Kernel: Cleanup accessing userspace memory

Instead of doing page validiation and loading manually we just do simple
memcpy and handle the possible page faults

.gitignore

@@ -3,3 +3,4 @@
 build/
 base/
 script/fakeroot-context
+tools/update-image-perms

BAN/include/BAN/Array.h

@@ -18,78 +18,78 @@ namespace BAN
 		using const_iterator = ConstIteratorSimple<T, Array>;
 
 	public:
-		Array() = default;
-		Array(const T&);
+		constexpr Array() = default;
+		constexpr Array(const T&);
 
 		iterator begin() { return iterator(m_data); }
 		iterator end() { return iterator(m_data + size()); }
 		const_iterator begin() const { return const_iterator(m_data); }
 		const_iterator end() const { return const_iterator(m_data + size()); }
 
-		const T& operator[](size_type) const;
-		T& operator[](size_type);
+		constexpr const T& operator[](size_type) const;
+		constexpr T& operator[](size_type);
 
-		const T& back() const;
-		T& back();
-		const T& front() const;
-		T& front();
+		constexpr const T& back() const;
+		constexpr T& back();
+		constexpr const T& front() const;
+		constexpr T& front();
 
 		Span<T> span() { return Span(m_data, size()); }
-		const Span<T> span() const { return Span(m_data, size()); }
+		Span<const T> span() const { return Span(m_data, size()); }
 
 		constexpr size_type size() const;
 
-		const T* data() const { return m_data; }
-		T* data() { return m_data; }
+		constexpr const T* data() const { return m_data; }
+		constexpr T* data() { return m_data; }
 
 	private:
 		T m_data[S] {};
 	};
 
 	template<typename T, size_t S>
-	Array<T, S>::Array(const T& value)
+	constexpr Array<T, S>::Array(const T& value)
 	{
 		for (size_type i = 0; i < S; i++)
 			m_data[i] = value;
 	}
 
 	template<typename T, size_t S>
-	const T& Array<T, S>::operator[](size_type index) const
+	constexpr const T& Array<T, S>::operator[](size_type index) const
 	{
 		ASSERT(index < S);
 		return m_data[index];
 	}
 
 	template<typename T, size_t S>
-	T& Array<T, S>::operator[](size_type index)
+	constexpr T& Array<T, S>::operator[](size_type index)
 	{
 		ASSERT(index < S);
 		return m_data[index];
 	}
 
 	template<typename T, size_t S>
-	const T& Array<T, S>::back() const
+	constexpr const T& Array<T, S>::back() const
 	{
 		ASSERT(S != 0);
 		return m_data[S - 1];
 	}
 
 	template<typename T, size_t S>
-	T& Array<T, S>::back()
+	constexpr T& Array<T, S>::back()
 	{
 		ASSERT(S != 0);
 		return m_data[S - 1];
 	}
 
 	template<typename T, size_t S>
-	const T& Array<T, S>::front() const
+	constexpr const T& Array<T, S>::front() const
 	{
 		ASSERT(S != 0);
 		return m_data[0];
 	}
 
 	template<typename T, size_t S>
-	T& Array<T, S>::front()
+	constexpr T& Array<T, S>::front()
 	{
 		ASSERT(S != 0);
 		return m_data[0];

BAN/include/BAN/CircularQueue.h

@@ -34,6 +34,11 @@ namespace BAN
 		const T& back() const;
 		T& back();
 
+		const T& operator[](size_t index) const;
+		T& operator[](size_t index);
+
+		void clear();
+
 		size_type size() const { return m_size; }
 		bool empty() const { return size() == 0; }
 		bool full() const { return size() == capacity(); }
@@ -53,8 +58,7 @@ namespace BAN
 	template<typename T, size_t S>
 	CircularQueue<T, S>::~CircularQueue()
 	{
-		for (size_type i = 0; i < m_size; i++)
-			element_at((m_first + i) % capacity())->~T();
+		clear();
 	}
 
 	template<typename T, size_t S>
@@ -115,6 +119,28 @@ namespace BAN
 		return *element_at((m_first + m_size - 1) % capacity());
 	}
 
+	template<typename T, size_t S>
+	const T& CircularQueue<T, S>::operator[](size_t index) const
+	{
+		ASSERT(index < m_size);
+		return *element_at((m_first + index) % capacity());
+	}
+
+	template<typename T, size_t S>
+	T& CircularQueue<T, S>::operator[](size_t index)
+	{
+		ASSERT(index < m_size);
+		return *element_at((m_first + index) % capacity());
+	}
+
+	template<typename T, size_t S>
+	void CircularQueue<T, S>::clear()
+	{
+		for (size_type i = 0; i < m_size; i++)
+			element_at((m_first + i) % capacity())->~T();
+		m_size = 0;
+	}
+
 	template<typename T, size_t S>
 	const T* CircularQueue<T, S>::element_at(size_type index) const
 	{

BAN/include/BAN/Debug.h

@@ -9,29 +9,35 @@
 #include <BAN/Formatter.h>
 #include <stdio.h>
 
-#define __debug_putchar [](int c) { putc(c, stddbg); }
+#define __debug_putchar [](int c) { putc_unlocked(c, stddbg); }
 
 #define dprintln(...)											\
 	do {														\
+		flockfile(stddbg);										\
 		BAN::Formatter::print(__debug_putchar, __VA_ARGS__);	\
-		BAN::Formatter::print(__debug_putchar,"\r\n");			\
+		BAN::Formatter::print(__debug_putchar,"\n");			\
 		fflush(stddbg);											\
+		funlockfile(stddbg);									\
 	} while (false)
 
 #define dwarnln(...)											\
 	do {														\
+		flockfile(stddbg);										\
 		BAN::Formatter::print(__debug_putchar, "\e[33m");		\
 		BAN::Formatter::print(__debug_putchar, __VA_ARGS__);	\
-		BAN::Formatter::print(__debug_putchar, "\e[m\r\n");		\
+		BAN::Formatter::print(__debug_putchar, "\e[m\n");		\
 		fflush(stddbg);											\
+		funlockfile(stddbg);									\
 	} while(false)
 
 #define derrorln(...)											\
 	do {														\
+		flockfile(stddbg);										\
 		BAN::Formatter::print(__debug_putchar, "\e[31m");		\
 		BAN::Formatter::print(__debug_putchar, __VA_ARGS__);	\
-		BAN::Formatter::print(__debug_putchar, "\e[m\r\n");		\
+		BAN::Formatter::print(__debug_putchar, "\e[m\n");		\
 		fflush(stddbg);											\
+		funlockfile(stddbg);									\
 	} while(false)
 
 #define dprintln_if(cond, ...)		\

BAN/include/BAN/Heap.h

@@ -0,0 +1,89 @@
+#pragma once
+
+#include <BAN/Iterators.h>
+#include <BAN/Swap.h>
+#include <BAN/Traits.h>
+
+namespace BAN
+{
+
+	namespace detail
+	{
+
+		template<typename It, typename Comp>
+		void heapify_up(It begin, size_t index, Comp comp)
+		{
+			size_t parent = (index - 1) / 2;
+			while (parent < index)
+			{
+				if (comp(*(begin + index), *(begin + parent)))
+					break;
+				swap(*(begin + parent), *(begin + index));
+				index = parent;
+				parent = (index - 1) / 2;
+			}
+		}
+
+		template<typename It, typename Comp>
+		void heapify_down(It begin, size_t index, size_t len, Comp comp)
+		{
+			for (;;)
+			{
+				const size_t lchild = 2 * index + 1;
+				const size_t rchild = 2 * index + 2;
+
+				size_t child = 0;
+				if (lchild < len && !comp(*(begin + lchild), *(begin + index)))
+				{
+					if (rchild < len && !comp(*(begin + rchild), *(begin + lchild)))
+						child = rchild;
+					else
+						child = lchild;
+				}
+				else if (rchild < len && !comp(*(begin + rchild), *(begin + index)))
+					child = rchild;
+				else
+					break;
+
+				swap(*(begin + child), *(begin + index));
+				index = child;
+			}
+		}
+
+	}
+
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
+	void make_heap(It begin, It end, Comp comp = {})
+	{
+		const size_t len = distance(begin, end);
+		if (len <= 1)
+			return;
+
+		size_t index = (len - 2) / 2;
+		while (index < len)
+			detail::heapify_down(begin, index--, len, comp);
+	}
+
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
+	void push_heap(It begin, It end, Comp comp = {})
+	{
+		const size_t len = distance(begin, end);
+		detail::heapify_up(begin, len - 1, comp);
+	}
+
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
+	void pop_heap(It begin, It end, Comp comp = {})
+	{
+		const size_t len = distance(begin, end);
+		swap(*begin, *(begin + len - 1));
+		detail::heapify_down(begin, 0, len - 1, comp);
+	}
+
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
+	void sort_heap(It begin, It end, Comp comp = {})
+	{
+		while (begin != end)
+			pop_heap(begin, end--, comp);
+	}
+
+}

BAN/include/BAN/IPv4.h

@@ -9,6 +9,10 @@ namespace BAN
 
 	struct IPv4Address
 	{
+		constexpr IPv4Address()
+			: IPv4Address(0)
+		{ }
+
 		constexpr IPv4Address(uint32_t u32_address)
 		{
 			raw = u32_address;

BAN/include/BAN/Math.h

@@ -36,12 +36,11 @@ namespace BAN::Math
 	template<integral T>
 	inline constexpr T gcd(T a, T b)
 	{
-		T t;
 		while (b)
 		{
-			t = b;
+			T temp = b;
 			b = a % b;
-			a = t;
+			a = temp;
 		}
 		return a;
 	}
@@ -66,25 +65,20 @@ namespace BAN::Math
 		return (x & (x - 1)) == 0;
 	}
 
-	template<BAN::integral T>
-	static constexpr bool will_multiplication_overflow(T a, T b)
+	template<integral T>
+	__attribute__((always_inline))
+	inline constexpr bool will_multiplication_overflow(T a, T b)
 	{
-		if (a == 0 || b == 0)
-			return false;
-		if ((a > 0) == (b > 0))
-			return a > BAN::numeric_limits<T>::max() / b;
-		else
-			return a < BAN::numeric_limits<T>::min() / b;
+		T dummy;
+		return __builtin_mul_overflow(a, b, &dummy);
 	}
 
-	template<BAN::integral T>
-	static constexpr bool will_addition_overflow(T a, T b)
+	template<integral T>
+	__attribute__((always_inline))
+	inline constexpr bool will_addition_overflow(T a, T b)
 	{
-		if (a > 0 && b > 0)
-			return a > BAN::numeric_limits<T>::max() - b;
-		if (a < 0 && b < 0)
-			return a < BAN::numeric_limits<T>::min() - b;
-		return false;
+		T dummy;
+		return __builtin_add_overflow(a, b, &dummy);
 	}
 
 	template<typename T>
@@ -98,6 +92,19 @@ namespace BAN::Math
 		return sizeof(T) * 8 - __builtin_clzll(x) - 1;
 	}
 
+// This is ugly but my clangd does not like including
+// intrinsic headers at all
+#if !defined(__SSE__) || !defined(__SSE2__)
+#pragma GCC push_options
+#ifndef __SSE__
+#pragma GCC target("sse")
+#endif
+#ifndef __SSE2__
+#pragma GCC target("sse2")
+#endif
+#define BAN_MATH_POP_OPTIONS
+#endif
+
 	template<floating_point T>
 	inline constexpr T floor(T x)
 	{
@@ -159,7 +166,23 @@ namespace BAN::Math
 			"jne 1b;"
 			: "+t"(a)
 			: "u"(b)
-			: "ax"
+			: "ax", "cc"
+		);
+		return a;
+	}
+
+	template<floating_point T>
+	inline constexpr T remainder(T a, T b)
+	{
+		asm(
+			"1:"
+			"fprem1;"
+			"fnstsw %%ax;"
+			"testb $4, %%ah;"
+			"jne 1b;"
+			: "+t"(a)
+			: "u"(b)
+			: "ax", "cc"
 		);
 		return a;
 	}
@@ -167,7 +190,7 @@ namespace BAN::Math
 	template<floating_point T>
 	static T modf(T x, T* iptr)
 	{
-		const T frac = BAN::Math::fmod<T>(x, 1);
+		const T frac = BAN::Math::fmod<T>(x, (T)1.0);
 		*iptr = x - frac;
 		return frac;
 	}
@@ -175,15 +198,15 @@ namespace BAN::Math
 	template<floating_point T>
 	inline constexpr T frexp(T num, int* exp)
 	{
-		if (num == 0.0)
+		if (num == (T)0.0)
 		{
 			*exp = 0;
-			return 0.0;
+			return (T)0.0;
 		}
 
-		T _exp;
-		asm("fxtract" : "+t"(num), "=u"(_exp));
-		*exp = (int)_exp + 1;
+		T e;
+		asm("fxtract" : "+t"(num), "=u"(e));
+		*exp = (int)e + 1;
 		return num / (T)2.0;
 	}
 
@@ -251,6 +274,7 @@ namespace BAN::Math
 			"fstp %%st(1);"
 			: "+t"(x)
 		);
+
 		return x;
 	}
 
@@ -263,18 +287,9 @@ namespace BAN::Math
 	template<floating_point T>
 	inline constexpr T pow(T x, T y)
 	{
-		asm(
-			"fyl2x;"
-			"fld1;"
-			"fld %%st(1);"
-			"fprem;"
-			"f2xm1;"
-			"faddp;"
-			"fscale;"
-			: "+t"(x), "+u"(y)
-		);
-
-		return x;
+		if (x == (T)0.0)
+			return (T)0.0;
+		return exp2<T>(y * log2<T>(x));
 	}
 
 	template<floating_point T>
@@ -309,17 +324,28 @@ namespace BAN::Math
 
 	template<floating_point T>
 	inline constexpr T sqrt(T x)
+	{
+		if constexpr(BAN::is_same_v<T, float>)
+		{
+			using v4sf = float __attribute__((vector_size(16)));
+			return __builtin_ia32_sqrtss((v4sf) { x, 0.0f, 0.0f, 0.0f })[0];
+		}
+		else if constexpr(BAN::is_same_v<T, double>)
+		{
+			using v2df = double __attribute__((vector_size(16)));
+			return __builtin_ia32_sqrtsd((v2df) { x, 0.0 })[0];
+		}
+		else if constexpr(BAN::is_same_v<T, long double>)
 		{
 			asm("fsqrt" : "+t"(x));
 			return x;
 		}
+	}
 
 	template<floating_point T>
 	inline constexpr T cbrt(T value)
 	{
-		if (value == 0.0)
-			return 0.0;
-		return pow<T>(value, 1.0 / 3.0);
+		return pow<T>(value, (T)1.0 / (T)3.0);
 	}
 
 	template<floating_point T>
@@ -346,30 +372,21 @@ namespace BAN::Math
 	inline constexpr T tan(T x)
 	{
 		T one, ret;
-		asm(
-			"fptan"
-			: "=t"(one), "=u"(ret)
-			: "0"(x)
-		);
+		asm("fptan" : "=t"(one), "=u"(ret) : "0"(x));
 		return ret;
 	}
 
 	template<floating_point T>
 	inline constexpr T atan2(T y, T x)
 	{
-		asm(
-			"fpatan"
-			: "+t"(x)
-			: "u"(y)
-			: "st(1)"
-		);
+		asm("fpatan" : "+t"(x) : "u"(y) : "st(1)");
 		return x;
 	}
 
 	template<floating_point T>
 	inline constexpr T atan(T x)
 	{
-		return atan2<T>(x, 1.0);
+		return atan2<T>(x, (T)1.0);
 	}
 
 	template<floating_point T>
@@ -378,10 +395,10 @@ namespace BAN::Math
 		if (x == (T)0.0)
 			return (T)0.0;
 		if (x == (T)1.0)
-			return  numbers::pi_v<T> / (T)2.0;
+			return +numbers::pi_v<T> / (T)2.0;
 		if (x == (T)-1.0)
 			return -numbers::pi_v<T> / (T)2.0;
-		return (T)2.0 * atan<T>(x / (T(1.0) + sqrt<T>((T)1.0 - x * x)));
+		return (T)2.0 * atan<T>(x / ((T)1.0 + sqrt<T>((T)1.0 - x * x)));
 	}
 
 	template<floating_point T>
@@ -411,7 +428,7 @@ namespace BAN::Math
 	template<floating_point T>
 	inline constexpr T tanh(T x)
 	{
-		const T exp_px = exp<T>(x);
+		const T exp_px = exp<T>(+x);
 		const T exp_nx = exp<T>(-x);
 		return (exp_px - exp_nx) / (exp_px + exp_nx);
 	}
@@ -440,4 +457,9 @@ namespace BAN::Math
 		return sqrt<T>(x * x + y * y);
 	}
 
+#ifdef BAN_MATH_POP_OPTIONS
+#undef BAN_MATH_POP_OPTIONS
+#pragma GCC pop_options
+#endif
+
 }

BAN/include/BAN/New.h

@@ -9,10 +9,12 @@
 namespace BAN
 {
 	#if defined(__is_kernel)
-		static constexpr void*(&allocator)(size_t) = kmalloc;
-		static constexpr void(&deallocator)(void*) = kfree;
+		static constexpr void*(*allocator)(size_t) = kmalloc;
+		static constexpr void*(*reallocator)(void*, size_t) = nullptr;
+		static constexpr void(*deallocator)(void*) = kfree;
 	#else
-		static constexpr void*(&allocator)(size_t) = malloc;
-		static constexpr void(&deallocator)(void*) = free;
+		static constexpr void*(*allocator)(size_t) = malloc;
+		static constexpr void*(*reallocator)(void*, size_t) = realloc;
+		static constexpr void(*deallocator)(void*) = free;
 	#endif
 }

BAN/include/BAN/PriorityQueue.h

@@ -0,0 +1,64 @@
+#pragma once
+
+#include "BAN/Errors.h"
+#include <BAN/Vector.h>
+#include <BAN/Heap.h>
+
+namespace BAN
+{
+
+	template<typename T, typename Comp = less<T>>
+	class PriorityQueue
+	{
+	public:
+		PriorityQueue() = default;
+		PriorityQueue(Comp comp)
+			: m_comp(comp)
+		{ }
+
+		ErrorOr<void> push(const T& value)
+		{
+			TRY(m_data.push_back(value));
+			push_heap(m_data.begin(), m_data.end());
+			return {};
+		}
+
+		ErrorOr<void> push(T&& value)
+		{
+			TRY(m_data.push_back(move(value)));
+			push_heap(m_data.begin(), m_data.end());
+			return {};
+		}
+
+		template<typename... Args>
+		ErrorOr<void> emplace(Args&&... args) requires is_constructible_v<T, Args...>
+		{
+			TRY(m_data.emplace_back(forward<Args>(args)...));
+			push_heap(m_data.begin(), m_data.end());
+			return {};
+		}
+
+		void pop()
+		{
+			pop_heap(m_data.begin(), m_data.end());
+			m_data.pop_back();
+		}
+
+		BAN::ErrorOr<void> reserve(Vector<T>::size_type size)
+		{
+			return m_data.reserve(size);
+		}
+
+		T& top() { return m_data.front(); }
+		const T& top() const { return m_data.front(); }
+
+		bool empty() const { return m_data.empty(); }
+		Vector<T>::size_type size() const { return m_data.size(); }
+		Vector<T>::size_type capacity() const { return m_data.capacity(); }
+
+	private:
+		Comp m_comp;
+		Vector<T> m_data;
+	};
+
+}

BAN/include/BAN/Sort.h

@@ -1,5 +1,6 @@
 #pragma once
 
+#include <BAN/Heap.h>
 #include <BAN/Math.h>
 #include <BAN/Swap.h>
 #include <BAN/Traits.h>
@@ -8,7 +9,7 @@
 namespace BAN::sort
 {
 
-	template<typename It, typename Comp = less<typename It::value_type>>
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
 	void exchange_sort(It begin, It end, Comp comp = {})
 	{
 		for (It lhs = begin; lhs != end; ++lhs)
@@ -42,7 +43,7 @@ namespace BAN::sort
 
 	}
 
-	template<typename It, typename Comp = less<typename It::value_type>>
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
 	void quick_sort(It begin, It end, Comp comp = {})
 	{
 		if (distance(begin, end) <= 1)
@@ -52,14 +53,14 @@ namespace BAN::sort
 		quick_sort(++mid, end, comp);
 	}
 
-	template<typename It, typename Comp = less<typename It::value_type>>
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
 	void insertion_sort(It begin, It end, Comp comp = {})
 	{
 		if (distance(begin, end) <= 1)
 			return;
 		for (It it1 = next(begin, 1); it1 != end; ++it1)
 		{
-			typename It::value_type x = move(*it1);
+			auto x = move(*it1);
 			It it2 = it1;
 			for (; it2 != begin && comp(x, *prev(it2, 1)); --it2)
 				*it2 = move(*prev(it2, 1));
@@ -67,83 +68,7 @@ namespace BAN::sort
 		}
 	}
 
-	namespace detail
-	{
-
-		template<typename It, typename Comp>
-		void push_heap(It begin, size_t hole_index, size_t top_index, typename It::value_type value, Comp comp)
-		{
-			size_t parent = (hole_index - 1) / 2;
-			while (hole_index > top_index && comp(*next(begin, parent), value))
-			{
-				*next(begin, hole_index) = move(*next(begin, parent));
-				hole_index = parent;
-				parent = (hole_index - 1) / 2;
-			}
-			*next(begin, hole_index) = move(value);
-		}
-
-		template<typename It, typename Comp>
-		void adjust_heap(It begin, size_t hole_index, size_t len, typename It::value_type value, Comp comp)
-		{
-			const size_t top_index = hole_index;
-			size_t child = hole_index;
-			while (child < (len - 1) / 2)
-			{
-				child = 2 * (child + 1);
-				if (comp(*next(begin, child), *next(begin, child - 1)))
-					child--;
-				*next(begin, hole_index) = move(*next(begin, child));
-				hole_index = child;
-			}
-			if (len % 2 == 0 && child == (len - 2) / 2)
-			{
-				child = 2 * (child + 1);
-				*next(begin, hole_index) = move(*next(begin, child - 1));
-				hole_index = child - 1;
-			}
-			push_heap(begin, hole_index, top_index, move(value), comp);
-		}
-
-	}
-
-	template<typename It, typename Comp = less<typename It::value_type>>
-	void make_heap(It begin, It end, Comp comp = {})
-	{
-		const size_t len = distance(begin, end);
-		if (len <= 1)
-			return;
-
-		size_t parent = (len - 2) / 2;
-		while (true)
-		{
-			detail::adjust_heap(begin, parent, len, move(*next(begin, parent)), comp);
-
-			if (parent == 0)
-				break;
-
-			parent--;
-		}
-	}
-
-	template<typename It, typename Comp = less<typename It::value_type>>
-	void sort_heap(It begin, It end, Comp comp = {})
-	{
-		const size_t len = distance(begin, end);
-		if (len <= 1)
-			return;
-
-		size_t last = len;
-		while (last > 1)
-		{
-			last--;
-			typename It::value_type x = move(*next(begin, last));
-			*next(begin, last) = move(*begin);
-			detail::adjust_heap(begin, 0, last, move(x), comp);
-		}
-	}
-
-	template<typename It, typename Comp = less<typename It::value_type>>
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
 	void heap_sort(It begin, It end, Comp comp = {})
 	{
 		make_heap(begin, end, comp);
@@ -167,7 +92,7 @@ namespace BAN::sort
 
