GP Fault handler

Reviewed-on: #7

Makefile

@@ -1,4 +1,4 @@
-SOURCES = src/io/kbd/ps2.c src/io/serial/serial.c src/io/term/printf.c src/io/term/term.c src/idt/idt.c src/mem/gdt/gdt.c src/mem/misc/utils.c src/time/timer.c src/kmain.c
+SOURCES = src/mem/heap/kheap.c src/mem/paging/vmm.c src/mem/paging/paging.c src/mem/paging/pmm.c src/string/string.c src/io/kbd/ps2.c src/io/serial/serial.c src/io/term/printf.c src/io/term/term.c src/idt/idt.c src/mem/gdt/gdt.c src/mem/misc/utils.c src/time/timer.c src/kmain.c
 
 build:
 	rm -f *.o
@@ -30,7 +30,7 @@ build-iso: limine/limine build
 	./limine/limine bios-install pepper.iso
 
 debug:
-	qemu-system-x86_64 -drive file=pepper.iso -s -S -d int -no-reboot &
+	qemu-system-x86_64 -drive file=pepper.iso -s -S -d int -no-reboot -no-shutdown &
 	gdb pepperk --command=debug.gdb
 
 run: build-iso

README.md

@@ -7,6 +7,25 @@ First install the dependencies: `sudo apt install xorriso make qemu-system`
 Then, to compile the kernel and make an ISO image file: `make build-iso`
 To run it with QEMU, `make run`
 
+## TODO
+
+The basics that I'm targeting are:
+
+- Fix terminal driver (backspace issues, scrolling) OR add Flanterm or equivalent
+- Implement paging / see what Limine does at boot with memory management
+- Implement tasks, and task switching
+- Load an executable
+- Scheduler (round-robin using the PIT timer interrupt)
+- Filesystem (TAR for read-only initfs, then maybe read-write using FAT12/16/32
+- Getting to userspace (syscalls)
+- Porting musl libc or equivalent
+
+In the future, maybe?
+
+- SMP support
+- Parsing the ACPI tables and using them for something
+- Replacing the PIT timer with APIC
+
 ## Thanks
 
 PepperOS wouldn't be possible without the following freely-licensed software:

src/idt/idt.c

@@ -1,9 +1,10 @@
 #include "idt.h"
 #include <stdint.h>
 #include <stddef.h>
-#include "../io/serial/serial.h"
+#include "io/serial/serial.h"
-#include "../io/kbd/ps2.h"
+#include "io/kbd/ps2.h"
 #include <kernel.h>
+#include <stdbool.h>
 
 struct interrupt_descriptor idt[256];
 struct idtr idt_reg;
@@ -53,6 +54,64 @@ void idt_init()
     DEBUG("IDT initialized");
 }
 
+static inline uint64_t read_cr2(void)
+{
+    uint64_t val;
+    asm volatile ("mov %%cr2, %0" : "=r"(val));
+    return val;
+}
+
+static void page_fault_handler(struct cpu_status_t* ctx)
+{
+    // It could be used to remap pages etc. to fix the fault, but right now what I'm more
+    // interested in is getting more info out of those numbers cause i'm lost each time i have 
+    // to read all this mess
+    uint64_t cr2 = read_cr2();
+    
+    DEBUG("\x1b[38;5;231mPage Fault at rip=0x%p, err=%u (%s%s%s%s%s%s%s%s) when accessing addr=0x%p\x1b[0m", ctx->iret_rip, ctx->error_code,
+    CHECK_BIT(ctx->error_code, 0) ? "PAGE_PROTECTION_VIOLATION " : "PAGE_NOT_PRESENT ",
+    CHECK_BIT(ctx->error_code, 1) ? "ON_WRITE " : "ON_READ ",
+    CHECK_BIT(ctx->error_code, 2) ? "IN_USER_MODE" : "IN_KERNEL_MODE",
+    CHECK_BIT(ctx->error_code, 3) ? " WAS_RESERVED" : "",
+    CHECK_BIT(ctx->error_code, 4) ? " ON_INSTRUCTION_FETCH" : "",
+    CHECK_BIT(ctx->error_code, 5) ? " PK_VIOLATION" : "",
+    CHECK_BIT(ctx->error_code, 6) ? " ON_SHADOWSTACK_ACCESS" : "",
+    CHECK_BIT(ctx->error_code, 7) ? " SGX_VIOLATION" : "",
+    cr2);
+
+    /* if (CHECK_BIT(ctx->error_code, 0))
+    {
+        panic(ctx);
+    } */
+    panic(ctx);
+}
+
+static void gp_fault_handler(struct cpu_status_t* ctx)
+{
+    DEBUG("\x1b[38;5;231mGeneral Protection Fault at rip=0x%p, err=%u (%s)\x1b[0m",
+    ctx->iret_rip,
+    ctx->error_code,
+    (ctx->error_code == 0) ? "NOT_SEGMENT_RELATED" : "SEGMENT_RELATED");
+
+    // Segment-related
+    if (ctx->error_code != 0)
+    {
+        bool is_external = CHECK_BIT(ctx->error_code, 0);
+        // is it IDT, GDT, LDT?
+        uint8_t table = ctx->error_code & 0x6; // 0b110 (isolate table)
+        uint16_t index = ctx->error_code & 0xFFF8; // 13*1 1111111111111 + 000 = 1111111111111000
+    
+        char* table_names[4] = {"GDT", "IDT", "LDT", "IDT"};
+
+        DEBUG("\x1b[38;5;231m%s in %s index %u\x1b[0m",
+        is_external ? "EXTERNAL" : "INTERNAL",
+        table_names[table],
+        index);
+    }
+
+    panic(ctx);
+}
+
 struct cpu_status_t* interrupt_dispatch(struct cpu_status_t* context)
 {
     switch(context->vector_number)
@@ -97,10 +156,11 @@ struct cpu_status_t* interrupt_dispatch(struct cpu_status_t* context)
             DEBUG("Stack-Segment Fault!");
             break;
         case 13:
-            DEBUG("General Protection Fault!");
+            gp_fault_handler(context);
             break;
         case 14:
-            DEBUG("Page Fault!");
+            // Better debugging for page faults...
+            page_fault_handler(context);
             break;
         case 15:
             DEBUG("Intel Reserved Interrupt! (Achievement unlocked: How Did We Get Here?)");

