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_EXCLUSIVE_LOADS_DEFAULT_ON
49 bool "Shrink the zswap pool on memory !! 49 bool "Invalidate zswap entries when pages are loaded"
50 depends on ZSWAP 50 depends on ZSWAP
51 default n <<
52 help 51 help
53 If selected, the zswap shrinker will !! 52 If selected, exclusive loads for zswap will be enabled at boot,
54 stored in the zswap pool will become !! 53 otherwise it will be disabled.
55 written back to the backing swap dev !! 54
56 !! 55 If exclusive loads are enabled, when a page is loaded from zswap,
57 This means that zswap writeback coul !! 56 the zswap entry is invalidated at once, as opposed to leaving it
58 not yet full, or the cgroup zswap li !! 57 in zswap until the swap entry is freed.
59 reducing the chance that cold pages !! 58
60 and consume memory indefinitely. !! 59 This avoids having two copies of the same page in memory
>> 60 (compressed and uncompressed) after faulting in a page from zswap.
>> 61 The cost is that if the page was never dirtied and needs to be
>> 62 swapped out again, it will be re-compressed.
61 63
62 choice 64 choice
63 prompt "Default compressor" 65 prompt "Default compressor"
64 depends on ZSWAP 66 depends on ZSWAP
65 default ZSWAP_COMPRESSOR_DEFAULT_LZO 67 default ZSWAP_COMPRESSOR_DEFAULT_LZO
66 help 68 help
67 Selects the default compression algo 69 Selects the default compression algorithm for the compressed cache
68 for swap pages. 70 for swap pages.
69 71
70 For an overview what kind of perform 72 For an overview what kind of performance can be expected from
71 a particular compression algorithm p 73 a particular compression algorithm please refer to the benchmarks
72 available at the following LWN page: 74 available at the following LWN page:
73 https://lwn.net/Articles/751795/ 75 https://lwn.net/Articles/751795/
74 76
75 If in doubt, select 'LZO'. 77 If in doubt, select 'LZO'.
76 78
77 The selection made here can be overr 79 The selection made here can be overridden by using the kernel
78 command line 'zswap.compressor=' opt 80 command line 'zswap.compressor=' option.
79 81
80 config ZSWAP_COMPRESSOR_DEFAULT_DEFLATE 82 config ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
81 bool "Deflate" 83 bool "Deflate"
82 select CRYPTO_DEFLATE 84 select CRYPTO_DEFLATE
83 help 85 help
84 Use the Deflate algorithm as the def 86 Use the Deflate algorithm as the default compression algorithm.
85 87
86 config ZSWAP_COMPRESSOR_DEFAULT_LZO 88 config ZSWAP_COMPRESSOR_DEFAULT_LZO
87 bool "LZO" 89 bool "LZO"
88 select CRYPTO_LZO 90 select CRYPTO_LZO
89 help 91 help
90 Use the LZO algorithm as the default 92 Use the LZO algorithm as the default compression algorithm.
91 93
92 config ZSWAP_COMPRESSOR_DEFAULT_842 94 config ZSWAP_COMPRESSOR_DEFAULT_842
93 bool "842" 95 bool "842"
94 select CRYPTO_842 96 select CRYPTO_842
95 help 97 help
96 Use the 842 algorithm as the default 98 Use the 842 algorithm as the default compression algorithm.
97 99
98 config ZSWAP_COMPRESSOR_DEFAULT_LZ4 100 config ZSWAP_COMPRESSOR_DEFAULT_LZ4
99 bool "LZ4" 101 bool "LZ4"
100 select CRYPTO_LZ4 102 select CRYPTO_LZ4
101 help 103 help
102 Use the LZ4 algorithm as the default 104 Use the LZ4 algorithm as the default compression algorithm.
103 105
104 config ZSWAP_COMPRESSOR_DEFAULT_LZ4HC 106 config ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
105 bool "LZ4HC" 107 bool "LZ4HC"
106 select CRYPTO_LZ4HC 108 select CRYPTO_LZ4HC
107 help 109 help
108 Use the LZ4HC algorithm as the defau 110 Use the LZ4HC algorithm as the default compression algorithm.
109 111
110 config ZSWAP_COMPRESSOR_DEFAULT_ZSTD 112 config ZSWAP_COMPRESSOR_DEFAULT_ZSTD
111 bool "zstd" 113 bool "zstd"
112 select CRYPTO_ZSTD 114 select CRYPTO_ZSTD
113 help 115 help
114 Use the zstd algorithm as the defaul 116 Use the zstd algorithm as the default compression algorithm.
115 endchoice 117 endchoice
116 118
117 config ZSWAP_COMPRESSOR_DEFAULT 119 config ZSWAP_COMPRESSOR_DEFAULT
118 string 120 string
119 depends on ZSWAP 121 depends on ZSWAP
120 default "deflate" if ZSWAP_COMPRESSOR_D 122 default "deflate" if ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
121 default "lzo" if ZSWAP_COMPRESSOR_DEFAU 123 default "lzo" if ZSWAP_COMPRESSOR_DEFAULT_LZO
122 default "842" if ZSWAP_COMPRESSOR_DEFAU 124 default "842" if ZSWAP_COMPRESSOR_DEFAULT_842
123 default "lz4" if ZSWAP_COMPRESSOR_DEFAU 125 default "lz4" if ZSWAP_COMPRESSOR_DEFAULT_LZ4
124 default "lz4hc" if ZSWAP_COMPRESSOR_DEF 126 default "lz4hc" if ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
125 default "zstd" if ZSWAP_COMPRESSOR_DEFA 127 default "zstd" if ZSWAP_COMPRESSOR_DEFAULT_ZSTD
126 default "" 128 default ""
127 129
128 choice 130 choice
129 prompt "Default allocator" 131 prompt "Default allocator"
130 depends on ZSWAP 132 depends on ZSWAP
131 default ZSWAP_ZPOOL_DEFAULT_ZSMALLOC i <<
132 default ZSWAP_ZPOOL_DEFAULT_ZBUD 133 default ZSWAP_ZPOOL_DEFAULT_ZBUD
133 help 134 help
134 Selects the default allocator for th 135 Selects the default allocator for the compressed cache for
135 swap pages. 136 swap pages.
136 The default is 'zbud' for compatibil 137 The default is 'zbud' for compatibility, however please do
137 read the description of each of the 138 read the description of each of the allocators below before
138 making a right choice. 139 making a right choice.
139 140
140 The selection made here can be overr 141 The selection made here can be overridden by using the kernel
141 command line 'zswap.zpool=' option. 142 command line 'zswap.zpool=' option.
142 143
143 config ZSWAP_ZPOOL_DEFAULT_ZBUD 144 config ZSWAP_ZPOOL_DEFAULT_ZBUD
144 bool "zbud" 145 bool "zbud"
145 select ZBUD 146 select ZBUD
146 help 147 help
147 Use the zbud allocator as the defaul 148 Use the zbud allocator as the default allocator.
148 149
149 config ZSWAP_ZPOOL_DEFAULT_Z3FOLD_DEPRECATED 150 config ZSWAP_ZPOOL_DEFAULT_Z3FOLD_DEPRECATED
150 bool "z3foldi (DEPRECATED)" 151 bool "z3foldi (DEPRECATED)"
151 select Z3FOLD_DEPRECATED 152 select Z3FOLD_DEPRECATED
152 help 153 help
153 Use the z3fold allocator as the defa 154 Use the z3fold allocator as the default allocator.
154 155
155 Deprecated and scheduled for removal 156 Deprecated and scheduled for removal in a few cycles,
156 see CONFIG_Z3FOLD_DEPRECATED. 157 see CONFIG_Z3FOLD_DEPRECATED.
157 158
158 config ZSWAP_ZPOOL_DEFAULT_ZSMALLOC 159 config ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
159 bool "zsmalloc" 160 bool "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
202 config ZSMALLOC 203 config ZSMALLOC
203 tristate 204 tristate
204 prompt "N:1 compression allocator (zsm !! 205 prompt "N:1 compression allocator (zsmalloc)" if ZSWAP
205 depends on MMU 206 depends on MMU
206 help 207 help
207 zsmalloc is a slab-based memory allo 208 zsmalloc is a slab-based memory allocator designed to store
208 pages of various compression levels 209 pages of various compression levels efficiently. It achieves
209 the highest storage density with the 210 the highest storage density with the least amount of fragmentation.
210 211
211 config ZSMALLOC_STAT 212 config ZSMALLOC_STAT
212 bool "Export zsmalloc statistics" 213 bool "Export zsmalloc statistics"
213 depends on ZSMALLOC 214 depends on ZSMALLOC
214 select DEBUG_FS 215 select DEBUG_FS
215 help 216 help
216 This option enables code in the zsma 217 This option enables code in the zsmalloc to collect various
217 statistics about what's happening in 218 statistics about what's happening in zsmalloc and exports that
218 information to userspace via debugfs 219 information to userspace via debugfs.
219 If unsure, say N. 220 If unsure, say N.
220 221
221 config ZSMALLOC_CHAIN_SIZE 222 config ZSMALLOC_CHAIN_SIZE
222 int "Maximum number of physical pages 223 int "Maximum number of physical pages per-zspage"
223 default 8 224 default 8
224 range 4 16 225 range 4 16
225 depends on ZSMALLOC 226 depends on ZSMALLOC
226 help 227 help
227 This option sets the upper limit on 228 This option sets the upper limit on the number of physical pages
228 that a zmalloc page (zspage) can con 229 that a zmalloc page (zspage) can consist of. The optimal zspage
229 chain size is calculated for each si 230 chain size is calculated for each size class during the
230 initialization of the pool. 231 initialization of the pool.
231 232
232 Changing this option can alter the c 233 Changing this option can alter the characteristics of size classes,
233 such as the number of pages per zspa 234 such as the number of pages per zspage and the number of objects
234 per zspage. This can also result in 235 per zspage. This can also result in different configurations of
235 the pool, as zsmalloc merges size cl 236 the pool, as zsmalloc merges size classes with similar
236 characteristics. 237 characteristics.
