Meta: Move kernel and userland to their own folders.

This is to allow both the kernel and the userland to share certain
header files and to make the folder structure a bit more clear.

2 files changed, 693 insertions, 0 deletions

diff --git a/kernel/arch/i386/boot.s b/kernel/arch/i386/boot.s
new file mode 100644
index 0000000..61bbfbb
--- /dev/null
+++ b/kernel/arch/i386/boot.s
@@ -0,0 +1,128 @@
+.set ALIGN,    1<<0
+.set MEMINFO,  1<<1
+.set VIDEO_MODE,  1<<2
+.set FLAGS,    VIDEO_MODE|ALIGN | MEMINFO
+.set MAGIC,    0x1BADB002
+.set CHECKSUM, -(MAGIC + FLAGS)
+ 
+#.section .multiboot
+.align 16, 0
+.section .multiboot.data, "aw"
+	.align 4
+	.long MAGIC
+	.long FLAGS
+	.long CHECKSUM
+	# unused(padding)
+	.long 0,0,0,0,0
+	# Video mode
+	.long   0       # Linear graphics
+	.long   0       # Preferred width
+	.long   0       # Preferred height
+	.long   32      # Preferred pixel depth
+
+# Allocate the initial stack.
+.section .bootstrap_stack, "aw", @nobits
+stack_bottom:
+.skip 16384 # 16 KiB
+stack_top:
+
+.global boot_page_directory
+.global boot_page_table1
+.section .bss, "aw", @nobits
+	.align 4096
+boot_page_directory:
+	.skip 4096
+boot_page_table1:
+	.skip 4096
+
+.section .multiboot.text, "a"
+.global _start
+.extern _kernel_end
+.type _start, @function
+# Kernel entry
+_start:
+    # edi contains the buffer we wish to modify
+	movl $(boot_page_table1 - 0xC0000000), %edi
+
+    # for loop
+	movl $0, %esi
+	movl $1024, %ecx
+1:
+    # esi = address
+    # If we are in kernel range jump to "label 2"
+	cmpl $_kernel_start, %esi
+	jl 2f
+
+    # If we are past the kernel jump to the final stage
+    # "label 3"
+	cmpl $(_kernel_end - 0xC0000000), %esi
+	jge 3f
+    
+2:
+    # Add permission to the pages
+	movl %esi, %edx
+	orl $0x003, %edx
+	movl %edx, (%edi)
+
+	# Size of page is 4096 bytes.
+	addl $4096, %esi
+	# Size of entries in boot_page_table1 is 4 bytes.
+	addl $4, %edi
+	# Loop to the next entry if we haven't finished.
+	loop 1b
+
+3:
+	# Map the page table to both virtual addresses 0x00000000 and 0xC0000000.
+	movl $(boot_page_table1 - 0xC0000000 + 0x003), boot_page_directory - 0xC0000000 + 0
+	movl $(boot_page_table1 - 0xC0000000 + 0x003), boot_page_directory - 0xC0000000 + 768 * 4
+
+	# Set cr3 to the address of the boot_page_directory.
+	movl $(boot_page_directory - 0xC0000000), %ecx
+	movl %ecx, %cr3
+
+	# Enable paging
+	movl %cr0, %ecx
+	orl $0x80000000, %ecx
+	movl %ecx, %cr0
+
+	# Jump to higher half with an absolute jump. 
+	lea 4f, %ecx
+	jmp *%ecx
+
+.size _start, . - _start
+
+.section .text
+
+4:
+	# At this point, paging is fully set up and enabled.
+
+	# Unmap the identity mapping as it is now unnecessary. 
+	movl $0, boot_page_directory + 0
+
+	# Reload crc3 to force a TLB flush so the changes to take effect.
+	movl %cr3, %ecx
+	movl %ecx, %cr3
+
+	# Set up the stack.
+	mov $stack_top, %esp
+
+        # Reset EFLAGS.
