1 ================= 1 =================
2 Concepts overview 2 Concepts overview
3 ================= 3 =================
4 4
5 The memory management in Linux is a complex sy 5 The memory management in Linux is a complex system that evolved over the
6 years and included more and more functionality 6 years and included more and more functionality to support a variety of
7 systems from MMU-less microcontrollers to supe 7 systems from MMU-less microcontrollers to supercomputers. The memory
8 management for systems without an MMU is calle 8 management for systems without an MMU is called ``nommu`` and it
9 definitely deserves a dedicated document, whic 9 definitely deserves a dedicated document, which hopefully will be
10 eventually written. Yet, although some of the 10 eventually written. Yet, although some of the concepts are the same,
11 here we assume that an MMU is available and a 11 here we assume that an MMU is available and a CPU can translate a virtual
12 address to a physical address. 12 address to a physical address.
13 13
14 .. contents:: :local: 14 .. contents:: :local:
15 15
16 Virtual Memory Primer 16 Virtual Memory Primer
17 ===================== 17 =====================
18 18
19 The physical memory in a computer system is a 19 The physical memory in a computer system is a limited resource and
20 even for systems that support memory hotplug t 20 even for systems that support memory hotplug there is a hard limit on
21 the amount of memory that can be installed. Th 21 the amount of memory that can be installed. The physical memory is not
22 necessarily contiguous; it might be accessible 22 necessarily contiguous; it might be accessible as a set of distinct
23 address ranges. Besides, different CPU archite 23 address ranges. Besides, different CPU architectures, and even
24 different implementations of the same architec 24 different implementations of the same architecture have different views
25 of how these address ranges are defined. 25 of how these address ranges are defined.
26 26
27 All this makes dealing directly with physical 27 All this makes dealing directly with physical memory quite complex and
28 to avoid this complexity a concept of virtual 28 to avoid this complexity a concept of virtual memory was developed.
29 29
30 The virtual memory abstracts the details of ph 30 The virtual memory abstracts the details of physical memory from the
31 application software, allows to keep only need 31 application software, allows to keep only needed information in the
32 physical memory (demand paging) and provides a 32 physical memory (demand paging) and provides a mechanism for the
33 protection and controlled sharing of data betw 33 protection and controlled sharing of data between processes.
34 34
35 With virtual memory, each and every memory acc 35 With virtual memory, each and every memory access uses a virtual
36 address. When the CPU decodes an instruction t 36 address. When the CPU decodes an instruction that reads (or
37 writes) from (or to) the system memory, it tra 37 writes) from (or to) the system memory, it translates the `virtual`
38 address encoded in that instruction to a `phys 38 address encoded in that instruction to a `physical` address that the
39 memory controller can understand. 39 memory controller can understand.
40 40
41 The physical system memory is divided into pag 41 The physical system memory is divided into page frames, or pages. The
42 size of each page is architecture specific. So 42 size of each page is architecture specific. Some architectures allow
43 selection of the page size from several suppor 43 selection of the page size from several supported values; this
44 selection is performed at the kernel build tim 44 selection is performed at the kernel build time by setting an
45 appropriate kernel configuration option. 45 appropriate kernel configuration option.
46 46
47 Each physical memory page can be mapped as one 47 Each physical memory page can be mapped as one or more virtual
48 pages. These mappings are described by page ta 48 pages. These mappings are described by page tables that allow
49 translation from a virtual address used by pro 49 translation from a virtual address used by programs to the physical
50 memory address. The page tables are organized 50 memory address. The page tables are organized hierarchically.
51 51
52 The tables at the lowest level of the hierarch 52 The tables at the lowest level of the hierarchy contain physical
53 addresses of actual pages used by the software 53 addresses of actual pages used by the software. The tables at higher
54 levels contain physical addresses of the pages 54 levels contain physical addresses of the pages belonging to the lower
55 levels. The pointer to the top level page tabl 55 levels. The pointer to the top level page table resides in a
56 register. When the CPU performs the address tr 56 register. When the CPU performs the address translation, it uses this
57 register to access the top level page table. T 57 register to access the top level page table. The high bits of the
58 virtual address are used to index an entry in 58 virtual address are used to index an entry in the top level page
59 table. That entry is then used to access the n 59 table. That entry is then used to access the next level in the
60 hierarchy with the next bits of the virtual ad 60 hierarchy with the next bits of the virtual address as the index to
61 that level page table. The lowest bits in the 61 that level page table. The lowest bits in the virtual address define
62 the offset inside the actual page. 62 the offset inside the actual page.