 	}
 
-	template<typename It, typename Comp = less<typename It::value_type>>
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
 	void intro_sort(It begin, It end, Comp comp = {})
 	{
 		const size_t len = distance(begin, end);
@@ -190,10 +115,10 @@ namespace BAN::sort
 	}
 
 	template<typename It, size_t radix = 256>
-	requires is_unsigned_v<typename It::value_type> && (radix > 0 && (radix & (radix - 1)) == 0)
+	requires is_unsigned_v<it_value_type_t<It>> && (radix > 0 && (radix & (radix - 1)) == 0)
 	BAN::ErrorOr<void> radix_sort(It begin, It end)
 	{
-		using value_type = typename It::value_type;
+		using value_type = it_value_type_t<It>;
 
 		const size_t len = distance(begin, end);
 		if (len <= 1)
@@ -231,7 +156,7 @@ namespace BAN::sort
 		return {};
 	}
 
-	template<typename It, typename Comp = less<typename It::value_type>>
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
 	void sort(It begin, It end, Comp comp = {})
 	{
 		return intro_sort(begin, end, comp);

BAN/include/BAN/Span.h

@@ -69,7 +69,6 @@ namespace BAN
 
 		value_type* data() const
 		{
-			ASSERT(m_data);
 			return m_data;
 		}
 
@@ -84,7 +83,6 @@ namespace BAN
 
 		Span slice(size_type start, size_type length = ~size_type(0)) const
 		{
-			ASSERT(m_data);
 			ASSERT(start <= m_size);
 			if (length == ~size_type(0))
 				length = m_size - start;

BAN/include/BAN/String.h

@@ -14,6 +14,7 @@ namespace BAN
 	{
 	public:
 		using size_type = size_t;
+		using value_type = char;
 		using iterator = IteratorSimple<char, String>;
 		using const_iterator = ConstIteratorSimple<char, String>;
 		static constexpr size_type sso_capacity = 15;
@@ -352,10 +353,9 @@ namespace BAN::Formatter
 {
 
 	template<typename F>
-	void print_argument(F putc, const String& string, const ValueFormat&)
+	void print_argument(F putc, const String& string, const ValueFormat& format)
 	{
-		for (String::size_type i = 0; i < string.size(); i++)
-			putc(string[i]);
+		print_argument(putc, string.sv(), format);
 	}
 
 }

BAN/include/BAN/StringView.h

@@ -14,6 +14,7 @@ namespace BAN
 	{
 	public:
 		using size_type = size_t;
+		using value_type = char;
 		using const_iterator = ConstIteratorSimple<char, StringView>;
 
 	public:
@@ -21,7 +22,8 @@ namespace BAN
 		constexpr StringView(const char* string, size_type len = -1)
 		{
 			if (len == size_type(-1))
-				len = strlen(string);
+				for (len = 0; string[len];)
+					len++;
 			m_data = string;
 			m_size = len;
 		}
@@ -245,10 +247,12 @@ namespace BAN::Formatter
 {
 
 	template<typename F>
-	void print_argument(F putc, const StringView& sv, const ValueFormat&)
+	void print_argument(F putc, const StringView& sv, const ValueFormat& format)
 	{
 		for (StringView::size_type i = 0; i < sv.size(); i++)
 			putc(sv[i]);
+		for (int i = sv.size(); i < format.fill; i++)
+			putc(' ');
 	}
 
 }

BAN/include/BAN/Traits.h

@@ -61,6 +61,9 @@ namespace BAN
 	template<typename T> struct is_move_constructible { static constexpr bool value = is_constructible_v<T, T&&>; };
 	template<typename T> inline constexpr bool is_move_constructible_v = is_move_constructible<T>::value;
 
+	template<typename T> struct is_trivially_copyable { static constexpr bool value = __is_trivially_copyable(T); };
+	template<typename T> inline constexpr bool is_trivially_copyable_v = is_trivially_copyable<T>::value;
+
 	template<typename T> struct is_integral { static constexpr bool value = requires (T t, T* p, void (*f)(T)) { reinterpret_cast<T>(t); f(0); p + t; }; };
 	template<typename T> inline constexpr bool is_integral_v = is_integral<T>::value;
 	template<typename T> concept integral = is_integral_v<T>;
@@ -139,6 +142,10 @@ namespace BAN
 	template<typename T> using make_signed_t = typename make_signed<T>::type;
 #undef __BAN_TRAITS_MAKE_SIGNED_CV
 
+	template<typename T> struct it_value_type { using value_type = T::value_type; };
+	template<typename T> struct it_value_type<T*> { using value_type = T; };
+	template<typename T> using it_value_type_t = typename it_value_type<T>::value_type;
+
 	template<typename T> struct less	{ constexpr bool operator()(const T& lhs, const T& rhs) const { return lhs < rhs; } };
 	template<typename T> struct equal	{ constexpr bool operator()(const T& lhs, const T& rhs) const { return lhs == rhs; } };
 	template<typename T> struct greater	{ constexpr bool operator()(const T& lhs, const T& rhs) const { return lhs > rhs; } };

BAN/include/BAN/UTF8.h

@@ -18,7 +18,7 @@ namespace BAN::UTF8
 			return 3;
 		if ((first_byte & 0xF8) == 0xF0)
 			return 4;
-		return 0;
+		return UTF8::invalid;
 	}
 
 	template<typename T> requires (sizeof(T) == 1)

BAN/include/BAN/Variant.h

@@ -126,6 +126,7 @@ namespace BAN
 		Variant(Variant&& other)
 			: m_index(other.m_index)
 		{
+			if (other.has_value())
 				detail::move_construct<Ts...>(other.m_index, other.m_storage, m_storage);
 			other.clear();
 		}
@@ -133,6 +134,7 @@ namespace BAN
 		Variant(const Variant& other)
 			: m_index(other.m_index)
 		{
+			if (other.has_value())
 				detail::copy_construct<Ts...>(other.m_index, other.m_storage, m_storage);
 		}
 
@@ -157,11 +159,12 @@ namespace BAN
 
 		Variant& operator=(Variant&& other)
 		{
-			if (m_index == other.m_index)
+			if (m_index == other.m_index && m_index != invalid_index())
 				detail::move_assign<Ts...>(m_index, other.m_storage, m_storage);
 			else
 			{
 				clear();
+				if (other.has_value())
 					detail::move_construct<Ts...>(other.m_index, other.m_storage, m_storage);
 				m_index = other.m_index;
 			}
@@ -171,11 +174,12 @@ namespace BAN
 
 		Variant& operator=(const Variant& other)
 		{
-			if (m_index == other.m_index)
+			if (m_index == other.m_index && m_index != invalid_index())
 				detail::copy_assign<Ts...>(m_index, other.m_storage, m_storage);
 			else
 			{
 				clear();
+				if (other.has_value())
 					detail::copy_construct<Ts...>(other.m_index, other.m_storage, m_storage);
 				m_index = other.m_index;
 			}

BAN/include/BAN/Vector.h

@@ -56,7 +56,7 @@ namespace BAN
 		bool contains(const T&) const;
 
 		Span<T> span() { return Span(m_data, m_size); }
-		const Span<T> span() const { return Span(m_data, m_size); }
+		Span<const T> span() const { return Span(m_data, m_size); }
 
 		const T& operator[](size_type) const;
 		T& operator[](size_type);
@@ -381,10 +381,35 @@ namespace BAN
 	template<typename T>
 	ErrorOr<void> Vector<T>::ensure_capacity(size_type size)
 	{
+		static_assert(alignof(T) <= alignof(max_align_t), "over aligned types not supported");
+
 		if (m_capacity >= size)
 			return {};
-		size_type new_cap = BAN::Math::max<size_type>(size, m_capacity * 2);
-		T* new_data = (T*)BAN::allocator(new_cap * sizeof(T));
+
+		const size_type new_cap = BAN::Math::max<size_type>(size, m_capacity * 2);
+
+		if constexpr (BAN::is_trivially_copyable_v<T>)
+		{
+			if constexpr (BAN::reallocator)
+			{
+				auto* new_data = static_cast<T*>(BAN::reallocator(m_data, new_cap * sizeof(T)));
+				if (new_data == nullptr)
+					return Error::from_errno(ENOMEM);
+				m_data = new_data;
+			}
+			else
+			{
+				auto* new_data = static_cast<T*>(BAN::allocator(new_cap * sizeof(T)));
+				if (new_data == nullptr)
+					return Error::from_errno(ENOMEM);
+				memcpy(new_data, m_data, m_size * sizeof(T));
+				BAN::deallocator(m_data);
+				m_data = new_data;
+			}
+		}
+		else
+		{
+			auto* new_data = static_cast<T*>(BAN::allocator(new_cap * sizeof(T)));
 			if (new_data == nullptr)
 				return Error::from_errno(ENOMEM);
 			for (size_type i = 0; i < m_size; i++)
@@ -394,6 +419,8 @@ namespace BAN
 			}
 			BAN::deallocator(m_data);
 			m_data = new_data;
+		}
+
 		m_capacity = new_cap;
 		return {};
 	}

BAN/include/BAN/WeakPtr.h

@@ -20,7 +20,7 @@ namespace BAN
 	class WeakLink : public RefCounted<WeakLink<T>>
 	{
 	public:
-		RefPtr<T> try_lock()
+		RefPtr<T> try_lock() const
 		{
 #if __is_kernel
 			Kernel::SpinLockGuard _(m_weak_lock);
@@ -44,7 +44,7 @@ namespace BAN
 	private:
 		T* m_ptr;
 #if __is_kernel
-		Kernel::SpinLock m_weak_lock;
+		mutable Kernel::SpinLock m_weak_lock;
 #endif
 		friend class RefPtr<WeakLink<T>>;
 	};
@@ -99,7 +99,7 @@ namespace BAN
 			return *this;
 		}
 
-		RefPtr<T> lock()
+		RefPtr<T> lock() const
 		{
 			if (m_link)
 				return m_link->try_lock();

CMakeLists.txt

@@ -29,12 +29,12 @@ set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 
 # include headers of ${library} to ${target}
 function(banan_include_headers target library)
-	target_include_directories(${target} PRIVATE $<TARGET_PROPERTY:${library},SOURCE_DIR>/include)
+	target_include_directories(${target} PUBLIC $<TARGET_PROPERTY:${library},SOURCE_DIR>/include)
 endfunction()
 
 # include headers and link ${library} to ${target}
 function(banan_link_library target library)
-	target_link_libraries(${target} PRIVATE ${library})
+	target_link_libraries(${target} PUBLIC ${library})
 	banan_include_headers(${target} ${library})
 endfunction()
 

README.md

@@ -23,7 +23,7 @@ If you want to try out DOOM, you should first enter the GUI environment using th
 - [x] Basic graphical environment
   - [x] Terminal emulator
   - [x] Status bar
-  - [ ] Program launcher
+  - [x] Program launcher
   - [ ] Some nice apps
 - [x] ELF dynamic linking
 - [x] copy-on-write memory
@@ -118,6 +118,8 @@ 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 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
 ./bos image-full

kernel/CMakeLists.txt

@@ -1,11 +1,16 @@
 set(KERNEL_SOURCES
 	font/prefs.psf.o
 	kernel/ACPI/ACPI.cpp
-	kernel/ACPI/BatterySystem.cpp
 	kernel/ACPI/AML/Namespace.cpp
 	kernel/ACPI/AML/Node.cpp
 	kernel/ACPI/AML/OpRegion.cpp
+	kernel/ACPI/BatterySystem.cpp
+	kernel/ACPI/EmbeddedController.cpp
 	kernel/APIC.cpp
+	kernel/Audio/AC97/Controller.cpp
+	kernel/Audio/Controller.cpp
+	kernel/Audio/HDAudio/AudioFunctionGroup.cpp
+	kernel/Audio/HDAudio/Controller.cpp
 	kernel/BootInfo.cpp
 	kernel/CPUID.cpp
 	kernel/Credentials.cpp
@@ -17,8 +22,10 @@ set(KERNEL_SOURCES
 	kernel/Device/RandomDevice.cpp
 	kernel/Device/ZeroDevice.cpp
 	kernel/ELF.cpp
+	kernel/Epoll.cpp
 	kernel/Errors.cpp
 	kernel/FS/DevFS/FileSystem.cpp
+	kernel/FS/EventFD.cpp
 	kernel/FS/Ext2/FileSystem.cpp
 	kernel/FS/Ext2/Inode.cpp
 	kernel/FS/FAT/FileSystem.cpp
@@ -30,6 +37,7 @@ set(KERNEL_SOURCES
 	kernel/FS/ProcFS/Inode.cpp
 	kernel/FS/TmpFS/FileSystem.cpp
 	kernel/FS/TmpFS/Inode.cpp
+	kernel/FS/USTARModule.cpp
 	kernel/FS/VirtualFileSystem.cpp
 	kernel/GDT.cpp
 	kernel/IDT.cpp
@@ -42,6 +50,8 @@ set(KERNEL_SOURCES
 	kernel/Interruptable.cpp
 	kernel/InterruptController.cpp
 	kernel/kernel.cpp
+	kernel/Lock/SpinLock.cpp
+	kernel/Memory/ByteRingBuffer.cpp
 	kernel/Memory/DMARegion.cpp
 	kernel/Memory/FileBackedRegion.cpp
 	kernel/Memory/Heap.cpp
@@ -55,6 +65,7 @@ set(KERNEL_SOURCES
 	kernel/Networking/E1000/E1000.cpp
 	kernel/Networking/E1000/E1000E.cpp
 	kernel/Networking/IPv4Layer.cpp
+	kernel/Networking/Loopback.cpp
 	kernel/Networking/NetworkInterface.cpp
 	kernel/Networking/NetworkLayer.cpp
 	kernel/Networking/NetworkManager.cpp
@@ -71,7 +82,6 @@ set(KERNEL_SOURCES
 	kernel/Processor.cpp
 	kernel/Random.cpp
 	kernel/Scheduler.cpp
-	kernel/ThreadBlocker.cpp
 	kernel/SSP.cpp
 	kernel/Storage/ATA/AHCI/Controller.cpp
 	kernel/Storage/ATA/AHCI/Device.cpp
@@ -94,6 +104,7 @@ set(KERNEL_SOURCES
 	kernel/Terminal/TTY.cpp
 	kernel/Terminal/VirtualTTY.cpp
 	kernel/Thread.cpp
+	kernel/ThreadBlocker.cpp
 	kernel/Timer/HPET.cpp
 	kernel/Timer/PIT.cpp
 	kernel/Timer/RTC.cpp
@@ -101,6 +112,7 @@ set(KERNEL_SOURCES
 	kernel/USB/Controller.cpp
 	kernel/USB/Device.cpp
 	kernel/USB/HID/HIDDriver.cpp
+	kernel/USB/HID/Joystick.cpp
 	kernel/USB/HID/Keyboard.cpp
 	kernel/USB/HID/Mouse.cpp
 	kernel/USB/Hub/HubDriver.cpp
@@ -127,6 +139,8 @@ if("${BANAN_ARCH}" STREQUAL "x86_64")
 		arch/x86_64/Signal.S
 		arch/x86_64/Syscall.S
 		arch/x86_64/Thread.S
+		arch/x86_64/User.S
+		arch/x86_64/Yield.S
 	)
 elseif("${BANAN_ARCH}" STREQUAL "i686")
 	set(KERNEL_SOURCES
@@ -137,6 +151,8 @@ elseif("${BANAN_ARCH}" STREQUAL "i686")
 		arch/i686/Signal.S
 		arch/i686/Syscall.S
 		arch/i686/Thread.S
+		arch/i686/User.S
+		arch/i686/Yield.S
 	)
 else()
 	message(FATAL_ERROR "unsupported architecure ${BANAN_ARCH}")
@@ -158,6 +174,12 @@ set(KLIBC_SOURCES
 	../userspace/libraries/LibC/arch/${BANAN_ARCH}/string.S
 )
 
+set(LIBDEFLATE_SOURCE
+	../userspace/libraries/LibDEFLATE/Compressor.cpp
+	../userspace/libraries/LibDEFLATE/Decompressor.cpp
+	../userspace/libraries/LibDEFLATE/HuffmanTree.cpp
+)
+
 set(LIBFONT_SOURCES
 	../userspace/libraries/LibFont/Font.cpp
 	../userspace/libraries/LibFont/PSF.cpp
@@ -168,18 +190,25 @@ set(LIBINPUT_SOURCE
 	../userspace/libraries/LibInput/KeyEvent.cpp
 )
 
+set(LIBQR_SOURCE
+	../userspace/libraries/LibQR/QRCode.cpp
+)
+
 set(KERNEL_SOURCES
 	${KERNEL_SOURCES}
 	${BAN_SOURCES}
 	${KLIBC_SOURCES}
+	${LIBDEFLATE_SOURCE}
 	${LIBFONT_SOURCES}
 	${LIBINPUT_SOURCE}
+	${LIBQR_SOURCE}
 )
 
 add_executable(kernel ${KERNEL_SOURCES})
 
 target_compile_definitions(kernel PRIVATE __is_kernel)
 target_compile_definitions(kernel PRIVATE __arch=${BANAN_ARCH})
+target_compile_definitions(kernel PRIVATE LIBDEFLATE_AVOID_STACK=1)
 
 target_compile_options(kernel PRIVATE
 	-O2 -g
@@ -233,9 +262,11 @@ add_custom_command(
 
 banan_include_headers(kernel ban)
 banan_include_headers(kernel libc)
-banan_include_headers(kernel libfont)
+banan_include_headers(kernel libdeflate)
 banan_include_headers(kernel libelf)
+banan_include_headers(kernel libfont)
 banan_include_headers(kernel libinput)
+banan_include_headers(kernel libqr)
 
 banan_install_headers(kernel)
 set_target_properties(kernel PROPERTIES OUTPUT_NAME banan-os.kernel)

kernel/arch/i686/PageTable.cpp

@@ -206,6 +206,7 @@ namespace Kernel
 	{
 		ASSERT(s_kernel);
 		ASSERT(paddr);
+		ASSERT(paddr % PAGE_SIZE == 0);
 
 		ASSERT(s_fast_page_lock.current_processor_has_lock());
 
@@ -220,7 +221,7 @@ namespace Kernel
 		ASSERT(!(pt[pte] & Flags::Present));
 		pt[pte] = paddr | Flags::ReadWrite | Flags::Present;
 
-		invalidate(fast_page(), false);
+		asm volatile("invlpg (%0)" :: "r"(fast_page()) : "memory");
 	}
 
 	void PageTable::unmap_fast_page()
@@ -240,7 +241,7 @@ namespace Kernel
 		ASSERT(pt[pte] & Flags::Present);
 		pt[pte] = 0;
 
-		invalidate(fast_page(), false);
+		asm volatile("invlpg (%0)" :: "r"(fast_page()) : "memory");
 	}
 
 	BAN::ErrorOr<PageTable*> PageTable::create_userspace()
@@ -313,7 +314,8 @@ namespace Kernel
 				.type = Processor::SMPMessage::Type::FlushTLB,
 				.flush_tlb = {
 					.vaddr      = vaddr,
-					.page_count = 1
+					.page_count = 1,
+					.page_table = vaddr < KERNEL_OFFSET ? this : nullptr,
 				}
 			});
 		}
@@ -342,7 +344,11 @@ namespace Kernel
 		uint64_t* pd   = reinterpret_cast<uint64_t*>(P2V(pdpt[pdpte] & PAGE_ADDR_MASK));
 		uint64_t* pt   = reinterpret_cast<uint64_t*>(P2V(pd[pde] & PAGE_ADDR_MASK));
 
+		const paddr_t old_paddr = pt[pte] & PAGE_ADDR_MASK;
+
 		pt[pte] = 0;
+
+		if (old_paddr != 0)
 			invalidate(vaddr, send_smp_message);
 	}
 
@@ -360,7 +366,8 @@ namespace Kernel
 			.type = Processor::SMPMessage::Type::FlushTLB,
 			.flush_tlb = {
 				.vaddr      = vaddr,
-				.page_count = page_count
+				.page_count = page_count,
+				.page_table = vaddr < KERNEL_OFFSET ? this : nullptr,
 			}
 		});
 	}
@@ -416,8 +423,12 @@ namespace Kernel
 			uwr_flags &= ~Flags::Present;
 
 		uint64_t* pt = reinterpret_cast<uint64_t*>(P2V(pd[pde] & PAGE_ADDR_MASK));
+
+		const paddr_t old_paddr = pt[pte] & PAGE_ADDR_MASK;
+
 		pt[pte] = paddr | uwr_flags | extra_flags;
 
+		if (old_paddr != 0)
 			invalidate(vaddr, send_smp_message);
 	}
 
@@ -437,7 +448,8 @@ namespace Kernel
 			.type = Processor::SMPMessage::Type::FlushTLB,
 			.flush_tlb = {
 				.vaddr      = vaddr,
-				.page_count = page_count
+				.page_count = page_count,
+				.page_table = vaddr < KERNEL_OFFSET ? this : nullptr,
 			}
 		});
 	}
@@ -496,13 +508,13 @@ namespace Kernel
 		return true;
 	}
 
-	bool PageTable::reserve_page(vaddr_t vaddr, bool only_free)
+	bool PageTable::reserve_page(vaddr_t vaddr, bool only_free, bool send_smp_message)
 	{
 		SpinLockGuard _(m_lock);
 		ASSERT(vaddr % PAGE_SIZE == 0);
 		if (only_free && !is_page_free(vaddr))
 			return false;
-		map_page_at(0, vaddr, Flags::Reserved);
+		map_page_at(0, vaddr, Flags::Reserved, MemoryType::Normal, send_smp_message);
 		return true;
 	}
 