src/io/kbd/ps2.c

@@ -1,9 +1,9 @@
 // PS/2 Keyboard support
 
-#include "../serial/serial.h"
+#include "io/serial/serial.h"
 #include "ps2.h"
 #include <stdint.h>
-#include "../term/term.h"
+#include "io/term/term.h"
 #include <kernel.h>
 
 // The key status bitfield will be used to see if ALT, CONTROL, or SHIFT is pressed

src/io/term/term.c

@@ -1,11 +1,17 @@
 // Terminal output
+/*
+There are a couple of bugs here and there but for now I don't care too much
+because this shitty implementation will be replaced one day by Flanterm
+(once memory management is okay: paging & kernel malloc)
+*/
 
 #include <limine.h>
 #include <stddef.h>
 #include <kernel.h>
 #include "term.h"
+#include "mem/misc/utils.h"
 
-extern struct limine_framebuffer* framebuffer;
+extern struct boot_context boot_ctx;
 
 // Importing the PSF object file
 extern unsigned char _binary_zap_light16_psf_start[];
@@ -28,14 +34,19 @@ Cursor cursor = {0};
 
 unsigned char* fb;
 
+struct limine_framebuffer* framebuffer;
+
+uint8_t lines_length[MAX_LINES];
+
 int term_init()
 {
 	// Get framebuffer address from Limine struct
 	
-	if (framebuffer)
+	if (boot_ctx.fb)
 	{
-		fb = framebuffer->address;
+		fb = boot_ctx.fb->address;
-		DEBUG("terminal initialized");
+		framebuffer = boot_ctx.fb;
+		DEBUG("terminal initialized, fb=0x%p (width=%u height=%u pitch=%u bpp=%u)", fb, framebuffer->width, framebuffer->height, framebuffer->pitch, framebuffer->bpp);
 		return 0;
 	}
 	return -ENOMEM;
@@ -79,10 +90,56 @@ static void erase_char(int px, int py)
 	}
 }
 
+static inline size_t term_max_lines(void)
+{
+	return framebuffer->height / FONT_HEIGHT;
+}
+
+void term_scroll()
+{
+	const size_t row_height = FONT_HEIGHT;
+	const size_t row_bytes  = framebuffer->pitch;
+	const size_t screen_rows = framebuffer->height;
+
+	// Move framebuffer up by one text row 
+	//memmove(fb, fb + row_height * row_bytes, (screen_rows - row_height) * row_bytes);
+
+	for (size_t i = 0; i < (screen_rows - row_height) * row_bytes; i++)
+	{
+		fb[i] = fb[i + row_height * row_bytes];
+	}
+
+	// Clear last text row
+	size_t clear_start = (screen_rows - row_height) * row_bytes;
+	memset(fb + clear_start, 0, row_height * row_bytes);
+
+	// Shift line lengths by 1 (for backspace handling)
+	size_t max_lines = term_max_lines();
+	for (size_t i = 1; i < max_lines; i++)
+	{
+		lines_length[i - 1] = lines_length[i];
+	}
+
+	lines_length[max_lines - 1] = 0;
+
+	if (cursor.y > 0)
+	{
+		cursor.y--;
+	}
+}
+
+
 void putchar(char c)
 {
+	if ((c == '\n') && ((cursor.y+1)*FONT_HEIGHT >= framebuffer->height))
+	{
+		term_scroll();
+		return;
+	} 
+
 	if (c == '\n')
 	{
+		lines_length[cursor.y] = cursor.x;
 		cursor.x = 0;
 		cursor.y++;
 		return;
@@ -102,7 +159,8 @@ void putchar(char c)
 		if (cursor.x == 0)
 		{
 			cursor.y--;
-			cursor.x = (framebuffer->width / FONT_WIDTH) -1; // here
+			// cursor.x = (framebuffer->width / FONT_WIDTH) -1; // here
+			cursor.x = lines_length[cursor.y];
 		}
 		else {
 			cursor.x--;
@@ -120,6 +178,11 @@ void putchar(char c)
 		cursor.y++;
 	}
 
+	if ((cursor.y+1)*FONT_HEIGHT >= framebuffer->height)
+	{
+		term_scroll();
+	}
+
 	int px = cursor.x * FONT_WIDTH;
 	int py = cursor.y * FONT_HEIGHT;
 	draw_char(c, px, py, WHITE, BLACK);

src/io/term/term.h

@@ -11,6 +11,8 @@ enum TermColors
     WHITE = 0xffffff
 };
 
