info.si_addr = reinterpret_cast<void*>(regs->cr2);
 				break;
 			default:
 				dwarnln("Unhandled exception");
@@ -435,7 +421,7 @@ namespace Kernel
 		if (Processor::current_is_bsp())
 			Process::update_alarm_queue();
 
-		Processor::scheduler().on_timer_interrupt();
+		Processor::scheduler().timer_interrupt();
 	}
 
 	extern "C" void cpp_irq_handler(uint32_t irq)

kernel/kernel/Input/InputDevice.cpp

@@ -197,6 +197,9 @@ namespace Kernel
 					}
 					else switch (key_event.keycode)
 					{
+						case LibInput::keycode_function(1):
+							Processor::toggle_should_print_cpu_load();
+							break;
 						case LibInput::keycode_function(11):
 							DevFileSystem::get().initiate_disk_cache_drop();
 							break;

kernel/kernel/Input/PS2/Keyboard.cpp

@@ -151,18 +151,16 @@ namespace Kernel::Input
 		uint16_t toggle_mask = 0;
 		switch (dummy_event.key)
 		{
-			case Key::LeftShift:  modifier_mask = KeyModifier::LShift; break;
+			case Key::LeftShift:	modifier_mask = KeyModifier::LShift;	break;
-			case Key::RightShift: modifier_mask = KeyModifier::RShift; break;
+			case Key::RightShift:	modifier_mask = KeyModifier::RShift;	break;
-			case Key::LeftCtrl:   modifier_mask = KeyModifier::LCtrl;  break;
+			case Key::LeftCtrl:		modifier_mask = KeyModifier::LCtrl;		break;
-			case Key::RightCtrl:  modifier_mask = KeyModifier::RCtrl;  break;
+			case Key::RightCtrl:	modifier_mask = KeyModifier::RCtrl;		break;
-			case Key::LeftAlt:    modifier_mask = KeyModifier::LAlt;   break;
+			case Key::LeftAlt:		modifier_mask = KeyModifier::LAlt;		break;
-			case Key::RightAlt:   modifier_mask = KeyModifier::RAlt;   break;
+			case Key::RightAlt:		modifier_mask = KeyModifier::RAlt;		break;
-			case Key::LeftSuper:  modifier_mask = KeyModifier::LSuper; break;
-			case Key::RightSuper: modifier_mask = KeyModifier::RSuper; break;
 
-			case Key::ScrollLock: toggle_mask = KeyModifier::ScrollLock; break;
+			case Key::ScrollLock:	toggle_mask = KeyModifier::ScrollLock;	break;
-			case Key::NumLock:    toggle_mask = KeyModifier::NumLock;    break;
+			case Key::NumLock:		toggle_mask = KeyModifier::NumLock;		break;
-			case Key::CapsLock:   toggle_mask = KeyModifier::CapsLock;   break;
+			case Key::CapsLock:		toggle_mask = KeyModifier::CapsLock;	break;
 
 			default: break;
 		}

kernel/kernel/Input/PS2/Keymap.cpp

@@ -95,9 +95,7 @@ namespace Kernel::Input
 		m_scancode_to_keycode_normal[0x38]   = keycode_normal(4,  2);
 		m_scancode_to_keycode_normal[0x39]   = keycode_normal(4,  3);
 		m_scancode_to_keycode_extended[0x38] = keycode_normal(4,  4);
-		m_scancode_to_keycode_extended[0x5C] = keycode_normal(4,  5);
+		m_scancode_to_keycode_extended[0x1D] = keycode_normal(4,  5);
-		m_scancode_to_keycode_extended[0x5D] = keycode_normal(4,  6);
-		m_scancode_to_keycode_extended[0x1D] = keycode_normal(4,  7);
 
 		m_scancode_to_keycode_normal[0x45]   = keycode_numpad(0, 0);
 		m_scancode_to_keycode_extended[0x35] = keycode_numpad(0, 1);
@@ -210,9 +208,7 @@ namespace Kernel::Input
 		m_scancode_to_keycode_normal[0x11]   = keycode_normal(4,  2);
 		m_scancode_to_keycode_normal[0x29]   = keycode_normal(4,  3);
 		m_scancode_to_keycode_extended[0x11] = keycode_normal(4,  4);
-		m_scancode_to_keycode_extended[0x27] = keycode_normal(4,  5);
+		m_scancode_to_keycode_extended[0x14] = keycode_normal(4,  5);
-		m_scancode_to_keycode_extended[0x2F] = keycode_normal(4,  6);
-		m_scancode_to_keycode_extended[0x14] = keycode_normal(4,  7);
 
 		m_scancode_to_keycode_normal[0x77]   = keycode_numpad(0, 0);
 		m_scancode_to_keycode_extended[0x4A] = keycode_numpad(0, 1);
@@ -325,9 +321,7 @@ namespace Kernel::Input
 		m_scancode_to_keycode_normal[0x19] = keycode_normal(4,  2);
 		m_scancode_to_keycode_normal[0x29] = keycode_normal(4,  3);
 		m_scancode_to_keycode_normal[0x39] = keycode_normal(4,  4);
-		m_scancode_to_keycode_normal[0x8C] = keycode_normal(4,  5);
+		m_scancode_to_keycode_normal[0x58] = keycode_normal(4,  5);
-		m_scancode_to_keycode_normal[0x8D] = keycode_normal(4,  6);
-		m_scancode_to_keycode_normal[0x58] = keycode_normal(4,  7);
 
 		m_scancode_to_keycode_normal[0x76] = keycode_numpad(0, 0);
 		m_scancode_to_keycode_normal[0x4A] = keycode_numpad(0, 1);

kernel/kernel/Memory/FileBackedRegion.cpp

@@ -93,7 +93,7 @@ namespace Kernel
 
 	void SharedFileData::sync_no_lock(size_t page_index)
 	{
-		if (pages[page_index] == 0 || BAN::atomic_load(writers[page_index]) > 0)
+		if (pages[page_index] == 0 || BAN::atomic_load(writers[page_index]) == 0)
 			return;
 
 		uint8_t page_buffer[PAGE_SIZE];
@@ -108,9 +108,8 @@ namespace Kernel
 
 	BAN::ErrorOr<void> FileBackedRegion::msync(vaddr_t address, size_t size, int flags)
 	{
-		// TODO: maybe do something with the flags :)
+		if (flags != MS_SYNC)
-		(void)flags;
+			dprintln("async file backed mmap msync");
-
 		if (m_type != Type::SHARED)
 			return {};
 

kernel/kernel/Memory/SharedMemoryObject.cpp

@@ -112,8 +112,8 @@ namespace Kernel
 			paddr = Heap::get().take_free_page();
 			if (paddr == 0)
 				return BAN::Error::from_errno(ENOMEM);
-			PageTable::with_per_cpu_fast_page(paddr, [&](void* addr) {
+			PageTable::with_fast_page(paddr, [&] {
-				memset(addr, 0x00, PAGE_SIZE);
+				memset(PageTable::fast_page_as_ptr(), 0x00, PAGE_SIZE);
 			});
 			m_object->paddrs[(vaddr - m_vaddr) / PAGE_SIZE] = paddr;
 		}

kernel/kernel/Networking/ARPTable.cpp

@@ -12,8 +12,8 @@ namespace Kernel
 		Reply = 2,
 	};
 
-	static constexpr BAN::IPv4Address s_broadcast_ipv4 { 0xFFFFFFFF };
+	static constexpr BAN::IPv4Address	s_broadcast_ipv4 { 0xFFFFFFFF };
-	static constexpr BAN::MACAddress  s_broadcast_mac  {{ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF }};
+	static constexpr BAN::MACAddress	s_broadcast_mac {{ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF }};
 
 	BAN::ErrorOr<BAN::UniqPtr<ARPTable>> ARPTable::create()
 	{
@@ -50,19 +50,18 @@ namespace Kernel
 				return it->value;
 		}
 
-		const auto arp_request = ARPPacket {
+		ARPPacket arp_request;
-			.htype = 0x0001,
+		arp_request.htype = 0x0001;
-			.ptype = EtherType::IPv4,
+		arp_request.ptype = EtherType::IPv4;
-			.hlen = 0x06,
+		arp_request.hlen = 0x06;
-			.plen = 0x04,
+		arp_request.plen = 0x04;
-			.oper = ARPOperation::Request,
+		arp_request.oper = ARPOperation::Request;
-			.sha = interface.get_mac_address(),
+		arp_request.sha = interface.get_mac_address();
-			.spa = interface.get_ipv4_address(),
+		arp_request.spa = interface.get_ipv4_address();
-			.tha = {{ 0, 0, 0, 0, 0, 0 }},
+		arp_request.tha = {{ 0, 0, 0, 0, 0, 0 }};
-			.tpa = ipv4_address,
+		arp_request.tpa = ipv4_address;
-		};
 
-		TRY(interface.send_with_ethernet_header(s_broadcast_mac, EtherType::ARP, BAN::ConstByteSpan::from(arp_request)));
+		TRY(interface.send_bytes(s_broadcast_mac, EtherType::ARP, BAN::ConstByteSpan::from(arp_request)));
 
 		uint64_t timeout = SystemTimer::get().ms_since_boot() + 1'000;
 		while (SystemTimer::get().ms_since_boot() < timeout)
@@ -101,7 +100,7 @@ namespace Kernel
 			{
 				if (packet.tpa == interface.get_ipv4_address())
 				{
-					const auto arp_reply = ARPPacket {
+					const ARPPacket arp_reply {
 						.htype = 0x0001,
 						.ptype = EtherType::IPv4,
 						.hlen = 0x06,
@@ -112,7 +111,7 @@ namespace Kernel
 						.tha = packet.sha,
 						.tpa = packet.spa,
 					};
-					TRY(interface.send_with_ethernet_header(packet.sha, EtherType::ARP, BAN::ConstByteSpan::from(arp_reply)));
+					TRY(interface.send_bytes(packet.sha, EtherType::ARP, BAN::ConstByteSpan::from(arp_reply)));
 				}
 				break;
 			}

kernel/kernel/Networking/E1000/E1000.cpp

@@ -202,8 +202,6 @@ namespace Kernel
 		rctrl |= RCTL_BSIZE_8192;
 		write32(REG_RCTL, rctrl);
 
-		write32(REG_RXCSUM, RXSUM_IPOFLD | RXSUM_TUOFLD);
-
 		return {};
 	}
 
@@ -281,13 +279,9 @@ namespace Kernel
 		return {};
 	}
 
-	BAN::ErrorOr<void> E1000::send_raw_bytes(BAN::Span<const BAN::ConstByteSpan> buffers)
+	BAN::ErrorOr<void> E1000::send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::Span<const BAN::ConstByteSpan> payload)
 	{
-		const auto interrupt_state = Processor::get_interrupt_state();
+		const uint32_t tx_current = m_tx_head1.fetch_add(1) % E1000_TX_DESCRIPTOR_COUNT;
-		Processor::set_interrupt_state(InterruptState::Disabled);
-
-		const uint32_t tx_current_nowrap = m_tx_head.fetch_add(1);
-		const uint32_t tx_current = tx_current_nowrap % E1000_TX_DESCRIPTOR_COUNT;
 
 		auto& descriptor = reinterpret_cast<volatile e1000_tx_desc*>(m_tx_descriptor_region->vaddr())[tx_current];
 		while (descriptor.status == 0)
@@ -295,10 +289,15 @@ namespace Kernel
 
 		auto* tx_buffer = reinterpret_cast<uint8_t*>(m_tx_buffer_region->vaddr() + E1000_TX_BUFFER_SIZE * tx_current);
 
