1 # SPDX-License-Identifier: GPL-2.0-only 1 # SPDX-License-Identifier: GPL-2.0-only
2 2
3 menu "Memory Management options" 3 menu "Memory Management options"
4 4
5 # 5 #
6 # For some reason microblaze and nios2 hard co 6 # For some reason microblaze and nios2 hard code SWAP=n. Hopefully we can
7 # add proper SWAP support to them, in which ca 7 # add proper SWAP support to them, in which case this can be remove.
8 # 8 #
9 config ARCH_NO_SWAP 9 config ARCH_NO_SWAP
10 bool 10 bool
11 11
12 config ZPOOL 12 config ZPOOL
13 bool 13 bool
14 14
15 menuconfig SWAP 15 menuconfig SWAP
16 bool "Support for paging of anonymous 16 bool "Support for paging of anonymous memory (swap)"
17 depends on MMU && BLOCK && !ARCH_NO_SW 17 depends on MMU && BLOCK && !ARCH_NO_SWAP
18 default y 18 default y
19 help 19 help
20 This option allows you to choose whe 20 This option allows you to choose whether you want to have support
21 for so called swap devices or swap f 21 for so called swap devices or swap files in your kernel that are
22 used to provide more virtual memory 22 used to provide more virtual memory than the actual RAM present
23 in your computer. If unsure say Y. 23 in your computer. If unsure say Y.
24 24
25 config ZSWAP 25 config ZSWAP
26 bool "Compressed cache for swap pages" 26 bool "Compressed cache for swap pages"
27 depends on SWAP 27 depends on SWAP
28 select CRYPTO 28 select CRYPTO
29 select ZPOOL 29 select ZPOOL
30 help 30 help
31 A lightweight compressed cache for s 31 A lightweight compressed cache for swap pages. It takes
32 pages that are in the process of bei 32 pages that are in the process of being swapped out and attempts to
33 compress them into a dynamically all 33 compress them into a dynamically allocated RAM-based memory pool.
34 This can result in a significant I/O 34 This can result in a significant I/O reduction on swap device and,
35 in the case where decompressing from 35 in the case where decompressing from RAM is faster than swap device
36 reads, can also improve workload per 36 reads, can also improve workload performance.
37 37
38 config ZSWAP_DEFAULT_ON 38 config ZSWAP_DEFAULT_ON
39 bool "Enable the compressed cache for 39 bool "Enable the compressed cache for swap pages by default"
40 depends on ZSWAP 40 depends on ZSWAP
41 help 41 help
42 If selected, the compressed cache fo 42 If selected, the compressed cache for swap pages will be enabled
43 at boot, otherwise it will be disabl 43 at boot, otherwise it will be disabled.
44 44
45 The selection made here can be overr 45 The selection made here can be overridden by using the kernel
46 command line 'zswap.enabled=' option 46 command line 'zswap.enabled=' option.
47 47
48 config ZSWAP_SHRINKER_DEFAULT_ON 48 config ZSWAP_SHRINKER_DEFAULT_ON
49 bool "Shrink the zswap pool on memory 49 bool "Shrink the zswap pool on memory pressure"
50 depends on ZSWAP 50 depends on ZSWAP
51 default n 51 default n
52 help 52 help
53 If selected, the zswap shrinker will 53 If selected, the zswap shrinker will be enabled, and the pages
54 stored in the zswap pool will become 54 stored in the zswap pool will become available for reclaim (i.e
55 written back to the backing swap dev 55 written back to the backing swap device) on memory pressure.
56 56
57 This means that zswap writeback coul 57 This means that zswap writeback could happen even if the pool is
58 not yet full, or the cgroup zswap li 58 not yet full, or the cgroup zswap limit has not been reached,
59 reducing the chance that cold pages 59 reducing the chance that cold pages will reside in the zswap pool
60 and consume memory indefinitely. 60 and consume memory indefinitely.
61 61
62 choice 62 choice
63 prompt "Default compressor" 63 prompt "Default compressor"
64 depends on ZSWAP 64 depends on ZSWAP
65 default ZSWAP_COMPRESSOR_DEFAULT_LZO 65 default ZSWAP_COMPRESSOR_DEFAULT_LZO
66 help 66 help
67 Selects the default compression algo 67 Selects the default compression algorithm for the compressed cache
68 for swap pages. 68 for swap pages.
69 69
70 For an overview what kind of perform 70 For an overview what kind of performance can be expected from
71 a particular compression algorithm p 71 a particular compression algorithm please refer to the benchmarks
72 available at the following LWN page: 72 available at the following LWN page:
73 https://lwn.net/Articles/751795/ 73 https://lwn.net/Articles/751795/
74 74
75 If in doubt, select 'LZO'. 75 If in doubt, select 'LZO'.
76 76
77 The selection made here can be overr 77 The selection made here can be overridden by using the kernel
78 command line 'zswap.compressor=' opt 78 command line 'zswap.compressor=' option.
79 79
80 config ZSWAP_COMPRESSOR_DEFAULT_DEFLATE 80 config ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
81 bool "Deflate" 81 bool "Deflate"
82 select CRYPTO_DEFLATE 82 select CRYPTO_DEFLATE
83 help 83 help
84 Use the Deflate algorithm as the def 84 Use the Deflate algorithm as the default compression algorithm.
85 85
86 config ZSWAP_COMPRESSOR_DEFAULT_LZO 86 config ZSWAP_COMPRESSOR_DEFAULT_LZO
87 bool "LZO" 87 bool "LZO"
88 select CRYPTO_LZO 88 select CRYPTO_LZO
89 help 89 help
90 Use the LZO algorithm as the default 90 Use the LZO algorithm as the default compression algorithm.
91 91
92 config ZSWAP_COMPRESSOR_DEFAULT_842 92 config ZSWAP_COMPRESSOR_DEFAULT_842
93 bool "842" 93 bool "842"
94 select CRYPTO_842 94 select CRYPTO_842
95 help 95 help
96 Use the 842 algorithm as the default 96 Use the 842 algorithm as the default compression algorithm.
97 97
98 config ZSWAP_COMPRESSOR_DEFAULT_LZ4 98 config ZSWAP_COMPRESSOR_DEFAULT_LZ4
99 bool "LZ4" 99 bool "LZ4"
100 select CRYPTO_LZ4 100 select CRYPTO_LZ4
101 help 101 help
102 Use the LZ4 algorithm as the default 102 Use the LZ4 algorithm as the default compression algorithm.
103 103
104 config ZSWAP_COMPRESSOR_DEFAULT_LZ4HC 104 config ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
105 bool "LZ4HC" 105 bool "LZ4HC"
106 select CRYPTO_LZ4HC 106 select CRYPTO_LZ4HC
107 help 107 help
108 Use the LZ4HC algorithm as the defau 108 Use the LZ4HC algorithm as the default compression algorithm.
109 109
110 config ZSWAP_COMPRESSOR_DEFAULT_ZSTD 110 config ZSWAP_COMPRESSOR_DEFAULT_ZSTD
111 bool "zstd" 111 bool "zstd"
112 select CRYPTO_ZSTD 112 select CRYPTO_ZSTD
113 help 113 help
114 Use the zstd algorithm as the defaul 114 Use the zstd algorithm as the default compression algorithm.
115 endchoice 115 endchoice
116 116
117 config ZSWAP_COMPRESSOR_DEFAULT 117 config ZSWAP_COMPRESSOR_DEFAULT
118 string 118 string
119 depends on ZSWAP 119 depends on ZSWAP
120 default "deflate" if ZSWAP_COMPRESSOR_D 120 default "deflate" if ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
121 default "lzo" if ZSWAP_COMPRESSOR_DEFAU 121 default "lzo" if ZSWAP_COMPRESSOR_DEFAULT_LZO
122 default "842" if ZSWAP_COMPRESSOR_DEFAU 122 default "842" if ZSWAP_COMPRESSOR_DEFAULT_842
123 default "lz4" if ZSWAP_COMPRESSOR_DEFAU 123 default "lz4" if ZSWAP_COMPRESSOR_DEFAULT_LZ4
124 default "lz4hc" if ZSWAP_COMPRESSOR_DEF 124 default "lz4hc" if ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
125 default "zstd" if ZSWAP_COMPRESSOR_DEFA 125 default "zstd" if ZSWAP_COMPRESSOR_DEFAULT_ZSTD
126 default "" 126 default ""
127 127
128 choice 128 choice
129 prompt "Default allocator" 129 prompt "Default allocator"
130 depends on ZSWAP 130 depends on ZSWAP
131 default ZSWAP_ZPOOL_DEFAULT_ZSMALLOC i !! 131 default ZSWAP_ZPOOL_DEFAULT_ZSMALLOC if HAVE_ZSMALLOC
132 default ZSWAP_ZPOOL_DEFAULT_ZBUD 132 default ZSWAP_ZPOOL_DEFAULT_ZBUD
133 help 133 help
134 Selects the default allocator for th 134 Selects the default allocator for the compressed cache for
135 swap pages. 135 swap pages.
136 The default is 'zbud' for compatibil 136 The default is 'zbud' for compatibility, however please do
137 read the description of each of the 137 read the description of each of the allocators below before
138 making a right choice. 138 making a right choice.
139 139
140 The selection made here can be overr 140 The selection made here can be overridden by using the kernel
141 command line 'zswap.zpool=' option. 141 command line 'zswap.zpool=' option.
142 142
143 config ZSWAP_ZPOOL_DEFAULT_ZBUD 143 config ZSWAP_ZPOOL_DEFAULT_ZBUD
144 bool "zbud" 144 bool "zbud"
145 select ZBUD 145 select ZBUD
146 help 146 help
147 Use the zbud allocator as the defaul 147 Use the zbud allocator as the default allocator.
148 148
149 config ZSWAP_ZPOOL_DEFAULT_Z3FOLD_DEPRECATED 149 config ZSWAP_ZPOOL_DEFAULT_Z3FOLD_DEPRECATED
150 bool "z3foldi (DEPRECATED)" 150 bool "z3foldi (DEPRECATED)"
151 select Z3FOLD_DEPRECATED 151 select Z3FOLD_DEPRECATED
152 help 152 help
153 Use the z3fold allocator as the defa 153 Use the z3fold allocator as the default allocator.
154 154
155 Deprecated and scheduled for removal 155 Deprecated and scheduled for removal in a few cycles,
156 see CONFIG_Z3FOLD_DEPRECATED. 156 see CONFIG_Z3FOLD_DEPRECATED.