+        pushl   $0
+        popf
+
+	pushl %esp
+
+        /* Push the pointer to the Multiboot information structure. */
+        pushl   %ebx
+        /* Push the magic value. */
+        pushl   %eax
+
+	mov $_kernel_end, %eax
+        pushl   %eax
+	# Enter the high-level kernel.
+	call kernel_main
+
+	# Infinite loop if the system has nothing more to do.
+	cli
+1:	hlt
+	jmp 1b
diff --git a/kernel/arch/i386/mmu.c b/kernel/arch/i386/mmu.c
new file mode 100644
index 0000000..ccfe894
--- /dev/null
+++ b/kernel/arch/i386/mmu.c
@@ -0,0 +1,565 @@
+#include <assert.h>
+#include <ksbrk.h>
+#include <log.h>
+#include <mmu.h>
+
+#define INDEX_FROM_BIT(a) (a / (32))
+#define OFFSET_FROM_BIT(a) (a % (32))
+#define PAGE_SIZE ((uintptr_t)0x1000)
+
+#define PAGE_ALLOCATE 1
+#define PAGE_NO_ALLOCATE 0
+
+PageDirectory *kernel_directory;
+PageDirectory real_kernel_directory;
+PageDirectory *active_directory = 0;
+
+#define END_OF_MEMORY 0x8000000 * 15
+const uint32_t num_of_frames = END_OF_MEMORY / PAGE_SIZE;
+uint32_t frames[END_OF_MEMORY / PAGE_SIZE] = {0};
+uint32_t num_allocated_frames = 0;
+
+#define KERNEL_START 0xc0000000
+extern uintptr_t data_end;
+
+void write_to_frame(uint32_t frame_address, uint8_t on);
+
+void *ksbrk(size_t s) {
+  uintptr_t rc = (uintptr_t)align_page((void *)data_end);
+  data_end += s;
+  data_end = (uintptr_t)align_page((void *)data_end);
+
+  if (!get_active_pagedirectory()) {
+    // If there is no active pagedirectory we
+    // just assume that the memory is
+    // already mapped.
+    return (void *)rc;
+  }
+  // Determine whether we are approaching a unallocated table
+  int table_index = 1 + (rc / (1024 * 0x1000));
+  if (!active_directory->tables[table_index]) {
+    uint32_t physical;
+    active_directory->tables[table_index] = (PageTable *)0xDEADBEEF;
+    active_directory->tables[table_index] =
+        (PageTable *)ksbrk_physical(sizeof(PageTable), (void **)&physical);
+    memset(active_directory->tables[table_index], 0, sizeof(PageTable));
+    active_directory->physical_tables[table_index] = (uint32_t)physical | 0x3;
+
+    kernel_directory->tables[table_index] =
+        active_directory->tables[table_index];
+    kernel_directory->physical_tables[table_index] =
+        active_directory->physical_tables[table_index];
+    return ksbrk(s);
+  }
+  mmu_allocate_shared_kernel_region((void *)rc, (data_end - (uintptr_t)rc));
+  assert(((uintptr_t)rc % PAGE_SIZE) == 0);
+  memset((void *)rc, 0x00, s);
+
+  return (void *)rc;
+}
+
+void *ksbrk_physical(size_t s, void **physical) {
+  void *r = ksbrk(s);
+  if (physical) {
+    *physical = (void *)virtual_to_physical(r, 0);
+  }
+  return r;
+}
+
+uint32_t mmu_get_number_of_allocated_frames(void) {
+  return num_allocated_frames;
+}
+
+Page *get_page(void *ptr, PageDirectory *directory, int create_new_page,
+               int set_user) {
+  uintptr_t address = (uintptr_t)ptr;
+  if (!directory)
+    directory = get_active_pagedirectory();
+  address /= 0x1000;
+
+  uint32_t table_index = address / 1024;
+  if (!directory->tables[table_index]) {
+    if (!create_new_page)
+      return 0;
+
+    uint32_t physical;
+    directory->tables[table_index] =