@@ -516,7 +528,15 @@ namespace Kernel
 		if (only_free && !is_range_free(vaddr, bytes))
 			return false;
 		for (size_t offset = 0; offset < bytes; offset += PAGE_SIZE)
-			reserve_page(vaddr + offset);
+			reserve_page(vaddr + offset, true, false);
+		Processor::broadcast_smp_message({
+			.type = Processor::SMPMessage::Type::FlushTLB,
+			.flush_tlb = {
+				.vaddr      = vaddr,
+				.page_count = bytes / PAGE_SIZE,
+				.page_table = vaddr < KERNEL_OFFSET ? this : nullptr,
+			}
+		});
 		return true;
 	}
 

kernel/arch/i686/Signal.S

@@ -2,33 +2,72 @@
 
 // stack contains
 //    return address
+//    return stack
+//    return rflags
+//    siginfo_t
 //    signal number
 //    signal handler
 
 .global signal_trampoline
 signal_trampoline:
+	pushl %esi // gregs
+	pushl %edi
+	pushl %edx
+	pushl %ecx
+	pushl %ebx
+	pushl %eax
 	pushl %ebp
-	movl %esp, %ebp
 
-	pusha
+	// FIXME: populate these
+	xorl %eax, %eax
+	pushl %eax // stack
+	pushl %eax
+	pushl %eax
+	pushl %eax // sigset
+	pushl %eax
+	pushl %eax // link
 
-	movl 40(%esp), %edi
-	movl 36(%esp), %eax
+	movl    %esp,  %edx // ucontext
+	leal 60(%esp), %esi // siginfo
+	movl 56(%esp), %edi // signal number
+	movl 52(%esp), %eax // handlers
 
 	// align stack to 16 bytes
 	movl %esp, %ebp
 	andl $-16, %esp
 
-	subl $12, %esp
+	subl $512, %esp
+	fxsave (%esp)
+
+	subl $4, %esp
+	pushl %edx
+	pushl %esi
 	pushl %edi
 	call *%eax
 	addl $16, %esp
 
+	fxrstor (%esp)
+	addl $512, %esp
+
 	// restore stack
 	movl %ebp, %esp
-	popa
+	addl $24, %esp
 
-	leave
-	addl $8, %esp
+	// restore registers
+	popl %ebp
+	popl %eax
+	popl %ebx
+	popl %ecx
+	popl %edx
+	popl %edi
+	popl %esi
+
+	// skip handler, number, siginfo_t
+	addl $44, %esp
+
+	// restore flags
+	popf
+
+	movl (%esp), %esp
 
 	ret

kernel/arch/i686/Syscall.S

@@ -1,12 +1,6 @@
 // arguments in EAX, EBX, ECX, EDX, ESI, EDI
 .global asm_syscall_handler
 asm_syscall_handler:
-	# save segment registers
-	pushw %ds
-	pushw %es
-	pushw %fs
-	pushw %gs
-
 	# save general purpose registers
 	pushl %ebx
 	pushl %ecx
@@ -18,13 +12,10 @@ asm_syscall_handler:
 
 	# align stack
 	movl %esp, %ebp
-	subl $15, %esp
-	andl $0xFFFFFFF0, %esp
+	andl $-16, %esp
 
 	# push arguments
-	subl $4, %esp
-	pushl %ebp
-	addl $32, (%esp)
+	subl $8, %esp
 	pushl %edi
 	pushl %esi
 	pushl %edx
@@ -44,6 +35,15 @@ asm_syscall_handler:
 
 	movl %ebp, %esp
 
+	# restore userspace segments
+	movw $(0x20 | 3), %bx
+	movw %bx, %ds
+	movw %bx, %es
+	movw $(0x30 | 3), %bx
+	movw %bx, %fs
+	movw $(0x38 | 3), %bx
+	movw %bx, %gs
+
 	# restore general purpose registers
 	popl %ebp
 	popl %esi
@@ -52,12 +52,6 @@ asm_syscall_handler:
 	popl %ecx
 	popl %ebx
 
-	# restore segment registers
-	popw %gs
-	popw %fs
-	popw %es
-	popw %ds
-
 	iret
 
 .global sys_fork_trampoline

kernel/arch/i686/Thread.S

@@ -31,25 +31,18 @@ start_kernel_thread:
 	subl $12, %esp
 	pushl %edi
 	call *%esi
-	addl $16, %esp
-
 
 .global start_userspace_thread
 start_userspace_thread:
-	call load_thread_sse
-
 	call get_thread_start_sp
 	movl %eax, %esp
 
-	# ds, es = user data
 	movw $(0x20 | 3), %bx
 	movw %bx, %ds
 	movw %bx, %es
-	# gs = thread local
 	movw $(0x30 | 3), %bx
-	movw %bx, %gs
-	# fs = 0
-	xorw %bx, %bx
 	movw %bx, %fs
+	movw $(0x38 | 3), %bx
+	movw %bx, %gs
 
 	iret

kernel/arch/i686/User.S

@@ -0,0 +1,54 @@
+# bool safe_user_memcpy(void*, const void*, size_t)
+.global safe_user_memcpy
+.global safe_user_memcpy_end
+.global safe_user_memcpy_fault
+safe_user_memcpy:
+	xorl %eax, %eax
+	xchgl 4(%esp), %edi
+	xchgl 8(%esp), %esi
+	movl 12(%esp), %ecx
+	movl %edi, %edx
+	rep movsb
+	movl 4(%esp), %edi
+	movl 8(%esp), %esi
+	incl %eax
+safe_user_memcpy_fault:
+	ret
+safe_user_memcpy_end:
+
+# bool safe_user_strncpy(void*, const void*, size_t)
+.global safe_user_strncpy
+.global safe_user_strncpy_end
+.global safe_user_strncpy_fault
+safe_user_strncpy:
+	xchgl 4(%esp), %edi
+	xchgl 8(%esp), %esi
+	movl 12(%esp), %ecx
+
+	testl %ecx, %ecx
+	jz safe_user_strncpy_fault
+
+ .safe_user_strncpy_loop:
+	movb (%esi), %al
+	movb %al, (%edi)
+	testb %al, %al
+	jz .safe_user_strncpy_done
+
+	incl %edi
+	incl %esi
+	decl %ecx
+	jnz .safe_user_strncpy_loop
+
+ safe_user_strncpy_fault:
+	xorl %eax, %eax
+	jmp .safe_user_strncpy_return
+
+ .safe_user_strncpy_done:
+	movl $1, %eax
+
+ .safe_user_strncpy_return:
+	movl 4(%esp), %edi
+	movl 8(%esp), %esi
+	ret
+
+safe_user_strncpy_end:

kernel/arch/i686/Yield.S

@@ -0,0 +1,25 @@
+.global asm_yield_trampoline
+asm_yield_trampoline:
+	leal 4(%esp), %ecx
+	movl 4(%esp), %esp
+
+	pushl -4(%ecx)
+	pushl %ecx
+	pushl %eax
+	pushl %ebx
+	pushl %esi
+	pushl %edi
+	pushl %ebp
+
+	pushl %esp
+	call scheduler_on_yield
+	addl $4, %esp
+
+	popl %ebp
+	popl %edi
+	popl %esi
+	popl %ebx
+	popl %eax
+	movl 4(%esp), %ecx
+	movl 0(%esp), %esp
+	jmp *%ecx

kernel/arch/i686/boot.S

@@ -14,6 +14,23 @@
 # multiboot2 header
 .section .multiboot, "aw"
 	.align 8
+multiboot_start:
+	.long 0x1BADB002
+	.long (1 << 2) # page align modules
+	.long -(0x1BADB002 + (1 << 2))
+
+	.long 0
+	.long 0
+	.long 0
+	.long 0
+	.long 0
+
+	.long 0
+	.long FB_WIDTH
+	.long FB_HEIGHT
+	.long FB_BPP
+multiboot_end:
+	.align 8
 multiboot2_start:
 	.long 0xE85250D6
 	.long 0
@@ -36,6 +53,12 @@ multiboot2_start:
 	.long 12
 	.long V2P(_start)
 
+	# page align modules
+	.align 8
+	.short 6
+	.short 0
+	.long 8
+
 	.align 8
 	.short 0
 	.short 0
@@ -53,10 +76,10 @@ bananboot_start:
 bananboot_end:
 
 .section .bss, "aw", @nobits
-	.align 4096
-	boot_stack_bottom:
+    .global g_boot_stack_top
+	g_boot_stack_bottom:
 		.skip 4096 * 4
-	boot_stack_top:
+	g_boot_stack_top:
 
 	.global g_kernel_cmdline
 	g_kernel_cmdline:
@@ -161,6 +184,13 @@ enable_sse:
 	movl %eax, %cr4
 	ret
 
+enable_tsc:
+	# allow userspace to use RDTSC
+	movl %cr4, %ecx
+	andl $0xFFFFFFFB, %ecx
+	movl %ecx, %cr4
+	ret
+
 initialize_paging:
 	# enable PAE
 	movl %cr4, %ecx
@@ -188,7 +218,7 @@ _start:
 	movl %ebx, V2P(bootloader_info)
 
 	# load boot stack
-	movl $V2P(boot_stack_top), %esp
+	movl $V2P(g_boot_stack_top), %esp
 
 	# load boot GDT
 	lgdt V2P(boot_gdtr)
@@ -203,10 +233,11 @@ gdt_flush:
 	# do processor initialization
 	call check_requirements
 	call enable_sse
+	call enable_tsc
 	call initialize_paging
 
 	# load higher half stack pointer
-	movl $boot_stack_top, %esp
+	movl $g_boot_stack_top, %esp
 
 	# jump to higher half
 	leal higher_half, %ecx
@@ -279,6 +310,7 @@ ap_protected_mode:
 	movb $1, AP_V2P(ap_stack_loaded)
 
 	leal V2P(enable_sse),        %ecx; call *%ecx
+	leal V2P(enable_tsc),        %ecx; call *%ecx
 	leal V2P(initialize_paging), %ecx; call *%ecx
 
 	# load boot gdt and enter long mode

kernel/arch/i686/interrupts.S

@@ -1,32 +1,33 @@
-.macro push_userspace
-	pushw %gs
-	pushw %fs
-	pushw %es
-	pushw %ds
-	pushal
-.endm
+.macro maybe_load_kernel_segments, n
+	testb $3, \n(%esp)
+	jz 1f; jnp 1f
 
-.macro load_kernel_segments
 	movw $0x10, %ax
 	movw %ax, %ds
 	movw %ax, %es
 	movw %ax, %fs
-
 	movw $0x28, %ax
 	movw %ax, %gs
+1:
 .endm
 
-.macro pop_userspace
-	popal
-	popw %ds
-	popw %es
-	popw %fs
-	popw %gs
+.macro maybe_load_userspace_segments, n
+	testb $3, \n(%esp)
+	jz 1f; jnp 1f
+
+	movw $(0x20 | 3), %bx
+	movw %bx, %ds
+	movw %bx, %es
+	movw $(0x30 | 3), %bx
+	movw %bx, %fs
+	movw $(0x38 | 3), %bx
+	movw %bx, %gs
+1:
 .endm
 
 isr_stub:
-	push_userspace
-	load_kernel_segments
+	pushal
+	maybe_load_kernel_segments 44
 	cld
 
 	movl %cr0, %eax; pushl %eax
@@ -34,33 +35,39 @@ isr_stub:
 	movl %cr3, %eax; pushl %eax
 	movl %cr4, %eax; pushl %eax
 
-	movl %esp,     %eax // register ptr
-	leal 64(%esp), %ebx // interrupt stack ptr
-	movl 60(%esp), %ecx // error code
-	movl 56(%esp), %edx // isr number
+	movl 48(%esp), %edi // isr number
+	movl 52(%esp), %esi // error code
+	leal 56(%esp), %edx // interrupt stack ptr
+	movl %esp,     %ecx // register ptr
+
+	# stack frame for stack trace
+	leal 56(%esp), %eax
+	pushl (%eax)
+	pushl %ebp
 
 	movl %esp, %ebp
 	andl $-16, %esp
 
-	pushl %eax
-	pushl %ebx
 	pushl %ecx
 	pushl %edx
+	pushl %esi
+	pushl %edi
 	call cpp_isr_handler
 
 	movl %ebp, %esp
-	addl $16, %esp
+	addl $24, %esp
 
-	pop_userspace
+	maybe_load_userspace_segments 44
+	popal
 	addl $8, %esp
 	iret
 
 irq_stub:
-	push_userspace
-	load_kernel_segments
+	pushal
+	maybe_load_kernel_segments 44
 	cld
 
-	movl 40(%esp), %edi # interrupt number
+	movl 32(%esp), %edi # interrupt number
 
 	movl %esp, %ebp
 	andl $-16, %esp
@@ -71,36 +78,15 @@ irq_stub:
 
 	movl %ebp, %esp
 
-	pop_userspace
-	addl $8, %esp
-	iret
-
-.global asm_yield_handler
-asm_yield_handler:
-	# This can only be called from kernel, so no segment saving is needed
-	pushal
-	cld
-
-	leal 32(%esp), %edi # interrupt stack ptr
-	movl %esp,     %esi # interrupt registers ptr
-
-	movl %esp, %ebp
-	andl $-16, %esp
-
-	subl $8, %esp
-	pushl %esi
-	pushl %edi
-	call cpp_yield_handler
-
-	movl %ebp, %esp
-
+	maybe_load_userspace_segments 44
 	popal
+	addl $8, %esp
 	iret
 
 .global asm_ipi_handler
 asm_ipi_handler:
-	push_userspace
-	load_kernel_segments
+	pushal
+	maybe_load_kernel_segments 36
 	cld
 
 	movl %esp, %ebp
@@ -110,13 +96,14 @@ asm_ipi_handler:
 
 	movl %ebp, %esp
 
-	pop_userspace
+	maybe_load_userspace_segments 36
+	popal
 	iret
 
 .global asm_timer_handler
 asm_timer_handler:
-	push_userspace
-	load_kernel_segments
+	pushal
+	maybe_load_kernel_segments 36
 	cld
 
 	movl %esp, %ebp
@@ -126,7 +113,8 @@ asm_timer_handler:
 
 	movl %ebp, %esp
 
-	pop_userspace
+	maybe_load_userspace_segments 36
+	popal
 	iret
 
 .macro isr n

kernel/arch/x86_64/PageTable.cpp

@@ -503,6 +503,7 @@ namespace Kernel
 	{
 		ASSERT(s_kernel);
 		ASSERT(paddr);
+		ASSERT(paddr % PAGE_SIZE == 0);
 
 		ASSERT(s_fast_page_lock.current_processor_has_lock());
 
@@ -520,7 +521,7 @@ namespace Kernel
 		ASSERT(!(pt[pte] & Flags::Present));
 		pt[pte] = paddr | Flags::ReadWrite | Flags::Present;
 
-		invalidate(fast_page(), false);
+		asm volatile("invlpg (%0)" :: "r"(fast_page()) : "memory");
 	}
 
 	void PageTable::unmap_fast_page()
@@ -543,7 +544,7 @@ namespace Kernel
 		ASSERT(pt[pte] & Flags::Present);
 		pt[pte] = 0;
 
-		invalidate(fast_page(), false);
+		asm volatile("invlpg (%0)" :: "r"(fast_page()) : "memory");
 	}
 
 	BAN::ErrorOr<PageTable*> PageTable::create_userspace()
@@ -622,7 +623,8 @@ namespace Kernel
 				.type = Processor::SMPMessage::Type::FlushTLB,
 				.flush_tlb = {
 					.vaddr      = vaddr,
-					.page_count = 1
+					.page_count = 1,
+					.page_table = vaddr < KERNEL_OFFSET ? this : nullptr,
 				}
 			});
 		}
@@ -657,7 +659,11 @@ namespace Kernel
 		uint64_t* pd   = P2V(pdpt[pdpte] & s_page_addr_mask);
 		uint64_t* pt   = P2V(pd[pde]     & s_page_addr_mask);
 
+		const paddr_t old_paddr = pt[pte] & PAGE_ADDR_MASK;
+
 		pt[pte] = 0;
+
+		if (old_paddr != 0)
 			invalidate(vaddr, send_smp_message);
 	}
 
@@ -675,7 +681,8 @@ namespace Kernel
 			.type = Processor::SMPMessage::Type::FlushTLB,
 			.flush_tlb = {
 				.vaddr      = vaddr,
-				.page_count = page_count
+				.page_count = page_count,
+				.page_table = vaddr < KERNEL_OFFSET ? this : nullptr,
 			}
 		});
 	}
@@ -741,8 +748,11 @@ namespace Kernel
 		if (!(flags & Flags::Present))
 			uwr_flags &= ~Flags::Present;
 
+		const paddr_t old_paddr = pt[pte] & PAGE_ADDR_MASK;
+
 		pt[pte] = paddr | uwr_flags | extra_flags;
 
+		if (old_paddr != 0)
 			invalidate(vaddr, send_smp_message);
 	}
 
@@ -764,7 +774,8 @@ namespace Kernel
 			.type = Processor::SMPMessage::Type::FlushTLB,
 			.flush_tlb = {
 				.vaddr      = vaddr,
-				.page_count = page_count
+				.page_count = page_count,
+				.page_table = vaddr < KERNEL_OFFSET ? this : nullptr,
 			}
 		});
 	}
@@ -813,13 +824,13 @@ namespace Kernel
 		return page_data & s_page_addr_mask;
 	}
 
-	bool PageTable::reserve_page(vaddr_t vaddr, bool only_free)
+	bool PageTable::reserve_page(vaddr_t vaddr, bool only_free, bool send_smp_message)
 	{
 		SpinLockGuard _(m_lock);
 		ASSERT(vaddr % PAGE_SIZE == 0);
 		if (only_free && !is_page_free(vaddr))
 			return false;
-		map_page_at(0, vaddr, Flags::Reserved);
+		map_page_at(0, vaddr, Flags::Reserved, MemoryType::Normal, send_smp_message);
 		return true;
 	}
 
@@ -833,7 +844,15 @@ namespace Kernel
 		if (only_free && !is_range_free(vaddr, bytes))
 			return false;
 		for (size_t offset = 0; offset < bytes; offset += PAGE_SIZE)
-			reserve_page(vaddr + offset);
+			reserve_page(vaddr + offset, true, false);
+		Processor::broadcast_smp_message({
+			.type = Processor::SMPMessage::Type::FlushTLB,
+			.flush_tlb = {
+				.vaddr      = vaddr,
+				.page_count = bytes / PAGE_SIZE,
+				.page_table = vaddr < KERNEL_OFFSET ? this : nullptr,
+			}
+		});
 		return true;
 	}
 
@@ -847,7 +866,7 @@ namespace Kernel
 			last_address -= rem;
 
 		ASSERT(is_canonical(first_address));
-		ASSERT(is_canonical(last_address));
+		ASSERT(is_canonical(last_address - 1));
 		const vaddr_t uc_vaddr_start = uncanonicalize(first_address);
 		const vaddr_t uc_vaddr_end   = uncanonicalize(last_address);
 
@@ -929,7 +948,7 @@ namespace Kernel
 			last_address -= rem;
 
 		ASSERT(is_canonical(first_address));
-		ASSERT(is_canonical(last_address));
+		ASSERT(is_canonical(last_address - 1));
 
 		SpinLockGuard _(m_lock);
 

kernel/arch/x86_64/Signal.S

@@ -2,55 +2,83 @@
 
 // stack contains
 //    return address
+//    return stack
+//    return rflags
+//    siginfo_t
 //    signal number
 //    signal handler
 
 .global signal_trampoline
 signal_trampoline:
-	pushq %rax
-	pushq %rbx
-	pushq %rcx
-	pushq %rdx
-	pushq %rbp
-	pushq %rdi
-	pushq %rsi
-	pushq %r8
-	pushq %r9
-	pushq %r10
-	pushq %r11
-	pushq %r12
-	pushq %r13
+	pushq %r15 // gregs
 	pushq %r14
-	pushq %r15
+	pushq %r13
+	pushq %r12
+	pushq %r11
+	pushq %r10
+	pushq %r9
+	pushq %r8
+	pushq %rsi
+	pushq %rdi
+	pushq %rdx
+	pushq %rcx
+	pushq %rbx
+	pushq %rax
+	pushq %rbp
 
-	movq 128(%rsp), %rdi
-	movq 120(%rsp), %rax
+	// FIXME: populate these
+	xorq %rax, %rax
+	pushq %rax // stack
+	pushq %rax
+	pushq %rax
+	pushq %rax // sigset
+	pushq %rax // link
+
+	movq  %rsp,     %rdx // ucontext
+	leaq 176(%rsp), %rsi // siginfo
+	movq 168(%rsp), %rdi // signal number
+	movq 160(%rsp), %rax // handler
 
 	// align stack to 16 bytes
 	movq %rsp, %rbp
 	andq $-16, %rsp
 
+	subq $512, %rsp
+	fxsave64 (%rsp)
+
 	call *%rax
 
+	fxrstor64 (%rsp)
+	addq $512, %rsp
+
 	// restore stack
 	movq %rbp, %rsp
-	popq %r15
-	popq %r14
-	popq %r13
-	popq %r12
-	popq %r11
-	popq %r10
-	popq %r9
-	popq %r8
-	popq %rsi
-	popq %rdi
+	addq $40, %rsp
+
+	// restore registers
 	popq %rbp
-	popq %rdx
-	popq %rcx
-	popq %rbx
 	popq %rax
+	popq %rbx
+	popq %rcx
+	popq %rdx
+	popq %rdi
+	popq %rsi
+	popq %r8
+	popq %r9
+	popq %r10
+	popq %r11
+	popq %r12
+	popq %r13
+	popq %r14
+	popq %r15
 
-	addq $16, %rsp
+	// skip handler, number, siginfo_t
+	addq $72, %rsp
 
-	// return over red-zone
+	// restore flags
+	popfq
+
+	movq (%rsp), %rsp
+
+	// return over red-zone and siginfo_t
 	ret $128

kernel/arch/x86_64/Syscall.S

@@ -1,49 +1,26 @@
-// arguments in RAX, RBX, RCX, RDX, RSI, RDI
-// System V ABI: RDI, RSI, RDX, RCX, R8, R9
 .global asm_syscall_handler
 asm_syscall_handler:
-	pushq %rbx
-	pushq %rcx
-	pushq %rdx
-	pushq %rdi
-	pushq %rsi
-	pushq %rbp
-	pushq %r8
-	pushq %r9
-	pushq %r10
+	swapgs
+
+	movq %rsp, %rax
+	movq %gs:8, %rsp
+
+	pushq $(0x20 | 3)
+	pushq %rax
 	pushq %r11
-	pushq %r12
-	pushq %r13
-	pushq %r14
-	pushq %r15
-	cld
+	pushq $(0x28 | 3)
+	pushq %rcx
+	subq $8, %rsp
 
