tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1# SPDX-License-Identifier: GPL-2.0-only
2
3menu "Memory Management options"
4
5#
6# For some reason microblaze and nios2 hard code SWAP=n. Hopefully we can
7# add proper SWAP support to them, in which case this can be remove.
8#
9config ARCH_NO_SWAP
10 bool
11
12config ZPOOL
13 bool
14
15menuconfig SWAP
16 bool "Support for paging of anonymous memory (swap)"
17 depends on MMU && BLOCK && !ARCH_NO_SWAP
18 default y
19 help
20 This option allows you to choose whether you want to have support
21 for so called swap devices or swap files in your kernel that are
22 used to provide more virtual memory than the actual RAM present
23 in your computer. If unsure say Y.
24
25config 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 swap pages. It takes
32 pages that are in the process of being swapped out and attempts to
33 compress them into a dynamically allocated RAM-based memory pool.
34 This can result in a significant I/O reduction on swap device and,
35 in the case where decompressing from RAM is faster than swap device
36 reads, can also improve workload performance.
37
38config ZSWAP_DEFAULT_ON
39 bool "Enable the compressed cache for swap pages by default"
40 depends on ZSWAP
41 help
42 If selected, the compressed cache for swap pages will be enabled
43 at boot, otherwise it will be disabled.
44
45 The selection made here can be overridden by using the kernel
46 command line 'zswap.enabled=' option.
47
48config ZSWAP_EXCLUSIVE_LOADS_DEFAULT_ON
49 bool "Invalidate zswap entries when pages are loaded"
50 depends on ZSWAP
51 help
52 If selected, exclusive loads for zswap will be enabled at boot,
53 otherwise it will be disabled.
54
55 If exclusive loads are enabled, when a page is loaded from zswap,
56 the zswap entry is invalidated at once, as opposed to leaving it
57 in zswap until the swap entry is freed.
58
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.
63
64choice
65 prompt "Default compressor"
66 depends on ZSWAP
67 default ZSWAP_COMPRESSOR_DEFAULT_LZO
68 help
69 Selects the default compression algorithm for the compressed cache
70 for swap pages.
71
72 For an overview what kind of performance can be expected from
73 a particular compression algorithm please refer to the benchmarks
74 available at the following LWN page:
75 https://lwn.net/Articles/751795/
76
77 If in doubt, select 'LZO'.
78
79 The selection made here can be overridden by using the kernel
80 command line 'zswap.compressor=' option.
81
82config ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
83 bool "Deflate"
84 select CRYPTO_DEFLATE
85 help
86 Use the Deflate algorithm as the default compression algorithm.
87
88config ZSWAP_COMPRESSOR_DEFAULT_LZO
89 bool "LZO"
90 select CRYPTO_LZO
91 help
92 Use the LZO algorithm as the default compression algorithm.
93
94config ZSWAP_COMPRESSOR_DEFAULT_842
95 bool "842"
96 select CRYPTO_842
97 help
98 Use the 842 algorithm as the default compression algorithm.
99
100config ZSWAP_COMPRESSOR_DEFAULT_LZ4
101 bool "LZ4"
102 select CRYPTO_LZ4
103 help
104 Use the LZ4 algorithm as the default compression algorithm.
105
106config ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
107 bool "LZ4HC"
108 select CRYPTO_LZ4HC
109 help
110 Use the LZ4HC algorithm as the default compression algorithm.
111
112config ZSWAP_COMPRESSOR_DEFAULT_ZSTD
113 bool "zstd"
114 select CRYPTO_ZSTD
115 help
116 Use the zstd algorithm as the default compression algorithm.
117endchoice
118
119config ZSWAP_COMPRESSOR_DEFAULT
120 string
121 depends on ZSWAP
122 default "deflate" if ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
123 default "lzo" if ZSWAP_COMPRESSOR_DEFAULT_LZO
124 default "842" if ZSWAP_COMPRESSOR_DEFAULT_842
125 default "lz4" if ZSWAP_COMPRESSOR_DEFAULT_LZ4
126 default "lz4hc" if ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
127 default "zstd" if ZSWAP_COMPRESSOR_DEFAULT_ZSTD
128 default ""
129
130choice
131 prompt "Default allocator"
132 depends on ZSWAP
133 default ZSWAP_ZPOOL_DEFAULT_ZBUD
134 help
135 Selects the default allocator for the compressed cache for
136 swap pages.
137 The default is 'zbud' for compatibility, however please do
138 read the description of each of the allocators below before
139 making a right choice.
140
141 The selection made here can be overridden by using the kernel
142 command line 'zswap.zpool=' option.
143
144config ZSWAP_ZPOOL_DEFAULT_ZBUD
145 bool "zbud"
146 select ZBUD
147 help
148 Use the zbud allocator as the default allocator.
149
150config ZSWAP_ZPOOL_DEFAULT_Z3FOLD
151 bool "z3fold"
152 select Z3FOLD
153 help
154 Use the z3fold allocator as the default allocator.
155
156config ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
157 bool "zsmalloc"
158 select ZSMALLOC
159 help
160 Use the zsmalloc allocator as the default allocator.
161endchoice
162
163config ZSWAP_ZPOOL_DEFAULT
164 string
165 depends on ZSWAP
166 default "zbud" if ZSWAP_ZPOOL_DEFAULT_ZBUD
167 default "z3fold" if ZSWAP_ZPOOL_DEFAULT_Z3FOLD
168 default "zsmalloc" if ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
169 default ""
170
171config ZBUD
172 tristate "2:1 compression allocator (zbud)"
173 depends on ZSWAP
174 help
175 A special purpose allocator for storing compressed pages.
176 It is designed to store up to two compressed pages per physical
177 page. While this design limits storage density, it has simple and
178 deterministic reclaim properties that make it preferable to a higher
179 density approach when reclaim will be used.