63 63
64 Huge Pages 64 Huge Pages
65 ========== 65 ==========
66 66
67 The address translation requires several memor 67 The address translation requires several memory accesses and memory
68 accesses are slow relatively to CPU speed. To 68 accesses are slow relatively to CPU speed. To avoid spending precious
69 processor cycles on the address translation, C 69 processor cycles on the address translation, CPUs maintain a cache of
70 such translations called Translation Lookaside 70 such translations called Translation Lookaside Buffer (or
71 TLB). Usually TLB is pretty scarce resource an 71 TLB). Usually TLB is pretty scarce resource and applications with
72 large memory working set will experience perfo 72 large memory working set will experience performance hit because of
73 TLB misses. 73 TLB misses.
74 74
75 Many modern CPU architectures allow mapping of 75 Many modern CPU architectures allow mapping of the memory pages
76 directly by the higher levels in the page tabl 76 directly by the higher levels in the page table. For instance, on x86,
77 it is possible to map 2M and even 1G pages usi 77 it is possible to map 2M and even 1G pages using entries in the second
78 and the third level page tables. In Linux such 78 and the third level page tables. In Linux such pages are called
79 `huge`. Usage of huge pages significantly redu 79 `huge`. Usage of huge pages significantly reduces pressure on TLB,
80 improves TLB hit-rate and thus improves overal 80 improves TLB hit-rate and thus improves overall system performance.
81 81
82 There are two mechanisms in Linux that enable 82 There are two mechanisms in Linux that enable mapping of the physical
83 memory with the huge pages. The first one is ` 83 memory with the huge pages. The first one is `HugeTLB filesystem`, or
84 hugetlbfs. It is a pseudo filesystem that uses 84 hugetlbfs. It is a pseudo filesystem that uses RAM as its backing
85 store. For the files created in this filesyste 85 store. For the files created in this filesystem the data resides in
86 the memory and mapped using huge pages. The hu 86 the memory and mapped using huge pages. The hugetlbfs is described at
87 Documentation/admin-guide/mm/hugetlbpage.rst. 87 Documentation/admin-guide/mm/hugetlbpage.rst.
88 88
89 Another, more recent, mechanism that enables u 89 Another, more recent, mechanism that enables use of the huge pages is
90 called `Transparent HugePages`, or THP. Unlike 90 called `Transparent HugePages`, or THP. Unlike the hugetlbfs that
91 requires users and/or system administrators to 91 requires users and/or system administrators to configure what parts of
92 the system memory should and can be mapped by 92 the system memory should and can be mapped by the huge pages, THP
93 manages such mappings transparently to the use 93 manages such mappings transparently to the user and hence the
94 name. See Documentation/admin-guide/mm/transhu 94 name. See Documentation/admin-guide/mm/transhuge.rst for more details
95 about THP. 95 about THP.
96 96
97 Zones 97 Zones
98 ===== 98 =====
99 99
100 Often hardware poses restrictions on how diffe 100 Often hardware poses restrictions on how different physical memory
101 ranges can be accessed. In some cases, devices 101 ranges can be accessed. In some cases, devices cannot perform DMA to
102 all the addressable memory. In other cases, th 102 all the addressable memory. In other cases, the size of the physical
103 memory exceeds the maximal addressable size of 103 memory exceeds the maximal addressable size of virtual memory and
104 special actions are required to access portion 104 special actions are required to access portions of the memory. Linux
105 groups memory pages into `zones` according to 105 groups memory pages into `zones` according to their possible
106 usage. For example, ZONE_DMA will contain memo 106 usage. For example, ZONE_DMA will contain memory that can be used by
107 devices for DMA, ZONE_HIGHMEM will contain mem 107 devices for DMA, ZONE_HIGHMEM will contain memory that is not
108 permanently mapped into kernel's address space 108 permanently mapped into kernel's address space and ZONE_NORMAL will
109 contain normally addressed pages. 109 contain normally addressed pages.
110 110
111 The actual layout of the memory zones is hardw 111 The actual layout of the memory zones is hardware dependent as not all
112 architectures define all zones, and requiremen 112 architectures define all zones, and requirements for DMA are different
113 for different platforms. 113 for different platforms.