237 238
238 For more information, see zsmalloc d 239 For more information, see zsmalloc documentation.
239 240
240 menu "Slab allocator options" !! 241 menu "SLAB allocator options"
>> 242
>> 243 choice
>> 244 prompt "Choose SLAB allocator"
>> 245 default SLUB
>> 246 help
>> 247 This option allows to select a slab allocator.
>> 248
>> 249 config SLAB_DEPRECATED
>> 250 bool "SLAB (DEPRECATED)"
>> 251 depends on !PREEMPT_RT
>> 252 help
>> 253 Deprecated and scheduled for removal in a few cycles. Replaced by
>> 254 SLUB.
>> 255
>> 256 If you cannot migrate to SLUB, please contact linux-mm@kvack.org
>> 257 and the people listed in the SLAB ALLOCATOR section of MAINTAINERS
>> 258 file, explaining why.
>> 259
>> 260 The regular slab allocator that is established and known to work
>> 261 well in all environments. It organizes cache hot objects in
>> 262 per cpu and per node queues.
241 263
242 config SLUB 264 config SLUB
243 def_bool y !! 265 bool "SLUB (Unqueued Allocator)"
>> 266 help
>> 267 SLUB is a slab allocator that minimizes cache line usage
>> 268 instead of managing queues of cached objects (SLAB approach).
>> 269 Per cpu caching is realized using slabs of objects instead
>> 270 of queues of objects. SLUB can use memory efficiently
>> 271 and has enhanced diagnostics. SLUB is the default choice for
>> 272 a slab allocator.
>> 273
>> 274 endchoice
>> 275
>> 276 config SLAB
>> 277 bool
>> 278 default y
>> 279 depends on SLAB_DEPRECATED
244 280
245 config SLUB_TINY 281 config SLUB_TINY
246 bool "Configure for minimal memory foo !! 282 bool "Configure SLUB for minimal memory footprint"
247 depends on EXPERT !! 283 depends on SLUB && EXPERT
248 select SLAB_MERGE_DEFAULT 284 select SLAB_MERGE_DEFAULT
249 help 285 help
250 Configures the slab allocator in a !! 286 Configures the SLUB allocator in a way to achieve minimal memory
251 footprint, sacrificing scalability, 287 footprint, sacrificing scalability, debugging and other features.
252 This is intended only for the small 288 This is intended only for the smallest system that had used the
253 SLOB allocator and is not recommend 289 SLOB allocator and is not recommended for systems with more than
254 16MB RAM. 290 16MB RAM.
255 291
256 If unsure, say N. 292 If unsure, say N.
257 293
258 config SLAB_MERGE_DEFAULT 294 config SLAB_MERGE_DEFAULT
259 bool "Allow slab caches to be merged" 295 bool "Allow slab caches to be merged"
260 default y 296 default y
>> 297 depends on SLAB || SLUB
261 help 298 help
262 For reduced kernel memory fragmentat 299 For reduced kernel memory fragmentation, slab caches can be
263 merged when they share the same size 300 merged when they share the same size and other characteristics.
264 This carries a risk of kernel heap o 301 This carries a risk of kernel heap overflows being able to
265 overwrite objects from merged caches 302 overwrite objects from merged caches (and more easily control
266 cache layout), which makes such heap 303 cache layout), which makes such heap attacks easier to exploit
267 by attackers. By keeping caches unme 304 by attackers. By keeping caches unmerged, these kinds of exploits
268 can usually only damage objects in t 305 can usually only damage objects in the same cache. To disable
269 merging at runtime, "slab_nomerge" c 306 merging at runtime, "slab_nomerge" can be passed on the kernel
270 command line. 307 command line.
271 308
272 config SLAB_FREELIST_RANDOM 309 config SLAB_FREELIST_RANDOM
273 bool "Randomize slab freelist" 310 bool "Randomize slab freelist"
274 depends on !SLUB_TINY !! 311 depends on SLAB || (SLUB && !SLUB_TINY)
275 help 312 help
276 Randomizes the freelist order used o 313 Randomizes the freelist order used on creating new pages. This
277 security feature reduces the predict 314 security feature reduces the predictability of the kernel slab
278 allocator against heap overflows. 315 allocator against heap overflows.
279 316
280 config SLAB_FREELIST_HARDENED 317 config SLAB_FREELIST_HARDENED
281 bool "Harden slab freelist metadata" 318 bool "Harden slab freelist metadata"
282 depends on !SLUB_TINY !! 319 depends on SLAB || (SLUB && !SLUB_TINY)
283 help 320 help
284 Many kernel heap attacks try to targ 321 Many kernel heap attacks try to target slab cache metadata and
285 other infrastructure. This options m 322 other infrastructure. This options makes minor performance
286 sacrifices to harden the kernel slab 323 sacrifices to harden the kernel slab allocator against common
287 freelist exploit methods. !! 324 freelist exploit methods. Some slab implementations have more
288 !! 325 sanity-checking than others. This option is most effective with
289 config SLAB_BUCKETS !! 326 CONFIG_SLUB.
290 bool "Support allocation from separate <<
291 depends on !SLUB_TINY <<
292 default SLAB_FREELIST_HARDENED <<
293 help <<
294 Kernel heap attacks frequently depen <<
295 specifically-sized allocations with <<
296 that will be allocated into the same <<
297 target object. To avoid sharing thes <<
298 provide an explicitly separated set <<
299 user-controlled allocations. This ma <<
300 memory fragmentation, though in prac <<
301 of extra pages since the bulk of use <<
302 are relatively long-lived. <<
303 <<
304 If unsure, say Y. <<
305 327
306 config SLUB_STATS 328 config SLUB_STATS
307 default n 329 default n
308 bool "Enable performance statistics" !! 330 bool "Enable SLUB performance statistics"
309 depends on SYSFS && !SLUB_TINY !! 331 depends on SLUB && SYSFS && !SLUB_TINY
310 help 332 help
311 The statistics are useful to debug s !! 333 SLUB statistics are useful to debug SLUBs allocation behavior in
312 order find ways to optimize the allo 334 order find ways to optimize the allocator. This should never be
313 enabled for production use since kee 335 enabled for production use since keeping statistics slows down
314 the allocator by a few percentage po 336 the allocator by a few percentage points. The slabinfo command
315 supports the determination of the mo 337 supports the determination of the most active slabs to figure
316 out which slabs are relevant to a pa 338 out which slabs are relevant to a particular load.
317 Try running: slabinfo -DA 339 Try running: slabinfo -DA
318 340
319 config SLUB_CPU_PARTIAL 341 config SLUB_CPU_PARTIAL
320 default y 342 default y
321 depends on SMP && !SLUB_TINY !! 343 depends on SLUB && SMP && !SLUB_TINY
322 bool "Enable per cpu partial caches" !! 344 bool "SLUB per cpu partial cache"
323 help 345 help
324 Per cpu partial caches accelerate ob 346 Per cpu partial caches accelerate objects allocation and freeing
325 that is local to a processor at the 347 that is local to a processor at the price of more indeterminism
326 in the latency of the free. On overf 348 in the latency of the free. On overflow these caches will be cleared
327 which requires the taking of locks t 349 which requires the taking of locks that may cause latency spikes.
328 Typically one would choose no for a 350 Typically one would choose no for a realtime system.
329 351
330 config RANDOM_KMALLOC_CACHES 352 config RANDOM_KMALLOC_CACHES
331 default n 353 default n
332 depends on !SLUB_TINY !! 354 depends on SLUB && !SLUB_TINY
333 bool "Randomize slab caches for normal 355 bool "Randomize slab caches for normal kmalloc"
334 help 356 help
335 A hardening feature that creates mul 357 A hardening feature that creates multiple copies of slab caches for
336 normal kmalloc allocation and makes 358 normal kmalloc allocation and makes kmalloc randomly pick one based
337 on code address, which makes the att 359 on code address, which makes the attackers more difficult to spray
338 vulnerable memory objects on the hea 360 vulnerable memory objects on the heap for the purpose of exploiting
339 memory vulnerabilities. 361 memory vulnerabilities.
340 362
341 Currently the number of copies is se 363 Currently the number of copies is set to 16, a reasonably large value
342 that effectively diverges the memory 364 that effectively diverges the memory objects allocated for different
343 subsystems or modules into different 365 subsystems or modules into different caches, at the expense of a
344 limited degree of memory and CPU ove 366 limited degree of memory and CPU overhead that relates to hardware and
345 system workload. 367 system workload.
346 368
347 endmenu # Slab allocator options !! 369 endmenu # SLAB allocator options
348 370
349 config SHUFFLE_PAGE_ALLOCATOR 371 config SHUFFLE_PAGE_ALLOCATOR
350 bool "Page allocator randomization" 372 bool "Page allocator randomization"
351 default SLAB_FREELIST_RANDOM && ACPI_N 373 default SLAB_FREELIST_RANDOM && ACPI_NUMA
352 help 374 help
353 Randomization of the page allocator 375 Randomization of the page allocator improves the average
354 utilization of a direct-mapped memor 376 utilization of a direct-mapped memory-side-cache. See section
355 5.2.27 Heterogeneous Memory Attribut 377 5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI
356 6.2a specification for an example of 378 6.2a specification for an example of how a platform advertises
357 the presence of a memory-side-cache. 379 the presence of a memory-side-cache. There are also incidental
358 security benefits as it reduces the 380 security benefits as it reduces the predictability of page
359 allocations to compliment SLAB_FREEL 381 allocations to compliment SLAB_FREELIST_RANDOM, but the
360 default granularity of shuffling on !! 382 default granularity of shuffling on the MAX_ORDER i.e, 10th
361 order of pages is selected based on 383 order of pages is selected based on cache utilization benefits
362 on x86. 384 on x86.
363 385
364 While the randomization improves cac 386 While the randomization improves cache utilization it may
365 negatively impact workloads on platf 387 negatively impact workloads on platforms without a cache. For
366 this reason, by default, the randomi !! 388 this reason, by default, the randomization is enabled only
367 if SHUFFLE_PAGE_ALLOCATOR=y. The ran !! 389 after runtime detection of a direct-mapped memory-side-cache.