+        (PageTable *)ksbrk_physical(sizeof(PageTable), (void **)&physical);
+    memset(directory->tables[table_index], 0, sizeof(PageTable));
+    directory->physical_tables[table_index] =
+        (uint32_t)physical | ((set_user) ? 0x7 : 0x3);
+
+    if (!set_user) {
+      kernel_directory->tables[table_index] = directory->tables[table_index];
+      kernel_directory->physical_tables[table_index] =
+          directory->physical_tables[table_index];
+    }
+  }
+  Page *p = &directory->tables[table_index]->pages[address % 1024];
+  if (create_new_page) {
+    p->present = 0;
+  }
+  return &directory->tables[table_index]->pages[address % 1024];
+}
+
+void mmu_free_pages(void *a, uint32_t n) {
+  for (; n > 0; n--) {
+    Page *p = get_page(a, NULL, PAGE_NO_ALLOCATE, 0);
+    p->present = 0;
+    write_to_frame(p->frame * 0x1000, 0);
+    a += 0x1000;
+  }
+}
+
+void *next_page(void *ptr) {
+  uintptr_t a = (uintptr_t)ptr;
+  return (void *)(a + (PAGE_SIZE - ((uint32_t)a & (PAGE_SIZE - 1))));
+}
+
+void *align_page(void *a) {
+  if ((uintptr_t)a & (PAGE_SIZE - 1))
+    return next_page(a);
+
+  return a;
+}
+
+void flush_tlb(void) {
+  asm volatile("\
+	mov %cr3, %eax;\
+	mov %eax, %cr3");
+}
+
+uint32_t first_free_frame(void) {
+  for (uint32_t i = 1; i < INDEX_FROM_BIT(num_of_frames); i++) {
+    if (frames[i] == 0xFFFFFFFF)
+      continue;
+
+    for (uint32_t c = 0; c < 32; c++)
+      if (!(frames[i] & ((uint32_t)1 << c)))
+        return i * 32 + c;
+  }
+
+  kprintf("ERROR Num frames: %x\n", mmu_get_number_of_allocated_frames());
+  klog("No free frames, uh oh.", LOG_ERROR);
+  assert(0);
+  return 0;
+}
+
+void write_to_frame(uint32_t frame_address, uint8_t on) {
+  uint32_t frame = frame_address / 0x1000;
+  assert(INDEX_FROM_BIT(frame) < sizeof(frames) / sizeof(frames[0]));
+  if (on) {
+    num_allocated_frames++;
+    frames[INDEX_FROM_BIT(frame)] |= ((uint32_t)0x1 << OFFSET_FROM_BIT(frame));
+    return;
+  }
+  num_allocated_frames--;
+  frames[INDEX_FROM_BIT(frame)] &= ~((uint32_t)0x1 << OFFSET_FROM_BIT(frame));
+}
+
+PageDirectory *get_active_pagedirectory(void) { return active_directory; }
+
+PageTable *clone_table(uint32_t src_index, PageDirectory *src_directory,
+                       uint32_t *physical_address) {
+  PageTable *new_table =
+      ksbrk_physical(sizeof(PageTable), (void **)physical_address);
+  PageTable *src = src_directory->tables[src_index];
+
+  // Copy all the pages
+  for (uint16_t i = 0; i < 1024; i++) {
+    if (!src->pages[i].present) {
+      new_table->pages[i].present = 0;
+      continue;
+    }
+    uint32_t frame_address = first_free_frame();
+    write_to_frame(frame_address * 0x1000, 1);
+    new_table->pages[i].frame = frame_address;
+
+    new_table->pages[i].present |= src->pages[i].present;
+    new_table->pages[i].rw |= src->pages[i].rw;
+    new_table->pages[i].user |= src->pages[i].user;
+    new_table->pages[i].accessed |= src->pages[i].accessed;
+    new_table->pages[i].dirty |= src->pages[i].dirty;
+  }
+
+  // Now copy all of the data to the new table. This is done by creating a
+  // virutal pointer to this newly created tables physical frame so we can
+  // copy data to it.