180
181config Z3FOLD
182 tristate "3:1 compression allocator (z3fold)"
183 depends on ZSWAP
184 help
185 A special purpose allocator for storing compressed pages.
186 It is designed to store up to three compressed pages per physical
187 page. It is a ZBUD derivative so the simplicity and determinism are
188 still there.
189
190config ZSMALLOC
191 tristate
192 prompt "N:1 compression allocator (zsmalloc)" if ZSWAP
193 depends on MMU
194 help
195 zsmalloc is a slab-based memory allocator designed to store
196 pages of various compression levels efficiently. It achieves
197 the highest storage density with the least amount of fragmentation.
198
199config ZSMALLOC_STAT
200 bool "Export zsmalloc statistics"
201 depends on ZSMALLOC
202 select DEBUG_FS
203 help
204 This option enables code in the zsmalloc to collect various
205 statistics about what's happening in zsmalloc and exports that
206 information to userspace via debugfs.
207 If unsure, say N.
208
209config ZSMALLOC_CHAIN_SIZE
210 int "Maximum number of physical pages per-zspage"
211 default 8
212 range 4 16
213 depends on ZSMALLOC
214 help
215 This option sets the upper limit on the number of physical pages
216 that a zmalloc page (zspage) can consist of. The optimal zspage
217 chain size is calculated for each size class during the
218 initialization of the pool.
219
220 Changing this option can alter the characteristics of size classes,
221 such as the number of pages per zspage and the number of objects
222 per zspage. This can also result in different configurations of
223 the pool, as zsmalloc merges size classes with similar
224 characteristics.
225
226 For more information, see zsmalloc documentation.
227
228menu "SLAB allocator options"
229
230choice
231 prompt "Choose SLAB allocator"
232 default SLUB
233 help
234 This option allows to select a slab allocator.
235
236config SLAB_DEPRECATED
237 bool "SLAB (DEPRECATED)"
238 depends on !PREEMPT_RT
239 help
240 Deprecated and scheduled for removal in a few cycles. Replaced by
241 SLUB.
242
243 If you cannot migrate to SLUB, please contact linux-mm@kvack.org
244 and the people listed in the SLAB ALLOCATOR section of MAINTAINERS
245 file, explaining why.
246
247 The regular slab allocator that is established and known to work
248 well in all environments. It organizes cache hot objects in
249 per cpu and per node queues.
250
251config SLUB
252 bool "SLUB (Unqueued Allocator)"
253 help
254 SLUB is a slab allocator that minimizes cache line usage
255 instead of managing queues of cached objects (SLAB approach).
256 Per cpu caching is realized using slabs of objects instead
257 of queues of objects. SLUB can use memory efficiently
258 and has enhanced diagnostics. SLUB is the default choice for
259 a slab allocator.
260
261endchoice
262
263config SLAB
264 bool
265 default y
266 depends on SLAB_DEPRECATED
267
268config SLUB_TINY
269 bool "Configure SLUB for minimal memory footprint"
270 depends on SLUB && EXPERT
271 select SLAB_MERGE_DEFAULT
272 help
273 Configures the SLUB allocator in a way to achieve minimal memory
274 footprint, sacrificing scalability, debugging and other features.
275 This is intended only for the smallest system that had used the
276 SLOB allocator and is not recommended for systems with more than
277 16MB RAM.
278
279 If unsure, say N.
280
281config SLAB_MERGE_DEFAULT
282 bool "Allow slab caches to be merged"
283 default y
284 depends on SLAB || SLUB
285 help
286 For reduced kernel memory fragmentation, slab caches can be
287 merged when they share the same size and other characteristics.
288 This carries a risk of kernel heap overflows being able to
289 overwrite objects from merged caches (and more easily control
290 cache layout), which makes such heap attacks easier to exploit
291 by attackers. By keeping caches unmerged, these kinds of exploits
292 can usually only damage objects in the same cache. To disable
293 merging at runtime, "slab_nomerge" can be passed on the kernel
294 command line.
295
296config SLAB_FREELIST_RANDOM
297 bool "Randomize slab freelist"
298 depends on SLAB || (SLUB && !SLUB_TINY)
299 help
300 Randomizes the freelist order used on creating new pages. This
301 security feature reduces the predictability of the kernel slab
302 allocator against heap overflows.
303
304config SLAB_FREELIST_HARDENED
305 bool "Harden slab freelist metadata"
306 depends on SLAB || (SLUB && !SLUB_TINY)
307 help
308 Many kernel heap attacks try to target slab cache metadata and
309 other infrastructure. This options makes minor performance
310 sacrifices to harden the kernel slab allocator against common
311 freelist exploit methods. Some slab implementations have more
312 sanity-checking than others. This option is most effective with
313 CONFIG_SLUB.
314
315config SLUB_STATS
316 default n
317 bool "Enable SLUB performance statistics"
318 depends on SLUB && SYSFS && !SLUB_TINY
319 help
320 SLUB statistics are useful to debug SLUBs allocation behavior in
321 order find ways to optimize the allocator. This should never be
322 enabled for production use since keeping statistics slows down
323 the allocator by a few percentage points. The slabinfo command
324 supports the determination of the most active slabs to figure
325 out which slabs are relevant to a particular load.
