1 # SPDX-License-Identifier: GPL-2.0-only <<
2 <<
3 menu "Memory Management options" <<
4 <<
5 # <<
6 # For some reason microblaze and nios2 hard co <<
7 # add proper SWAP support to them, in which ca <<
8 # <<
9 config ARCH_NO_SWAP <<
10 bool <<
11 <<
12 config ZPOOL <<
13 bool <<
14 <<
15 menuconfig SWAP <<
16 bool "Support for paging of anonymous <<
17 depends on MMU && BLOCK && !ARCH_NO_SW <<
18 default y <<
19 help <<
20 This option allows you to choose whe <<
21 for so called swap devices or swap f <<
22 used to provide more virtual memory <<
23 in your computer. If unsure say Y. <<
24 <<
25 config ZSWAP <<
26 bool "Compressed cache for swap pages" <<
27 depends on SWAP <<
28 select CRYPTO <<
29 select ZPOOL <<
30 help <<
31 A lightweight compressed cache for s <<
32 pages that are in the process of bei <<
33 compress them into a dynamically all <<
34 This can result in a significant I/O <<
35 in the case where decompressing from <<
36 reads, can also improve workload per <<
37 <<
38 config ZSWAP_DEFAULT_ON <<
39 bool "Enable the compressed cache for <<
40 depends on ZSWAP <<
41 help <<
42 If selected, the compressed cache fo <<
43 at boot, otherwise it will be disabl <<
44 <<
45 The selection made here can be overr <<
46 command line 'zswap.enabled=' option <<
47 <<
48 config ZSWAP_SHRINKER_DEFAULT_ON <<
49 bool "Shrink the zswap pool on memory <<
50 depends on ZSWAP <<
51 default n <<
52 help <<
53 If selected, the zswap shrinker will <<
54 stored in the zswap pool will become <<
55 written back to the backing swap dev <<
56 <<
57 This means that zswap writeback coul <<
58 not yet full, or the cgroup zswap li <<
59 reducing the chance that cold pages <<
60 and consume memory indefinitely. <<
61 <<
62 choice <<
63 prompt "Default compressor" <<
64 depends on ZSWAP <<
65 default ZSWAP_COMPRESSOR_DEFAULT_LZO <<
66 help <<
67 Selects the default compression algo <<
68 for swap pages. <<
69 <<
70 For an overview what kind of perform <<
71 a particular compression algorithm p <<
72 available at the following LWN page: <<
73 https://lwn.net/Articles/751795/ <<
74 <<
75 If in doubt, select 'LZO'. <<
76 <<
77 The selection made here can be overr <<
78 command line 'zswap.compressor=' opt <<
79 <<
80 config ZSWAP_COMPRESSOR_DEFAULT_DEFLATE <<
81 bool "Deflate" <<
82 select CRYPTO_DEFLATE <<
83 help <<
84 Use the Deflate algorithm as the def <<
85 <<
86 config ZSWAP_COMPRESSOR_DEFAULT_LZO <<
87 bool "LZO" <<
88 select CRYPTO_LZO <<
89 help <<
90 Use the LZO algorithm as the default <<
91 <<
92 config ZSWAP_COMPRESSOR_DEFAULT_842 <<
93 bool "842" <<
94 select CRYPTO_842 <<
95 help <<
96 Use the 842 algorithm as the default <<
97 <<
98 config ZSWAP_COMPRESSOR_DEFAULT_LZ4 <<
99 bool "LZ4" <<
100 select CRYPTO_LZ4 <<
101 help <<
102 Use the LZ4 algorithm as the default <<
103 <<
104 config ZSWAP_COMPRESSOR_DEFAULT_LZ4HC <<
105 bool "LZ4HC" <<
106 select CRYPTO_LZ4HC <<
107 help <<
108 Use the LZ4HC algorithm as the defau <<
109 <<
110 config ZSWAP_COMPRESSOR_DEFAULT_ZSTD <<
111 bool "zstd" <<
112 select CRYPTO_ZSTD <<
113 help <<
114 Use the zstd algorithm as the defaul <<
115 endchoice <<
116 <<
117 config ZSWAP_COMPRESSOR_DEFAULT <<
118 string <<
119 depends on ZSWAP <<
120 default "deflate" if ZSWAP_COMPRESSOR_D <<
121 default "lzo" if ZSWAP_COMPRESSOR_DEFAU <<
122 default "842" if ZSWAP_COMPRESSOR_DEFAU <<
123 default "lz4" if ZSWAP_COMPRESSOR_DEFAU <<
124 default "lz4hc" if ZSWAP_COMPRESSOR_DEF <<
125 default "zstd" if ZSWAP_COMPRESSOR_DEFA <<
126 default "" <<
127 <<
128 choice <<
129 prompt "Default allocator" <<
130 depends on ZSWAP <<
131 default ZSWAP_ZPOOL_DEFAULT_ZSMALLOC i <<
132 default ZSWAP_ZPOOL_DEFAULT_ZBUD <<
133 help <<
134 Selects the default allocator for th <<
135 swap pages. <<
136 The default is 'zbud' for compatibil <<
137 read the description of each of the <<
138 making a right choice. <<
139 <<
140 The selection made here can be overr <<
141 command line 'zswap.zpool=' option. <<
142 <<
143 config ZSWAP_ZPOOL_DEFAULT_ZBUD <<
144 bool "zbud" <<
145 select ZBUD <<
146 help <<
147 Use the zbud allocator as the defaul <<
148 <<
149 config ZSWAP_ZPOOL_DEFAULT_Z3FOLD_DEPRECATED <<
150 bool "z3foldi (DEPRECATED)" <<
151 select Z3FOLD_DEPRECATED <<
152 help <<
153 Use the z3fold allocator as the defa <<
154 <<
155 Deprecated and scheduled for removal <<
156 see CONFIG_Z3FOLD_DEPRECATED. <<
157 <<
158 config ZSWAP_ZPOOL_DEFAULT_ZSMALLOC <<
159 bool "zsmalloc" <<
160 select ZSMALLOC <<
161 help <<
162 Use the zsmalloc allocator as the de <<
163 endchoice <<
164 <<
165 config ZSWAP_ZPOOL_DEFAULT <<
166 string <<
167 depends on ZSWAP <<
168 default "zbud" if ZSWAP_ZPOOL_DEFAULT_Z <<
169 default "z3fold" if ZSWAP_ZPOOL_DEFAULT <<
170 default "zsmalloc" if ZSWAP_ZPOOL_DEFAU <<
171 default "" <<
172 <<
173 config ZBUD <<
174 tristate "2:1 compression allocator (z <<
175 depends on ZSWAP <<
176 help <<
177 A special purpose allocator for stor <<
178 It is designed to store up to two co <<
179 page. While this design limits stor <<
180 deterministic reclaim properties tha <<
181 density approach when reclaim will b <<
182 <<
183 config Z3FOLD_DEPRECATED <<
184 tristate "3:1 compression allocator (z <<
185 depends on ZSWAP <<
186 help <<
187 Deprecated and scheduled for removal <<
188 a good reason for using Z3FOLD over <<
189 linux-mm@kvack.org and the zswap mai <<
190 <<
191 A special purpose allocator for stor <<
192 It is designed to store up to three <<
193 page. It is a ZBUD derivative so the <<
194 still there. <<
195 <<
196 config Z3FOLD <<
197 tristate <<
198 default y if Z3FOLD_DEPRECATED=y <<
199 default m if Z3FOLD_DEPRECATED=m <<
200 depends on Z3FOLD_DEPRECATED <<
201 <<
202 config ZSMALLOC <<
203 tristate <<
204 prompt "N:1 compression allocator (zsm <<
205 depends on MMU <<
206 help <<
207 zsmalloc is a slab-based memory allo <<
208 pages of various compression levels <<
209 the highest storage density with the <<
210 <<
211 config ZSMALLOC_STAT <<
212 bool "Export zsmalloc statistics" <<
213 depends on ZSMALLOC <<
214 select DEBUG_FS <<
215 help <<
216 This option enables code in the zsma <<
217 statistics about what's happening in <<
218 information to userspace via debugfs <<
219 If unsure, say N. <<
220 <<
221 config ZSMALLOC_CHAIN_SIZE <<
222 int "Maximum number of physical pages <<
223 default 8 <<
224 range 4 16 <<
225 depends on ZSMALLOC <<
226 help <<
227 This option sets the upper limit on <<
228 that a zmalloc page (zspage) can con <<
229 chain size is calculated for each si <<
230 initialization of the pool. <<
231 <<
232 Changing this option can alter the c <<
233 such as the number of pages per zspa <<
234 per zspage. This can also result in <<
235 the pool, as zsmalloc merges size cl <<
236 characteristics. <<
237 <<
238 For more information, see zsmalloc d <<
239 <<
240 menu "Slab allocator options" <<
241 <<
242 config SLUB <<
243 def_bool y <<
244 <<
245 config SLUB_TINY <<
246 bool "Configure for minimal memory foo <<
247 depends on EXPERT <<
248 select SLAB_MERGE_DEFAULT <<
249 help <<
250 Configures the slab allocator in a <<
251 footprint, sacrificing scalability, <<
252 This is intended only for the small <<
253 SLOB allocator and is not recommend <<
254 16MB RAM. <<
255 <<
256 If unsure, say N. <<
257 <<
258 config SLAB_MERGE_DEFAULT <<
259 bool "Allow slab caches to be merged" <<