+#define MAX_LINES 256
+
 #define PSF1_FONT_MAGIC 0x0436
 
 typedef struct

src/kernel.h

@@ -1,6 +1,10 @@
 #ifndef KERNEL_H
 #define KERNEL_H
 
+#define PEPPEROS_VERSION_MAJOR "0"
+#define PEPPEROS_VERSION_MINOR "0"
+#define PEPPEROS_VERSION_PATCH "1"
+
 enum ErrorCodes
 {
 	ENOMEM,
@@ -12,10 +16,24 @@ enum ErrorCodes
 
 #include "io/serial/serial.h"
 #include "io/term/printf.h"
+#include "idt/idt.h"
 
-// Still lacks print formatting...
+#define DEBUG(log, ...) fctprintf((void*)&skputc, 0, "debug: [%s]: " log "\r\n", __FILE__, ##__VA_ARGS__)
-#define DEBUG(log, ...) \
+
-	printf("debug: [%s]: " log "\n", __FILE__, ##__VA_ARGS__); \
+#define CHECK_BIT(var,pos) ((var) & (1<<(pos)))
-	fctprintf((void*)&skputc, 0, "debug: [%s]: %s\n", __FILE__, log)
+
+// printf("debug: [%s]: " log "\n", __FILE__, ##__VA_ARGS__);
+
+void panic(struct cpu_status_t* ctx);
+void hcf();
+#define assert(check) do { if(!(check)) hcf(); } while(0)
+
+struct boot_context
+{
+	struct limine_framebuffer* fb;
+	struct limine_memmap_response* mmap;
+	struct limine_hhdm_response* hhdm;
+	struct limine_kernel_address_response* kaddr;
+};
 
 #endif

src/kmain.c

@@ -10,6 +10,10 @@
 #include "kernel.h"
 #include "time/timer.h"
 #include "io/kbd/ps2.h"
+#include "mem/paging/pmm.h"
+#include "mem/paging/paging.h"
+#include "mem/paging/vmm.h"
+#include "mem/heap/kheap.h"
 
 // Limine version used
 __attribute__((used, section(".limine_requests")))
@@ -22,16 +26,35 @@ static volatile struct limine_framebuffer_request framebuffer_request = {
 	.revision = 0
 };
 
+// Memory map request
+__attribute__((used, section(".limine_requests")))
+static volatile struct limine_memmap_request memmap_request = {
+	.id = LIMINE_MEMMAP_REQUEST,
+	.revision = 0
+};
+
+// Higher Half Direct Map
+__attribute__((used, section(".limine_requests")))
+static volatile struct limine_hhdm_request hhdm_request = {
+	.id = LIMINE_HHDM_REQUEST,
+	.revision = 0
+};
+
+// Executable Address/Kernel Address (find base phys/virt address of kernel)
+__attribute__((used, section(".limine_requests")))
+static volatile struct limine_kernel_address_request kerneladdr_request = {
+	.id = LIMINE_KERNEL_ADDRESS_REQUEST,
+	.revision = 0
+};
+
 __attribute__((used, section(".limine_requests_start")))
 static volatile LIMINE_REQUESTS_START_MARKER;
 
 __attribute__((used, section(".limine_requests_end")))
 static volatile LIMINE_REQUESTS_END_MARKER;
 
-struct limine_framebuffer* framebuffer;
+// Panic (should dump registers etc. in the future)
-
+void hcf()
-// Panic 
-static void hcf()
 {
 	for (;;)
 	{
@@ -39,29 +62,67 @@ static void hcf()
 	}
 }
 
+void panic(struct cpu_status_t* ctx)
+{
+	DEBUG("\x1b[38;5;231m\x1b[48;5;196mKernel panic!!!\x1b[0m at rip=%p\nSomething went horribly wrong! vect=0x%.2x errcode=0x%x\nrax=%p rbx=%p rcx=%p rdx=%p\nrsi=%p rdi=%p  r8=%p  r9=%p\nr10=%p r11=%p r12=%p r13=%p\nr14=%p r15=%p\n\nflags=%p\nstack at rbp=%p\nHalting...",
+		ctx->iret_rip,
+		ctx->vector_number, ctx->error_code, ctx->rax, ctx->rbx, ctx->rcx, ctx->rdx, ctx->rsi, ctx->rdi,
+		ctx->r8, ctx->r9, ctx->r10, ctx->r11, ctx->r12, ctx->r13, ctx->r14, ctx->r15, ctx->iret_flags,
+		ctx->rbp);
+	hcf();
+}
+
+const char* splash = "pepperOS version "PEPPEROS_VERSION_MAJOR"."PEPPEROS_VERSION_MINOR"."PEPPEROS_VERSION_PATCH"\n";
+
+struct boot_context boot_ctx;
+
 // This is our entry point
 void kmain()
 {
 	if (!LIMINE_BASE_REVISION_SUPPORTED) hcf();
-	if (framebuffer_request.response == NULL || framebuffer_request.response->framebuffer_count < 1) hcf();
 
-	// Get the first framebuffer from the response
+	// Populate boot context
-	framebuffer = framebuffer_request.response->framebuffers[0];
+ 	boot_ctx.fb = framebuffer_request.response ? framebuffer_request.response->framebuffers[0] : NULL;
+	boot_ctx.mmap = memmap_request.response ? memmap_request.response : NULL;
+	boot_ctx.hhdm = hhdm_request.response ? hhdm_request.response : NULL;
+	boot_ctx.kaddr = kerneladdr_request.response ? kerneladdr_request.response : NULL;
 