326 Try running: slabinfo -DA
327
328config SLUB_CPU_PARTIAL
329 default y
330 depends on SLUB && SMP && !SLUB_TINY
331 bool "SLUB per cpu partial cache"
332 help
333 Per cpu partial caches accelerate objects allocation and freeing
334 that is local to a processor at the price of more indeterminism
335 in the latency of the free. On overflow these caches will be cleared
336 which requires the taking of locks that may cause latency spikes.
337 Typically one would choose no for a realtime system.
338
339config RANDOM_KMALLOC_CACHES
340 default n
341 depends on SLUB && !SLUB_TINY
342 bool "Randomize slab caches for normal kmalloc"
343 help
344 A hardening feature that creates multiple copies of slab caches for
345 normal kmalloc allocation and makes kmalloc randomly pick one based
346 on code address, which makes the attackers more difficult to spray
347 vulnerable memory objects on the heap for the purpose of exploiting
348 memory vulnerabilities.
349
350 Currently the number of copies is set to 16, a reasonably large value
351 that effectively diverges the memory objects allocated for different
352 subsystems or modules into different caches, at the expense of a
353 limited degree of memory and CPU overhead that relates to hardware and
354 system workload.
355
356endmenu # SLAB allocator options
357
358config SHUFFLE_PAGE_ALLOCATOR
359 bool "Page allocator randomization"
360 default SLAB_FREELIST_RANDOM && ACPI_NUMA
361 help
362 Randomization of the page allocator improves the average
363 utilization of a direct-mapped memory-side-cache. See section
364 5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI
365 6.2a specification for an example of how a platform advertises
366 the presence of a memory-side-cache. There are also incidental
367 security benefits as it reduces the predictability of page
368 allocations to compliment SLAB_FREELIST_RANDOM, but the
369 default granularity of shuffling on the MAX_ORDER i.e, 10th
370 order of pages is selected based on cache utilization benefits
371 on x86.
372
373 While the randomization improves cache utilization it may
374 negatively impact workloads on platforms without a cache. For
375 this reason, by default, the randomization is enabled only
376 after runtime detection of a direct-mapped memory-side-cache.
377 Otherwise, the randomization may be force enabled with the
378 'page_alloc.shuffle' kernel command line parameter.
379
380 Say Y if unsure.
381
382config COMPAT_BRK
383 bool "Disable heap randomization"
384 default y
385 help
386 Randomizing heap placement makes heap exploits harder, but it
387 also breaks ancient binaries (including anything libc5 based).
388 This option changes the bootup default to heap randomization
389 disabled, and can be overridden at runtime by setting
390 /proc/sys/kernel/randomize_va_space to 2.
391
392 On non-ancient distros (post-2000 ones) N is usually a safe choice.
393
394config MMAP_ALLOW_UNINITIALIZED
395 bool "Allow mmapped anonymous memory to be uninitialized"
396 depends on EXPERT && !MMU
397 default n
398 help
399 Normally, and according to the Linux spec, anonymous memory obtained
400 from mmap() has its contents cleared before it is passed to
401 userspace. Enabling this config option allows you to request that
402 mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus
403 providing a huge performance boost. If this option is not enabled,
404 then the flag will be ignored.
405
406 This is taken advantage of by uClibc's malloc(), and also by
407 ELF-FDPIC binfmt's brk and stack allocator.
408
409 Because of the obvious security issues, this option should only be
410 enabled on embedded devices where you control what is run in
411 userspace. Since that isn't generally a problem on no-MMU systems,
412 it is normally safe to say Y here.
413
414 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
415
416config SELECT_MEMORY_MODEL
417 def_bool y
418 depends on ARCH_SELECT_MEMORY_MODEL
419
420choice
421 prompt "Memory model"
422 depends on SELECT_MEMORY_MODEL
423 default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT
424 default FLATMEM_MANUAL
425 help
426 This option allows you to change some of the ways that
427 Linux manages its memory internally. Most users will
428 only have one option here selected by the architecture
429 configuration. This is normal.
430
431config FLATMEM_MANUAL
432 bool "Flat Memory"
433 depends on !ARCH_SPARSEMEM_ENABLE || ARCH_FLATMEM_ENABLE
434 help
435 This option is best suited for non-NUMA systems with
436 flat address space. The FLATMEM is the most efficient
437 system in terms of performance and resource consumption
438 and it is the best option for smaller systems.
439
440 For systems that have holes in their physical address
441 spaces and for features like NUMA and memory hotplug,
442 choose "Sparse Memory".
443
444 If unsure, choose this option (Flat Memory) over any other.
445
446config SPARSEMEM_MANUAL
447 bool "Sparse Memory"
448 depends on ARCH_SPARSEMEM_ENABLE
449 help
450 This will be the only option for some systems, including
451 memory hot-plug systems. This is normal.
452
453 This option provides efficient support for systems with
454 holes is their physical address space and allows memory
455 hot-plug and hot-remove.
456
457 If unsure, choose "Flat Memory" over this option.
458
459endchoice
460
461config SPARSEMEM
462 def_bool y
463 depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL
464
465config FLATMEM
466 def_bool y
467 depends on !SPARSEMEM || FLATMEM_MANUAL
468
469#
470# SPARSEMEM_EXTREME (which is the default) does some bootmem
471# allocations when sparse_init() is called. If this cannot
472# be done on your architecture, select this option. However,
473# statically allocating the mem_section[] array can potentially
474# consume vast quantities of .bss, so be careful.
475#
476# This option will also potentially produce smaller runtime code
477# with gcc 3.4 and later.