368 with the 'page_alloc.shuffle' kernel !! 390 Otherwise, the randomization may be force enabled with the
>> 391 'page_alloc.shuffle' kernel command line parameter.
369 392
370 Say Y if unsure. 393 Say Y if unsure.
371 394
372 config COMPAT_BRK 395 config COMPAT_BRK
373 bool "Disable heap randomization" 396 bool "Disable heap randomization"
374 default y 397 default y
375 help 398 help
376 Randomizing heap placement makes hea 399 Randomizing heap placement makes heap exploits harder, but it
377 also breaks ancient binaries (includ 400 also breaks ancient binaries (including anything libc5 based).
378 This option changes the bootup defau 401 This option changes the bootup default to heap randomization
379 disabled, and can be overridden at r 402 disabled, and can be overridden at runtime by setting
380 /proc/sys/kernel/randomize_va_space 403 /proc/sys/kernel/randomize_va_space to 2.
381 404
382 On non-ancient distros (post-2000 on 405 On non-ancient distros (post-2000 ones) N is usually a safe choice.
383 406
384 config MMAP_ALLOW_UNINITIALIZED 407 config MMAP_ALLOW_UNINITIALIZED
385 bool "Allow mmapped anonymous memory t 408 bool "Allow mmapped anonymous memory to be uninitialized"
386 depends on EXPERT && !MMU 409 depends on EXPERT && !MMU
387 default n 410 default n
388 help 411 help
389 Normally, and according to the Linux 412 Normally, and according to the Linux spec, anonymous memory obtained
390 from mmap() has its contents cleared 413 from mmap() has its contents cleared before it is passed to
391 userspace. Enabling this config opt 414 userspace. Enabling this config option allows you to request that
392 mmap() skip that if it is given an M 415 mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus
393 providing a huge performance boost. 416 providing a huge performance boost. If this option is not enabled,
394 then the flag will be ignored. 417 then the flag will be ignored.
395 418
396 This is taken advantage of by uClibc 419 This is taken advantage of by uClibc's malloc(), and also by
397 ELF-FDPIC binfmt's brk and stack all 420 ELF-FDPIC binfmt's brk and stack allocator.
398 421
399 Because of the obvious security issu 422 Because of the obvious security issues, this option should only be
400 enabled on embedded devices where yo 423 enabled on embedded devices where you control what is run in
401 userspace. Since that isn't general 424 userspace. Since that isn't generally a problem on no-MMU systems,
402 it is normally safe to say Y here. 425 it is normally safe to say Y here.
403 426
404 See Documentation/admin-guide/mm/nom 427 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
405 428
406 config SELECT_MEMORY_MODEL 429 config SELECT_MEMORY_MODEL
407 def_bool y 430 def_bool y
408 depends on ARCH_SELECT_MEMORY_MODEL 431 depends on ARCH_SELECT_MEMORY_MODEL
409 432
410 choice 433 choice
411 prompt "Memory model" 434 prompt "Memory model"
412 depends on SELECT_MEMORY_MODEL 435 depends on SELECT_MEMORY_MODEL
413 default SPARSEMEM_MANUAL if ARCH_SPARS 436 default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT
414 default FLATMEM_MANUAL 437 default FLATMEM_MANUAL
415 help 438 help
416 This option allows you to change som 439 This option allows you to change some of the ways that
417 Linux manages its memory internally. 440 Linux manages its memory internally. Most users will
418 only have one option here selected b 441 only have one option here selected by the architecture
419 configuration. This is normal. 442 configuration. This is normal.
420 443
421 config FLATMEM_MANUAL 444 config FLATMEM_MANUAL
422 bool "Flat Memory" 445 bool "Flat Memory"
423 depends on !ARCH_SPARSEMEM_ENABLE || A 446 depends on !ARCH_SPARSEMEM_ENABLE || ARCH_FLATMEM_ENABLE
424 help 447 help
425 This option is best suited for non-N 448 This option is best suited for non-NUMA systems with
426 flat address space. The FLATMEM is t 449 flat address space. The FLATMEM is the most efficient
427 system in terms of performance and r 450 system in terms of performance and resource consumption
428 and it is the best option for smalle 451 and it is the best option for smaller systems.
429 452
430 For systems that have holes in their 453 For systems that have holes in their physical address
431 spaces and for features like NUMA an 454 spaces and for features like NUMA and memory hotplug,
432 choose "Sparse Memory". 455 choose "Sparse Memory".
433 456
434 If unsure, choose this option (Flat 457 If unsure, choose this option (Flat Memory) over any other.
435 458
436 config SPARSEMEM_MANUAL 459 config SPARSEMEM_MANUAL
437 bool "Sparse Memory" 460 bool "Sparse Memory"
438 depends on ARCH_SPARSEMEM_ENABLE 461 depends on ARCH_SPARSEMEM_ENABLE
439 help 462 help
440 This will be the only option for som 463 This will be the only option for some systems, including
441 memory hot-plug systems. This is no 464 memory hot-plug systems. This is normal.
442 465
443 This option provides efficient suppo 466 This option provides efficient support for systems with
444 holes is their physical address spac 467 holes is their physical address space and allows memory
445 hot-plug and hot-remove. 468 hot-plug and hot-remove.
446 469
447 If unsure, choose "Flat Memory" over 470 If unsure, choose "Flat Memory" over this option.
448 471
449 endchoice 472 endchoice
450 473
451 config SPARSEMEM 474 config SPARSEMEM
452 def_bool y 475 def_bool y
453 depends on (!SELECT_MEMORY_MODEL && AR 476 depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL
454 477
455 config FLATMEM 478 config FLATMEM
456 def_bool y 479 def_bool y
457 depends on !SPARSEMEM || FLATMEM_MANUA 480 depends on !SPARSEMEM || FLATMEM_MANUAL
458 481
459 # 482 #
460 # SPARSEMEM_EXTREME (which is the default) doe 483 # SPARSEMEM_EXTREME (which is the default) does some bootmem
461 # allocations when sparse_init() is called. I 484 # allocations when sparse_init() is called. If this cannot
462 # be done on your architecture, select this op 485 # be done on your architecture, select this option. However,
463 # statically allocating the mem_section[] arra 486 # statically allocating the mem_section[] array can potentially
464 # consume vast quantities of .bss, so be caref 487 # consume vast quantities of .bss, so be careful.
465 # 488 #
466 # This option will also potentially produce sm 489 # This option will also potentially produce smaller runtime code
467 # with gcc 3.4 and later. 490 # with gcc 3.4 and later.
468 # 491 #
469 config SPARSEMEM_STATIC 492 config SPARSEMEM_STATIC
470 bool 493 bool
471 494
472 # 495 #
473 # Architecture platforms which require a two l 496 # Architecture platforms which require a two level mem_section in SPARSEMEM
474 # must select this option. This is usually for 497 # must select this option. This is usually for architecture platforms with
475 # an extremely sparse physical address space. 498 # an extremely sparse physical address space.
476 # 499 #
477 config SPARSEMEM_EXTREME 500 config SPARSEMEM_EXTREME
478 def_bool y 501 def_bool y
479 depends on SPARSEMEM && !SPARSEMEM_STA 502 depends on SPARSEMEM && !SPARSEMEM_STATIC
480 503
481 config SPARSEMEM_VMEMMAP_ENABLE 504 config SPARSEMEM_VMEMMAP_ENABLE
482 bool 505 bool
483 506
484 config SPARSEMEM_VMEMMAP 507 config SPARSEMEM_VMEMMAP
485 bool "Sparse Memory virtual memmap" 508 bool "Sparse Memory virtual memmap"
486 depends on SPARSEMEM && SPARSEMEM_VMEM 509 depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE
487 default y 510 default y
488 help 511 help
489 SPARSEMEM_VMEMMAP uses a virtually m 512 SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise
490 pfn_to_page and page_to_pfn operatio 513 pfn_to_page and page_to_pfn operations. This is the most
491 efficient option when sufficient ker 514 efficient option when sufficient kernel resources are available.
492 # 515 #
493 # Select this config option from the architect 516 # Select this config option from the architecture Kconfig, if it is preferred
494 # to enable the feature of HugeTLB/dev_dax vme 517 # to enable the feature of HugeTLB/dev_dax vmemmap optimization.
495 # 518 #
496 config ARCH_WANT_OPTIMIZE_DAX_VMEMMAP 519 config ARCH_WANT_OPTIMIZE_DAX_VMEMMAP
497 bool 520 bool
498 521
499 config ARCH_WANT_OPTIMIZE_HUGETLB_VMEMMAP 522 config ARCH_WANT_OPTIMIZE_HUGETLB_VMEMMAP
500 bool 523 bool
501 524
502 config HAVE_MEMBLOCK_PHYS_MAP 525 config HAVE_MEMBLOCK_PHYS_MAP
503 bool 526 bool
504 527
505 config HAVE_GUP_FAST !! 528 config HAVE_FAST_GUP
506 depends on MMU 529 depends on MMU
507 bool 530 bool
508 531
509 # Don't discard allocated memory used to track 532 # Don't discard allocated memory used to track "memory" and "reserved" memblocks
510 # after early boot, so it can still be used to 533 # after early boot, so it can still be used to test for validity of memory.
511 # Also, memblocks are updated with memory hot( 534 # Also, memblocks are updated with memory hot(un)plug.
512 config ARCH_KEEP_MEMBLOCK 535 config ARCH_KEEP_MEMBLOCK
513 bool 536 bool
514 537
515 # Keep arch NUMA mapping infrastructure post-i 538 # Keep arch NUMA mapping infrastructure post-init.
516 config NUMA_KEEP_MEMINFO 539 config NUMA_KEEP_MEMINFO
517 bool 540 bool
518 541
519 config MEMORY_ISOLATION 542 config MEMORY_ISOLATION
520 bool 543 bool
521 544
522 # IORESOURCE_SYSTEM_RAM regions in the kernel 545 # IORESOURCE_SYSTEM_RAM regions in the kernel resource tree that are marked
523 # IORESOURCE_EXCLUSIVE cannot be mapped to use 546 # IORESOURCE_EXCLUSIVE cannot be mapped to user space, for example, via
524 # /dev/mem. 547 # /dev/mem.
