xref: /linux/mm/Kconfig (revision 078f80f909ba9fa3060e89dc634ff4b1defc43d3)
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
12menuconfig SWAP
13	bool "Support for paging of anonymous memory (swap)"
14	depends on MMU && BLOCK && !ARCH_NO_SWAP
15	default y
16	help
17	  This option allows you to choose whether you want to have support
18	  for so called swap devices or swap files in your kernel that are
19	  used to provide more virtual memory than the actual RAM present
20	  in your computer.  If unsure say Y.
21
22config ZSWAP
23	bool "Compressed cache for swap pages"
24	depends on SWAP
25	select CRYPTO
26	select ZSMALLOC
27	help
28	  A lightweight compressed cache for swap pages.  It takes
29	  pages that are in the process of being swapped out and attempts to
30	  compress them into a dynamically allocated RAM-based memory pool.
31	  This can result in a significant I/O reduction on swap device and,
32	  in the case where decompressing from RAM is faster than swap device
33	  reads, can also improve workload performance.
34
35config ZSWAP_DEFAULT_ON
36	bool "Enable the compressed cache for swap pages by default"
37	depends on ZSWAP
38	help
39	  If selected, the compressed cache for swap pages will be enabled
40	  at boot, otherwise it will be disabled.
41
42	  The selection made here can be overridden by using the kernel
43	  command line 'zswap.enabled=' option.
44
45config ZSWAP_SHRINKER_DEFAULT_ON
46	bool "Shrink the zswap pool on memory pressure"
47	depends on ZSWAP
48	default n
49	help
50	  If selected, the zswap shrinker will be enabled, and the pages
51	  stored in the zswap pool will become available for reclaim (i.e
52	  written back to the backing swap device) on memory pressure.
53
54	  This means that zswap writeback could happen even if the pool is
55	  not yet full, or the cgroup zswap limit has not been reached,
56	  reducing the chance that cold pages will reside in the zswap pool
57	  and consume memory indefinitely.
58
59choice
60	prompt "Default compressor"
61	depends on ZSWAP
62	default ZSWAP_COMPRESSOR_DEFAULT_LZO
63	help
64	  Selects the default compression algorithm for the compressed cache
65	  for swap pages.
66
67	  For an overview what kind of performance can be expected from
68	  a particular compression algorithm please refer to the benchmarks
69	  available at the following LWN page:
70	  https://lwn.net/Articles/751795/
71
72	  If in doubt, select 'LZO'.
73
74	  The selection made here can be overridden by using the kernel
75	  command line 'zswap.compressor=' option.
76
77config ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
78	bool "Deflate"
79	select CRYPTO_DEFLATE
80	help
81	  Use the Deflate algorithm as the default compression algorithm.
82
83config ZSWAP_COMPRESSOR_DEFAULT_LZO
84	bool "LZO"
85	select CRYPTO_LZO
86	help
87	  Use the LZO algorithm as the default compression algorithm.
88
89config ZSWAP_COMPRESSOR_DEFAULT_842
90	bool "842"
91	select CRYPTO_842
92	help
93	  Use the 842 algorithm as the default compression algorithm.
94
95config ZSWAP_COMPRESSOR_DEFAULT_LZ4
96	bool "LZ4"
97	select CRYPTO_LZ4
98	help
99	  Use the LZ4 algorithm as the default compression algorithm.
100
101config ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
102	bool "LZ4HC"
103	select CRYPTO_LZ4HC
104	help
105	  Use the LZ4HC algorithm as the default compression algorithm.
106
107config ZSWAP_COMPRESSOR_DEFAULT_ZSTD
108	bool "zstd"
109	select CRYPTO_ZSTD
110	help
111	  Use the zstd algorithm as the default compression algorithm.
112endchoice
113
114config ZSWAP_COMPRESSOR_DEFAULT
115       string
116       depends on ZSWAP
117       default "deflate" if ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
118       default "lzo" if ZSWAP_COMPRESSOR_DEFAULT_LZO
119       default "842" if ZSWAP_COMPRESSOR_DEFAULT_842
120       default "lz4" if ZSWAP_COMPRESSOR_DEFAULT_LZ4
121       default "lz4hc" if ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
122       default "zstd" if ZSWAP_COMPRESSOR_DEFAULT_ZSTD
123       default ""
124
125config ZSMALLOC
126	tristate
127
128if ZSMALLOC
129
130menu "Zsmalloc allocator options"
131	depends on ZSMALLOC
132
133comment "Zsmalloc is a common backend allocator for zswap & zram"
134
135config ZSMALLOC_STAT
136	bool "Export zsmalloc statistics"
137	select DEBUG_FS
138	help
139	  This option enables code in the zsmalloc to collect various
140	  statistics about what's happening in zsmalloc and exports that
141	  information to userspace via debugfs.
142	  If unsure, say N.
143
144config ZSMALLOC_CHAIN_SIZE
145	int "Maximum number of physical pages per-zspage"
146	default 8
147	range 4 16
148	help
149	  This option sets the upper limit on the number of physical pages
150	  that a zmalloc page (zspage) can consist of. The optimal zspage
151	  chain size is calculated for each size class during the
152	  initialization of the pool.
153
154	  Changing this option can alter the characteristics of size classes,
155	  such as the number of pages per zspage and the number of objects
156	  per zspage. This can also result in different configurations of
157	  the pool, as zsmalloc merges size classes with similar
158	  characteristics.