478#
479config SPARSEMEM_STATIC
480 bool
481
482#
483# Architecture platforms which require a two level mem_section in SPARSEMEM
484# must select this option. This is usually for architecture platforms with
485# an extremely sparse physical address space.
486#
487config SPARSEMEM_EXTREME
488 def_bool y
489 depends on SPARSEMEM && !SPARSEMEM_STATIC
490
491config SPARSEMEM_VMEMMAP_ENABLE
492 bool
493
494config SPARSEMEM_VMEMMAP
495 bool "Sparse Memory virtual memmap"
496 depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE
497 default y
498 help
499 SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise
500 pfn_to_page and page_to_pfn operations. This is the most
501 efficient option when sufficient kernel resources are available.
502#
503# Select this config option from the architecture Kconfig, if it is preferred
504# to enable the feature of HugeTLB/dev_dax vmemmap optimization.
505#
506config ARCH_WANT_OPTIMIZE_DAX_VMEMMAP
507 bool
508
509config ARCH_WANT_OPTIMIZE_HUGETLB_VMEMMAP
510 bool
511
512config HAVE_MEMBLOCK_PHYS_MAP
513 bool
514
515config HAVE_FAST_GUP
516 depends on MMU
517 bool
518
519# Don't discard allocated memory used to track "memory" and "reserved" memblocks
520# after early boot, so it can still be used to test for validity of memory.
521# Also, memblocks are updated with memory hot(un)plug.
522config ARCH_KEEP_MEMBLOCK
523 bool
524
525# Keep arch NUMA mapping infrastructure post-init.
526config NUMA_KEEP_MEMINFO
527 bool
528
529config MEMORY_ISOLATION
530 bool
531
532# IORESOURCE_SYSTEM_RAM regions in the kernel resource tree that are marked
533# IORESOURCE_EXCLUSIVE cannot be mapped to user space, for example, via
534# /dev/mem.
535config EXCLUSIVE_SYSTEM_RAM
536 def_bool y
537 depends on !DEVMEM || STRICT_DEVMEM
538
539#
540# Only be set on architectures that have completely implemented memory hotplug
541# feature. If you are not sure, don't touch it.
542#
543config HAVE_BOOTMEM_INFO_NODE
544 def_bool n
545
546config ARCH_ENABLE_MEMORY_HOTPLUG
547 bool
548
549config ARCH_ENABLE_MEMORY_HOTREMOVE
550 bool
551
552# eventually, we can have this option just 'select SPARSEMEM'
553menuconfig MEMORY_HOTPLUG
554 bool "Memory hotplug"
555 select MEMORY_ISOLATION
556 depends on SPARSEMEM
557 depends on ARCH_ENABLE_MEMORY_HOTPLUG
558 depends on 64BIT
559 select NUMA_KEEP_MEMINFO if NUMA
560
561if MEMORY_HOTPLUG
562
563config MEMORY_HOTPLUG_DEFAULT_ONLINE
564 bool "Online the newly added memory blocks by default"
565 depends on MEMORY_HOTPLUG
566 help
567 This option sets the default policy setting for memory hotplug
568 onlining policy (/sys/devices/system/memory/auto_online_blocks) which
569 determines what happens to newly added memory regions. Policy setting
570 can always be changed at runtime.
571 See Documentation/admin-guide/mm/memory-hotplug.rst for more information.
572
573 Say Y here if you want all hot-plugged memory blocks to appear in
574 'online' state by default.
575 Say N here if you want the default policy to keep all hot-plugged
576 memory blocks in 'offline' state.
577
578config MEMORY_HOTREMOVE
579 bool "Allow for memory hot remove"
580 select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64)
581 depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE
582 depends on MIGRATION
583
584config MHP_MEMMAP_ON_MEMORY
585 def_bool y
586 depends on MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP
587 depends on ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
588
589endif # MEMORY_HOTPLUG
590
591config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
592 bool
593
594# Heavily threaded applications may benefit from splitting the mm-wide
595# page_table_lock, so that faults on different parts of the user address
596# space can be handled with less contention: split it at this NR_CPUS.
597# Default to 4 for wider testing, though 8 might be more appropriate.
598# ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
599# PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
600# SPARC32 allocates multiple pte tables within a single page, and therefore
601# a per-page lock leads to problems when multiple tables need to be locked
602# at the same time (e.g. copy_page_range()).
603# DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
604#
605config SPLIT_PTLOCK_CPUS
606 int
607 default "999999" if !MMU
608 default "999999" if ARM && !CPU_CACHE_VIPT
609 default "999999" if PARISC && !PA20
610 default "999999" if SPARC32
611 default "4"
612
613config ARCH_ENABLE_SPLIT_PMD_PTLOCK
614 bool
615
616#
617# support for memory balloon
618config MEMORY_BALLOON
619 bool
620
621#
622# support for memory balloon compaction
623config BALLOON_COMPACTION
624 bool "Allow for balloon memory compaction/migration"
625 def_bool y
626 depends on COMPACTION && MEMORY_BALLOON
627 help
628 Memory fragmentation introduced by ballooning might reduce
629 significantly the number of 2MB contiguous memory blocks that can be
630 used within a guest, thus imposing performance penalties associated
631 with the reduced number of transparent huge pages that could be used
632 by the guest workload. Allowing the compaction & migration for memory
633 pages enlisted as being part of memory balloon devices avoids the
634 scenario aforementioned and helps improving memory defragmentation.