525 config EXCLUSIVE_SYSTEM_RAM 548 config EXCLUSIVE_SYSTEM_RAM
526 def_bool y 549 def_bool y
527 depends on !DEVMEM || STRICT_DEVMEM 550 depends on !DEVMEM || STRICT_DEVMEM
528 551
529 # 552 #
530 # Only be set on architectures that have compl 553 # Only be set on architectures that have completely implemented memory hotplug
531 # feature. If you are not sure, don't touch it 554 # feature. If you are not sure, don't touch it.
532 # 555 #
533 config HAVE_BOOTMEM_INFO_NODE 556 config HAVE_BOOTMEM_INFO_NODE
534 def_bool n 557 def_bool n
535 558
536 config ARCH_ENABLE_MEMORY_HOTPLUG 559 config ARCH_ENABLE_MEMORY_HOTPLUG
537 bool 560 bool
538 561
539 config ARCH_ENABLE_MEMORY_HOTREMOVE 562 config ARCH_ENABLE_MEMORY_HOTREMOVE
540 bool 563 bool
541 564
542 # eventually, we can have this option just 'se 565 # eventually, we can have this option just 'select SPARSEMEM'
543 menuconfig MEMORY_HOTPLUG 566 menuconfig MEMORY_HOTPLUG
544 bool "Memory hotplug" 567 bool "Memory hotplug"
545 select MEMORY_ISOLATION 568 select MEMORY_ISOLATION
546 depends on SPARSEMEM 569 depends on SPARSEMEM
547 depends on ARCH_ENABLE_MEMORY_HOTPLUG 570 depends on ARCH_ENABLE_MEMORY_HOTPLUG
548 depends on 64BIT 571 depends on 64BIT
549 select NUMA_KEEP_MEMINFO if NUMA 572 select NUMA_KEEP_MEMINFO if NUMA
550 573
551 if MEMORY_HOTPLUG 574 if MEMORY_HOTPLUG
552 575
553 config MEMORY_HOTPLUG_DEFAULT_ONLINE 576 config MEMORY_HOTPLUG_DEFAULT_ONLINE
554 bool "Online the newly added memory bl 577 bool "Online the newly added memory blocks by default"
555 depends on MEMORY_HOTPLUG 578 depends on MEMORY_HOTPLUG
556 help 579 help
557 This option sets the default policy 580 This option sets the default policy setting for memory hotplug
558 onlining policy (/sys/devices/system 581 onlining policy (/sys/devices/system/memory/auto_online_blocks) which
559 determines what happens to newly add 582 determines what happens to newly added memory regions. Policy setting
560 can always be changed at runtime. 583 can always be changed at runtime.
561 See Documentation/admin-guide/mm/mem 584 See Documentation/admin-guide/mm/memory-hotplug.rst for more information.
562 585
563 Say Y here if you want all hot-plugg 586 Say Y here if you want all hot-plugged memory blocks to appear in
564 'online' state by default. 587 'online' state by default.
565 Say N here if you want the default p 588 Say N here if you want the default policy to keep all hot-plugged
566 memory blocks in 'offline' state. 589 memory blocks in 'offline' state.
567 590
568 config MEMORY_HOTREMOVE 591 config MEMORY_HOTREMOVE
569 bool "Allow for memory hot remove" 592 bool "Allow for memory hot remove"
570 select HAVE_BOOTMEM_INFO_NODE if (X86_ 593 select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64)
571 depends on MEMORY_HOTPLUG && ARCH_ENAB 594 depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE
572 depends on MIGRATION 595 depends on MIGRATION
573 596
574 config MHP_MEMMAP_ON_MEMORY 597 config MHP_MEMMAP_ON_MEMORY
575 def_bool y 598 def_bool y
576 depends on MEMORY_HOTPLUG && SPARSEMEM 599 depends on MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP
577 depends on ARCH_MHP_MEMMAP_ON_MEMORY_E 600 depends on ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
578 601
579 endif # MEMORY_HOTPLUG 602 endif # MEMORY_HOTPLUG
580 603
581 config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE 604 config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
582 bool 605 bool
583 606
584 # Heavily threaded applications may benefit fr 607 # Heavily threaded applications may benefit from splitting the mm-wide
585 # page_table_lock, so that faults on different 608 # page_table_lock, so that faults on different parts of the user address
586 # space can be handled with less contention: s 609 # space can be handled with less contention: split it at this NR_CPUS.
587 # Default to 4 for wider testing, though 8 mig 610 # Default to 4 for wider testing, though 8 might be more appropriate.
588 # ARM's adjust_pte (unused if VIPT) depends on 611 # ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
589 # PA-RISC 7xxx's spinlock_t would enlarge stru 612 # PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
590 # SPARC32 allocates multiple pte tables within 613 # SPARC32 allocates multiple pte tables within a single page, and therefore
591 # a per-page lock leads to problems when multi 614 # a per-page lock leads to problems when multiple tables need to be locked
592 # at the same time (e.g. copy_page_range()). 615 # at the same time (e.g. copy_page_range()).
593 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock 616 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
594 # 617 #
595 config SPLIT_PTE_PTLOCKS !! 618 config SPLIT_PTLOCK_CPUS
596 def_bool y !! 619 int
597 depends on MMU !! 620 default "999999" if !MMU
598 depends on SMP !! 621 default "999999" if ARM && !CPU_CACHE_VIPT
599 depends on NR_CPUS >= 4 !! 622 default "999999" if PARISC && !PA20
600 depends on !ARM || CPU_CACHE_VIPT !! 623 default "999999" if SPARC32
601 depends on !PARISC || PA20 !! 624 default "4"
602 depends on !SPARC32 <<
603 625
604 config ARCH_ENABLE_SPLIT_PMD_PTLOCK 626 config ARCH_ENABLE_SPLIT_PMD_PTLOCK
605 bool 627 bool
606 628
607 config SPLIT_PMD_PTLOCKS <<
608 def_bool y <<
609 depends on SPLIT_PTE_PTLOCKS && ARCH_E <<
610 <<
611 # 629 #
612 # support for memory balloon 630 # support for memory balloon
613 config MEMORY_BALLOON 631 config MEMORY_BALLOON
614 bool 632 bool
615 633
616 # 634 #
617 # support for memory balloon compaction 635 # support for memory balloon compaction
618 config BALLOON_COMPACTION 636 config BALLOON_COMPACTION
619 bool "Allow for balloon memory compact 637 bool "Allow for balloon memory compaction/migration"
620 default y !! 638 def_bool y
621 depends on COMPACTION && MEMORY_BALLOO 639 depends on COMPACTION && MEMORY_BALLOON
622 help 640 help
623 Memory fragmentation introduced by b 641 Memory fragmentation introduced by ballooning might reduce
624 significantly the number of 2MB cont 642 significantly the number of 2MB contiguous memory blocks that can be
625 used within a guest, thus imposing p 643 used within a guest, thus imposing performance penalties associated
626 with the reduced number of transpare 644 with the reduced number of transparent huge pages that could be used
627 by the guest workload. Allowing the 645 by the guest workload. Allowing the compaction & migration for memory
628 pages enlisted as being part of memo 646 pages enlisted as being part of memory balloon devices avoids the
629 scenario aforementioned and helps im 647 scenario aforementioned and helps improving memory defragmentation.
630 648
631 # 649 #
632 # support for memory compaction 650 # support for memory compaction
633 config COMPACTION 651 config COMPACTION
634 bool "Allow for memory compaction" 652 bool "Allow for memory compaction"
635 default y !! 653 def_bool y
636 select MIGRATION 654 select MIGRATION
637 depends on MMU 655 depends on MMU
638 help 656 help
639 Compaction is the only memory manage 657 Compaction is the only memory management component to form
640 high order (larger physically contig 658 high order (larger physically contiguous) memory blocks
641 reliably. The page allocator relies 659 reliably. The page allocator relies on compaction heavily and
642 the lack of the feature can lead to 660 the lack of the feature can lead to unexpected OOM killer
643 invocations for high order memory re 661 invocations for high order memory requests. You shouldn't
644 disable this option unless there rea 662 disable this option unless there really is a strong reason for
645 it and then we would be really inter 663 it and then we would be really interested to hear about that at
646 linux-mm@kvack.org. 664 linux-mm@kvack.org.
647 665
648 config COMPACT_UNEVICTABLE_DEFAULT 666 config COMPACT_UNEVICTABLE_DEFAULT
649 int 667 int
650 depends on COMPACTION 668 depends on COMPACTION
651 default 0 if PREEMPT_RT 669 default 0 if PREEMPT_RT
652 default 1 670 default 1
653 671
654 # 672 #
655 # support for free page reporting 673 # support for free page reporting
656 config PAGE_REPORTING 674 config PAGE_REPORTING
657 bool "Free page reporting" 675 bool "Free page reporting"
>> 676 def_bool n
658 help 677 help
659 Free page reporting allows for the i 678 Free page reporting allows for the incremental acquisition of
660 free pages from the buddy allocator 679 free pages from the buddy allocator for the purpose of reporting
661 those pages to another entity, such 680 those pages to another entity, such as a hypervisor, so that the
662 memory can be freed within the host 681 memory can be freed within the host for other uses.
663 682
664 # 683 #
665 # support for page migration 684 # support for page migration
666 # 685 #
667 config MIGRATION 686 config MIGRATION
668 bool "Page migration" 687 bool "Page migration"
669 default y !! 688 def_bool y
670 depends on (NUMA || ARCH_ENABLE_MEMORY 689 depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU
671 help 690 help
672 Allows the migration of the physical 691 Allows the migration of the physical location of pages of processes
673 while the virtual addresses are not 692 while the virtual addresses are not changed. This is useful in
674 two situations. The first is on NUMA 693 two situations. The first is on NUMA systems to put pages nearer
675 to the processors accessing. The sec 694 to the processors accessing. The second is when allocating huge
676 pages as migration can relocate page 695 pages as migration can relocate pages to satisfy a huge page
677 allocation instead of reclaiming. 696 allocation instead of reclaiming.