-		size_t packet_size = 0;
+		auto& ethernet_header = *reinterpret_cast<EthernetHeader*>(tx_buffer);
-		for (const auto& buffer : buffers)
+		ethernet_header.dst_mac = destination;
+		ethernet_header.src_mac = get_mac_address();
+		ethernet_header.ether_type = protocol;
+
+		size_t packet_size = sizeof(EthernetHeader);
+		for (const auto& buffer : payload)
 		{
-			ASSERT(packet_size + buffer.size() <= E1000_TX_BUFFER_SIZE);
+			ASSERT(packet_size + buffer.size() < E1000_TX_BUFFER_SIZE);
 			memcpy(tx_buffer + packet_size, buffer.data(), buffer.size());
 			packet_size += buffer.size();
 		}
@@ -307,12 +306,8 @@ namespace Kernel
 		descriptor.status = 0;
 		descriptor.cmd = CMD_EOP | CMD_IFCS | CMD_RS;
 
-		while (tx_current_nowrap != m_tx_commit.load())
+		if (tx_current == m_tx_head2.fetch_add(1) % E1000_TX_DESCRIPTOR_COUNT)
-			Processor::pause();
+			write32(REG_TDT, (tx_current + 1) % E1000_TX_DESCRIPTOR_COUNT);
-		write32(REG_TDT, (tx_current + 1) % E1000_TX_DESCRIPTOR_COUNT);
-		m_tx_commit.add_fetch(1);
-
-		Processor::set_interrupt_state(interrupt_state);
 
 		dprintln_if(DEBUG_E1000, "sent {} bytes", packet_size);
 
@@ -321,66 +316,35 @@ namespace Kernel
 
 	void E1000::receive_thread()
 	{
-		SpinLockGuard guard(m_rx_lock);
+		SpinLockGuard _(m_rx_lock);
-
-		uint32_t rx_current = 0;
 
 		while (!m_thread_should_die)
 		{
-			uint32_t rx_tail = rx_current;
-
 			for (;;)
 			{
+				const uint32_t rx_current = (read32(REG_RDT0) + 1) % E1000_RX_DESCRIPTOR_COUNT;
+
 				auto& descriptor = reinterpret_cast<volatile e1000_rx_desc*>(m_rx_descriptor_region->vaddr())[rx_current];
-
+				if (!(descriptor.status & 1))
-				const auto status = descriptor.status;
-				if (!(status & RX_STS_DD))
 					break;
+				ASSERT(descriptor.length <= E1000_RX_BUFFER_SIZE);
 
-				const auto errors = descriptor.errors;
+				dprintln_if(DEBUG_E1000, "got {} bytes", (uint16_t)descriptor.length);
-				if (!(status & RX_STS_EOP))
-					dwarnln("multi descriptor packet??");
-				else if (errors & (RX_ERR_CE | RX_ERR_SE | RX_ERR_RXE))
-					dwarnln("descriptor error {2h}", errors);
-				else if ((status & RX_STS_IPCS) && (errors & RX_ERR_IPE))
-					dwarnln("IPv4 checksum error");
-				else if ((status & RX_STS_TCPCS) && (errors & RX_ERR_TCPE))
-					dwarnln("TCP checkum error");
-				else if ((status & RX_STS_UDPCS) && (errors & RX_ERR_TCPE))
-					dwarnln("UDP checksum error");
-				else
-				{
-					m_rx_lock.unlock(InterruptState::Enabled);
 
-					const uint32_t packet_length = descriptor.length;
+				m_rx_lock.unlock(InterruptState::Enabled);
-					ASSERT(packet_length <= E1000_RX_BUFFER_SIZE);
 
-					dprintln_if(DEBUG_E1000, "got {} bytes", packet_length);
+				NetworkManager::get().on_receive(*this, BAN::ConstByteSpan {
+					reinterpret_cast<const uint8_t*>(m_rx_buffer_region->vaddr() + rx_current * E1000_RX_BUFFER_SIZE),
+					descriptor.length
+				});
 
-					uint32_t validated_cksums = 0;
+				m_rx_lock.lock();
-					if ((status & RX_STS_IPCS) && !(errors & RX_ERR_IPE))
-						validated_cksums |= CKSUM_IPV4;
-					if ((status & RX_STS_TCPCS) && !(errors & RX_ERR_TCPE))
-						validated_cksums |= CKSUM_TCP;
-					if ((status & RX_STS_UDPCS) && !(errors & RX_ERR_TCPE))
-						validated_cksums |= CKSUM_UDP;
-
-					const uint8_t* packet_data = reinterpret_cast<const uint8_t*>(m_rx_buffer_region->vaddr() + rx_current * E1000_RX_BUFFER_SIZE);
-					NetworkManager::get().on_receive(*this, BAN::ConstByteSpan { packet_data, packet_length }, validated_cksums);
-
-					m_rx_lock.lock();
-				}
 
 				descriptor.status = 0;
-
+				write32(REG_RDT0, rx_current);
-				rx_tail = rx_current;
-				rx_current = (rx_current + 1) % E1000_RX_DESCRIPTOR_COUNT;
 			}
 
-			if (rx_current != rx_tail)
+			SpinLockAsMutex smutex(m_rx_lock, InterruptState::Enabled);
-				write32(REG_RDT0, rx_tail);
-
-			SpinLockGuardAsMutex smutex(guard);
 			m_rx_blocker.block_indefinite(&smutex);
 		}
 

kernel/kernel/Networking/IPv4Layer.cpp

@@ -180,7 +180,7 @@ namespace Kernel
 		return {};
 	}
 
-	BAN::ErrorOr<void> IPv4Layer::sendto(NetworkSocket& socket, BAN::ConstByteSpan payload, const sockaddr* address, socklen_t address_len)
+	BAN::ErrorOr<size_t> IPv4Layer::sendto(NetworkSocket& socket, BAN::ConstByteSpan payload, const sockaddr* address, socklen_t address_len)
 	{
 		if (address->sa_family != AF_INET)
 			return BAN::Error::from_errno(EINVAL);
@@ -224,21 +224,23 @@ namespace Kernel
 		};
 
 		uint8_t protocol_header_buffer[32];
+		ASSERT(socket.protocol_header_size() < sizeof(protocol_header_buffer));
+
 		auto protocol_header = BAN::ByteSpan::from(protocol_header_buffer).slice(0, socket.protocol_header_size());
 		socket.get_protocol_header(protocol_header, payload, dst_port, pseudo_header);
 
-		const BAN::ConstByteSpan buffers[] {
+		BAN::ConstByteSpan buffers[] {
 			BAN::ConstByteSpan::from(ipv4_header),
 			protocol_header,
 			payload,
 		};
 
-		TRY(interface->send_with_ethernet_header(dst_mac, EtherType::IPv4, buffers));
+		TRY(interface->send_bytes(dst_mac, EtherType::IPv4, { buffers, sizeof(buffers) / sizeof(*buffers) }));
 
-		return {};
+		return payload.size();
 	}
 
-	BAN::ErrorOr<void> IPv4Layer::handle_ipv4_packet(NetworkInterface& interface, BAN::ConstByteSpan packet, uint32_t validated_cksums)
+	BAN::ErrorOr<void> IPv4Layer::handle_ipv4_packet(NetworkInterface& interface, BAN::ConstByteSpan packet)
 	{
 		if (packet.size() < sizeof(IPv4Header))
 		{
@@ -247,7 +249,7 @@ namespace Kernel
 		}
 
 		auto& ipv4_header = packet.as<const IPv4Header>();
-		if (!(validated_cksums & CKSUM_IPV4) && calculate_internet_checksum(BAN::ConstByteSpan::from(ipv4_header)) != 0)
+		if (calculate_internet_checksum(BAN::ConstByteSpan::from(ipv4_header)) != 0)
 		{
 			dwarnln_if(DEBUG_IPV4, "IPv4 packet checksum failed");
 			return {};
@@ -282,33 +284,34 @@ namespace Kernel
 				{
 					case ICMPType::EchoRequest:
 					{
-						const auto dst_mac = TRY(m_arp_table->get_mac_from_ipv4(interface, src_ipv4));
+						auto dst_mac = TRY(m_arp_table->get_mac_from_ipv4(interface, src_ipv4));
 
-						const auto send_ipv4_header = get_ipv4_header(
+						auto send_ipv4_header = get_ipv4_header(
 							ipv4_data.size(),
 							interface.get_ipv4_address(),
 							src_ipv4,
 							NetworkProtocol::ICMP
 						);
 
-						auto send_icmp_header = ICMPHeader {
+						ICMPHeader send_icmp_header {
 							.type = ICMPType::EchoReply,
 							.code = icmp_header.code,
 							.checksum = 0,
 							.rest = icmp_header.rest,
 						};
 
+						auto send_payload = ipv4_data.slice(sizeof(ICMPHeader));
+
 						const BAN::ConstByteSpan send_buffers[] {
 							BAN::ConstByteSpan::from(send_ipv4_header),
 							BAN::ConstByteSpan::from(send_icmp_header),
-							ipv4_data.slice(sizeof(ICMPHeader))
+							send_payload
 						};
+						auto send_buffers_span = BAN::Span { send_buffers, sizeof(send_buffers) / sizeof(*send_buffers) };
 
-						send_icmp_header.checksum = calculate_internet_checksum({
+						send_icmp_header.checksum = calculate_internet_checksum(send_buffers_span.slice(1));
-							send_buffers + 1, sizeof(send_buffers) / sizeof(*send_buffers) - 1
-						});
 
-						TRY(interface.send_with_ethernet_header(dst_mac, EtherType::IPv4, send_buffers));
+						TRY(interface.send_bytes(dst_mac, EtherType::IPv4, send_buffers_span));
 
 						break;
 					}
@@ -386,7 +389,7 @@ namespace Kernel
 		sender.sin_family = AF_INET;
 		sender.sin_port = BAN::host_to_network_endian(src_port);
 		sender.sin_addr.s_addr = src_ipv4.raw;
-		bound_socket->receive_packet(ipv4_data, reinterpret_cast<const sockaddr*>(&sender), sizeof(sender), validated_cksums);
+		bound_socket->receive_packet(ipv4_data, reinterpret_cast<const sockaddr*>(&sender), sizeof(sender));
 
 		return {};
 	}

kernel/kernel/Networking/Loopback.cpp

@@ -1,4 +1,4 @@
-#include <kernel/Lock/SpinLockAsMutex.h>
+#include <kernel/Lock/LockGuard.h>
 #include <kernel/Networking/Loopback.h>
 #include <kernel/Networking/NetworkManager.h>
 