159
160	  For more information, see zsmalloc documentation.
161
162endmenu
163
164endif
165
166menu "Slab allocator options"
167
168config SLUB
169	def_bool y
170	select IRQ_WORK
171
172config KVFREE_RCU_BATCHED
173	def_bool y
174	depends on !SLUB_TINY && !TINY_RCU
175
176config SLUB_TINY
177	bool "Configure for minimal memory footprint"
178	depends on EXPERT && !COMPILE_TEST
179	select SLAB_MERGE_DEFAULT
180	help
181	   Configures the slab allocator in a way to achieve minimal memory
182	   footprint, sacrificing scalability, debugging and other features.
183	   This is intended only for the smallest system that had used the
184	   SLOB allocator and is not recommended for systems with more than
185	   16MB RAM.
186
187	   If unsure, say N.
188
189config SLAB_MERGE_DEFAULT
190	bool "Allow slab caches to be merged"
191	default y
192	help
193	  For reduced kernel memory fragmentation, slab caches can be
194	  merged when they share the same size and other characteristics.
195	  This carries a risk of kernel heap overflows being able to
196	  overwrite objects from merged caches (and more easily control
197	  cache layout), which makes such heap attacks easier to exploit
198	  by attackers. By keeping caches unmerged, these kinds of exploits
199	  can usually only damage objects in the same cache. To disable
200	  merging at runtime, "slab_nomerge" can be passed on the kernel
201	  command line.
202
203config SLAB_FREELIST_RANDOM
204	bool "Randomize slab freelist"
205	depends on !SLUB_TINY
206	help
207	  Randomizes the freelist order used on creating new pages. This
208	  security feature reduces the predictability of the kernel slab
209	  allocator against heap overflows.
210
211config SLAB_FREELIST_HARDENED
212	bool "Harden slab freelist metadata"
213	depends on !SLUB_TINY
214	help
215	  Many kernel heap attacks try to target slab cache metadata and
216	  other infrastructure. This options makes minor performance
217	  sacrifices to harden the kernel slab allocator against common
218	  freelist exploit methods.
219
220config SLAB_BUCKETS
221	bool "Support allocation from separate kmalloc buckets"
222	depends on !SLUB_TINY
223	default SLAB_FREELIST_HARDENED
224	help
225	  Kernel heap attacks frequently depend on being able to create
226	  specifically-sized allocations with user-controlled contents
227	  that will be allocated into the same kmalloc bucket as a
228	  target object. To avoid sharing these allocation buckets,
229	  provide an explicitly separated set of buckets to be used for
230	  user-controlled allocations. This may very slightly increase
231	  memory fragmentation, though in practice it's only a handful
232	  of extra pages since the bulk of user-controlled allocations
233	  are relatively long-lived.
234
235	  If unsure, say Y.
236
237config SLUB_STATS
238	default n
239	bool "Enable performance statistics"
240	depends on SYSFS && !SLUB_TINY
241	help
242	  The statistics are useful to debug slab allocation behavior in
243	  order find ways to optimize the allocator. This should never be
244	  enabled for production use since keeping statistics slows down
245	  the allocator by a few percentage points. The slabinfo command
246	  supports the determination of the most active slabs to figure
247	  out which slabs are relevant to a particular load.
248	  Try running: slabinfo -DA
249
250config RANDOM_KMALLOC_CACHES
251	default n
252	depends on !SLUB_TINY
253	bool "Randomize slab caches for normal kmalloc"
254	help
255	  A hardening feature that creates multiple copies of slab caches for
256	  normal kmalloc allocation and makes kmalloc randomly pick one based
257	  on code address, which makes the attackers more difficult to spray
258	  vulnerable memory objects on the heap for the purpose of exploiting
259	  memory vulnerabilities.
260
261	  Currently the number of copies is set to 16, a reasonably large value
262	  that effectively diverges the memory objects allocated for different
263	  subsystems or modules into different caches, at the expense of a
264	  limited degree of memory and CPU overhead that relates to hardware and
265	  system workload.
266
267endmenu # Slab allocator options
268
269config SHUFFLE_PAGE_ALLOCATOR
270	bool "Page allocator randomization"
271	default SLAB_FREELIST_RANDOM && ACPI_NUMA
272	help
273	  Randomization of the page allocator improves the average
274	  utilization of a direct-mapped memory-side-cache. See section
275	  5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI
276	  6.2a specification for an example of how a platform advertises
277	  the presence of a memory-side-cache. There are also incidental
278	  security benefits as it reduces the predictability of page
279	  allocations to compliment SLAB_FREELIST_RANDOM, but the
280	  default granularity of shuffling on the MAX_PAGE_ORDER i.e, 10th
281	  order of pages is selected based on cache utilization benefits
282	  on x86.
283
284	  While the randomization improves cache utilization it may
285	  negatively impact workloads on platforms without a cache. For
286	  this reason, by default, the randomization is not enabled even
287	  if SHUFFLE_PAGE_ALLOCATOR=y. The randomization may be force enabled
288	  with the 'page_alloc.shuffle' kernel command line parameter.
289
290	  Say Y if unsure.
291
292config COMPAT_BRK
293	bool "Disable heap randomization"
294	default y
295	help
296	  Randomizing heap placement makes heap exploits harder, but it
297	  also breaks ancient binaries (including anything libc5 based).
298	  This option changes the bootup default to heap randomization
299	  disabled, and can be overridden at runtime by setting
300	  /proc/sys/kernel/randomize_va_space to 2.
301
302	  On non-ancient distros (post-2000 ones) N is usually a safe choice.
303
304config MMAP_ALLOW_UNINITIALIZED
305	bool "Allow mmapped anonymous memory to be uninitialized"
306	depends on EXPERT && !MMU
307	default n
308	help
309	  Normally, and according to the Linux spec, anonymous memory obtained
310	  from mmap() has its contents cleared before it is passed to
311	  userspace.  Enabling this config option allows you to request that
312	  mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus
313	  providing a huge performance boost.  If this option is not enabled,
314	  then the flag will be ignored.