157 157
158 config ZSWAP_ZPOOL_DEFAULT_ZSMALLOC 158 config ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
159 bool "zsmalloc" 159 bool "zsmalloc"
>> 160 depends on HAVE_ZSMALLOC
160 select ZSMALLOC 161 select ZSMALLOC
161 help 162 help
162 Use the zsmalloc allocator as the de 163 Use the zsmalloc allocator as the default allocator.
163 endchoice 164 endchoice
164 165
165 config ZSWAP_ZPOOL_DEFAULT 166 config ZSWAP_ZPOOL_DEFAULT
166 string 167 string
167 depends on ZSWAP 168 depends on ZSWAP
168 default "zbud" if ZSWAP_ZPOOL_DEFAULT_Z 169 default "zbud" if ZSWAP_ZPOOL_DEFAULT_ZBUD
169 default "z3fold" if ZSWAP_ZPOOL_DEFAULT 170 default "z3fold" if ZSWAP_ZPOOL_DEFAULT_Z3FOLD_DEPRECATED
170 default "zsmalloc" if ZSWAP_ZPOOL_DEFAU 171 default "zsmalloc" if ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
171 default "" 172 default ""
172 173
173 config ZBUD 174 config ZBUD
174 tristate "2:1 compression allocator (z 175 tristate "2:1 compression allocator (zbud)"
175 depends on ZSWAP 176 depends on ZSWAP
176 help 177 help
177 A special purpose allocator for stor 178 A special purpose allocator for storing compressed pages.
178 It is designed to store up to two co 179 It is designed to store up to two compressed pages per physical
179 page. While this design limits stor 180 page. While this design limits storage density, it has simple and
180 deterministic reclaim properties tha 181 deterministic reclaim properties that make it preferable to a higher
181 density approach when reclaim will b 182 density approach when reclaim will be used.
182 183
183 config Z3FOLD_DEPRECATED 184 config Z3FOLD_DEPRECATED
184 tristate "3:1 compression allocator (z 185 tristate "3:1 compression allocator (z3fold) (DEPRECATED)"
185 depends on ZSWAP 186 depends on ZSWAP
186 help 187 help
187 Deprecated and scheduled for removal 188 Deprecated and scheduled for removal in a few cycles. If you have
188 a good reason for using Z3FOLD over 189 a good reason for using Z3FOLD over ZSMALLOC, please contact
189 linux-mm@kvack.org and the zswap mai 190 linux-mm@kvack.org and the zswap maintainers.
190 191
191 A special purpose allocator for stor 192 A special purpose allocator for storing compressed pages.
192 It is designed to store up to three 193 It is designed to store up to three compressed pages per physical
193 page. It is a ZBUD derivative so the 194 page. It is a ZBUD derivative so the simplicity and determinism are
194 still there. 195 still there.
195 196
196 config Z3FOLD 197 config Z3FOLD
197 tristate 198 tristate
198 default y if Z3FOLD_DEPRECATED=y 199 default y if Z3FOLD_DEPRECATED=y
199 default m if Z3FOLD_DEPRECATED=m 200 default m if Z3FOLD_DEPRECATED=m
200 depends on Z3FOLD_DEPRECATED 201 depends on Z3FOLD_DEPRECATED
201 202
>> 203 config HAVE_ZSMALLOC
>> 204 def_bool y
>> 205 depends on MMU
>> 206 depends on PAGE_SIZE_LESS_THAN_256KB # we want <= 64 KiB
>> 207
202 config ZSMALLOC 208 config ZSMALLOC
203 tristate 209 tristate
204 prompt "N:1 compression allocator (zsm !! 210 prompt "N:1 compression allocator (zsmalloc)" if ZSWAP
205 depends on MMU !! 211 depends on HAVE_ZSMALLOC
206 help 212 help
207 zsmalloc is a slab-based memory allo 213 zsmalloc is a slab-based memory allocator designed to store
208 pages of various compression levels 214 pages of various compression levels efficiently. It achieves
209 the highest storage density with the 215 the highest storage density with the least amount of fragmentation.
210 216
211 config ZSMALLOC_STAT 217 config ZSMALLOC_STAT
212 bool "Export zsmalloc statistics" 218 bool "Export zsmalloc statistics"
213 depends on ZSMALLOC 219 depends on ZSMALLOC
214 select DEBUG_FS 220 select DEBUG_FS
215 help 221 help
216 This option enables code in the zsma 222 This option enables code in the zsmalloc to collect various
217 statistics about what's happening in 223 statistics about what's happening in zsmalloc and exports that
218 information to userspace via debugfs 224 information to userspace via debugfs.
219 If unsure, say N. 225 If unsure, say N.
220 226
221 config ZSMALLOC_CHAIN_SIZE 227 config ZSMALLOC_CHAIN_SIZE
222 int "Maximum number of physical pages 228 int "Maximum number of physical pages per-zspage"
223 default 8 229 default 8
224 range 4 16 230 range 4 16
225 depends on ZSMALLOC 231 depends on ZSMALLOC
226 help 232 help
227 This option sets the upper limit on 233 This option sets the upper limit on the number of physical pages
228 that a zmalloc page (zspage) can con 234 that a zmalloc page (zspage) can consist of. The optimal zspage
229 chain size is calculated for each si 235 chain size is calculated for each size class during the
230 initialization of the pool. 236 initialization of the pool.
231 237
232 Changing this option can alter the c 238 Changing this option can alter the characteristics of size classes,
233 such as the number of pages per zspa 239 such as the number of pages per zspage and the number of objects
234 per zspage. This can also result in 240 per zspage. This can also result in different configurations of
235 the pool, as zsmalloc merges size cl 241 the pool, as zsmalloc merges size classes with similar
236 characteristics. 242 characteristics.
237 243
238 For more information, see zsmalloc d 244 For more information, see zsmalloc documentation.
239 245
240 menu "Slab allocator options" 246 menu "Slab allocator options"
241 247
242 config SLUB 248 config SLUB
243 def_bool y 249 def_bool y
244 250
245 config SLUB_TINY 251 config SLUB_TINY
246 bool "Configure for minimal memory foo 252 bool "Configure for minimal memory footprint"
247 depends on EXPERT 253 depends on EXPERT
248 select SLAB_MERGE_DEFAULT 254 select SLAB_MERGE_DEFAULT
249 help 255 help
250 Configures the slab allocator in a 256 Configures the slab allocator in a way to achieve minimal memory
251 footprint, sacrificing scalability, 257 footprint, sacrificing scalability, debugging and other features.
252 This is intended only for the small 258 This is intended only for the smallest system that had used the
253 SLOB allocator and is not recommend 259 SLOB allocator and is not recommended for systems with more than
254 16MB RAM. 260 16MB RAM.
255 261
256 If unsure, say N. 262 If unsure, say N.
257 263
258 config SLAB_MERGE_DEFAULT 264 config SLAB_MERGE_DEFAULT
259 bool "Allow slab caches to be merged" 265 bool "Allow slab caches to be merged"
260 default y 266 default y
261 help 267 help
262 For reduced kernel memory fragmentat 268 For reduced kernel memory fragmentation, slab caches can be
263 merged when they share the same size 269 merged when they share the same size and other characteristics.
264 This carries a risk of kernel heap o 270 This carries a risk of kernel heap overflows being able to
265 overwrite objects from merged caches 271 overwrite objects from merged caches (and more easily control
266 cache layout), which makes such heap 272 cache layout), which makes such heap attacks easier to exploit
267 by attackers. By keeping caches unme 273 by attackers. By keeping caches unmerged, these kinds of exploits
268 can usually only damage objects in t 274 can usually only damage objects in the same cache. To disable
269 merging at runtime, "slab_nomerge" c 275 merging at runtime, "slab_nomerge" can be passed on the kernel
270 command line. 276 command line.
271 277
272 config SLAB_FREELIST_RANDOM 278 config SLAB_FREELIST_RANDOM
273 bool "Randomize slab freelist" 279 bool "Randomize slab freelist"
274 depends on !SLUB_TINY 280 depends on !SLUB_TINY
275 help 281 help
276 Randomizes the freelist order used o 282 Randomizes the freelist order used on creating new pages. This
277 security feature reduces the predict 283 security feature reduces the predictability of the kernel slab
278 allocator against heap overflows. 284 allocator against heap overflows.
279 285
280 config SLAB_FREELIST_HARDENED 286 config SLAB_FREELIST_HARDENED
281 bool "Harden slab freelist metadata" 287 bool "Harden slab freelist metadata"
282 depends on !SLUB_TINY 288 depends on !SLUB_TINY
283 help 289 help
284 Many kernel heap attacks try to targ 290 Many kernel heap attacks try to target slab cache metadata and
285 other infrastructure. This options m 291 other infrastructure. This options makes minor performance
286 sacrifices to harden the kernel slab 292 sacrifices to harden the kernel slab allocator against common
287 freelist exploit methods. 293 freelist exploit methods.
288 294
289 config SLAB_BUCKETS 295 config SLAB_BUCKETS
290 bool "Support allocation from separate 296 bool "Support allocation from separate kmalloc buckets"
291 depends on !SLUB_TINY 297 depends on !SLUB_TINY
292 default SLAB_FREELIST_HARDENED 298 default SLAB_FREELIST_HARDENED
293 help 299 help
294 Kernel heap attacks frequently depen 300 Kernel heap attacks frequently depend on being able to create
295 specifically-sized allocations with 301 specifically-sized allocations with user-controlled contents
296 that will be allocated into the same 302 that will be allocated into the same kmalloc bucket as a
297 target object. To avoid sharing thes 303 target object. To avoid sharing these allocation buckets,
298 provide an explicitly separated set 304 provide an explicitly separated set of buckets to be used for
299 user-controlled allocations. This ma 305 user-controlled allocations. This may very slightly increase
300 memory fragmentation, though in prac 306 memory fragmentation, though in practice it's only a handful
301 of extra pages since the bulk of use 307 of extra pages since the bulk of user-controlled allocations
302 are relatively long-lived. 308 are relatively long-lived.
303 309
304 If unsure, say Y. 310 If unsure, say Y.
305 311
306 config SLUB_STATS 312 config SLUB_STATS
307 default n 313 default n
308 bool "Enable performance statistics" 314 bool "Enable performance statistics"
309 depends on SYSFS && !SLUB_TINY 315 depends on SYSFS && !SLUB_TINY
310 help 316 help
311 The statistics are useful to debug s 317 The statistics are useful to debug slab allocation behavior in
312 order find ways to optimize the allo 318 order find ways to optimize the allocator. This should never be
313 enabled for production use since kee 319 enabled for production use since keeping statistics slows down
314 the allocator by a few percentage po 320 the allocator by a few percentage points. The slabinfo command
315 supports the determination of the mo 321 supports the determination of the most active slabs to figure
316 out which slabs are relevant to a pa 322 out which slabs are relevant to a particular load.