260 default y <<
261 help <<
262 For reduced kernel memory fragmentat <<
263 merged when they share the same size <<
264 This carries a risk of kernel heap o <<
265 overwrite objects from merged caches <<
266 cache layout), which makes such heap <<
267 by attackers. By keeping caches unme <<
268 can usually only damage objects in t <<
269 merging at runtime, "slab_nomerge" c <<
270 command line. <<
271 <<
272 config SLAB_FREELIST_RANDOM <<
273 bool "Randomize slab freelist" <<
274 depends on !SLUB_TINY <<
275 help <<
276 Randomizes the freelist order used o <<
277 security feature reduces the predict <<
278 allocator against heap overflows. <<
279 <<
280 config SLAB_FREELIST_HARDENED <<
281 bool "Harden slab freelist metadata" <<
282 depends on !SLUB_TINY <<
283 help <<
284 Many kernel heap attacks try to targ <<
285 other infrastructure. This options m <<
286 sacrifices to harden the kernel slab <<
287 freelist exploit methods. <<
288 <<
289 config SLAB_BUCKETS <<
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 <<
306 config SLUB_STATS <<
307 default n <<
308 bool "Enable performance statistics" <<
309 depends on SYSFS && !SLUB_TINY <<
310 help <<
311 The statistics are useful to debug s <<
312 order find ways to optimize the allo <<
313 enabled for production use since kee <<
314 the allocator by a few percentage po <<
315 supports the determination of the mo <<
316 out which slabs are relevant to a pa <<
317 Try running: slabinfo -DA <<
318 <<
319 config SLUB_CPU_PARTIAL <<
320 default y <<
321 depends on SMP && !SLUB_TINY <<
322 bool "Enable per cpu partial caches" <<
323 help <<
324 Per cpu partial caches accelerate ob <<
325 that is local to a processor at the <<
326 in the latency of the free. On overf <<
327 which requires the taking of locks t <<
328 Typically one would choose no for a <<
329 <<
330 config RANDOM_KMALLOC_CACHES <<
331 default n <<
332 depends on !SLUB_TINY <<
333 bool "Randomize slab caches for normal <<
334 help <<
335 A hardening feature that creates mul <<
336 normal kmalloc allocation and makes <<
337 on code address, which makes the att <<
338 vulnerable memory objects on the hea <<
339 memory vulnerabilities. <<
340 <<
341 Currently the number of copies is se <<
342 that effectively diverges the memory <<
343 subsystems or modules into different <<
344 limited degree of memory and CPU ove <<
345 system workload. <<
346 <<
347 endmenu # Slab allocator options <<
348 <<
349 config SHUFFLE_PAGE_ALLOCATOR <<
350 bool "Page allocator randomization" <<
351 default SLAB_FREELIST_RANDOM && ACPI_N <<
352 help <<
353 Randomization of the page allocator <<
354 utilization of a direct-mapped memor <<
355 5.2.27 Heterogeneous Memory Attribut <<
356 6.2a specification for an example of <<
357 the presence of a memory-side-cache. <<
358 security benefits as it reduces the <<
359 allocations to compliment SLAB_FREEL <<
360 default granularity of shuffling on <<
361 order of pages is selected based on <<
362 on x86. <<
363 <<
364 While the randomization improves cac <<
365 negatively impact workloads on platf <<
366 this reason, by default, the randomi <<
367 if SHUFFLE_PAGE_ALLOCATOR=y. The ran <<
368 with the 'page_alloc.shuffle' kernel <<
369 <<
370 Say Y if unsure. <<
371 <<
372 config COMPAT_BRK <<
373 bool "Disable heap randomization" <<
374 default y <<
375 help <<
376 Randomizing heap placement makes hea <<
377 also breaks ancient binaries (includ <<
378 This option changes the bootup defau <<
379 disabled, and can be overridden at r <<
380 /proc/sys/kernel/randomize_va_space <<
381 <<
382 On non-ancient distros (post-2000 on <<
383 <<
384 config MMAP_ALLOW_UNINITIALIZED <<
385 bool "Allow mmapped anonymous memory t <<
386 depends on EXPERT && !MMU <<
387 default n <<
388 help <<
389 Normally, and according to the Linux <<
390 from mmap() has its contents cleared <<
391 userspace. Enabling this config opt <<
392 mmap() skip that if it is given an M <<
393 providing a huge performance boost. <<
394 then the flag will be ignored. <<
395 <<
396 This is taken advantage of by uClibc <<
397 ELF-FDPIC binfmt's brk and stack all <<
398 <<
399 Because of the obvious security issu <<
400 enabled on embedded devices where yo <<
401 userspace. Since that isn't general <<
402 it is normally safe to say Y here. <<
403 <<
404 See Documentation/admin-guide/mm/nom <<
405 <<
406 config SELECT_MEMORY_MODEL 1 config SELECT_MEMORY_MODEL
407 def_bool y 2 def_bool y
408 depends on ARCH_SELECT_MEMORY_MODEL 3 depends on ARCH_SELECT_MEMORY_MODEL
409 4
410 choice 5 choice
411 prompt "Memory model" 6 prompt "Memory model"
412 depends on SELECT_MEMORY_MODEL 7 depends on SELECT_MEMORY_MODEL
>> 8 default DISCONTIGMEM_MANUAL if ARCH_DISCONTIGMEM_DEFAULT
413 default SPARSEMEM_MANUAL if ARCH_SPARS 9 default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT
414 default FLATMEM_MANUAL 10 default FLATMEM_MANUAL
415 help <<
416 This option allows you to change som <<
417 Linux manages its memory internally. <<
418 only have one option here selected b <<
419 configuration. This is normal. <<
420 11
421 config FLATMEM_MANUAL 12 config FLATMEM_MANUAL
422 bool "Flat Memory" 13 bool "Flat Memory"
423 depends on !ARCH_SPARSEMEM_ENABLE || A !! 14 depends on !(ARCH_DISCONTIGMEM_ENABLE || ARCH_SPARSEMEM_ENABLE) || ARCH_FLATMEM_ENABLE
424 help 15 help
425 This option is best suited for non-N !! 16 This option allows you to change some of the ways that
426 flat address space. The FLATMEM is t !! 17 Linux manages its memory internally. Most users will
427 system in terms of performance and r !! 18 only have one option here: FLATMEM. This is normal
428 and it is the best option for smalle !! 19 and a correct option.
429 !! 20
430 For systems that have holes in their !! 21 Some users of more advanced features like NUMA and
431 spaces and for features like NUMA an !! 22 memory hotplug may have different options here.
432 choose "Sparse Memory". !! 23 DISCONTIGMEM is a more mature, better tested system,
>> 24 but is incompatible with memory hotplug and may suffer
>> 25 decreased performance over SPARSEMEM. If unsure between
>> 26 "Sparse Memory" and "Discontiguous Memory", choose
>> 27 "Discontiguous Memory".
433 28
434 If unsure, choose this option (Flat 29 If unsure, choose this option (Flat Memory) over any other.
435 30
>> 31 config DISCONTIGMEM_MANUAL
>> 32 bool "Discontiguous Memory"
>> 33 depends on ARCH_DISCONTIGMEM_ENABLE
>> 34 help
>> 35 This option provides enhanced support for discontiguous
>> 36 memory systems, over FLATMEM. These systems have holes
>> 37 in their physical address spaces, and this option provides
>> 38 more efficient handling of these holes. However, the vast
>> 39 majority of hardware has quite flat address spaces, and
>> 40 can have degraded performance from the extra overhead that
>> 41 this option imposes.
>> 42
>> 43 Many NUMA configurations will have this as the only option.
>> 44
>> 45 If unsure, choose "Flat Memory" over this option.
>> 46
436 config SPARSEMEM_MANUAL 47 config SPARSEMEM_MANUAL
437 bool "Sparse Memory" 48 bool "Sparse Memory"
438 depends on ARCH_SPARSEMEM_ENABLE 49 depends on ARCH_SPARSEMEM_ENABLE
439 help 50 help
440 This will be the only option for som 51 This will be the only option for some systems, including
441 memory hot-plug systems. This is no !! 52 memory hotplug systems. This is normal.