315
316	  This is taken advantage of by uClibc's malloc(), and also by
317	  ELF-FDPIC binfmt's brk and stack allocator.
318
319	  Because of the obvious security issues, this option should only be
320	  enabled on embedded devices where you control what is run in
321	  userspace.  Since that isn't generally a problem on no-MMU systems,
322	  it is normally safe to say Y here.
323
324	  See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
325
326config SELECT_MEMORY_MODEL
327	def_bool y
328	depends on ARCH_SELECT_MEMORY_MODEL
329
330choice
331	prompt "Memory model"
332	depends on SELECT_MEMORY_MODEL
333	default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT
334	default FLATMEM_MANUAL
335	help
336	  This option allows you to change some of the ways that
337	  Linux manages its memory internally. Most users will
338	  only have one option here selected by the architecture
339	  configuration. This is normal.
340
341config FLATMEM_MANUAL
342	bool "Flat Memory"
343	depends on !ARCH_SPARSEMEM_ENABLE || ARCH_FLATMEM_ENABLE
344	help
345	  This option is best suited for non-NUMA systems with
346	  flat address space. The FLATMEM is the most efficient
347	  system in terms of performance and resource consumption
348	  and it is the best option for smaller systems.
349
350	  For systems that have holes in their physical address
351	  spaces and for features like NUMA and memory hotplug,
352	  choose "Sparse Memory".
353
354	  If unsure, choose this option (Flat Memory) over any other.
355
356config SPARSEMEM_MANUAL
357	bool "Sparse Memory"
358	depends on ARCH_SPARSEMEM_ENABLE
359	help
360	  This will be the only option for some systems, including
361	  memory hot-plug systems.  This is normal.
362
363	  This option provides efficient support for systems with
364	  holes is their physical address space and allows memory
365	  hot-plug and hot-remove.
366
367	  If unsure, choose "Flat Memory" over this option.
368
369endchoice
370
371config SPARSEMEM
372	def_bool y
373	depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL
374
375config FLATMEM
376	def_bool y
377	depends on !SPARSEMEM || FLATMEM_MANUAL
378
379#
380# SPARSEMEM_EXTREME (which is the default) does some bootmem
381# allocations when sparse_init() is called.  If this cannot
382# be done on your architecture, select this option.  However,
383# statically allocating the mem_section[] array can potentially
384# consume vast quantities of .bss, so be careful.
385#
386# This option will also potentially produce smaller runtime code
387# with gcc 3.4 and later.
388#
389config SPARSEMEM_STATIC
390	bool
391
392#
393# Architecture platforms which require a two level mem_section in SPARSEMEM
394# must select this option. This is usually for architecture platforms with
395# an extremely sparse physical address space.
396#
397config SPARSEMEM_EXTREME
398	def_bool y
399	depends on SPARSEMEM && !SPARSEMEM_STATIC
400
401config SPARSEMEM_VMEMMAP_ENABLE
402	bool
403
404config SPARSEMEM_VMEMMAP
405	def_bool y
406	depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE
407	help
408	  SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise
409	  pfn_to_page and page_to_pfn operations.  This is the most
410	  efficient option when sufficient kernel resources are available.
411
412config SPARSEMEM_VMEMMAP_PREINIT
413	bool
414#
415# Select this config option from the architecture Kconfig, if it is preferred
416# to enable the feature of HugeTLB/dev_dax vmemmap optimization.
417#
418config ARCH_WANT_OPTIMIZE_DAX_VMEMMAP
419	bool
420
421config ARCH_WANT_OPTIMIZE_HUGETLB_VMEMMAP
422	bool
423
424config ARCH_WANT_HUGETLB_VMEMMAP_PREINIT
425	bool
426
427config HAVE_MEMBLOCK_PHYS_MAP
428	bool
429
430config HAVE_GUP_FAST
431	depends on MMU
432	bool
433
434# Enable memblock support for scratch memory which is needed for kexec handover
435config MEMBLOCK_KHO_SCRATCH
436	bool
437
438# Don't discard allocated memory used to track "memory" and "reserved" memblocks
439# after early boot, so it can still be used to test for validity of memory.
440# Also, memblocks are updated with memory hot(un)plug.
441config ARCH_KEEP_MEMBLOCK
442	bool
443
444# Keep arch NUMA mapping infrastructure post-init.
445config NUMA_KEEP_MEMINFO
446	bool
447
448config MEMORY_ISOLATION
449	bool
450
451# IORESOURCE_SYSTEM_RAM regions in the kernel resource tree that are marked
452# IORESOURCE_EXCLUSIVE cannot be mapped to user space, for example, via
453# /dev/mem.
454config EXCLUSIVE_SYSTEM_RAM
455	def_bool y
456	depends on !DEVMEM || STRICT_DEVMEM
457
458#
459# Only be set on architectures that have completely implemented memory hotplug
460# feature. If you are not sure, don't touch it.
461#
462config HAVE_BOOTMEM_INFO_NODE
463	def_bool n
464
465config ARCH_ENABLE_MEMORY_HOTPLUG
466	bool
467
468# eventually, we can have this option just 'select SPARSEMEM'
469menuconfig MEMORY_HOTPLUG
470	bool "Memory hotplug"
471	select MEMORY_ISOLATION
472	depends on SPARSEMEM_VMEMMAP
473	depends on ARCH_ENABLE_MEMORY_HOTPLUG
474	depends on 64BIT
475	select NUMA_KEEP_MEMINFO if NUMA
476
477if MEMORY_HOTPLUG
478
479choice
480	prompt "Memory Hotplug Default Online Type"
481	default MHP_DEFAULT_ONLINE_TYPE_OFFLINE
482	help
483	  Default memory type for hotplugged memory.
484
485	  This option sets the default policy setting for memory hotplug
486	  onlining policy (/sys/devices/system/memory/auto_online_blocks) which
487	  determines what happens to newly added memory regions. Policy setting
488	  can always be changed at runtime.