317 Try running: slabinfo -DA 323 Try running: slabinfo -DA
318 324
319 config SLUB_CPU_PARTIAL 325 config SLUB_CPU_PARTIAL
320 default y 326 default y
321 depends on SMP && !SLUB_TINY 327 depends on SMP && !SLUB_TINY
322 bool "Enable per cpu partial caches" 328 bool "Enable per cpu partial caches"
323 help 329 help
324 Per cpu partial caches accelerate ob 330 Per cpu partial caches accelerate objects allocation and freeing
325 that is local to a processor at the 331 that is local to a processor at the price of more indeterminism
326 in the latency of the free. On overf 332 in the latency of the free. On overflow these caches will be cleared
327 which requires the taking of locks t 333 which requires the taking of locks that may cause latency spikes.
328 Typically one would choose no for a 334 Typically one would choose no for a realtime system.
329 335
330 config RANDOM_KMALLOC_CACHES 336 config RANDOM_KMALLOC_CACHES
331 default n 337 default n
332 depends on !SLUB_TINY 338 depends on !SLUB_TINY
333 bool "Randomize slab caches for normal 339 bool "Randomize slab caches for normal kmalloc"
334 help 340 help
335 A hardening feature that creates mul 341 A hardening feature that creates multiple copies of slab caches for
336 normal kmalloc allocation and makes 342 normal kmalloc allocation and makes kmalloc randomly pick one based
337 on code address, which makes the att 343 on code address, which makes the attackers more difficult to spray
338 vulnerable memory objects on the hea 344 vulnerable memory objects on the heap for the purpose of exploiting
339 memory vulnerabilities. 345 memory vulnerabilities.
340 346
341 Currently the number of copies is se 347 Currently the number of copies is set to 16, a reasonably large value
342 that effectively diverges the memory 348 that effectively diverges the memory objects allocated for different
343 subsystems or modules into different 349 subsystems or modules into different caches, at the expense of a
344 limited degree of memory and CPU ove 350 limited degree of memory and CPU overhead that relates to hardware and
345 system workload. 351 system workload.
346 352
347 endmenu # Slab allocator options 353 endmenu # Slab allocator options
348 354
349 config SHUFFLE_PAGE_ALLOCATOR 355 config SHUFFLE_PAGE_ALLOCATOR
350 bool "Page allocator randomization" 356 bool "Page allocator randomization"
351 default SLAB_FREELIST_RANDOM && ACPI_N 357 default SLAB_FREELIST_RANDOM && ACPI_NUMA
352 help 358 help
353 Randomization of the page allocator 359 Randomization of the page allocator improves the average
354 utilization of a direct-mapped memor 360 utilization of a direct-mapped memory-side-cache. See section
355 5.2.27 Heterogeneous Memory Attribut 361 5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI
356 6.2a specification for an example of 362 6.2a specification for an example of how a platform advertises
357 the presence of a memory-side-cache. 363 the presence of a memory-side-cache. There are also incidental
358 security benefits as it reduces the 364 security benefits as it reduces the predictability of page
359 allocations to compliment SLAB_FREEL 365 allocations to compliment SLAB_FREELIST_RANDOM, but the
360 default granularity of shuffling on 366 default granularity of shuffling on the MAX_PAGE_ORDER i.e, 10th
361 order of pages is selected based on 367 order of pages is selected based on cache utilization benefits
362 on x86. 368 on x86.
363 369
364 While the randomization improves cac 370 While the randomization improves cache utilization it may
365 negatively impact workloads on platf 371 negatively impact workloads on platforms without a cache. For
366 this reason, by default, the randomi 372 this reason, by default, the randomization is not enabled even
367 if SHUFFLE_PAGE_ALLOCATOR=y. The ran 373 if SHUFFLE_PAGE_ALLOCATOR=y. The randomization may be force enabled
368 with the 'page_alloc.shuffle' kernel 374 with the 'page_alloc.shuffle' kernel command line parameter.
369 375
370 Say Y if unsure. 376 Say Y if unsure.
371 377
372 config COMPAT_BRK 378 config COMPAT_BRK
373 bool "Disable heap randomization" 379 bool "Disable heap randomization"
374 default y 380 default y
375 help 381 help
376 Randomizing heap placement makes hea 382 Randomizing heap placement makes heap exploits harder, but it
377 also breaks ancient binaries (includ 383 also breaks ancient binaries (including anything libc5 based).
378 This option changes the bootup defau 384 This option changes the bootup default to heap randomization
379 disabled, and can be overridden at r 385 disabled, and can be overridden at runtime by setting
380 /proc/sys/kernel/randomize_va_space 386 /proc/sys/kernel/randomize_va_space to 2.
381 387
382 On non-ancient distros (post-2000 on 388 On non-ancient distros (post-2000 ones) N is usually a safe choice.
383 389
384 config MMAP_ALLOW_UNINITIALIZED 390 config MMAP_ALLOW_UNINITIALIZED
385 bool "Allow mmapped anonymous memory t 391 bool "Allow mmapped anonymous memory to be uninitialized"
386 depends on EXPERT && !MMU 392 depends on EXPERT && !MMU
387 default n 393 default n
388 help 394 help
389 Normally, and according to the Linux 395 Normally, and according to the Linux spec, anonymous memory obtained
390 from mmap() has its contents cleared 396 from mmap() has its contents cleared before it is passed to
391 userspace. Enabling this config opt 397 userspace. Enabling this config option allows you to request that
392 mmap() skip that if it is given an M 398 mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus
393 providing a huge performance boost. 399 providing a huge performance boost. If this option is not enabled,
394 then the flag will be ignored. 400 then the flag will be ignored.
395 401
396 This is taken advantage of by uClibc 402 This is taken advantage of by uClibc's malloc(), and also by
397 ELF-FDPIC binfmt's brk and stack all 403 ELF-FDPIC binfmt's brk and stack allocator.
398 404
399 Because of the obvious security issu 405 Because of the obvious security issues, this option should only be
400 enabled on embedded devices where yo 406 enabled on embedded devices where you control what is run in
401 userspace. Since that isn't general 407 userspace. Since that isn't generally a problem on no-MMU systems,
402 it is normally safe to say Y here. 408 it is normally safe to say Y here.
403 409
404 See Documentation/admin-guide/mm/nom 410 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
405 411
406 config SELECT_MEMORY_MODEL 412 config SELECT_MEMORY_MODEL
407 def_bool y 413 def_bool y
408 depends on ARCH_SELECT_MEMORY_MODEL 414 depends on ARCH_SELECT_MEMORY_MODEL
409 415
410 choice 416 choice
411 prompt "Memory model" 417 prompt "Memory model"
412 depends on SELECT_MEMORY_MODEL 418 depends on SELECT_MEMORY_MODEL
413 default SPARSEMEM_MANUAL if ARCH_SPARS 419 default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT
414 default FLATMEM_MANUAL 420 default FLATMEM_MANUAL
415 help 421 help
416 This option allows you to change som 422 This option allows you to change some of the ways that
417 Linux manages its memory internally. 423 Linux manages its memory internally. Most users will
418 only have one option here selected b 424 only have one option here selected by the architecture
419 configuration. This is normal. 425 configuration. This is normal.
420 426
421 config FLATMEM_MANUAL 427 config FLATMEM_MANUAL
422 bool "Flat Memory" 428 bool "Flat Memory"
423 depends on !ARCH_SPARSEMEM_ENABLE || A 429 depends on !ARCH_SPARSEMEM_ENABLE || ARCH_FLATMEM_ENABLE
424 help 430 help
425 This option is best suited for non-N 431 This option is best suited for non-NUMA systems with
426 flat address space. The FLATMEM is t 432 flat address space. The FLATMEM is the most efficient
427 system in terms of performance and r 433 system in terms of performance and resource consumption
428 and it is the best option for smalle 434 and it is the best option for smaller systems.
429 435
430 For systems that have holes in their 436 For systems that have holes in their physical address
431 spaces and for features like NUMA an 437 spaces and for features like NUMA and memory hotplug,
432 choose "Sparse Memory". 438 choose "Sparse Memory".
433 439
434 If unsure, choose this option (Flat 440 If unsure, choose this option (Flat Memory) over any other.
435 441
436 config SPARSEMEM_MANUAL 442 config SPARSEMEM_MANUAL
437 bool "Sparse Memory" 443 bool "Sparse Memory"
438 depends on ARCH_SPARSEMEM_ENABLE 444 depends on ARCH_SPARSEMEM_ENABLE
439 help 445 help
440 This will be the only option for som 446 This will be the only option for some systems, including
441 memory hot-plug systems. This is no 447 memory hot-plug systems. This is normal.
442 448
443 This option provides efficient suppo 449 This option provides efficient support for systems with
444 holes is their physical address spac 450 holes is their physical address space and allows memory
445 hot-plug and hot-remove. 451 hot-plug and hot-remove.
446 452
447 If unsure, choose "Flat Memory" over 453 If unsure, choose "Flat Memory" over this option.
448 454
449 endchoice 455 endchoice
450 456
451 config SPARSEMEM 457 config SPARSEMEM
452 def_bool y 458 def_bool y
453 depends on (!SELECT_MEMORY_MODEL && AR 459 depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL
454 460
455 config FLATMEM 461 config FLATMEM
456 def_bool y 462 def_bool y
457 depends on !SPARSEMEM || FLATMEM_MANUA 463 depends on !SPARSEMEM || FLATMEM_MANUAL
458 464
459 # 465 #
460 # SPARSEMEM_EXTREME (which is the default) doe 466 # SPARSEMEM_EXTREME (which is the default) does some bootmem
461 # allocations when sparse_init() is called. I 467 # allocations when sparse_init() is called. If this cannot
462 # be done on your architecture, select this op 468 # be done on your architecture, select this option. However,
463 # statically allocating the mem_section[] arra 469 # statically allocating the mem_section[] array can potentially
464 # consume vast quantities of .bss, so be caref 470 # consume vast quantities of .bss, so be careful.
465 # 471 #
466 # This option will also potentially produce sm 472 # This option will also potentially produce smaller runtime code
467 # with gcc 3.4 and later. 473 # with gcc 3.4 and later.
468 # 474 #
469 config SPARSEMEM_STATIC 475 config SPARSEMEM_STATIC
470 bool 476 bool
471 477
472 # 478 #
473 # Architecture platforms which require a two l 479 # Architecture platforms which require a two level mem_section in SPARSEMEM
474 # must select this option. This is usually for 480 # must select this option. This is usually for architecture platforms with
475 # an extremely sparse physical address space. 481 # an extremely sparse physical address space.