-	movq %rsi, %r8
-	movq %rdi, %r9
-	movq %rax, %rdi
-	movq %rbx, %rsi
-	xchgq %rcx, %rdx
-	leaq 112(%rsp), %rbx
-	pushq %rbx
+	movq %r10, %rcx
 	call cpp_syscall_handler
-	addq $8, %rsp
 
-	popq %r15
-	popq %r14
-	popq %r13
-	popq %r12
-	popq %r11
-	popq %r10
-	popq %r9
-	popq %r8
-	popq %rbp
-	popq %rsi
-	popq %rdi
-	popq %rdx
-	popq %rcx
-	popq %rbx
-	iretq
+	movq  8(%rsp), %rcx
+	movq 24(%rsp), %r11
+	movq 32(%rsp), %rsp
 
+	swapgs
+	sysretq
 
 .global sys_fork_trampoline
 sys_fork_trampoline:
@@ -56,7 +33,7 @@ sys_fork_trampoline:
 
 	call read_ip
 	testq %rax, %rax
-	je .reload_stack
+	je .done
 
 	movq %rax, %rsi
 	movq %rsp, %rdi
@@ -70,9 +47,3 @@ sys_fork_trampoline:
 	popq %rbp
 	popq %rbx
 	ret
-
- .reload_stack:
-	call get_thread_start_sp
-	movq %rax, %rsp
-	xorq %rax, %rax
-	jmp .done

kernel/arch/x86_64/Thread.S

@@ -27,9 +27,9 @@ start_kernel_thread:
 
 .global start_userspace_thread
 start_userspace_thread:
-	call load_thread_sse
-
 	call get_thread_start_sp
 	movq %rax, %rsp
 
+	swapgs
+
 	iretq

kernel/arch/x86_64/User.S

@@ -0,0 +1,87 @@
+# bool safe_user_memcpy(void*, const void*, size_t)
+.global safe_user_memcpy
+.global safe_user_memcpy_end
+.global safe_user_memcpy_fault
+safe_user_memcpy:
+	xorq %rax, %rax
+	movq %rdx, %rcx
+	rep movsb
+	incq %rax
+safe_user_memcpy_fault:
+	ret
+safe_user_memcpy_end:
+
+# bool safe_user_strncpy(void*, const void*, size_t)
+.global safe_user_strncpy
+.global safe_user_strncpy_end
+.global safe_user_strncpy_fault
+safe_user_strncpy:
+	movq %rdx, %rcx
+	testq %rcx, %rcx
+	jz safe_user_strncpy_fault
+
+ .safe_user_strncpy_align_loop:
+	testb $0x7, %sil
+	jz .safe_user_strncpy_align_done
+
+	movb (%rsi), %al
+	movb %al, (%rdi)
+	testb %al, %al
+	jz .safe_user_strncpy_done
+
+	incq %rdi
+	incq %rsi
+	decq %rcx
+	jnz .safe_user_strncpy_align_loop
+	jmp safe_user_strncpy_fault
+
+ .safe_user_strncpy_align_done:
+	movq $0x0101010101010101, %r8
+	movq $0x8080808080808080, %r9
+
+ .safe_user_strncpy_qword_loop:
+	cmpq $8, %rcx
+	jb .safe_user_strncpy_qword_done
+
+	movq (%rsi), %rax
+	movq %rax, %r10
+	movq %rax, %r11
+
+	# https://graphics.stanford.edu/~seander/bithacks.html#ZeroInWord
+	subq %r8, %r10
+	notq %r11
+	andq %r11, %r10
+	andq %r9, %r10
+	jnz .safe_user_strncpy_byte_loop
+
+	movq %rax, (%rdi)
+
+	addq $8, %rdi
+	addq $8, %rsi
+	subq $8, %rcx
+	jnz .safe_user_strncpy_qword_loop
+	jmp safe_user_strncpy_fault
+
+ .safe_user_strncpy_qword_done:
+	testq %rcx, %rcx
+	jz safe_user_strncpy_fault
+
+ .safe_user_strncpy_byte_loop:
+	movb (%rsi), %al
+	movb %al, (%rdi)
+	testb %al, %al
+	jz .safe_user_strncpy_done
+
+	incq %rdi
+	incq %rsi
+	decq %rcx
+	jnz .safe_user_strncpy_byte_loop
+
+ safe_user_strncpy_fault:
+	xorq %rax, %rax
+	ret
+
+ .safe_user_strncpy_done:
+	movb $1, %al
+	ret
+safe_user_strncpy_end:

kernel/arch/x86_64/Yield.S

@@ -0,0 +1,29 @@
+.global asm_yield_trampoline
+asm_yield_trampoline:
+	leaq 8(%rsp), %rcx
+	movq %rdi, %rsp
+
+	subq $8, %rsp
+	pushq -8(%rcx)
+	pushq %rcx
+	pushq %rax
+	pushq %rbx
+	pushq %rbp
+	pushq %r12
+	pushq %r13
+	pushq %r14
+	pushq %r15
+
+	movq %rsp, %rdi
+	call scheduler_on_yield
+
+	popq %r15
+	popq %r14
+	popq %r13
+	popq %r12
+	popq %rbp
+	popq %rbx
+	popq %rax
+	movq 8(%rsp), %rcx
+	movq 0(%rsp), %rsp
+	jmp *%rcx

kernel/arch/x86_64/boot.S

@@ -14,6 +14,23 @@
 # multiboot2 header
 .section .multiboot, "aw"
 	.align 8
+multiboot_start:
+	.long 0x1BADB002
+	.long (1 << 2) # page align modules
+	.long -(0x1BADB002 + (1 << 2))
+
+	.long 0
+	.long 0
+	.long 0
+	.long 0
+	.long 0
+
+	.long 0
+	.long FB_WIDTH
+	.long FB_HEIGHT
+	.long FB_BPP
+multiboot_end:
+	.align 8
 multiboot2_start:
 	.long 0xE85250D6
 	.long 0
@@ -36,6 +53,12 @@ multiboot2_start:
 	.long 12
 	.long V2P(_start)
 
+	# page align modules
+	.align 8
+	.short 6
+	.short 0
+	.long 8
+
 	.align 8
 	.short 0
 	.short 0
@@ -53,9 +76,10 @@ bananboot_start:
 bananboot_end:
 
 .section .bss, "aw", @nobits
-	boot_stack_bottom:
-		.skip 4096 * 64
-	boot_stack_top:
+    .global g_boot_stack_top
+	g_boot_stack_bottom:
+		.skip 4096 * 4
+	g_boot_stack_top:
 
 	.global g_kernel_cmdline
 	g_kernel_cmdline:
@@ -155,6 +179,13 @@ enable_sse:
 	movl %eax, %cr4
 	ret
 
+enable_tsc:
+	# allow userspace to use RDTSC
+	movl %cr4, %ecx
+	andl $0xFFFFFFFB, %ecx
+	movl %ecx, %cr4
+	ret
+
 initialize_paging:
 	# enable PAE
 	movl %cr4, %ecx
@@ -187,10 +218,11 @@ _start:
 	movl %eax, V2P(bootloader_magic)
 	movl %ebx, V2P(bootloader_info)
 
-	movl $V2P(boot_stack_top), %esp
+	movl $V2P(g_boot_stack_top), %esp
 
 	call check_requirements
 	call enable_sse
+	call enable_tsc
 	call initialize_paging
 
 	# flush gdt and jump to 64 bit
@@ -277,6 +309,7 @@ ap_protected_mode:
 	movb $1, AP_V2P(ap_stack_loaded)
 
 	leal V2P(enable_sse),        %ecx; call *%ecx
+	leal V2P(enable_tsc),        %ecx; call *%ecx
 	leal V2P(initialize_paging), %ecx; call *%ecx
 
 	# load boot gdt and enter long mode

kernel/arch/x86_64/interrupts.S

@@ -1,4 +1,12 @@
-.macro pushaq
+.macro swapgs_if_necessary, n
+	testb $3, \n(%rsp)
+	jz 1f; jnp 1f
+	swapgs
+1:
+.endm
+
+.macro pushaq, n
+	swapgs_if_necessary \n
 	pushq %rax
 	pushq %rcx
 	pushq %rdx
@@ -16,7 +24,7 @@
 	pushq %r15
 .endm
 
-.macro popaq
+.macro popaq, n
 	popq %r15
 	popq %r14
 	popq %r13
@@ -32,10 +40,11 @@
 	popq %rdx
 	popq %rcx
 	popq %rax
+	swapgs_if_necessary \n
 .endm
 
 isr_stub:
-	pushaq
+	pushaq 24
 	cld
 	movq %cr0, %rax; pushq %rax
 	movq %cr2, %rax; pushq %rax
@@ -49,43 +58,33 @@ isr_stub:
 	call cpp_isr_handler
 	addq $32, %rsp
 
-	popaq
+	popaq 24
 	addq $16, %rsp
 	iretq
 
 irq_stub:
-	pushaq
+	pushaq 24
 	cld
 	movq 120(%rsp), %rdi	# irq number
 	call cpp_irq_handler
-	popaq
+	popaq 24
 	addq $16, %rsp
 	iretq
 
-.global asm_yield_handler
-asm_yield_handler:
-	pushaq
-	cld
-	leaq 120(%rsp), %rdi # interrupt stack ptr
-	movq     %rsp,  %rsi # interrupt register ptr
-	call cpp_yield_handler
-	popaq
-	iretq
-
 .global asm_ipi_handler
 asm_ipi_handler:
-	pushaq
+	pushaq 8
 	cld
 	call cpp_ipi_handler
-	popaq
+	popaq 8
 	iretq
 
 .global asm_timer_handler
 asm_timer_handler:
-	pushaq
+	pushaq 8
 	cld
 	call cpp_timer_handler
-	popaq
+	popaq 8
 	iretq
 
 .macro isr n

kernel/include/kernel/ACPI/ACPI.h

@@ -2,6 +2,7 @@
 
 #include <BAN/Vector.h>
 #include <kernel/ACPI/AML/Namespace.h>
+#include <kernel/ACPI/EmbeddedController.h>
 #include <kernel/ACPI/Headers.h>
 #include <kernel/Memory/Types.h>
 #include <kernel/ThreadBlocker.h>
@@ -35,8 +36,12 @@ namespace Kernel::ACPI
 		BAN::ErrorOr<void> poweroff();
 		BAN::ErrorOr<void> reset();
 
+		BAN::ErrorOr<void> register_gpe_handler(uint8_t gpe, void (*callback)(void*), void* argument);
+
 		void handle_irq() override;
 
+		BAN::Span<BAN::UniqPtr<EmbeddedController>> embedded_controllers() { return m_embedded_controllers.span(); }
+
 	private:
 		ACPI() = default;
 		BAN::ErrorOr<void> initialize_impl();
@@ -50,6 +55,12 @@ namespace Kernel::ACPI
 
 		BAN::ErrorOr<void> route_interrupt_link_device(const AML::Scope& device, uint64_t& routed_irq_mask);
 
+		BAN::ErrorOr<void> initialize_embedded_controller(const AML::Scope& embedded_controller);
+		BAN::ErrorOr<void> initialize_embedded_controllers();
+
+		BAN::Optional<GAS> find_gpe_block(size_t index);
+		bool enable_gpe(uint8_t gpe);
+
 	private:
 		paddr_t m_header_table_paddr = 0;
 		vaddr_t m_header_table_vaddr = 0;
@@ -68,8 +79,23 @@ namespace Kernel::ACPI
 
 		ThreadBlocker m_event_thread_blocker;
 
+		BAN::Vector<BAN::UniqPtr<EmbeddedController>> m_embedded_controllers;
+
+		struct GPEHandler
+		{
+			bool has_callback { false };
+			union {
+				AML::Reference* method;
+				struct
+				{
+					void (*callback)(void*);
+					void* argument;
+				};
+			};
+		};
+		bool m_has_any_gpes { false };
 		AML::Scope m_gpe_scope;
-		BAN::Array<AML::Reference*, 0xFF> m_gpe_methods { nullptr };
+		BAN::Array<GPEHandler, 0xFF> m_gpe_methods;
 
 		bool m_hardware_reduced { false };
 		AML::Namespace* m_namespace { nullptr };

kernel/include/kernel/ACPI/AML/Namespace.h

@@ -50,6 +50,7 @@ namespace Kernel::ACPI::AML
 		BAN::ErrorOr<void> for_each_child(const Scope&, const BAN::Function<BAN::Iteration(const Scope&, Reference*)>&);
 
 		BAN::ErrorOr<BAN::Vector<Scope>> find_device_with_eisa_id(BAN::StringView eisa_id);
+		BAN::ErrorOr<BAN::Vector<Scope>> find_device_with_eisa_id(BAN::Span<BAN::StringView> eisa_ids);
 
 	private:
 		BAN::ErrorOr<Scope> resolve_path(const Scope& scope, const NameString& name_string);

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

@@ -89,6 +89,9 @@ namespace Kernel::ACPI::AML
 		GAS::AddressSpaceID address_space;
 		uint64_t offset;
 		uint64_t length;
+		alignas(Scope) uint8_t scope_storage[sizeof(Scope)];
+		Scope& scope() { return *reinterpret_cast<Scope*>(scope_storage); }
+		const Scope& scope() const { return *reinterpret_cast<const Scope*>(scope_storage); }
 	};
 
 	struct FieldUnit

kernel/include/kernel/ACPI/EmbeddedController.h

@@ -0,0 +1,69 @@
+#pragma once
+
+#include <BAN/Atomic.h>
+#include <BAN/UniqPtr.h>
+
+#include <kernel/ACPI/AML/Scope.h>
+#include <kernel/Lock/Mutex.h>
+#include <kernel/Thread.h>
+
+namespace Kernel::ACPI
+{
+
+	class EmbeddedController
+	{
+	public:
+		static BAN::ErrorOr<BAN::UniqPtr<EmbeddedController>> create(AML::Scope&& scope, uint16_t command_port, uint16_t data_port, BAN::Optional<uint8_t> gpe);
+		~EmbeddedController();
+
+		BAN::ErrorOr<uint8_t> read_byte(uint8_t offset);
+		BAN::ErrorOr<void> write_byte(uint8_t offset, uint8_t value);
+
+		const AML::Scope& scope() const { return m_scope; }
+
+	private:
+		EmbeddedController(AML::Scope&& scope, uint16_t command_port, uint16_t data_port, bool has_gpe)
+			: m_scope(BAN::move(scope))
+			, m_command_port(command_port)
+			, m_data_port(data_port)
+			, m_has_gpe(has_gpe)
+		{ }
+
+	private:
+		void wait_status_bit(uint8_t bit, uint8_t value);
+
+		uint8_t read_one(uint16_t port);
+		void write_one(uint16_t port, uint8_t value);
+
+		static void handle_gpe_wrapper(void*);
+		void handle_gpe();
+
+		BAN::ErrorOr<void> call_query_method(uint8_t notification);
+
+		void thread_task();
+
+		struct Command
+		{
+			uint8_t command;
+			BAN::Optional<uint8_t> data1;
+			BAN::Optional<uint8_t> data2;
+			uint8_t* response;
+			BAN::Atomic<bool> done;
+		};
+		BAN::ErrorOr<void> send_command(Command& command);
+
+	private:
+		const AML::Scope m_scope;
+		const uint16_t m_command_port;
+		const uint16_t m_data_port;
+		const bool m_has_gpe;
+
+		Mutex m_mutex;
+		ThreadBlocker m_thread_blocker;
+
+		BAN::Optional<Command*> m_queued_command;
+
+		Thread* m_thread { nullptr };
+	};
+
+}

kernel/include/kernel/ACPI/Resource.h

@@ -0,0 +1,179 @@
+#pragma once
+
+#include <BAN/ByteSpan.h>
+#include <BAN/Debug.h>
+#include <BAN/Optional.h>
+
+namespace Kernel::ACPI
+{
+
+	struct ResourceData
+	{
+		enum class Type
+		{
+			IRQ,
+			DMA,
+			IOPort,
+			FixedIOPort,
+			FixedDMA,
+
+			// TODO: large stuff the stuff :)
+		};
+
+		union {
+			struct {
+				uint16_t irq_mask;
+				union {
+					struct {
+						uint8_t edge_triggered : 1;
+						uint8_t                : 2;
+						uint8_t active_low     : 1;
+						uint8_t shared         : 1;
+						uint8_t wake_capable   : 1;
+						uint8_t                : 2;
+					};
+					uint8_t raw;
+				};
+			} irq;
+			struct {
+				uint8_t channel_mask;
+				union {
+					struct {
+						uint8_t type          : 2; // 0: 8 bit, 1: 8 and 16 bit, 2: 16 bit only
+						uint8_t bus_master    : 1;
+						uint8_t               : 2;
+						uint8_t channel_speed : 2; // 0: compatibility, 1: type A, 2: type B, 3: type F
+						uint8_t               : 1;
+					};
+					uint8_t raw;
+				};
+			} dma;
+			struct {
+				uint16_t range_min_base;
+				uint16_t range_max_base;
+				uint8_t base_alignment;
+				uint8_t range_length;
+			} io_port;
+			struct {
+				uint16_t range_base;
+				uint8_t range_length;
+			} fixed_io_port;
+			struct {
+				uint16_t request_line;
+				uint16_t channel;
+				uint8_t transfer_width; // 0: 8 bit, 1: 16 bit, 2: 32 bit, 3: 64 bit, 4: 128 bit
+			} fixed_dma;
+		} as;
+		Type type;
+	};
+
+	class ResourceParser
+	{
+	public:
+		ResourceParser(BAN::ConstByteSpan buffer)
+			: m_buffer(buffer)
+		{}
+
+		BAN::Optional<ResourceData> get_next()
+		{
+			for (;;)
+			{
+				if (m_buffer.empty())
+					return {};
+
+				if (m_buffer[0] & 0x80)
+				{
+					dprintln("Skipping large resource 0x{2H}", m_buffer[0] & 0x7F);
+					const uint16_t length = (m_buffer[2] << 8) | m_buffer[1];
+					if (m_buffer.size() < static_cast<size_t>(3 + length))
+						return {};
+					m_buffer = m_buffer.slice(3 + length);
+					continue;
+				}
+
+				const uint8_t length = m_buffer[0] & 0x07;
+				if (m_buffer.size() < static_cast<size_t>(1 + length))
+					return {};
+
+				BAN::Optional<ResourceData> result;
+				switch ((m_buffer[0] >> 3) & 0x0F)
+				{
+					case 0x04:
+						if (length < 2)
+							break;
+						result = ResourceData {
+							.as = { .irq = {
+								.irq_mask = static_cast<uint16_t>((m_buffer[2] << 8) | m_buffer[1]),
+								.raw = (length >= 3) ? m_buffer[3] : static_cast<uint8_t>(1),
+							}},
+							.type = ResourceData::Type::IRQ,
+						};
+						break;
+					case 0x05:
+						if (length < 2)
+							break;
+						result = ResourceData {
+							.as = { .dma = {
+								.channel_mask = m_buffer[1],
+								.raw = m_buffer[2],
+							}},
+							.type = ResourceData::Type::DMA,
+						};
+						break;
+					case 0x08:
+						if (length < 7)
+							break;
+						result = ResourceData {
+							.as = { .io_port = {
+								.range_min_base = static_cast<uint16_t>(((m_buffer[3] << 8) | m_buffer[2]) & ((m_buffer[1] & 1) ? 0xFFFF : 0x03FF)),
+								.range_max_base = static_cast<uint16_t>(((m_buffer[5] << 8) | m_buffer[4]) & ((m_buffer[1] & 1) ? 0xFFFF : 0x03FF)),
+								.base_alignment = m_buffer[6],
+								.range_length = m_buffer[7],
+							}},
+							.type = ResourceData::Type::IOPort,
+						};
+						break;
+					case 0x09:
+						if (length < 3)
+							break;
+						result = ResourceData {
+							.as = { .fixed_io_port = {
+								.range_base = static_cast<uint16_t>(((m_buffer[2] << 8) | m_buffer[1]) & 0x03FF),
+								.range_length = m_buffer[3],
+							}},
+							.type = ResourceData::Type::FixedIOPort,
+						};
+						break;
+					case 0x0A:
+						if (length < 5)
+							break;
+						result = ResourceData {
+							.as = { .fixed_dma = {
+								.request_line = static_cast<uint16_t>((m_buffer[2] << 8) | m_buffer[1]),
+								.channel      = static_cast<uint16_t>((m_buffer[4] << 8) | m_buffer[3]),
+								.transfer_width = m_buffer[5],
+							}},
+							.type = ResourceData::Type::FixedDMA,
+						};
+						break;
+					case 0x0F:
+						// End tag
+						return {};
+					case 0x06:
+					case 0x07:
+					case 0x0E:
+						dprintln("Skipping short resource 0x{2H}", (m_buffer[0] >> 3) & 0x0F);
+						break;
+				}
+
+				m_buffer = m_buffer.slice(1 + length);
+				if (result.has_value())
+					return result.release_value();
+			}
+		}
+
+	private:
+		BAN::ConstByteSpan m_buffer;
+	};
+
+}

kernel/include/kernel/API/SharedPage.h

@@ -0,0 +1,37 @@
+#pragma once
+
+#include <stdint.h>
+
+namespace Kernel::API
+{
+
+	enum SharedPageFeature : uint32_t
+	{
+		SPF_GETTIME = 1 << 0,
+	};
+
+	struct SharedPage
+	{
+		uint8_t __sequence[0x100];
+
+		uint32_t features;
+
+		struct
+		{
+			uint8_t shift;
+			uint64_t mult;
+			uint64_t realtime_seconds;
+		} gettime_shared;
+
+		struct
+		{
+			struct
+			{
+				uint32_t seq;
+				uint64_t last_ns;
+				uint64_t last_tsc;
+			} gettime_local;
+		} cpus[];
+	};
+
+}

kernel/include/kernel/API/Syscall.h

@@ -0,0 +1,44 @@
+#pragma once
+
+#include <kernel/Attributes.h>
+#include <kernel/IDT.h>
+#include <stdint.h>
+#include <sys/syscall.h>
+
+namespace Kernel
+{
+
+	ALWAYS_INLINE long syscall(int syscall, uintptr_t arg1 = 0, uintptr_t arg2 = 0, uintptr_t arg3 = 0, uintptr_t arg4 = 0, uintptr_t arg5 = 0)
+	{
+		long ret;
+#if ARCH(x86_64)
+		register uintptr_t r10 asm("r10") = arg3;
+		register uintptr_t r8  asm( "r8") = arg4;
+		register uintptr_t r9  asm( "r9") = arg5;
+		asm volatile(
+			"syscall"
+			: "=a"(ret)
+			, "+D"(syscall)
+			, "+S"(arg1)
+			, "+d"(arg2)
+			, "+r"(r10)
+			, "+r"(r8)
+			, "+r"(r9)
+			:: "rcx", "r11", "memory");
+#elif ARCH(i686)
+		asm volatile(
+			"int %[irq]"
+			: "=a"(ret)
+			: [irq]"i"(static_cast<int>(IRQ_SYSCALL)) // WTF GCC 15
+			, "a"(syscall)
+			, "b"(arg1)
+			, "c"(arg2)
+			, "d"(arg3)
+			, "S"(arg4)
+			, "D"(arg5)
+			: "memory");
+#endif
+		return ret;
+	}
+
+}