442 53
443 This option provides efficient suppo !! 54 For many other systems, this will be an alternative to
444 holes is their physical address spac !! 55 "Discontiguous Memory". This option provides some potential
445 hot-plug and hot-remove. !! 56 performance benefits, along with decreased code complexity,
>> 57 but it is newer, and more experimental.
446 58
447 If unsure, choose "Flat Memory" over !! 59 If unsure, choose "Discontiguous Memory" or "Flat Memory"
>> 60 over this option.
448 61
449 endchoice 62 endchoice
450 63
>> 64 config DISCONTIGMEM
>> 65 def_bool y
>> 66 depends on (!SELECT_MEMORY_MODEL && ARCH_DISCONTIGMEM_ENABLE) || DISCONTIGMEM_MANUAL
>> 67
451 config SPARSEMEM 68 config SPARSEMEM
452 def_bool y 69 def_bool y
453 depends on (!SELECT_MEMORY_MODEL && AR 70 depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL
454 71
455 config FLATMEM 72 config FLATMEM
456 def_bool y 73 def_bool y
457 depends on !SPARSEMEM || FLATMEM_MANUA !! 74 depends on (!DISCONTIGMEM && !SPARSEMEM) || FLATMEM_MANUAL
>> 75
>> 76 config FLAT_NODE_MEM_MAP
>> 77 def_bool y
>> 78 depends on !SPARSEMEM
>> 79
>> 80 #
>> 81 # Both the NUMA code and DISCONTIGMEM use arrays of pg_data_t's
>> 82 # to represent different areas of memory. This variable allows
>> 83 # those dependencies to exist individually.
>> 84 #
>> 85 config NEED_MULTIPLE_NODES
>> 86 def_bool y
>> 87 depends on DISCONTIGMEM || NUMA
>> 88
>> 89 config HAVE_MEMORY_PRESENT
>> 90 def_bool y
>> 91 depends on ARCH_HAVE_MEMORY_PRESENT || SPARSEMEM
458 92
459 # 93 #
460 # SPARSEMEM_EXTREME (which is the default) doe 94 # SPARSEMEM_EXTREME (which is the default) does some bootmem
461 # allocations when sparse_init() is called. I !! 95 # allocations when memory_present() is called. If this cannot
462 # be done on your architecture, select this op 96 # be done on your architecture, select this option. However,
463 # statically allocating the mem_section[] arra 97 # statically allocating the mem_section[] array can potentially
464 # consume vast quantities of .bss, so be caref 98 # consume vast quantities of .bss, so be careful.
465 # 99 #
466 # This option will also potentially produce sm 100 # This option will also potentially produce smaller runtime code
467 # with gcc 3.4 and later. 101 # with gcc 3.4 and later.
468 # 102 #
469 config SPARSEMEM_STATIC 103 config SPARSEMEM_STATIC
470 bool 104 bool
471 105
472 # 106 #
473 # Architecture platforms which require a two l 107 # Architecture platforms which require a two level mem_section in SPARSEMEM
474 # must select this option. This is usually for 108 # must select this option. This is usually for architecture platforms with
475 # an extremely sparse physical address space. 109 # an extremely sparse physical address space.
476 # 110 #
477 config SPARSEMEM_EXTREME 111 config SPARSEMEM_EXTREME
478 def_bool y 112 def_bool y
479 depends on SPARSEMEM && !SPARSEMEM_STA 113 depends on SPARSEMEM && !SPARSEMEM_STATIC
480 114
481 config SPARSEMEM_VMEMMAP_ENABLE 115 config SPARSEMEM_VMEMMAP_ENABLE
482 bool 116 bool
483 117
>> 118 config SPARSEMEM_ALLOC_MEM_MAP_TOGETHER
>> 119 def_bool y
>> 120 depends on SPARSEMEM && X86_64
>> 121
484 config SPARSEMEM_VMEMMAP 122 config SPARSEMEM_VMEMMAP
485 bool "Sparse Memory virtual memmap" 123 bool "Sparse Memory virtual memmap"
486 depends on SPARSEMEM && SPARSEMEM_VMEM 124 depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE
487 default y 125 default y
488 help 126 help
489 SPARSEMEM_VMEMMAP uses a virtually m !! 127 SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise
490 pfn_to_page and page_to_pfn operatio !! 128 pfn_to_page and page_to_pfn operations. This is the most
491 efficient option when sufficient ker !! 129 efficient option when sufficient kernel resources are available.
492 # !! 130
493 # Select this config option from the architect !! 131 config HAVE_MEMBLOCK
494 # to enable the feature of HugeTLB/dev_dax vme <<
495 # <<
496 config ARCH_WANT_OPTIMIZE_DAX_VMEMMAP <<
497 bool 132 bool
498 133
499 config ARCH_WANT_OPTIMIZE_HUGETLB_VMEMMAP !! 134 config HAVE_MEMBLOCK_NODE_MAP
500 bool 135 bool
501 136
502 config HAVE_MEMBLOCK_PHYS_MAP 137 config HAVE_MEMBLOCK_PHYS_MAP
503 bool 138 bool
504 139
505 config HAVE_GUP_FAST !! 140 config HAVE_GENERIC_RCU_GUP
506 depends on MMU <<
507 bool 141 bool
508 142
509 # Don't discard allocated memory used to track !! 143 config ARCH_DISCARD_MEMBLOCK
510 # after early boot, so it can still be used to <<
511 # Also, memblocks are updated with memory hot( <<
512 config ARCH_KEEP_MEMBLOCK <<
513 bool 144 bool
514 145
515 # Keep arch NUMA mapping infrastructure post-i !! 146 config NO_BOOTMEM
516 config NUMA_KEEP_MEMINFO <<
517 bool 147 bool
518 148
519 config MEMORY_ISOLATION 149 config MEMORY_ISOLATION
520 bool 150 bool
521 151
522 # IORESOURCE_SYSTEM_RAM regions in the kernel !! 152 config MOVABLE_NODE
523 # IORESOURCE_EXCLUSIVE cannot be mapped to use !! 153 bool "Enable to assign a node which has only movable memory"
524 # /dev/mem. !! 154 depends on HAVE_MEMBLOCK
525 config EXCLUSIVE_SYSTEM_RAM !! 155 depends on NO_BOOTMEM
526 def_bool y !! 156 depends on X86_64
527 depends on !DEVMEM || STRICT_DEVMEM !! 157 depends on NUMA
>> 158 default n
>> 159 help
>> 160 Allow a node to have only movable memory. Pages used by the kernel,
>> 161 such as direct mapping pages cannot be migrated. So the corresponding
>> 162 memory device cannot be hotplugged. This option allows the following
>> 163 two things:
>> 164 - When the system is booting, node full of hotpluggable memory can
>> 165 be arranged to have only movable memory so that the whole node can
>> 166 be hot-removed. (need movable_node boot option specified).
>> 167 - After the system is up, the option allows users to online all the
>> 168 memory of a node as movable memory so that the whole node can be
>> 169 hot-removed.
>> 170
>> 171 Users who don't use the memory hotplug feature are fine with this
>> 172 option on since they don't specify movable_node boot option or they
>> 173 don't online memory as movable.
>> 174
>> 175 Say Y here if you want to hotplug a whole node.
>> 176 Say N here if you want kernel to use memory on all nodes evenly.
528 177
529 # 178 #
530 # Only be set on architectures that have compl 179 # Only be set on architectures that have completely implemented memory hotplug
531 # feature. If you are not sure, don't touch it 180 # feature. If you are not sure, don't touch it.
532 # 181 #
533 config HAVE_BOOTMEM_INFO_NODE 182 config HAVE_BOOTMEM_INFO_NODE
534 def_bool n 183 def_bool n
535 184
536 config ARCH_ENABLE_MEMORY_HOTPLUG <<
537 bool <<
538 <<
539 config ARCH_ENABLE_MEMORY_HOTREMOVE <<
540 bool <<
541 <<
542 # eventually, we can have this option just 'se 185 # eventually, we can have this option just 'select SPARSEMEM'
543 menuconfig MEMORY_HOTPLUG !! 186 config MEMORY_HOTPLUG
544 bool "Memory hotplug" !! 187 bool "Allow for memory hot-add"
545 select MEMORY_ISOLATION !! 188 depends on SPARSEMEM || X86_64_ACPI_NUMA
546 depends on SPARSEMEM <<
547 depends on ARCH_ENABLE_MEMORY_HOTPLUG 189 depends on ARCH_ENABLE_MEMORY_HOTPLUG
548 depends on 64BIT !! 190 depends on COMPILE_TEST || !KASAN
549 select NUMA_KEEP_MEMINFO if NUMA <<
550 191
551 if MEMORY_HOTPLUG !! 192 config MEMORY_HOTPLUG_SPARSE
>> 193 def_bool y
>> 194 depends on SPARSEMEM && MEMORY_HOTPLUG
552 195
553 config MEMORY_HOTPLUG_DEFAULT_ONLINE 196 config MEMORY_HOTPLUG_DEFAULT_ONLINE
554 bool "Online the newly added memory bl !! 197 bool "Online the newly added memory blocks by default"
555 depends on MEMORY_HOTPLUG !! 198 default n
556 help !! 199 depends on MEMORY_HOTPLUG
>> 200 help
557 This option sets the default policy 201 This option sets the default policy setting for memory hotplug
558 onlining policy (/sys/devices/system 202 onlining policy (/sys/devices/system/memory/auto_online_blocks) which
559 determines what happens to newly add 203 determines what happens to newly added memory regions. Policy setting
560 can always be changed at runtime. 204 can always be changed at runtime.