678 697
679 config DEVICE_MIGRATION 698 config DEVICE_MIGRATION
680 def_bool MIGRATION && ZONE_DEVICE 699 def_bool MIGRATION && ZONE_DEVICE
681 700
682 config ARCH_ENABLE_HUGEPAGE_MIGRATION 701 config ARCH_ENABLE_HUGEPAGE_MIGRATION
683 bool 702 bool
684 703
685 config ARCH_ENABLE_THP_MIGRATION 704 config ARCH_ENABLE_THP_MIGRATION
686 bool 705 bool
687 706
688 config HUGETLB_PAGE_SIZE_VARIABLE 707 config HUGETLB_PAGE_SIZE_VARIABLE
689 def_bool n 708 def_bool n
690 help 709 help
691 Allows the pageblock_order value to 710 Allows the pageblock_order value to be dynamic instead of just standard
692 HUGETLB_PAGE_ORDER when there are mu 711 HUGETLB_PAGE_ORDER when there are multiple HugeTLB page sizes available
693 on a platform. 712 on a platform.
694 713
695 Note that the pageblock_order cannot !! 714 Note that the pageblock_order cannot exceed MAX_ORDER and will be
696 clamped down to MAX_PAGE_ORDER. !! 715 clamped down to MAX_ORDER.
697 716
698 config CONTIG_ALLOC 717 config CONTIG_ALLOC
699 def_bool (MEMORY_ISOLATION && COMPACTI 718 def_bool (MEMORY_ISOLATION && COMPACTION) || CMA
700 719
701 config PCP_BATCH_SCALE_MAX 720 config PCP_BATCH_SCALE_MAX
702 int "Maximum scale factor of PCP (Per- 721 int "Maximum scale factor of PCP (Per-CPU pageset) batch allocate/free"
703 default 5 722 default 5
704 range 0 6 723 range 0 6
705 help 724 help
706 In page allocator, PCP (Per-CPU page 725 In page allocator, PCP (Per-CPU pageset) is refilled and drained in
707 batches. The batch number is scaled 726 batches. The batch number is scaled automatically to improve page
708 allocation/free throughput. But too 727 allocation/free throughput. But too large scale factor may hurt
709 latency. This option sets the upper 728 latency. This option sets the upper limit of scale factor to limit
710 the maximum latency. 729 the maximum latency.
711 730
712 config PHYS_ADDR_T_64BIT 731 config PHYS_ADDR_T_64BIT
713 def_bool 64BIT 732 def_bool 64BIT
714 733
715 config BOUNCE 734 config BOUNCE
716 bool "Enable bounce buffers" 735 bool "Enable bounce buffers"
717 default y 736 default y
718 depends on BLOCK && MMU && HIGHMEM 737 depends on BLOCK && MMU && HIGHMEM
719 help 738 help
720 Enable bounce buffers for devices th 739 Enable bounce buffers for devices that cannot access the full range of
721 memory available to the CPU. Enabled 740 memory available to the CPU. Enabled by default when HIGHMEM is
722 selected, but you may say n to overr 741 selected, but you may say n to override this.
723 742
724 config MMU_NOTIFIER 743 config MMU_NOTIFIER
725 bool 744 bool
726 select INTERVAL_TREE 745 select INTERVAL_TREE
727 746
728 config KSM 747 config KSM
729 bool "Enable KSM for page merging" 748 bool "Enable KSM for page merging"
730 depends on MMU 749 depends on MMU
731 select XXHASH 750 select XXHASH
732 help 751 help
733 Enable Kernel Samepage Merging: KSM 752 Enable Kernel Samepage Merging: KSM periodically scans those areas
734 of an application's address space th 753 of an application's address space that an app has advised may be
735 mergeable. When it finds pages of i 754 mergeable. When it finds pages of identical content, it replaces
736 the many instances by a single page 755 the many instances by a single page with that content, so
737 saving memory until one or another a 756 saving memory until one or another app needs to modify the content.
738 Recommended for use with KVM, or wit 757 Recommended for use with KVM, or with other duplicative applications.
739 See Documentation/mm/ksm.rst for mor 758 See Documentation/mm/ksm.rst for more information: KSM is inactive
740 until a program has madvised that an 759 until a program has madvised that an area is MADV_MERGEABLE, and
741 root has set /sys/kernel/mm/ksm/run 760 root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set).
742 761
743 config DEFAULT_MMAP_MIN_ADDR 762 config DEFAULT_MMAP_MIN_ADDR
744 int "Low address space to protect from 763 int "Low address space to protect from user allocation"
745 depends on MMU 764 depends on MMU
746 default 4096 765 default 4096
747 help 766 help
748 This is the portion of low virtual m 767 This is the portion of low virtual memory which should be protected
749 from userspace allocation. Keeping 768 from userspace allocation. Keeping a user from writing to low pages
750 can help reduce the impact of kernel 769 can help reduce the impact of kernel NULL pointer bugs.
751 770
752 For most arm64, ppc64 and x86 users !! 771 For most ia64, ppc64 and x86 users with lots of address space
753 a value of 65536 is reasonable and s 772 a value of 65536 is reasonable and should cause no problems.
754 On arm and other archs it should not 773 On arm and other archs it should not be higher than 32768.
755 Programs which use vm86 functionalit 774 Programs which use vm86 functionality or have some need to map
756 this low address space will need CAP 775 this low address space will need CAP_SYS_RAWIO or disable this
757 protection by setting the value to 0 776 protection by setting the value to 0.
758 777
759 This value can be changed after boot 778 This value can be changed after boot using the
760 /proc/sys/vm/mmap_min_addr tunable. 779 /proc/sys/vm/mmap_min_addr tunable.
761 780
762 config ARCH_SUPPORTS_MEMORY_FAILURE 781 config ARCH_SUPPORTS_MEMORY_FAILURE
763 bool 782 bool
764 783
765 config MEMORY_FAILURE 784 config MEMORY_FAILURE
766 depends on MMU 785 depends on MMU
767 depends on ARCH_SUPPORTS_MEMORY_FAILUR 786 depends on ARCH_SUPPORTS_MEMORY_FAILURE
768 bool "Enable recovery from hardware me 787 bool "Enable recovery from hardware memory errors"
769 select MEMORY_ISOLATION 788 select MEMORY_ISOLATION
770 select RAS 789 select RAS
771 help 790 help
772 Enables code to recover from some me 791 Enables code to recover from some memory failures on systems
773 with MCA recovery. This allows a sys 792 with MCA recovery. This allows a system to continue running
774 even when some of its memory has unc 793 even when some of its memory has uncorrected errors. This requires
775 special hardware support and typical 794 special hardware support and typically ECC memory.
776 795
777 config HWPOISON_INJECT 796 config HWPOISON_INJECT
778 tristate "HWPoison pages injector" 797 tristate "HWPoison pages injector"
779 depends on MEMORY_FAILURE && DEBUG_KER 798 depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS
780 select PROC_PAGE_MONITOR 799 select PROC_PAGE_MONITOR
781 800
782 config NOMMU_INITIAL_TRIM_EXCESS 801 config NOMMU_INITIAL_TRIM_EXCESS
783 int "Turn on mmap() excess space trimm 802 int "Turn on mmap() excess space trimming before booting"
784 depends on !MMU 803 depends on !MMU
785 default 1 804 default 1
786 help 805 help
787 The NOMMU mmap() frequently needs to 806 The NOMMU mmap() frequently needs to allocate large contiguous chunks
788 of memory on which to store mappings 807 of memory on which to store mappings, but it can only ask the system
789 allocator for chunks in 2^N*PAGE_SIZ 808 allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently
790 more than it requires. To deal with 809 more than it requires. To deal with this, mmap() is able to trim off
791 the excess and return it to the allo 810 the excess and return it to the allocator.
792 811
793 If trimming is enabled, the excess i 812 If trimming is enabled, the excess is trimmed off and returned to the
794 system allocator, which can cause ex 813 system allocator, which can cause extra fragmentation, particularly
795 if there are a lot of transient proc 814 if there are a lot of transient processes.
796 815
797 If trimming is disabled, the excess 816 If trimming is disabled, the excess is kept, but not used, which for
798 long-term mappings means that the sp 817 long-term mappings means that the space is wasted.
799 818
800 Trimming can be dynamically controll 819 Trimming can be dynamically controlled through a sysctl option
801 (/proc/sys/vm/nr_trim_pages) which s 820 (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of
802 excess pages there must be before tr 821 excess pages there must be before trimming should occur, or zero if
803 no trimming is to occur. 822 no trimming is to occur.
804 823
805 This option specifies the initial va 824 This option specifies the initial value of this option. The default
806 of 1 says that all excess pages shou 825 of 1 says that all excess pages should be trimmed.
807 826
808 See Documentation/admin-guide/mm/nom 827 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
809 828
810 config ARCH_WANT_GENERAL_HUGETLB 829 config ARCH_WANT_GENERAL_HUGETLB
811 bool 830 bool
812 831
813 config ARCH_WANTS_THP_SWAP 832 config ARCH_WANTS_THP_SWAP
814 def_bool n 833 def_bool n
815 834
816 menuconfig TRANSPARENT_HUGEPAGE 835 menuconfig TRANSPARENT_HUGEPAGE
817 bool "Transparent Hugepage Support" 836 bool "Transparent Hugepage Support"
818 depends on HAVE_ARCH_TRANSPARENT_HUGEP 837 depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT
819 select COMPACTION 838 select COMPACTION
820 select XARRAY_MULTI 839 select XARRAY_MULTI
821 help 840 help
822 Transparent Hugepages allows the ker 841 Transparent Hugepages allows the kernel to use huge pages and
823 huge tlb transparently to the applic 842 huge tlb transparently to the applications whenever possible.