@@ -54,15 +54,12 @@ namespace Kernel
 			Processor::yield();
 	}
 
-	BAN::ErrorOr<void> LoopbackInterface::send_raw_bytes(BAN::Span<const BAN::ConstByteSpan> buffers)
+	BAN::ErrorOr<void> LoopbackInterface::send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::Span<const BAN::ConstByteSpan> payload)
 	{
-		const auto interrupt_state = Processor::get_interrupt_state();
-		Processor::set_interrupt_state(InterruptState::Disabled);
-
 		auto& descriptor =
 			[&]() -> Descriptor&
 			{
-				SpinLockGuard guard(m_buffer_lock);
+				LockGuard _(m_buffer_lock);
 				for (;;)
 				{
 					auto& descriptor = m_descriptors[m_buffer_head];
@@ -72,31 +69,34 @@ namespace Kernel
 						descriptor.state = 1;
 						return descriptor;
 					}
-					SpinLockGuardAsMutex smutex(guard);
+					m_thread_blocker.block_indefinite(&m_buffer_lock);
-					m_thread_blocker.block_indefinite(&smutex);
 				}
 			}();
 
-		size_t packet_size = 0;
+		auto& ethernet_header = *reinterpret_cast<EthernetHeader*>(descriptor.addr);
-		for (const auto& buffer : buffers)
+		ethernet_header.dst_mac = destination;
+		ethernet_header.src_mac = get_mac_address();
+		ethernet_header.ether_type = protocol;
+
+		size_t packet_size = sizeof(EthernetHeader);
+		for (const auto& buffer : payload)
 		{
 			ASSERT(packet_size + buffer.size() <= buffer_size);
 			memcpy(descriptor.addr + packet_size, buffer.data(), buffer.size());
 			packet_size += buffer.size();
 		}
 
-		m_buffer_lock.lock();
+		LockGuard _(m_buffer_lock);
 		descriptor.size = packet_size;
 		descriptor.state = 2;
 		m_thread_blocker.unblock();
-		m_buffer_lock.unlock(interrupt_state);
 
 		return {};
 	}
 
 	void LoopbackInterface::receive_thread()
 	{
-		SpinLockGuard guard(m_buffer_lock);
+		LockGuard _(m_buffer_lock);
 
 		while (!m_thread_should_die)
 		{
@@ -107,9 +107,12 @@ namespace Kernel
 					break;
 				m_buffer_tail = (m_buffer_tail + 1) % buffer_count;
 
-				m_buffer_lock.unlock(InterruptState::Enabled);
+				m_buffer_lock.unlock();
 
-				NetworkManager::get().on_receive(*this, BAN::ConstByteSpan { descriptor.addr, descriptor.size }, 0);
+				NetworkManager::get().on_receive(*this, {
+					descriptor.addr,
+					descriptor.size,
+				});
 
 				m_buffer_lock.lock();
 
@@ -118,8 +121,7 @@ namespace Kernel
 				m_thread_blocker.unblock();
 			}
 
-			SpinLockGuardAsMutex smutex(guard);
+			m_thread_blocker.block_indefinite(&m_buffer_lock);
-			m_thread_blocker.block_indefinite(&smutex);
 		}
 
 		m_thread_is_dead = true;

kernel/kernel/Networking/NetworkInterface.cpp

@@ -3,8 +3,6 @@
 #include <kernel/FS/DevFS/FileSystem.h>
 #include <kernel/Networking/NetworkInterface.h>
 
-#include <netinet/in.h>
-#include <net/if.h>
 #include <sys/sysmacros.h>
 #include <string.h>
 
@@ -27,7 +25,7 @@ namespace Kernel
 	}
 
 	NetworkInterface::NetworkInterface(Type type)
-		: CharacterDevice(0664, 0, 0)
+		: CharacterDevice(0400, 0, 0)
 		, m_type(type)
 	{
 		m_rdev = get_rdev(type);
@@ -45,95 +43,4 @@ namespace Kernel
 		}
 	}
 
-	BAN::ErrorOr<long> NetworkInterface::ioctl_impl(unsigned long request, void* arg)
-	{
-		if (arg == nullptr)
-		{
-			dprintln("No argument provided");
-			return BAN::Error::from_errno(EINVAL);
-		}
-
-		auto* ifreq = reinterpret_cast<struct ifreq*>(arg);
-
-		switch (request)
-		{
-			case SIOCGIFADDR:
-			{
-				auto& ifru_addr = *reinterpret_cast<sockaddr_in*>(&ifreq->ifr_ifru.ifru_addr);
-				ifru_addr.sin_family = AF_INET;
-				ifru_addr.sin_addr.s_addr = get_ipv4_address().raw;
-				return 0;
-			}
-			case SIOCSIFADDR:
-			{
-				auto& ifru_addr = *reinterpret_cast<const sockaddr_in*>(&ifreq->ifr_ifru.ifru_addr);
-				if (ifru_addr.sin_family != AF_INET)
-					return BAN::Error::from_errno(EADDRNOTAVAIL);
-				set_ipv4_address(BAN::IPv4Address { ifru_addr.sin_addr.s_addr });
-				dprintln("IPv4 address set to {}", get_ipv4_address());
-				return 0;
-			}
-			case SIOCGIFNETMASK:
-			{
-				auto& ifru_netmask = *reinterpret_cast<sockaddr_in*>(&ifreq->ifr_ifru.ifru_netmask);
-				ifru_netmask.sin_family = AF_INET;
-				ifru_netmask.sin_addr.s_addr = get_netmask().raw;
-				return 0;
-			}
-			case SIOCSIFNETMASK:
-			{
-				auto& ifru_netmask = *reinterpret_cast<const sockaddr_in*>(&ifreq->ifr_ifru.ifru_netmask);
-				if (ifru_netmask.sin_family != AF_INET)
-					return BAN::Error::from_errno(EADDRNOTAVAIL);
-				set_netmask(BAN::IPv4Address { ifru_netmask.sin_addr.s_addr });
-				dprintln("Netmask set to {}", get_netmask());
-				return 0;
-			}
-			case SIOCGIFGWADDR:
-			{
-				auto& ifru_gwaddr = *reinterpret_cast<sockaddr_in*>(&ifreq->ifr_ifru.ifru_gwaddr);
-				ifru_gwaddr.sin_family = AF_INET;
-				ifru_gwaddr.sin_addr.s_addr = get_gateway().raw;
-				return 0;
-			}
-			case SIOCSIFGWADDR:
-			{
-				auto& ifru_gwaddr = *reinterpret_cast<const sockaddr_in*>(&ifreq->ifr_ifru.ifru_gwaddr);
-				if (ifru_gwaddr.sin_family != AF_INET)
-					return BAN::Error::from_errno(EADDRNOTAVAIL);
-				set_gateway(BAN::IPv4Address { ifru_gwaddr.sin_addr.s_addr });
-				dprintln("Gateway set to {}", get_gateway());
-				return 0;
-			}
-			case SIOCGIFHWADDR:
-			{
-				auto mac_address = get_mac_address();
-				ifreq->ifr_ifru.ifru_hwaddr.sa_family = AF_INET;
-				memcpy(ifreq->ifr_ifru.ifru_hwaddr.sa_data, &mac_address, sizeof(mac_address));
-				return 0;
-			}
-			case SIOCGIFNAME:
-			{
-				auto& ifrn_name = ifreq->ifr_ifrn.ifrn_name;
-				ASSERT(name().size() < sizeof(ifrn_name));
-				memcpy(ifrn_name, name().data(), name().size());
-				ifrn_name[name().size()] = '\0';
-				return 0;
-			}
-			case SIOCGIFFLAGS:
-			{
-				int flags = 0;
-				if (link_up())
-					flags |= IFF_UP | IFF_RUNNING;
-				if (m_type == Type::Loopback)
-					flags |= IFF_LOOPBACK;
-				ifreq->ifr_ifru.ifru_flags = flags;
-				return 0;
-			}
-		}
-
-		return CharacterDevice::ioctl_impl(request, arg);
-	}
-
-
 }

kernel/kernel/Networking/NetworkManager.cpp

@@ -148,7 +148,7 @@ namespace Kernel
 		return {};
 	}
 
-	void NetworkManager::on_receive(NetworkInterface& interface, BAN::ConstByteSpan packet, uint32_t validated_cksums)
+	void NetworkManager::on_receive(NetworkInterface& interface, BAN::ConstByteSpan packet)
 	{
 		if (packet.size() < sizeof(EthernetHeader))
 			return;
@@ -163,7 +163,7 @@ namespace Kernel
 					dwarnln("ARP: {}", ret.error());
 				break;
 			case EtherType::IPv4:
-				if (auto ret = m_ipv4_layer->handle_ipv4_packet(interface, packet_data, validated_cksums); ret.is_error())
+				if (auto ret = m_ipv4_layer->handle_ipv4_packet(interface, packet_data); ret.is_error())
 					dwarnln("IPv4; {}", ret.error());
 				break;
 			default:

kernel/kernel/Networking/NetworkSocket.cpp

@@ -108,22 +108,85 @@ namespace Kernel
 		m_address_len = addr_len;
 	}
 
-	BAN::ErrorOr<long> NetworkSocket::ioctl_impl(unsigned long request, void* arg)
+	BAN::ErrorOr<long> NetworkSocket::ioctl_impl(int request, void* arg)
 	{
+		if (!arg)
+		{
+			dprintln("No argument provided");
+			return BAN::Error::from_errno(EINVAL);
+		}
+
+		auto interface = TRY(this->interface(nullptr, 0));
+		auto* ifreq = reinterpret_cast<struct ifreq*>(arg);
+
 		switch (request)
 		{
 			case SIOCGIFADDR:
+			{
+				auto& ifru_addr = *reinterpret_cast<sockaddr_in*>(&ifreq->ifr_ifru.ifru_addr);
+				ifru_addr.sin_family = AF_INET;
+				ifru_addr.sin_addr.s_addr = interface->get_ipv4_address().raw;
+				return 0;
+			}
 			case SIOCSIFADDR:
+			{
+				auto& ifru_addr = *reinterpret_cast<const sockaddr_in*>(&ifreq->ifr_ifru.ifru_addr);
+				if (ifru_addr.sin_family != AF_INET)
+					return BAN::Error::from_errno(EADDRNOTAVAIL);
+				interface->set_ipv4_address(BAN::IPv4Address { ifru_addr.sin_addr.s_addr });
+				dprintln("IPv4 address set to {}", interface->get_ipv4_address());
+				return 0;
+			}
 			case SIOCGIFNETMASK:
+			{
+				auto& ifru_netmask = *reinterpret_cast<sockaddr_in*>(&ifreq->ifr_ifru.ifru_netmask);
+				ifru_netmask.sin_family = AF_INET;
+				ifru_netmask.sin_addr.s_addr = interface->get_netmask().raw;
+				return 0;
+			}
 			case SIOCSIFNETMASK:
+			{
+				auto& ifru_netmask = *reinterpret_cast<const sockaddr_in*>(&ifreq->ifr_ifru.ifru_netmask);
+				if (ifru_netmask.sin_family != AF_INET)
+					return BAN::Error::from_errno(EADDRNOTAVAIL);
+				interface->set_netmask(BAN::IPv4Address { ifru_netmask.sin_addr.s_addr });
+				dprintln("Netmask set to {}", interface->get_netmask());
+				return 0;
+			}
 			case SIOCGIFGWADDR:
+			{
+				auto& ifru_gwaddr = *reinterpret_cast<sockaddr_in*>(&ifreq->ifr_ifru.ifru_gwaddr);
+				ifru_gwaddr.sin_family = AF_INET;
+				ifru_gwaddr.sin_addr.s_addr = interface->get_gateway().raw;
+				return 0;
+			}
 			case SIOCSIFGWADDR:
+			{
+				auto& ifru_gwaddr = *reinterpret_cast<const sockaddr_in*>(&ifreq->ifr_ifru.ifru_gwaddr);
+				if (ifru_gwaddr.sin_family != AF_INET)
+					return BAN::Error::from_errno(EADDRNOTAVAIL);
+				interface->set_gateway(BAN::IPv4Address { ifru_gwaddr.sin_addr.s_addr });
+				dprintln("Gateway set to {}", interface->get_gateway());
+				return 0;
+			}
 			case SIOCGIFHWADDR:
+			{
+				auto mac_address = interface->get_mac_address();
+				ifreq->ifr_ifru.ifru_hwaddr.sa_family = AF_INET;
+				memcpy(ifreq->ifr_ifru.ifru_hwaddr.sa_data, &mac_address, sizeof(mac_address));
+				return 0;
+			}
 			case SIOCGIFNAME:
-				return TRY(interface(nullptr, 0))->ioctl(request, arg);
+			{
+				auto& ifrn_name = ifreq->ifr_ifrn.ifrn_name;
+				ASSERT(interface->name().size() < sizeof(ifrn_name));
+				memcpy(ifrn_name, interface->name().data(), interface->name().size());
+				ifrn_name[interface->name().size()] = '\0';
+				return 0;
+			}
+			default:
+				return BAN::Error::from_errno(EINVAL);
 		}
-
-		return Socket::ioctl_impl(request, arg);
 	}
 