489
490	  The default is 'offline'.
491
492	  Select offline to defer onlining to drivers and user policy.
493	  Select auto to let the kernel choose what zones to utilize.
494	  Select online_kernel to generally allow kernel usage of this memory.
495	  Select online_movable to generally disallow kernel usage of this memory.
496
497	  Example kernel usage would be page structs and page tables.
498
499	  See Documentation/admin-guide/mm/memory-hotplug.rst for more information.
500
501config MHP_DEFAULT_ONLINE_TYPE_OFFLINE
502	bool "offline"
503	help
504	  Hotplugged memory will not be onlined by default.
505	  Choose this for systems with drivers and user policy that
506	  handle onlining of hotplug memory policy.
507
508config MHP_DEFAULT_ONLINE_TYPE_ONLINE_AUTO
509	bool "auto"
510	help
511	  Select this if you want the kernel to automatically online
512	  hotplugged memory into the zone it thinks is reasonable.
513	  This memory may be utilized for kernel data.
514
515config MHP_DEFAULT_ONLINE_TYPE_ONLINE_KERNEL
516	bool "kernel"
517	help
518	  Select this if you want the kernel to automatically online
519	  hotplugged memory into a zone capable of being used for kernel
520	  data. This typically means ZONE_NORMAL.
521
522config MHP_DEFAULT_ONLINE_TYPE_ONLINE_MOVABLE
523	bool "movable"
524	help
525	  Select this if you want the kernel to automatically online
526	  hotplug memory into ZONE_MOVABLE. This memory will generally
527	  not be utilized for kernel data.
528
529	  This should only be used when the admin knows sufficient
530	  ZONE_NORMAL memory is available to describe hotplug memory,
531	  otherwise hotplug memory may fail to online. For example,
532	  sufficient kernel-capable memory (ZONE_NORMAL) must be
533	  available to allocate page structs to describe ZONE_MOVABLE.
534
535endchoice
536
537config MEMORY_HOTREMOVE
538	bool "Allow for memory hot remove"
539	select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64)
540	depends on MEMORY_HOTPLUG
541	select MIGRATION
542
543config MHP_MEMMAP_ON_MEMORY
544	def_bool y
545	depends on MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP
546	depends on ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
547
548endif # MEMORY_HOTPLUG
549
550config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
551       bool
552
553# Heavily threaded applications may benefit from splitting the mm-wide
554# page_table_lock, so that faults on different parts of the user address
555# space can be handled with less contention: split it at this NR_CPUS.
556# Default to 4 for wider testing, though 8 might be more appropriate.
557# ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
558# PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
559# SPARC32 allocates multiple pte tables within a single page, and therefore
560# a per-page lock leads to problems when multiple tables need to be locked
561# at the same time (e.g. copy_page_range()).
562# DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
563#
564config SPLIT_PTE_PTLOCKS
565	def_bool y
566	depends on MMU
567	depends on SMP
568	depends on NR_CPUS >= 4
569	depends on !ARM || CPU_CACHE_VIPT
570	depends on !PARISC || PA20
571	depends on !SPARC32
572
573config ARCH_ENABLE_SPLIT_PMD_PTLOCK
574	bool
575
576config SPLIT_PMD_PTLOCKS
577	def_bool y
578	depends on SPLIT_PTE_PTLOCKS && ARCH_ENABLE_SPLIT_PMD_PTLOCK
579
580#
581# support for memory balloon
582config BALLOON
583	bool
584
585#
586# support for memory balloon page migration
587config BALLOON_MIGRATION
588	bool "Allow for balloon memory migration"
589	default y
590	depends on MIGRATION && BALLOON
591	help
592	  Allow for migration of pages inflated in a memory balloon such that
593	  they can be allocated from memory areas only available for movable
594	  allocations (e.g., ZONE_MOVABLE, CMA) and such that they can be
595	  migrated for memory defragmentation purposes by memory compaction.
596
597#
598# support for memory compaction
599config COMPACTION
600	bool "Allow for memory compaction"
601	default y
602	select MIGRATION
603	depends on MMU
604	help
605	  Compaction is the only memory management component to form
606	  high order (larger physically contiguous) memory blocks
607	  reliably. The page allocator relies on compaction heavily and
608	  the lack of the feature can lead to unexpected OOM killer
609	  invocations for high order memory requests. You shouldn't
610	  disable this option unless there really is a strong reason for
611	  it and then we would be really interested to hear about that at
612	  linux-mm@kvack.org.
613
614config COMPACT_UNEVICTABLE_DEFAULT
615	int
616	depends on COMPACTION
617	default 0 if PREEMPT_RT
618	default 1
619
620#
621# support for free page reporting
622config PAGE_REPORTING
623	bool "Free page reporting"
624	help
625	  Free page reporting allows for the incremental acquisition of
626	  free pages from the buddy allocator for the purpose of reporting
627	  those pages to another entity, such as a hypervisor, so that the
628	  memory can be freed within the host for other uses.