635
636#
637# support for memory compaction
638config COMPACTION
639 bool "Allow for memory compaction"
640 def_bool y
641 select MIGRATION
642 depends on MMU
643 help
644 Compaction is the only memory management component to form
645 high order (larger physically contiguous) memory blocks
646 reliably. The page allocator relies on compaction heavily and
647 the lack of the feature can lead to unexpected OOM killer
648 invocations for high order memory requests. You shouldn't
649 disable this option unless there really is a strong reason for
650 it and then we would be really interested to hear about that at
651 linux-mm@kvack.org.
652
653config COMPACT_UNEVICTABLE_DEFAULT
654 int
655 depends on COMPACTION
656 default 0 if PREEMPT_RT
657 default 1
658
659#
660# support for free page reporting
661config PAGE_REPORTING
662 bool "Free page reporting"
663 def_bool n
664 help
665 Free page reporting allows for the incremental acquisition of
666 free pages from the buddy allocator for the purpose of reporting
667 those pages to another entity, such as a hypervisor, so that the
668 memory can be freed within the host for other uses.
669
670#
671# support for page migration
672#
673config MIGRATION
674 bool "Page migration"
675 def_bool y
676 depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU
677 help
678 Allows the migration of the physical location of pages of processes
679 while the virtual addresses are not changed. This is useful in
680 two situations. The first is on NUMA systems to put pages nearer
681 to the processors accessing. The second is when allocating huge
682 pages as migration can relocate pages to satisfy a huge page
683 allocation instead of reclaiming.
684
685config DEVICE_MIGRATION
686 def_bool MIGRATION && ZONE_DEVICE
687
688config ARCH_ENABLE_HUGEPAGE_MIGRATION
689 bool
690
691config ARCH_ENABLE_THP_MIGRATION
692 bool
693
694config HUGETLB_PAGE_SIZE_VARIABLE
695 def_bool n
696 help
697 Allows the pageblock_order value to be dynamic instead of just standard
698 HUGETLB_PAGE_ORDER when there are multiple HugeTLB page sizes available
699 on a platform.
700
701 Note that the pageblock_order cannot exceed MAX_ORDER and will be
702 clamped down to MAX_ORDER.
703
704config CONTIG_ALLOC
705 def_bool (MEMORY_ISOLATION && COMPACTION) || CMA
706
707config PHYS_ADDR_T_64BIT
708 def_bool 64BIT
709
710config BOUNCE
711 bool "Enable bounce buffers"
712 default y
713 depends on BLOCK && MMU && HIGHMEM
714 help
715 Enable bounce buffers for devices that cannot access the full range of
716 memory available to the CPU. Enabled by default when HIGHMEM is
717 selected, but you may say n to override this.
718
719config MMU_NOTIFIER
720 bool
721 select INTERVAL_TREE
722
723config KSM
724 bool "Enable KSM for page merging"
725 depends on MMU
726 select XXHASH
727 help
728 Enable Kernel Samepage Merging: KSM periodically scans those areas
729 of an application's address space that an app has advised may be
730 mergeable. When it finds pages of identical content, it replaces
731 the many instances by a single page with that content, so
732 saving memory until one or another app needs to modify the content.
733 Recommended for use with KVM, or with other duplicative applications.
734 See Documentation/mm/ksm.rst for more information: KSM is inactive
735 until a program has madvised that an area is MADV_MERGEABLE, and
736 root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set).
737
738config DEFAULT_MMAP_MIN_ADDR
739 int "Low address space to protect from user allocation"
740 depends on MMU
741 default 4096
742 help
743 This is the portion of low virtual memory which should be protected
744 from userspace allocation. Keeping a user from writing to low pages
745 can help reduce the impact of kernel NULL pointer bugs.
746
747 For most ia64, ppc64 and x86 users with lots of address space
748 a value of 65536 is reasonable and should cause no problems.
749 On arm and other archs it should not be higher than 32768.
750 Programs which use vm86 functionality or have some need to map
751 this low address space will need CAP_SYS_RAWIO or disable this
752 protection by setting the value to 0.
753
754 This value can be changed after boot using the
755 /proc/sys/vm/mmap_min_addr tunable.
756
757config ARCH_SUPPORTS_MEMORY_FAILURE
758 bool
759
760config MEMORY_FAILURE
761 depends on MMU
762 depends on ARCH_SUPPORTS_MEMORY_FAILURE
763 bool "Enable recovery from hardware memory errors"
764 select MEMORY_ISOLATION
765 select RAS
766 help
767 Enables code to recover from some memory failures on systems
768 with MCA recovery. This allows a system to continue running
769 even when some of its memory has uncorrected errors. This requires
770 special hardware support and typically ECC memory.
771
772config HWPOISON_INJECT
773 tristate "HWPoison pages injector"
774 depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS
775 select PROC_PAGE_MONITOR
776
777config NOMMU_INITIAL_TRIM_EXCESS
778 int "Turn on mmap() excess space trimming before booting"
779 depends on !MMU
780 default 1
781 help
782 The NOMMU mmap() frequently needs to allocate large contiguous chunks
783 of memory on which to store mappings, but it can only ask the system
784 allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently
785 more than it requires. To deal with this, mmap() is able to trim off
786 the excess and return it to the allocator.
787
788 If trimming is enabled, the excess is trimmed off and returned to the
789 system allocator, which can cause extra fragmentation, particularly
790 if there are a lot of transient processes.
791
792 If trimming is disabled, the excess is kept, but not used, which for
793 long-term mappings means that the space is wasted.
794
795 Trimming can be dynamically controlled through a sysctl option
796 (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of
797 excess pages there must be before trimming should occur, or zero if
798 no trimming is to occur.