+  for (uint32_t i = 0; i < 1024; i++) {
+    // Find a unused table
+    if (src_directory->tables[i])
+      continue;
+
+    // Link the table to the new table temporarily
+    src_directory->tables[i] = new_table;
+    src_directory->physical_tables[i] = *physical_address | 0x7;
+    PageDirectory *tmp = get_active_pagedirectory();
+    switch_page_directory(src_directory);
+
+    // For each page in the table copy all the data over.
+    for (uint32_t c = 0; c < 1024; c++) {
+      // Only copy pages that are used.
+      if (!src->pages[c].frame || !src->pages[c].present)
+        continue;
+
+      uint32_t table_data_pointer = i << 22 | c << 12;
+      uint32_t src_data_pointer = src_index << 22 | c << 12;
+      memcpy((void *)table_data_pointer, (void *)src_data_pointer, 0x1000);
+    }
+    src_directory->tables[i] = 0;
+    src_directory->physical_tables[i] = 0;
+    switch_page_directory(tmp);
+    return new_table;
+  }
+  ASSERT_NOT_REACHED;
+  return 0;
+}
+
+PageTable *copy_table(PageTable *src, uint32_t *physical_address) {
+  PageTable *new_table =
+      ksbrk_physical(sizeof(PageTable), (void **)physical_address);
+
+  // copy all the pages
+  for (uint16_t i = 0; i < 1024; i++) {
+    if (!src->pages[i].present) {
+      new_table->pages[i].present = 0;
+      continue;
+    }
+    new_table->pages[i].frame = src->pages[i].frame;
+    new_table->pages[i].present = src->pages[i].present;
+    new_table->pages[i].rw = src->pages[i].rw;
+    new_table->pages[i].user = src->pages[i].user;
+    new_table->pages[i].accessed = src->pages[i].accessed;
+    new_table->pages[i].dirty = src->pages[i].dirty;
+  }
+  return new_table;
+}
+
+PageDirectory *clone_directory(PageDirectory *original) {
+  if (!original)
+    original = get_active_pagedirectory();
+
+  uint32_t physical_address;
+  PageDirectory *new_directory =
+      ksbrk_physical(sizeof(PageDirectory), (void **)&physical_address);
+  uint32_t offset =
+      (uint32_t)new_directory->physical_tables - (uint32_t)new_directory;
+  new_directory->physical_address = physical_address + offset;
+
+  for (int i = 0; i < 1024; i++) {
+    if (!original->tables[i]) {
+      new_directory->tables[i] = NULL;
+      new_directory->physical_tables[i] = (uint32_t)NULL;
+      continue;
+    }
+
+    // Make sure to copy instead of cloning the stack.