kernel/include/kernel/Audio/AC97/Controller.h

@@ -0,0 +1,59 @@
+#pragma once
+
+#include <kernel/Audio/Controller.h>
+#include <kernel/Memory/DMARegion.h>
+
+namespace Kernel
+{
+
+	class AC97AudioController final : public AudioController, public Interruptable
+	{
+	public:
+		static BAN::ErrorOr<void> create(PCI::Device& pci_device);
+
+		void handle_irq() override;
+
+	protected:
+		void handle_new_data() override;
+
+		uint32_t get_channels() const override { return 2; }
+		uint32_t get_sample_rate() const override { return 48000; }
+
+		uint32_t get_total_pins() const override { return 1; }
+		uint32_t get_current_pin() const override { return 0; }
+		BAN::ErrorOr<void> set_current_pin(uint32_t pin) override { if (pin != 0) return BAN::Error::from_errno(EINVAL); return {}; }
+
+		BAN::ErrorOr<void> set_volume_mdB(int32_t) override;
+
+	private:
+		AC97AudioController(PCI::Device& pci_device)
+			: m_pci_device(pci_device)
+		{ }
+
+		uint32_t get_volume_data() const;
+
+		BAN::ErrorOr<void> initialize();
+		BAN::ErrorOr<void> initialize_bld();
+		BAN::ErrorOr<void> initialize_interrupts();
+
+		bool queue_samples_to_bld();
+
+	private:
+		static constexpr size_t m_bdl_entries = 32;
+		static constexpr size_t m_samples_per_entry = 0x1000;
+
+		// We only store samples in 2 BDL entries at a time to reduce the amount of samples queued.
+		// This is to reduce latency as you cannot remove data already passed to the BDLs
+		static constexpr size_t m_used_bdl_entries = 2;
+
+		PCI::Device& m_pci_device;
+		BAN::UniqPtr<PCI::BarRegion> m_mixer;
+		BAN::UniqPtr<PCI::BarRegion> m_bus_master;
+
+		BAN::UniqPtr<DMARegion> m_bdl_region;
+
+		uint32_t m_bdl_tail { 0 };
+		uint32_t m_bdl_head { 0 };
+	};
+
+}

kernel/include/kernel/Audio/AC97/Definitions.h

@@ -0,0 +1,15 @@
+#pragma once
+
+#include <stdint.h>
+
+namespace Kernel::AC97
+{
+
+	struct BufferDescriptorListEntry
+	{
+		uint32_t address;
+		uint16_t samples;
+		uint16_t flags; // bit 14: last entry, bit 15: IOC
+	};
+
+}

kernel/include/kernel/Audio/Controller.h

@@ -0,0 +1,58 @@
+#pragma once
+
+#include <kernel/Device/Device.h>
+#include <kernel/Memory/ByteRingBuffer.h>
+#include <kernel/PCI.h>
+
+#include <sys/ioctl.h>
+
+namespace Kernel
+{
+
+	class AudioController : public CharacterDevice
+	{
+	public:
+		static BAN::ErrorOr<void> create(PCI::Device& pci_device);
+
+		dev_t rdev() const override { return m_rdev; }
+		BAN::StringView name() const override { return m_name; }
+
+	protected:
+		AudioController();
+		BAN::ErrorOr<void> initialize();
+
+		virtual void handle_new_data() = 0;
+
+		virtual uint32_t get_channels() const = 0;
+		virtual uint32_t get_sample_rate() const = 0;
+
+		virtual uint32_t get_total_pins() const = 0;
+		virtual uint32_t get_current_pin() const = 0;
+		virtual BAN::ErrorOr<void> set_current_pin(uint32_t) = 0;
+
+		virtual BAN::ErrorOr<void> set_volume_mdB(int32_t) = 0;
+
+		bool can_read_impl() const override { return false; }
+		bool can_write_impl() const override { SpinLockGuard _(m_spinlock); return !m_sample_data->full(); }
+		bool has_error_impl() const override { return false; }
+		bool has_hungup_impl() const override { return false; }
+
+		BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override;
+
+		BAN::ErrorOr<long> ioctl_impl(int cmd, void* arg) override;
+
+	protected:
+		ThreadBlocker m_sample_data_blocker;
+		mutable SpinLock m_spinlock;
+
+		static constexpr size_t m_sample_data_capacity = 1 << 20;
+		BAN::UniqPtr<ByteRingBuffer> m_sample_data;
+
+		snd_volume_info m_volume_info {};
+
+	private:
+		const dev_t m_rdev;
+		char m_name[10] {};
+	};
+
+}

kernel/include/kernel/Audio/HDAudio/AudioFunctionGroup.h

@@ -0,0 +1,80 @@
+#pragma once
+
+#include <kernel/Audio/Controller.h>
+#include <kernel/Audio/HDAudio/Controller.h>
+
+namespace Kernel
+{
+
+	class HDAudioController;
+
+	class HDAudioFunctionGroup final : public AudioController
+	{
+	public:
+		static BAN::ErrorOr<BAN::RefPtr<HDAudioFunctionGroup>> create(BAN::RefPtr<HDAudioController>, uint8_t cid, HDAudio::AFGNode&&);
+
+		void on_stream_interrupt(uint8_t stream_index);
+
+	protected:
+		// FIXME: allow setting these :D
+		uint32_t get_channels() const override { return 2; }
+		uint32_t get_sample_rate() const override { return 48000; }
+
+		uint32_t get_total_pins() const override;
+		uint32_t get_current_pin() const override;
+		BAN::ErrorOr<void> set_current_pin(uint32_t) override;
+
+		BAN::ErrorOr<void> set_volume_mdB(int32_t) override;
+
+		void handle_new_data() override;
+
+	private:
+		HDAudioFunctionGroup(BAN::RefPtr<HDAudioController> controller, uint8_t cid, HDAudio::AFGNode&& afg_node)
+			: m_controller(controller)
+			, m_afg_node(BAN::move(afg_node))
+			, m_cid(cid)
+		{ }
+		~HDAudioFunctionGroup();
+
+		BAN::ErrorOr<void> initialize();
+		BAN::ErrorOr<void> initialize_stream();
+		BAN::ErrorOr<void> initialize_output();
+
+		BAN::ErrorOr<void> enable_output_path(uint8_t index);
+		BAN::ErrorOr<void> disable_output_path(uint8_t index);
+
+		void reset_stream();
+
+		BAN::ErrorOr<void> recurse_output_paths(const HDAudio::AFGWidget& widget, BAN::Vector<const HDAudio::AFGWidget*>& path);
+
+		uint16_t get_format_data() const;
+
+		size_t bdl_offset() const;
+
+		void queue_bdl_data();
+
+	private:
+		// use 6 512 sample BDL entries
+		// each entry is ~10.7 ms at 48 kHz
+		// -> total buffered audio is 64 ms
+		static constexpr size_t m_bdl_entry_sample_frames = 512;
+		static constexpr size_t m_bdl_entry_count         = 6;
+
+		BAN::RefPtr<HDAudioController> m_controller;
+		const HDAudio::AFGNode m_afg_node;
+		const uint8_t m_cid;
+
+		BAN::Vector<BAN::Vector<const HDAudio::AFGWidget*>> m_output_paths;
+		BAN::Vector<const HDAudio::AFGWidget*> m_output_pins;
+		size_t m_output_path_index { SIZE_MAX };
+
+		uint8_t m_stream_id    { 0xFF };
+		uint8_t m_stream_index { 0xFF };
+		BAN::UniqPtr<DMARegion> m_bdl_region;
+
+		size_t m_bdl_head { 0 };
+		size_t m_bdl_tail { 0 };
+		bool m_stream_running { false };
+	};
+
+}

kernel/include/kernel/Audio/HDAudio/Controller.h

@@ -0,0 +1,77 @@
+#pragma once
+
+#include <kernel/Audio/Controller.h>
+#include <kernel/Audio/HDAudio/Definitions.h>
+#include <kernel/Memory/DMARegion.h>
+
+namespace Kernel
+{
+
+	class HDAudioController : public Interruptable, public BAN::RefCounted<HDAudioController>
+	{
+	public:
+		static BAN::ErrorOr<void> create(PCI::Device& pci_device);
+
+		BAN::ErrorOr<uint32_t> send_command(HDAudio::CORBEntry);
+
+		uint8_t get_stream_index(HDAudio::StreamType type, uint8_t index) const;
+
+		BAN::ErrorOr<uint8_t> allocate_stream_id();
+		void deallocate_stream_id(uint8_t id);
+
+		BAN::ErrorOr<uint8_t> allocate_stream(HDAudio::StreamType type, void* afg);
+		void deallocate_stream(uint8_t index);
+
+		PCI::BarRegion& bar0() { return *m_bar0; }
+
+		bool is_64bit() const { return m_is64bit; }
+
+		void handle_irq() override;
+
+	private:
+		HDAudioController(PCI::Device& pci_device)
+			: m_pci_device(pci_device)
+		{ }
+
+		BAN::ErrorOr<void> initialize();
+		BAN::ErrorOr<void> initialize_ring_buffers();
+
+		BAN::ErrorOr<void> reset_controller();
+
+		BAN::ErrorOr<HDAudio::Codec> initialize_codec(uint8_t codec);
+		BAN::ErrorOr<HDAudio::AFGNode> initialize_node(uint8_t codec, uint8_t node);
+		BAN::ErrorOr<HDAudio::AFGWidget> initialize_widget(uint8_t codec, uint8_t node);
+
+	private:
+		struct RingBuffer
+		{
+			vaddr_t vaddr;
+			uint32_t index;
+			uint32_t size;
+		};
+
+	private:
+		PCI::Device& m_pci_device;
+		BAN::UniqPtr<PCI::BarRegion> m_bar0;
+		bool m_is64bit { false };
+
+		bool m_use_immediate_command { false };
+
+		uint8_t m_output_streams { 0 };
+		uint8_t m_input_streams { 0 };
+		uint8_t m_bidir_streams { 0 };
+		void* m_allocated_streams[30] {};
+
+		// NOTE: stream ids are from 1 to 15
+		uint16_t m_allocated_stream_ids { 0 };
+
+		Mutex m_command_mutex;
+		SpinLock m_rb_lock;
+		ThreadBlocker m_rb_blocker;
+
+		RingBuffer m_corb;
+		RingBuffer m_rirb;
+		BAN::UniqPtr<DMARegion> m_ring_buffer_region;
+	};
+
+}

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

@@ -0,0 +1,90 @@
+#pragma once
+
+#include <BAN/Vector.h>
+
+namespace Kernel::HDAudio
+{
+
+	struct CORBEntry
+	{
+		union {
+			struct {
+				uint32_t data          : 8;
+				uint32_t command       : 12;
+				uint32_t node_index    : 8;
+				uint32_t codec_address : 4;
+			};
+			uint32_t raw;
+		};
+	};
+	static_assert(sizeof(CORBEntry) == sizeof(uint32_t));
+
+	struct BDLEntry
+	{
+		paddr_t address;
+		uint32_t length;
+		uint32_t ioc;
+	};
+	static_assert(sizeof(BDLEntry) == 16);
+
+	struct AFGWidget
+	{
+		enum class Type
+		{
+			OutputConverter,
+			InputConverter,
+			Mixer,
+			Selector,
+			PinComplex,
+			Power,
+			VolumeKnob,
+			BeepGenerator,
+		};
+
+		Type type;
+		uint8_t id;
+
+		union
+		{
+			struct
+			{
+				bool input;
+				bool output;
+				bool display; // HDMI or DP
+				uint32_t config;
+			} pin_complex;
+		};
+
+		struct Amplifier
+		{
+			uint8_t offset;
+			uint8_t num_steps;
+			uint8_t step_size;
+			bool mute;
+		};
+
+		BAN::Optional<Amplifier> output_amplifier;
+
+		BAN::Vector<uint16_t> connections;
+	};
+
+	struct AFGNode
+	{
+		uint8_t id;
+		BAN::Vector<AFGWidget> widgets;
+	};
+
+	struct Codec
+	{
+		uint8_t id;
+		BAN::Vector<AFGNode> nodes;
+	};
+
+	enum class StreamType
+	{
+		Input,
+		Output,
+		Bidirectional,
+	};
+
+}

kernel/include/kernel/Audio/HDAudio/Registers.h

@@ -0,0 +1,50 @@
+#pragma once
+
+#include <stdint.h>
+
+namespace Kernel::HDAudio
+{
+
+	enum Regs : uint8_t
+	{
+		GCAP = 0x00,
+		VMIN = 0x02,
+		VMAJ = 0x03,
+		GCTL = 0x08,
+		STATESTS = 0x0E,
+
+		INTCTL = 0x20,
+		INTSTS = 0x24,
+
+		CORBLBASE = 0x40,
+		CORBUBASE = 0x44,
+		CORBWP    = 0x48,
+		CORBRP    = 0x4A,
+		CORBCTL   = 0x4C,
+		CORBSTS   = 0x4D,
+		CORBSIZE  = 0x4E,
+
+		RIRBLBASE = 0x50,
+		RIRBUBASE = 0x54,
+		RIRBWP    = 0x58,
+		RINTCNT   = 0x5A,
+		RIRBCTL   = 0x5C,
+		RIRBSTS   = 0x5D,
+		RIRBSIZE  = 0x5E,
+
+		ICOI = 0x60,
+		ICII = 0x64,
+		ICIS = 0x68,
+
+		SDCTL   = 0x00,
+		SDSTS   = 0x03,
+		SDLPIB  = 0x04,
+		SDCBL   = 0x08,
+		SDLVI   = 0x0C,
+		SDFIFOD = 0x10,
+		SDFMT   = 0x12,
+		SDBDPL  = 0x18,
+		SDBDPU  = 0x1C,
+	};
+
+}

kernel/include/kernel/BootInfo.h

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

kernel/include/kernel/CPUID.h

@@ -81,5 +81,6 @@ namespace CPUID
 	bool has_pge();
 	bool has_pat();
 	bool has_1gib_pages();
+	bool has_invariant_tsc();
 
 }

kernel/include/kernel/Credentials.h

@@ -35,10 +35,8 @@ namespace Kernel
 
 		bool has_egid(gid_t) const;
 
-		BAN::ErrorOr<void> initialize_supplementary_groups();
-
-	private:
-		BAN::ErrorOr<BAN::String> find_username() const;
+		BAN::Span<const gid_t> groups() const { return m_supplementary.span(); }
+		BAN::ErrorOr<void> set_groups(BAN::Span<const gid_t> groups);
 
 	private:
 		uid_t m_ruid, m_euid, m_suid;

kernel/include/kernel/Debug.h

@@ -70,10 +70,15 @@
 #define DEBUG_USB_MOUSE 0
 #define DEBUG_USB_MASS_STORAGE 0
 
+#define DEBUG_HDAUDIO 0
+
 
 namespace Debug
 {
 	void dump_stack_trace();
+	void dump_stack_trace(uintptr_t ip, uintptr_t bp);
+	void dump_qr_code();
+
 	void putchar(char);
 	void print_prefix(const char*, int);
 

kernel/include/kernel/Device/DebugDevice.h

@@ -3,7 +3,7 @@
 namespace Kernel
 {
 
-	class DebugDevice : public CharacterDevice
+	class DebugDevice final : public CharacterDevice
 	{
 	public:
 		static BAN::ErrorOr<BAN::RefPtr<DebugDevice>> create(mode_t, uid_t, gid_t);
@@ -24,6 +24,7 @@ namespace Kernel
 		virtual bool can_read_impl() const override { return false; }
 		virtual bool can_write_impl() const override { return true; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		const dev_t m_rdev;

kernel/include/kernel/Device/Device.h

@@ -16,7 +16,11 @@ namespace Kernel
 		virtual bool is_partition() const { return false; }
 		virtual bool is_storage_device() const { return false; }
 
-		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> mmap_region(PageTable&, off_t offset, size_t len, AddressRange, MemoryRegion::Type, PageTable::flags_t) { (void)offset; (void)len; return BAN::Error::from_errno(EINVAL); }
+		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> mmap_region(PageTable&, off_t offset, size_t len, AddressRange, MemoryRegion::Type, PageTable::flags_t, int status_flags)
+		{
+			(void)offset; (void)len; (void)status_flags;
+			return BAN::Error::from_errno(ENOTSUP);
+		}
 
 		virtual dev_t rdev() const override = 0;
 

kernel/include/kernel/Device/DeviceNumbers.h

@@ -16,11 +16,14 @@ namespace Kernel
 		Debug,
 		Keyboard,
 		Mouse,
+		Joystick,
 		SCSI,
 		NVMeController,
 		NVMeNamespace,
 		Ethernet,
+		Loopback,
 		TmpFS,
+		AudioController,
 	};
 
 }

kernel/include/kernel/Device/FramebufferDevice.h

@@ -10,6 +10,7 @@ namespace Kernel
 	{
 	public:
 		static BAN::ErrorOr<BAN::RefPtr<FramebufferDevice>> create_from_boot_framebuffer();
+		static BAN::RefPtr<FramebufferDevice> boot_framebuffer();
 		~FramebufferDevice();
 
 		uint32_t width() const { return m_width; }
@@ -17,6 +18,7 @@ namespace Kernel
 
 		uint32_t get_pixel(uint32_t x, uint32_t y) const;
 		void set_pixel(uint32_t x, uint32_t y, uint32_t rgb);
+		void fill(uint32_t rgb);
 
 		// positive rows -> empty pixels on bottom
 		// negative rows -> empty pixels on top
@@ -26,7 +28,7 @@ namespace Kernel
 		void sync_pixels_linear(uint32_t first_pixel, uint32_t pixel_count);
 		void sync_pixels_rectangle(uint32_t top_right_x, uint32_t top_right_y, uint32_t width, uint32_t height);
 
-		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> mmap_region(PageTable&, off_t offset, size_t len, AddressRange, MemoryRegion::Type, PageTable::flags_t) override;
+		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> mmap_region(PageTable&, off_t offset, size_t len, AddressRange, MemoryRegion::Type, PageTable::flags_t, int status_flags) override;
 
 		virtual dev_t rdev() const override { return m_rdev; }
 		virtual BAN::StringView name() const override { return m_name.sv(); }
@@ -38,6 +40,7 @@ namespace Kernel
 		virtual bool can_read_impl() const override { return true; }
 		virtual bool can_write_impl() const override { return true; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		FramebufferDevice(mode_t mode, uid_t uid, gid_t gid, dev_t rdev, paddr_t paddr, uint32_t width, uint32_t height, uint32_t pitch, uint8_t bpp);

kernel/include/kernel/Device/NullDevice.h

@@ -5,7 +5,7 @@
 namespace Kernel
 {
 
-	class NullDevice : public CharacterDevice
+	class NullDevice final : public CharacterDevice
 	{
 	public:
 		static BAN::ErrorOr<BAN::RefPtr<NullDevice>> create(mode_t, uid_t, gid_t);
@@ -26,6 +26,7 @@ namespace Kernel
 		virtual bool can_read_impl() const override { return false; }
 		virtual bool can_write_impl() const override { return true; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		const dev_t m_rdev;

kernel/include/kernel/Device/RandomDevice.h

@@ -3,7 +3,7 @@
 namespace Kernel
 {
 
-	class RandomDevice : public CharacterDevice
+	class RandomDevice final : public CharacterDevice
 	{
 	public:
 		static BAN::ErrorOr<BAN::RefPtr<RandomDevice>> create(mode_t, uid_t, gid_t);
@@ -24,6 +24,7 @@ namespace Kernel
 		virtual bool can_read_impl() const override { return true; }
 		virtual bool can_write_impl() const override { return false; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		const dev_t m_rdev;

kernel/include/kernel/Device/ZeroDevice.h

@@ -3,7 +3,7 @@
 namespace Kernel
 {
 
-	class ZeroDevice : public CharacterDevice
+	class ZeroDevice final : public CharacterDevice
 	{
 	public:
 		static BAN::ErrorOr<BAN::RefPtr<ZeroDevice>> create(mode_t, uid_t, gid_t);
@@ -24,6 +24,7 @@ namespace Kernel
 		virtual bool can_read_impl() const override { return true; }
 		virtual bool can_write_impl() const override { return false; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		const dev_t m_rdev;

kernel/include/kernel/ELF.h

@@ -20,6 +20,6 @@ namespace Kernel::ELF
 		BAN::Vector<BAN::UniqPtr<MemoryRegion>> regions;
 	};
 
-	BAN::ErrorOr<LoadResult> load_from_inode(BAN::RefPtr<Inode>, const Credentials&, PageTable&);
+	BAN::ErrorOr<LoadResult> load_from_inode(BAN::RefPtr<Inode> root, BAN::RefPtr<Inode> inode, const Credentials&, PageTable&);
 