561 See Documentation/admin-guide/mm/mem !! 205 See Documentation/memory-hotplug.txt for more information.
562 206
563 Say Y here if you want all hot-plugg 207 Say Y here if you want all hot-plugged memory blocks to appear in
564 'online' state by default. 208 'online' state by default.
565 Say N here if you want the default p 209 Say N here if you want the default policy to keep all hot-plugged
566 memory blocks in 'offline' state. 210 memory blocks in 'offline' state.
567 211
568 config MEMORY_HOTREMOVE 212 config MEMORY_HOTREMOVE
569 bool "Allow for memory hot remove" 213 bool "Allow for memory hot remove"
>> 214 select MEMORY_ISOLATION
570 select HAVE_BOOTMEM_INFO_NODE if (X86_ 215 select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64)
571 depends on MEMORY_HOTPLUG && ARCH_ENAB 216 depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE
572 depends on MIGRATION 217 depends on MIGRATION
573 218
574 config MHP_MEMMAP_ON_MEMORY <<
575 def_bool y <<
576 depends on MEMORY_HOTPLUG && SPARSEMEM <<
577 depends on ARCH_MHP_MEMMAP_ON_MEMORY_E <<
578 <<
579 endif # MEMORY_HOTPLUG <<
580 <<
581 config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE <<
582 bool <<
583 <<
584 # Heavily threaded applications may benefit fr 219 # Heavily threaded applications may benefit from splitting the mm-wide
585 # page_table_lock, so that faults on different 220 # page_table_lock, so that faults on different parts of the user address
586 # space can be handled with less contention: s 221 # space can be handled with less contention: split it at this NR_CPUS.
587 # Default to 4 for wider testing, though 8 mig 222 # Default to 4 for wider testing, though 8 might be more appropriate.
588 # ARM's adjust_pte (unused if VIPT) depends on 223 # ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
589 # PA-RISC 7xxx's spinlock_t would enlarge stru 224 # PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
590 # SPARC32 allocates multiple pte tables within <<
591 # a per-page lock leads to problems when multi <<
592 # at the same time (e.g. copy_page_range()). <<
593 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock 225 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
594 # 226 #
595 config SPLIT_PTE_PTLOCKS !! 227 config SPLIT_PTLOCK_CPUS
596 def_bool y !! 228 int
597 depends on MMU !! 229 default "999999" if !MMU
598 depends on SMP !! 230 default "999999" if ARM && !CPU_CACHE_VIPT
599 depends on NR_CPUS >= 4 !! 231 default "999999" if PARISC && !PA20
600 depends on !ARM || CPU_CACHE_VIPT !! 232 default "4"
601 depends on !PARISC || PA20 <<
602 depends on !SPARC32 <<
603 233
604 config ARCH_ENABLE_SPLIT_PMD_PTLOCK 234 config ARCH_ENABLE_SPLIT_PMD_PTLOCK
605 bool 235 bool
606 236
607 config SPLIT_PMD_PTLOCKS <<
608 def_bool y <<
609 depends on SPLIT_PTE_PTLOCKS && ARCH_E <<
610 <<
611 # 237 #
612 # support for memory balloon 238 # support for memory balloon
613 config MEMORY_BALLOON 239 config MEMORY_BALLOON
614 bool 240 bool
615 241
616 # 242 #
617 # support for memory balloon compaction 243 # support for memory balloon compaction
618 config BALLOON_COMPACTION 244 config BALLOON_COMPACTION
619 bool "Allow for balloon memory compact 245 bool "Allow for balloon memory compaction/migration"
620 default y !! 246 def_bool y
621 depends on COMPACTION && MEMORY_BALLOO 247 depends on COMPACTION && MEMORY_BALLOON
622 help 248 help
623 Memory fragmentation introduced by b 249 Memory fragmentation introduced by ballooning might reduce
624 significantly the number of 2MB cont 250 significantly the number of 2MB contiguous memory blocks that can be
625 used within a guest, thus imposing p 251 used within a guest, thus imposing performance penalties associated
626 with the reduced number of transpare 252 with the reduced number of transparent huge pages that could be used
627 by the guest workload. Allowing the 253 by the guest workload. Allowing the compaction & migration for memory
628 pages enlisted as being part of memo 254 pages enlisted as being part of memory balloon devices avoids the
629 scenario aforementioned and helps im 255 scenario aforementioned and helps improving memory defragmentation.
630 256
631 # 257 #
632 # support for memory compaction 258 # support for memory compaction
633 config COMPACTION 259 config COMPACTION
634 bool "Allow for memory compaction" 260 bool "Allow for memory compaction"
635 default y !! 261 def_bool y
636 select MIGRATION 262 select MIGRATION
637 depends on MMU 263 depends on MMU
638 help 264 help
639 Compaction is the only memory manage !! 265 Compaction is the only memory management component to form
640 high order (larger physically contig !! 266 high order (larger physically contiguous) memory blocks
641 reliably. The page allocator relies !! 267 reliably. The page allocator relies on compaction heavily and
642 the lack of the feature can lead to !! 268 the lack of the feature can lead to unexpected OOM killer
643 invocations for high order memory re !! 269 invocations for high order memory requests. You shouldn't
644 disable this option unless there rea !! 270 disable this option unless there really is a strong reason for
645 it and then we would be really inter !! 271 it and then we would be really interested to hear about that at
646 linux-mm@kvack.org. !! 272 linux-mm@kvack.org.
647 <<
648 config COMPACT_UNEVICTABLE_DEFAULT <<
649 int <<
650 depends on COMPACTION <<
651 default 0 if PREEMPT_RT <<
652 default 1 <<
653 <<
654 # <<
655 # support for free page reporting <<
656 config PAGE_REPORTING <<
657 bool "Free page reporting" <<
658 help <<
659 Free page reporting allows for the i <<
660 free pages from the buddy allocator <<
661 those pages to another entity, such <<
662 memory can be freed within the host <<
663 273
664 # 274 #
665 # support for page migration 275 # support for page migration
666 # 276 #
667 config MIGRATION 277 config MIGRATION
668 bool "Page migration" 278 bool "Page migration"
669 default y !! 279 def_bool y
670 depends on (NUMA || ARCH_ENABLE_MEMORY 280 depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU
671 help 281 help
672 Allows the migration of the physical 282 Allows the migration of the physical location of pages of processes
673 while the virtual addresses are not 283 while the virtual addresses are not changed. This is useful in
674 two situations. The first is on NUMA 284 two situations. The first is on NUMA systems to put pages nearer
675 to the processors accessing. The sec 285 to the processors accessing. The second is when allocating huge
676 pages as migration can relocate page 286 pages as migration can relocate pages to satisfy a huge page
677 allocation instead of reclaiming. 287 allocation instead of reclaiming.
678 288
679 config DEVICE_MIGRATION <<
680 def_bool MIGRATION && ZONE_DEVICE <<
681 <<
682 config ARCH_ENABLE_HUGEPAGE_MIGRATION 289 config ARCH_ENABLE_HUGEPAGE_MIGRATION
683 bool 290 bool
684 291
685 config ARCH_ENABLE_THP_MIGRATION <<
686 bool <<
687 <<
688 config HUGETLB_PAGE_SIZE_VARIABLE <<
689 def_bool n <<
690 help <<
691 Allows the pageblock_order value to <<
692 HUGETLB_PAGE_ORDER when there are mu <<
693 on a platform. <<
694 <<
695 Note that the pageblock_order cannot <<
696 clamped down to MAX_PAGE_ORDER. <<
697 <<
698 config CONTIG_ALLOC <<
699 def_bool (MEMORY_ISOLATION && COMPACTI <<
700 <<
701 config PCP_BATCH_SCALE_MAX <<
702 int "Maximum scale factor of PCP (Per- <<
703 default 5 <<
704 range 0 6 <<
705 help <<
706 In page allocator, PCP (Per-CPU page <<
707 batches. The batch number is scaled <<
708 allocation/free throughput. But too <<
709 latency. This option sets the upper <<
710 the maximum latency. <<
711 <<
712 config PHYS_ADDR_T_64BIT 292 config PHYS_ADDR_T_64BIT
713 def_bool 64BIT !! 293 def_bool 64BIT || ARCH_PHYS_ADDR_T_64BIT
714 294
715 config BOUNCE 295 config BOUNCE
716 bool "Enable bounce buffers" 296 bool "Enable bounce buffers"
717 default y 297 default y
718 depends on BLOCK && MMU && HIGHMEM !! 298 depends on BLOCK && MMU && (ZONE_DMA || HIGHMEM)
>> 299 help
>> 300 Enable bounce buffers for devices that cannot access
>> 301 the full range of memory available to the CPU. Enabled
>> 302 by default when ZONE_DMA or HIGHMEM is selected, but you
>> 303 may say n to override this.