476 # 482 #
477 config SPARSEMEM_EXTREME 483 config SPARSEMEM_EXTREME
478 def_bool y 484 def_bool y
479 depends on SPARSEMEM && !SPARSEMEM_STA 485 depends on SPARSEMEM && !SPARSEMEM_STATIC
480 486
481 config SPARSEMEM_VMEMMAP_ENABLE 487 config SPARSEMEM_VMEMMAP_ENABLE
482 bool 488 bool
483 489
484 config SPARSEMEM_VMEMMAP 490 config SPARSEMEM_VMEMMAP
485 bool "Sparse Memory virtual memmap" 491 bool "Sparse Memory virtual memmap"
486 depends on SPARSEMEM && SPARSEMEM_VMEM 492 depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE
487 default y 493 default y
488 help 494 help
489 SPARSEMEM_VMEMMAP uses a virtually m 495 SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise
490 pfn_to_page and page_to_pfn operatio 496 pfn_to_page and page_to_pfn operations. This is the most
491 efficient option when sufficient ker 497 efficient option when sufficient kernel resources are available.
492 # 498 #
493 # Select this config option from the architect 499 # Select this config option from the architecture Kconfig, if it is preferred
494 # to enable the feature of HugeTLB/dev_dax vme 500 # to enable the feature of HugeTLB/dev_dax vmemmap optimization.
495 # 501 #
496 config ARCH_WANT_OPTIMIZE_DAX_VMEMMAP 502 config ARCH_WANT_OPTIMIZE_DAX_VMEMMAP
497 bool 503 bool
498 504
499 config ARCH_WANT_OPTIMIZE_HUGETLB_VMEMMAP 505 config ARCH_WANT_OPTIMIZE_HUGETLB_VMEMMAP
500 bool 506 bool
501 507
502 config HAVE_MEMBLOCK_PHYS_MAP 508 config HAVE_MEMBLOCK_PHYS_MAP
503 bool 509 bool
504 510
505 config HAVE_GUP_FAST 511 config HAVE_GUP_FAST
506 depends on MMU 512 depends on MMU
507 bool 513 bool
508 514
509 # Don't discard allocated memory used to track 515 # Don't discard allocated memory used to track "memory" and "reserved" memblocks
510 # after early boot, so it can still be used to 516 # after early boot, so it can still be used to test for validity of memory.
511 # Also, memblocks are updated with memory hot( 517 # Also, memblocks are updated with memory hot(un)plug.
512 config ARCH_KEEP_MEMBLOCK 518 config ARCH_KEEP_MEMBLOCK
513 bool 519 bool
514 520
515 # Keep arch NUMA mapping infrastructure post-i 521 # Keep arch NUMA mapping infrastructure post-init.
516 config NUMA_KEEP_MEMINFO 522 config NUMA_KEEP_MEMINFO
517 bool 523 bool
518 524
519 config MEMORY_ISOLATION 525 config MEMORY_ISOLATION
520 bool 526 bool
521 527
522 # IORESOURCE_SYSTEM_RAM regions in the kernel 528 # IORESOURCE_SYSTEM_RAM regions in the kernel resource tree that are marked
523 # IORESOURCE_EXCLUSIVE cannot be mapped to use 529 # IORESOURCE_EXCLUSIVE cannot be mapped to user space, for example, via
524 # /dev/mem. 530 # /dev/mem.
525 config EXCLUSIVE_SYSTEM_RAM 531 config EXCLUSIVE_SYSTEM_RAM
526 def_bool y 532 def_bool y
527 depends on !DEVMEM || STRICT_DEVMEM 533 depends on !DEVMEM || STRICT_DEVMEM
528 534
529 # 535 #
530 # Only be set on architectures that have compl 536 # Only be set on architectures that have completely implemented memory hotplug
531 # feature. If you are not sure, don't touch it 537 # feature. If you are not sure, don't touch it.
532 # 538 #
533 config HAVE_BOOTMEM_INFO_NODE 539 config HAVE_BOOTMEM_INFO_NODE
534 def_bool n 540 def_bool n
535 541
536 config ARCH_ENABLE_MEMORY_HOTPLUG 542 config ARCH_ENABLE_MEMORY_HOTPLUG
537 bool 543 bool
538 544
539 config ARCH_ENABLE_MEMORY_HOTREMOVE 545 config ARCH_ENABLE_MEMORY_HOTREMOVE
540 bool 546 bool
541 547
542 # eventually, we can have this option just 'se 548 # eventually, we can have this option just 'select SPARSEMEM'
543 menuconfig MEMORY_HOTPLUG 549 menuconfig MEMORY_HOTPLUG
544 bool "Memory hotplug" 550 bool "Memory hotplug"
545 select MEMORY_ISOLATION 551 select MEMORY_ISOLATION
546 depends on SPARSEMEM 552 depends on SPARSEMEM
547 depends on ARCH_ENABLE_MEMORY_HOTPLUG 553 depends on ARCH_ENABLE_MEMORY_HOTPLUG
548 depends on 64BIT 554 depends on 64BIT
549 select NUMA_KEEP_MEMINFO if NUMA 555 select NUMA_KEEP_MEMINFO if NUMA
550 556
551 if MEMORY_HOTPLUG 557 if MEMORY_HOTPLUG
552 558
553 config MEMORY_HOTPLUG_DEFAULT_ONLINE 559 config MEMORY_HOTPLUG_DEFAULT_ONLINE
554 bool "Online the newly added memory bl 560 bool "Online the newly added memory blocks by default"
555 depends on MEMORY_HOTPLUG 561 depends on MEMORY_HOTPLUG
556 help 562 help
557 This option sets the default policy 563 This option sets the default policy setting for memory hotplug
558 onlining policy (/sys/devices/system 564 onlining policy (/sys/devices/system/memory/auto_online_blocks) which
559 determines what happens to newly add 565 determines what happens to newly added memory regions. Policy setting
560 can always be changed at runtime. 566 can always be changed at runtime.
561 See Documentation/admin-guide/mm/mem 567 See Documentation/admin-guide/mm/memory-hotplug.rst for more information.
562 568
563 Say Y here if you want all hot-plugg 569 Say Y here if you want all hot-plugged memory blocks to appear in
564 'online' state by default. 570 'online' state by default.
565 Say N here if you want the default p 571 Say N here if you want the default policy to keep all hot-plugged
566 memory blocks in 'offline' state. 572 memory blocks in 'offline' state.
567 573
568 config MEMORY_HOTREMOVE 574 config MEMORY_HOTREMOVE
569 bool "Allow for memory hot remove" 575 bool "Allow for memory hot remove"
570 select HAVE_BOOTMEM_INFO_NODE if (X86_ 576 select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64)
571 depends on MEMORY_HOTPLUG && ARCH_ENAB 577 depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE
572 depends on MIGRATION 578 depends on MIGRATION
573 579
574 config MHP_MEMMAP_ON_MEMORY 580 config MHP_MEMMAP_ON_MEMORY
575 def_bool y 581 def_bool y
576 depends on MEMORY_HOTPLUG && SPARSEMEM 582 depends on MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP
577 depends on ARCH_MHP_MEMMAP_ON_MEMORY_E 583 depends on ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
578 584
579 endif # MEMORY_HOTPLUG 585 endif # MEMORY_HOTPLUG
580 586
581 config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE 587 config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
582 bool 588 bool
583 589
584 # Heavily threaded applications may benefit fr 590 # Heavily threaded applications may benefit from splitting the mm-wide
585 # page_table_lock, so that faults on different 591 # page_table_lock, so that faults on different parts of the user address
586 # space can be handled with less contention: s 592 # space can be handled with less contention: split it at this NR_CPUS.
587 # Default to 4 for wider testing, though 8 mig 593 # Default to 4 for wider testing, though 8 might be more appropriate.
588 # ARM's adjust_pte (unused if VIPT) depends on 594 # ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
589 # PA-RISC 7xxx's spinlock_t would enlarge stru 595 # PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
590 # SPARC32 allocates multiple pte tables within 596 # SPARC32 allocates multiple pte tables within a single page, and therefore
591 # a per-page lock leads to problems when multi 597 # a per-page lock leads to problems when multiple tables need to be locked
592 # at the same time (e.g. copy_page_range()). 598 # at the same time (e.g. copy_page_range()).
593 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock 599 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
594 # 600 #
595 config SPLIT_PTE_PTLOCKS !! 601 config SPLIT_PTLOCK_CPUS
596 def_bool y !! 602 int
597 depends on MMU !! 603 default "999999" if !MMU
598 depends on SMP !! 604 default "999999" if ARM && !CPU_CACHE_VIPT
599 depends on NR_CPUS >= 4 !! 605 default "999999" if PARISC && !PA20
600 depends on !ARM || CPU_CACHE_VIPT !! 606 default "999999" if SPARC32
601 depends on !PARISC || PA20 !! 607 default "4"
602 depends on !SPARC32 <<
603 608
604 config ARCH_ENABLE_SPLIT_PMD_PTLOCK 609 config ARCH_ENABLE_SPLIT_PMD_PTLOCK
605 bool 610 bool
606 611
607 config SPLIT_PMD_PTLOCKS <<
608 def_bool y <<
609 depends on SPLIT_PTE_PTLOCKS && ARCH_E <<
610 <<
611 # 612 #
612 # support for memory balloon 613 # support for memory balloon
613 config MEMORY_BALLOON 614 config MEMORY_BALLOON
614 bool 615 bool
615 616
616 # 617 #
617 # support for memory balloon compaction 618 # support for memory balloon compaction
618 config BALLOON_COMPACTION 619 config BALLOON_COMPACTION
619 bool "Allow for balloon memory compact 620 bool "Allow for balloon memory compaction/migration"
620 default y 621 default y
621 depends on COMPACTION && MEMORY_BALLOO 622 depends on COMPACTION && MEMORY_BALLOON
622 help 623 help
623 Memory fragmentation introduced by b 624 Memory fragmentation introduced by ballooning might reduce
624 significantly the number of 2MB cont 625 significantly the number of 2MB contiguous memory blocks that can be
625 used within a guest, thus imposing p 626 used within a guest, thus imposing performance penalties associated
626 with the reduced number of transpare 627 with the reduced number of transparent huge pages that could be used
627 by the guest workload. Allowing the 628 by the guest workload. Allowing the compaction & migration for memory
628 pages enlisted as being part of memo 629 pages enlisted as being part of memory balloon devices avoids the
629 scenario aforementioned and helps im 630 scenario aforementioned and helps improving memory defragmentation.
630 631
631 # 632 #
632 # support for memory compaction 633 # support for memory compaction
633 config COMPACTION 634 config COMPACTION
634 bool "Allow for memory compaction" 635 bool "Allow for memory compaction"
635 default y 636 default y
636 select MIGRATION 637 select MIGRATION
637 depends on MMU 638 depends on MMU
638 help 639 help
639 Compaction is the only memory manage 640 Compaction is the only memory management component to form
640 high order (larger physically contig 641 high order (larger physically contiguous) memory blocks
641 reliably. The page allocator relies 642 reliably. The page allocator relies on compaction heavily and
642 the lack of the feature can lead to 643 the lack of the feature can lead to unexpected OOM killer
643 invocations for high order memory re 644 invocations for high order memory requests. You shouldn't
644 disable this option unless there rea 645 disable this option unless there really is a strong reason for
645 it and then we would be really inter 646 it and then we would be really interested to hear about that at
646 linux-mm@kvack.org. 647 linux-mm@kvack.org.