114 114
115 Nodes 115 Nodes
116 ===== 116 =====
117 117
118 Many multi-processor machines are NUMA - Non-U 118 Many multi-processor machines are NUMA - Non-Uniform Memory Access -
119 systems. In such systems the memory is arrange 119 systems. In such systems the memory is arranged into banks that have
120 different access latency depending on the "dis 120 different access latency depending on the "distance" from the
121 processor. Each bank is referred to as a `node 121 processor. Each bank is referred to as a `node` and for each node Linux
122 constructs an independent memory management su 122 constructs an independent memory management subsystem. A node has its
123 own set of zones, lists of free and used pages 123 own set of zones, lists of free and used pages and various statistics
124 counters. You can find more details about NUMA 124 counters. You can find more details about NUMA in
125 Documentation/mm/numa.rst` and in 125 Documentation/mm/numa.rst` and in
126 Documentation/admin-guide/mm/numa_memory_polic 126 Documentation/admin-guide/mm/numa_memory_policy.rst.
127 127
128 Page cache 128 Page cache
129 ========== 129 ==========
130 130
131 The physical memory is volatile and the common 131 The physical memory is volatile and the common case for getting data
132 into the memory is to read it from files. When 132 into the memory is to read it from files. Whenever a file is read, the
133 data is put into the `page cache` to avoid exp 133 data is put into the `page cache` to avoid expensive disk access on
134 the subsequent reads. Similarly, when one writ 134 the subsequent reads. Similarly, when one writes to a file, the data
135 is placed in the page cache and eventually get 135 is placed in the page cache and eventually gets into the backing
136 storage device. The written pages are marked a 136 storage device. The written pages are marked as `dirty` and when Linux
137 decides to reuse them for other purposes, it m 137 decides to reuse them for other purposes, it makes sure to synchronize
138 the file contents on the device with the updat 138 the file contents on the device with the updated data.
139 139
140 Anonymous Memory 140 Anonymous Memory
141 ================ 141 ================
142 142
143 The `anonymous memory` or `anonymous mappings` 143 The `anonymous memory` or `anonymous mappings` represent memory that
144 is not backed by a filesystem. Such mappings a 144 is not backed by a filesystem. Such mappings are implicitly created
145 for program's stack and heap or by explicit ca 145 for program's stack and heap or by explicit calls to mmap(2) system
146 call. Usually, the anonymous mappings only def 146 call. Usually, the anonymous mappings only define virtual memory areas
147 that the program is allowed to access. The rea 147 that the program is allowed to access. The read accesses will result
148 in creation of a page table entry that referen 148 in creation of a page table entry that references a special physical
149 page filled with zeroes. When the program perf 149 page filled with zeroes. When the program performs a write, a regular
150 physical page will be allocated to hold the wr 150 physical page will be allocated to hold the written data. The page
151 will be marked dirty and if the kernel decides 151 will be marked dirty and if the kernel decides to repurpose it,
152 the dirty page will be swapped out. 152 the dirty page will be swapped out.
153 153
154 Reclaim 154 Reclaim
155 ======= 155 =======
156 156
157 Throughout the system lifetime, a physical pag 157 Throughout the system lifetime, a physical page can be used for storing
158 different types of data. It can be kernel inte 158 different types of data. It can be kernel internal data structures,
159 DMA'able buffers for device drivers use, data 159 DMA'able buffers for device drivers use, data read from a filesystem,
160 memory allocated by user space processes etc. 160 memory allocated by user space processes etc.
161 161
162 Depending on the page usage it is treated diff 162 Depending on the page usage it is treated differently by the Linux
163 memory management. The pages that can be freed 163 memory management. The pages that can be freed at any time, either
164 because they cache the data available elsewher 164 because they cache the data available elsewhere, for instance, on a
165 hard disk, or because they can be swapped out, 165 hard disk, or because they can be swapped out, again, to the hard
166 disk, are called `reclaimable`. The most notab 166 disk, are called `reclaimable`. The most notable categories of the
167 reclaimable pages are page cache and anonymous 167 reclaimable pages are page cache and anonymous memory.