 	BAN::ErrorOr<void> NetworkSocket::getsockname_impl(sockaddr* address, socklen_t* address_len)

kernel/kernel/Networking/RTL8169/RTL8169.cpp

@@ -47,29 +47,30 @@ namespace Kernel
 
 		TRY(reset());
 
-		// disable rx, tx, checksum offload
-		m_io_bar_region->write16(RTL8169_IO_CPlusCR, 0x0000);
-		m_io_bar_region->write8 (RTL8169_IO_CR,      0x00);
-
 		dprintln("  reset done");
 
 		for (size_t i = 0; i < 6; i++)
 			m_mac_address.address[i] = m_io_bar_region->read8(RTL8169_IO_IDR0 + i);
 		dprintln("  MAC {}", m_mac_address);
 
+		// unlock config registers
+		m_io_bar_region->write8(RTL8169_IO_9346CR, RTL8169_9346CR_MODE_CONFIG);
+
 		TRY(initialize_rx());
 		TRY(initialize_tx());
 		m_io_bar_region->write8(RTL8169_IO_CR, RTL8169_CR_RE | RTL8169_CR_TE);
 		dprintln("  descriptors initialized");
 
 		m_link_up = m_io_bar_region->read8(RTL8169_IO_PHYSts) & RTL8169_PHYSts_LinkSts;
-		dprintln("  link status {}", link_up() ? "UP" : "DOWN");
+		if (m_link_up)
-		if (link_up())
 			dprintln("  link speed {}", link_speed());
 
 		TRY(enable_interrupt());
 		dprintln("  interrupts enabled");
 
+		// lock config registers
+		m_io_bar_region->write8(RTL8169_IO_9346CR, RTL8169_9346CR_MODE_NORMAL);
+
 		auto* thread = TRY(Thread::create_kernel([](void* rtl8169_ptr) {
 			static_cast<RTL8169*>(rtl8169_ptr)->receive_thread();
 		}, this));
@@ -78,17 +79,17 @@ namespace Kernel
 			delete thread;
 			return ret.release_error();
 		}
-		m_rx_thread_is_dead = false;
+		m_thread_is_dead = false;
 
 		return {};
 	}
 
 	RTL8169::~RTL8169()
 	{
-		m_rx_thread_should_die = true;
+		m_thread_should_die = true;
-		m_rx_blocker.unblock();
+		m_thread_blocker.unblock();
 
-		while (!m_rx_thread_is_dead)
+		while (!m_thread_is_dead)
 			Processor::yield();
 	}
 
@@ -109,33 +110,35 @@ namespace Kernel
 		m_rx_buffer_region = TRY(DMARegion::create(m_rx_descriptor_count * s_buffer_size, PageTable::MemoryType::Normal));
 		m_rx_descriptor_region = TRY(DMARegion::create(m_rx_descriptor_count * sizeof(RTL8169Descriptor)));
 
-		auto* rx_descriptors = reinterpret_cast<volatile RTL8169Descriptor*>(m_rx_descriptor_region->vaddr());
 		for (size_t i = 0; i < m_rx_descriptor_count; i++)
 		{
 			const paddr_t rx_buffer_paddr = m_rx_buffer_region->paddr() + i * s_buffer_size;
-			rx_descriptors[i].command     = 0x1FF8 | RTL8169_DESC_CMD_OWN;
+
-			rx_descriptors[i].vlan        = 0;
+			uint32_t command = 0x1FF8 | RTL8169_DESC_CMD_OWN;
-			rx_descriptors[i].buffer_low  = rx_buffer_paddr & 0xFFFFFFFF;
+			if (i == m_rx_descriptor_count - 1)
-			rx_descriptors[i].buffer_high = rx_buffer_paddr >> 32;
+				command |= RTL8169_DESC_CMD_EOR;
+
+			auto& rx_descriptor = reinterpret_cast<volatile RTL8169Descriptor*>(m_rx_descriptor_region->vaddr())[i];
+			rx_descriptor.command     = command;
+			rx_descriptor.vlan        = 0;
+			rx_descriptor.buffer_low  = rx_buffer_paddr & 0xFFFFFFFF;
+			rx_descriptor.buffer_high = rx_buffer_paddr >> 32;
 		}
-		rx_descriptors[m_rx_descriptor_count - 1].command |= RTL8169_DESC_CMD_EOR;
 
 		// configure rx descriptor addresses
-		m_io_bar_region->write32(RTL8169_IO_RDSAR + 4, m_rx_descriptor_region->paddr() >> 32);
 		m_io_bar_region->write32(RTL8169_IO_RDSAR + 0, m_rx_descriptor_region->paddr() & 0xFFFFFFFF);
+		m_io_bar_region->write32(RTL8169_IO_RDSAR + 4, m_rx_descriptor_region->paddr() >> 32);
 
-		// configure receive control (no fifo threshold, max dma burst unlimited, broadcast, multicast, accept physical match)
+		// configure receibe control (no fifo threshold, max dma burst 1024, accept physical match, broadcast, multicast)
 		m_io_bar_region->write32(RTL8169_IO_RCR,
-			RTL8169_RCR_RXFTH_NO | RTL8169_RCR_MXDMA_UNLIMITED | RTL8169_RCR_AB | RTL8169_RCR_AM | RTL8169_RCR_APM
+			RTL8169_RCR_RXFTH_NO | RTL8169_RCR_MXDMA_1024 | RTL8169_RCR_AB | RTL8169_RCR_AM | RTL8169_RCR_APM
 		);
 
+		m_io_bar_region->write32(0x44, 0b111111 | (0b111 << 8) | (0b111 << 13));
+
 		// configure max rx packet size
 		m_io_bar_region->write16(RTL8169_IO_RMS, RTL8169_RMS_MAX);
 
-		// enable ip/tcp/udp checksum offloading
-		// TODO: is this supported on all cards?
-		m_io_bar_region->write16(RTL8169_IO_CPlusCR, 1 << 5);
-
 		return {};
 	}
 
@@ -144,23 +147,27 @@ namespace Kernel
 		m_tx_buffer_region = TRY(DMARegion::create(m_tx_descriptor_count * s_buffer_size, PageTable::MemoryType::Normal));
 		m_tx_descriptor_region = TRY(DMARegion::create(m_tx_descriptor_count * sizeof(RTL8169Descriptor)));
 
-		auto* tx_descriptors = reinterpret_cast<volatile RTL8169Descriptor*>(m_tx_descriptor_region->vaddr());
 		for (size_t i = 0; i < m_tx_descriptor_count; i++)
 		{
 			const paddr_t tx_buffer_paddr = m_tx_buffer_region->paddr() + i * s_buffer_size;
-			tx_descriptors[i].command     = 0;
+
-			tx_descriptors[i].vlan        = 0;
+			uint32_t command = 0;
-			tx_descriptors[i].buffer_low  = tx_buffer_paddr & 0xFFFFFFFF;
+			if (i == m_tx_descriptor_count - 1)
-			tx_descriptors[i].buffer_high = tx_buffer_paddr >> 32;
+				command |= RTL8169_DESC_CMD_EOR;
+
+			auto& tx_descriptor = reinterpret_cast<volatile RTL8169Descriptor*>(m_tx_descriptor_region->vaddr())[i];
+			tx_descriptor.command     = command;
+			tx_descriptor.vlan        = 0;
+			tx_descriptor.buffer_low  = tx_buffer_paddr & 0xFFFFFFFF;
+			tx_descriptor.buffer_high = tx_buffer_paddr >> 32;
 		}
-		tx_descriptors[m_tx_descriptor_count - 1].command |= RTL8169_DESC_CMD_EOR;
 
 		// configure tx descriptor addresses
-		m_io_bar_region->write32(RTL8169_IO_TNPDS + 4, m_tx_descriptor_region->paddr() >> 32);
 		m_io_bar_region->write32(RTL8169_IO_TNPDS + 0, m_tx_descriptor_region->paddr() & 0xFFFFFFFF);
+		m_io_bar_region->write32(RTL8169_IO_TNPDS + 4, m_tx_descriptor_region->paddr() >> 32);
 
-		// configure transmit control (standard ifg, max dma burst unlimited)
+		// configure transmit control (standard ifg, max dma burst 1024)
-		m_io_bar_region->write32(RTL8169_IO_TCR, RTL8169_TCR_IFG_0 | RTL8169_TCR_MXDMA_UNLIMITED);
+		m_io_bar_region->write32(RTL8169_IO_TCR, RTL8169_TCR_IFG_0 | RTL8169_TCR_MXDMA_1024);
 
 		// configure max tx packet size
 		m_io_bar_region->write8(RTL8169_IO_MTPS, RTL8169_MTPS_MAX);
@@ -174,13 +181,14 @@ namespace Kernel
 		m_pci_device.enable_interrupt(0, *this);
 
 		m_io_bar_region->write16(RTL8169_IO_IMR,
-			RTL8169_IR_ROK |
+			RTL8169_IR_ROK
-			RTL8169_IR_RER |
+			| RTL8169_IR_RER
-			RTL8169_IR_TOK |
+			| RTL8169_IR_TOK
-			RTL8169_IR_TER |
+			| RTL8169_IR_TER
-			RTL8169_IR_RDU |
+			| RTL8169_IR_RDU
-			RTL8169_IR_LinkChg |
+			| RTL8169_IR_LinkChg
-			RTL8169_IR_FVOW
+			| RTL8169_IR_FVOW
+			| RTL8169_IR_TDU
 		);
 		m_io_bar_region->write16(RTL8169_IO_ISR, 0xFFFF);
 