647 648
648 config COMPACT_UNEVICTABLE_DEFAULT 649 config COMPACT_UNEVICTABLE_DEFAULT
649 int 650 int
650 depends on COMPACTION 651 depends on COMPACTION
651 default 0 if PREEMPT_RT 652 default 0 if PREEMPT_RT
652 default 1 653 default 1
653 654
654 # 655 #
655 # support for free page reporting 656 # support for free page reporting
656 config PAGE_REPORTING 657 config PAGE_REPORTING
657 bool "Free page reporting" 658 bool "Free page reporting"
658 help 659 help
659 Free page reporting allows for the i 660 Free page reporting allows for the incremental acquisition of
660 free pages from the buddy allocator 661 free pages from the buddy allocator for the purpose of reporting
661 those pages to another entity, such 662 those pages to another entity, such as a hypervisor, so that the
662 memory can be freed within the host 663 memory can be freed within the host for other uses.
663 664
664 # 665 #
665 # support for page migration 666 # support for page migration
666 # 667 #
667 config MIGRATION 668 config MIGRATION
668 bool "Page migration" 669 bool "Page migration"
669 default y 670 default y
670 depends on (NUMA || ARCH_ENABLE_MEMORY 671 depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU
671 help 672 help
672 Allows the migration of the physical 673 Allows the migration of the physical location of pages of processes
673 while the virtual addresses are not 674 while the virtual addresses are not changed. This is useful in
674 two situations. The first is on NUMA 675 two situations. The first is on NUMA systems to put pages nearer
675 to the processors accessing. The sec 676 to the processors accessing. The second is when allocating huge
676 pages as migration can relocate page 677 pages as migration can relocate pages to satisfy a huge page
677 allocation instead of reclaiming. 678 allocation instead of reclaiming.
678 679
679 config DEVICE_MIGRATION 680 config DEVICE_MIGRATION
680 def_bool MIGRATION && ZONE_DEVICE 681 def_bool MIGRATION && ZONE_DEVICE
681 682
682 config ARCH_ENABLE_HUGEPAGE_MIGRATION 683 config ARCH_ENABLE_HUGEPAGE_MIGRATION
683 bool 684 bool
684 685
685 config ARCH_ENABLE_THP_MIGRATION 686 config ARCH_ENABLE_THP_MIGRATION
686 bool 687 bool
687 688
688 config HUGETLB_PAGE_SIZE_VARIABLE 689 config HUGETLB_PAGE_SIZE_VARIABLE
689 def_bool n 690 def_bool n
690 help 691 help
691 Allows the pageblock_order value to 692 Allows the pageblock_order value to be dynamic instead of just standard
692 HUGETLB_PAGE_ORDER when there are mu 693 HUGETLB_PAGE_ORDER when there are multiple HugeTLB page sizes available
693 on a platform. 694 on a platform.
694 695
695 Note that the pageblock_order cannot 696 Note that the pageblock_order cannot exceed MAX_PAGE_ORDER and will be
696 clamped down to MAX_PAGE_ORDER. 697 clamped down to MAX_PAGE_ORDER.
697 698
698 config CONTIG_ALLOC 699 config CONTIG_ALLOC
699 def_bool (MEMORY_ISOLATION && COMPACTI 700 def_bool (MEMORY_ISOLATION && COMPACTION) || CMA
700 701
701 config PCP_BATCH_SCALE_MAX 702 config PCP_BATCH_SCALE_MAX
702 int "Maximum scale factor of PCP (Per- 703 int "Maximum scale factor of PCP (Per-CPU pageset) batch allocate/free"
703 default 5 704 default 5
704 range 0 6 705 range 0 6
705 help 706 help
706 In page allocator, PCP (Per-CPU page 707 In page allocator, PCP (Per-CPU pageset) is refilled and drained in
707 batches. The batch number is scaled 708 batches. The batch number is scaled automatically to improve page
708 allocation/free throughput. But too 709 allocation/free throughput. But too large scale factor may hurt
709 latency. This option sets the upper 710 latency. This option sets the upper limit of scale factor to limit
710 the maximum latency. 711 the maximum latency.
711 712
712 config PHYS_ADDR_T_64BIT 713 config PHYS_ADDR_T_64BIT
713 def_bool 64BIT 714 def_bool 64BIT
714 715
715 config BOUNCE 716 config BOUNCE
716 bool "Enable bounce buffers" 717 bool "Enable bounce buffers"
717 default y 718 default y
718 depends on BLOCK && MMU && HIGHMEM 719 depends on BLOCK && MMU && HIGHMEM
719 help 720 help
720 Enable bounce buffers for devices th 721 Enable bounce buffers for devices that cannot access the full range of
721 memory available to the CPU. Enabled 722 memory available to the CPU. Enabled by default when HIGHMEM is
722 selected, but you may say n to overr 723 selected, but you may say n to override this.
723 724
724 config MMU_NOTIFIER 725 config MMU_NOTIFIER
725 bool 726 bool
726 select INTERVAL_TREE 727 select INTERVAL_TREE
727 728
728 config KSM 729 config KSM
729 bool "Enable KSM for page merging" 730 bool "Enable KSM for page merging"
730 depends on MMU 731 depends on MMU
731 select XXHASH 732 select XXHASH
732 help 733 help
733 Enable Kernel Samepage Merging: KSM 734 Enable Kernel Samepage Merging: KSM periodically scans those areas
734 of an application's address space th 735 of an application's address space that an app has advised may be
735 mergeable. When it finds pages of i 736 mergeable. When it finds pages of identical content, it replaces
736 the many instances by a single page 737 the many instances by a single page with that content, so
737 saving memory until one or another a 738 saving memory until one or another app needs to modify the content.
738 Recommended for use with KVM, or wit 739 Recommended for use with KVM, or with other duplicative applications.
739 See Documentation/mm/ksm.rst for mor 740 See Documentation/mm/ksm.rst for more information: KSM is inactive
740 until a program has madvised that an 741 until a program has madvised that an area is MADV_MERGEABLE, and
741 root has set /sys/kernel/mm/ksm/run 742 root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set).
742 743
743 config DEFAULT_MMAP_MIN_ADDR 744 config DEFAULT_MMAP_MIN_ADDR
744 int "Low address space to protect from 745 int "Low address space to protect from user allocation"
745 depends on MMU 746 depends on MMU
746 default 4096 747 default 4096
747 help 748 help
748 This is the portion of low virtual m 749 This is the portion of low virtual memory which should be protected
749 from userspace allocation. Keeping 750 from userspace allocation. Keeping a user from writing to low pages
750 can help reduce the impact of kernel 751 can help reduce the impact of kernel NULL pointer bugs.
751 752
752 For most arm64, ppc64 and x86 users 753 For most arm64, ppc64 and x86 users with lots of address space
753 a value of 65536 is reasonable and s 754 a value of 65536 is reasonable and should cause no problems.
754 On arm and other archs it should not 755 On arm and other archs it should not be higher than 32768.
755 Programs which use vm86 functionalit 756 Programs which use vm86 functionality or have some need to map
756 this low address space will need CAP 757 this low address space will need CAP_SYS_RAWIO or disable this
757 protection by setting the value to 0 758 protection by setting the value to 0.
758 759
759 This value can be changed after boot 760 This value can be changed after boot using the
760 /proc/sys/vm/mmap_min_addr tunable. 761 /proc/sys/vm/mmap_min_addr tunable.
761 762
762 config ARCH_SUPPORTS_MEMORY_FAILURE 763 config ARCH_SUPPORTS_MEMORY_FAILURE
763 bool 764 bool
764 765
765 config MEMORY_FAILURE 766 config MEMORY_FAILURE
766 depends on MMU 767 depends on MMU
767 depends on ARCH_SUPPORTS_MEMORY_FAILUR 768 depends on ARCH_SUPPORTS_MEMORY_FAILURE
768 bool "Enable recovery from hardware me 769 bool "Enable recovery from hardware memory errors"
769 select MEMORY_ISOLATION 770 select MEMORY_ISOLATION
770 select RAS 771 select RAS
771 help 772 help
772 Enables code to recover from some me 773 Enables code to recover from some memory failures on systems
773 with MCA recovery. This allows a sys 774 with MCA recovery. This allows a system to continue running
774 even when some of its memory has unc 775 even when some of its memory has uncorrected errors. This requires
775 special hardware support and typical 776 special hardware support and typically ECC memory.
776 777
777 config HWPOISON_INJECT 778 config HWPOISON_INJECT
778 tristate "HWPoison pages injector" 779 tristate "HWPoison pages injector"
779 depends on MEMORY_FAILURE && DEBUG_KER 780 depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS
780 select PROC_PAGE_MONITOR 781 select PROC_PAGE_MONITOR
781 782
782 config NOMMU_INITIAL_TRIM_EXCESS 783 config NOMMU_INITIAL_TRIM_EXCESS
783 int "Turn on mmap() excess space trimm 784 int "Turn on mmap() excess space trimming before booting"
784 depends on !MMU 785 depends on !MMU
785 default 1 786 default 1
786 help 787 help
787 The NOMMU mmap() frequently needs to 788 The NOMMU mmap() frequently needs to allocate large contiguous chunks
788 of memory on which to store mappings 789 of memory on which to store mappings, but it can only ask the system
789 allocator for chunks in 2^N*PAGE_SIZ 790 allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently
790 more than it requires. To deal with 791 more than it requires. To deal with this, mmap() is able to trim off
791 the excess and return it to the allo 792 the excess and return it to the allocator.
792 793
793 If trimming is enabled, the excess i 794 If trimming is enabled, the excess is trimmed off and returned to the
794 system allocator, which can cause ex 795 system allocator, which can cause extra fragmentation, particularly
795 if there are a lot of transient proc 796 if there are a lot of transient processes.
796 797
797 If trimming is disabled, the excess 798 If trimming is disabled, the excess is kept, but not used, which for
798 long-term mappings means that the sp 799 long-term mappings means that the space is wasted.
799 800
800 Trimming can be dynamically controll 801 Trimming can be dynamically controlled through a sysctl option
801 (/proc/sys/vm/nr_trim_pages) which s 802 (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of
802 excess pages there must be before tr 803 excess pages there must be before trimming should occur, or zero if
803 no trimming is to occur. 804 no trimming is to occur.