-	term_init();
 	serial_init();
 
+	memmap_display(boot_ctx.mmap);
+	hhdm_display(boot_ctx.hhdm);
+	DEBUG("kernel: phys_base=0x%p virt_base=0x%p", boot_ctx.kaddr->physical_base, boot_ctx.kaddr->virtual_base);
+
 	CLEAR_INTERRUPTS;
 	gdt_init();
 	idt_init();
 	timer_init();
 	SET_INTERRUPTS;
 
+	pmm_init(boot_ctx.mmap, boot_ctx.hhdm);
+
+	// Remap kernel , HHDM and framebuffer
+	paging_init(boot_ctx.kaddr, boot_ctx.fb);
+	
+	kheap_init();
+
+	void* ptr = kmalloc(10); DEBUG("(KMALLOC TEST) Allocated 10 bytes at 0x%p", ptr);
+	void* ptr2 = kmalloc(200); DEBUG("(KMALLOC TEST) Allocated 200 bytes at 0x%p", ptr2);
+	kfree(ptr);
+	void* ptr3 = kmalloc(5); DEBUG("(KMALLOC TEST) Allocated 5 bytes at 0x%p", ptr3);
+
+	vmm_init();
+
 	keyboard_init(FR);
 
-	// Draw something
+	term_init();
-	printf("%s, %s!\n", "Hello", "world");
+	kputs(splash);
-	// Yoohoooooo!
+
-	DEBUG("kernel initialized successfully! hanging... wow=%d", 42);
+	for (int i=0; i<50; i++)
+	{
+		printf("testing, attention please %d\n", i);
+		timer_wait(1000);
+	}
+
 	hcf();
 }

src/mem/gdt/gdt.c

@@ -1,6 +1,6 @@
 #include "gdt.h"
 #include <stdint.h>
-#include "../../io/serial/serial.h"
+#include "io/serial/serial.h"
 #include <kernel.h>
 
 // Descriptors are 8-byte wide (64bits)

src/mem/heap/kheap.c

@@ -0,0 +1,106 @@
+#include "kheap.h"
+#include "mem/paging/paging.h"
+#include "mem/paging/pmm.h"
+#include <stddef.h>
+#include <kernel.h>
+
+extern uint64_t kernel_phys_base;
+extern uint64_t kernel_virt_base;
+
+uintptr_t kheap_start;
+
+static struct heap_block_t* head = NULL;
+static uintptr_t end;
+
+// Kernel root table (level 4)
+extern uint64_t *kernel_pml4;
+
+static void kheap_map_page()
+{
+    uintptr_t phys = pmm_alloc();
+    paging_map_page(kernel_pml4, end, phys, PTE_PRESENT | PTE_WRITABLE | PTE_NOEXEC);
+    end += PAGE_SIZE;
+    DEBUG("Mapped first kheap page");
+}
+
+void kheap_init()
+{
+    kheap_start = ALIGN_UP(kernel_virt_base + KERNEL_SIZE, PAGE_SIZE);
+    end = kheap_start;
+
+    // At least 1 page must be mapped for it to work
+    kheap_map_page();
+
+    // Give linked list head its properties
+    head = (struct heap_block_t*)kheap_start;
+    head->size = PAGE_SIZE - sizeof(struct heap_block_t);
+    head->free = true;
+    head->next = NULL;
+    DEBUG("kheap initialized, head=0x%p, size=%u", head, head->size);
+}
+
+void* kmalloc(size_t size)
+{
+    // No size, no memory allocated!
+    if (!size) return NULL;
+
+    struct heap_block_t* curr = head;
+
+    while (curr)
+    {
+        // Is block free and big enough for us?
+        if (curr->free && curr->size >= size)
+        {
+            // We split the block if it is big enough
+            if (curr->size > size + sizeof(struct heap_block_t))
+            {
+                struct heap_block_t* new_block = (struct heap_block_t*)((uintptr_t)curr + sizeof(struct heap_block_t) + size);
+                // We have to subtract the size of our block struct
+                new_block->size = curr->size - size - sizeof(struct heap_block_t);
+                new_block->free = true;
+
+                // Then we chain up the block in the list
+                new_block->next = curr->next;
+                curr->next = new_block;
+
+                curr->size = size;
+            }
+
+            // Found a good block, we return it
+            curr->free = false;
+            return (void*)((uintptr_t)curr + sizeof(struct heap_block_t));
+        }
+        // Continue browsing the list if nothing good was found yet
+        curr = curr->next;
+    }
+
+    // If we're hear it means we didn't have enough memory
+    // for the block allocation. So we will allocate more..
+    uintptr_t old_end = end;
+    kheap_map_page();
+
+    struct heap_block_t* block = (struct heap_block_t*)old_end;
+    block->size = PAGE_SIZE - sizeof(struct heap_block_t);
+    block->free = false;
+    block->next = NULL;
+
+    // Put the block at the end of the list
+    curr = head;
+    while (curr->next)
+    {
+        curr = curr->next;
+    }
+    curr->next = block;
+
+    return (void*)((uintptr_t)block + sizeof(struct heap_block_t));
+}
+
+void kfree(void* ptr)
+{
+    // Nothing to free
+    if (!ptr) return;
+
+    // Set it free!
+    struct heap_block_t* block = (struct heap_block_t*)((uintptr_t)ptr - sizeof(struct heap_block_t));
+    block->free = true;
+}