629
630#
631# support for page migration
632#
633config MIGRATION
634	bool "Page migration"
635	default y
636	depends on (NUMA || MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU
637	help
638	  Allows the migration of the physical location of pages of processes
639	  while the virtual addresses are not changed. This is useful in
640	  two situations. The first is on NUMA systems to put pages nearer
641	  to the processors accessing. The second is when allocating huge
642	  pages as migration can relocate pages to satisfy a huge page
643	  allocation instead of reclaiming.
644
645config DEVICE_MIGRATION
646	def_bool MIGRATION && ZONE_DEVICE
647
648config ARCH_ENABLE_HUGEPAGE_MIGRATION
649	bool
650
651config ARCH_ENABLE_THP_MIGRATION
652	bool
653
654config HUGETLB_PAGE_SIZE_VARIABLE
655	def_bool n
656	help
657	  Allows the pageblock_order value to be dynamic instead of just standard
658	  HUGETLB_PAGE_ORDER when there are multiple HugeTLB page sizes available
659	  on a platform.
660
661	  Note that the pageblock_order cannot exceed MAX_PAGE_ORDER and will be
662	  clamped down to MAX_PAGE_ORDER.
663
664config CONTIG_ALLOC
665	def_bool (MEMORY_ISOLATION && COMPACTION) || CMA
666
667config PCP_BATCH_SCALE_MAX
668	int "Maximum scale factor of PCP (Per-CPU pageset) batch allocate/free"
669	default 5
670	range 0 6
671	help
672	  In page allocator, PCP (Per-CPU pageset) is refilled and drained in
673	  batches.  The batch number is scaled automatically to improve page
674	  allocation/free throughput.  But too large scale factor may hurt
675	  latency.  This option sets the upper limit of scale factor to limit
676	  the maximum latency.
677
678config PHYS_ADDR_T_64BIT
679	def_bool 64BIT
680
681config MMU_NOTIFIER
682	bool
683	select INTERVAL_TREE
684
685config KSM
686	bool "Enable KSM for page merging"
687	depends on MMU
688	select XXHASH
689	help
690	  Enable Kernel Samepage Merging: KSM periodically scans those areas
691	  of an application's address space that an app has advised may be
692	  mergeable.  When it finds pages of identical content, it replaces
693	  the many instances by a single page with that content, so
694	  saving memory until one or another app needs to modify the content.
695	  Recommended for use with KVM, or with other duplicative applications.
696	  See Documentation/mm/ksm.rst for more information: KSM is inactive
697	  until a program has madvised that an area is MADV_MERGEABLE, and
698	  root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set).
699
700config DEFAULT_MMAP_MIN_ADDR
701	int "Low address space to protect from user allocation"
702	depends on MMU
703	default 4096
704	help
705	  This is the portion of low virtual memory which should be protected
706	  from userspace allocation.  Keeping a user from writing to low pages
707	  can help reduce the impact of kernel NULL pointer bugs.
708
709	  For most arm64, ppc64 and x86 users with lots of address space
710	  a value of 65536 is reasonable and should cause no problems.
711	  On arm and other archs it should not be higher than 32768.
712	  Programs which use vm86 functionality or have some need to map
713	  this low address space will need CAP_SYS_RAWIO or disable this
714	  protection by setting the value to 0.
715
716	  This value can be changed after boot using the
717	  /proc/sys/vm/mmap_min_addr tunable.
718
719config ARCH_SUPPORTS_MEMORY_FAILURE
720	bool
721
722config MEMORY_FAILURE
723	depends on MMU
724	depends on ARCH_SUPPORTS_MEMORY_FAILURE
725	bool "Enable recovery from hardware memory errors"
726	select INTERVAL_TREE
727	help
728	  Enables code to recover from some memory failures on systems
729	  with MCA recovery. This allows a system to continue running
730	  even when some of its memory has uncorrected errors. This requires
731	  special hardware support and typically ECC memory.
732
733config HWPOISON_INJECT
734	tristate "HWPoison pages injector"
735	depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS
736	select PROC_PAGE_MONITOR
737
738config NOMMU_INITIAL_TRIM_EXCESS
739	int "Turn on mmap() excess space trimming before booting"
740	depends on !MMU
741	default 1
742	help
743	  The NOMMU mmap() frequently needs to allocate large contiguous chunks
744	  of memory on which to store mappings, but it can only ask the system
745	  allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently
746	  more than it requires.  To deal with this, mmap() is able to trim off
747	  the excess and return it to the allocator.
748
749	  If trimming is enabled, the excess is trimmed off and returned to the
750	  system allocator, which can cause extra fragmentation, particularly
751	  if there are a lot of transient processes.
752
753	  If trimming is disabled, the excess is kept, but not used, which for
754	  long-term mappings means that the space is wasted.
755
756	  Trimming can be dynamically controlled through a sysctl option
757	  (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of
758	  excess pages there must be before trimming should occur, or zero if
759	  no trimming is to occur.
760
761	  This option specifies the initial value of this option.  The default
762	  of 1 says that all excess pages should be trimmed.
763
764	  See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
765
766config ARCH_WANT_GENERAL_HUGETLB
767	bool
768
769config ARCH_WANTS_THP_SWAP
770	def_bool n
771
772config PERSISTENT_HUGE_ZERO_FOLIO
773	bool "Allocate a PMD sized folio for zeroing"
774	depends on TRANSPARENT_HUGEPAGE
775	help
776	  Enable this option to reduce the runtime refcounting overhead
777	  of the huge zero folio and expand the places in the kernel
778	  that can use huge zero folios. For instance, block I/O benefits
779	  from access to large folios for zeroing memory.
780
781	  With this option enabled, the huge zero folio is allocated
782	  once and never freed. One full huge page's worth of memory shall
783	  be used.
