Thuc notes - OS Kernel Assembly

  • First part of the kernel. Will be called by the bootloader.
  • Define a mechanism that the bootloader can recognize and load the system kernel.
  • It implements the logic that can’t handle from a high-level kernel.
  • Basic functions:
    • Define a function _start that will be jumped to from bootloader as an entry-point.
    • Responsible to allocate stack to use.
    • Transfer to the high-level kernel by calling to function written by C, C++, Go, Rust etc.
    • Keep the computer in an infinite loop.

1. Multiboot standard layout

  • An interface standard that’s used by the bootloader to recognize and load the system kernel.
  • Bootloader will search multiboot headers (magic numbers) to recognize the kernel as multiboot compatible.

multiboot header

OffsetTypeField
0unsigned 32-bitmagic number
4unsigned 32-bitflags
8unsigned 32-bitchecksum
12unsigned 32-bitheader_addr
16unsigned 32-bitload_addr
20unsigned 32-bitload_end_addr
24unsigned 32-bitbss_end_addr
28unsigned 32-bitentry_addr
32unsigned 32-bitmode_type
36unsigned 32-bitwidth
40unsigned 32-bitheight
44unsigned 32-bitdepth
  • Include: magic number, flags and checksum

Magic number

  • The hexadecimal value always is 0x1BADB002

Flags:

  • Control address fields and graphics fields.
  • Bits 0-15 is a required feature. If the boot loader can’t understand the flag or fulfil them, it fails to load the OS image.
  • Bits 16-31 is optional features. If the boot loader can’t understand the flag or can’t fulfil them, it’s ignored.
BitDescription
0If it’s set, all boot modules loaded with must be aligned on 4KB page boundaries
1If it’s set, mem_lower and mem_upper in multiboot information structure, that’s passed by the boot loader to OS, must be included. mmap_length and mmap_addr are had in case the boot loader can pass it to.
2If it’s set, information about video mode table must be available in multiboot information structure, the boot loader passes that to OS, must be included
16If it’s set, header address fields are used by boot loader instead of the fields in the actual executable header. Don’t need to provide if the kernel image is in ELF format, required if it’s a.out format or in some other format

checksum:

  • Value’s always -(magic number + flags)
  • All of the address fields enabled by flag bit 16 are physical addresses.
  • Include: header_addr, load_addr, load_end_addr, bss_end_addr and entry_addr
  • header_addr: the physical address of the beginning of the multiboot header.
  • load_addr: the physical address of the beginning of text segment.
  • load_end_addr: the physical address of the end of the data segment. So the text and data segment are consecutive in the OS image.
  • bss_end_addr: the physical address of the end of the BSS segment.
  • entry_addr: the physical address to jump to start OS from the boot loader.
  • All of the graphics fields are enabled by flag bit 2.
  • Include: mode_type, width, height and depth
  • Specify the preferred graphics mode
  • Recommended mode by OS image, the boot loader may choose a different mode if it sees fit.
  • mode_type: 0 = linear graphics mode, 1 = EGA-standard text mode
  • width: The number of columns
  • height: The number of the lines
  • depth: The number of bits per pixel in a graphics mode, zero in a text mode.

1.2. Stack definition

  • Create a section; allocate 16 Kb; create a symbol at the top by setting the esp register. The stack grows downwards on x86.
  • Aligned 16-byte according to the System V ABI standard and de-facto extensions

1.3. Infinite loop

  • Disable CLI interrupts
  • Use hlt to wait for the next interrupt to archive.
  • Jump back to the waiting step for hlt again if the computer wakes up by any interrupting.

1.4. Multiboot information structure

  • Multiboot information structure is data that the boot loader communicates to the OS.
  • The EBX register contains the physical address of the multiboot information structure.
OffsetSizeFieldDescription
04flags
44mem_lowerMemory definition
84mem_upperMemory definition
124boot_deviceBoot device information
164cmdlineCommand line information
204mods_countLoadded modules
244mods_addrLoadded modules
28-4016symsDetail fields between those addresses are controlled by the flag
444mmap_lengthMemory map
484mmap_lengthMemory map
524drives_lengthDrives
564drives_addrDrives
604config_tableConfig table
644boot_loader_nameBoot loader name
684apm_tableAdvanced Power Management table
7216 (2 for each)vbe_*
8821framebuffer_*framebuffer

Flags:

  • Indicates features are defined in the following fields will be used.
BitDescription
0Enable mem_lower and mem_upper fields to control memory boundary
1Enable to set the boot_device fields
2Enable to set the cmdline fields
3Enable to set the modules fields
4Defile structure offset 28-40. 4 or 5 is set, another is unset.
6Enable to set the memory map fields
7Enable to set the drives fields
8Enable to set the config table field
9Enable to set the boot loader name field
10Enable to set the Advanced Power Management table fields
11Enable to set the VBE fields
12Enable to set the framebuffer fields

Memory:

  • mem_lower, mem_upper.
  • Define total available memory that can use.

Boot device:

  • boot_device.
  • Indicate boot device which BIOSK disk device the boot loader loaded the OS image from
  • If the OS is not booted by BIOSK disk, it should be clear.

Command line:

  • cmdline, byte offset = 16, size = 4.
  • Define the physical address of the command line that is passed to the kernel.

Modules:

  • mods_count, mods_addr.
  • Indicate what boot modules were loaded along with the kernel image.
  • mods_count = number of modules loaded.
  • mods_addr = physical address of the first module.
  • Each module will have the following structure
OffsetSizeFieldDescription
04mod_startphysical starting address of module
44mod_endphysical ending address of module
84mod_endarbitrary string of module. It can be command line, pathname
124reservedset 0 by the boot loader and ignored by OS

Symbol table:

  • If bit 4 of the flag are set, the following symbol location info is used
OffsetSizeFieldDescription
284tabsizesize parameter
324strsizesize parameter
364addrphysical address of and array of a.out format nlist structures
404reservedjust for reversed
  • If bit 5 of the flag are set, the following s section header table from ELF info is used
OffsetSizeFieldDescription
284numnumber of entries
324sizesize of each entry
364addrPhysical address fields of the ELF section header
404shndx

Memory map:

  • mmap_length, mmap_addr
  • Indicate a buffer containing a memory map starting address and its length. BIOS provides the buffer having a memory map. The map provided is guaranteed to list all standard RAM that should be available for usage.
  • Each item in the memory map has structure:
OffsetSizeFieldDescription
-44sizesize of memory map item
08base_addrphysical address of the memory that the map define to
88lengththe size of the memory region
168typethe variety of address range represented. 1 = available RAM, 3 = usable memory holding ACPI information, 4 = reserved memory which needs to be preserved on hibernation, 5 = occupied by defective RAM modules and all other values currently indicated a reserved area

Drives:

  • drives_length, drives_addr
  • Indicate the address of the physical address of the first drive structure and the size of drive structures.

Config table:

  • config_table
  • Indicates the address of the ROM configuration table returned by the GET CONFIGURATION BIOS call

Boot loader name:

  • boot_loader_name
  • Contains the physical address of the name of a boot loader booting the kernel

Advanced Power Management table:

VBE

Graphics

2. Multiboot2 standard layout

  • TBD.
  • Magic number: 0xE85250D6

3. References