Higher Half x86 Bare Bones (Backup)
Difficulty level |
---|
Medium |
Here is some sample code for a kernel that is loaded by GRUB and is mapped in the upper half of memory. In this case, the kernel is loaded at 1MB in the physical address space (0x00100000), but is mapped at 3GB + 1MB in the virtual address space (0xC0100000). It is recommended that you have a firm grasp of the contents within the Bare bones tutorial and Paging before attempting this.
loader.asm
This piece of code is taking over control from the Multiboot bootloader. It sets up a page directory with page table entries that identity map the first 4MB, and also map the first 4MB to virtual 3GB. After setting up paging, it unmaps the identity mapping so that the kernel is entirely in the higher half and jumps into the kernel proper.
global _loader ; Make entry point visible to linker.
extern _main ; _main is defined elsewhere
; setting up the Multiboot header - see GRUB docs for details
MODULEALIGN equ 1<<0 ; align loaded modules on page boundaries
MEMINFO equ 1<<1 ; provide memory map
FLAGS equ MODULEALIGN | MEMINFO ; this is the Multiboot 'flag' field
MAGIC equ 0x1BADB002 ; 'magic number' lets bootloader find the header
CHECKSUM equ -(MAGIC + FLAGS) ; checksum required
; This is the virtual base address of kernel space. It must be used to convert virtual
; addresses into physical addresses until paging is enabled. Note that this is not
; the virtual address where the kernel image itself is loaded -- just the amount that must
; be subtracted from a virtual address to get a physical address.
KERNEL_VIRTUAL_BASE equ 0xC0000000 ; 3GB
KERNEL_PAGE_NUMBER equ (KERNEL_VIRTUAL_BASE >> 22) ; Page directory index of kernel's 4MB PTE.
section .data
align 0x1000
BootPageDirectory:
; This page directory entry identity-maps the first 4MB of the 32-bit physical address space.
; All bits are clear except the following:
; bit 7: PS The kernel page is 4MB.
; bit 1: RW The kernel page is read/write.
; bit 0: P The kernel page is present.
; This entry must be here -- otherwise the kernel will crash immediately after paging is
; enabled because it can't fetch the next instruction! It's ok to unmap this page later.
dd 0x00000083
times (KERNEL_PAGE_NUMBER - 1) dd 0 ; Pages before kernel space.
; This page directory entry defines a 4MB page containing the kernel.
dd 0x00000083
times (1024 - KERNEL_PAGE_NUMBER - 1) dd 0 ; Pages after the kernel image.
section .text
align 4
MultiBootHeader:
dd MAGIC
dd FLAGS
dd CHECKSUM
; reserve initial kernel stack space -- that's 16k.
STACKSIZE equ 0x4000
; setting up entry point for linker
loader equ (_loader - 0xC0000000)
global loader
_loader:
; NOTE: Until paging is set up, the code must be position-independent and use physical
; addresses, not virtual ones!
mov ecx, (BootPageDirectory - KERNEL_VIRTUAL_BASE)
mov cr3, ecx ; Load Page Directory Base Register.
mov ecx, cr4
or ecx, 0x00000010 ; Set PSE bit in CR4 to enable 4MB pages.
mov cr4, ecx
mov ecx, cr0
or ecx, 0x80000000 ; Set PG bit in CR0 to enable paging.
mov cr0, ecx
; Start fetching instructions in kernel space.
; Since eip at this point holds the physical address of this command (approximately 0x00100000)
; we need to do a long jump to the correct virtual address of StartInHigherHalf which is
; approximately 0xC0100000.
lea ecx, [StartInHigherHalf]
jmp ecx ; NOTE: Must be absolute jump!
StartInHigherHalf:
; Unmap the identity-mapped first 4MB of physical address space. It should not be needed
; anymore.
mov dword [BootPageDirectory], 0
invlpg [0]
; NOTE: From now on, paging should be enabled. The first 4MB of physical address space is
; mapped starting at KERNEL_VIRTUAL_BASE. Everything is linked to this address, so no more
; position-independent code or funny business with virtual-to-physical address translation
; should be necessary. We now have a higher-half kernel.
mov esp, stack+STACKSIZE ; set up the stack
push eax ; pass Multiboot magic number
; pass Multiboot info structure -- WARNING: This is a physical address and may not be
; in the first 4MB!
push ebx
call _main ; call kernel proper
hlt ; halt machine should kernel return
section .bss
align 32
stack:
resb STACKSIZE ; reserve 16k stack on a uint64_t boundary
linker.ld
This is a little trickier than it was for the C kernel tutorial, since you need to distinguish between virtual addresses (which will be in the higher half) and load addresses, which GRUB needs to decide where to put your kernel.
ENTRY(loader)
OUTPUT_FORMAT(elf32-i386)
SECTIONS {
/* The kernel will live at 3GB + 1MB in the virtual
address space, which will be mapped to 1MB in the
physical address space. */
. = 0xC0100000;
.text : AT(ADDR(.text) - 0xC0000000) {
*(.text)
*(.rodata*)
}
.data ALIGN (0x1000) : AT(ADDR(.data) - 0xC0000000) {
*(.data)
}
.bss : AT(ADDR(.bss) - 0xC0000000) {
_sbss = .;
*(COMMON)
*(.bss)
_ebss = .;
}
}
Note that we use loader (and not _loader) as our entry point. This is due to the fact that _loader's address is approximately 0xC0100000, if we try to set our eip to that address it will not find our loader function. Also note our entry point is not being converted to physical address. GRUB does this conversion when calculating starting value of EIP, and if you attempt to do the translation, you may get your execution when you don't want it or get "entry point isn't in a segment" error.
kernel.c
Using the kernel.c code from the original bare bones tutorial will work fine, with one small change. On the fourth line in terminal_initialize(), change:
terminal_buffer = (uint16_t*) 0xB8000;
to
terminal_buffer = (uint16_t*) 0xC00B8000;
This accomodates for the kernel's new offset into higher-half space. Any direct memory access by the kernel at this point should take place with respect to this offset where necessary.
Troubleshooting
- I got a page fault (#PF) when accessing my GRUB Multiboot info structure
The address passed by loader.s is physical, you have to make it virtual to add your virtual base to it. For example, in your loader.s:
add ebx, KERNEL_VIRTUAL_BASE ; make the address virtual
push ebx ; push it on the stack for _main()
Don't forget to make all addresses pointing to memory locations in the Multiboot info structure also virtual.
See Also
Articles
- Bare bones
- An alternate higher half Bare Bones.