784
785	  Say Y if your system has lots of memory. Say N if you are
786	  memory constrained.
787
788config MM_ID
789	def_bool n
790
791menuconfig TRANSPARENT_HUGEPAGE
792	bool "Transparent Hugepage Support"
793	depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT
794	select COMPACTION
795	select XARRAY_MULTI
796	select MM_ID
797	help
798	  Transparent Hugepages allows the kernel to use huge pages and
799	  huge tlb transparently to the applications whenever possible.
800	  This feature can improve computing performance to certain
801	  applications by speeding up page faults during memory
802	  allocation, by reducing the number of tlb misses and by speeding
803	  up the pagetable walking.
804
805	  If memory constrained on embedded, you may want to say N.
806
807if TRANSPARENT_HUGEPAGE
808
809choice
810	prompt "Transparent Hugepage Support sysfs defaults"
811	depends on TRANSPARENT_HUGEPAGE
812	default TRANSPARENT_HUGEPAGE_ALWAYS
813	help
814	  Selects the sysfs defaults for Transparent Hugepage Support.
815
816	config TRANSPARENT_HUGEPAGE_ALWAYS
817		bool "always"
818	help
819	  Enabling Transparent Hugepage always, can increase the
820	  memory footprint of applications without a guaranteed
821	  benefit but it will work automatically for all applications.
822
823	config TRANSPARENT_HUGEPAGE_MADVISE
824		bool "madvise"
825	help
826	  Enabling Transparent Hugepage madvise, will only provide a
827	  performance improvement benefit to the applications using
828	  madvise(MADV_HUGEPAGE) but it won't risk to increase the
829	  memory footprint of applications without a guaranteed
830	  benefit.
831
832	config TRANSPARENT_HUGEPAGE_NEVER
833		bool "never"
834	help
835	  Disable Transparent Hugepage by default. It can still be
836	  enabled at runtime via sysfs.
837endchoice
838
839choice
840	prompt "Shmem hugepage allocation defaults"
841	depends on TRANSPARENT_HUGEPAGE
842	default TRANSPARENT_HUGEPAGE_SHMEM_HUGE_NEVER
843	help
844	  Selects the hugepage allocation policy defaults for
845	  the internal shmem mount.
846
847	  The selection made here can be overridden by using the kernel
848	  command line 'transparent_hugepage_shmem=' option.
849
850	config TRANSPARENT_HUGEPAGE_SHMEM_HUGE_NEVER
851		bool "never"
852	help
853	  Disable hugepage allocation for shmem mount by default. It can
854	  still be enabled with the kernel command line
855	  'transparent_hugepage_shmem=' option or at runtime via sysfs
856	  knob. Note that madvise(MADV_COLLAPSE) can still cause
857	  transparent huge pages to be obtained even if this mode is
858	  specified.
859
860	config TRANSPARENT_HUGEPAGE_SHMEM_HUGE_ALWAYS
861		bool "always"
862	help
863	  Always attempt to allocate hugepage for shmem mount, can
864	  increase the memory footprint of applications without a
865	  guaranteed benefit but it will work automatically for all
866	  applications.
867
868	config TRANSPARENT_HUGEPAGE_SHMEM_HUGE_WITHIN_SIZE
869		bool "within_size"
870	help
871	  Enable hugepage allocation for shmem mount if the allocation
872	  will be fully within the i_size. This configuration also takes
873	  into account any madvise(MADV_HUGEPAGE) hints that may be
874	  provided by the applications.
875
876	config TRANSPARENT_HUGEPAGE_SHMEM_HUGE_ADVISE
877		bool "advise"
878	help
879	  Enable hugepage allocation for the shmem mount exclusively when
880	  applications supply the madvise(MADV_HUGEPAGE) hint.
881	  This ensures that hugepages are used only in response to explicit
882	  requests from applications.
883endchoice
884
885choice
886	prompt "Tmpfs hugepage allocation defaults"
887	depends on TRANSPARENT_HUGEPAGE
888	default TRANSPARENT_HUGEPAGE_TMPFS_HUGE_NEVER
889	help
890	  Selects the hugepage allocation policy defaults for
891	  the tmpfs mount.
892
893	  The selection made here can be overridden by using the kernel
894	  command line 'transparent_hugepage_tmpfs=' option.
895
896	config TRANSPARENT_HUGEPAGE_TMPFS_HUGE_NEVER
897		bool "never"
898	help
899	  Disable hugepage allocation for tmpfs mount by default. It can
900	  still be enabled with the kernel command line
901	  'transparent_hugepage_tmpfs=' option. Note that
902	  madvise(MADV_COLLAPSE) can still cause transparent huge pages
903	  to be obtained even if this mode is specified.
904
905	config TRANSPARENT_HUGEPAGE_TMPFS_HUGE_ALWAYS
906		bool "always"
907	help
908	  Always attempt to allocate hugepage for tmpfs mount, can
909	  increase the memory footprint of applications without a
910	  guaranteed benefit but it will work automatically for all
911	  applications.
912
913	config TRANSPARENT_HUGEPAGE_TMPFS_HUGE_WITHIN_SIZE
914		bool "within_size"
915	help
916	  Enable hugepage allocation for tmpfs mount if the allocation
917	  will be fully within the i_size. This configuration also takes
918	  into account any madvise(MADV_HUGEPAGE) hints that may be
919	  provided by the applications.
920
921	config TRANSPARENT_HUGEPAGE_TMPFS_HUGE_ADVISE
922		bool "advise"
923	help
924	  Enable hugepage allocation for the tmpfs mount exclusively when
925	  applications supply the madvise(MADV_HUGEPAGE) hint.
