mm.txt 2.1 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647
  1. <previous description obsolete, deleted>
  2. Virtual memory map with 4 level page tables:
  3. 0000000000000000 - 00007fffffffffff (=47 bits) user space, different per mm
  4. hole caused by [48:63] sign extension
  5. ffff800000000000 - ffff87ffffffffff (=43 bits) guard hole, reserved for hypervisor
  6. ffff880000000000 - ffffc7ffffffffff (=64 TB) direct mapping of all phys. memory
  7. ffffc80000000000 - ffffc8ffffffffff (=40 bits) hole
  8. ffffc90000000000 - ffffe8ffffffffff (=45 bits) vmalloc/ioremap space
  9. ffffe90000000000 - ffffe9ffffffffff (=40 bits) hole
  10. ffffea0000000000 - ffffeaffffffffff (=40 bits) virtual memory map (1TB)
  11. ... unused hole ...
  12. ffffec0000000000 - fffffbffffffffff (=44 bits) kasan shadow memory (16TB)
  13. ... unused hole ...
  14. ffffff0000000000 - ffffff7fffffffff (=39 bits) %esp fixup stacks
  15. ... unused hole ...
  16. ffffffef00000000 - fffffffeffffffff (=64 GB) EFI region mapping space
  17. ... unused hole ...
  18. ffffffff80000000 - ffffffff9fffffff (=512 MB) kernel text mapping, from phys 0
  19. ffffffffa0000000 - ffffffffff5fffff (=1526 MB) module mapping space
  20. ffffffffff600000 - ffffffffffdfffff (=8 MB) vsyscalls
  21. ffffffffffe00000 - ffffffffffffffff (=2 MB) unused hole
  22. The direct mapping covers all memory in the system up to the highest
  23. memory address (this means in some cases it can also include PCI memory
  24. holes).
  25. vmalloc space is lazily synchronized into the different PML4 pages of
  26. the processes using the page fault handler, with init_level4_pgt as
  27. reference.
  28. Current X86-64 implementations support up to 46 bits of address space (64 TB),
  29. which is our current limit. This expands into MBZ space in the page tables.
  30. We map EFI runtime services in the 'efi_pgd' PGD in a 64Gb large virtual
  31. memory window (this size is arbitrary, it can be raised later if needed).
  32. The mappings are not part of any other kernel PGD and are only available
  33. during EFI runtime calls.
  34. Note that if CONFIG_RANDOMIZE_MEMORY is enabled, the direct mapping of all
  35. physical memory, vmalloc/ioremap space and virtual memory map are randomized.
  36. Their order is preserved but their base will be offset early at boot time.
  37. -Andi Kleen, Jul 2004