src/mem/heap/kheap.h

@@ -0,0 +1,26 @@
+#ifndef KHEAP_H
+#define KHEAP_H
+
+// We need some kind of simple kernel heap to make our linked list
+// for the VMM, as we need "malloc" and "free" for that data structure.
+// When the kernel heap is ready, we can alloc our VM object linked list
+// and then continue working on the VMM.
+
+// 16MB should be enough for some linked lists
+#define KHEAP_SIZE (16*1024*1024)
+
+#include <stdbool.h>
+#include <stddef.h>
+
+struct heap_block_t
+{
+    size_t size;
+    bool free;
+    struct heap_block_t* next;
+};
+
+void kheap_init();
+void* kmalloc(size_t size);
+void kfree(void* ptr);
+
+#endif

src/mem/misc/utils.c

@@ -1,5 +1,8 @@
 #include <stddef.h>
 #include <stdint.h>
+#include <limine.h>
+#include "kernel.h"
+#include "string/string.h"
 
 // We won't be linked to standard library, but still need the basic mem* functions
 // so everything goes allright with the compiler
@@ -68,3 +71,52 @@ int memcmp(const void* s1, const void* s2, size_t n)
 
 	return 0;
 }
+
+// Display the memmap so we see how the memory is laid out at handoff
+void memmap_display(struct limine_memmap_response* response)
+{
+	DEBUG("Got memory map from Limine: revision %u, %u entries", response->revision, response->entry_count);
+
+	for (size_t i=0; i<response->entry_count; i++)
+	{
+		struct limine_memmap_entry* entry = response->entries[i];
+		char type[32] = {0};
+		switch(entry->type)
+		{
+			case LIMINE_MEMMAP_USABLE:
+				strcpy(type, "USABLE");
+				break;
+			case LIMINE_MEMMAP_RESERVED:
+				strcpy(type, "RESERVED");
+				break;
+			case LIMINE_MEMMAP_ACPI_RECLAIMABLE:
+				strcpy(type, "ACPI_RECLAIMABLE");
+				break;
+			case LIMINE_MEMMAP_ACPI_NVS:
+				strcpy(type, "ACPI_NVS");
+				break;
+			case LIMINE_MEMMAP_BAD_MEMORY:
+				strcpy(type, "BAD_MEMORY");
+				break;
+			case LIMINE_MEMMAP_BOOTLOADER_RECLAIMABLE:
+				strcpy(type, "BOOTLOADER_RECLAIMABLE");
+				break;
+			case LIMINE_MEMMAP_KERNEL_AND_MODULES:
+				strcpy(type, "KERNEL_AND_MODULES");
+				break;
+			case LIMINE_MEMMAP_FRAMEBUFFER:
+				strcpy(type, "FRAMEBUFFER");
+				break;
+			default:
+				strcpy(type, "UNKNOWN");
+				break;
+		}
+		DEBUG("entry %02u: [0x%016x | %016u bytes] - %s", i, entry->base, entry->length, type);
+	}
+}
+
+// Display the HHDM
+void hhdm_display(struct limine_hhdm_response* hhdm)
+{
+	DEBUG("Got HHDM revision=%u offset=0x%p", hhdm->revision, hhdm->offset);
+}

src/mem/misc/utils.h

@@ -8,4 +8,7 @@ void* memset(void* s, int c, size_t n);
 void* memmove(void *dest, const void* src, size_t n);
 int memcmp(const void* s1, const void* s2, size_t n);
 
+void memmap_display(struct limine_memmap_response* response);
+void hhdm_display(struct limine_hhdm_response* hhdm);
+
 #endif