799
800 This option specifies the initial value of this option. The default
801 of 1 says that all excess pages should be trimmed.
802
803 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
804
805config ARCH_WANT_GENERAL_HUGETLB
806 bool
807
808config ARCH_WANTS_THP_SWAP
809 def_bool n
810
811menuconfig TRANSPARENT_HUGEPAGE
812 bool "Transparent Hugepage Support"
813 depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT
814 select COMPACTION
815 select XARRAY_MULTI
816 help
817 Transparent Hugepages allows the kernel to use huge pages and
818 huge tlb transparently to the applications whenever possible.
819 This feature can improve computing performance to certain
820 applications by speeding up page faults during memory
821 allocation, by reducing the number of tlb misses and by speeding
822 up the pagetable walking.
823
824 If memory constrained on embedded, you may want to say N.
825
826if TRANSPARENT_HUGEPAGE
827
828choice
829 prompt "Transparent Hugepage Support sysfs defaults"
830 depends on TRANSPARENT_HUGEPAGE
831 default TRANSPARENT_HUGEPAGE_ALWAYS
832 help
833 Selects the sysfs defaults for Transparent Hugepage Support.
834
835 config TRANSPARENT_HUGEPAGE_ALWAYS
836 bool "always"
837 help
838 Enabling Transparent Hugepage always, can increase the
839 memory footprint of applications without a guaranteed
840 benefit but it will work automatically for all applications.
841
842 config TRANSPARENT_HUGEPAGE_MADVISE
843 bool "madvise"
844 help
845 Enabling Transparent Hugepage madvise, will only provide a
846 performance improvement benefit to the applications using
847 madvise(MADV_HUGEPAGE) but it won't risk to increase the
848 memory footprint of applications without a guaranteed
849 benefit.
850endchoice
851
852config THP_SWAP
853 def_bool y
854 depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP && 64BIT
855 help
856 Swap transparent huge pages in one piece, without splitting.
857 XXX: For now, swap cluster backing transparent huge page
858 will be split after swapout.
859
860 For selection by architectures with reasonable THP sizes.
861
862config READ_ONLY_THP_FOR_FS
863 bool "Read-only THP for filesystems (EXPERIMENTAL)"
864 depends on TRANSPARENT_HUGEPAGE && SHMEM
865
866 help
867 Allow khugepaged to put read-only file-backed pages in THP.
868
869 This is marked experimental because it is a new feature. Write
870 support of file THPs will be developed in the next few release
871 cycles.
872
873endif # TRANSPARENT_HUGEPAGE
874
875#
876# UP and nommu archs use km based percpu allocator
877#
878config NEED_PER_CPU_KM
879 depends on !SMP || !MMU
880 bool
881 default y
882
883config NEED_PER_CPU_EMBED_FIRST_CHUNK
884 bool
885
886config NEED_PER_CPU_PAGE_FIRST_CHUNK
887 bool
888
889config USE_PERCPU_NUMA_NODE_ID
890 bool
891
892config HAVE_SETUP_PER_CPU_AREA
893 bool
894
895config CMA
896 bool "Contiguous Memory Allocator"
897 depends on MMU
898 select MIGRATION
899 select MEMORY_ISOLATION
900 help
901 This enables the Contiguous Memory Allocator which allows other
902 subsystems to allocate big physically-contiguous blocks of memory.
903 CMA reserves a region of memory and allows only movable pages to
904 be allocated from it. This way, the kernel can use the memory for
905 pagecache and when a subsystem requests for contiguous area, the
906 allocated pages are migrated away to serve the contiguous request.
907
908 If unsure, say "n".
909
910config CMA_DEBUG
911 bool "CMA debug messages (DEVELOPMENT)"
912 depends on DEBUG_KERNEL && CMA
913 help
914 Turns on debug messages in CMA. This produces KERN_DEBUG
915 messages for every CMA call as well as various messages while
916 processing calls such as dma_alloc_from_contiguous().
917 This option does not affect warning and error messages.
918
919config CMA_DEBUGFS
920 bool "CMA debugfs interface"
921 depends on CMA && DEBUG_FS
922 help
923 Turns on the DebugFS interface for CMA.
924
925config CMA_SYSFS
926 bool "CMA information through sysfs interface"
927 depends on CMA && SYSFS
928 help
929 This option exposes some sysfs attributes to get information
930 from CMA.
931
932config CMA_AREAS
933 int "Maximum count of the CMA areas"
934 depends on CMA
935 default 19 if NUMA
936 default 7
937 help
938 CMA allows to create CMA areas for particular purpose, mainly,
939 used as device private area. This parameter sets the maximum
940 number of CMA area in the system.
941
942 If unsure, leave the default value "7" in UMA and "19" in NUMA.
943
944config MEM_SOFT_DIRTY
945 bool "Track memory changes"
946 depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS
947 select PROC_PAGE_MONITOR
948 help
949 This option enables memory changes tracking by introducing a
950 soft-dirty bit on pte-s. This bit it set when someone writes
951 into a page just as regular dirty bit, but unlike the latter
952 it can be cleared by hands.
953
954 See Documentation/admin-guide/mm/soft-dirty.rst for more details.
955
956config GENERIC_EARLY_IOREMAP
957 bool
958
959config STACK_MAX_DEFAULT_SIZE_MB
960 int "Default maximum user stack size for 32-bit processes (MB)"
961 default 100
962 range 8 2048
963 depends on STACK_GROWSUP && (!64BIT || COMPAT)
964 help
965 This is the maximum stack size in Megabytes in the VM layout of 32-bit
966 user processes when the stack grows upwards (currently only on parisc
967 arch) when the RLIMIT_STACK hard limit is unlimited.