>> 304
>> 305 # On the 'tile' arch, USB OHCI needs the bounce pool since tilegx will often
>> 306 # have more than 4GB of memory, but we don't currently use the IOTLB to present
>> 307 # a 32-bit address to OHCI. So we need to use a bounce pool instead.
>> 308 config NEED_BOUNCE_POOL
>> 309 bool
>> 310 default y if TILE && USB_OHCI_HCD
>> 311
>> 312 config NR_QUICK
>> 313 int
>> 314 depends on QUICKLIST
>> 315 default "2" if AVR32
>> 316 default "1"
>> 317
>> 318 config VIRT_TO_BUS
>> 319 bool
719 help 320 help
720 Enable bounce buffers for devices th !! 321 An architecture should select this if it implements the
721 memory available to the CPU. Enabled !! 322 deprecated interface virt_to_bus(). All new architectures
722 selected, but you may say n to overr !! 323 should probably not select this.
>> 324
723 325
724 config MMU_NOTIFIER 326 config MMU_NOTIFIER
725 bool 327 bool
726 select INTERVAL_TREE !! 328 select SRCU
727 329
728 config KSM 330 config KSM
729 bool "Enable KSM for page merging" 331 bool "Enable KSM for page merging"
730 depends on MMU 332 depends on MMU
731 select XXHASH <<
732 help 333 help
733 Enable Kernel Samepage Merging: KSM 334 Enable Kernel Samepage Merging: KSM periodically scans those areas
734 of an application's address space th 335 of an application's address space that an app has advised may be
735 mergeable. When it finds pages of i 336 mergeable. When it finds pages of identical content, it replaces
736 the many instances by a single page 337 the many instances by a single page with that content, so
737 saving memory until one or another a 338 saving memory until one or another app needs to modify the content.
738 Recommended for use with KVM, or wit 339 Recommended for use with KVM, or with other duplicative applications.
739 See Documentation/mm/ksm.rst for mor !! 340 See Documentation/vm/ksm.txt for more information: KSM is inactive
740 until a program has madvised that an 341 until a program has madvised that an area is MADV_MERGEABLE, and
741 root has set /sys/kernel/mm/ksm/run 342 root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set).
742 343
743 config DEFAULT_MMAP_MIN_ADDR 344 config DEFAULT_MMAP_MIN_ADDR
744 int "Low address space to protect from !! 345 int "Low address space to protect from user allocation"
745 depends on MMU 346 depends on MMU
746 default 4096 !! 347 default 4096
747 help !! 348 help
748 This is the portion of low virtual m 349 This is the portion of low virtual memory which should be protected
749 from userspace allocation. Keeping 350 from userspace allocation. Keeping a user from writing to low pages
750 can help reduce the impact of kernel 351 can help reduce the impact of kernel NULL pointer bugs.
751 352
752 For most arm64, ppc64 and x86 users !! 353 For most ia64, ppc64 and x86 users with lots of address space
753 a value of 65536 is reasonable and s 354 a value of 65536 is reasonable and should cause no problems.
754 On arm and other archs it should not 355 On arm and other archs it should not be higher than 32768.
755 Programs which use vm86 functionalit 356 Programs which use vm86 functionality or have some need to map
756 this low address space will need CAP 357 this low address space will need CAP_SYS_RAWIO or disable this
757 protection by setting the value to 0 358 protection by setting the value to 0.
758 359
759 This value can be changed after boot 360 This value can be changed after boot using the
760 /proc/sys/vm/mmap_min_addr tunable. 361 /proc/sys/vm/mmap_min_addr tunable.
761 362
762 config ARCH_SUPPORTS_MEMORY_FAILURE 363 config ARCH_SUPPORTS_MEMORY_FAILURE
763 bool 364 bool
764 365
765 config MEMORY_FAILURE 366 config MEMORY_FAILURE
766 depends on MMU 367 depends on MMU
767 depends on ARCH_SUPPORTS_MEMORY_FAILUR 368 depends on ARCH_SUPPORTS_MEMORY_FAILURE
768 bool "Enable recovery from hardware me 369 bool "Enable recovery from hardware memory errors"
769 select MEMORY_ISOLATION 370 select MEMORY_ISOLATION
770 select RAS 371 select RAS
771 help 372 help
772 Enables code to recover from some me 373 Enables code to recover from some memory failures on systems
773 with MCA recovery. This allows a sys 374 with MCA recovery. This allows a system to continue running
774 even when some of its memory has unc 375 even when some of its memory has uncorrected errors. This requires
775 special hardware support and typical 376 special hardware support and typically ECC memory.
776 377
777 config HWPOISON_INJECT 378 config HWPOISON_INJECT
778 tristate "HWPoison pages injector" 379 tristate "HWPoison pages injector"
779 depends on MEMORY_FAILURE && DEBUG_KER 380 depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS
780 select PROC_PAGE_MONITOR 381 select PROC_PAGE_MONITOR
781 382
782 config NOMMU_INITIAL_TRIM_EXCESS 383 config NOMMU_INITIAL_TRIM_EXCESS
783 int "Turn on mmap() excess space trimm 384 int "Turn on mmap() excess space trimming before booting"
784 depends on !MMU 385 depends on !MMU
785 default 1 386 default 1
786 help 387 help
787 The NOMMU mmap() frequently needs to 388 The NOMMU mmap() frequently needs to allocate large contiguous chunks
788 of memory on which to store mappings 389 of memory on which to store mappings, but it can only ask the system
789 allocator for chunks in 2^N*PAGE_SIZ 390 allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently
790 more than it requires. To deal with 391 more than it requires. To deal with this, mmap() is able to trim off
791 the excess and return it to the allo 392 the excess and return it to the allocator.
792 393
793 If trimming is enabled, the excess i 394 If trimming is enabled, the excess is trimmed off and returned to the
794 system allocator, which can cause ex 395 system allocator, which can cause extra fragmentation, particularly
795 if there are a lot of transient proc 396 if there are a lot of transient processes.
796 397
797 If trimming is disabled, the excess 398 If trimming is disabled, the excess is kept, but not used, which for
798 long-term mappings means that the sp 399 long-term mappings means that the space is wasted.
799 400
800 Trimming can be dynamically controll 401 Trimming can be dynamically controlled through a sysctl option
801 (/proc/sys/vm/nr_trim_pages) which s 402 (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of
802 excess pages there must be before tr 403 excess pages there must be before trimming should occur, or zero if
803 no trimming is to occur. 404 no trimming is to occur.
804 405
805 This option specifies the initial va 406 This option specifies the initial value of this option. The default
806 of 1 says that all excess pages shou 407 of 1 says that all excess pages should be trimmed.
807 408
808 See Documentation/admin-guide/mm/nom !! 409 See Documentation/nommu-mmap.txt for more information.
809 <<
810 config ARCH_WANT_GENERAL_HUGETLB <<
811 bool <<
812 410
813 config ARCH_WANTS_THP_SWAP !! 411 config TRANSPARENT_HUGEPAGE
814 def_bool n <<
815 <<
816 menuconfig TRANSPARENT_HUGEPAGE <<
817 bool "Transparent Hugepage Support" 412 bool "Transparent Hugepage Support"
818 depends on HAVE_ARCH_TRANSPARENT_HUGEP !! 413 depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE
819 select COMPACTION 414 select COMPACTION
820 select XARRAY_MULTI !! 415 select RADIX_TREE_MULTIORDER
821 help 416 help
822 Transparent Hugepages allows the ker 417 Transparent Hugepages allows the kernel to use huge pages and
823 huge tlb transparently to the applic 418 huge tlb transparently to the applications whenever possible.
824 This feature can improve computing p 419 This feature can improve computing performance to certain
825 applications by speeding up page fau 420 applications by speeding up page faults during memory
826 allocation, by reducing the number o 421 allocation, by reducing the number of tlb misses and by speeding
827 up the pagetable walking. 422 up the pagetable walking.