 }

kernel/include/kernel/Epoll.h

@@ -0,0 +1,109 @@
+#pragma once
+
+#include <BAN/Array.h>
+#include <BAN/CircularQueue.h>
+#include <BAN/HashMap.h>
+#include <BAN/HashSet.h>
+#include <kernel/FS/Inode.h>
+
+#include <limits.h>
+#include <sys/epoll.h>
+
+namespace Kernel
+{
+
+	class Epoll final : public Inode
+	{
+	public:
+		static BAN::ErrorOr<BAN::RefPtr<Epoll>> create();
+		~Epoll();
+
+		BAN::ErrorOr<void> ctl(int op, int fd, BAN::RefPtr<Inode> inode, epoll_event event);
+		BAN::ErrorOr<size_t> wait(BAN::Span<epoll_event> events, uint64_t waketime_ns);
+
+		void notify(BAN::RefPtr<Inode> inode, uint32_t event);
+
+	private:
+		Epoll() = default;
+
+	public:
+		ino_t ino() const override         { return 0; }
+		Mode mode() const override         { return { Mode::IRUSR | Mode::IWUSR }; }
+		nlink_t nlink() const override     { return 0; }
+		uid_t uid() const override         { return 0; }
+		gid_t gid() const override         { return 0; }
+		off_t size() const override        { return 0; }
+		timespec atime() const override    { return {}; }
+		timespec mtime() const override    { return {}; }
+		timespec ctime() const override    { return {}; }
+		blksize_t blksize() const override { return PAGE_SIZE; }
+		blkcnt_t blocks() const override   { return 0; }
+		dev_t dev() const override         { return 0; }
+		dev_t rdev() const override        { return 0; }
+
+		bool is_epoll() const override { return true; }
+
+		const FileSystem* filesystem() const override { return nullptr; }
+
+		bool can_read_impl() const override { return false; }
+		bool can_write_impl() const override { return false; }
+		bool has_error_impl() const override { return false; }
+		bool has_hungup_impl() const override { return false; }
+
+		BAN::ErrorOr<void> fsync_impl() override { return {}; }
+
+	private:
+		struct InodeRefPtrHash
+		{
+			BAN::hash_t operator()(const BAN::RefPtr<Inode>& inode)
+			{
+				return BAN::hash<const Inode*>()(inode.ptr());
+			}
+		};
+
+		struct ListenEventList
+		{
+			BAN::Array<epoll_event, OPEN_MAX> events;
+			uint32_t bitmap[(OPEN_MAX + 31) / 32] {};
+
+			bool has_fd(int fd) const
+			{
+				// For some reason having (fd < 0 || ...) makes GCC 15.1.0
+				// think bitmap access can be out of bounds...
+				if (static_cast<size_t>(fd) >= events.size())
+					return false;
+				return bitmap[fd / 32] & (1u << (fd % 32));
+			}
+
+			bool empty() const
+			{
+				for (auto val : bitmap)
+					if (val != 0)
+						return false;
+				return true;
+			}
+
+			void add_fd(int fd, epoll_event event)
+			{
+				ASSERT(!has_fd(fd));
+				bitmap[fd / 32] |= (1u << (fd % 32));
+				events[fd] = event;
+			}
+
+			void remove_fd(int fd)
+			{
+				ASSERT(has_fd(fd));
+				bitmap[fd / 32] &= ~(1u << (fd % 32));
+				events[fd] = {};
+			}
+		};
+
+	private:
+		ThreadBlocker m_thread_blocker;
+		SpinLock m_ready_lock;
+		BAN::HashMap<BAN::RefPtr<Inode>, uint32_t,        InodeRefPtrHash> m_ready_events;
+		BAN::HashMap<BAN::RefPtr<Inode>, uint32_t,        InodeRefPtrHash> m_processing_events;
+		BAN::HashMap<BAN::RefPtr<Inode>, ListenEventList, InodeRefPtrHash> m_listening_events;
+	};
+
+}

kernel/include/kernel/FS/DevFS/FileSystem.h

@@ -22,6 +22,7 @@ namespace Kernel
 
 		void add_inode(BAN::StringView path, BAN::RefPtr<TmpInode>);
 
+		void initiate_disk_cache_drop();
 		void initiate_sync(bool should_block);
 
 	private:
@@ -37,6 +38,10 @@ namespace Kernel
 		ThreadBlocker m_sync_done;
 		ThreadBlocker m_sync_thread_blocker;
 		volatile bool m_should_sync { false };
+
+		SpinLock m_disk_cache_lock;
+		ThreadBlocker m_disk_cache_thread_blocker;
+		BAN::Atomic<bool> m_should_drop_disk_cache { false };
 	};
 
 }

kernel/include/kernel/FS/EventFD.h

@@ -0,0 +1,52 @@
+#pragma once
+
+#include <kernel/FS/Inode.h>
+
+namespace Kernel
+{
+
+	class EventFD final : public Inode
+	{
+	public:
+		static BAN::ErrorOr<BAN::RefPtr<Inode>> create(uint64_t initval, bool semaphore);
+
+		ino_t ino() const override { return 0; }
+		Mode mode() const override { return { Mode::IFCHR | Mode::IRUSR | Mode::IWUSR }; }
+		nlink_t nlink() const override { return ref_count(); }
+		uid_t uid() const override { return 0; }
+		gid_t gid() const override { return 0; }
+		off_t size() const override { return 0; }
+		timespec atime() const override { return {}; }
+		timespec mtime() const override { return {}; }
+		timespec ctime() const override { return {}; }
+		blksize_t blksize() const override { return 8; }
+		blkcnt_t blocks() const override { return 0; }
+		dev_t dev() const override { return 0; }
+		dev_t rdev() const override { return 0; }
+
+		const FileSystem* filesystem() const override { return nullptr; }
+
+	protected:
+		BAN::ErrorOr<size_t> read_impl(off_t, BAN::ByteSpan) override;
+		BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override;
+		BAN::ErrorOr<void> fsync_impl() final override { return {}; }
+
+		bool can_read_impl() const override { return m_value > 0; }
+		bool can_write_impl() const override { return m_value < UINT64_MAX - 1; }
+		bool has_error_impl() const override { return false; }
+		bool has_hungup_impl() const override { return false; }
+
+	private:
+		EventFD(uint64_t initval, bool is_semaphore)
+			: m_is_semaphore(is_semaphore)
+			, m_value(initval)
+		{ }
+
+	private:
+		const bool m_is_semaphore;
+		uint64_t m_value;
+
+		ThreadBlocker m_thread_blocker;
+	};
+
+}

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

@@ -26,18 +26,32 @@ namespace Kernel
 		class BlockBufferWrapper
 		{
 			BAN_NON_COPYABLE(BlockBufferWrapper);
-			BAN_NON_MOVABLE(BlockBufferWrapper);
 
 		public:
-			BlockBufferWrapper(BAN::Span<uint8_t> buffer, bool& used)
+			BlockBufferWrapper(BAN::Span<uint8_t> buffer, void (*callback)(void*, const uint8_t*), void* argument)
 				: m_buffer(buffer)
-				, m_used(used)
-			{
-				ASSERT(m_used);
-			}
+				, m_callback(callback)
+				, m_argument(argument)
+			{ }
+			BlockBufferWrapper(BlockBufferWrapper&& other) { *this = BAN::move(other); }
 			~BlockBufferWrapper()
 			{
-				m_used = false;
+				if (m_callback == nullptr)
+					return;
+				m_callback(m_argument, m_buffer.data());
+			}
+
+			BlockBufferWrapper& operator=(BlockBufferWrapper&& other)
+			{
+				this->m_buffer = other.m_buffer;
+				this->m_callback = other.m_callback;
+				this->m_argument = other.m_argument;
+
+				other.m_buffer = {};
+				other.m_callback = nullptr;
+				other.m_argument = nullptr;
+
+				return *this;
 			}
 
 			size_t size() const { return m_buffer.size(); }
@@ -53,7 +67,8 @@ namespace Kernel
 
 		private:
 			BAN::Span<uint8_t> m_buffer;
-			bool& m_used;
+			void (*m_callback)(void*, const uint8_t*);
+			void* m_argument;
 		};
 
 	public:
@@ -62,8 +77,6 @@ namespace Kernel
 
 		virtual BAN::RefPtr<Inode> root_inode() override { return m_root_inode; }
 
-		virtual dev_t dev() const override { return m_block_device->rdev(); };
-
 	private:
 		Ext2FS(BAN::RefPtr<BlockDevice> block_device)
 			: m_block_device(block_device)
@@ -81,7 +94,7 @@ namespace Kernel
 		BAN::ErrorOr<void> sync_superblock();
 		BAN::ErrorOr<void> sync_block(uint32_t block);
 
-		BlockBufferWrapper get_block_buffer();
+		BAN::ErrorOr<BlockBufferWrapper> get_block_buffer();
 
 		BAN::ErrorOr<uint32_t> reserve_free_block(uint32_t primary_bgd);
 		BAN::ErrorOr<void> release_block(uint32_t block);
@@ -104,10 +117,13 @@ namespace Kernel
 		{
 		public:
 			BlockBufferManager() = default;
-			BlockBufferWrapper get_buffer();
+			BAN::ErrorOr<BlockBufferWrapper> get_buffer();
 
 			BAN::ErrorOr<void> initialize(size_t block_size);
 
+		private:
+			void destroy_callback(const uint8_t* buffer_ptr);
+
 		private:
 			struct BlockBuffer
 			{
@@ -115,8 +131,20 @@ namespace Kernel
 				bool used { false };
 			};
 
+			struct ThreadInfo
+			{
+				pid_t tid { 0 };
+				size_t buffers { 0 };
+			};
+
 		private:
-			BAN::Array<BlockBuffer, 10> m_buffers;
+			static constexpr size_t max_threads = 8;
+			static constexpr size_t max_buffers_per_thread = 6;
+
+			Mutex m_buffer_mutex;
+			ThreadBlocker m_buffer_blocker;
+			BAN::Array<BlockBuffer, max_threads * max_buffers_per_thread> m_buffers;
+			BAN::Array<ThreadInfo, max_threads> m_thread_infos;
 		};
 
 	private:

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

@@ -37,6 +37,7 @@ namespace Kernel
 		virtual BAN::ErrorOr<void> create_file_impl(BAN::StringView, mode_t, uid_t, gid_t) override;
 		virtual BAN::ErrorOr<void> create_directory_impl(BAN::StringView, mode_t, uid_t, gid_t) override;
 		virtual BAN::ErrorOr<void> link_inode_impl(BAN::StringView, BAN::RefPtr<Inode>) override;
+		virtual BAN::ErrorOr<void> rename_inode_impl(BAN::RefPtr<Inode>, BAN::StringView, BAN::StringView) override;
 		virtual BAN::ErrorOr<void> unlink_impl(BAN::StringView) override;
 
 		virtual BAN::ErrorOr<BAN::String> link_target_impl() override;
@@ -46,29 +47,32 @@ namespace Kernel
 		virtual BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override;
 		virtual BAN::ErrorOr<void> truncate_impl(size_t) override;
 		virtual BAN::ErrorOr<void> chmod_impl(mode_t) override;
+		virtual BAN::ErrorOr<void> chown_impl(uid_t, gid_t) override;
+		virtual BAN::ErrorOr<void> utimens_impl(const timespec[2]) override;
 		virtual BAN::ErrorOr<void> fsync_impl() override;
 
 		virtual bool can_read_impl() const override { return true; }
 		virtual bool can_write_impl() const override { return true; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		// Returns maximum number of data blocks in use
 		// NOTE: the inode might have more blocks than what this suggests if it has been shrinked
 		uint32_t max_used_data_block_count() const { return size() / blksize(); }
 
-		BAN::ErrorOr<BAN::Optional<uint32_t>> block_from_indirect_block(uint32_t block, uint32_t index, uint32_t depth);
-		BAN::ErrorOr<BAN::Optional<uint32_t>> fs_block_of_data_block_index(uint32_t data_block_index);
+		BAN::ErrorOr<BAN::Optional<uint32_t>> block_from_indirect_block(uint32_t& block, uint32_t index, uint32_t depth, bool allocate);
+		BAN::ErrorOr<BAN::Optional<uint32_t>> fs_block_of_data_block_index(uint32_t data_block_index, bool allocate);
 
 		BAN::ErrorOr<void> link_inode_to_directory(Ext2Inode&, BAN::StringView name);
+		BAN::ErrorOr<void> remove_inode_from_directory(BAN::StringView name, bool cleanup_directory);
 		BAN::ErrorOr<bool> is_directory_empty();
 
 		BAN::ErrorOr<void> cleanup_indirect_block(uint32_t block, uint32_t depth);
 		BAN::ErrorOr<void> cleanup_default_links();
+		BAN::ErrorOr<void> cleanup_data_blocks();
 		BAN::ErrorOr<void> cleanup_from_fs();
 
-		BAN::ErrorOr<uint32_t> allocate_new_block_to_indirect_block(uint32_t& block, uint32_t index, uint32_t depth);
-		BAN::ErrorOr<uint32_t> allocate_new_block(uint32_t data_block_index);
 		BAN::ErrorOr<void> sync();
 
 		uint32_t block_group() const;
@@ -81,6 +85,26 @@ namespace Kernel
 		{}
 		static BAN::ErrorOr<BAN::RefPtr<Ext2Inode>> create(Ext2FS&, uint32_t);
 
+	private:
+		struct ScopedSync
+		{
+			ScopedSync(Ext2Inode& inode)
+				: inode(inode)
+				, inode_info(inode.m_inode)
+			{ }
+
+			~ScopedSync()
+			{
+				if (memcmp(&inode.m_inode, &inode_info, sizeof(Ext2::Inode)) == 0)
+					return;
+				if (auto ret = inode.sync(); ret.is_error())
+					dwarnln("failed to sync inode: {}", ret.error());
+			}
+
+			Ext2Inode& inode;
+			Ext2::Inode inode_info;
+		};
+
 	private:
 		Ext2FS& m_fs;
 		Ext2::Inode m_inode;

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

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

kernel/include/kernel/FS/FAT/Inode.h

@@ -44,11 +44,13 @@ namespace Kernel
 		//virtual BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override;
 		//virtual BAN::ErrorOr<void> truncate_impl(size_t) override;
 		//virtual BAN::ErrorOr<void> chmod_impl(mode_t) override;
+		//virtual BAN::ErrorOr<void> utimens_impl(const timespec[2]) override;
 		virtual BAN::ErrorOr<void> fsync_impl() override { return {}; }
 
 		virtual bool can_read_impl() const override { return true; }
 		virtual bool can_write_impl() const override { return true; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		FATInode(FATFS& fs, const FAT::DirectoryEntry& entry, ino_t ino, uint32_t block_count)

kernel/include/kernel/FS/FileSystem.h

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

kernel/include/kernel/FS/Inode.h

@@ -1,6 +1,7 @@
 #pragma once
 
 #include <BAN/ByteSpan.h>
+#include <BAN/LinkedList.h>
 #include <BAN/RefPtr.h>
 #include <BAN/String.h>
 #include <BAN/StringView.h>
@@ -19,10 +20,9 @@
 namespace Kernel
 {
 
-	class FileSystem;
-
 	class FileBackedRegion;
-	class SharedFileData;
+	class FileSystem;
+	struct SharedFileData;
 
 	class Inode : public BAN::RefCounted<Inode>
 	{
@@ -85,6 +85,7 @@ namespace Kernel
 		virtual dev_t rdev() const = 0;
 
 		virtual bool is_device() const { return false; }
+		virtual bool is_epoll() const { return false; }
 		virtual bool is_pipe() const { return false; }
 		virtual bool is_tty() const { return false; }
 
@@ -96,6 +97,7 @@ namespace Kernel
 		BAN::ErrorOr<void> create_file(BAN::StringView, mode_t, uid_t, gid_t);
 		BAN::ErrorOr<void> create_directory(BAN::StringView, mode_t, uid_t, gid_t);
 		BAN::ErrorOr<void> link_inode(BAN::StringView, BAN::RefPtr<Inode>);
+		BAN::ErrorOr<void> rename_inode(BAN::RefPtr<Inode>, BAN::StringView, BAN::StringView);
 		BAN::ErrorOr<void> unlink(BAN::StringView);
 
 		// Link API
@@ -107,9 +109,12 @@ namespace Kernel
 		BAN::ErrorOr<void> bind(const sockaddr* address, socklen_t address_len);
 		BAN::ErrorOr<void> connect(const sockaddr* address, socklen_t address_len);
 		BAN::ErrorOr<void> listen(int backlog);
-		BAN::ErrorOr<size_t> sendto(BAN::ConstByteSpan message, const sockaddr* address, socklen_t address_len);
-		BAN::ErrorOr<size_t> recvfrom(BAN::ByteSpan buffer, sockaddr* address, socklen_t* address_len);
+		BAN::ErrorOr<size_t> sendmsg(const msghdr& message, int flags);
+		BAN::ErrorOr<size_t> recvmsg(msghdr& message, int flags);
 		BAN::ErrorOr<void> getsockname(sockaddr* address, socklen_t* address_len);
+		BAN::ErrorOr<void> getpeername(sockaddr* address, socklen_t* address_len);
+		BAN::ErrorOr<void> getsockopt(int level, int option, void* value, socklen_t* value_len);
+		BAN::ErrorOr<void> setsockopt(int level, int option, const void* value, socklen_t value_len);
 
 		// General API
 		BAN::ErrorOr<size_t> read(off_t, BAN::ByteSpan buffer);
@@ -117,15 +122,24 @@ namespace Kernel
 		BAN::ErrorOr<void> truncate(size_t);
 		BAN::ErrorOr<void> chmod(mode_t);
 		BAN::ErrorOr<void> chown(uid_t, gid_t);
+		BAN::ErrorOr<void> utimens(const timespec[2]);
 		BAN::ErrorOr<void> fsync();
 
 		// Select/Non blocking API
 		bool can_read() const;
 		bool can_write() const;
 		bool has_error() const;
+		bool has_hungup() const;
 
 		BAN::ErrorOr<long> ioctl(int request, void* arg);
 
+		BAN::ErrorOr<void> add_epoll(class Epoll*);
+		void del_epoll(class Epoll*);
+		void epoll_notify(uint32_t event);
+
+		virtual void on_close(int status_flags) { (void)status_flags; }
+		virtual void on_clone(int status_flags) { (void)status_flags; }
+
 	protected:
 		// Directory API
 		virtual BAN::ErrorOr<BAN::RefPtr<Inode>> find_inode_impl(BAN::StringView)							{ return BAN::Error::from_errno(ENOTSUP); }
@@ -133,6 +147,7 @@ namespace Kernel
 		virtual BAN::ErrorOr<void> create_file_impl(BAN::StringView, mode_t, uid_t, gid_t)					{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> create_directory_impl(BAN::StringView, mode_t, uid_t, gid_t)				{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> link_inode_impl(BAN::StringView, BAN::RefPtr<Inode>)						{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<void> rename_inode_impl(BAN::RefPtr<Inode>, BAN::StringView, BAN::StringView)	{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> unlink_impl(BAN::StringView)												{ return BAN::Error::from_errno(ENOTSUP); }
 
 		// Link API
@@ -144,9 +159,12 @@ namespace Kernel
 		virtual BAN::ErrorOr<void> connect_impl(const sockaddr*, socklen_t)							{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> listen_impl(int)													{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> bind_impl(const sockaddr*, socklen_t)							{ return BAN::Error::from_errno(ENOTSUP); }
-		virtual BAN::ErrorOr<size_t> sendto_impl(BAN::ConstByteSpan, const sockaddr*, socklen_t)	{ return BAN::Error::from_errno(ENOTSUP); }
-		virtual BAN::ErrorOr<size_t> recvfrom_impl(BAN::ByteSpan, sockaddr*, socklen_t*)			{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<size_t> recvmsg_impl(msghdr&, int)										{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<size_t> sendmsg_impl(const msghdr&, int)								{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> getsockname_impl(sockaddr*, socklen_t*)							{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<void> getpeername_impl(sockaddr*, socklen_t*)							{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<void> getsockopt_impl(int, int, void*, socklen_t*)						{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<void> setsockopt_impl(int, int, const void*, socklen_t)				{ return BAN::Error::from_errno(ENOTSUP); }
 
 		// General API
 		virtual BAN::ErrorOr<size_t> read_impl(off_t, BAN::ByteSpan)		{ return BAN::Error::from_errno(ENOTSUP); }
@@ -154,12 +172,14 @@ namespace Kernel
 		virtual BAN::ErrorOr<void> truncate_impl(size_t)					{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> chmod_impl(mode_t)						{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> chown_impl(uid_t, gid_t)					{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<void> utimens_impl(const timespec[2])			{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> fsync_impl() = 0;
 
 		// Select/Non blocking API
 		virtual bool can_read_impl() const = 0;
 		virtual bool can_write_impl() const = 0;
 		virtual bool has_error_impl() const = 0;
+		virtual bool has_hungup_impl() const = 0;
 
 		virtual BAN::ErrorOr<long> ioctl_impl(int, void*) { return BAN::Error::from_errno(ENOTSUP); }
 
@@ -168,9 +188,12 @@ namespace Kernel
 
 	private:
 		BAN::WeakPtr<SharedFileData> m_shared_region;
+		SpinLock m_epoll_lock;
+		BAN::LinkedList<class Epoll*> m_epolls;
+		friend class Epoll;
 		friend class FileBackedRegion;
 		friend class OpenFileDescriptorSet;
-		friend class SharedFileData;
+		friend struct SharedFileData;
 		friend class TTY;
 	};
 

kernel/include/kernel/FS/Pipe.h

@@ -2,6 +2,7 @@
 
 #include <BAN/Array.h>
 #include <kernel/FS/Inode.h>
+#include <kernel/Memory/ByteRingBuffer.h>
 #include <kernel/ThreadBlocker.h>
 
 namespace Kernel
@@ -13,8 +14,9 @@ namespace Kernel
 		static BAN::ErrorOr<BAN::RefPtr<Inode>> create(const Credentials&);
 
 		virtual bool is_pipe() const override { return true; }
-		void clone_writing();
-		void close_writing();
+
+		void on_close(int status_flags) override;
+		void on_clone(int status_flags) override;
 
 		virtual ino_t ino() const override { return 0; } // FIXME
 		virtual Mode mode() const override { return { Mode::IFIFO | Mode::IRUSR | Mode::IWUSR }; }
@@ -37,9 +39,10 @@ namespace Kernel
 		virtual BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override;
 		virtual BAN::ErrorOr<void> fsync_impl() final override { return {}; }
 
-		virtual bool can_read_impl() const override { return m_buffer_size > 0; }
+		virtual bool can_read_impl() const override { return !m_buffer->empty(); }
 		virtual bool can_write_impl() const override { return true; }
-		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_error_impl() const override { return m_reading_count == 0; }
+		virtual bool has_hungup_impl() const override { return m_writing_count == 0; }
 
 	private:
 		Pipe(const Credentials&);
@@ -52,11 +55,10 @@ namespace Kernel
 		timespec m_ctime {};
 		ThreadBlocker m_thread_blocker;
 
-		BAN::Array<uint8_t, PAGE_SIZE> m_buffer;
-		BAN::Atomic<size_t> m_buffer_size { 0 };
-		size_t m_buffer_tail { 0 };
+		BAN::UniqPtr<ByteRingBuffer> m_buffer;
 
 		BAN::Atomic<uint32_t> m_writing_count { 1 };
+		BAN::Atomic<uint32_t> m_reading_count { 1 };
 	};
 
 }

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

@@ -47,6 +47,7 @@ namespace Kernel
 		virtual bool can_read_impl() const override { return true; }
 		virtual bool can_write_impl() const override { return false; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		ProcROProcessInode(Process&, size_t (Process::*)(off_t, BAN::ByteSpan) const, TmpFileSystem&, const TmpInodeInfo&);
@@ -56,6 +57,31 @@ namespace Kernel
 		size_t (Process::*m_callback)(off_t, BAN::ByteSpan) const;
 	};
 
