mm.rst (11020B)
1.. SPDX-License-Identifier: GPL-2.0 2 3================= 4Memory Management 5================= 6 7Complete virtual memory map with 4-level page tables 8==================================================== 9 10.. note:: 11 12 - Negative addresses such as "-23 TB" are absolute addresses in bytes, counted down 13 from the top of the 64-bit address space. It's easier to understand the layout 14 when seen both in absolute addresses and in distance-from-top notation. 15 16 For example 0xffffe90000000000 == -23 TB, it's 23 TB lower than the top of the 17 64-bit address space (ffffffffffffffff). 18 19 Note that as we get closer to the top of the address space, the notation changes 20 from TB to GB and then MB/KB. 21 22 - "16M TB" might look weird at first sight, but it's an easier way to visualize size 23 notation than "16 EB", which few will recognize at first sight as 16 exabytes. 24 It also shows it nicely how incredibly large 64-bit address space is. 25 26:: 27 28 ======================================================================================================================== 29 Start addr | Offset | End addr | Size | VM area description 30 ======================================================================================================================== 31 | | | | 32 0000000000000000 | 0 | 00007fffffffffff | 128 TB | user-space virtual memory, different per mm 33 __________________|____________|__________________|_________|___________________________________________________________ 34 | | | | 35 0000800000000000 | +128 TB | ffff7fffffffffff | ~16M TB | ... huge, almost 64 bits wide hole of non-canonical 36 | | | | virtual memory addresses up to the -128 TB 37 | | | | starting offset of kernel mappings. 38 __________________|____________|__________________|_________|___________________________________________________________ 39 | 40 | Kernel-space virtual memory, shared between all processes: 41 ____________________________________________________________|___________________________________________________________ 42 | | | | 43 ffff800000000000 | -128 TB | ffff87ffffffffff | 8 TB | ... guard hole, also reserved for hypervisor 44 ffff880000000000 | -120 TB | ffff887fffffffff | 0.5 TB | LDT remap for PTI 45 ffff888000000000 | -119.5 TB | ffffc87fffffffff | 64 TB | direct mapping of all physical memory (page_offset_base) 46 ffffc88000000000 | -55.5 TB | ffffc8ffffffffff | 0.5 TB | ... unused hole 47 ffffc90000000000 | -55 TB | ffffe8ffffffffff | 32 TB | vmalloc/ioremap space (vmalloc_base) 48 ffffe90000000000 | -23 TB | ffffe9ffffffffff | 1 TB | ... unused hole 49 ffffea0000000000 | -22 TB | ffffeaffffffffff | 1 TB | virtual memory map (vmemmap_base) 50 ffffeb0000000000 | -21 TB | ffffebffffffffff | 1 TB | ... unused hole 51 ffffec0000000000 | -20 TB | fffffbffffffffff | 16 TB | KASAN shadow memory 52 __________________|____________|__________________|_________|____________________________________________________________ 53 | 54 | Identical layout to the 56-bit one from here on: 55 ____________________________________________________________|____________________________________________________________ 56 | | | | 57 fffffc0000000000 | -4 TB | fffffdffffffffff | 2 TB | ... unused hole 58 | | | | vaddr_end for KASLR 59 fffffe0000000000 | -2 TB | fffffe7fffffffff | 0.5 TB | cpu_entry_area mapping 60 fffffe8000000000 | -1.5 TB | fffffeffffffffff | 0.5 TB | ... unused hole 61 ffffff0000000000 | -1 TB | ffffff7fffffffff | 0.5 TB | %esp fixup stacks 62 ffffff8000000000 | -512 GB | ffffffeeffffffff | 444 GB | ... unused hole 63 ffffffef00000000 | -68 GB | fffffffeffffffff | 64 GB | EFI region mapping space 64 ffffffff00000000 | -4 GB | ffffffff7fffffff | 2 GB | ... unused hole 65 ffffffff80000000 | -2 GB | ffffffff9fffffff | 512 MB | kernel text mapping, mapped to physical address 0 66 ffffffff80000000 |-2048 MB | | | 67 ffffffffa0000000 |-1536 MB | fffffffffeffffff | 1520 MB | module mapping space 68 ffffffffff000000 | -16 MB | | | 69 FIXADDR_START | ~-11 MB | ffffffffff5fffff | ~0.5 MB | kernel-internal fixmap range, variable size and offset 70 ffffffffff600000 | -10 MB | ffffffffff600fff | 4 kB | legacy vsyscall ABI 71 ffffffffffe00000 | -2 MB | ffffffffffffffff | 2 MB | ... unused hole 72 __________________|____________|__________________|_________|___________________________________________________________ 73 74 75Complete virtual memory map with 5-level page tables 76==================================================== 77 78.. note:: 79 80 - With 56-bit addresses, user-space memory gets expanded by a factor of 512x, 81 from 0.125 PB to 64 PB. All kernel mappings shift down to the -64 PB starting 82 offset and many of the regions expand to support the much larger physical 83 memory supported. 