828 423
829 If memory constrained on embedded, y 424 If memory constrained on embedded, you may want to say N.
830 425
831 if TRANSPARENT_HUGEPAGE <<
832 <<
833 choice 426 choice
834 prompt "Transparent Hugepage Support s 427 prompt "Transparent Hugepage Support sysfs defaults"
835 depends on TRANSPARENT_HUGEPAGE 428 depends on TRANSPARENT_HUGEPAGE
836 default TRANSPARENT_HUGEPAGE_ALWAYS 429 default TRANSPARENT_HUGEPAGE_ALWAYS
837 help 430 help
838 Selects the sysfs defaults for Trans 431 Selects the sysfs defaults for Transparent Hugepage Support.
839 432
840 config TRANSPARENT_HUGEPAGE_ALWAYS 433 config TRANSPARENT_HUGEPAGE_ALWAYS
841 bool "always" 434 bool "always"
842 help 435 help
843 Enabling Transparent Hugepage always 436 Enabling Transparent Hugepage always, can increase the
844 memory footprint of applications wit 437 memory footprint of applications without a guaranteed
845 benefit but it will work automatical 438 benefit but it will work automatically for all applications.
846 439
847 config TRANSPARENT_HUGEPAGE_MADVISE 440 config TRANSPARENT_HUGEPAGE_MADVISE
848 bool "madvise" 441 bool "madvise"
849 help 442 help
850 Enabling Transparent Hugepage madvis 443 Enabling Transparent Hugepage madvise, will only provide a
851 performance improvement benefit to t 444 performance improvement benefit to the applications using
852 madvise(MADV_HUGEPAGE) but it won't 445 madvise(MADV_HUGEPAGE) but it won't risk to increase the
853 memory footprint of applications wit 446 memory footprint of applications without a guaranteed
854 benefit. 447 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 448 endchoice
862 449
863 config THP_SWAP <<
864 def_bool y <<
865 depends on TRANSPARENT_HUGEPAGE && ARC <<
866 help <<
867 Swap transparent huge pages in one p <<
868 XXX: For now, swap cluster backing t <<
869 will be split after swapout. <<
870 <<
871 For selection by architectures with <<
872 <<
873 config READ_ONLY_THP_FOR_FS <<
874 bool "Read-only THP for filesystems (E <<
875 depends on TRANSPARENT_HUGEPAGE && SHM <<
876 <<
877 help <<
878 Allow khugepaged to put read-only fi <<
879 <<
880 This is marked experimental because <<
881 support of file THPs will be develop <<
882 cycles. <<
883 <<
884 endif # TRANSPARENT_HUGEPAGE <<
885 <<
886 # 450 #
887 # The architecture supports pgtable leaves tha !! 451 # We don't deposit page tables on file THP mapping,
>> 452 # but Power makes use of them to address MMU quirk.
888 # 453 #
889 config PGTABLE_HAS_HUGE_LEAVES !! 454 config TRANSPARENT_HUGE_PAGECACHE
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 455 def_bool y
899 depends on ARCH_SUPPORTS_HUGE_PFNMAP & !! 456 depends on TRANSPARENT_HUGEPAGE && !PPC
900 <<
901 config ARCH_SUPPORTS_PUD_PFNMAP <<
902 def_bool y <<
903 depends on ARCH_SUPPORTS_HUGE_PFNMAP & <<
904 457
905 # 458 #
906 # UP and nommu archs use km based percpu alloc 459 # UP and nommu archs use km based percpu allocator
907 # 460 #
908 config NEED_PER_CPU_KM 461 config NEED_PER_CPU_KM
909 depends on !SMP || !MMU !! 462 depends on !SMP
910 bool 463 bool
911 default y 464 default y
912 465
913 config NEED_PER_CPU_EMBED_FIRST_CHUNK !! 466 config CLEANCACHE
914 bool !! 467 bool "Enable cleancache driver to cache clean pages if tmem is present"
>> 468 default n
>> 469 help
>> 470 Cleancache can be thought of as a page-granularity victim cache
>> 471 for clean pages that the kernel's pageframe replacement algorithm
>> 472 (PFRA) would like to keep around, but can't since there isn't enough
>> 473 memory. So when the PFRA "evicts" a page, it first attempts to use
>> 474 cleancache code to put the data contained in that page into
>> 475 "transcendent memory", memory that is not directly accessible or
>> 476 addressable by the kernel and is of unknown and possibly
>> 477 time-varying size. And when a cleancache-enabled
>> 478 filesystem wishes to access a page in a file on disk, it first
>> 479 checks cleancache to see if it already contains it; if it does,
>> 480 the page is copied into the kernel and a disk access is avoided.
>> 481 When a transcendent memory driver is available (such as zcache or
>> 482 Xen transcendent memory), a significant I/O reduction
>> 483 may be achieved. When none is available, all cleancache calls
>> 484 are reduced to a single pointer-compare-against-NULL resulting
>> 485 in a negligible performance hit.
915 486
916 config NEED_PER_CPU_PAGE_FIRST_CHUNK !! 487 If unsure, say Y to enable cleancache
917 bool <<
918 488
919 config USE_PERCPU_NUMA_NODE_ID !! 489 config FRONTSWAP
920 bool !! 490 bool "Enable frontswap to cache swap pages if tmem is present"
>> 491 depends on SWAP
>> 492 default n
>> 493 help
>> 494 Frontswap is so named because it can be thought of as the opposite
>> 495 of a "backing" store for a swap device. The data is stored into
>> 496 "transcendent memory", memory that is not directly accessible or
>> 497 addressable by the kernel and is of unknown and possibly
>> 498 time-varying size. When space in transcendent memory is available,
>> 499 a significant swap I/O reduction may be achieved. When none is
>> 500 available, all frontswap calls are reduced to a single pointer-
>> 501 compare-against-NULL resulting in a negligible performance hit
>> 502 and swap data is stored as normal on the matching swap device.
921 503
922 config HAVE_SETUP_PER_CPU_AREA !! 504 If unsure, say Y to enable frontswap.
923 bool <<
924 505
925 config CMA 506 config CMA
926 bool "Contiguous Memory Allocator" 507 bool "Contiguous Memory Allocator"
927 depends on MMU !! 508 depends on HAVE_MEMBLOCK && MMU
928 select MIGRATION 509 select MIGRATION
929 select MEMORY_ISOLATION 510 select MEMORY_ISOLATION
930 help 511 help
931 This enables the Contiguous Memory A 512 This enables the Contiguous Memory Allocator which allows other
932 subsystems to allocate big physicall 513 subsystems to allocate big physically-contiguous blocks of memory.
933 CMA reserves a region of memory and 514 CMA reserves a region of memory and allows only movable pages to
934 be allocated from it. This way, the 515 be allocated from it. This way, the kernel can use the memory for
935 pagecache and when a subsystem reque 516 pagecache and when a subsystem requests for contiguous area, the
936 allocated pages are migrated away to 517 allocated pages are migrated away to serve the contiguous request.
937 518
938 If unsure, say "n". 519 If unsure, say "n".
939 520
>> 521 config CMA_DEBUG
>> 522 bool "CMA debug messages (DEVELOPMENT)"
>> 523 depends on DEBUG_KERNEL && CMA
>> 524 help
>> 525 Turns on debug messages in CMA. This produces KERN_DEBUG
>> 526 messages for every CMA call as well as various messages while
>> 527 processing calls such as dma_alloc_from_contiguous().
>> 528 This option does not affect warning and error messages.
>> 529
940 config CMA_DEBUGFS 530 config CMA_DEBUGFS
941 bool "CMA debugfs interface" 531 bool "CMA debugfs interface"
942 depends on CMA && DEBUG_FS 532 depends on CMA && DEBUG_FS
943 help 533 help
944 Turns on the DebugFS interface for C 534 Turns on the DebugFS interface for CMA.
945 535
946 config CMA_SYSFS <<
947 bool "CMA information through sysfs in <<
948 depends on CMA && SYSFS <<
949 help <<
950 This option exposes some sysfs attri <<
951 from CMA. <<
952 <<
953 config CMA_AREAS 536 config CMA_AREAS
954 int "Maximum count of the CMA areas" 537 int "Maximum count of the CMA areas"
955 depends on CMA 538 depends on CMA
956 default 20 if NUMA !! 539 default 7
957 default 8 <<
958 help 540 help
959 CMA allows to create CMA areas for p 541 CMA allows to create CMA areas for particular purpose, mainly,
960 used as device private area. This pa 542 used as device private area. This parameter sets the maximum
961 number of CMA area in the system. 543 number of CMA area in the system.