+    if (i >= 635 && i <= 641) {
+      uint32_t physical;
+      new_directory->tables[i] = clone_table(i, original, &physical);
+      new_directory->physical_tables[i] =
+          physical | (original->physical_tables[i] & 0xFFF);
+      continue;
+    }
+
+    //    if (original->tables[i] == kernel_directory->tables[i]) {
+    if (i > 641) {
+      new_directory->tables[i] = kernel_directory->tables[i];
+      new_directory->physical_tables[i] = kernel_directory->physical_tables[i];
+      continue;
+    }
+
+    uint32_t physical;
+    new_directory->tables[i] = clone_table(i, original, &physical);
+    new_directory->physical_tables[i] =
+        physical | (original->physical_tables[i] & 0xFFF);
+  }
+
+  return new_directory;
+}
+
+void mmu_allocate_shared_kernel_region(void *rc, size_t n) {
+  size_t num_pages = n / PAGE_SIZE;
+  for (size_t i = 0; i <= num_pages; i++) {
+    Page *p = get_page((void *)(rc + i * 0x1000), NULL, PAGE_NO_ALLOCATE, 0);
+    if (!p) {
+      kprintf("don't have: %x\n", rc + i * 0x1000);
+      p = get_page((void *)(rc + i * 0x1000), NULL, PAGE_ALLOCATE, 0);
+    }
+    if (!p->present || !p->frame)
+      allocate_frame(p, 0, 1);
+  }
+}
+
+void mmu_remove_virtual_physical_address_mapping(void *ptr, size_t length) {
+  size_t num_pages = (uintptr_t)align_page((void *)length) / PAGE_SIZE;
+  for (size_t i = 0; i < num_pages; i++) {
+    Page *p = get_page(ptr + (i * PAGE_SIZE), NULL, PAGE_NO_ALLOCATE, 0);
+    if (!p)
+      return;
+    p->frame = 0;
+    p->present = 0;
+  }
+}
+
+void *mmu_find_unallocated_virtual_range(void *addr, size_t length) {
+  addr = align_page(addr);
+  // Check if the pages already exist
+  for (size_t i = 0; i < length; i += 0x1000) {
+    if (get_page(addr + i, NULL, PAGE_NO_ALLOCATE, 0)) {
+      // Page already exists
+      return mmu_find_unallocated_virtual_range(addr + 0x1000, length);
+    }
+  }
+  return addr;
+}
+
+int mmu_allocate_region(void *ptr, size_t n, mmu_flags flags,
+                        PageDirectory *pd) {
+  pd = (pd) ? pd : get_active_pagedirectory();
+  size_t num_pages = n / 0x1000;
+  for (size_t i = 0; i <= num_pages; i++) {
+    Page *p = get_page((void *)(ptr + i * 0x1000), pd, PAGE_ALLOCATE, 1);
+    assert(p);
+    int rw = (flags & MMU_FLAG_RW);
+    int kernel = (flags & MMU_FLAG_KERNEL);
+    if (!allocate_frame(p, rw, kernel)) {
+      klog("MMU: Frame allocation failed", LOG_WARN);
+      return 0;
+    }
+  }
+  return 1;
+}
+
+void *mmu_map_frames(void *const ptr, size_t s) {
+  void *const r = mmu_find_unallocated_virtual_range((void *)0xEF000000, s);
+  kprintf("r: %x\n", r);
+  size_t num_pages = s / 0x1000;
+  for (size_t i = 0; i <= num_pages; i++) {
+    Page *p = get_page((void *)(r + i * 0x1000), NULL, PAGE_ALLOCATE, 1);
+    assert(p);
+    int rw = 1;
+    int is_kernel = 1;
+    p->present = 1;
+    p->rw = rw;
+    p->user = !is_kernel;
+    p->frame = (uintptr_t)(ptr + i * 0x1000) / 0x1000;
+    write_to_frame((uintptr_t)ptr + i * 0x1000, 1);
+  }
+  flush_tlb();
+  return r;
+}
+
+void *allocate_frame(Page *page, int rw, int is_kernel) {
+  if (page->present) {
+    dump_backtrace(5);
+    klog("Page is already set", 1);
+    for (;;)
+      ;
+    return 0;
+  }
+  uint32_t frame_address = first_free_frame();
+  assert(frame_address < sizeof(frames) / sizeof(frames[0]));
+  write_to_frame(frame_address * 0x1000, 1);
+
+  page->present = 1;
+  page->rw = rw;
+  page->user = !is_kernel;
+  page->frame = frame_address;
+  return (void *)(frame_address * 0x1000);
+}
+
