Architecture: x86_64

x86_64

The x86_64 implementation uses GRUB as bootloader.

config file

# This is the default architecture so we don't need to specify `ARCH`.
 
# This is only useful when QEMU is used.
CONFIG_SEMIHOSTING = 1

Boot options

The GRUB configuration offers two choices: the normal mode and the kernel mode. The normal (or "default") mode will load the /bin/init program in user mode ("ring 3"). The kernel mode will load the kshell instead and will stay in kernel mode (as its name implies).

It is possible to change the command line by passing the CMDLINE variable to make (or add it to your config file):

$ make clean run-debug CMDLINE="/bin/init foo bar"

Note

It is also possible to set KERNEL_CMDLINE, which is passed to the kernel in... kernel mode.

Early boot sequence

1. The BIOS loads the bootloader (GRUB) from the ISO
2. The bootloader reads the kernel executable and finds the multiboot header
3. The bootloader copies the .boot and .text sections to memory (to addresses 0x100000 and 0x100020)
4. The bootloader jumps to the entry point (at 0x100020)
5. Various checks are performed to make sure we can continue the boot process
6. Page table and paging are configured and enabled
7. Load a 64-bit Global Descriptor Table
8. Load a Task State Segment
9. Reload all the data segment registers
10. Call the C kmain() function
11. Eventually switch to user mode and call /bin/init by default

Memory management

During the early boot sequence, we identity map the first gigabyte of our kernel with 512 2MiB pages. When kmain() is called, we call paging_init() to update the mapping:

1. Memory from 0x00000000 to 0x000A0000 is identity mapped (present)
2. Memory from 0x000A0000 to 0x00100000 is identity mapped (present + writable)
3. The VBE framebuffer might be identity mapped as well (present + writable, and only if ENABLE_FRAMEBUFFER=1 is passed to make)
4. The kernel sections are identity mapped with the correct flags for each section
5. The multiboot information is identity mapped
6. The multiboot modules are identity mapped with the correct flags (in our case, it is the init ramdisk)
7. The memory allocated for the frame allocator is then identity mapped

In addition, paging_init() will:

1. enable the write protection
2. enable the no-execute feature
3. create a guard page

Memory map

Virtual address space

• 0x00000000 to 0x00100000: the first 1MiB area is reserved
• 0x00100000: the kernel code and data and multiboot modules
• 0x0020xxxx: after the end of the previous area (which is computed at runtime using the multiboot information), we allocate some space for the frame allocator
• 0x10000000: kernel heap area (we currently have a fixed heap size)
• 0x40000000: user space