Kernel::x86_64: Re-copy boot.S and add longmode in

Now we should have the posibility to run in longmode, after enabling it
in the gdt

Kernel/Arch/x86_64/Boot/boot.S

@@ -1,3 +1,4 @@
+.code32
 .set MULTIBOOT_MAGIC,         0x1badb002
 .set MULTIBOOT_PAGE_ALIGN,    0x1
 .set MULTIBOOT_MEMORY_INFO,   0x2
@@ -61,12 +62,42 @@ boot_pd3_pt1023:
 .global start
 .type start, @function
 
-.extern init
+.extern init 
 .type init, @function
 
 .extern multiboot_info_ptr
 .type multiboot_info_ptr, @object
 
+/*
+    construct the following (32-bit PAE) page table layout:
+
+pdpt
+
+    0: boot_pd0 (0-1GB)
+    1: n/a      (1-2GB)
+    2: n/a      (2-3GB)
+    3: boot_pd3 (3-4GB)
+
+boot_pd0 : 512 pde's
+
+    0: boot_pd0_pt0 (0-2MB) (id 512 4KB pages)
+
+boot_pd3 : 512 pde's
+
+    0: boot_pd3_pts[0] (3072-3074MB) (pseudo 512 4KB pages)
+    1: boot_pd3_pts[1] (3074-3076MB) (pseudo 512 4KB pages)
+    2: boot_pd3_pts[2] (3076-3078MB) (pseudo 512 4KB pages)
+    3: boot_pd3_pts[3] (3078-3080MB) (pseudo 512 4KB pages)
+    4: boot_pd3_pts[4] (3080-3082MB) (pseudo 512 4KB pages)
+    5: boot_pd3_pts[5] (3082-3084MB) (pseudo 512 4KB pages)
+    6: boot_pd3_pts[6] (3084-3086MB) (pseudo 512 4KB pages)
+    7: boot_pd3_pts[7] (3086-3088MB) (pseudo 512 4KB pages)
+    8: boot_pd3_pt1023 (4094-4096MB) (for page table mappings)
+
+the 9 page tables each contain 512 pte's that map individual 4KB pages
+
+*/
+
 start:
     cli
     cld
@@ -84,6 +115,139 @@ start:
     movl $(kernel_cmdline - 0xc0000000), %edi
     rep movsl
 
+    /* clear pdpt */
+    movl $(boot_pdpt - 0xc0000000), %edi
+    movl $1024, %ecx
+    xorl %eax, %eax
+    rep stosl
+
+    /* set up pdpt[0] and pdpt[3] */
+    movl $(boot_pdpt - 0xc0000000), %edi
+    movl $((boot_pd0 - 0xc0000000) + 1), 0(%edi)
+    movl $((boot_pd3 - 0xc0000000) + 1), 24(%edi)
+
+    /* clear pd0 */
+    movl $(boot_pd0 - 0xc0000000), %edi
+    movl $1024, %ecx
+    xorl %eax, %eax
+    rep stosl
+
+    /* clear pd3 */
+    movl $(boot_pd3 - 0xc0000000), %edi
+    movl $1024, %ecx
+    xorl %eax, %eax
+    rep stosl
+
+    /* clear pd0's pt's */
+    movl $(boot_pd0_pt0 - 0xc0000000), %edi
+    movl $(1024 * 4), %ecx
+    xorl %eax, %eax
+    rep stosl
+
+    /* clear pd3's pt's */
+    movl $(boot_pd3_pts - 0xc0000000), %edi
+    movl $(1024 * 9), %ecx
+    xorl %eax, %eax
+    rep stosl
+
+    /* add boot_pd0_pt0 to boot_pd0 */
+    movl $(boot_pd0 - 0xc0000000), %edi
+    movl $(boot_pd0_pt0 - 0xc0000000), %eax
+    movl %eax, 0(%edi)
+    /* R/W + Present */
+    orl $0x3, 0(%edi)
+
+    /* add boot_pd3_pts to boot_pd3 */
+    movl $8, %ecx
+    movl $(boot_pd3 - 0xc0000000), %edi
+    movl $(boot_pd3_pts - 0xc0000000), %eax
+
+1:
+    movl %eax, 0(%edi)
+    /* R/W + Present */
+    orl $0x3, 0(%edi)
+    addl $8, %edi
+    addl $4096, %eax
+    loop 1b
+
+    /* identity map the 0 to 2MB range */
+    movl $512, %ecx
+    movl $(boot_pd0_pt0 - 0xc0000000), %edi
+    xorl %eax, %eax
+
+1:
+    movl %eax, 0(%edi)
+    /* R/W + Present */
+    orl $0x3, 0(%edi)
+    addl $8, %edi
+    addl $4096, %eax
+    loop 1b
+
+    /* pseudo identity map the 3072-3090MB range */
+    movl $(512 * 8), %ecx
+    movl $(boot_pd3_pts - 0xc0000000), %edi
+    xorl %eax, %eax
+
+1:
+    movl %eax, 0(%edi)
+    /* R/W + Present */
+    orl $0x3, 0(%edi)
+    addl $8, %edi
+    addl $4096, %eax
+    loop 1b
+
+    /* create an empty page table for the top 2MB at the 4GB mark */
+    movl $(boot_pd3 - 0xc0000000), %edi
+    movl $(boot_pd3_pt1023 - 0xc0000000), 4088(%edi)
+    orl $0x3, 4088(%edi)
+    movl $0, 4092(%edi)
+
+    /* point CR3 to PDPT */
+    movl $(boot_pdpt - 0xc0000000), %eax
+    movl %eax, %cr3
+
+    /* enable PAE + PSE */
+    movl %cr4, %eax
+    orl $0x60, %eax
+    movl %eax, %cr4
+
+1:
+    /* Enter Long-mode! ref(https://wiki.osdev.org/Setting_Up_Long_Mode)*/
+    mov $0xC0000080, %ecx           /* Set the C-register to 0xC0000080, which is the EFER MSR.*/
+    rdmsr                           /* Read from the model-specific register.*/
+    or $(1 << 8), %eax              /* Set the LM-bit which is the 9th bit (bit 8).*/
+    wrmsr                           /* Write to the model-specific register.*/
+
+    /* enable PG */
+    movl %cr0, %eax
+    orl $0x80000000, %eax
+    movl %eax, %cr0
+    /* Now we are in 32-bit compatablity mode, We still need to load a 64-bit GDT */
+
+    /* set up stack */
+    mov $stack_top, %esp
+    and $-16, %esp
+
+    /* jmp to an address above the 3GB mark */
+    movl $1f,%eax
+    jmp *%eax
+1:
+    movl %cr3, %eax
+    movl %eax, %cr3
+
+    /* unmap the 0-1MB range, which isn't used after jmp-ing up here */
+    movl $256, %ecx
+    movl $(boot_pd0_pt0 - 0xc0000000), %edi
+    xorl %eax, %eax
+1:
+    movl %eax, 0(%edi)
+    addl $8, %edi
+    loop 1b
+
+    /* jump into C++ land */
+    addl $0xc0000000, %ebx
+    movl %ebx, multiboot_info_ptr
+
     call init
     add $4, %esp
 