src/mem/paging/paging.c

@@ -0,0 +1,159 @@
+#include "paging.h"
+#include "pmm.h"
+#include <kernel.h>
+#include <stddef.h>
+#include <limine.h>
+
+/*
+Paging on x86 uses four different page table levels:
+cr3 register contains the phys address for the PML4 (root directory)
+
+Each directory/table is made of 512 entries, each one uint64_t
+Each of these entries have special bits (PRESENT/WRITEABLE/USER/etc.)
+that dictates their attributes. Also these bits fall back on children tables.
+
+If we use 1GB huge pages: PML4 -> PDPT -> 1gb pages
+2MB huge pages: PML4 -> PDPT -> PD -> 2mb pages
+4KB (regular size): PML4 -> PDPT -> PD -> PT -> 4kb pages
+*/
+
+static inline void load_cr3(uint64_t value) {
+    asm volatile ("mov %0, %%cr3" :: "r"(value) : "memory");
+}
+
+// To flush TLB
+static inline void invlpg(void *addr)
+{
+    asm volatile("invlpg (%0)" :: "r"(addr) : "memory");
+}
+
+// Allocates a 512-entry 64bit page table/directory/whatever (zeroed)
+static uint64_t* alloc_page_table()
+{
+    uint64_t* virt = (uint64_t*)PHYS_TO_VIRT(pmm_alloc());
+
+    for (size_t i=0; i<512; i++)
+    {
+        virt[i] = 0;
+    }
+    return virt;
+}
+
+// Kernel paging root table, that will be placed in cr3
+__attribute__((aligned(4096)))
+uint64_t *kernel_pml4;
+
+// Map a page, taking virt and phys address. This will go through the paging structures
+// beginning at the given root table, translate the virtual address in indexes in
+// page table/directories, and then mapping the correct page table entry with the
+// given physical address + flags
+void paging_map_page(uint64_t* root_table, uint64_t virt, uint64_t phys, uint64_t flags)
+{
+    virt = PAGE_ALIGN_DOWN(virt);
+    phys = PAGE_ALIGN_DOWN(phys);
+
+    // Translate the virt address into page table indexes
+    uint64_t pml4_i = PML4_INDEX(virt);
+    uint64_t pdpt_i = PDPT_INDEX(virt);
+    uint64_t pd_i = PD_INDEX(virt);
+    uint64_t pt_i = PT_INDEX(virt);
+
+    uint64_t *pdpt, *pd, *pt;
+
+    // PML4
+    // If the entry at index is not present, allocate enough space for it
+    // then populate the entry with correct addr + flags
+    if (!(root_table[pml4_i] & PTE_PRESENT))
+    {
+        pdpt = alloc_page_table();
+        root_table[pml4_i] = VIRT_TO_PHYS(pdpt) | PTE_PRESENT | PTE_WRITABLE;
+    }
+    else {
+        pdpt = (uint64_t *)PHYS_TO_VIRT(root_table[pml4_i] & ~0xFFFULL);
+    }
+
+    // PDPT: same here
+    if (!(pdpt[pdpt_i] & PTE_PRESENT))
+    {
+        pd = alloc_page_table();
+        pdpt[pdpt_i] = VIRT_TO_PHYS(pd) | PTE_PRESENT | PTE_WRITABLE;
+    } 
+    else {
+        pd = (uint64_t *)PHYS_TO_VIRT(pdpt[pdpt_i] & ~0xFFFULL);
+    }
+
+    // PD: and here
+    if (!(pd[pd_i] & PTE_PRESENT))
+    {
+        pt = alloc_page_table();
+        pd[pd_i] = VIRT_TO_PHYS(pt) | PTE_PRESENT | PTE_WRITABLE;
+    }
+    else {
+        pt = (uint64_t *)PHYS_TO_VIRT(pd[pd_i] & ~0xFFFULL);
+    }
+
+    // PT: finally, populate the page table entry
+    pt[pt_i] = phys | flags | PTE_PRESENT;
+
+    // Flush TLB (apply changes)
+    invlpg((void *)virt);
+}
+
+uint64_t kernel_phys_base;
+uint64_t kernel_virt_base;
+
+void paging_init(struct limine_kernel_address_response* kaddr, struct limine_framebuffer* fb)
+{
+    // We should map the kernel, GDT, IDT, stack, framebuffer.
+    // Optionally we could map ACPI tables (we can find them in the Limine memmap)
+
+    kernel_phys_base = kaddr->physical_base;
+    kernel_virt_base = kaddr->virtual_base;
+
+    DEBUG("Kernel lives at virt=0x%p phys=0x%p", kernel_virt_base, kernel_phys_base);
+
+    kernel_pml4 = alloc_page_table();
+
+    // for debug
+    uint64_t page_count = 0;
+
+    // HHDM map first 1 GB using given offset
+    for (uint64_t i=0; i<0x40000000; i += PAGE_SIZE)
+    {
+        //paging_kmap_page(i+hhdm_off, i, PTE_WRITABLE);
+        paging_map_page(kernel_pml4, i+hhdm_off, i, PTE_WRITABLE);
+        page_count++;
+    }
+    DEBUG("Mapped %u pages for first 1GB (HHDM)", page_count); page_count = 0;
+
+    // Map the kernel (according to virt/phys_base given by Limine)
+    // SOME DAY when we want a safer kernel we should map .text as Read/Exec
+    // .rodata as Read and .data as Read/Write
+    // For now who gives a shit, let's RWX all kernel
+    for (uint64_t i = 0; i < KERNEL_SIZE; i += PAGE_SIZE)
+    {
+        //paging_kmap_page(kernel_virt_base+i, kernel_phys_base+i, PTE_WRITABLE);
+        paging_map_page(kernel_pml4, kernel_virt_base+i, kernel_phys_base+i, PTE_WRITABLE);
+        page_count++;
+    }
+    DEBUG("Mapped %u pages for kernel", page_count); page_count = 0;
+
+    // Get the framebuffer phys/virt address, and size
+    uint64_t fb_virt = (uint64_t)fb->address;
+    uint64_t fb_phys = VIRT_TO_PHYS(fb_virt);
+    uint64_t fb_size = fb->pitch * fb->height;
+    uint64_t fb_pages = (fb_size + PAGE_SIZE-1)/PAGE_SIZE;
+
+    // Map the framebuffer (with cache-disable & write-through)
+    for (uint64_t i=0; i<fb_pages; i++)
+    {
+        //paging_kmap_page(fb_virt+i*PAGE_SIZE, fb_phys+i*PAGE_SIZE, PTE_WRITABLE | PTE_PCD | PTE_PWT);
+        paging_map_page(kernel_pml4, fb_virt+i*PAGE_SIZE, fb_phys+i*PAGE_SIZE, PTE_WRITABLE | PTE_PCD | PTE_PWT);
+        page_count++;
+    }
+    DEBUG("Mapped %u pages for framebuffer", page_count);
+
+    // Finally, we load the physical address of our PML4 (root table) into cr3
+    load_cr3(VIRT_TO_PHYS(kernel_pml4));
+    DEBUG("cr3 loaded, we're still alive");
+}