926	  This ensures that hugepages are used only in response to explicit
927	  requests from applications.
928endchoice
929
930config THP_SWAP
931	def_bool y
932	depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP && 64BIT
933	help
934	  Swap transparent huge pages in one piece, without splitting.
935	  XXX: For now, swap cluster backing transparent huge page
936	  will be split after swapout.
937
938	  For selection by architectures with reasonable THP sizes.
939
940config READ_ONLY_THP_FOR_FS
941	bool "Read-only THP for filesystems (EXPERIMENTAL)"
942	depends on TRANSPARENT_HUGEPAGE
943
944	help
945	  Allow khugepaged to put read-only file-backed pages in THP.
946
947	  This is marked experimental because it is a new feature. Write
948	  support of file THPs will be developed in the next few release
949	  cycles.
950
951config NO_PAGE_MAPCOUNT
952	bool "No per-page mapcount (EXPERIMENTAL)"
953	help
954	  Do not maintain per-page mapcounts for pages part of larger
955	  allocations, such as transparent huge pages.
956
957	  When this config option is enabled, some interfaces that relied on
958	  this information will rely on less-precise per-allocation information
959	  instead: for example, using the average per-page mapcount in such
960	  a large allocation instead of the per-page mapcount.
961
962	  EXPERIMENTAL because the impact of some changes is still unclear.
963
964endif # TRANSPARENT_HUGEPAGE
965
966# simple helper to make the code a bit easier to read
967config PAGE_MAPCOUNT
968	def_bool !NO_PAGE_MAPCOUNT
969
970#
971# The architecture supports pgtable leaves that is larger than PAGE_SIZE
972#
973config PGTABLE_HAS_HUGE_LEAVES
974	def_bool TRANSPARENT_HUGEPAGE || HUGETLB_PAGE
975
976#
977# We can end up creating gigantic folio.
978#
979config HAVE_GIGANTIC_FOLIOS
980	def_bool (HUGETLB_PAGE && ARCH_HAS_GIGANTIC_PAGE) || \
981		 (ZONE_DEVICE && HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD)
982
983config ASYNC_KERNEL_PGTABLE_FREE
984	def_bool n
985
986# TODO: Allow to be enabled without THP
987config ARCH_SUPPORTS_HUGE_PFNMAP
988	def_bool n
989	depends on TRANSPARENT_HUGEPAGE
990
991config ARCH_SUPPORTS_PMD_PFNMAP
992	def_bool y
993	depends on ARCH_SUPPORTS_HUGE_PFNMAP && HAVE_ARCH_TRANSPARENT_HUGEPAGE
994
995config ARCH_SUPPORTS_PUD_PFNMAP
996	def_bool y
997	depends on ARCH_SUPPORTS_HUGE_PFNMAP && HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
998
999#
1000# Architectures that always use weak definitions for percpu
1001# variables in modules should set this.
1002#
1003config ARCH_MODULE_NEEDS_WEAK_PER_CPU
1004       bool
1005
1006#
1007# UP and nommu archs use km based percpu allocator
1008#
1009config NEED_PER_CPU_KM
1010	depends on !SMP || !MMU
1011	bool
1012	default y
1013
1014config NEED_PER_CPU_EMBED_FIRST_CHUNK
1015	bool
1016
1017config NEED_PER_CPU_PAGE_FIRST_CHUNK
1018	bool
1019
1020config USE_PERCPU_NUMA_NODE_ID
1021	bool
1022
1023config HAVE_SETUP_PER_CPU_AREA
1024	bool
1025
1026config CMA
1027	bool "Contiguous Memory Allocator"
1028	depends on MMU
1029	select MIGRATION
1030	select MEMORY_ISOLATION
1031	help
1032	  This enables the Contiguous Memory Allocator which allows other
1033	  subsystems to allocate big physically-contiguous blocks of memory.
1034	  CMA reserves a region of memory and allows only movable pages to
1035	  be allocated from it. This way, the kernel can use the memory for
1036	  pagecache and when a subsystem requests for contiguous area, the
1037	  allocated pages are migrated away to serve the contiguous request.
1038
1039	  If unsure, say "n".
1040
1041config CMA_DEBUGFS
1042	bool "CMA debugfs interface"
1043	depends on CMA && DEBUG_FS
1044	help
1045	  Turns on the DebugFS interface for CMA.
1046
1047config CMA_SYSFS
1048	bool "CMA information through sysfs interface"
1049	depends on CMA && SYSFS
1050	help
1051	  This option exposes some sysfs attributes to get information
1052	  from CMA.
1053
1054config CMA_AREAS
1055	int "Maximum count of the CMA areas"
1056	depends on CMA
1057	default 20 if NUMA
1058	default 8
1059	help
1060	  CMA allows to create CMA areas for particular purpose, mainly,
1061	  used as device private area. This parameter sets the maximum
1062	  number of CMA area in the system.
1063
1064	  If unsure, leave the default value "8" in UMA and "20" in NUMA.
1065
1066#
1067# Select this config option from the architecture Kconfig, if available, to set
1068# the max page order for physically contiguous allocations.
1069#
1070config ARCH_FORCE_MAX_ORDER
1071	int
1072
1073#
1074# When ARCH_FORCE_MAX_ORDER is not defined,
1075# the default page block order is MAX_PAGE_ORDER (10) as per
1076# include/linux/mmzone.h.