@@ -201,156 +209,122 @@ namespace Kernel
 		return 0;
 	}
 
-	BAN::ErrorOr<void> RTL8169::send_raw_bytes(BAN::Span<const BAN::ConstByteSpan> buffers)
+	BAN::ErrorOr<void> RTL8169::send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::Span<const BAN::ConstByteSpan> payload)
 	{
 		if (!link_up())
 			return BAN::Error::from_errno(EADDRNOTAVAIL);
 
-		const auto interrupt_state = Processor::get_interrupt_state();
+		auto state = m_lock.lock();
-		Processor::set_interrupt_state(InterruptState::Disabled);
 
-		const uint32_t tx_current_nowrap = m_tx_head.fetch_add(1);
+		const uint32_t tx_current = m_tx_current;
-		const uint32_t tx_current = tx_current_nowrap % m_tx_descriptor_count;
+		m_tx_current = (m_tx_current + 1) % m_tx_descriptor_count;
 
 		auto& descriptor = reinterpret_cast<volatile RTL8169Descriptor*>(m_tx_descriptor_region->vaddr())[tx_current];
-		if (descriptor.command & RTL8169_DESC_CMD_OWN)
+		while (descriptor.command & RTL8169_DESC_CMD_OWN)
 		{
-			SpinLockGuard guard(m_tx_lock);
+			SpinLockAsMutex smutex(m_lock, state);
-			while (descriptor.command & RTL8169_DESC_CMD_OWN)
+			m_thread_blocker.block_indefinite(&smutex);
-			{
-				SpinLockGuardAsMutex smutex(guard);
-				m_tx_blocker.block_indefinite(&smutex);
-			}
 		}
 
+		m_lock.unlock(state);
+
 		auto* tx_buffer = reinterpret_cast<uint8_t*>(m_tx_buffer_region->vaddr() + tx_current * s_buffer_size);
 
 		// write packet
-		size_t packet_size = 0;
+		auto& ethernet_header = *reinterpret_cast<EthernetHeader*>(tx_buffer);
-		for (const auto& buffer : buffers)
+		ethernet_header.dst_mac = destination;
+		ethernet_header.src_mac = get_mac_address();
+		ethernet_header.ether_type = protocol;
+
+		size_t packet_size = sizeof(EthernetHeader);
+		for (const auto& buffer : payload)
 		{
-			ASSERT(packet_size + buffer.size() <= s_buffer_size);
 			memcpy(tx_buffer + packet_size, buffer.data(), buffer.size());
 			packet_size += buffer.size();
 		}
 
 		// give packet ownership to NIC
 		uint32_t command = packet_size | RTL8169_DESC_CMD_OWN | RTL8169_DESC_CMD_LS | RTL8169_DESC_CMD_FS;
-		if (tx_current == m_tx_descriptor_count - 1)
+		if (tx_current >= m_tx_descriptor_count - 1)
 			command |= RTL8169_DESC_CMD_EOR;
 		descriptor.command = command;
 
-		// ring tx queue doorbell
+		// notify NIC about new packet
-		if (tx_current_nowrap == m_tx_commit.load())
+		m_io_bar_region->write8(RTL8169_IO_TPPoll, RTL8169_TPPoll_NPQ);
-			m_io_bar_region->write8(RTL8169_IO_TPPoll, RTL8169_TPPoll_NPQ);
-		while (tx_current_nowrap != m_tx_commit.load())
-			Processor::pause();
-		m_tx_commit.add_fetch(1);
-
-		Processor::set_interrupt_state(interrupt_state);
 
 		return {};
 	}
 
 	void RTL8169::receive_thread()
 	{
-		SpinLockGuard rx_lock_guard(m_rx_lock);
+		SpinLockGuard _(m_lock);
 
-		while (!m_rx_thread_should_die)
+		while (!m_thread_should_die)
 		{
 			for (;;)
 			{
-				auto& descriptor = reinterpret_cast<volatile RTL8169Descriptor*>(m_rx_descriptor_region->vaddr())[m_rx_head];
+				auto& descriptor = reinterpret_cast<volatile RTL8169Descriptor*>(m_rx_descriptor_region->vaddr())[m_rx_current];
-
-				const auto command = descriptor.command;
 				if (descriptor.command & RTL8169_DESC_CMD_OWN)
 					break;
 
 				// packet buffer can only hold single packet, so we should not receive any multi-descriptor packets
-				ASSERT((command & RTL8169_DESC_CMD_LS) && (command & RTL8169_DESC_CMD_FS));
+				ASSERT((descriptor.command & RTL8169_DESC_CMD_LS) && (descriptor.command & RTL8169_DESC_CMD_FS));
 
-				const uint8_t protocol = (command >> 17) & 3;
+				const uint16_t packet_length = descriptor.command & 0x3FFF;
-				const uint16_t packet_length = command & 0x3FFF;
 				if (packet_length > s_buffer_size)
 					dwarnln("Got {} bytes to {} byte buffer", packet_length, s_buffer_size);
-				else if (command & (1u << 21))
+				else if (descriptor.command & (1u << 21))
-					dwarnln("descriptor error {4h}", command);
+					; // descriptor has an error
-				else if (protocol == 1 && (command & (1u << 14)))
-					dwarnln("TCP checksum error");
-				else if (protocol == 2 && (command & (1u << 15)))
-					dwarnln("UDP checksum error");
-				else if (protocol != 0 && (command & (1u << 16)))
-					dwarnln("IPv4 checksum error");
 				else
 				{
-					m_rx_lock.unlock(InterruptState::Enabled);
+					m_lock.unlock(InterruptState::Enabled);
 
-					uint32_t validated_cksums;
+					NetworkManager::get().on_receive(*this, BAN::ConstByteSpan {
-					switch (protocol)
+						reinterpret_cast<const uint8_t*>(m_rx_buffer_region->vaddr() + m_rx_current * s_buffer_size),
-					{
+						packet_length
-						case 0: validated_cksums = 0;                      break;
+					});
-						case 1: validated_cksums = CKSUM_IPV4 | CKSUM_TCP; break;
-						case 2: validated_cksums = CKSUM_IPV4 | CKSUM_UDP; break;
-						case 3: validated_cksums = CKSUM_IPV4;             break;
-					}
 
-					const uint8_t* packet_data = reinterpret_cast<const uint8_t*>(m_rx_buffer_region->vaddr() + m_rx_head * s_buffer_size);
+					m_lock.lock();
-					NetworkManager::get().on_receive(*this, BAN::ConstByteSpan { packet_data, packet_length }, validated_cksums);
-
-					m_rx_lock.lock();
 				}
 
-				uint32_t new_command = 0x1FF8 | RTL8169_DESC_CMD_OWN;
+				m_rx_current = (m_rx_current + 1) % m_rx_descriptor_count;
-				if (m_rx_head == m_rx_descriptor_count - 1)
-					new_command |= RTL8169_DESC_CMD_EOR;
-				descriptor.command = new_command;
 
-				m_rx_head = (m_rx_head + 1) % m_rx_descriptor_count;
+				descriptor.command = descriptor.command | RTL8169_DESC_CMD_OWN;
 			}
 
-			SpinLockGuardAsMutex smutex(rx_lock_guard);
+			SpinLockAsMutex smutex(m_lock, InterruptState::Enabled);
-			m_rx_blocker.block_indefinite(&smutex);
+			m_thread_blocker.block_indefinite(&smutex);
 		}
 
-		m_rx_thread_is_dead = true;
+		m_thread_is_dead = true;
 	}
 
 	void RTL8169::handle_irq()
 	{
-		uint16_t isr;
+		const uint16_t interrupt_status = m_io_bar_region->read16(RTL8169_IO_ISR);
-		while ((isr = m_io_bar_region->read16(RTL8169_IO_ISR)))
+		m_io_bar_region->write16(RTL8169_IO_ISR, interrupt_status);
+
+		if (interrupt_status & RTL8169_IR_LinkChg)
 		{
-			m_io_bar_region->write16(RTL8169_IO_ISR, isr);
+			m_link_up = m_io_bar_region->read8(RTL8169_IO_PHYSts) & RTL8169_PHYSts_LinkSts;
-
+			dprintln("link status -> {}", m_link_up.load());
-			if (isr & RTL8169_IR_LinkChg)
-			{
-				m_link_up = m_io_bar_region->read8(RTL8169_IO_PHYSts) & RTL8169_PHYSts_LinkSts;
-				dprintln("link status {}", link_up() ? "UP" : "DOWN");
-				if (link_up())
-					dprintln("link speed {}", link_speed());
-			}
-
-			if (isr & (RTL8169_IR_TER | RTL8169_IR_TOK))
-			{
-				SpinLockGuard _(m_tx_lock);
-				m_tx_blocker.unblock();
-			}
-
-			if (isr & (RTL8169_IR_RER | RTL8169_IR_ROK))
-			{
-				SpinLockGuard _(m_rx_lock);
-				m_rx_blocker.unblock();
-			}
-
-			if (isr & RTL8169_IR_RER)
-				dwarnln("Rx error");
-			if (isr & RTL8169_IR_TER)
-				dwarnln("Tx error");
-			if (isr & RTL8169_IR_RDU)
-				dwarnln("Rx descriptor not available");
-			if (isr & RTL8169_IR_FVOW)
-				dwarnln("Rx FIFO overflow");
 		}
+
+		if (interrupt_status & (RTL8169_IR_TOK | RTL8169_IR_ROK))
+		{
+			SpinLockGuard _(m_lock);
+			m_thread_blocker.unblock();
+		}
+
+		if (interrupt_status & RTL8169_IR_RER)
+			dwarnln("Rx error");
+		if (interrupt_status & RTL8169_IR_TER)
+			dwarnln("Tx error");
+		if (interrupt_status & RTL8169_IR_RDU)
+			dwarnln("Rx descriptor not available");
+		if (interrupt_status & RTL8169_IR_FVOW)
+			dwarnln("Rx FIFO overflow");
+		// dont log TDU is sent after each sent packet
 	}
 
 }

kernel/kernel/Networking/TCPSocket.cpp

@@ -402,7 +402,7 @@ namespace Kernel
 		return {};
 	}
 
-	BAN::ErrorOr<long> TCPSocket::ioctl_impl(unsigned long request, void* argument)
+	BAN::ErrorOr<long> TCPSocket::ioctl_impl(int request, void* argument)
 	{
 		switch (request)
 		{
@@ -572,7 +572,7 @@ namespace Kernel
 		dprintln_if(DEBUG_TCP, "  seq {}", (uint32_t)header.seq_number);
 	}
 