824 This feature can improve computing p 843 This feature can improve computing performance to certain
825 applications by speeding up page fau 844 applications by speeding up page faults during memory
826 allocation, by reducing the number o 845 allocation, by reducing the number of tlb misses and by speeding
827 up the pagetable walking. 846 up the pagetable walking.
828 847
829 If memory constrained on embedded, y 848 If memory constrained on embedded, you may want to say N.
830 849
831 if TRANSPARENT_HUGEPAGE 850 if TRANSPARENT_HUGEPAGE
832 851
833 choice 852 choice
834 prompt "Transparent Hugepage Support s 853 prompt "Transparent Hugepage Support sysfs defaults"
835 depends on TRANSPARENT_HUGEPAGE 854 depends on TRANSPARENT_HUGEPAGE
836 default TRANSPARENT_HUGEPAGE_ALWAYS 855 default TRANSPARENT_HUGEPAGE_ALWAYS
837 help 856 help
838 Selects the sysfs defaults for Trans 857 Selects the sysfs defaults for Transparent Hugepage Support.
839 858
840 config TRANSPARENT_HUGEPAGE_ALWAYS 859 config TRANSPARENT_HUGEPAGE_ALWAYS
841 bool "always" 860 bool "always"
842 help 861 help
843 Enabling Transparent Hugepage always 862 Enabling Transparent Hugepage always, can increase the
844 memory footprint of applications wit 863 memory footprint of applications without a guaranteed
845 benefit but it will work automatical 864 benefit but it will work automatically for all applications.
846 865
847 config TRANSPARENT_HUGEPAGE_MADVISE 866 config TRANSPARENT_HUGEPAGE_MADVISE
848 bool "madvise" 867 bool "madvise"
849 help 868 help
850 Enabling Transparent Hugepage madvis 869 Enabling Transparent Hugepage madvise, will only provide a
851 performance improvement benefit to t 870 performance improvement benefit to the applications using
852 madvise(MADV_HUGEPAGE) but it won't 871 madvise(MADV_HUGEPAGE) but it won't risk to increase the
853 memory footprint of applications wit 872 memory footprint of applications without a guaranteed
854 benefit. 873 benefit.
855 <<
856 config TRANSPARENT_HUGEPAGE_NEVER <<
857 bool "never" <<
858 help <<
859 Disable Transparent Hugepage by defa <<
860 enabled at runtime via sysfs. <<
861 endchoice 874 endchoice
862 875
863 config THP_SWAP 876 config THP_SWAP
864 def_bool y 877 def_bool y
865 depends on TRANSPARENT_HUGEPAGE && ARC 878 depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP && 64BIT
866 help 879 help
867 Swap transparent huge pages in one p 880 Swap transparent huge pages in one piece, without splitting.
868 XXX: For now, swap cluster backing t 881 XXX: For now, swap cluster backing transparent huge page
869 will be split after swapout. 882 will be split after swapout.
870 883
871 For selection by architectures with 884 For selection by architectures with reasonable THP sizes.
872 885
873 config READ_ONLY_THP_FOR_FS 886 config READ_ONLY_THP_FOR_FS
874 bool "Read-only THP for filesystems (E 887 bool "Read-only THP for filesystems (EXPERIMENTAL)"
875 depends on TRANSPARENT_HUGEPAGE && SHM 888 depends on TRANSPARENT_HUGEPAGE && SHMEM
876 889
877 help 890 help
878 Allow khugepaged to put read-only fi 891 Allow khugepaged to put read-only file-backed pages in THP.
879 892
880 This is marked experimental because 893 This is marked experimental because it is a new feature. Write
881 support of file THPs will be develop 894 support of file THPs will be developed in the next few release
882 cycles. 895 cycles.
883 896
884 endif # TRANSPARENT_HUGEPAGE 897 endif # TRANSPARENT_HUGEPAGE
885 898
886 # 899 #
887 # The architecture supports pgtable leaves tha <<
888 # <<
889 config PGTABLE_HAS_HUGE_LEAVES <<
890 def_bool TRANSPARENT_HUGEPAGE || HUGET <<
891 <<
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 # <<
906 # UP and nommu archs use km based percpu alloc 900 # UP and nommu archs use km based percpu allocator
907 # 901 #
908 config NEED_PER_CPU_KM 902 config NEED_PER_CPU_KM
909 depends on !SMP || !MMU 903 depends on !SMP || !MMU
910 bool 904 bool
911 default y 905 default y
912 906
913 config NEED_PER_CPU_EMBED_FIRST_CHUNK 907 config NEED_PER_CPU_EMBED_FIRST_CHUNK
914 bool 908 bool
915 909
916 config NEED_PER_CPU_PAGE_FIRST_CHUNK 910 config NEED_PER_CPU_PAGE_FIRST_CHUNK
917 bool 911 bool
918 912
919 config USE_PERCPU_NUMA_NODE_ID 913 config USE_PERCPU_NUMA_NODE_ID
920 bool 914 bool
921 915
922 config HAVE_SETUP_PER_CPU_AREA 916 config HAVE_SETUP_PER_CPU_AREA
923 bool 917 bool
924 918
925 config CMA 919 config CMA
926 bool "Contiguous Memory Allocator" 920 bool "Contiguous Memory Allocator"
927 depends on MMU 921 depends on MMU
928 select MIGRATION 922 select MIGRATION
929 select MEMORY_ISOLATION 923 select MEMORY_ISOLATION
930 help 924 help
931 This enables the Contiguous Memory A 925 This enables the Contiguous Memory Allocator which allows other
932 subsystems to allocate big physicall 926 subsystems to allocate big physically-contiguous blocks of memory.
933 CMA reserves a region of memory and 927 CMA reserves a region of memory and allows only movable pages to
934 be allocated from it. This way, the 928 be allocated from it. This way, the kernel can use the memory for
935 pagecache and when a subsystem reque 929 pagecache and when a subsystem requests for contiguous area, the
936 allocated pages are migrated away to 930 allocated pages are migrated away to serve the contiguous request.
937 931
938 If unsure, say "n". 932 If unsure, say "n".
939 933
>> 934 config CMA_DEBUG
>> 935 bool "CMA debug messages (DEVELOPMENT)"
>> 936 depends on DEBUG_KERNEL && CMA
>> 937 help
>> 938 Turns on debug messages in CMA. This produces KERN_DEBUG
>> 939 messages for every CMA call as well as various messages while
>> 940 processing calls such as dma_alloc_from_contiguous().
>> 941 This option does not affect warning and error messages.
>> 942
940 config CMA_DEBUGFS 943 config CMA_DEBUGFS
941 bool "CMA debugfs interface" 944 bool "CMA debugfs interface"
942 depends on CMA && DEBUG_FS 945 depends on CMA && DEBUG_FS
943 help 946 help
944 Turns on the DebugFS interface for C 947 Turns on the DebugFS interface for CMA.
945 948
946 config CMA_SYSFS 949 config CMA_SYSFS
947 bool "CMA information through sysfs in 950 bool "CMA information through sysfs interface"
948 depends on CMA && SYSFS 951 depends on CMA && SYSFS
949 help 952 help
950 This option exposes some sysfs attri 953 This option exposes some sysfs attributes to get information
951 from CMA. 954 from CMA.
952 955
953 config CMA_AREAS 956 config CMA_AREAS
954 int "Maximum count of the CMA areas" 957 int "Maximum count of the CMA areas"
955 depends on CMA 958 depends on CMA
956 default 20 if NUMA !! 959 default 19 if NUMA
957 default 8 !! 960 default 7
958 help 961 help
959 CMA allows to create CMA areas for p 962 CMA allows to create CMA areas for particular purpose, mainly,
960 used as device private area. This pa 963 used as device private area. This parameter sets the maximum
961 number of CMA area in the system. 964 number of CMA area in the system.
962 965
963 If unsure, leave the default value " !! 966 If unsure, leave the default value "7" in UMA and "19" in NUMA.
964 967
965 config MEM_SOFT_DIRTY 968 config MEM_SOFT_DIRTY
966 bool "Track memory changes" 969 bool "Track memory changes"
967 depends on CHECKPOINT_RESTORE && HAVE_ 970 depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS
968 select PROC_PAGE_MONITOR 971 select PROC_PAGE_MONITOR
969 help 972 help
970 This option enables memory changes t 973 This option enables memory changes tracking by introducing a
971 soft-dirty bit on pte-s. This bit it 974 soft-dirty bit on pte-s. This bit it set when someone writes
972 into a page just as regular dirty bi 975 into a page just as regular dirty bit, but unlike the latter
973 it can be cleared by hands. 976 it can be cleared by hands.
974 977
975 See Documentation/admin-guide/mm/sof 978 See Documentation/admin-guide/mm/soft-dirty.rst for more details.
976 979
977 config GENERIC_EARLY_IOREMAP 980 config GENERIC_EARLY_IOREMAP
978 bool 981 bool
979 982
980 config STACK_MAX_DEFAULT_SIZE_MB 983 config STACK_MAX_DEFAULT_SIZE_MB
981 int "Default maximum user stack size f 984 int "Default maximum user stack size for 32-bit processes (MB)"
982 default 100 985 default 100
983 range 8 2048 986 range 8 2048
984 depends on STACK_GROWSUP && (!64BIT || 987 depends on STACK_GROWSUP && (!64BIT || COMPAT)
985 help 988 help
986 This is the maximum stack size in Me 989 This is the maximum stack size in Megabytes in the VM layout of 32-bit
987 user processes when the stack grows 990 user processes when the stack grows upwards (currently only on parisc
988 arch) when the RLIMIT_STACK hard lim 991 arch) when the RLIMIT_STACK hard limit is unlimited.