+	class ProcSymlinkProcessInode final : public TmpInode
+	{
+	public:
+		static BAN::ErrorOr<BAN::RefPtr<ProcSymlinkProcessInode>> create_new(Process& process, BAN::ErrorOr<BAN::String> (Process::*callback)() const, TmpFileSystem&, mode_t);
+		~ProcSymlinkProcessInode() = default;
+
+		virtual uid_t uid() const override { return m_process.credentials().ruid(); }
+		virtual gid_t gid() const override { return m_process.credentials().rgid(); }
+
+	protected:
+		virtual BAN::ErrorOr<BAN::String> link_target_impl() override;
+
+		virtual bool can_read_impl() const override { return false; }
+		virtual bool can_write_impl() const override { return false; }
+		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
+
+	private:
+		ProcSymlinkProcessInode(Process& process, BAN::ErrorOr<BAN::String> (Process::*)() const, TmpFileSystem&, const TmpInodeInfo&);
+
+	private:
+		Process& m_process;
+		BAN::ErrorOr<BAN::String> (Process::*m_callback)() const;
+	};
+
 	class ProcROInode final : public TmpInode
 	{
 	public:
@@ -72,6 +98,7 @@ namespace Kernel
 		virtual bool can_read_impl() const override { return true; }
 		virtual bool can_write_impl() const override { return false; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		ProcROInode(size_t (*callback)(off_t, BAN::ByteSpan), TmpFileSystem&, const TmpInodeInfo&);
@@ -80,4 +107,57 @@ namespace Kernel
 		size_t (*m_callback)(off_t, BAN::ByteSpan);
 	};
 
+	class ProcSymlinkInode final : public TmpInode
+	{
+	public:
+		static BAN::ErrorOr<BAN::RefPtr<ProcSymlinkInode>> create_new(BAN::ErrorOr<BAN::String> (*)(void*), void (*)(void*), void* data, TmpFileSystem&, mode_t, uid_t, gid_t);
+		~ProcSymlinkInode();
+
+	protected:
+		virtual BAN::ErrorOr<BAN::String> link_target_impl() override;
+
+		virtual bool can_read_impl() const override { return false; }
+		virtual bool can_write_impl() const override { return false; }
+		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
+
+	private:
+		ProcSymlinkInode(BAN::ErrorOr<BAN::String> (*callback)(void*), void (*destructor)(void*), void* data, TmpFileSystem&, const TmpInodeInfo&);
+
+	private:
+		BAN::ErrorOr<BAN::String> (*m_callback)(void*);
+		void (*m_destructor)(void*);
+		void* m_data;
+	};
+
+	class ProcFDDirectoryInode final : public TmpInode
+	{
+	public:
+		static BAN::ErrorOr<BAN::RefPtr<ProcFDDirectoryInode>> create_new(Process&, TmpFileSystem&, mode_t);
+		~ProcFDDirectoryInode() = default;
+
+		virtual uid_t uid() const override { return m_process.credentials().ruid(); }
+		virtual gid_t gid() const override { return m_process.credentials().rgid(); }
+
+	protected:
+		virtual BAN::ErrorOr<BAN::RefPtr<Inode>> find_inode_impl(BAN::StringView) override;
+		virtual BAN::ErrorOr<size_t> list_next_inodes_impl(off_t, struct dirent*, size_t) override;
+		virtual BAN::ErrorOr<void> create_file_impl(BAN::StringView, mode_t, uid_t, gid_t) override                 { return BAN::Error::from_errno(EPERM); }
+		virtual BAN::ErrorOr<void> create_directory_impl(BAN::StringView, mode_t, uid_t, gid_t) override            { return BAN::Error::from_errno(EPERM); }
+		virtual BAN::ErrorOr<void> link_inode_impl(BAN::StringView, BAN::RefPtr<Inode>) override                    { return BAN::Error::from_errno(EPERM); }
+		virtual BAN::ErrorOr<void> rename_inode_impl(BAN::RefPtr<Inode>, BAN::StringView, BAN::StringView) override { return BAN::Error::from_errno(EPERM); }
+		virtual BAN::ErrorOr<void> unlink_impl(BAN::StringView) override                                            { return BAN::Error::from_errno(EPERM); }
+
+		virtual bool can_read_impl() const override { return false; }
+		virtual bool can_write_impl() const override { return false; }
+		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
+
+	private:
+		ProcFDDirectoryInode(Process&, TmpFileSystem&, const TmpInodeInfo&);
+
+	private:
+		Process& m_process;
+	};
+
 }

kernel/include/kernel/FS/Socket.h

@@ -51,8 +51,6 @@ namespace Kernel
 			: m_info(info)
 		{}
 
-		BAN::ErrorOr<size_t> read_impl(off_t, BAN::ByteSpan buffer) override { return recvfrom_impl(buffer, nullptr, nullptr); }
-		BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan buffer) override { return sendto_impl(buffer, nullptr, 0); }
 		BAN::ErrorOr<void> fsync_impl() final override { return {}; }
 
 	private:

kernel/include/kernel/FS/TmpFS/Definitions.h

@@ -28,7 +28,7 @@ namespace Kernel
 		// 1x singly indirect
 		// 1x doubly indirect
 		// 1x triply indirect
-		BAN::Array<paddr_t, 5> block;
+		BAN::Array<size_t, 5> block;
 		static constexpr size_t direct_block_count = 2;
 #elif ARCH(i686)
 		uint32_t __padding;
@@ -36,8 +36,8 @@ namespace Kernel
 		// 1x singly indirect
 		// 1x doubly indirect
 		// 1x triply indirect
-		BAN::Array<paddr_t, 8> block;
-		static constexpr size_t direct_block_count = 5;
+		BAN::Array<size_t, 16> block;
+		static constexpr size_t direct_block_count = 13;
 #else
 	#error
 #endif

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

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

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

@@ -47,14 +47,22 @@ namespace Kernel
 	protected:
 		TmpInode(TmpFileSystem&, ino_t, const TmpInodeInfo&);
 
+		virtual BAN::ErrorOr<void> chmod_impl(mode_t) override;
+		virtual BAN::ErrorOr<void> chown_impl(uid_t, gid_t) override;
+		virtual BAN::ErrorOr<void> utimens_impl(const timespec[2]) override;
 		virtual BAN::ErrorOr<void> fsync_impl() override { return {}; }
 
 		void sync();
-		void free_all_blocks();
 		virtual BAN::ErrorOr<void> prepare_unlink() { return {}; };
 
+		void free_all_blocks();
+		void free_indirect_blocks(size_t block, uint32_t depth);
+
 		BAN::Optional<size_t> block_index(size_t data_block_index);
+		BAN::Optional<size_t> block_index_from_indirect(size_t block, size_t index, uint32_t depth);
+
 		BAN::ErrorOr<size_t> block_index_with_allocation(size_t data_block_index);
+		BAN::ErrorOr<size_t> block_index_from_indirect_with_allocation(size_t& block, size_t index, uint32_t depth);
 
 	protected:
 		TmpFileSystem& m_fs;
@@ -75,11 +83,11 @@ 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;
 		virtual BAN::ErrorOr<void> truncate_impl(size_t) override;
-		virtual BAN::ErrorOr<void> chmod_impl(mode_t) override;
 
 		virtual bool can_read_impl() const override { return true; }
 		virtual bool can_write_impl() const override { return true; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		TmpFileInode(TmpFileSystem&, ino_t, const TmpInodeInfo&);
@@ -97,11 +105,11 @@ namespace Kernel
 		virtual BAN::ErrorOr<size_t> read_impl(off_t, BAN::ByteSpan) override { return BAN::Error::from_errno(ENODEV); }
 		virtual BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override { return BAN::Error::from_errno(ENODEV); }
 		virtual BAN::ErrorOr<void> truncate_impl(size_t) override { return BAN::Error::from_errno(ENODEV); }
-		virtual BAN::ErrorOr<void> chmod_impl(mode_t) override;
 
 		virtual bool can_read_impl() const override { return false; }
 		virtual bool can_write_impl() const override { return false; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		TmpSocketInode(TmpFileSystem&, ino_t, const TmpInodeInfo&);
@@ -115,14 +123,14 @@ namespace Kernel
 		static BAN::ErrorOr<BAN::RefPtr<TmpSymlinkInode>> create_new(TmpFileSystem&, mode_t, uid_t, gid_t, BAN::StringView target);
 		~TmpSymlinkInode();
 
-		BAN::ErrorOr<void> set_link_target(BAN::StringView);
-
 	protected:
-		virtual BAN::ErrorOr<BAN::String> link_target_impl() override;
+		BAN::ErrorOr<BAN::String> link_target_impl() override;
+		BAN::ErrorOr<void> set_link_target_impl(BAN::StringView) override;
 
 		virtual bool can_read_impl() const override { return false; }
 		virtual bool can_write_impl() const override { return false; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		TmpSymlinkInode(TmpFileSystem&, ino_t, const TmpInodeInfo&);
@@ -148,16 +156,21 @@ namespace Kernel
 		virtual BAN::ErrorOr<size_t> list_next_inodes_impl(off_t, struct dirent*, size_t) override final;
 		virtual BAN::ErrorOr<void> create_file_impl(BAN::StringView, mode_t, uid_t, gid_t) override final;
 		virtual BAN::ErrorOr<void> create_directory_impl(BAN::StringView, mode_t, uid_t, gid_t) override final;
+		virtual BAN::ErrorOr<void> link_inode_impl(BAN::StringView, BAN::RefPtr<Inode>) override final;
+		virtual BAN::ErrorOr<void> rename_inode_impl(BAN::RefPtr<Inode>, BAN::StringView, BAN::StringView) override final;
 		virtual BAN::ErrorOr<void> unlink_impl(BAN::StringView) override;
 
 		virtual bool can_read_impl() const override { return false; }
 		virtual bool can_write_impl() const override { return false; }
 		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
 
 	private:
 		template<TmpFuncs::for_each_valid_entry_callback F>
 		void for_each_valid_entry(F callback);
 
+		BAN::ErrorOr<void> unlink_inode(BAN::StringView, bool cleanup);
+
 		friend class TmpInode;
 	};
 

kernel/include/kernel/FS/USTARModule.h

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

kernel/include/kernel/FS/VirtualFileSystem.h

@@ -29,9 +29,6 @@ namespace Kernel
 
 		virtual BAN::RefPtr<Inode> root_inode() override { return m_root_fs->root_inode(); }
 
-		// FIXME:
-		virtual dev_t dev() const override { return 0; }
-
 		BAN::ErrorOr<void> mount(const Credentials&, BAN::StringView, BAN::StringView);
 		BAN::ErrorOr<void> mount(const Credentials&, BAN::RefPtr<FileSystem>, BAN::StringView);
 
@@ -74,13 +71,13 @@ namespace Kernel
 
 		File root_file()
 		{
-			return File(root_inode(), "/"_sv);
+			return File { root_inode(), "/"_sv };
 		}
 
-		BAN::ErrorOr<File> file_from_relative_path(const File& parent, const Credentials&, BAN::StringView, int);
-		BAN::ErrorOr<File> file_from_absolute_path(const Credentials& credentials, BAN::StringView path, int flags)
+		BAN::ErrorOr<File> file_from_relative_path(BAN::RefPtr<Inode> root_inode, const File& parent, const Credentials&, BAN::StringView, int);
+		BAN::ErrorOr<File> file_from_absolute_path(BAN::RefPtr<Inode> root_inode, const Credentials& credentials, BAN::StringView path, int flags)
 		{
-			return file_from_relative_path(root_file(), credentials, path, flags);
+			return file_from_relative_path(root_inode, File { root_inode, "/"_sv }, credentials, path, flags);
 		}
 
 	private:

kernel/include/kernel/GDT.h

@@ -129,7 +129,8 @@ namespace Kernel
 		}
 
 #if ARCH(i686)
-		void set_tls(uintptr_t addr);
+		void set_fsbase(uintptr_t addr);
+		void set_gsbase(uintptr_t addr);
 #endif
 
 	private:
@@ -150,11 +151,11 @@ namespace Kernel
 
 	private:
 #if ARCH(x86_64)
-		BAN::Array<SegmentDescriptor, 7> m_gdt; // null, kernel code, kernel data, user code, user data, tss low, tss high
-		static constexpr uint16_t m_tss_offset = 0x28;
+		BAN::Array<SegmentDescriptor, 8> m_gdt; // null, kernel code, kernel data, user code (32 bit), user data, user code (64 bit), tss low, tss high
+		static constexpr uint16_t m_tss_offset = 0x30;
 #elif ARCH(i686)
-		BAN::Array<SegmentDescriptor, 8> m_gdt; // null, kernel code, kernel data, user code, user data, processor data, tls, tss
-		static constexpr uint16_t m_tss_offset = 0x38;
+		BAN::Array<SegmentDescriptor, 9> m_gdt; // null, kernel code, kernel data, user code, user data, processor data, fsbase, gsbase, tss
+		static constexpr uint16_t m_tss_offset = 0x40;
 #endif
 		TaskStateSegment m_tss;
 		const GDTR m_gdtr {

kernel/include/kernel/IDT.h

@@ -18,10 +18,12 @@ namespace Kernel
 
 	constexpr uint8_t IRQ_VECTOR_BASE   = 0x20;
 	constexpr uint8_t IRQ_MSI_BASE      = 0x80;
+	constexpr uint8_t IRQ_MSI_END       = 0xF0;
+#if ARCH(i686)
 	constexpr uint8_t IRQ_SYSCALL       = 0xF0;
-	constexpr uint8_t IRQ_YIELD         = 0xF1;
-	constexpr uint8_t IRQ_IPI           = 0xF2;
-	constexpr uint8_t IRQ_TIMER         = 0xF3;
+#endif
+	constexpr uint8_t IRQ_IPI           = 0xF1;
+	constexpr uint8_t IRQ_TIMER         = 0xF2;
 
 #if ARCH(x86_64)
 	struct GateDescriptor
@@ -75,7 +77,7 @@ namespace Kernel
 	private:
 		IDT() = default;
 
-		void register_interrupt_handler(uint8_t index, void (*handler)());
+		void register_interrupt_handler(uint8_t index, void (*handler)(), uint8_t ist = 0);
 		void register_syscall_handler(uint8_t index, void (*handler)());
 
 	private:

kernel/include/kernel/Input/InputDevice.h

@@ -15,6 +15,7 @@ namespace Kernel
 		{
 			Mouse,
 			Keyboard,
+			Joystick,
 		};
 
 	public:
@@ -31,7 +32,7 @@ namespace Kernel
 		bool can_read_impl() const override { SpinLockGuard _(m_event_lock); return m_event_count > 0; }
 		bool can_write_impl() const override { return false; }
 		bool has_error_impl() const override { return false; }
-
+		bool has_hungup_impl() const override { return false; }
 
 	private:
 		BAN::ErrorOr<size_t> read_non_block(BAN::ByteSpan);
@@ -64,7 +65,7 @@ namespace Kernel
 	public:
 		static BAN::ErrorOr<BAN::RefPtr<KeyboardDevice>> create(mode_t mode, uid_t uid, gid_t gid);
 
-		void notify() { m_thread_blocker.unblock(); }
+		void notify();
 
 	private:
 		KeyboardDevice(mode_t mode, uid_t uid, gid_t gid);
@@ -73,6 +74,7 @@ namespace Kernel
 		bool can_read_impl() const override;
 		bool can_write_impl() const override { return false; }
 		bool has_error_impl() const override { return false; }
+		bool has_hungup_impl() const override { return false; }
 
 		BAN::StringView name() const final override { return m_name; }
 		dev_t rdev() const final override { return m_rdev; }
@@ -90,7 +92,7 @@ namespace Kernel
 	public:
 		static BAN::ErrorOr<BAN::RefPtr<MouseDevice>> create(mode_t mode, uid_t uid, gid_t gid);
 
-		void notify() { m_thread_blocker.unblock(); }
+		void notify();
 
 	private:
 		MouseDevice(mode_t mode, uid_t uid, gid_t gid);
@@ -99,6 +101,7 @@ namespace Kernel
 		bool can_read_impl() const override;
 		bool can_write_impl() const override { return false; }
 		bool has_error_impl() const override { return false; }
+		bool has_hungup_impl() const override { return false; }
 
 		BAN::StringView name() const final override { return m_name; }
 		dev_t rdev() const final override { return m_rdev; }

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

@@ -14,6 +14,15 @@ namespace Kernel::Input
 
 	class PS2Controller
 	{
+	public:
+		enum class DeviceType
+		{
+			None,
+			Unknown,
+			Keyboard,
+			Mouse,
+		};
+
 	public:
 		static BAN::ErrorOr<void> initialize(uint8_t scancode_set);
 		static PS2Controller& get();
@@ -24,10 +33,14 @@ namespace Kernel::Input
 		// Returns true, if byte is used as command, if returns false, byte is meant to device
 		bool handle_command_byte(PS2Device*, uint8_t);
 
+		uint8_t data_port() const { return m_data_port; }
+
 	private:
 		PS2Controller() = default;
 		BAN::ErrorOr<void> initialize_impl(uint8_t scancode_set);
-		BAN::ErrorOr<void> initialize_device(uint8_t, uint8_t scancode_set);
+		BAN::ErrorOr<DeviceType> identify_device(uint8_t);
+
+		void device_initialize_task(void*);
 
 		BAN::ErrorOr<uint8_t> read_byte();
 		BAN::ErrorOr<void> send_byte(uint16_t port, uint8_t byte);
@@ -59,6 +72,9 @@ namespace Kernel::Input
 		};
 
 	private:
+		uint16_t m_command_port { PS2::IOPort::COMMAND };
+		uint16_t m_data_port    { PS2::IOPort::DATA };
+
 		BAN::RefPtr<PS2Device> m_devices[2];
 
 		Mutex m_mutex;

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

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

kernel/include/kernel/InterruptStack.h

@@ -27,7 +27,6 @@ namespace Kernel
 		uintptr_t r10;
 		uintptr_t r9;
 		uintptr_t r8;
-
 		uintptr_t rdi;
 		uintptr_t rsi;
 		uintptr_t rbp;
@@ -36,6 +35,18 @@ namespace Kernel
 		uintptr_t rcx;
 		uintptr_t rax;
 	};
+	struct YieldRegisters
+	{
+		uintptr_t r15;
+		uintptr_t r14;
+		uintptr_t r13;
+		uintptr_t r12;
+		uintptr_t rbp;
+		uintptr_t rbx;
+		uintptr_t ret;
+		uintptr_t sp;
+		uintptr_t ip;
+	};
 #elif ARCH(i686)
 	struct InterruptRegisters
 	{
@@ -48,6 +59,16 @@ namespace Kernel
 		uintptr_t ecx;
 		uintptr_t eax;
 	};
+	struct YieldRegisters
+	{
+		uintptr_t ebp;
+		uintptr_t edi;
+		uintptr_t esi;
+		uintptr_t ebx;
+		uintptr_t ret;
+		uintptr_t sp;
+		uintptr_t ip;
+	};
 #endif
 
 }

kernel/include/kernel/Lock/LockGuard.h

@@ -29,30 +29,4 @@ namespace Kernel
 		Lock& m_lock;
 	};
 
-	template<typename Lock>
-	class LockFreeGuard
-	{
-		BAN_NON_COPYABLE(LockFreeGuard);
-		BAN_NON_MOVABLE(LockFreeGuard);
-
-	public:
-		LockFreeGuard(Lock& lock)
-			: m_lock(lock)
-			, m_depth(lock.lock_depth())
-		{
-			for (uint32_t i = 0; i < m_depth; i++)
-				m_lock.unlock();
-		}
-
-		~LockFreeGuard()
-		{
-			for (uint32_t i = 0; i < m_depth; i++)
-				m_lock.lock();
-		}
-
-	private:
-		Lock& m_lock;
-		const uint32_t m_depth;
-	};
-
 }

kernel/include/kernel/Lock/Mutex.h

@@ -9,7 +9,19 @@
 namespace Kernel
 {
 
-	class Mutex
+	class BaseMutex
+	{
+	public:
+		virtual void lock() = 0;
+		virtual bool try_lock() = 0;
+		virtual void unlock() = 0;
+
+		virtual pid_t locker() const = 0;
+		virtual bool is_locked() const = 0;
+		virtual uint32_t lock_depth() const = 0;
+	};
+
+	class Mutex : public BaseMutex
 	{
 		BAN_NON_COPYABLE(Mutex);
 		BAN_NON_MOVABLE(Mutex);
@@ -17,13 +29,14 @@ namespace Kernel
 	public:
 		Mutex() = default;
 
-		void lock()
+		void lock() override
 		{
 			const auto tid = Thread::current_tid();
 			if (tid == m_locker)
 				ASSERT(m_lock_depth > 0);
 			else
 			{
+				ASSERT(!tid || !Thread::current().has_spinlock());
 				pid_t expected = -1;
 				while (!m_locker.compare_exchange(expected, tid))
 				{
@@ -37,7 +50,7 @@ namespace Kernel
 			m_lock_depth++;
 		}
 
-		bool try_lock()
+		bool try_lock() override
 		{
 			const auto tid = Thread::current_tid();
 			if (tid == m_locker)
@@ -55,7 +68,7 @@ namespace Kernel
 			return true;
 		}
 
-		void unlock()
+		void unlock() override
 		{
 			const auto tid = Thread::current_tid();
 			ASSERT(m_locker == tid);
@@ -68,16 +81,16 @@ namespace Kernel
 			}
 		}
 
-		pid_t locker() const { return m_locker; }
-		bool is_locked() const { return m_locker != -1; }
-		uint32_t lock_depth() const { return m_lock_depth; }
+		pid_t locker() const override { return m_locker; }
+		bool is_locked() const override { return m_locker != -1; }
+		uint32_t lock_depth() const override { return m_lock_depth; }
 
 	private:
 		BAN::Atomic<pid_t>	m_locker		{ -1 };
 		uint32_t			m_lock_depth	{  0 };
 	};
 
-	class PriorityMutex
+	class PriorityMutex : public BaseMutex
 	{
 		BAN_NON_COPYABLE(PriorityMutex);
 		BAN_NON_MOVABLE(PriorityMutex);
@@ -85,13 +98,15 @@ namespace Kernel
 	public:
 		PriorityMutex() = default;
 
-		void lock()
+		void lock() override
 		{
 			const auto tid = Thread::current_tid();
+
 			if (tid == m_locker)
 				ASSERT(m_lock_depth > 0);
 			else
 			{
+				ASSERT(!tid || !Thread::current().has_spinlock());
 				bool has_priority = tid ? !Thread::current().is_userspace() : true;
 				if (has_priority)
 					m_queue_length++;
@@ -108,9 +123,10 @@ namespace Kernel
 			m_lock_depth++;
 		}
 