@@ -131,8 +295,92 @@ apic_ap_start:
     ljmpl $8, $(apic_ap_start32 - apic_ap_start + 0x8000)
 apic_ap_start32:
 .code32
-    cli
-    hlt
+    mov $0x10, %ax
+    mov %ax, %ss
+    mov %ax, %ds
+    mov %ax, %es
+    mov %ax, %fs
+    mov %ax, %gs
+
+    movl $0x8000, %ebp
+
+    /* generate a unique ap cpu id (0 means 1st ap, not bsp!) */
+    xorl %eax, %eax
+    incl %eax
+    lock; xaddl %eax, (ap_cpu_id - apic_ap_start)(%ebp) /* avoid relocation entries */
+    movl %eax, %esi
+
+    /* find our allocated stack based on the generated id */
+    movl (ap_cpu_init_stacks - apic_ap_start)(%ebp, %eax, 4), %esp
+
+    /* check if we support NX and enable it if we do */
+    movl $0x80000001, %eax
+    cpuid
+    testl $0x100000, %edx
+    je (1f - apic_ap_start + 0x8000)
+    /* turn on IA32_EFER.NXE */
+    movl $0xc0000080, %ecx
+    rdmsr
+    orl $0x800, %eax
+    wrmsr
+1:
+    
+    /* load the bsp's cr3 value */
+    movl (ap_cpu_init_cr3 - apic_ap_start)(%ebp), %eax
+    movl %eax, %cr3
+    
+    /* enable PAE + PSE */
+    movl %cr4, %eax
+    orl $0x60, %eax
+    movl %eax, %cr4
+
+    /* enable PG */
+    movl %cr0, %eax
+    orl $0x80000000, %eax
+    movl %eax, %cr0
+
+    /* load a second temporary gdt that points above 3GB */
+    lgdt (ap_cpu_gdtr_initial2 - apic_ap_start + 0xc0008000)
+
+    /* jump above 3GB into our identity mapped area now */
+    ljmp $8, $(apic_ap_start32_2 - apic_ap_start + 0xc0008000)
+apic_ap_start32_2:
+    /* flush the TLB */
+    movl %cr3, %eax
+    movl %eax, %cr3
+    
+    movl $0xc0008000, %ebp
+    
+    /* now load the final gdt and idt from the identity mapped area */
+    movl (ap_cpu_gdtr - apic_ap_start)(%ebp), %eax
+    lgdt (%eax)
+    movl (ap_cpu_idtr - apic_ap_start)(%ebp), %eax
+    lidt (%eax)
+
+    /* set same cr0 and cr4 values as the BSP */
+    movl (ap_cpu_init_cr0 - apic_ap_start)(%ebp), %eax
+    movl %eax, %cr0
+    movl (ap_cpu_init_cr4 - apic_ap_start)(%ebp), %eax
+    movl %eax, %cr4
+    
+    /* push the Processor pointer this CPU is going to use */
+    movl (ap_cpu_init_processor_info_array - apic_ap_start)(%ebp), %eax
+    addl $0xc0000000, %eax
+    movl 0(%eax, %esi, 4), %eax
+    push %eax
+    
+    /* push the cpu id, 0 representing the bsp and call into c++ */
+    incl %esi
+    push %esi
+    
+    xor %ebp, %ebp
+    cld
+    
+    /* We are in identity mapped P0x8000 and the BSP will unload this code
+       once all APs are initialized, so call init_ap but return to our
+       infinite loop */
+    push $loop
+    ljmp $8, $init_ap
 
 .align 4
 .global apic_ap_start_size