989 992
990 A sane initial value is 100 MB. 993 A sane initial value is 100 MB.
991 994
992 config DEFERRED_STRUCT_PAGE_INIT 995 config DEFERRED_STRUCT_PAGE_INIT
993 bool "Defer initialisation of struct p 996 bool "Defer initialisation of struct pages to kthreads"
994 depends on SPARSEMEM 997 depends on SPARSEMEM
995 depends on !NEED_PER_CPU_KM 998 depends on !NEED_PER_CPU_KM
996 depends on 64BIT 999 depends on 64BIT
997 depends on !KMSAN <<
998 select PADATA 1000 select PADATA
999 help 1001 help
1000 Ordinarily all struct pages are ini 1002 Ordinarily all struct pages are initialised during early boot in a
1001 single thread. On very large machin 1003 single thread. On very large machines this can take a considerable
1002 amount of time. If this option is s 1004 amount of time. If this option is set, large machines will bring up
1003 a subset of memmap at boot and then 1005 a subset of memmap at boot and then initialise the rest in parallel.
1004 This has a potential performance im 1006 This has a potential performance impact on tasks running early in the
1005 lifetime of the system until these 1007 lifetime of the system until these kthreads finish the
1006 initialisation. 1008 initialisation.
1007 1009
1008 config PAGE_IDLE_FLAG 1010 config PAGE_IDLE_FLAG
1009 bool 1011 bool
1010 select PAGE_EXTENSION if !64BIT 1012 select PAGE_EXTENSION if !64BIT
1011 help 1013 help
1012 This adds PG_idle and PG_young flag 1014 This adds PG_idle and PG_young flags to 'struct page'. PTE Accessed
1013 bit writers can set the state of th 1015 bit writers can set the state of the bit in the flags so that PTE
1014 Accessed bit readers may avoid dist 1016 Accessed bit readers may avoid disturbance.
1015 1017
1016 config IDLE_PAGE_TRACKING 1018 config IDLE_PAGE_TRACKING
1017 bool "Enable idle page tracking" 1019 bool "Enable idle page tracking"
1018 depends on SYSFS && MMU 1020 depends on SYSFS && MMU
1019 select PAGE_IDLE_FLAG 1021 select PAGE_IDLE_FLAG
1020 help 1022 help
1021 This feature allows to estimate the 1023 This feature allows to estimate the amount of user pages that have
1022 not been touched during a given per 1024 not been touched during a given period of time. This information can
1023 be useful to tune memory cgroup lim 1025 be useful to tune memory cgroup limits and/or for job placement
1024 within a compute cluster. 1026 within a compute cluster.
1025 1027
1026 See Documentation/admin-guide/mm/id 1028 See Documentation/admin-guide/mm/idle_page_tracking.rst for
1027 more details. 1029 more details.
1028 1030
1029 # Architectures which implement cpu_dcache_is <<
1030 # whether the data caches are aliased (VIVT o <<
1031 # aliasing) need to select this. <<
1032 config ARCH_HAS_CPU_CACHE_ALIASING <<
1033 bool <<
1034 <<
1035 config ARCH_HAS_CACHE_LINE_SIZE 1031 config ARCH_HAS_CACHE_LINE_SIZE
1036 bool 1032 bool
1037 1033
1038 config ARCH_HAS_CURRENT_STACK_POINTER 1034 config ARCH_HAS_CURRENT_STACK_POINTER
1039 bool 1035 bool
1040 help 1036 help
1041 In support of HARDENED_USERCOPY per 1037 In support of HARDENED_USERCOPY performing stack variable lifetime
1042 checking, an architecture-agnostic 1038 checking, an architecture-agnostic way to find the stack pointer
1043 is needed. Once an architecture def 1039 is needed. Once an architecture defines an unsigned long global
1044 register alias named "current_stack 1040 register alias named "current_stack_pointer", this config can be
1045 selected. 1041 selected.
1046 1042
1047 config ARCH_HAS_PTE_DEVMAP 1043 config ARCH_HAS_PTE_DEVMAP
1048 bool 1044 bool
1049 1045
1050 config ARCH_HAS_ZONE_DMA_SET 1046 config ARCH_HAS_ZONE_DMA_SET
1051 bool 1047 bool
1052 1048
1053 config ZONE_DMA 1049 config ZONE_DMA
1054 bool "Support DMA zone" if ARCH_HAS_Z 1050 bool "Support DMA zone" if ARCH_HAS_ZONE_DMA_SET
1055 default y if ARM64 || X86 1051 default y if ARM64 || X86
1056 1052
1057 config ZONE_DMA32 1053 config ZONE_DMA32
1058 bool "Support DMA32 zone" if ARCH_HAS 1054 bool "Support DMA32 zone" if ARCH_HAS_ZONE_DMA_SET
1059 depends on !X86_32 1055 depends on !X86_32
1060 default y if ARM64 1056 default y if ARM64
1061 1057
1062 config ZONE_DEVICE 1058 config ZONE_DEVICE
1063 bool "Device memory (pmem, HMM, etc.. 1059 bool "Device memory (pmem, HMM, etc...) hotplug support"
1064 depends on MEMORY_HOTPLUG 1060 depends on MEMORY_HOTPLUG
1065 depends on MEMORY_HOTREMOVE 1061 depends on MEMORY_HOTREMOVE
1066 depends on SPARSEMEM_VMEMMAP 1062 depends on SPARSEMEM_VMEMMAP
1067 depends on ARCH_HAS_PTE_DEVMAP 1063 depends on ARCH_HAS_PTE_DEVMAP
1068 select XARRAY_MULTI 1064 select XARRAY_MULTI
1069 1065
1070 help 1066 help
1071 Device memory hotplug support allow 1067 Device memory hotplug support allows for establishing pmem,
1072 or other device driver discovered m 1068 or other device driver discovered memory regions, in the
1073 memmap. This allows pfn_to_page() l 1069 memmap. This allows pfn_to_page() lookups of otherwise
1074 "device-physical" addresses which i 1070 "device-physical" addresses which is needed for using a DAX
1075 mapping in an O_DIRECT operation, a 1071 mapping in an O_DIRECT operation, among other things.
1076 1072
1077 If FS_DAX is enabled, then say Y. 1073 If FS_DAX is enabled, then say Y.
1078 1074
1079 # 1075 #
1080 # Helpers to mirror range of the CPU page tab 1076 # Helpers to mirror range of the CPU page tables of a process into device page
1081 # tables. 1077 # tables.
1082 # 1078 #
1083 config HMM_MIRROR 1079 config HMM_MIRROR
1084 bool 1080 bool
1085 depends on MMU 1081 depends on MMU
1086 1082
1087 config GET_FREE_REGION 1083 config GET_FREE_REGION
>> 1084 depends on SPARSEMEM
1088 bool 1085 bool
1089 1086
1090 config DEVICE_PRIVATE 1087 config DEVICE_PRIVATE
1091 bool "Unaddressable device memory (GP 1088 bool "Unaddressable device memory (GPU memory, ...)"
1092 depends on ZONE_DEVICE 1089 depends on ZONE_DEVICE
1093 select GET_FREE_REGION 1090 select GET_FREE_REGION
1094 1091
1095 help 1092 help
1096 Allows creation of struct pages to 1093 Allows creation of struct pages to represent unaddressable device
1097 memory; i.e., memory that is only a 1094 memory; i.e., memory that is only accessible from the device (or
1098 group of devices). You likely also 1095 group of devices). You likely also want to select HMM_MIRROR.
1099 1096
1100 config VMAP_PFN 1097 config VMAP_PFN
1101 bool 1098 bool
1102 1099
1103 config ARCH_USES_HIGH_VMA_FLAGS 1100 config ARCH_USES_HIGH_VMA_FLAGS
1104 bool 1101 bool
1105 config ARCH_HAS_PKEYS 1102 config ARCH_HAS_PKEYS
1106 bool 1103 bool
1107 1104
1108 config ARCH_USES_PG_ARCH_2 !! 1105 config ARCH_USES_PG_ARCH_X
1109 bool <<
1110 config ARCH_USES_PG_ARCH_3 <<
1111 bool 1106 bool
>> 1107 help
>> 1108 Enable the definition of PG_arch_x page flags with x > 1. Only
>> 1109 suitable for 64-bit architectures with CONFIG_FLATMEM or
>> 1110 CONFIG_SPARSEMEM_VMEMMAP enabled, otherwise there may not be
>> 1111 enough room for additional bits in page->flags.
1112 1112
1113 config VM_EVENT_COUNTERS 1113 config VM_EVENT_COUNTERS
1114 default y 1114 default y
1115 bool "Enable VM event counters for /p 1115 bool "Enable VM event counters for /proc/vmstat" if EXPERT
1116 help 1116 help
1117 VM event counters are needed for ev 1117 VM event counters are needed for event counts to be shown.
1118 This option allows the disabling of 1118 This option allows the disabling of the VM event counters
1119 on EXPERT systems. /proc/vmstat wi 1119 on EXPERT systems. /proc/vmstat will only show page counts
1120 if VM event counters are disabled. 1120 if VM event counters are disabled.
1121 1121
1122 config PERCPU_STATS 1122 config PERCPU_STATS
1123 bool "Collect percpu memory statistic 1123 bool "Collect percpu memory statistics"
1124 help 1124 help
1125 This feature collects and exposes s 1125 This feature collects and exposes statistics via debugfs. The
1126 information includes global and per 1126 information includes global and per chunk statistics, which can
1127 be used to help understand percpu m 1127 be used to help understand percpu memory usage.
1128 1128
1129 config GUP_TEST 1129 config GUP_TEST
1130 bool "Enable infrastructure for get_u 1130 bool "Enable infrastructure for get_user_pages()-related unit tests"
1131 depends on DEBUG_FS 1131 depends on DEBUG_FS
1132 help 1132 help
1133 Provides /sys/kernel/debug/gup_test 1133 Provides /sys/kernel/debug/gup_test, which in turn provides a way
1134 to make ioctl calls that can launch 1134 to make ioctl calls that can launch kernel-based unit tests for
1135 the get_user_pages*() and pin_user_ 1135 the get_user_pages*() and pin_user_pages*() family of API calls.
1136 1136
1137 These tests include benchmark testi 1137 These tests include benchmark testing of the _fast variants of
1138 get_user_pages*() and pin_user_page 1138 get_user_pages*() and pin_user_pages*(), as well as smoke tests of
1139 the non-_fast variants. 1139 the non-_fast variants.