-	void TCPSocket::receive_packet(BAN::ConstByteSpan packet, const sockaddr* sender, socklen_t sender_len, uint32_t validated_cksums)
+	void TCPSocket::receive_packet(BAN::ConstByteSpan buffer, const sockaddr* sender, socklen_t sender_len)
 	{
 		if (m_state == State::Listen)
 		{
@@ -585,10 +585,9 @@ namespace Kernel
 				}();
 
 			if (socket)
-				return socket->receive_packet(packet, sender, sender_len, validated_cksums);
+				return socket->receive_packet(buffer, sender, sender_len);
 		}
 
-		if (!(validated_cksums & CKSUM_TCP))
 		{
 			uint16_t checksum = 0;
 
@@ -604,23 +603,23 @@ namespace Kernel
 					.src_ipv4 = BAN::IPv4Address(addr_in.sin_addr.s_addr),
 					.dst_ipv4 = interface->get_ipv4_address(),
 					.protocol = NetworkProtocol::TCP,
-					.length = packet.size(),
+					.length = buffer.size(),
 				};
 				const BAN::ConstByteSpan buffers[] {
 					BAN::ConstByteSpan::from(pseudo_header),
-					packet
+					buffer
 				};
 				checksum = calculate_internet_checksum({ buffers, sizeof(buffers) / sizeof(*buffers) });
 			}
 			else
 			{
-				dwarnln("no TCP checksum validation for socket family {}", sender->sa_family);
+				dwarnln("No tcp checksum validation for socket family {}", sender->sa_family);
 				return;
 			}
 
 			if (checksum != 0)
 			{
-				dwarnln("checksum does not match");
+				dprintln("Checksum does not match");
 				return;
 			}
 		}
@@ -629,7 +628,7 @@ namespace Kernel
 
 		const bool hungup_before = has_hungup_impl();
 
-		const auto& header = packet.as<const TCPHeader>();
+		auto& header = buffer.as<const TCPHeader>();
 
 		dprintln_if(DEBUG_TCP, "receiving {} {8b}", (uint8_t)m_state, header.flags);
 		dprintln_if(DEBUG_TCP, "  ack {}", (uint32_t)header.ack_number);
@@ -769,7 +768,7 @@ namespace Kernel
 			if (header.ack_number > m_send_window.current_ack)
 				m_send_window.current_ack = header.ack_number;
 
-			auto payload = packet.slice(header.data_offset * sizeof(uint32_t));
+			auto payload = buffer.slice(header.data_offset * sizeof(uint32_t));
 
 			if (header.seq_number < expected_seq)
 			{
@@ -968,9 +967,6 @@ namespace Kernel
 			if (m_last_sent_window_size == 0)
 				m_should_send_zero_window = false;
 
-			if (m_last_sent_window_size == m_recv_window.buffer->free())
-				m_should_send_window_update = false;
-
 			if (m_should_send_zero_window || m_should_send_window_update)
 			{
 				ASSERT(m_connection_info.has_value());

kernel/kernel/Networking/UDPSocket.cpp

@@ -57,69 +57,21 @@ namespace Kernel
 			header.checksum = 0xFFFF;
 	}
 
-	void UDPSocket::receive_packet(BAN::ConstByteSpan packet, const sockaddr* sender, socklen_t sender_len, uint32_t validated_cksums)
+	void UDPSocket::receive_packet(BAN::ConstByteSpan packet, const sockaddr* sender, socklen_t sender_len)
 	{
-		const auto header = packet.as<const UDPHeader>();
+		auto payload = packet.slice(sizeof(UDPHeader));
-
-		if (!(validated_cksums & CKSUM_UDP) && header.checksum)
-		{
-			uint16_t checksum = 0;
-
-			if (sender->sa_family == AF_INET)
-			{
-				auto interface_or_error = interface(sender, sender_len);
-				if (interface_or_error.is_error())
-					return;
-				auto interface = interface_or_error.release_value();
-
-				auto& addr_in = *reinterpret_cast<const sockaddr_in*>(sender);
-				const PseudoHeader pseudo_header {
-					.src_ipv4 = BAN::IPv4Address(addr_in.sin_addr.s_addr),
-					.dst_ipv4 = interface->get_ipv4_address(),
-					.protocol = NetworkProtocol::UDP,
-					.length = packet.size(),
-				};
-				const UDPHeader udp_header {
-					.src_port = header.src_port,
-					.dst_port = header.dst_port,
-					.length   = header.length,
-					.checksum = 0,
-				};
-				const BAN::ConstByteSpan buffers[] {
-					BAN::ConstByteSpan::from(pseudo_header),
-					BAN::ConstByteSpan::from(udp_header),
-					packet.slice(sizeof(UDPHeader))
-				};
-				checksum = calculate_internet_checksum({ buffers, sizeof(buffers) / sizeof(*buffers) });
-			}
-			else
-			{
-				dwarnln("no UDP checksum validation for socket family {}", sender->sa_family);
-				return;
-			}
-
-			const bool checksum_valid1 = (header.checksum == checksum);
-			const bool checksum_valid2 = (header.checksum == 0xFFFF && checksum == 0);
-			if (!checksum_valid1 && !checksum_valid2)
-			{
-				dwarnln("checksum does not match {4h}", checksum);
-				return;
-			}
-		}
 
 		SpinLockGuard _(m_packet_lock);
 
 		if (m_packets.full())
 		{
-			dwarnln("Packet buffer full, dropping packet");
+			dprintln("Packet buffer full, dropping packet");
 			return;
 		}
 
-		const auto payload = packet.slice(sizeof(UDPHeader));
-
 		if (m_packet_total_size + payload.size() > m_packet_buffer->size())
 		{
-			dwarnln("Packet buffer full, dropping packet");
+			dprintln("Packet buffer full, dropping packet");
 			return;
 		}
 
@@ -272,9 +224,7 @@ namespace Kernel
 			address_len = message.msg_namelen;
 		}
 
-		TRY(m_network_layer.sendto(*this, buffer.span(), address, address_len));
+		return TRY(m_network_layer.sendto(*this, buffer.span(), address, address_len));
-
-		return buffer.size();
 	}
 
 	BAN::ErrorOr<void> UDPSocket::getsockopt_impl(int level, int option, void* value, socklen_t* value_len)
@@ -336,7 +286,7 @@ namespace Kernel
 		return {};
 	}
 
-	BAN::ErrorOr<long> UDPSocket::ioctl_impl(unsigned long request, void* argument)
+	BAN::ErrorOr<long> UDPSocket::ioctl_impl(int request, void* argument)
 	{
 		switch (request)
 		{

kernel/kernel/Networking/UNIX/Socket.cpp

@@ -704,25 +704,6 @@ namespace Kernel
 		}
 	}
 
-	BAN::ErrorOr<void> UnixDomainSocket::getsockname_impl(sockaddr* address, socklen_t* address_len)
-	{
-		sockaddr_un sa_un {
-			.sun_family = AF_UNIX,
-			.sun_path = {},
-		};
-
-		{
-			LockGuard _(m_bind_mutex);
-			strcpy(sa_un.sun_path, m_bound_file.canonical_path.data());
-		}
-
-		const size_t to_copy = BAN::Math::min<socklen_t>(sizeof(sockaddr_un), *address_len);
-		memcpy(address, &sa_un, to_copy);
-		*address_len = to_copy;
-		return {};
-	}
-
-
 	BAN::ErrorOr<void> UnixDomainSocket::getpeername_impl(sockaddr* address, socklen_t* address_len)
 	{
 		if (!m_info.has<ConnectionInfo>())
@@ -732,7 +713,20 @@ namespace Kernel
 		if (!connection)
 			return BAN::Error::from_errno(ENOTCONN);
 
-		return connection->getsockname_impl(address, address_len);
+		sockaddr_un sa_un {
+			.sun_family = AF_UNIX,
+			.sun_path = {},
+		};
+
+		{
+			LockGuard _(m_bind_mutex);
+			strcpy(sa_un.sun_path, connection->m_bound_file.canonical_path.data());
+		}
+
+		const size_t to_copy = BAN::Math::min<socklen_t>(sizeof(sockaddr_un), *address_len);
+		memcpy(address, &sa_un, to_copy);
+		*address_len = to_copy;
+		return {};
 	}
 
 	BAN::ErrorOr<void> UnixDomainSocket::getsockopt_impl(int level, int option, void* value, socklen_t* value_len)
@@ -749,24 +743,12 @@ namespace Kernel
 			case SO_ERROR:
 				result = 0;
 				break;
-			case SO_KEEPALIVE:
-				result = 1;
-				break;
 			case SO_SNDBUF:
 				result = m_sndbuf;
 				break;
 			case SO_RCVBUF:
 				result = m_packet_buffer->size();
 				break;
-			case SO_TYPE:
-				switch (m_socket_type)
-				{
-					case Type::STREAM:    result = SOCK_STREAM;    break;
-					case Type::DGRAM:     result = SOCK_DGRAM;     break;
-					case Type::SEQPACKET: result = SOCK_SEQPACKET; break;
-					default: ASSERT_NOT_REACHED();
-				}
-				break;
 			default:
 				dwarnln("getsockopt(SOL_SOCKET, {})", option);
 				return BAN::Error::from_errno(ENOTSUP);

kernel/kernel/OpenFileDescriptorSet.cpp

@@ -401,19 +401,11 @@ namespace Kernel
 
 	BAN::ErrorOr<void> OpenFileDescriptorSet::close(int fd)
 	{
-		BAN::RefPtr<OpenFileDescription> open_file;
+		LockGuard _(m_mutex);
 
-		{
+		TRY(validate_fd(fd));
-			LockGuard _(m_mutex);
 
-			TRY(validate_fd(fd));
+		auto& open_file = m_open_files[fd];
-
-			open_file = m_open_files[fd];
-
-			m_open_files[fd]->file.inode->on_close(m_open_files[fd]->status_flags);
-			m_open_files[fd] = {};
-			remove_cloexec(fd);
-		}
 
 		if (auto& flock = open_file->flock; Thread::current().has_process() && flock.lockers.contains(Process::current().pid()))
 		{
@@ -437,6 +429,10 @@ namespace Kernel
 			}
 		}
 
+		open_file->file.inode->on_close(open_file->status_flags);
+		open_file = {};
+		remove_cloexec(fd);
+
 		return {};
 	}
 
@@ -857,15 +853,6 @@ namespace Kernel
 		return inode->sendmsg(message, flags | (is_nonblock ? MSG_DONTWAIT : 0));
 	}
 