84 85:: 86 87 ======================================================================================================================== 88 Start addr | Offset | End addr | Size | VM area description 89 ======================================================================================================================== 90 | | | | 91 0000000000000000 | 0 | 00ffffffffffffff | 64 PB | user-space virtual memory, different per mm 92 __________________|____________|__________________|_________|___________________________________________________________ 93 | | | | 94 0100000000000000 | +64 PB | feffffffffffffff | ~16K PB | ... huge, still almost 64 bits wide hole of non-canonical 95 | | | | virtual memory addresses up to the -64 PB 96 | | | | starting offset of kernel mappings. 97 __________________|____________|__________________|_________|___________________________________________________________ 98 | 99 | Kernel-space virtual memory, shared between all processes: 100 ____________________________________________________________|___________________________________________________________ 101 | | | | 102 ff00000000000000 | -64 PB | ff0fffffffffffff | 4 PB | ... guard hole, also reserved for hypervisor 103 ff10000000000000 | -60 PB | ff10ffffffffffff | 0.25 PB | LDT remap for PTI 104 ff11000000000000 | -59.75 PB | ff90ffffffffffff | 32 PB | direct mapping of all physical memory (page_offset_base) 105 ff91000000000000 | -27.75 PB | ff9fffffffffffff | 3.75 PB | ... unused hole 106 ffa0000000000000 | -24 PB | ffd1ffffffffffff | 12.5 PB | vmalloc/ioremap space (vmalloc_base) 107 ffd2000000000000 | -11.5 PB | ffd3ffffffffffff | 0.5 PB | ... unused hole 108 ffd4000000000000 | -11 PB | ffd5ffffffffffff | 0.5 PB | virtual memory map (vmemmap_base) 109 ffd6000000000000 | -10.5 PB | ffdeffffffffffff | 2.25 PB | ... unused hole 110 ffdf000000000000 | -8.25 PB | fffffbffffffffff | ~8 PB | KASAN shadow memory 111 __________________|____________|__________________|_________|____________________________________________________________ 112 | 113 | Identical layout to the 47-bit one from here on: 114 ____________________________________________________________|____________________________________________________________ 115 | | | | 116 fffffc0000000000 | -4 TB | fffffdffffffffff | 2 TB | ... unused hole 117 | | | | vaddr_end for KASLR 118 fffffe0000000000 | -2 TB | fffffe7fffffffff | 0.5 TB | cpu_entry_area mapping 119 fffffe8000000000 | -1.5 TB | fffffeffffffffff | 0.5 TB | ... unused hole 120 ffffff0000000000 | -1 TB | ffffff7fffffffff | 0.5 TB | %esp fixup stacks 121 ffffff8000000000 | -512 GB | ffffffeeffffffff | 444 GB | ... unused hole 122 ffffffef00000000 | -68 GB | fffffffeffffffff | 64 GB | EFI region mapping space 123 ffffffff00000000 | -4 GB | ffffffff7fffffff | 2 GB | ... unused hole 124 ffffffff80000000 | -2 GB | ffffffff9fffffff | 512 MB | kernel text mapping, mapped to physical address 0 125 ffffffff80000000 |-2048 MB | | | 126 ffffffffa0000000 |-1536 MB | fffffffffeffffff | 1520 MB | module mapping space 127 ffffffffff000000 | -16 MB | | | 128 FIXADDR_START | ~-11 MB | ffffffffff5fffff | ~0.5 MB | kernel-internal fixmap range, variable size and offset 129 ffffffffff600000 | -10 MB | ffffffffff600fff | 4 kB | legacy vsyscall ABI 130 ffffffffffe00000 | -2 MB | ffffffffffffffff | 2 MB | ... unused hole 131 __________________|____________|__________________|_________|___________________________________________________________ 132 133Architecture defines a 64-bit virtual address. Implementations can support 134less. Currently supported are 48- and 57-bit virtual addresses. Bits 63 135through to the most-significant implemented bit are sign extended. 136This causes hole between user space and kernel addresses if you interpret them 137as unsigned. 138 139The direct mapping covers all memory in the system up to the highest 140memory address (this means in some cases it can also include PCI memory 141holes). 142 143We map EFI runtime services in the 'efi_pgd' PGD in a 64Gb large virtual 144memory window (this size is arbitrary, it can be raised later if needed). 145The mappings are not part of any other kernel PGD and are only available 146during EFI runtime calls. 147 148Note that if CONFIG_RANDOMIZE_MEMORY is enabled, the direct mapping of all 149physical memory, vmalloc/ioremap space and virtual memory map are randomized. 150Their order is preserved but their base will be offset early at boot time. 151 152Be very careful vs. KASLR when changing anything here. The KASLR address 153range must not overlap with anything except the KASAN shadow area, which is 154correct as KASAN disables KASLR. 155 156For both 4- and 5-level layouts, the STACKLEAK_POISON value in the last 2MB 157hole: ffffffffffff4111