The physical address space is divided into fixed regions established by the bootloader and kernel initialisation code. All addresses are physical (paging is identity-mapped for the kernel range; user space is virtually addressed through the two-level page tables).
Physical address Size Contents
─────────────────────────────────────────────────────────────────
0x00000000 – 0x00001FFF 8 KB BIOS data area (IVT, BDA)
0x00002000 – 0x00006FFF 20 KB Boot stack (grows ↓ from 0x7000)
0x00007C00 – 0x00007DFF 512 B Boot sector (MBR, first 512 bytes of disk)
0x00008000 – 0x000117FF ~40 KB Kernel image (80 sectors loaded by bootloader)
0x0000A000 (entry) kernel_gate entry point called by bootloader
0x00050000 – 0x000CFFFF 512 KB Kernel heap (initialised before any subsystem)
0x00100000 – 0x00EFFFFF 14 MB Paging pool (page directories and page tables)
0x01000000 – 0x0FFFFFFF 240 MB User space (frame-allocated via bitmap)
Established by the bootloader before entering protected mode.
stack_pointer = 0x7000, stack_size = 0x5000 (20 KB, grows downward).
The stack region is mapped as a user-accessible page so that future user-mode tasks can share the initial stack layout.
Initialised as the very first kernel subsystem — before interrupts, before paging.
This allows the allocator (ralloc/zalloc) to be used by every subsequent subsystem.
Size: kernel_heap_size = 0x80000 (512 KB).
The heap uses a best-fit strategy with intrusive block headers; size_of(ptr) recovers the allocated size from any pointer.
Managed by pages.c as a bitmap of 4 KB frames.
Used exclusively for page directory and page table allocation (allocate_page_folder()).
The pool bitmap itself is allocated from the kernel heap at initialise_paging() time.
256 MB of virtually-addressed user space, managed by a second bitmap frame allocator (allocate_frame()).
Each process receives its own page directory (cloned from the kernel directory by clone_directory() during fork()).
Kernel mappings are shared (same physical pointers); user pages are deep-copied.