-	int OpenFileDescriptorSet::get_max_open_fd() const
-	{
-		LockGuard _(m_mutex);
-		for (int fd = m_open_files.size() - 1; fd > 0; fd--)
-			if (m_open_files[fd])
-				return fd;
-		return -1;
-	}
-
 	BAN::ErrorOr<VirtualFileSystem::File> OpenFileDescriptorSet::file_of(int fd) const
 	{
 		LockGuard _(m_mutex);

kernel/kernel/PCI.cpp

@@ -235,13 +235,11 @@ namespace Kernel::PCI
 						switch (pci_device.subclass())
 						{
 							case 0x01:
+							case 0x05:
+							case 0x06:
 								if (auto res = ATAController::create(pci_device); res.is_error())
 									dprintln("ATA: {}", res.error());
 								break;
-							case 0x06:
-								if (auto res = AHCIController::create(pci_device); res.is_error())
-									dprintln("AHCI: {}", res.error());
-								break;
 							case 0x08:
 								if (auto res = NVMeController::create(pci_device); res.is_error())
 									dprintln("NVMe: {}", res.error());

kernel/kernel/Processor.cpp

@@ -1,4 +1,3 @@
-#include <kernel/CPUID.h>
 #include <kernel/InterruptController.h>
 #include <kernel/Memory/Heap.h>
 #include <kernel/Memory/kmalloc.h>
@@ -21,11 +20,10 @@ namespace Kernel
 	static constexpr uint32_t MSR_IA32_FMASK = 0xC0000084;
 #endif
 
-	static constexpr uint32_t MSR_IA32_TSC_AUX = 0xC0000103;
-
 	ProcessorID          Processor::s_bsp_id                     { PROCESSOR_NONE };
 	BAN::Atomic<uint8_t> Processor::s_processor_count            { 0 };
 	BAN::Atomic<bool>    Processor::s_is_smp_enabled             { false };
+	BAN::Atomic<bool>    Processor::s_should_print_cpu_load      { false };
 	paddr_t              Processor::s_shared_page_paddr          { 0 };
 	vaddr_t              Processor::s_shared_page_vaddr          { 0 };
 
@@ -222,9 +220,6 @@ namespace Kernel
 		shared_page.gdt_cpu_offset = GDT::cpu_index_offset();
 		shared_page.features = 0;
 
-		if (CPUID::has_rdtscp())
-			shared_page.features |= API::SPF_RDTSCP;
-
 		ASSERT(Processor::count() + sizeof(Kernel::API::SharedPage) <= PAGE_SIZE);
 	}
 
@@ -295,11 +290,13 @@ namespace Kernel
 		}
 	}
 
-	void Processor::initialize_tsc(uint64_t realtime_seconds)
+	void Processor::initialize_tsc(uint8_t shift, uint64_t mult, uint64_t realtime_seconds)
 	{
 		auto& shared_page = Processor::shared_page();
-		shared_page.gettime_shared.realtime_s = realtime_seconds;
+
-		shared_page.gettime_shared.realtime_ns = 0;
+		shared_page.gettime_shared.shift = shift;
+		shared_page.gettime_shared.mult = mult;
+		shared_page.gettime_shared.realtime_seconds = realtime_seconds;
 
 		update_tsc();
 
@@ -326,63 +323,23 @@ namespace Kernel
 
 	void Processor::update_tsc()
 	{
-		auto& lgettime = shared_page().cpus[current_index()].gettime_local;
+		auto& sgettime = shared_page().cpus[current_index()].gettime_local;
-
+		sgettime.seq = sgettime.seq + 1;
-		const auto seq = BAN::atomic_load(lgettime.seq, BAN::memory_order_relaxed);
+		sgettime.last_ns = SystemTimer::get().ns_since_boot_no_tsc();
-
+		sgettime.last_tsc = __builtin_ia32_rdtsc();
-		BAN::atomic_store(lgettime.seq, seq + 1, BAN::memory_order_release);
+		sgettime.seq = sgettime.seq + 1;
-
-		if (lgettime.seq == 1)
-		{
-			const auto tsc_info = SystemTimer::get().tsc_info();
-			lgettime.shift = tsc_info.shift;
-			lgettime.mult  = tsc_info.mult;
-			lgettime.last_ns  = SystemTimer::get().ns_since_boot_no_tsc();
-			lgettime.last_tsc = __builtin_ia32_rdtsc();
-
-			if (CPUID::has_rdtscp())
-				asm volatile("wrmsr" :: "d"(0x00000000), "a"(current_index()), "c"(MSR_IA32_TSC_AUX));
-		}
-		else
-		{
-			const auto current_ns = SystemTimer::get().ns_since_boot_no_tsc();
-			const auto current_tsc = __builtin_ia32_rdtsc();
-
-			auto delta_ns = current_tsc - lgettime.last_tsc;
-			if (lgettime.shift >= 0)
-				delta_ns <<= lgettime.shift;
-			else
-				delta_ns >>= -lgettime.shift;
-			delta_ns = (delta_ns * lgettime.mult) >> 32;
-
-			lgettime.last_ns += delta_ns;
-			lgettime.last_tsc = current_tsc;
-
-			// scale mult by [-0.25%, 0.25%] to fix for clock drift
-			const auto error_ns = static_cast<int64_t>(current_ns) - static_cast<int64_t>(lgettime.last_ns);
-			const auto correction_ppm = BAN::Math::clamp<int64_t>(error_ns * 1'000'000 / 1'000'000'000, -100, 100);
-			const auto correction_delta = -lgettime.mult * correction_ppm / 1'000'000;
-			lgettime.mult += correction_delta;
-		}
-
-		BAN::atomic_store(lgettime.seq, seq + 2, BAN::memory_order_release);
 	}
 
 	uint64_t Processor::ns_since_boot_tsc()
 	{
 		const auto& shared_page = Processor::shared_page();
+		const auto& sgettime = shared_page.gettime_shared;
 		const auto& lgettime = shared_page.cpus[current_index()].gettime_local;
 
 		auto state = get_interrupt_state();
 		set_interrupt_state(InterruptState::Disabled);
 
-		uint64_t current_ns = __builtin_ia32_rdtsc() - lgettime.last_tsc;
+		const auto current_ns = lgettime.last_ns + (((__builtin_ia32_rdtsc() - lgettime.last_tsc) * sgettime.mult) >> sgettime.shift);
-		if (lgettime.shift >= 0)
-			current_ns <<= lgettime.shift;
-		else
-			current_ns >>= -lgettime.shift;
-		current_ns = (current_ns * lgettime.mult) >> 32;
-		current_ns += lgettime.last_ns;
 
 		set_interrupt_state(state);
 
@@ -641,26 +598,6 @@ namespace Kernel
 		set_interrupt_state(state);
 	}
 
-	Processor::LoadStats Processor::get_load_stats(size_t index)
-	{
-		ASSERT(index < Processor::count());
-
-		auto& processor = s_processors[s_processor_ids[index].as_u32()];
-
-		bool expected = false;
-		while (!processor.m_load_stat_lock.compare_exchange(expected, true))
-		{
-			Processor::pause();
-			expected = false;
-		}
-
-		const auto load_stats = processor.m_load_stats;
-
-		processor.m_load_stat_lock.store(false);
-
-		return load_stats;
-	}
-
 	void Processor::yield()
 	{
 		auto state = get_interrupt_state();
@@ -668,32 +605,47 @@ namespace Kernel
 
 		ASSERT(!Thread::current().has_spinlock());
 
-		auto& processor = s_processors[current_id().as_u32()];
+		auto& processor_info = s_processors[current_id().as_u32()];
 
 		{
-			bool expected = false;
+			constexpr uint64_t load_update_interval_ns = 1'000'000'000;
-			while (!processor.m_load_stat_lock.compare_exchange(expected, true))
-			{
-				Processor::pause();
-				expected = false;
-			}
 
-			const uint64_t elapsed_ns = SystemTimer::get().ns_since_boot() - processor.m_load_start_ns;
+			const uint64_t current_ns = SystemTimer::get().ns_since_boot();
 
-			auto& load_stats = processor.m_load_stats;
 			if (scheduler().is_idle())
-				load_stats.ns_idle += elapsed_ns;
+				processor_info.m_idle_ns += current_ns - processor_info.m_start_ns;
-			load_stats.ns_total += elapsed_ns;
 
-			processor.m_load_stat_lock.store(false);
+			if (current_ns >= processor_info.m_next_update_ns)
+			{
+				if (s_should_print_cpu_load && g_terminal_driver)
+				{
+					const uint64_t duration_ns = current_ns - processor_info.m_last_update_ns;
+					const uint64_t load_x1000  = 100'000 * (duration_ns - BAN::Math::min(processor_info.m_idle_ns, duration_ns)) / duration_ns;
+
+					uint32_t x = g_terminal_driver->width() - 16;
+					uint32_t y = current_id().as_u32();
+					const auto proc_putc =
+						[&x, y](char ch)
+						{
+							if (x < g_terminal_driver->width() && y < g_terminal_driver->height())
+								g_terminal_driver->putchar_at(ch, x++, y, TerminalColor::WHITE, TerminalColor::BLACK);
+						};
+
+					BAN::Formatter::print(proc_putc, "CPU { 2}: { 3}.{3}%", current_id(), load_x1000 / 1000, load_x1000 % 1000);
+				}
+
+				processor_info.m_idle_ns         = 0;
+				processor_info.m_last_update_ns  = current_ns;
+				processor_info.m_next_update_ns += load_update_interval_ns;
+			}
 		}
 
 		if (!scheduler().is_idle())
 			Thread::current().set_cpu_time_stop();
 
-		asm_yield_trampoline(processor.stack_top_vaddr());
+		asm_yield_trampoline(processor_info.stack_top_vaddr());
 
-		processor.m_load_start_ns = SystemTimer::get().ns_since_boot();
+		processor_info.m_start_ns = SystemTimer::get().ns_since_boot();
 
 		if (!scheduler().is_idle())
 			Thread::current().set_cpu_time_start();

kernel/kernel/Scheduler.cpp

@@ -1,6 +1,5 @@
 #include <BAN/Optional.h>
 #include <BAN/Sort.h>
-#include <kernel/APIC.h>
 #include <kernel/InterruptController.h>
 #include <kernel/Lock/Mutex.h>
 #include <kernel/Process.h>
@@ -46,15 +45,12 @@ namespace Kernel
 		m_tail = node;
 	}
 
-	bool SchedulerQueue::add_thread_with_wake_time(Node* node)
+	void SchedulerQueue::add_thread_with_wake_time(Node* node)
 	{
 		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
 
 		if (m_tail == nullptr || node->wake_time_ns >= m_tail->wake_time_ns)
-		{
+			return add_thread_to_back(node);
-			add_thread_to_back(node);
-			return node == m_head;
-		}
 
 		Node* next = m_head;
 		Node* prev = nullptr;
@@ -68,8 +64,6 @@ namespace Kernel
 		node->prev = prev;
 		(next ? next->prev : m_tail) = node;
 		(prev ? prev->next : m_head) = node;
-
-		return node == m_head;
 	}
 
 	template<typename F>
@@ -128,9 +122,15 @@ namespace Kernel
 		m_idle_thread = TRY(Thread::create_kernel([](void*) { asm volatile("1: hlt; jmp 1b"); }, nullptr));
 		ASSERT(m_idle_thread);
 
+		size_t processor_index = 0;
+		for (; processor_index < Processor::count(); processor_index++)
+			if (Processor::id_from_index(processor_index) == Processor::current_id())
+				break;
+		ASSERT(processor_index < Processor::count());
+
 		// each CPU does load balance at different times. This calulates the offset to other CPUs
-		m_last_load_balance_ns = s_load_balance_interval_ns * Processor::current_index() / Processor::count();
+		m_last_load_balance_ns = s_load_balance_interval_ns * processor_index / Processor::count();
-		m_idle_ns              = m_last_load_balance_ns;
+		m_idle_ns              = -m_last_load_balance_ns;
 
 		s_schedulers_initialized++;
 		while (s_schedulers_initialized < Processor::count())
@@ -139,6 +139,8 @@ namespace Kernel
 		if (Processor::count() > 1)
 			Processor::set_smp_enabled();
 