-		bool try_lock()
+		bool try_lock() override
 		{
 			const auto tid = Thread::current_tid();
+
 			if (tid == m_locker)
 				ASSERT(m_lock_depth > 0);
 			else
@@ -129,7 +145,7 @@ namespace Kernel
 			return true;
 		}
 
-		void unlock()
+		void unlock() override
 		{
 			const auto tid = Thread::current_tid();
 			ASSERT(m_locker == tid);
@@ -145,9 +161,9 @@ namespace Kernel
 			}
 		}
 
-		pid_t locker() const { return m_locker; }
-		bool is_locked() const { return m_locker != -1; }
-		uint32_t lock_depth() const { return m_lock_depth; }
+		pid_t locker() const override { return m_locker; }
+		bool is_locked() const override { return m_locker != -1; }
+		uint32_t lock_depth() const override { return m_lock_depth; }
 
 	private:
 		BAN::Atomic<pid_t>		m_locker		{ -1 };

kernel/include/kernel/Lock/RWLock.h

@@ -0,0 +1,97 @@
+#pragma once
+
+#include <kernel/Lock/Mutex.h>
+#include <kernel/Lock/LockGuard.h>
+
+namespace Kernel
+{
+
+	class RWLock
+	{
+		BAN_NON_COPYABLE(RWLock);
+		BAN_NON_MOVABLE(RWLock);
+	public:
+		RWLock() = default;
+
+		void rd_lock()
+		{
+			LockGuard _(m_mutex);
+			while (m_writers_waiting > 0 || m_writer_active)
+				m_thread_blocker.block_indefinite(&m_mutex);
+			m_readers_active++;
+		}
+
+		void rd_unlock()
+		{
+			LockGuard _(m_mutex);
+			if (--m_readers_active == 0)
+				m_thread_blocker.unblock();
+		}
+
+		void wr_lock()
+		{
+			LockGuard _(m_mutex);
+			m_writers_waiting++;
+			while (m_readers_active > 0 || m_writer_active)
+				m_thread_blocker.block_indefinite(&m_mutex);
+			m_writers_waiting--;
+			m_writer_active = true;
+		}
+
+
+		void wr_unlock()
+		{
+			LockGuard _(m_mutex);
+			m_writer_active = false;
+			m_thread_blocker.unblock();
+		}
+
+	private:
+		Mutex m_mutex;
+		ThreadBlocker m_thread_blocker;
+		uint32_t m_readers_active  { 0 };
+		uint32_t m_writers_waiting { 0 };
+		bool m_writer_active { false };
+	};
+
+	class RWLockRDGuard
+	{
+		BAN_NON_COPYABLE(RWLockRDGuard);
+		BAN_NON_MOVABLE(RWLockRDGuard);
+	public:
+		RWLockRDGuard(RWLock& lock)
+			: m_lock(lock)
+		{
+			m_lock.rd_lock();
+		}
+
+		~RWLockRDGuard()
+		{
+			m_lock.rd_unlock();
+		}
+
+	private:
+		RWLock& m_lock;
+	};
+
+	class RWLockWRGuard
+	{
+		BAN_NON_COPYABLE(RWLockWRGuard);
+		BAN_NON_MOVABLE(RWLockWRGuard);
+	public:
+		RWLockWRGuard(RWLock& lock)
+			: m_lock(lock)
+		{
+			m_lock.wr_lock();
+		}
+
+		~RWLockWRGuard()
+		{
+			m_lock.wr_unlock();
+		}
+
+	private:
+		RWLock& m_lock;
+	};
+
+}

kernel/include/kernel/Lock/SpinLock.h

@@ -18,42 +18,13 @@ namespace Kernel
 	public:
 		SpinLock() = default;
 
-		InterruptState lock()
-		{
-			auto state = Processor::get_interrupt_state();
-			Processor::set_interrupt_state(InterruptState::Disabled);
+		InterruptState lock();
 
-			auto id = Processor::current_id().as_u32();
-			ASSERT(m_locker.load(BAN::MemoryOrder::memory_order_relaxed) != id);
+		bool try_lock_interrupts_disabled();
 
-			auto expected = PROCESSOR_NONE.as_u32();
-			while (!m_locker.compare_exchange(expected, id, BAN::MemoryOrder::memory_order_acquire))
-			{
-				Processor::pause();
-				expected = PROCESSOR_NONE.as_u32();
-			}
+		void unlock(InterruptState state);
 
-			return state;
-		}
-
-		bool try_lock_interrupts_disabled()
-		{
-			ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
-
-			auto id = Processor::current_id().as_u32();
-			ASSERT(m_locker.load(BAN::MemoryOrder::memory_order_relaxed) != id);
-
-			auto expected = PROCESSOR_NONE.as_u32();
-			return m_locker.compare_exchange(expected, id, BAN::MemoryOrder::memory_order_acquire);
-		}
-
-		void unlock(InterruptState state)
-		{
-			ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
-			ASSERT(current_processor_has_lock());
-			m_locker.store(PROCESSOR_NONE.as_u32(), BAN::MemoryOrder::memory_order_release);
-			Processor::set_interrupt_state(state);
-		}
+		uint32_t lock_depth() const { return current_processor_has_lock(); }
 
 		bool current_processor_has_lock() const
 		{
@@ -72,36 +43,13 @@ namespace Kernel
 	public:
 		RecursiveSpinLock() = default;
 
-		InterruptState lock()
-		{
-			auto state = Processor::get_interrupt_state();
-			Processor::set_interrupt_state(InterruptState::Disabled);
+		InterruptState lock();
 
-			auto id = Processor::current_id().as_u32();
+		bool try_lock_interrupts_disabled();
 
-			ProcessorID::value_type expected = PROCESSOR_NONE.as_u32();
-			while (!m_locker.compare_exchange(expected, id, BAN::MemoryOrder::memory_order_acq_rel))
-			{
-				if (expected == id)
-					break;
-				Processor::pause();
-				expected = PROCESSOR_NONE.as_u32();
-			}
+		void unlock(InterruptState state);
 
-			m_lock_depth++;
-
-			return state;
-		}
-
-		void unlock(InterruptState state)
-		{
-			ASSERT(Processor::get_interrupt_state() == InterruptState::Disabled);
-			ASSERT(current_processor_has_lock());
-			ASSERT(m_lock_depth > 0);
-			if (--m_lock_depth == 0)
-				m_locker.store(PROCESSOR_NONE.as_u32(), BAN::MemoryOrder::memory_order_release);
-			Processor::set_interrupt_state(state);
-		}
+		uint32_t lock_depth() const { return m_lock_depth; }
 
 		bool current_processor_has_lock() const
 		{
@@ -113,6 +61,9 @@ namespace Kernel
 		uint32_t                             m_lock_depth { 0 };
 	};
 
+	template<typename Lock>
+	class SpinLockGuardAsMutex;
+
 	template<typename Lock>
 	class SpinLockGuard
 	{
@@ -134,6 +85,7 @@ namespace Kernel
 	private:
 		Lock& m_lock;
 		InterruptState m_state;
+		friend class SpinLockGuardAsMutex<Lock>;
 	};
 
 }

kernel/include/kernel/Lock/SpinLockAsMutex.h

@@ -0,0 +1,64 @@
+#pragma once
+
+#include <kernel/Lock/SpinLock.h>
+#include <kernel/Lock/Mutex.h>
+
+namespace Kernel
+{
+
+	// FIXME: These classes are HACKS to allow passing spinlock
+	//        to unblock functions. Write a better API that either
+	//        allows passing spinlocks or do something cleaner that
+	//        whatever shit this is
+
+	template<typename Lock>
+	class SpinLockAsMutex : public BaseMutex
+	{
+	public:
+		SpinLockAsMutex(Lock& lock, InterruptState state)
+			: m_lock(lock)
+			, m_lock_depth(lock.lock_depth())
+			, m_state(state)
+			, m_locker(Thread::current_tid())
+		{
+			ASSERT(m_lock.current_processor_has_lock());
+		}
+
+		void lock() override
+		{
+			m_lock.lock();
+			m_lock_depth++;
+		}
+
+		bool try_lock() override
+		{
+			lock();
+			return true;
+		}
+
+		void unlock() override
+		{
+			m_lock.unlock(--m_lock_depth ? InterruptState::Disabled : m_state);
+		}
+
+		pid_t locker() const override { return is_locked() ? m_locker : -1; }
+		bool is_locked() const override { return m_lock_depth; }
+		uint32_t lock_depth() const override { return m_lock_depth; }
+
+	private:
+		SpinLock& m_lock;
+		uint32_t m_lock_depth { 0 };
+		InterruptState m_state;
+		const pid_t m_locker;
+	};
+
+	template<typename Lock>
+	class SpinLockGuardAsMutex : public SpinLockAsMutex<Lock>
+	{
+	public:
+		SpinLockGuardAsMutex(SpinLockGuard<Lock>& guard)
+			: SpinLockAsMutex<Lock>(guard.m_lock, guard.m_state)
+		{}
+	};
+
+}

kernel/include/kernel/Memory/ByteRingBuffer.h

@@ -0,0 +1,76 @@
+#pragma once
+
+#include <BAN/ByteSpan.h>
+#include <BAN/UniqPtr.h>
+#include <BAN/Vector.h>
+
+#include <kernel/Memory/Types.h>
+
+namespace Kernel
+{
+
+	class ByteRingBuffer
+	{
+	public:
+		static BAN::ErrorOr<BAN::UniqPtr<ByteRingBuffer>> create(size_t size);
+		~ByteRingBuffer();
+
+		void push(BAN::ConstByteSpan data)
+		{
+			ASSERT(data.size() + m_size <= m_capacity);
+			uint8_t* buffer_head = reinterpret_cast<uint8_t*>(m_vaddr) + (m_tail + m_size) % m_capacity;
+			memcpy(buffer_head, data.data(), data.size());
+			m_size += data.size();
+		}
+
+		void pop(size_t size)
+		{
+			ASSERT(size <= m_size);
+			m_tail = (m_tail + size) % m_capacity;
+			m_size -= size;
+		}
+
+		void pop_back(size_t size)
+		{
+			ASSERT(size <= m_size);
+			m_size -= size;
+		}
+
+		BAN::ConstByteSpan get_data() const
+		{
+			const uint8_t* base = reinterpret_cast<const uint8_t*>(m_vaddr);
+			return { base + m_tail, m_size };
+		}
+
+		uint8_t front() const
+		{
+			ASSERT(!empty());
+			return reinterpret_cast<const uint8_t*>(m_vaddr)[m_tail];
+		}
+
+		uint8_t back() const
+		{
+			ASSERT(!empty());
+			return reinterpret_cast<const uint8_t*>(m_vaddr)[m_tail + m_size];
+		}
+
+		bool empty() const { return m_size == 0; }
+		bool full() const { return m_size == m_capacity; }
+		size_t free() const { return m_capacity - m_size; }
+		size_t size() const { return m_size; }
+		size_t capacity() const { return m_capacity; }
+
+	private:
+		ByteRingBuffer(size_t capacity)
+			: m_capacity(capacity)
+		{ }
+
+	private:
+		size_t m_size { 0 };
+		size_t m_tail { 0 };
+		const size_t m_capacity;
+
+		vaddr_t m_vaddr { 0 };
+	};
+
+}

kernel/include/kernel/Memory/FileBackedRegion.h

@@ -27,18 +27,19 @@ namespace Kernel
 		BAN_NON_MOVABLE(FileBackedRegion);
 
 	public:
-		static BAN::ErrorOr<BAN::UniqPtr<FileBackedRegion>> create(BAN::RefPtr<Inode>, PageTable&, off_t offset, size_t size, AddressRange address_range, Type, PageTable::flags_t);
+		static BAN::ErrorOr<BAN::UniqPtr<FileBackedRegion>> create(BAN::RefPtr<Inode>, PageTable&, off_t offset, size_t size, AddressRange address_range, Type, PageTable::flags_t, int status_flags);
 		~FileBackedRegion();
 
-		virtual BAN::ErrorOr<void> msync(vaddr_t, size_t, int) override;
+		BAN::ErrorOr<void> msync(vaddr_t, size_t, int) override;
 
-		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) override;
+		BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) override;
+		BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> split(size_t offset) override;
 
 	protected:
-		virtual BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t vaddr, bool wants_write) override;
+		BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t vaddr, bool wants_write) override;
 
 	private:
-		FileBackedRegion(BAN::RefPtr<Inode>, PageTable&, off_t offset, ssize_t size, Type flags, PageTable::flags_t page_flags);
+		FileBackedRegion(BAN::RefPtr<Inode>, PageTable&, off_t offset, ssize_t size, Type type, PageTable::flags_t flags, int status_flags);
 
 	private:
 		BAN::RefPtr<Inode> m_inode;

kernel/include/kernel/Memory/MemoryBackedRegion.h

@@ -11,22 +11,23 @@ namespace Kernel
 		BAN_NON_MOVABLE(MemoryBackedRegion);
 
 	public:
-		static BAN::ErrorOr<BAN::UniqPtr<MemoryBackedRegion>> create(PageTable&, size_t size, AddressRange, Type, PageTable::flags_t);
+		static BAN::ErrorOr<BAN::UniqPtr<MemoryBackedRegion>> create(PageTable&, size_t size, AddressRange, Type, PageTable::flags_t, int status_flags);
 		~MemoryBackedRegion();
 
-		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) override;
+		BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) override;
+		BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> split(size_t offset) override;
 
-		virtual BAN::ErrorOr<void> msync(vaddr_t, size_t, int) override { return {}; }
+		BAN::ErrorOr<void> msync(vaddr_t, size_t, int) override { return {}; }
 
 		// Copy data from buffer into this region
 		// This can fail if no memory is mapped and no free memory was available
 		BAN::ErrorOr<void> copy_data_to_region(size_t offset_into_region, const uint8_t* buffer, size_t buffer_size);
 
 	protected:
-		virtual BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t vaddr, bool wants_write) override;
+		BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t vaddr, bool wants_write) override;
 
 	private:
-		MemoryBackedRegion(PageTable&, size_t size, Type, PageTable::flags_t);
+		MemoryBackedRegion(PageTable&, size_t size, Type, PageTable::flags_t, int status_flags);
 	};
 
 }

kernel/include/kernel/Memory/MemoryRegion.h

@@ -1,10 +1,10 @@
 #pragma once
 
 #include <BAN/UniqPtr.h>
+#include <kernel/Lock/Mutex.h>
 #include <kernel/Memory/PageTable.h>
 #include <kernel/Memory/Types.h>
-
-#include <stddef.h>
+#include <kernel/ThreadBlocker.h>
 
 namespace Kernel
 {
@@ -33,15 +33,25 @@ namespace Kernel
 		bool contains(vaddr_t address) const;
 		bool contains_fully(vaddr_t address, size_t size) const;
 		bool overlaps(vaddr_t address, size_t size) const;
+		bool is_contained_by(vaddr_t address, size_t size) const;
 
 		bool writable() const { return m_flags & PageTable::Flags::ReadWrite; }
 
 		size_t size() const { return m_size; }
 		vaddr_t vaddr() const { return m_vaddr; }
 
+		int status_flags() const { return m_status_flags; }
+		Type type() const { return m_type; }
+		PageTable::flags_t flags() const { return m_flags; }
+
 		size_t virtual_page_count() const { return BAN::Math::div_round_up<size_t>(m_size, PAGE_SIZE); }
 		size_t physical_page_count() const { return m_physical_page_count; }
 
+		void pin();
+		void unpin();
+		void wait_not_pinned();
+
+		BAN::ErrorOr<void> mprotect(PageTable::flags_t);
 		virtual BAN::ErrorOr<void> msync(vaddr_t, size_t, int) = 0;
 
 		// Returns error if no memory was available
@@ -50,20 +60,26 @@ namespace Kernel
 		BAN::ErrorOr<bool> allocate_page_containing(vaddr_t address, bool wants_write);
 
 		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) = 0;
+		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> split(size_t offset) = 0;
 
 	protected:
-		MemoryRegion(PageTable&, size_t size, Type type, PageTable::flags_t flags);
+		MemoryRegion(PageTable&, size_t size, Type type, PageTable::flags_t flags, int status_flags);
 		BAN::ErrorOr<void> initialize(AddressRange);
 
 		virtual BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t address, bool wants_write) = 0;
 
 	protected:
 		PageTable& m_page_table;
-		const size_t m_size;
+		size_t m_size { 0 };
 		const Type m_type;
-		const PageTable::flags_t m_flags;
+		PageTable::flags_t m_flags;
+		const int m_status_flags;
 		vaddr_t m_vaddr { 0 };
 		size_t m_physical_page_count { 0 };
+
+		Mutex m_pinned_mutex;
+		BAN::Atomic<size_t> m_pinned_count { 0 };
+		ThreadBlocker m_pinned_blocker;
 	};
 
 }

kernel/include/kernel/Memory/PageTable.h

@@ -112,7 +112,7 @@ namespace Kernel
 		bool is_page_free(vaddr_t) const;
 		bool is_range_free(vaddr_t, size_t bytes) const;
 
-		bool reserve_page(vaddr_t, bool only_free = true);
+		bool reserve_page(vaddr_t, bool only_free = true, bool send_smp_message = true);
 		bool reserve_range(vaddr_t, size_t bytes, bool only_free = true);
 
 		vaddr_t reserve_free_page(vaddr_t first_address, vaddr_t last_address = UINTPTR_MAX);
@@ -133,7 +133,7 @@ namespace Kernel
 		void map_kernel_memory();
 		void prepare_fast_page();
 
-		static void invalidate(vaddr_t, bool send_smp_message);
+		void invalidate(vaddr_t, bool send_smp_message);
 
 		static void map_fast_page(paddr_t);
 		static void unmap_fast_page();

kernel/include/kernel/Memory/SharedMemoryObject.h

@@ -6,6 +6,8 @@
 #include <kernel/Lock/SpinLock.h>
 #include <kernel/Memory/MemoryRegion.h>
 
+#include <fcntl.h>
+
 namespace Kernel
 {
 
@@ -55,15 +57,17 @@ namespace Kernel
 	public:
 		static BAN::ErrorOr<BAN::UniqPtr<SharedMemoryObject>> create(BAN::RefPtr<SharedMemoryObjectManager::Object>, PageTable&, AddressRange);
 
-		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) override;
-		virtual BAN::ErrorOr<void> msync(vaddr_t, size_t, int) override { return {}; }
+		BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) override;
+		BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> split(size_t offset) override;
+
+		BAN::ErrorOr<void> msync(vaddr_t, size_t, int) override { return {}; }
 
 	protected:
-		virtual BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t vaddr, bool wants_write) override;
+		BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t vaddr, bool wants_write) override;
 
 	private:
 		SharedMemoryObject(BAN::RefPtr<SharedMemoryObjectManager::Object> object, PageTable& page_table)
-			: MemoryRegion(page_table, object->size, MemoryRegion::Type::SHARED, object->flags)
+			: MemoryRegion(page_table, object->size, MemoryRegion::Type::SHARED, object->flags, O_EXEC | O_RDWR)
 			, m_object(object)
 		{ }
 

kernel/include/kernel/Memory/Types.h

@@ -4,7 +4,7 @@
 
 #if ARCH(x86_64)
 	#define KERNEL_OFFSET 0xFFFFFFFF80000000
-	#define USERSPACE_END 0xFFFF800000000000
+	#define USERSPACE_END 0x800000000000
 #elif ARCH(i686)
 	#define KERNEL_OFFSET 0xC0000000
 	#define USERSPACE_END 0xC0000000
@@ -12,6 +12,7 @@
 	#error
 #endif
 
+#undef PAGE_SIZE
 #define PAGE_SIZE ((uintptr_t)4096)
 #define PAGE_SIZE_SHIFT 12
 #define PAGE_ADDR_MASK (~(uintptr_t)0xFFF)

kernel/include/kernel/Memory/VirtualRange.h

@@ -15,21 +15,21 @@ namespace Kernel
 
 	public:
 		// Create virtual range to fixed virtual address
-		static BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> create_to_vaddr(PageTable&, vaddr_t, size_t, PageTable::flags_t flags, bool preallocate_pages);
+		static BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> create_to_vaddr(PageTable&, vaddr_t, size_t, PageTable::flags_t flags, bool preallocate_pages, bool add_guard_pages);
 		// Create virtual range to virtual address range
-		static BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> create_to_vaddr_range(PageTable&, vaddr_t vaddr_start, vaddr_t vaddr_end, size_t, PageTable::flags_t flags, bool preallocate_pages);
+		static BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> create_to_vaddr_range(PageTable&, vaddr_t vaddr_start, vaddr_t vaddr_end, size_t, PageTable::flags_t flags, bool preallocate_pages, bool add_guard_pages);
 		~VirtualRange();
 
 		BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> clone(PageTable&);
 
-		vaddr_t vaddr() const { return m_vaddr; }
-		size_t size() const { return m_size; }
+		vaddr_t vaddr() const { return m_vaddr + (m_has_guard_pages ? PAGE_SIZE : 0); }
+		size_t size() const { return m_size - (m_has_guard_pages ? 2 * PAGE_SIZE : 0); }
 		PageTable::flags_t flags() const { return m_flags; }
 
 		paddr_t paddr_of(vaddr_t vaddr) const
 		{
 			ASSERT(vaddr % PAGE_SIZE == 0);
-			const size_t index = (vaddr - m_vaddr) / PAGE_SIZE;
+			const size_t index = (vaddr - this->vaddr()) / PAGE_SIZE;
 			ASSERT(index < m_paddrs.size());
 			const paddr_t paddr = m_paddrs[index];
 			ASSERT(paddr);
@@ -38,15 +38,16 @@ namespace Kernel
 
 		bool contains(vaddr_t address) const { return vaddr() <= address && address < vaddr() + size(); }
 
-		BAN::ErrorOr<void> allocate_page_for_demand_paging(vaddr_t address);
+		BAN::ErrorOr<bool> allocate_page_for_demand_paging(vaddr_t address);
 
 	private:
-		VirtualRange(PageTable&, bool preallocated, vaddr_t, size_t, PageTable::flags_t);
+		VirtualRange(PageTable&, bool preallocated, bool has_guard_pages, vaddr_t, size_t, PageTable::flags_t);
 		BAN::ErrorOr<void> initialize();
 
 	private:
 		PageTable&               m_page_table;
 		const bool               m_preallocated;
+		const bool               m_has_guard_pages;
 		const vaddr_t            m_vaddr;
 		const size_t             m_size;
 		const PageTable::flags_t m_flags;