1140 1140
1141 There is also a sub-test that allow 1141 There is also a sub-test that allows running dump_page() on any
1142 of up to eight pages (selected by c 1142 of up to eight pages (selected by command line args) within the
1143 range of user-space addresses. Thes 1143 range of user-space addresses. These pages are either pinned via
1144 pin_user_pages*(), or pinned via ge 1144 pin_user_pages*(), or pinned via get_user_pages*(), as specified
1145 by other command line arguments. 1145 by other command line arguments.
1146 1146
1147 See tools/testing/selftests/mm/gup_ 1147 See tools/testing/selftests/mm/gup_test.c
1148 1148
1149 comment "GUP_TEST needs to have DEBUG_FS enab 1149 comment "GUP_TEST needs to have DEBUG_FS enabled"
1150 depends on !GUP_TEST && !DEBUG_FS 1150 depends on !GUP_TEST && !DEBUG_FS
1151 1151
1152 config GUP_GET_PXX_LOW_HIGH 1152 config GUP_GET_PXX_LOW_HIGH
1153 bool 1153 bool
1154 1154
1155 config DMAPOOL_TEST 1155 config DMAPOOL_TEST
1156 tristate "Enable a module to run time 1156 tristate "Enable a module to run time tests on dma_pool"
1157 depends on HAS_DMA 1157 depends on HAS_DMA
1158 help 1158 help
1159 Provides a test module that will al 1159 Provides a test module that will allocate and free many blocks of
1160 various sizes and report how long i 1160 various sizes and report how long it takes. This is intended to
1161 provide a consistent way to measure 1161 provide a consistent way to measure how changes to the
1162 dma_pool_alloc/free routines affect 1162 dma_pool_alloc/free routines affect performance.
1163 1163
1164 config ARCH_HAS_PTE_SPECIAL 1164 config ARCH_HAS_PTE_SPECIAL
1165 bool 1165 bool
1166 1166
>> 1167 #
>> 1168 # Some architectures require a special hugepage directory format that is
>> 1169 # required to support multiple hugepage sizes. For example a4fe3ce76
>> 1170 # "powerpc/mm: Allow more flexible layouts for hugepage pagetables"
>> 1171 # introduced it on powerpc. This allows for a more flexible hugepage
>> 1172 # pagetable layouts.
>> 1173 #
>> 1174 config ARCH_HAS_HUGEPD
>> 1175 bool
>> 1176
1167 config MAPPING_DIRTY_HELPERS 1177 config MAPPING_DIRTY_HELPERS
1168 bool 1178 bool
1169 1179
1170 config KMAP_LOCAL 1180 config KMAP_LOCAL
1171 bool 1181 bool
1172 1182
1173 config KMAP_LOCAL_NON_LINEAR_PTE_ARRAY 1183 config KMAP_LOCAL_NON_LINEAR_PTE_ARRAY
1174 bool 1184 bool
1175 1185
1176 # struct io_mapping based helper. Selected b 1186 # struct io_mapping based helper. Selected by drivers that need them
1177 config IO_MAPPING 1187 config IO_MAPPING
1178 bool 1188 bool
1179 1189
1180 config MEMFD_CREATE 1190 config MEMFD_CREATE
1181 bool "Enable memfd_create() system ca 1191 bool "Enable memfd_create() system call" if EXPERT
1182 1192
1183 config SECRETMEM 1193 config SECRETMEM
1184 default y 1194 default y
1185 bool "Enable memfd_secret() system ca 1195 bool "Enable memfd_secret() system call" if EXPERT
1186 depends on ARCH_HAS_SET_DIRECT_MAP 1196 depends on ARCH_HAS_SET_DIRECT_MAP
1187 help 1197 help
1188 Enable the memfd_secret() system ca 1198 Enable the memfd_secret() system call with the ability to create
1189 memory areas visible only in the co 1199 memory areas visible only in the context of the owning process and
1190 not mapped to other processes and o 1200 not mapped to other processes and other kernel page tables.
1191 1201
1192 config ANON_VMA_NAME 1202 config ANON_VMA_NAME
1193 bool "Anonymous VMA name support" 1203 bool "Anonymous VMA name support"
1194 depends on PROC_FS && ADVISE_SYSCALLS 1204 depends on PROC_FS && ADVISE_SYSCALLS && MMU
1195 1205
1196 help 1206 help
1197 Allow naming anonymous virtual memo 1207 Allow naming anonymous virtual memory areas.
1198 1208
1199 This feature allows assigning names 1209 This feature allows assigning names to virtual memory areas. Assigned
1200 names can be later retrieved from / 1210 names can be later retrieved from /proc/pid/maps and /proc/pid/smaps
1201 and help identifying individual ano 1211 and help identifying individual anonymous memory areas.
1202 Assigning a name to anonymous virtu 1212 Assigning a name to anonymous virtual memory area might prevent that
1203 area from being merged with adjacen 1213 area from being merged with adjacent virtual memory areas due to the
1204 difference in their name. 1214 difference in their name.
1205 1215
>> 1216 config USERFAULTFD
>> 1217 bool "Enable userfaultfd() system call"
>> 1218 depends on MMU
>> 1219 help
>> 1220 Enable the userfaultfd() system call that allows to intercept and
>> 1221 handle page faults in userland.
>> 1222
1206 config HAVE_ARCH_USERFAULTFD_WP 1223 config HAVE_ARCH_USERFAULTFD_WP
1207 bool 1224 bool
1208 help 1225 help
1209 Arch has userfaultfd write protecti 1226 Arch has userfaultfd write protection support
1210 1227
1211 config HAVE_ARCH_USERFAULTFD_MINOR 1228 config HAVE_ARCH_USERFAULTFD_MINOR
1212 bool 1229 bool
1213 help 1230 help
1214 Arch has userfaultfd minor fault su 1231 Arch has userfaultfd minor fault support
1215 1232
1216 menuconfig USERFAULTFD <<
1217 bool "Enable userfaultfd() system cal <<
1218 depends on MMU <<
1219 help <<
1220 Enable the userfaultfd() system cal <<
1221 handle page faults in userland. <<
1222 <<
1223 if USERFAULTFD <<
1224 config PTE_MARKER_UFFD_WP 1233 config PTE_MARKER_UFFD_WP
1225 bool "Userfaultfd write protection su 1234 bool "Userfaultfd write protection support for shmem/hugetlbfs"
1226 default y 1235 default y
1227 depends on HAVE_ARCH_USERFAULTFD_WP 1236 depends on HAVE_ARCH_USERFAULTFD_WP
1228 1237
1229 help 1238 help
1230 Allows to create marker PTEs for us 1239 Allows to create marker PTEs for userfaultfd write protection
1231 purposes. It is required to enable 1240 purposes. It is required to enable userfaultfd write protection on
1232 file-backed memory types like shmem 1241 file-backed memory types like shmem and hugetlbfs.
1233 endif # USERFAULTFD <<
1234 1242
1235 # multi-gen LRU { 1243 # multi-gen LRU {
1236 config LRU_GEN 1244 config LRU_GEN
1237 bool "Multi-Gen LRU" 1245 bool "Multi-Gen LRU"
1238 depends on MMU 1246 depends on MMU
1239 # make sure folio->flags has enough s 1247 # make sure folio->flags has enough spare bits
1240 depends on 64BIT || !SPARSEMEM || SPA 1248 depends on 64BIT || !SPARSEMEM || SPARSEMEM_VMEMMAP
1241 help 1249 help
1242 A high performance LRU implementati 1250 A high performance LRU implementation to overcommit memory. See
1243 Documentation/admin-guide/mm/multig 1251 Documentation/admin-guide/mm/multigen_lru.rst for details.
1244 1252
1245 config LRU_GEN_ENABLED 1253 config LRU_GEN_ENABLED
1246 bool "Enable by default" 1254 bool "Enable by default"
1247 depends on LRU_GEN 1255 depends on LRU_GEN
1248 help 1256 help
1249 This option enables the multi-gen L 1257 This option enables the multi-gen LRU by default.
1250 1258
1251 config LRU_GEN_STATS 1259 config LRU_GEN_STATS
1252 bool "Full stats for debugging" 1260 bool "Full stats for debugging"
1253 depends on LRU_GEN 1261 depends on LRU_GEN
1254 help 1262 help
1255 Do not enable this option unless yo 1263 Do not enable this option unless you plan to look at historical stats
1256 from evicted generations for debugg 1264 from evicted generations for debugging purpose.
1257 1265
1258 This option has a per-memcg and per 1266 This option has a per-memcg and per-node memory overhead.
1259 <<
1260 config LRU_GEN_WALKS_MMU <<
1261 def_bool y <<
1262 depends on LRU_GEN && ARCH_HAS_HW_PTE <<
1263 # } 1267 # }
1264 1268
1265 config ARCH_SUPPORTS_PER_VMA_LOCK 1269 config ARCH_SUPPORTS_PER_VMA_LOCK
1266 def_bool n 1270 def_bool n
1267 1271
1268 config PER_VMA_LOCK 1272 config PER_VMA_LOCK
1269 def_bool y 1273 def_bool y
1270 depends on ARCH_SUPPORTS_PER_VMA_LOCK 1274 depends on ARCH_SUPPORTS_PER_VMA_LOCK && MMU && SMP
1271 help 1275 help
1272 Allow per-vma locking during page f 1276 Allow per-vma locking during page fault handling.
1273 1277
1274 This feature allows locking each vi 1278 This feature allows locking each virtual memory area separately when
1275 handling page faults instead of tak 1279 handling page faults instead of taking mmap_lock.
1276 1280
1277 config LOCK_MM_AND_FIND_VMA 1281 config LOCK_MM_AND_FIND_VMA
1278 bool 1282 bool
1279 depends on !STACK_GROWSUP 1283 depends on !STACK_GROWSUP
1280 <<
1281 config IOMMU_MM_DATA <<
1282 bool <<
1283 <<
1284 config EXECMEM <<
1285 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 1284
1298 source "mm/damon/Kconfig" 1285 source "mm/damon/Kconfig"
1299 1286
1300 endmenu 1287 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.