+void mmu_free_address_range(void *ptr, size_t length) {
+  size_t num_pages = (size_t)align_page((void *)length) / PAGE_SIZE;
+  for (size_t i = 0; i < num_pages; i++, ptr += PAGE_SIZE) {
+    Page *page = get_page(ptr, NULL, PAGE_NO_ALLOCATE, 0);
+    if (!page)
+      continue;
+    if (!page->present)
+      continue;
+    if (!page->frame)
+      continue;
+    // Sanity check that none of the frames are invalid
+    assert(page->frame < sizeof(frames) / sizeof(frames[0]));
+    write_to_frame(((uint32_t)page->frame) * 0x1000, 0);
+    page->present = 0;
+    page->rw = 0;
+    page->user = 0;
+    page->frame = 0;
+  }
+}
+
+void mmu_map_directories(void *dst, PageDirectory *d, void *src,
+                         PageDirectory *s, size_t length) {
+  d = (!d) ? get_active_pagedirectory() : d;
+  s = (!s) ? get_active_pagedirectory() : s;
+  size_t num_pages = (uint32_t)align_page((void *)length) / 0x1000;
+  for (size_t i = 0; i < num_pages; i++, dst += 0x1000, src += 0x1000) {
+    Page *p = get_page(dst, d, PAGE_ALLOCATE, 1);
+    p->present = 1;
+    p->rw = 1;
+    p->user = 1;
+    void *physical = virtual_to_physical(src, s);
+    p->frame = (uint32_t)physical / PAGE_SIZE;
+  }
+  flush_tlb();
+}
+
+void mmu_map_physical(void *dst, PageDirectory *d, void *physical,
+                      size_t length) {
+  d = (!d) ? get_active_pagedirectory() : d;
+  size_t num_pages = (uint32_t)align_page((void *)length) / 0x1000;
+  for (size_t i = 0; i < num_pages; i++, dst += 0x1000, physical += 0x1000) {
+    Page *p = get_page(dst, d, PAGE_ALLOCATE, 1);
+    p->present = 1;
+    p->rw = 1;
+    p->user = 1;
+    p->frame = (uintptr_t)physical / PAGE_SIZE;
+    write_to_frame((uintptr_t)physical, 1);
+  }
+  flush_tlb();
+}
+
+void *virtual_to_physical(void *address, PageDirectory *directory) {
+  if (0 == directory)
+    directory = get_active_pagedirectory();
+  return (void *)((get_page((void *)address, directory, PAGE_NO_ALLOCATE, 0)
+                       ->frame *
+                   0x1000) +
+                  (((uintptr_t)address) & 0xFFF));
+}
+
+extern uint32_t inital_esp;
+void __attribute__((optimize("O0")))
+move_stack(uint32_t new_stack_address, uint32_t size) {
+  mmu_allocate_region((void *)(new_stack_address - size), size,
+                      MMU_FLAG_KERNEL, NULL);
+
+  uint32_t old_stack_pointer, old_base_pointer;
+
+  register uint32_t eax asm("eax");
+  asm volatile("mov %esp, %eax");
+  old_stack_pointer = eax;
+  asm volatile("mov %ebp, %eax");
+  old_base_pointer = eax;
+
+  uint32_t new_stack_pointer =
+      old_stack_pointer + ((uint32_t)new_stack_address - inital_esp);
+  uint32_t new_base_pointer =
+      old_base_pointer + ((uint32_t)new_stack_address - inital_esp);
+
+  // Copy the stack
+  memcpy((void *)new_stack_pointer, (void *)old_stack_pointer,
+         inital_esp - old_stack_pointer);
+  for (uint32_t i = (uint32_t)new_stack_address; i > new_stack_address - size;
+       i -= 4) {
+    uint32_t tmp = *(uint32_t *)i;
+    if (old_stack_pointer < tmp && tmp < inital_esp) {
+      tmp = tmp + (new_stack_address - inital_esp);
+      uint32_t *tmp2 = (uint32_t *)i;
+      *tmp2 = tmp;
+    }
+  }
+
+  inital_esp = new_stack_pointer;
+  // Actually change the stack
+  eax = new_stack_pointer;
+  asm volatile("mov %eax, %esp");
+  eax = new_base_pointer;
+  asm volatile("mov %eax, %ebp");
+}
+
+// C strings have a unknown length so it does not makes sense to check
+// for a size on the pointer. Instead we check whether the page it
+// resides in is accessible to the user.