src/mem/paging/paging.h

@@ -0,0 +1,44 @@
+#ifndef PAGING_H
+#define PAGING_H
+
+#define PAGE_SIZE 4096
+#define BITS_PER_ROW 64
+
+#include <stdint.h>
+#include <limine.h>
+
+void paging_init(struct limine_kernel_address_response* kaddr, struct limine_framebuffer* fb);
+void paging_map_page(uint64_t* root_table, uint64_t virt, uint64_t phys, uint64_t flags);
+
+extern uint64_t hhdm_off;
+
+#define PHYS_TO_VIRT(x) ((void*)((uintptr_t)(x) + hhdm_off))
+#define VIRT_TO_PHYS(x) ((uintptr_t)(x) - hhdm_off)
+
+// Stole it
+#define ALIGN_UP(x, align)   (((x) + ((align) - 1)) & ~((align) - 1))
+#define ALIGN_DOWN(x, align) ((x) & ~((align) - 1))
+#define PAGE_ALIGN_DOWN(x) ((x) & ~0xFFFULL)
+
+#define PML4_INDEX(x) (((x) >> 39) & 0x1FF)
+#define PDPT_INDEX(x) (((x) >> 30) & 0x1FF)
+#define PD_INDEX(x)   (((x) >> 21) & 0x1FF)
+#define PT_INDEX(x)   (((x) >> 12) & 0x1FF)
+
+// Page entry special bits
+// Bits set on a parent (directory, table) fall back to their children
+#define PTE_PRESENT     (1ULL << 0)
+#define PTE_WRITABLE    (1ULL << 1)
+#define PTE_USER        (1ULL << 2)
+#define PTE_PWT         (1ULL << 3)
+#define PTE_PCD         (1ULL << 4)
+#define PTE_HUGE        (1ULL << 7)
+#define PTE_NOEXEC      (1ULL << 63)
+
+// Specified in linker.ld
+#define KERNEL_BASE 0xFFFFFFFF80000000ULL
+
+// 2 MB should be enough (as of now, the whole kernel ELF is around 75kb)
+#define KERNEL_SIZE 0x200000
+
+#endif

src/mem/paging/pmm.c

@@ -0,0 +1,103 @@
+// OMG here we are. I'm cooked.
+
+/*
+pmm - Physical Memory Manager
+will manage 4kb pages physically
+it will probably need to get some info from Limine,
+to see which pages are used by kernel/bootloader/mmio/fb etc.
+*/
+
+#include "paging.h"
+#include <limine.h>
+#include <stddef.h>
+#include <stdint.h>
+#include <kernel.h>
+#include "mem/misc/utils.h"
+#include "pmm.h"
+
+/*
+First we'll have to discover the physical memory layout,
+and for that we can use a Limine request.
+*/
+
+/*
+We will look for the biggest usable physical memory region
+and use this for the bitmap. The reserved memory will be ignored.
+*/
+
+struct limine_memmap_entry* biggest_entry;
+
+static void pmm_find_biggest_usable_region(struct limine_memmap_response* memmap, struct limine_hhdm_response* hhdm)
+{
+    // Max length of a usable memory region
+    uint64_t length_max = 0;
+    uint64_t offset = hhdm->offset;
+
+    DEBUG("Usable Memory:");
+    for (size_t i=0; i<memmap->entry_count; i++)
+    {
+        struct limine_memmap_entry* entry = memmap->entries[i];
+
+        if (entry->type == LIMINE_MEMMAP_USABLE)
+        {
+            DEBUG("0x%p-0x%p mapped at 0x%p-0x%p", entry->base, entry->base+entry->length,
+                        entry->base+offset, entry->base+entry->length+offset);
+            if (entry->length > length_max)
+            {
+                length_max = entry->length;
+                biggest_entry = entry;
+            }
+        }
+    }
+
+    DEBUG("Biggest usable memory region:");
+    DEBUG("0x%p-0x%p mapped at 0x%p-0x%p", biggest_entry->base, biggest_entry->base + biggest_entry->length,
+            biggest_entry->base+offset, biggest_entry->base+biggest_entry->length+offset);
+}
+
+// Offset from Higher Half Direct Map
+uint64_t hhdm_off;
+
+static uintptr_t g_freelist = 0;
+
+uintptr_t pmm_alloc()
+{
+    if (!g_freelist) return 0;
+    uintptr_t addr = g_freelist;
+    g_freelist = *(uintptr_t*) PHYS_TO_VIRT(g_freelist);
+    return addr;
+}
+
+void pmm_free(uintptr_t addr)
+{
+    *(uintptr_t*) PHYS_TO_VIRT(addr) = g_freelist;
+    g_freelist = addr;
+}
+
+static void pmm_init_freelist()
+{
+    // We simply call pmm_free() on each page that is marked USABLE
+    // in our big memory region.
+    uint64_t base = ALIGN_UP(biggest_entry->base, PAGE_SIZE);
+    uint64_t end = ALIGN_DOWN(biggest_entry->base + biggest_entry->length, PAGE_SIZE);
+
+    uint64_t page_count=0;
+    for (uint64_t addr = base; addr < end; addr += PAGE_SIZE)
+    {
+        pmm_free(addr);
+        //DEBUG("page %u lives at phys 0x%p (virt 0x%p)", page_count, addr, PHYS_TO_VIRT(addr));
+        page_count++;
+    }
+    DEBUG("%u frames in freelist, available for use (%u bytes)", page_count, page_count*PAGE_SIZE);
+}
+
+void pmm_init(struct limine_memmap_response* memmap, struct limine_hhdm_response* hhdm)
+{
+    hhdm_off = hhdm->offset;
+    pmm_find_biggest_usable_region(memmap, hhdm);
+    //pmm_allocate_bitmap(hhdm); too complicated for my small brain
+    
+    // Now we have biggest USABLE region,
+    // so to populate the free list we just iterate through it
+    pmm_init_freelist();
+}