804 805
805 This option specifies the initial va 806 This option specifies the initial value of this option. The default
806 of 1 says that all excess pages shou 807 of 1 says that all excess pages should be trimmed.
807 808
808 See Documentation/admin-guide/mm/nom 809 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
809 810
810 config ARCH_WANT_GENERAL_HUGETLB 811 config ARCH_WANT_GENERAL_HUGETLB
811 bool 812 bool
812 813
813 config ARCH_WANTS_THP_SWAP 814 config ARCH_WANTS_THP_SWAP
814 def_bool n 815 def_bool n
815 816
816 menuconfig TRANSPARENT_HUGEPAGE 817 menuconfig TRANSPARENT_HUGEPAGE
817 bool "Transparent Hugepage Support" 818 bool "Transparent Hugepage Support"
818 depends on HAVE_ARCH_TRANSPARENT_HUGEP 819 depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT
819 select COMPACTION 820 select COMPACTION
820 select XARRAY_MULTI 821 select XARRAY_MULTI
821 help 822 help
822 Transparent Hugepages allows the ker 823 Transparent Hugepages allows the kernel to use huge pages and
823 huge tlb transparently to the applic 824 huge tlb transparently to the applications whenever possible.
824 This feature can improve computing p 825 This feature can improve computing performance to certain
825 applications by speeding up page fau 826 applications by speeding up page faults during memory
826 allocation, by reducing the number o 827 allocation, by reducing the number of tlb misses and by speeding
827 up the pagetable walking. 828 up the pagetable walking.
828 829
829 If memory constrained on embedded, y 830 If memory constrained on embedded, you may want to say N.
830 831
831 if TRANSPARENT_HUGEPAGE 832 if TRANSPARENT_HUGEPAGE
832 833
833 choice 834 choice
834 prompt "Transparent Hugepage Support s 835 prompt "Transparent Hugepage Support sysfs defaults"
835 depends on TRANSPARENT_HUGEPAGE 836 depends on TRANSPARENT_HUGEPAGE
836 default TRANSPARENT_HUGEPAGE_ALWAYS 837 default TRANSPARENT_HUGEPAGE_ALWAYS
837 help 838 help
838 Selects the sysfs defaults for Trans 839 Selects the sysfs defaults for Transparent Hugepage Support.
839 840
840 config TRANSPARENT_HUGEPAGE_ALWAYS 841 config TRANSPARENT_HUGEPAGE_ALWAYS
841 bool "always" 842 bool "always"
842 help 843 help
843 Enabling Transparent Hugepage always 844 Enabling Transparent Hugepage always, can increase the
844 memory footprint of applications wit 845 memory footprint of applications without a guaranteed
845 benefit but it will work automatical 846 benefit but it will work automatically for all applications.
846 847
847 config TRANSPARENT_HUGEPAGE_MADVISE 848 config TRANSPARENT_HUGEPAGE_MADVISE
848 bool "madvise" 849 bool "madvise"
849 help 850 help
850 Enabling Transparent Hugepage madvis 851 Enabling Transparent Hugepage madvise, will only provide a
851 performance improvement benefit to t 852 performance improvement benefit to the applications using
852 madvise(MADV_HUGEPAGE) but it won't 853 madvise(MADV_HUGEPAGE) but it won't risk to increase the
853 memory footprint of applications wit 854 memory footprint of applications without a guaranteed
854 benefit. 855 benefit.
855 856
856 config TRANSPARENT_HUGEPAGE_NEVER 857 config TRANSPARENT_HUGEPAGE_NEVER
857 bool "never" 858 bool "never"
858 help 859 help
859 Disable Transparent Hugepage by defa 860 Disable Transparent Hugepage by default. It can still be
860 enabled at runtime via sysfs. 861 enabled at runtime via sysfs.
861 endchoice 862 endchoice
862 863
863 config THP_SWAP 864 config THP_SWAP
864 def_bool y 865 def_bool y
865 depends on TRANSPARENT_HUGEPAGE && ARC 866 depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP && 64BIT
866 help 867 help
867 Swap transparent huge pages in one p 868 Swap transparent huge pages in one piece, without splitting.
868 XXX: For now, swap cluster backing t 869 XXX: For now, swap cluster backing transparent huge page
869 will be split after swapout. 870 will be split after swapout.
870 871
871 For selection by architectures with 872 For selection by architectures with reasonable THP sizes.
872 873
873 config READ_ONLY_THP_FOR_FS 874 config READ_ONLY_THP_FOR_FS
874 bool "Read-only THP for filesystems (E 875 bool "Read-only THP for filesystems (EXPERIMENTAL)"
875 depends on TRANSPARENT_HUGEPAGE && SHM 876 depends on TRANSPARENT_HUGEPAGE && SHMEM
876 877
877 help 878 help
878 Allow khugepaged to put read-only fi 879 Allow khugepaged to put read-only file-backed pages in THP.
879 880
880 This is marked experimental because 881 This is marked experimental because it is a new feature. Write
881 support of file THPs will be develop 882 support of file THPs will be developed in the next few release
882 cycles. 883 cycles.
883 884
884 endif # TRANSPARENT_HUGEPAGE 885 endif # TRANSPARENT_HUGEPAGE
885 886
886 # 887 #
887 # The architecture supports pgtable leaves tha 888 # The architecture supports pgtable leaves that is larger than PAGE_SIZE
888 # 889 #
889 config PGTABLE_HAS_HUGE_LEAVES 890 config PGTABLE_HAS_HUGE_LEAVES
890 def_bool TRANSPARENT_HUGEPAGE || HUGET 891 def_bool TRANSPARENT_HUGEPAGE || HUGETLB_PAGE
891 892
892 # TODO: Allow to be enabled without THP <<
893 config ARCH_SUPPORTS_HUGE_PFNMAP <<
894 def_bool n <<
895 depends on TRANSPARENT_HUGEPAGE <<
896 <<
897 config ARCH_SUPPORTS_PMD_PFNMAP <<
898 def_bool y <<
899 depends on ARCH_SUPPORTS_HUGE_PFNMAP & <<
900 <<
901 config ARCH_SUPPORTS_PUD_PFNMAP <<
902 def_bool y <<
903 depends on ARCH_SUPPORTS_HUGE_PFNMAP & <<
904 <<
905 # 893 #
906 # UP and nommu archs use km based percpu alloc 894 # UP and nommu archs use km based percpu allocator
907 # 895 #
908 config NEED_PER_CPU_KM 896 config NEED_PER_CPU_KM
909 depends on !SMP || !MMU 897 depends on !SMP || !MMU
910 bool 898 bool
911 default y 899 default y
912 900
913 config NEED_PER_CPU_EMBED_FIRST_CHUNK 901 config NEED_PER_CPU_EMBED_FIRST_CHUNK
914 bool 902 bool
915 903
916 config NEED_PER_CPU_PAGE_FIRST_CHUNK 904 config NEED_PER_CPU_PAGE_FIRST_CHUNK
917 bool 905 bool
918 906
919 config USE_PERCPU_NUMA_NODE_ID 907 config USE_PERCPU_NUMA_NODE_ID
920 bool 908 bool
921 909
922 config HAVE_SETUP_PER_CPU_AREA 910 config HAVE_SETUP_PER_CPU_AREA
923 bool 911 bool
924 912
925 config CMA 913 config CMA
926 bool "Contiguous Memory Allocator" 914 bool "Contiguous Memory Allocator"
927 depends on MMU 915 depends on MMU
928 select MIGRATION 916 select MIGRATION
929 select MEMORY_ISOLATION 917 select MEMORY_ISOLATION
930 help 918 help
931 This enables the Contiguous Memory A 919 This enables the Contiguous Memory Allocator which allows other
932 subsystems to allocate big physicall 920 subsystems to allocate big physically-contiguous blocks of memory.
933 CMA reserves a region of memory and 921 CMA reserves a region of memory and allows only movable pages to
934 be allocated from it. This way, the 922 be allocated from it. This way, the kernel can use the memory for
935 pagecache and when a subsystem reque 923 pagecache and when a subsystem requests for contiguous area, the
936 allocated pages are migrated away to 924 allocated pages are migrated away to serve the contiguous request.
937 925
938 If unsure, say "n". 926 If unsure, say "n".
939 927
940 config CMA_DEBUGFS 928 config CMA_DEBUGFS
941 bool "CMA debugfs interface" 929 bool "CMA debugfs interface"
942 depends on CMA && DEBUG_FS 930 depends on CMA && DEBUG_FS
943 help 931 help
944 Turns on the DebugFS interface for C 932 Turns on the DebugFS interface for CMA.
945 933
946 config CMA_SYSFS 934 config CMA_SYSFS
947 bool "CMA information through sysfs in 935 bool "CMA information through sysfs interface"
948 depends on CMA && SYSFS 936 depends on CMA && SYSFS
949 help 937 help
950 This option exposes some sysfs attri 938 This option exposes some sysfs attributes to get information
951 from CMA. 939 from CMA.
952 940
953 config CMA_AREAS 941 config CMA_AREAS
954 int "Maximum count of the CMA areas" 942 int "Maximum count of the CMA areas"
955 depends on CMA 943 depends on CMA
956 default 20 if NUMA 944 default 20 if NUMA
957 default 8 945 default 8
958 help 946 help
959 CMA allows to create CMA areas for p 947 CMA allows to create CMA areas for particular purpose, mainly,
960 used as device private area. This pa 948 used as device private area. This parameter sets the maximum
961 number of CMA area in the system. 949 number of CMA area in the system.
962 950
963 If unsure, leave the default value " 951 If unsure, leave the default value "8" in UMA and "20" in NUMA.
964 952
965 config MEM_SOFT_DIRTY 953 config MEM_SOFT_DIRTY
966 bool "Track memory changes" 954 bool "Track memory changes"
967 depends on CHECKPOINT_RESTORE && HAVE_ 955 depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS
968 select PROC_PAGE_MONITOR 956 select PROC_PAGE_MONITOR
969 help 957 help
970 This option enables memory changes t 958 This option enables memory changes tracking by introducing a
971 soft-dirty bit on pte-s. This bit it 959 soft-dirty bit on pte-s. This bit it set when someone writes
972 into a page just as regular dirty bi 960 into a page just as regular dirty bit, but unlike the latter
973 it can be cleared by hands. 961 it can be cleared by hands.