+void *is_valid_user_c_string(const char *ptr, size_t *size) {
+  void *r = (void *)ptr;
+  size_t s = 0;
+  for (;;) {
+    void *page = (void *)((uintptr_t)ptr & (uintptr_t)(~(PAGE_SIZE - 1)));
+    if (!is_valid_userpointer(page, PAGE_SIZE))
+      return NULL;
+    for (; (uintptr_t)ptr & (PAGE_SIZE - 1); ptr++, s++)
+      if (!*ptr) {
+        if (size)
+          *size = s;
+        return r;
+      }
+  }
+}
+
+void *is_valid_userpointer(const void *ptr, size_t s) {
+  uintptr_t t = (uintptr_t)ptr;
+  size_t num_pages = (uintptr_t)align_page((void *)s) / 0x1000;
+  for (size_t i = 0; i < num_pages; i++, t += 0x1000) {
+    Page *page = get_page((void *)t, NULL, PAGE_NO_ALLOCATE, 0);
+    if (!page)
+      return NULL;
+    if (!page->present)
+      return NULL;
+    if (!page->user)
+      return NULL;
+  }
+  return (void *)ptr;
+}
+
+void switch_page_directory(PageDirectory *directory) {
+  active_directory = directory;
+  asm("mov %0, %%cr3" ::"r"(directory->physical_address));
+}
+
+void enable_paging(void) {
+  asm("mov %cr0, %eax\n"
+      "or 0b10000000000000000000000000000000, %eax\n"
+      "mov %eax, %cr0\n");
+}
+
+extern uint32_t boot_page_directory;
+extern uint32_t boot_page_table1;
+
+void paging_init(void) {
+  uint32_t *cr3;
+  asm volatile("mov %%cr3, %0" : "=r"(cr3));
+  uint32_t *virtual = (uint32_t *)((uint32_t)cr3 + 0xC0000000);
+
+  kernel_directory = &real_kernel_directory;
+  kernel_directory->physical_address = (uint32_t)cr3;
+  for (uint32_t i = 0; i < 1024; i++) {
+    kernel_directory->physical_tables[i] = virtual[i];
+
+    if (!kernel_directory->physical_tables[i]) {
+      kernel_directory->tables[i] = NULL;
+      continue;
+    }
+
+    kernel_directory->tables[i] =
+        (PageTable *)(0xC0000000 + (virtual[i] & ~(0xFFF)));
+
+    // Loop through the pages in the table
+    PageTable *table = kernel_directory->tables[i];
+    for (size_t j = 0; j < 1024; j++) {
+      if (!table->pages[j].present)
+        continue;
+      // Add the frame to our bitmap to ensure it does not get used by
+      // another newly created page.
+      write_to_frame(table->pages[j].frame * 0x1000, 1);
+    }
+  }
+
+  switch_page_directory(kernel_directory);
+  kernel_directory = clone_directory(kernel_directory);
+
+  // Make null dereferences crash.
+  switch_page_directory(kernel_directory);
+  get_page(NULL, kernel_directory, PAGE_ALLOCATE, 0)->present = 0;
+  switch_page_directory(clone_directory(kernel_directory));
+  // FIXME: Really hacky solution. Since page table creation needs to
+  // allocate memory but memory allocation requires page table creation
+  // they depend on eachother. The bad/current solution is just to
+  // allocate a bunch of tables so the memory allocator has enough.
+  for (uintptr_t i = 0; i < 140; i++)
+    allocate_frame(
+        get_page((void *)((0x302 + i) * 0x1000 * 1024), NULL, PAGE_ALLOCATE, 0),
+        1, 1);
+  move_stack(0xA0000000, 0xC00000);
+}