168 168
169 In most cases, the pages holding internal kern 169 In most cases, the pages holding internal kernel data and used as DMA
170 buffers cannot be repurposed, and they remain 170 buffers cannot be repurposed, and they remain pinned until freed by
171 their user. Such pages are called `unreclaimab 171 their user. Such pages are called `unreclaimable`. However, in certain
172 circumstances, even pages occupied with kernel 172 circumstances, even pages occupied with kernel data structures can be
173 reclaimed. For instance, in-memory caches of f 173 reclaimed. For instance, in-memory caches of filesystem metadata can
174 be re-read from the storage device and therefo 174 be re-read from the storage device and therefore it is possible to
175 discard them from the main memory when system 175 discard them from the main memory when system is under memory
176 pressure. 176 pressure.
177 177
178 The process of freeing the reclaimable physica 178 The process of freeing the reclaimable physical memory pages and
179 repurposing them is called (surprise!) `reclai 179 repurposing them is called (surprise!) `reclaim`. Linux can reclaim
180 pages either asynchronously or synchronously, 180 pages either asynchronously or synchronously, depending on the state
181 of the system. When the system is not loaded, 181 of the system. When the system is not loaded, most of the memory is free
182 and allocation requests will be satisfied imme 182 and allocation requests will be satisfied immediately from the free
183 pages supply. As the load increases, the amoun 183 pages supply. As the load increases, the amount of the free pages goes
184 down and when it reaches a certain threshold ( 184 down and when it reaches a certain threshold (low watermark), an
185 allocation request will awaken the ``kswapd`` 185 allocation request will awaken the ``kswapd`` daemon. It will
186 asynchronously scan memory pages and either ju 186 asynchronously scan memory pages and either just free them if the data
187 they contain is available elsewhere, or evict 187 they contain is available elsewhere, or evict to the backing storage
188 device (remember those dirty pages?). As memor 188 device (remember those dirty pages?). As memory usage increases even
189 more and reaches another threshold - min water 189 more and reaches another threshold - min watermark - an allocation
190 will trigger `direct reclaim`. In this case al 190 will trigger `direct reclaim`. In this case allocation is stalled
191 until enough memory pages are reclaimed to sat 191 until enough memory pages are reclaimed to satisfy the request.
192 192
193 Compaction 193 Compaction
194 ========== 194 ==========
195 195
196 As the system runs, tasks allocate and free th 196 As the system runs, tasks allocate and free the memory and it becomes
197 fragmented. Although with virtual memory it is 197 fragmented. Although with virtual memory it is possible to present
198 scattered physical pages as virtually contiguo 198 scattered physical pages as virtually contiguous range, sometimes it is
199 necessary to allocate large physically contigu 199 necessary to allocate large physically contiguous memory areas. Such
200 need may arise, for instance, when a device dr 200 need may arise, for instance, when a device driver requires a large
201 buffer for DMA, or when THP allocates a huge p 201 buffer for DMA, or when THP allocates a huge page. Memory `compaction`
202 addresses the fragmentation issue. This mechan 202 addresses the fragmentation issue. This mechanism moves occupied pages
203 from the lower part of a memory zone to free p 203 from the lower part of a memory zone to free pages in the upper part
204 of the zone. When a compaction scan is finishe 204 of the zone. When a compaction scan is finished free pages are grouped
205 together at the beginning of the zone and allo 205 together at the beginning of the zone and allocations of large
206 physically contiguous areas become possible. 206 physically contiguous areas become possible.
207 207
208 Like reclaim, the compaction may happen asynch 208 Like reclaim, the compaction may happen asynchronously in the ``kcompactd``
209 daemon or synchronously as a result of a memor 209 daemon or synchronously as a result of a memory allocation request.
210 210
211 OOM killer 211 OOM killer
212 ========== 212 ==========
213 213
214 It is possible that on a loaded machine memory 214 It is possible that on a loaded machine memory will be exhausted and the
215 kernel will be unable to reclaim enough memory 215 kernel will be unable to reclaim enough memory to continue to operate. In
216 order to save the rest of the system, it invok 216 order to save the rest of the system, it invokes the `OOM killer`.
217 217
218 The `OOM killer` selects a task to sacrifice f 218 The `OOM killer` selects a task to sacrifice for the sake of the overall
219 system health. The selected task is killed in 219 system health. The selected task is killed in a hope that after it exits
220 enough memory will be freed to continue normal 220 enough memory will be freed to continue normal operation.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.