968
969 A sane initial value is 100 MB.
970
971config DEFERRED_STRUCT_PAGE_INIT
972 bool "Defer initialisation of struct pages to kthreads"
973 depends on SPARSEMEM
974 depends on !NEED_PER_CPU_KM
975 depends on 64BIT
976 select PADATA
977 help
978 Ordinarily all struct pages are initialised during early boot in a
979 single thread. On very large machines this can take a considerable
980 amount of time. If this option is set, large machines will bring up
981 a subset of memmap at boot and then initialise the rest in parallel.
982 This has a potential performance impact on tasks running early in the
983 lifetime of the system until these kthreads finish the
984 initialisation.
985
986config PAGE_IDLE_FLAG
987 bool
988 select PAGE_EXTENSION if !64BIT
989 help
990 This adds PG_idle and PG_young flags to 'struct page'. PTE Accessed
991 bit writers can set the state of the bit in the flags so that PTE
992 Accessed bit readers may avoid disturbance.
993
994config IDLE_PAGE_TRACKING
995 bool "Enable idle page tracking"
996 depends on SYSFS && MMU
997 select PAGE_IDLE_FLAG
998 help
999 This feature allows to estimate the amount of user pages that have
1000 not been touched during a given period of time. This information can
1001 be useful to tune memory cgroup limits and/or for job placement
1002 within a compute cluster.
1003
1004 See Documentation/admin-guide/mm/idle_page_tracking.rst for
1005 more details.
1006
1007config ARCH_HAS_CACHE_LINE_SIZE
1008 bool
1009
1010config ARCH_HAS_CURRENT_STACK_POINTER
1011 bool
1012 help
1013 In support of HARDENED_USERCOPY performing stack variable lifetime
1014 checking, an architecture-agnostic way to find the stack pointer
1015 is needed. Once an architecture defines an unsigned long global
1016 register alias named "current_stack_pointer", this config can be
1017 selected.
1018
1019config ARCH_HAS_PTE_DEVMAP
1020 bool
1021
1022config ARCH_HAS_ZONE_DMA_SET
1023 bool
1024
1025config ZONE_DMA
1026 bool "Support DMA zone" if ARCH_HAS_ZONE_DMA_SET
1027 default y if ARM64 || X86
1028
1029config ZONE_DMA32
1030 bool "Support DMA32 zone" if ARCH_HAS_ZONE_DMA_SET
1031 depends on !X86_32
1032 default y if ARM64
1033
1034config ZONE_DEVICE
1035 bool "Device memory (pmem, HMM, etc...) hotplug support"
1036 depends on MEMORY_HOTPLUG
1037 depends on MEMORY_HOTREMOVE
1038 depends on SPARSEMEM_VMEMMAP
1039 depends on ARCH_HAS_PTE_DEVMAP
1040 select XARRAY_MULTI
1041
1042 help
1043 Device memory hotplug support allows for establishing pmem,
1044 or other device driver discovered memory regions, in the
1045 memmap. This allows pfn_to_page() lookups of otherwise
1046 "device-physical" addresses which is needed for using a DAX
1047 mapping in an O_DIRECT operation, among other things.
1048
1049 If FS_DAX is enabled, then say Y.
1050
1051#
1052# Helpers to mirror range of the CPU page tables of a process into device page
1053# tables.
1054#
1055config HMM_MIRROR
1056 bool
1057 depends on MMU
1058
1059config GET_FREE_REGION
1060 depends on SPARSEMEM
1061 bool
1062
1063config DEVICE_PRIVATE
1064 bool "Unaddressable device memory (GPU memory, ...)"
1065 depends on ZONE_DEVICE
1066 select GET_FREE_REGION
1067
1068 help
1069 Allows creation of struct pages to represent unaddressable device
1070 memory; i.e., memory that is only accessible from the device (or
1071 group of devices). You likely also want to select HMM_MIRROR.
1072
1073config VMAP_PFN
1074 bool
1075
1076config ARCH_USES_HIGH_VMA_FLAGS
1077 bool
1078config ARCH_HAS_PKEYS
1079 bool
1080
1081config ARCH_USES_PG_ARCH_X
1082 bool
1083 help
1084 Enable the definition of PG_arch_x page flags with x > 1. Only
1085 suitable for 64-bit architectures with CONFIG_FLATMEM or
1086 CONFIG_SPARSEMEM_VMEMMAP enabled, otherwise there may not be
1087 enough room for additional bits in page->flags.
1088
1089config VM_EVENT_COUNTERS
1090 default y
1091 bool "Enable VM event counters for /proc/vmstat" if EXPERT
1092 help
1093 VM event counters are needed for event counts to be shown.
1094 This option allows the disabling of the VM event counters
1095 on EXPERT systems. /proc/vmstat will only show page counts
1096 if VM event counters are disabled.
1097
1098config PERCPU_STATS
1099 bool "Collect percpu memory statistics"
1100 help
1101 This feature collects and exposes statistics via debugfs. The
1102 information includes global and per chunk statistics, which can
1103 be used to help understand percpu memory usage.
1104
1105config GUP_TEST
1106 bool "Enable infrastructure for get_user_pages()-related unit tests"
1107 depends on DEBUG_FS
1108 help
1109 Provides /sys/kernel/debug/gup_test, which in turn provides a way
1110 to make ioctl calls that can launch kernel-based unit tests for
1111 the get_user_pages*() and pin_user_pages*() family of API calls.