962 544
963 If unsure, leave the default value " !! 545 If unsure, leave the default value "7".
964 546
965 config MEM_SOFT_DIRTY 547 config MEM_SOFT_DIRTY
966 bool "Track memory changes" 548 bool "Track memory changes"
967 depends on CHECKPOINT_RESTORE && HAVE_ 549 depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS
968 select PROC_PAGE_MONITOR 550 select PROC_PAGE_MONITOR
969 help 551 help
970 This option enables memory changes t 552 This option enables memory changes tracking by introducing a
971 soft-dirty bit on pte-s. This bit it 553 soft-dirty bit on pte-s. This bit it set when someone writes
972 into a page just as regular dirty bi 554 into a page just as regular dirty bit, but unlike the latter
973 it can be cleared by hands. 555 it can be cleared by hands.
974 556
975 See Documentation/admin-guide/mm/sof !! 557 See Documentation/vm/soft-dirty.txt for more details.
>> 558
>> 559 config ZSWAP
>> 560 bool "Compressed cache for swap pages (EXPERIMENTAL)"
>> 561 depends on FRONTSWAP && CRYPTO=y
>> 562 select CRYPTO_LZO
>> 563 select ZPOOL
>> 564 default n
>> 565 help
>> 566 A lightweight compressed cache for swap pages. It takes
>> 567 pages that are in the process of being swapped out and attempts to
>> 568 compress them into a dynamically allocated RAM-based memory pool.
>> 569 This can result in a significant I/O reduction on swap device and,
>> 570 in the case where decompressing from RAM is faster that swap device
>> 571 reads, can also improve workload performance.
>> 572
>> 573 This is marked experimental because it is a new feature (as of
>> 574 v3.11) that interacts heavily with memory reclaim. While these
>> 575 interactions don't cause any known issues on simple memory setups,
>> 576 they have not be fully explored on the large set of potential
>> 577 configurations and workloads that exist.
>> 578
>> 579 config ZPOOL
>> 580 tristate "Common API for compressed memory storage"
>> 581 default n
>> 582 help
>> 583 Compressed memory storage API. This allows using either zbud or
>> 584 zsmalloc.
>> 585
>> 586 config ZBUD
>> 587 tristate "Low (Up to 2x) density storage for compressed pages"
>> 588 default n
>> 589 help
>> 590 A special purpose allocator for storing compressed pages.
>> 591 It is designed to store up to two compressed pages per physical
>> 592 page. While this design limits storage density, it has simple and
>> 593 deterministic reclaim properties that make it preferable to a higher
>> 594 density approach when reclaim will be used.
>> 595
>> 596 config Z3FOLD
>> 597 tristate "Up to 3x density storage for compressed pages"
>> 598 depends on ZPOOL
>> 599 default n
>> 600 help
>> 601 A special purpose allocator for storing compressed pages.
>> 602 It is designed to store up to three compressed pages per physical
>> 603 page. It is a ZBUD derivative so the simplicity and determinism are
>> 604 still there.
>> 605
>> 606 config ZSMALLOC
>> 607 tristate "Memory allocator for compressed pages"
>> 608 depends on MMU
>> 609 default n
>> 610 help
>> 611 zsmalloc is a slab-based memory allocator designed to store
>> 612 compressed RAM pages. zsmalloc uses virtual memory mapping
>> 613 in order to reduce fragmentation. However, this results in a
>> 614 non-standard allocator interface where a handle, not a pointer, is
>> 615 returned by an alloc(). This handle must be mapped in order to
>> 616 access the allocated space.
>> 617
>> 618 config PGTABLE_MAPPING
>> 619 bool "Use page table mapping to access object in zsmalloc"
>> 620 depends on ZSMALLOC
>> 621 help
>> 622 By default, zsmalloc uses a copy-based object mapping method to
>> 623 access allocations that span two pages. However, if a particular
>> 624 architecture (ex, ARM) performs VM mapping faster than copying,
>> 625 then you should select this. This causes zsmalloc to use page table
>> 626 mapping rather than copying for object mapping.
>> 627
>> 628 You can check speed with zsmalloc benchmark:
>> 629 https://github.com/spartacus06/zsmapbench
>> 630
>> 631 config ZSMALLOC_STAT
>> 632 bool "Export zsmalloc statistics"
>> 633 depends on ZSMALLOC
>> 634 select DEBUG_FS
>> 635 help
>> 636 This option enables code in the zsmalloc to collect various
>> 637 statistics about whats happening in zsmalloc and exports that
>> 638 information to userspace via debugfs.
>> 639 If unsure, say N.
976 640
977 config GENERIC_EARLY_IOREMAP 641 config GENERIC_EARLY_IOREMAP
978 bool 642 bool
979 643
980 config STACK_MAX_DEFAULT_SIZE_MB !! 644 config MAX_STACK_SIZE_MB
981 int "Default maximum user stack size f !! 645 int "Maximum user stack size for 32-bit processes (MB)"
982 default 100 !! 646 default 80
>> 647 range 8 256 if METAG
983 range 8 2048 648 range 8 2048
984 depends on STACK_GROWSUP && (!64BIT || 649 depends on STACK_GROWSUP && (!64BIT || COMPAT)
985 help 650 help
986 This is the maximum stack size in Me 651 This is the maximum stack size in Megabytes in the VM layout of 32-bit
987 user processes when the stack grows 652 user processes when the stack grows upwards (currently only on parisc
988 arch) when the RLIMIT_STACK hard lim !! 653 and metag arch). The stack will be located at the highest memory
>> 654 address minus the given value, unless the RLIMIT_STACK hard limit is
>> 655 changed to a smaller value in which case that is used.
>> 656
>> 657 A sane initial value is 80 MB.
989 658
990 A sane initial value is 100 MB. !! 659 # For architectures that support deferred memory initialisation
>> 660 config ARCH_SUPPORTS_DEFERRED_STRUCT_PAGE_INIT
>> 661 bool
991 662
992 config DEFERRED_STRUCT_PAGE_INIT 663 config DEFERRED_STRUCT_PAGE_INIT
993 bool "Defer initialisation of struct p 664 bool "Defer initialisation of struct pages to kthreads"
994 depends on SPARSEMEM !! 665 default n
>> 666 depends on ARCH_SUPPORTS_DEFERRED_STRUCT_PAGE_INIT
>> 667 depends on NO_BOOTMEM && MEMORY_HOTPLUG
>> 668 depends on !FLATMEM
995 depends on !NEED_PER_CPU_KM 669 depends on !NEED_PER_CPU_KM
996 depends on 64BIT <<
997 depends on !KMSAN <<
998 select PADATA <<
999 help 670 help
1000 Ordinarily all struct pages are ini 671 Ordinarily all struct pages are initialised during early boot in a
1001 single thread. On very large machin 672 single thread. On very large machines this can take a considerable
1002 amount of time. If this option is s 673 amount of time. If this option is set, large machines will bring up
1003 a subset of memmap at boot and then !! 674 a subset of memmap at boot and then initialise the rest in parallel
1004 This has a potential performance im !! 675 by starting one-off "pgdatinitX" kernel thread for each node X. This
>> 676 has a potential performance impact on processes running early in the
1005 lifetime of the system until these 677 lifetime of the system until these kthreads finish the
1006 initialisation. 678 initialisation.
1007 679
1008 config PAGE_IDLE_FLAG <<
1009 bool <<
1010 select PAGE_EXTENSION if !64BIT <<
1011 help <<
1012 This adds PG_idle and PG_young flag <<
1013 bit writers can set the state of th <<
1014 Accessed bit readers may avoid dist <<
1015 <<
1016 config IDLE_PAGE_TRACKING 680 config IDLE_PAGE_TRACKING
1017 bool "Enable idle page tracking" 681 bool "Enable idle page tracking"
1018 depends on SYSFS && MMU 682 depends on SYSFS && MMU
1019 select PAGE_IDLE_FLAG !! 683 select PAGE_EXTENSION if !64BIT
1020 help 684 help
1021 This feature allows to estimate the 685 This feature allows to estimate the amount of user pages that have
1022 not been touched during a given per 686 not been touched during a given period of time. This information can
1023 be useful to tune memory cgroup lim 687 be useful to tune memory cgroup limits and/or for job placement
1024 within a compute cluster. 688 within a compute cluster.