+		m_next_reschedule_ns = SystemTimer::get().ns_since_boot();
+
 		return {};
 	}
 
@@ -230,14 +232,17 @@ namespace Kernel
 					m_thread_count--;
 					break;
 				case Thread::State::Executing:
+				{
+					const uint64_t current_ns = SystemTimer::get().ns_since_boot();
 					m_current->thread->yield_registers() = *yield_registers;
-					m_current->time_used_ns += SystemTimer::get().ns_since_boot() - m_current->last_start_ns;
+					m_current->time_used_ns += current_ns - m_current->last_start_ns;
 					add_current_to_most_loaded(m_current->blocked ? &m_block_queue : &m_run_queue);
 					if (!m_current->blocked)
 						m_run_queue.add_thread_to_back(m_current);
-					else if (m_block_queue.add_thread_with_wake_time(m_current))
+					else
-						update_wake_up_deadline();
+						m_block_queue.add_thread_with_wake_time(m_current);
 					break;
+				}
 				case Thread::State::NotStarted:
 					ASSERT(!m_current->blocked);
 					m_current->time_used_ns = 0;
@@ -316,48 +321,26 @@ namespace Kernel
 		}
 	}
 
-	void Scheduler::update_wake_up_deadline()
-	{
-		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
-
-		auto& interrupt_controller = InterruptController::get();
-
-		// TODO: support timer deadlines on non-apic timers and abstract it :)
-		if (!interrupt_controller.is_using_apic())
-			return;
-
-		uint64_t deadline_ns = m_next_reschedule_ns;
-		if (!m_block_queue.empty())
-			deadline_ns = BAN::Math::min(deadline_ns, m_block_queue.front()->wake_time_ns);
-		if (Processor::is_smp_enabled())
-			deadline_ns = BAN::Math::min(deadline_ns, m_last_load_balance_ns + s_load_balance_interval_ns);
-
-		static_cast<APIC&>(interrupt_controller).set_timer_dealine(deadline_ns);
-	}
-
 	void Scheduler::reschedule_if_idle()
 	{
 		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
 
-		if (m_current == nullptr && !m_run_queue.empty())
+		if (m_current != nullptr)
+			return;
+
+		if (m_run_queue.empty())
+			wake_up_sleeping_threads();
+
+		if (!m_run_queue.empty())
 			Processor::yield();
 	}
 
-	extern "C" void scheduler_on_yield_trampoline(YieldRegisters* yield_registers)
+	extern "C" void scheduler_on_yield(YieldRegisters* yield_registers)
 	{
-		Processor::scheduler().on_yield(yield_registers);
+		Processor::scheduler().reschedule(yield_registers);
 	}
 
-	void Scheduler::on_yield(YieldRegisters* yield_registers)
+	void Scheduler::timer_interrupt()
-	{
-		reschedule(yield_registers);
-		m_next_reschedule_ns = !is_idle()
-			? SystemTimer::get().ns_since_boot() + s_reschedule_interval_ns
-			: BAN::numeric_limits<uint64_t>::max();
-		update_wake_up_deadline();
-	}
-
-	void Scheduler::on_timer_interrupt()
 	{
 		ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
 
@@ -366,12 +349,11 @@ namespace Kernel
 
 		wake_up_sleeping_threads();
 
-		// NOTE: yield will update the timer deadline, but if we do not yield make sure
+		if (SystemTimer::get().ns_since_boot() >= m_next_reschedule_ns)
-		//       we set the next deadline or we won't get another timer interrupt
+		{
-		if (is_idle() || SystemTimer::get().ns_since_boot() >= m_next_reschedule_ns)
+			m_next_reschedule_ns += s_reschedule_interval_ns;
 			Processor::yield();
-		else
+		}
-			update_wake_up_deadline();
 	}
 
 	void Scheduler::unblock_thread(SchedulerQueue::Node* node)
@@ -412,8 +394,8 @@ namespace Kernel
 
 		if (!node->blocked)
 			m_run_queue.add_thread_to_back(node);
-		else if (m_block_queue.add_thread_with_wake_time(node))
+		else
-			update_wake_up_deadline();
+			m_block_queue.add_thread_with_wake_time(node);
 
 		if (auto* thread = node->thread; thread->is_userspace() && thread->has_process())
 			thread->update_processor_index_address();

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

@@ -7,16 +7,6 @@
 namespace Kernel
 {
 
-	BAN::ErrorOr<BAN::RefPtr<AHCIController>> AHCIController::create(PCI::Device& pci_device)
-	{
-		auto* controller_ptr = new AHCIController(pci_device);
-		if (controller_ptr == nullptr)
-			return BAN::Error::from_errno(ENOMEM);
-		auto controller = BAN::RefPtr<AHCIController>::adopt(controller_ptr);
-		TRY(controller->initialize());
-		return controller;
-	}
-
 	BAN::ErrorOr<void> AHCIController::initialize()
 	{
 		m_abar = TRY(m_pci_device.allocate_bar_region(5));

kernel/kernel/Storage/ATA/ATABus.cpp

@@ -66,7 +66,7 @@ namespace Kernel
 
 	static void select_delay()
 	{
-		SystemTimer::get().sleep_for_ns(400);
+		SystemTimer::get().sleep_ns(400);
 	}
 
 	void ATABus::select_device(bool is_secondary)
@@ -87,7 +87,7 @@ namespace Kernel
 		io_write(ATA_PORT_LBA1, 0);
 		io_write(ATA_PORT_LBA2, 0);
 		io_write(ATA_PORT_COMMAND, ATA_COMMAND_IDENTIFY);
-		SystemTimer::get().sleep_for_ms(1);
+		SystemTimer::get().sleep_ms(1);
 
 		// No device on port
 		if (io_read(ATA_PORT_STATUS) == 0)
@@ -112,7 +112,7 @@ namespace Kernel
 			}
 
 			io_write(ATA_PORT_COMMAND, ATA_COMMAND_IDENTIFY_PACKET);
-			SystemTimer::get().sleep_for_ms(1);
+			SystemTimer::get().sleep_ms(1);
 		}
 
 		TRY(wait(true));

kernel/kernel/Storage/ATA/ATAController.cpp

@@ -9,12 +9,29 @@
 namespace Kernel
 {
 
-	BAN::ErrorOr<BAN::RefPtr<ATAController>> ATAController::create(PCI::Device& pci_device)
+	BAN::ErrorOr<BAN::RefPtr<StorageController>> ATAController::create(PCI::Device& pci_device)
 	{
-		auto* controller_ptr = new ATAController(pci_device);
+		StorageController* controller_ptr = nullptr;
+
+		switch (pci_device.subclass())
+		{
+			case 0x01:
+				controller_ptr = new ATAController(pci_device);
+				break;
+			case 0x05:
+				dwarnln("unsupported DMA ATA Controller");
+				return BAN::Error::from_errno(ENOTSUP);
+			case 0x06:
+				controller_ptr = new AHCIController(pci_device);
+				break;
+			default:
+				ASSERT_NOT_REACHED();
+		}
+
 		if (controller_ptr == nullptr)
 			return BAN::Error::from_errno(ENOMEM);
-		auto controller = BAN::RefPtr<ATAController>::adopt(controller_ptr);
+
+		auto controller = BAN::RefPtr<StorageController>::adopt(controller_ptr);
 		TRY(controller->initialize());
 		return controller;
 	}

kernel/kernel/Storage/DiskCache.cpp

@@ -1,6 +1,5 @@
 #include <BAN/ScopeGuard.h>
 #include <kernel/BootInfo.h>
-#include <kernel/Lock/LockGuard.h>
 #include <kernel/Memory/Heap.h>
 #include <kernel/Memory/PageTable.h>
 #include <kernel/Storage/DiskCache.h>

kernel/kernel/Storage/NVMe/Controller.cpp

@@ -16,12 +16,12 @@ namespace Kernel
 		return minor++;
 	}
 
-	BAN::ErrorOr<BAN::RefPtr<NVMeController>> NVMeController::create(PCI::Device& pci_device)
+	BAN::ErrorOr<BAN::RefPtr<StorageController>> NVMeController::create(PCI::Device& pci_device)
 	{
 		auto* controller_ptr = new NVMeController(pci_device);
 		if (controller_ptr == nullptr)
 			return BAN::Error::from_errno(ENOMEM);
-		auto controller = BAN::RefPtr<NVMeController>::adopt(controller_ptr);
+		auto controller = BAN::RefPtr<StorageController>::adopt(controller_ptr);
 		TRY(controller->initialize());
 		return controller;
 	}
@@ -115,6 +115,8 @@ namespace Kernel
 
 		DevFileSystem::get().add_device(this);
 
+		StorageController::ref();
+
 		return {};
 	}
 

kernel/kernel/Terminal/PseudoTerminal.cpp

@@ -157,7 +157,7 @@ namespace Kernel
 			slave->m_write_blocker.unblock();
 	}
 
-	BAN::ErrorOr<long> PseudoTerminalMaster::ioctl_impl(unsigned long request, void* argument)
+	BAN::ErrorOr<long> PseudoTerminalMaster::ioctl_impl(int request, void* argument)
 	{
 		switch (request)
 		{
@@ -174,7 +174,7 @@ namespace Kernel
 				return BAN::Error::from_errno(ENODEV);
 		}
 
-		return CharacterDevice::ioctl_impl(request, argument);
+		return CharacterDevice::ioctl(request, argument);
 	}
 
 	PseudoTerminalSlave::PseudoTerminalSlave(BAN::String&& name, uint32_t number, mode_t mode, uid_t uid, gid_t gid)

kernel/kernel/Terminal/TTY.cpp

@@ -133,7 +133,7 @@ namespace Kernel
 		(void)Process::kill(-m_foreground_pgrp, SIGWINCH);
 	}
 
-	BAN::ErrorOr<long> TTY::ioctl_impl(unsigned long request, void* argument)
+	BAN::ErrorOr<long> TTY::ioctl_impl(int request, void* argument)
 	{
 		switch (request)
 		{
@@ -204,7 +204,7 @@ namespace Kernel
 			}
 		}
 
-		return CharacterDevice::ioctl_impl(request, argument);
+		return CharacterDevice::ioctl(request, argument);
 	}
 
 	void TTY::on_key_event(LibInput::RawKeyEvent event)
@@ -449,14 +449,9 @@ namespace Kernel
 	void TTY::putchar_current(uint8_t ch)
 	{
 		ASSERT(s_tty);
-
+		LockGuard _(s_tty->m_write_lock);
-		while (!s_tty->m_write_lock.try_lock())
-			Processor::pause();
-
 		s_tty->putchar(ch);
 		s_tty->after_write();
-
-		s_tty->m_write_lock.unlock();
 	}
 
 	bool TTY::is_initialized()