974 962
975 See Documentation/admin-guide/mm/sof 963 See Documentation/admin-guide/mm/soft-dirty.rst for more details.
976 964
977 config GENERIC_EARLY_IOREMAP 965 config GENERIC_EARLY_IOREMAP
978 bool 966 bool
979 967
980 config STACK_MAX_DEFAULT_SIZE_MB 968 config STACK_MAX_DEFAULT_SIZE_MB
981 int "Default maximum user stack size f 969 int "Default maximum user stack size for 32-bit processes (MB)"
982 default 100 970 default 100
983 range 8 2048 971 range 8 2048
984 depends on STACK_GROWSUP && (!64BIT || 972 depends on STACK_GROWSUP && (!64BIT || COMPAT)
985 help 973 help
986 This is the maximum stack size in Me 974 This is the maximum stack size in Megabytes in the VM layout of 32-bit
987 user processes when the stack grows 975 user processes when the stack grows upwards (currently only on parisc
988 arch) when the RLIMIT_STACK hard lim 976 arch) when the RLIMIT_STACK hard limit is unlimited.
989 977
990 A sane initial value is 100 MB. 978 A sane initial value is 100 MB.
991 979
992 config DEFERRED_STRUCT_PAGE_INIT 980 config DEFERRED_STRUCT_PAGE_INIT
993 bool "Defer initialisation of struct p 981 bool "Defer initialisation of struct pages to kthreads"
994 depends on SPARSEMEM 982 depends on SPARSEMEM
995 depends on !NEED_PER_CPU_KM 983 depends on !NEED_PER_CPU_KM
996 depends on 64BIT 984 depends on 64BIT
997 depends on !KMSAN 985 depends on !KMSAN
998 select PADATA 986 select PADATA
999 help 987 help
1000 Ordinarily all struct pages are ini 988 Ordinarily all struct pages are initialised during early boot in a
1001 single thread. On very large machin 989 single thread. On very large machines this can take a considerable
1002 amount of time. If this option is s 990 amount of time. If this option is set, large machines will bring up
1003 a subset of memmap at boot and then 991 a subset of memmap at boot and then initialise the rest in parallel.
1004 This has a potential performance im 992 This has a potential performance impact on tasks running early in the
1005 lifetime of the system until these 993 lifetime of the system until these kthreads finish the
1006 initialisation. 994 initialisation.
1007 995
1008 config PAGE_IDLE_FLAG 996 config PAGE_IDLE_FLAG
1009 bool 997 bool
1010 select PAGE_EXTENSION if !64BIT 998 select PAGE_EXTENSION if !64BIT
1011 help 999 help
1012 This adds PG_idle and PG_young flag 1000 This adds PG_idle and PG_young flags to 'struct page'. PTE Accessed
1013 bit writers can set the state of th 1001 bit writers can set the state of the bit in the flags so that PTE
1014 Accessed bit readers may avoid dist 1002 Accessed bit readers may avoid disturbance.
1015 1003
1016 config IDLE_PAGE_TRACKING 1004 config IDLE_PAGE_TRACKING
1017 bool "Enable idle page tracking" 1005 bool "Enable idle page tracking"
1018 depends on SYSFS && MMU 1006 depends on SYSFS && MMU
1019 select PAGE_IDLE_FLAG 1007 select PAGE_IDLE_FLAG
1020 help 1008 help
1021 This feature allows to estimate the 1009 This feature allows to estimate the amount of user pages that have
1022 not been touched during a given per 1010 not been touched during a given period of time. This information can
1023 be useful to tune memory cgroup lim 1011 be useful to tune memory cgroup limits and/or for job placement
1024 within a compute cluster. 1012 within a compute cluster.
1025 1013
1026 See Documentation/admin-guide/mm/id 1014 See Documentation/admin-guide/mm/idle_page_tracking.rst for
1027 more details. 1015 more details.
1028 1016
1029 # Architectures which implement cpu_dcache_is 1017 # Architectures which implement cpu_dcache_is_aliasing() to query
1030 # whether the data caches are aliased (VIVT o 1018 # whether the data caches are aliased (VIVT or VIPT with dcache
1031 # aliasing) need to select this. 1019 # aliasing) need to select this.
1032 config ARCH_HAS_CPU_CACHE_ALIASING 1020 config ARCH_HAS_CPU_CACHE_ALIASING
1033 bool 1021 bool
1034 1022
1035 config ARCH_HAS_CACHE_LINE_SIZE 1023 config ARCH_HAS_CACHE_LINE_SIZE
1036 bool 1024 bool
1037 1025
1038 config ARCH_HAS_CURRENT_STACK_POINTER 1026 config ARCH_HAS_CURRENT_STACK_POINTER
1039 bool 1027 bool
1040 help 1028 help
1041 In support of HARDENED_USERCOPY per 1029 In support of HARDENED_USERCOPY performing stack variable lifetime
1042 checking, an architecture-agnostic 1030 checking, an architecture-agnostic way to find the stack pointer
1043 is needed. Once an architecture def 1031 is needed. Once an architecture defines an unsigned long global
1044 register alias named "current_stack 1032 register alias named "current_stack_pointer", this config can be
1045 selected. 1033 selected.
1046 1034
1047 config ARCH_HAS_PTE_DEVMAP 1035 config ARCH_HAS_PTE_DEVMAP
1048 bool 1036 bool
1049 1037
1050 config ARCH_HAS_ZONE_DMA_SET 1038 config ARCH_HAS_ZONE_DMA_SET
1051 bool 1039 bool
1052 1040
1053 config ZONE_DMA 1041 config ZONE_DMA
1054 bool "Support DMA zone" if ARCH_HAS_Z 1042 bool "Support DMA zone" if ARCH_HAS_ZONE_DMA_SET
1055 default y if ARM64 || X86 1043 default y if ARM64 || X86
1056 1044
1057 config ZONE_DMA32 1045 config ZONE_DMA32
1058 bool "Support DMA32 zone" if ARCH_HAS 1046 bool "Support DMA32 zone" if ARCH_HAS_ZONE_DMA_SET
1059 depends on !X86_32 1047 depends on !X86_32
1060 default y if ARM64 1048 default y if ARM64
1061 1049
1062 config ZONE_DEVICE 1050 config ZONE_DEVICE
1063 bool "Device memory (pmem, HMM, etc.. 1051 bool "Device memory (pmem, HMM, etc...) hotplug support"
1064 depends on MEMORY_HOTPLUG 1052 depends on MEMORY_HOTPLUG
1065 depends on MEMORY_HOTREMOVE 1053 depends on MEMORY_HOTREMOVE
1066 depends on SPARSEMEM_VMEMMAP 1054 depends on SPARSEMEM_VMEMMAP
1067 depends on ARCH_HAS_PTE_DEVMAP 1055 depends on ARCH_HAS_PTE_DEVMAP
1068 select XARRAY_MULTI 1056 select XARRAY_MULTI
1069 1057
1070 help 1058 help
1071 Device memory hotplug support allow 1059 Device memory hotplug support allows for establishing pmem,
1072 or other device driver discovered m 1060 or other device driver discovered memory regions, in the
1073 memmap. This allows pfn_to_page() l 1061 memmap. This allows pfn_to_page() lookups of otherwise
1074 "device-physical" addresses which i 1062 "device-physical" addresses which is needed for using a DAX
1075 mapping in an O_DIRECT operation, a 1063 mapping in an O_DIRECT operation, among other things.
1076 1064
1077 If FS_DAX is enabled, then say Y. 1065 If FS_DAX is enabled, then say Y.
1078 1066
1079 # 1067 #
1080 # Helpers to mirror range of the CPU page tab 1068 # Helpers to mirror range of the CPU page tables of a process into device page
1081 # tables. 1069 # tables.
1082 # 1070 #
1083 config HMM_MIRROR 1071 config HMM_MIRROR
1084 bool 1072 bool
1085 depends on MMU 1073 depends on MMU
1086 1074
1087 config GET_FREE_REGION 1075 config GET_FREE_REGION
>> 1076 depends on SPARSEMEM
1088 bool 1077 bool
1089 1078
1090 config DEVICE_PRIVATE 1079 config DEVICE_PRIVATE
1091 bool "Unaddressable device memory (GP 1080 bool "Unaddressable device memory (GPU memory, ...)"
1092 depends on ZONE_DEVICE 1081 depends on ZONE_DEVICE
1093 select GET_FREE_REGION 1082 select GET_FREE_REGION
1094 1083
1095 help 1084 help
1096 Allows creation of struct pages to 1085 Allows creation of struct pages to represent unaddressable device
1097 memory; i.e., memory that is only a 1086 memory; i.e., memory that is only accessible from the device (or
1098 group of devices). You likely also 1087 group of devices). You likely also want to select HMM_MIRROR.
1099 1088
1100 config VMAP_PFN 1089 config VMAP_PFN
1101 bool 1090 bool
1102 1091
1103 config ARCH_USES_HIGH_VMA_FLAGS 1092 config ARCH_USES_HIGH_VMA_FLAGS
1104 bool 1093 bool
1105 config ARCH_HAS_PKEYS 1094 config ARCH_HAS_PKEYS
1106 bool 1095 bool
1107 1096
1108 config ARCH_USES_PG_ARCH_2 !! 1097 config ARCH_USES_PG_ARCH_X
1109 bool <<
1110 config ARCH_USES_PG_ARCH_3 <<
1111 bool 1098 bool
>> 1099 help
>> 1100 Enable the definition of PG_arch_x page flags with x > 1. Only
>> 1101 suitable for 64-bit architectures with CONFIG_FLATMEM or
>> 1102 CONFIG_SPARSEMEM_VMEMMAP enabled, otherwise there may not be
>> 1103 enough room for additional bits in page->flags.
1112 1104
1113 config VM_EVENT_COUNTERS 1105 config VM_EVENT_COUNTERS
1114 default y 1106 default y
1115 bool "Enable VM event counters for /p 1107 bool "Enable VM event counters for /proc/vmstat" if EXPERT
1116 help 1108 help
1117 VM event counters are needed for ev 1109 VM event counters are needed for event counts to be shown.
1118 This option allows the disabling of 1110 This option allows the disabling of the VM event counters
1119 on EXPERT systems. /proc/vmstat wi 1111 on EXPERT systems. /proc/vmstat will only show page counts
1120 if VM event counters are disabled. 1112 if VM event counters are disabled.
1121 1113
1122 config PERCPU_STATS 1114 config PERCPU_STATS
1123 bool "Collect percpu memory statistic 1115 bool "Collect percpu memory statistics"
1124 help 1116 help
1125 This feature collects and exposes s 1117 This feature collects and exposes statistics via debugfs. The
1126 information includes global and per 1118 information includes global and per chunk statistics, which can
1127 be used to help understand percpu m 1119 be used to help understand percpu memory usage.
1128 1120
1129 config GUP_TEST 1121 config GUP_TEST
1130 bool "Enable infrastructure for get_u 1122 bool "Enable infrastructure for get_user_pages()-related unit tests"
1131 depends on DEBUG_FS 1123 depends on DEBUG_FS
1132 help 1124 help
1133 Provides /sys/kernel/debug/gup_test 1125 Provides /sys/kernel/debug/gup_test, which in turn provides a way
1134 to make ioctl calls that can launch 1126 to make ioctl calls that can launch kernel-based unit tests for
1135 the get_user_pages*() and pin_user_ 1127 the get_user_pages*() and pin_user_pages*() family of API calls.