1025 689
1026 See Documentation/admin-guide/mm/id !! 690 See Documentation/vm/idle_page_tracking.txt for more details.
1027 more details. <<
1028 <<
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 <<
1036 bool <<
1037 <<
1038 config ARCH_HAS_CURRENT_STACK_POINTER <<
1039 bool <<
1040 help <<
1041 In support of HARDENED_USERCOPY per <<
1042 checking, an architecture-agnostic <<
1043 is needed. Once an architecture def <<
1044 register alias named "current_stack <<
1045 selected. <<
1046 <<
1047 config ARCH_HAS_PTE_DEVMAP <<
1048 bool <<
1049 <<
1050 config ARCH_HAS_ZONE_DMA_SET <<
1051 bool <<
1052 <<
1053 config ZONE_DMA <<
1054 bool "Support DMA zone" if ARCH_HAS_Z <<
1055 default y if ARM64 || X86 <<
1056 <<
1057 config ZONE_DMA32 <<
1058 bool "Support DMA32 zone" if ARCH_HAS <<
1059 depends on !X86_32 <<
1060 default y if ARM64 <<
1061 691
1062 config ZONE_DEVICE 692 config ZONE_DEVICE
1063 bool "Device memory (pmem, HMM, etc.. !! 693 bool "Device memory (pmem, etc...) hotplug support"
1064 depends on MEMORY_HOTPLUG 694 depends on MEMORY_HOTPLUG
1065 depends on MEMORY_HOTREMOVE 695 depends on MEMORY_HOTREMOVE
1066 depends on SPARSEMEM_VMEMMAP 696 depends on SPARSEMEM_VMEMMAP
1067 depends on ARCH_HAS_PTE_DEVMAP !! 697 depends on X86_64 #arch_add_memory() comprehends device memory
1068 select XARRAY_MULTI <<
1069 698
1070 help 699 help
1071 Device memory hotplug support allow 700 Device memory hotplug support allows for establishing pmem,
1072 or other device driver discovered m 701 or other device driver discovered memory regions, in the
1073 memmap. This allows pfn_to_page() l 702 memmap. This allows pfn_to_page() lookups of otherwise
1074 "device-physical" addresses which i 703 "device-physical" addresses which is needed for using a DAX
1075 mapping in an O_DIRECT operation, a 704 mapping in an O_DIRECT operation, among other things.
1076 705
1077 If FS_DAX is enabled, then say Y. 706 If FS_DAX is enabled, then say Y.
1078 707
1079 # !! 708 config FRAME_VECTOR
1080 # Helpers to mirror range of the CPU page tab <<
1081 # tables. <<
1082 # <<
1083 config HMM_MIRROR <<
1084 bool <<
1085 depends on MMU <<
1086 <<
1087 config GET_FREE_REGION <<
1088 bool <<
1089 <<
1090 config DEVICE_PRIVATE <<
1091 bool "Unaddressable device memory (GP <<
1092 depends on ZONE_DEVICE <<
1093 select GET_FREE_REGION <<
1094 <<
1095 help <<
1096 Allows creation of struct pages to <<
1097 memory; i.e., memory that is only a <<
1098 group of devices). You likely also <<
1099 <<
1100 config VMAP_PFN <<
1101 bool 709 bool
1102 710
1103 config ARCH_USES_HIGH_VMA_FLAGS 711 config ARCH_USES_HIGH_VMA_FLAGS
1104 bool 712 bool
1105 config ARCH_HAS_PKEYS 713 config ARCH_HAS_PKEYS
1106 bool 714 bool
1107 <<
1108 config ARCH_USES_PG_ARCH_2 <<
1109 bool <<
1110 config ARCH_USES_PG_ARCH_3 <<
1111 bool <<
1112 <<
1113 config VM_EVENT_COUNTERS <<
1114 default y <<
1115 bool "Enable VM event counters for /p <<
1116 help <<
1117 VM event counters are needed for ev <<
1118 This option allows the disabling of <<
1119 on EXPERT systems. /proc/vmstat wi <<
1120 if VM event counters are disabled. <<
1121 <<
1122 config PERCPU_STATS <<
1123 bool "Collect percpu memory statistic <<
1124 help <<
1125 This feature collects and exposes s <<
1126 information includes global and per <<
1127 be used to help understand percpu m <<
1128 <<
1129 config GUP_TEST <<
1130 bool "Enable infrastructure for get_u <<
1131 depends on DEBUG_FS <<
1132 help <<
1133 Provides /sys/kernel/debug/gup_test <<
1134 to make ioctl calls that can launch <<
1135 the get_user_pages*() and pin_user_ <<
1136 <<
1137 These tests include benchmark testi <<
1138 get_user_pages*() and pin_user_page <<
1139 the non-_fast variants. <<
1140 <<
1141 There is also a sub-test that allow <<
1142 of up to eight pages (selected by c <<
1143 range of user-space addresses. Thes <<
1144 pin_user_pages*(), or pinned via ge <<
1145 by other command line arguments. <<
1146 <<
1147 See tools/testing/selftests/mm/gup_ <<
1148 <<
1149 comment "GUP_TEST needs to have DEBUG_FS enab <<
1150 depends on !GUP_TEST && !DEBUG_FS <<
1151 <<
1152 config GUP_GET_PXX_LOW_HIGH <<
1153 bool <<
1154 <<
1155 config DMAPOOL_TEST <<
1156 tristate "Enable a module to run time <<
1157 depends on HAS_DMA <<
1158 help <<
1159 Provides a test module that will al <<
1160 various sizes and report how long i <<
1161 provide a consistent way to measure <<
1162 dma_pool_alloc/free routines affect <<
1163 <<
1164 config ARCH_HAS_PTE_SPECIAL <<
1165 bool <<
1166 <<
1167 config MAPPING_DIRTY_HELPERS <<
1168 bool <<
1169 <<
1170 config KMAP_LOCAL <<
1171 bool <<
1172 <<
1173 config KMAP_LOCAL_NON_LINEAR_PTE_ARRAY <<
1174 bool <<
1175 <<
1176 # struct io_mapping based helper. Selected b <<
1177 config IO_MAPPING <<
1178 bool <<
1179 <<
1180 config MEMFD_CREATE <<
1181 bool "Enable memfd_create() system ca <<
1182 <<
1183 config SECRETMEM <<
1184 default y <<
1185 bool "Enable memfd_secret() system ca <<
1186 depends on ARCH_HAS_SET_DIRECT_MAP <<
1187 help <<
1188 Enable the memfd_secret() system ca <<
1189 memory areas visible only in the co <<
1190 not mapped to other processes and o <<
1191 <<
1192 config ANON_VMA_NAME <<
1193 bool "Anonymous VMA name support" <<
1194 depends on PROC_FS && ADVISE_SYSCALLS <<
1195 <<
1196 help <<
1197 Allow naming anonymous virtual memo <<
1198 <<
1199 This feature allows assigning names <<
1200 names can be later retrieved from / <<
1201 and help identifying individual ano <<
1202 Assigning a name to anonymous virtu <<
1203 area from being merged with adjacen <<
1204 difference in their name. <<
1205 <<
1206 config HAVE_ARCH_USERFAULTFD_WP <<
1207 bool <<
1208 help <<
1209 Arch has userfaultfd write protecti <<
1210 <<
1211 config HAVE_ARCH_USERFAULTFD_MINOR <<
1212 bool <<
1213 help <<
1214 Arch has userfaultfd minor fault su <<
1215 <<
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 <<
1225 bool "Userfaultfd write protection su <<
1226 default y <<
1227 depends on HAVE_ARCH_USERFAULTFD_WP <<
1228 <<
1229 help <<
1230 Allows to create marker PTEs for us <<
1231 purposes. It is required to enable <<
1232 file-backed memory types like shmem <<
1233 endif # USERFAULTFD <<
1234 <<
1235 # multi-gen LRU { <<
1236 config LRU_GEN <<
1237 bool "Multi-Gen LRU" <<
1238 depends on MMU <<
1239 # make sure folio->flags has enough s <<
1240 depends on 64BIT || !SPARSEMEM || SPA <<
1241 help <<
1242 A high performance LRU implementati <<
1243 Documentation/admin-guide/mm/multig <<
1244 <<
1245 config LRU_GEN_ENABLED <<
1246 bool "Enable by default" <<
1247 depends on LRU_GEN <<
1248 help <<
1249 This option enables the multi-gen L <<
1250 <<
1251 config LRU_GEN_STATS <<
1252 bool "Full stats for debugging" <<
1253 depends on LRU_GEN <<
1254 help <<
1255 Do not enable this option unless yo <<
1256 from evicted generations for debugg <<
1257 <<
1258 This option has a per-memcg and per <<
1259 <<
1260 config LRU_GEN_WALKS_MMU <<
1261 def_bool y <<
1262 depends on LRU_GEN && ARCH_HAS_HW_PTE <<
1263 # } <<
1264 <<
1265 config ARCH_SUPPORTS_PER_VMA_LOCK <<
1266 def_bool n <<
1267 <<
1268 config PER_VMA_LOCK <<
1269 def_bool y <<
1270 depends on ARCH_SUPPORTS_PER_VMA_LOCK <<
1271 help <<
1272 Allow per-vma locking during page f <<
1273 <<
1274 This feature allows locking each vi <<
1275 handling page faults instead of tak <<
1276 <<
1277 config LOCK_MM_AND_FIND_VMA <<
1278 bool <<
1279 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 <<
1298 source "mm/damon/Kconfig" <<
1299 <<
1300 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.