1112
1113 These tests include benchmark testing of the _fast variants of
1114 get_user_pages*() and pin_user_pages*(), as well as smoke tests of
1115 the non-_fast variants.
1116
1117 There is also a sub-test that allows running dump_page() on any
1118 of up to eight pages (selected by command line args) within the
1119 range of user-space addresses. These pages are either pinned via
1120 pin_user_pages*(), or pinned via get_user_pages*(), as specified
1121 by other command line arguments.
1122
1123 See tools/testing/selftests/mm/gup_test.c
1124
1125comment "GUP_TEST needs to have DEBUG_FS enabled"
1126 depends on !GUP_TEST && !DEBUG_FS
1127
1128config GUP_GET_PXX_LOW_HIGH
1129 bool
1130
1131config DMAPOOL_TEST
1132 tristate "Enable a module to run time tests on dma_pool"
1133 depends on HAS_DMA
1134 help
1135 Provides a test module that will allocate and free many blocks of
1136 various sizes and report how long it takes. This is intended to
1137 provide a consistent way to measure how changes to the
1138 dma_pool_alloc/free routines affect performance.
1139
1140config ARCH_HAS_PTE_SPECIAL
1141 bool
1142
1143#
1144# Some architectures require a special hugepage directory format that is
1145# required to support multiple hugepage sizes. For example a4fe3ce76
1146# "powerpc/mm: Allow more flexible layouts for hugepage pagetables"
1147# introduced it on powerpc. This allows for a more flexible hugepage
1148# pagetable layouts.
1149#
1150config ARCH_HAS_HUGEPD
1151 bool
1152
1153config MAPPING_DIRTY_HELPERS
1154 bool
1155
1156config KMAP_LOCAL
1157 bool
1158
1159config KMAP_LOCAL_NON_LINEAR_PTE_ARRAY
1160 bool
1161
1162# struct io_mapping based helper. Selected by drivers that need them
1163config IO_MAPPING
1164 bool
1165
1166config MEMFD_CREATE
1167 bool "Enable memfd_create() system call" if EXPERT
1168
1169config SECRETMEM
1170 default y
1171 bool "Enable memfd_secret() system call" if EXPERT
1172 depends on ARCH_HAS_SET_DIRECT_MAP
1173 help
1174 Enable the memfd_secret() system call with the ability to create
1175 memory areas visible only in the context of the owning process and
1176 not mapped to other processes and other kernel page tables.
1177
1178config ANON_VMA_NAME
1179 bool "Anonymous VMA name support"
1180 depends on PROC_FS && ADVISE_SYSCALLS && MMU
1181
1182 help
1183 Allow naming anonymous virtual memory areas.
1184
1185 This feature allows assigning names to virtual memory areas. Assigned
1186 names can be later retrieved from /proc/pid/maps and /proc/pid/smaps
1187 and help identifying individual anonymous memory areas.
1188 Assigning a name to anonymous virtual memory area might prevent that
1189 area from being merged with adjacent virtual memory areas due to the
1190 difference in their name.
1191
1192config USERFAULTFD
1193 bool "Enable userfaultfd() system call"
1194 depends on MMU
1195 help
1196 Enable the userfaultfd() system call that allows to intercept and
1197 handle page faults in userland.
1198
1199config HAVE_ARCH_USERFAULTFD_WP
1200 bool
1201 help
1202 Arch has userfaultfd write protection support
1203
1204config HAVE_ARCH_USERFAULTFD_MINOR
1205 bool
1206 help
1207 Arch has userfaultfd minor fault support
1208
1209config PTE_MARKER_UFFD_WP
1210 bool "Userfaultfd write protection support for shmem/hugetlbfs"
1211 default y
1212 depends on HAVE_ARCH_USERFAULTFD_WP
1213
1214 help
1215 Allows to create marker PTEs for userfaultfd write protection
1216 purposes. It is required to enable userfaultfd write protection on
1217 file-backed memory types like shmem and hugetlbfs.
1218
1219# multi-gen LRU {
1220config LRU_GEN
1221 bool "Multi-Gen LRU"
1222 depends on MMU
1223 # make sure folio->flags has enough spare bits
1224 depends on 64BIT || !SPARSEMEM || SPARSEMEM_VMEMMAP
1225 help
1226 A high performance LRU implementation to overcommit memory. See
1227 Documentation/admin-guide/mm/multigen_lru.rst for details.
1228
1229config LRU_GEN_ENABLED
1230 bool "Enable by default"
1231 depends on LRU_GEN
1232 help
1233 This option enables the multi-gen LRU by default.
1234
1235config LRU_GEN_STATS
1236 bool "Full stats for debugging"
1237 depends on LRU_GEN
1238 help
1239 Do not enable this option unless you plan to look at historical stats
1240 from evicted generations for debugging purpose.
1241
1242 This option has a per-memcg and per-node memory overhead.
1243# }
1244
1245config ARCH_SUPPORTS_PER_VMA_LOCK
1246 def_bool n
1247
1248config PER_VMA_LOCK
1249 def_bool y
1250 depends on ARCH_SUPPORTS_PER_VMA_LOCK && MMU && SMP
1251 help
1252 Allow per-vma locking during page fault handling.
1253
1254 This feature allows locking each virtual memory area separately when
1255 handling page faults instead of taking mmap_lock.
1256
1257config LOCK_MM_AND_FIND_VMA
1258 bool
1259 depends on !STACK_GROWSUP
1260
1261source "mm/damon/Kconfig"
1262
1263endmenu