src/mem/paging/pmm.h

@@ -0,0 +1,10 @@
+#ifndef PAGING_PMM_H
+#define PAGING_PMM_H
+
+#include <limine.h>
+
+void pmm_init(struct limine_memmap_response* memmap, struct limine_hhdm_response* hhdm);
+void pmm_free(uintptr_t addr);
+uintptr_t pmm_alloc();
+
+#endif

src/mem/paging/vmm.c

@@ -0,0 +1,66 @@
+/*
+The VMM (virtual memory manager) will have two roles:
+- mapping pages
+- unmapping pages
+in a specified virtual space
+
+compared to the PMM which allocs/frees 4kb frames ("physical pages").
+*/
+
+#include "vmm.h"
+#include "paging.h"
+#include <stddef.h>
+#include "pmm.h"
+#include <kernel.h>
+
+void* vmm_pt_root = 0;
+
+// Linked list head for virtual memory objects
+struct vm_object* vm_objs = NULL;
+
+
+uint64_t convert_x86_vm_flags(size_t flags)
+{
+    uint64_t value = 0;
+    if (flags & VM_FLAG_WRITE)
+    {
+        value |= PTE_WRITABLE;
+    }
+    if (flags & VM_FLAG_USER)
+    {
+        value |= PTE_USER;
+    }
+    if ((flags & VM_FLAG_EXEC) == 0)
+    {
+        value |= PTE_NOEXEC;
+    }
+    return value;
+}
+
+extern uint64_t *kernel_pml4;
+
+void vmm_setup_pt_root()
+{
+    // We alloc a physical page (frame) for the pointer, then map it
+    // to virt (pointer)
+    uintptr_t phys = pmm_alloc();
+    vmm_pt_root = (void*)kernel_pml4;
+    paging_map_page(kernel_pml4, (uint64_t)vmm_pt_root, phys, convert_x86_vm_flags(VM_FLAG_WRITE | VM_FLAG_EXEC));
+    DEBUG("VMM setup: vmm_pt_root=0x%p (phys=0x%p)", vmm_pt_root, phys);
+}
+
+/* void* vmm_alloc(size_t length, size_t flags)
+{
+    // We will try to allocate at least length bytes, which have to be rounded UP to
+    // the next page so its coherent with the PMM
+    size_t len = ALIGN_UP(length, PAGE_SIZE);
+
+    // Need to implement this (as linked list)
+    // but for now kernel heap is sufficient
+    // The VMM will prob be more useful when we have userspace
+} */
+
+void vmm_init()
+{
+    vmm_setup_pt_root();
+}

src/mem/paging/vmm.h

@@ -0,0 +1,29 @@
+#ifndef VMM_H
+#define VMM_H
+
+#include <stdint.h>
+#include <stddef.h>
+
+/*
+This will be our linked list of virtual memory objects.
+Flags here aren't x86 flags, they are platform-agnostic
+kernel-defined flags.
+*/
+
+struct vm_object
+{
+    uintptr_t base;
+    size_t length;
+    size_t flags;
+    struct vm_object* next;
+};
+
+// Flags bitfield
+#define VM_FLAG_NONE 0
+#define VM_FLAG_WRITE (1 << 0)
+#define VM_FLAG_EXEC (1 << 1)
+#define VM_FLAG_USER (1 << 2)
+
+void vmm_init();
+
+#endif

src/string/string.c

@@ -0,0 +1,6 @@
+char* strcpy(char *dest, const char *src)
+{
+    char *temp = dest;
+    while((*dest++ = *src++));
+    return temp;
+}

src/string/string.h

@@ -0,0 +1,6 @@
+#ifndef STRING_H
+#define STRING_H
+
+char *strcpy(char *dest, const char *src);
+
+#endif

src/time/timer.c

@@ -1,5 +1,5 @@
 #include <stdint.h>
-#include "../io/serial/serial.h"
+#include "io/serial/serial.h"
 #include <kernel.h>
 
 /*
@@ -65,6 +65,17 @@ void pit_init()
     outb(0x40, (divisor >> 8) & 0xFF);
 }
 
+// Wait n ticks
+// Given that there's a tick every 1ms, wait n milliseconds
+void timer_wait(uint64_t wait_ticks)
+{
+    uint64_t then = ticks + wait_ticks;
+    while (ticks < then)
+    {
+        asm("hlt");
+    };
+}
+
 void timer_init()
 {
     // Remapping the PIC, because at startup it conflicts with

src/time/timer.h

@@ -2,5 +2,6 @@
 #define TIMER_H
 
 void timer_init();
+void timer_wait(unsigned int wait_ticks);
 
 #endif