1136 1128
1137 These tests include benchmark testi 1129 These tests include benchmark testing of the _fast variants of
1138 get_user_pages*() and pin_user_page 1130 get_user_pages*() and pin_user_pages*(), as well as smoke tests of
1139 the non-_fast variants. 1131 the non-_fast variants.
1140 1132
1141 There is also a sub-test that allow 1133 There is also a sub-test that allows running dump_page() on any
1142 of up to eight pages (selected by c 1134 of up to eight pages (selected by command line args) within the
1143 range of user-space addresses. Thes 1135 range of user-space addresses. These pages are either pinned via
1144 pin_user_pages*(), or pinned via ge 1136 pin_user_pages*(), or pinned via get_user_pages*(), as specified
1145 by other command line arguments. 1137 by other command line arguments.
1146 1138
1147 See tools/testing/selftests/mm/gup_ 1139 See tools/testing/selftests/mm/gup_test.c
1148 1140
1149 comment "GUP_TEST needs to have DEBUG_FS enab 1141 comment "GUP_TEST needs to have DEBUG_FS enabled"
1150 depends on !GUP_TEST && !DEBUG_FS 1142 depends on !GUP_TEST && !DEBUG_FS
1151 1143
1152 config GUP_GET_PXX_LOW_HIGH 1144 config GUP_GET_PXX_LOW_HIGH
1153 bool 1145 bool
1154 1146
1155 config DMAPOOL_TEST 1147 config DMAPOOL_TEST
1156 tristate "Enable a module to run time 1148 tristate "Enable a module to run time tests on dma_pool"
1157 depends on HAS_DMA 1149 depends on HAS_DMA
1158 help 1150 help
1159 Provides a test module that will al 1151 Provides a test module that will allocate and free many blocks of
1160 various sizes and report how long i 1152 various sizes and report how long it takes. This is intended to
1161 provide a consistent way to measure 1153 provide a consistent way to measure how changes to the
1162 dma_pool_alloc/free routines affect 1154 dma_pool_alloc/free routines affect performance.
1163 1155
1164 config ARCH_HAS_PTE_SPECIAL 1156 config ARCH_HAS_PTE_SPECIAL
1165 bool 1157 bool
1166 1158
1167 config MAPPING_DIRTY_HELPERS 1159 config MAPPING_DIRTY_HELPERS
1168 bool 1160 bool
1169 1161
1170 config KMAP_LOCAL 1162 config KMAP_LOCAL
1171 bool 1163 bool
1172 1164
1173 config KMAP_LOCAL_NON_LINEAR_PTE_ARRAY 1165 config KMAP_LOCAL_NON_LINEAR_PTE_ARRAY
1174 bool 1166 bool
1175 1167
1176 # struct io_mapping based helper. Selected b 1168 # struct io_mapping based helper. Selected by drivers that need them
1177 config IO_MAPPING 1169 config IO_MAPPING
1178 bool 1170 bool
1179 1171
1180 config MEMFD_CREATE 1172 config MEMFD_CREATE
1181 bool "Enable memfd_create() system ca 1173 bool "Enable memfd_create() system call" if EXPERT
1182 1174
1183 config SECRETMEM 1175 config SECRETMEM
1184 default y 1176 default y
1185 bool "Enable memfd_secret() system ca 1177 bool "Enable memfd_secret() system call" if EXPERT
1186 depends on ARCH_HAS_SET_DIRECT_MAP 1178 depends on ARCH_HAS_SET_DIRECT_MAP
1187 help 1179 help
1188 Enable the memfd_secret() system ca 1180 Enable the memfd_secret() system call with the ability to create
1189 memory areas visible only in the co 1181 memory areas visible only in the context of the owning process and
1190 not mapped to other processes and o 1182 not mapped to other processes and other kernel page tables.
1191 1183
1192 config ANON_VMA_NAME 1184 config ANON_VMA_NAME
1193 bool "Anonymous VMA name support" 1185 bool "Anonymous VMA name support"
1194 depends on PROC_FS && ADVISE_SYSCALLS 1186 depends on PROC_FS && ADVISE_SYSCALLS && MMU
1195 1187
1196 help 1188 help
1197 Allow naming anonymous virtual memo 1189 Allow naming anonymous virtual memory areas.
1198 1190
1199 This feature allows assigning names 1191 This feature allows assigning names to virtual memory areas. Assigned
1200 names can be later retrieved from / 1192 names can be later retrieved from /proc/pid/maps and /proc/pid/smaps
1201 and help identifying individual ano 1193 and help identifying individual anonymous memory areas.
1202 Assigning a name to anonymous virtu 1194 Assigning a name to anonymous virtual memory area might prevent that
1203 area from being merged with adjacen 1195 area from being merged with adjacent virtual memory areas due to the
1204 difference in their name. 1196 difference in their name.
1205 1197
1206 config HAVE_ARCH_USERFAULTFD_WP 1198 config HAVE_ARCH_USERFAULTFD_WP
1207 bool 1199 bool
1208 help 1200 help
1209 Arch has userfaultfd write protecti 1201 Arch has userfaultfd write protection support
1210 1202
1211 config HAVE_ARCH_USERFAULTFD_MINOR 1203 config HAVE_ARCH_USERFAULTFD_MINOR
1212 bool 1204 bool
1213 help 1205 help
1214 Arch has userfaultfd minor fault su 1206 Arch has userfaultfd minor fault support
1215 1207
1216 menuconfig USERFAULTFD 1208 menuconfig USERFAULTFD
1217 bool "Enable userfaultfd() system cal 1209 bool "Enable userfaultfd() system call"
1218 depends on MMU 1210 depends on MMU
1219 help 1211 help
1220 Enable the userfaultfd() system cal 1212 Enable the userfaultfd() system call that allows to intercept and
1221 handle page faults in userland. 1213 handle page faults in userland.
1222 1214
1223 if USERFAULTFD 1215 if USERFAULTFD
1224 config PTE_MARKER_UFFD_WP 1216 config PTE_MARKER_UFFD_WP
1225 bool "Userfaultfd write protection su 1217 bool "Userfaultfd write protection support for shmem/hugetlbfs"
1226 default y 1218 default y
1227 depends on HAVE_ARCH_USERFAULTFD_WP 1219 depends on HAVE_ARCH_USERFAULTFD_WP
1228 1220
1229 help 1221 help
1230 Allows to create marker PTEs for us 1222 Allows to create marker PTEs for userfaultfd write protection
1231 purposes. It is required to enable 1223 purposes. It is required to enable userfaultfd write protection on
1232 file-backed memory types like shmem 1224 file-backed memory types like shmem and hugetlbfs.
1233 endif # USERFAULTFD 1225 endif # USERFAULTFD
1234 1226
1235 # multi-gen LRU { 1227 # multi-gen LRU {
1236 config LRU_GEN 1228 config LRU_GEN
1237 bool "Multi-Gen LRU" 1229 bool "Multi-Gen LRU"
1238 depends on MMU 1230 depends on MMU
1239 # make sure folio->flags has enough s 1231 # make sure folio->flags has enough spare bits
1240 depends on 64BIT || !SPARSEMEM || SPA 1232 depends on 64BIT || !SPARSEMEM || SPARSEMEM_VMEMMAP
1241 help 1233 help
1242 A high performance LRU implementati 1234 A high performance LRU implementation to overcommit memory. See
1243 Documentation/admin-guide/mm/multig 1235 Documentation/admin-guide/mm/multigen_lru.rst for details.
1244 1236
1245 config LRU_GEN_ENABLED 1237 config LRU_GEN_ENABLED
1246 bool "Enable by default" 1238 bool "Enable by default"
1247 depends on LRU_GEN 1239 depends on LRU_GEN
1248 help 1240 help
1249 This option enables the multi-gen L 1241 This option enables the multi-gen LRU by default.
1250 1242
1251 config LRU_GEN_STATS 1243 config LRU_GEN_STATS
1252 bool "Full stats for debugging" 1244 bool "Full stats for debugging"
1253 depends on LRU_GEN 1245 depends on LRU_GEN
1254 help 1246 help
1255 Do not enable this option unless yo 1247 Do not enable this option unless you plan to look at historical stats
1256 from evicted generations for debugg 1248 from evicted generations for debugging purpose.
1257 1249
1258 This option has a per-memcg and per 1250 This option has a per-memcg and per-node memory overhead.
1259 1251
1260 config LRU_GEN_WALKS_MMU 1252 config LRU_GEN_WALKS_MMU
1261 def_bool y 1253 def_bool y
1262 depends on LRU_GEN && ARCH_HAS_HW_PTE 1254 depends on LRU_GEN && ARCH_HAS_HW_PTE_YOUNG
1263 # } 1255 # }
1264 1256
1265 config ARCH_SUPPORTS_PER_VMA_LOCK 1257 config ARCH_SUPPORTS_PER_VMA_LOCK
1266 def_bool n 1258 def_bool n
1267 1259
1268 config PER_VMA_LOCK 1260 config PER_VMA_LOCK
1269 def_bool y 1261 def_bool y
1270 depends on ARCH_SUPPORTS_PER_VMA_LOCK 1262 depends on ARCH_SUPPORTS_PER_VMA_LOCK && MMU && SMP
1271 help 1263 help
1272 Allow per-vma locking during page f 1264 Allow per-vma locking during page fault handling.
1273 1265
1274 This feature allows locking each vi 1266 This feature allows locking each virtual memory area separately when
1275 handling page faults instead of tak 1267 handling page faults instead of taking mmap_lock.
1276 1268
1277 config LOCK_MM_AND_FIND_VMA 1269 config LOCK_MM_AND_FIND_VMA
1278 bool 1270 bool
1279 depends on !STACK_GROWSUP 1271 depends on !STACK_GROWSUP
1280 1272
1281 config IOMMU_MM_DATA 1273 config IOMMU_MM_DATA
1282 bool 1274 bool
1283 1275
1284 config EXECMEM 1276 config EXECMEM
1285 bool 1277 bool
1286 <<
1287 config NUMA_MEMBLKS <<
1288 bool <<
1289 <<
1290 config NUMA_EMU <<
1291 bool "NUMA emulation" <<
1292 depends on NUMA_MEMBLKS <<
1293 help <<
1294 Enable NUMA emulation. A flat machi <<
1295 into virtual nodes when booted with <<
1296 number of nodes. This is only usefu <<
1297 1278
1298 source "mm/damon/Kconfig" 1279 source "mm/damon/Kconfig"
1299 1280
1300 endmenu 1281 endmenu
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.