1077#
1078config PAGE_BLOCK_MAX_ORDER
1079	int "Page Block Order Upper Limit"
1080	range 1 10 if ARCH_FORCE_MAX_ORDER = 0
1081	default 10 if ARCH_FORCE_MAX_ORDER = 0
1082	range 1 ARCH_FORCE_MAX_ORDER if ARCH_FORCE_MAX_ORDER != 0
1083	default ARCH_FORCE_MAX_ORDER if ARCH_FORCE_MAX_ORDER != 0
1084	help
1085	  The page block order refers to the power of two number of pages that
1086	  are physically contiguous and can have a migrate type associated to
1087	  them. The maximum size of the page block order is at least limited by
1088	  ARCH_FORCE_MAX_ORDER/MAX_PAGE_ORDER.
1089
1090	  This config adds a new upper limit of default page block
1091	  order when the page block order is required to be smaller than
1092	  ARCH_FORCE_MAX_ORDER/MAX_PAGE_ORDER or other limits
1093	  (see include/linux/pageblock-flags.h for details).
1094
1095	  Reducing pageblock order can negatively impact THP generation
1096	  success rate. If your workloads use THP heavily, please use this
1097	  option with caution.
1098
1099	  Don't change if unsure.
1100
1101config MEM_SOFT_DIRTY
1102	bool "Track memory changes"
1103	depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS
1104	select PROC_PAGE_MONITOR
1105	help
1106	  This option enables memory changes tracking by introducing a
1107	  soft-dirty bit on pte-s. This bit it set when someone writes
1108	  into a page just as regular dirty bit, but unlike the latter
1109	  it can be cleared by hands.
1110
1111	  See Documentation/admin-guide/mm/soft-dirty.rst for more details.
1112
1113config GENERIC_EARLY_IOREMAP
1114	bool
1115
1116config STACK_MAX_DEFAULT_SIZE_MB
1117	int "Default maximum user stack size for 32-bit processes (MB)"
1118	default 100
1119	range 8 2048
1120	depends on STACK_GROWSUP && (!64BIT || COMPAT)
1121	help
1122	  This is the maximum stack size in Megabytes in the VM layout of 32-bit
1123	  user processes when the stack grows upwards (currently only on parisc
1124	  arch) when the RLIMIT_STACK hard limit is unlimited.
1125
1126	  A sane initial value is 100 MB.
1127
1128config DEFERRED_STRUCT_PAGE_INIT
1129	bool "Defer initialisation of struct pages to kthreads"
1130	depends on SPARSEMEM
1131	depends on !NEED_PER_CPU_KM
1132	depends on 64BIT
1133	depends on !KMSAN
1134	select PADATA
1135	help
1136	  Ordinarily all struct pages are initialised during early boot in a
1137	  single thread. On very large machines this can take a considerable
1138	  amount of time. If this option is set, large machines will bring up
1139	  a subset of memmap at boot and then initialise the rest in parallel.
1140	  This has a potential performance impact on tasks running early in the
1141	  lifetime of the system until these kthreads finish the
1142	  initialisation.
1143
1144config PAGE_IDLE_FLAG
1145	bool
1146	select PAGE_EXTENSION if !64BIT
1147	help
1148	  This adds PG_idle and PG_young flags to 'struct page'.  PTE Accessed
1149	  bit writers can set the state of the bit in the flags so that PTE
1150	  Accessed bit readers may avoid disturbance.
1151
1152config IDLE_PAGE_TRACKING
1153	bool "Enable idle page tracking"
1154	depends on SYSFS && MMU
1155	select PAGE_IDLE_FLAG
1156	help
1157	  This feature allows to estimate the amount of user pages that have
1158	  not been touched during a given period of time. This information can
1159	  be useful to tune memory cgroup limits and/or for job placement
1160	  within a compute cluster.
1161
1162	  See Documentation/admin-guide/mm/idle_page_tracking.rst for
1163	  more details.
1164
1165# Architectures which implement cpu_dcache_is_aliasing() to query
1166# whether the data caches are aliased (VIVT or VIPT with dcache
1167# aliasing) need to select this.
1168config ARCH_HAS_CPU_CACHE_ALIASING
1169	bool
1170
1171config ARCH_HAS_CACHE_LINE_SIZE
1172	bool
1173
1174config ARCH_HAS_CURRENT_STACK_POINTER
1175	bool
1176	help
1177	  In support of HARDENED_USERCOPY performing stack variable lifetime
1178	  checking, an architecture-agnostic way to find the stack pointer
1179	  is needed. Once an architecture defines an unsigned long global
1180	  register alias named "current_stack_pointer", this config can be
1181	  selected.
1182
1183config ARCH_HAS_ZONE_DMA_SET
1184	bool
1185
1186config ZONE_DMA
1187	bool "Support DMA zone" if ARCH_HAS_ZONE_DMA_SET
1188	default y if ARM64 || X86
1189
1190config ZONE_DMA32
1191	bool "Support DMA32 zone" if ARCH_HAS_ZONE_DMA_SET
1192	depends on !X86_32
1193	default y if ARM64
1194
1195config ZONE_DEVICE
1196	bool "Device memory (pmem, HMM, etc...) hotplug support"
1197	depends on MEMORY_HOTPLUG
1198	depends on MEMORY_HOTREMOVE
1199	depends on SPARSEMEM_VMEMMAP
1200	select XARRAY_MULTI
1201
1202	help
1203	  Device memory hotplug support allows for establishing pmem,
1204	  or other device driver discovered memory regions, in the
1205	  memmap. This allows pfn_to_page() lookups of otherwise
1206	  "device-physical" addresses which is needed for DAX, PCI_P2PDMA, and
1207	  DEVICE_PRIVATE features among others.
1208
1209	  Enabling this option will reduce the entropy of x86 KASLR memory
1210	  regions. For example - on a 46 bit system, the entropy goes down
1211	  from 16 bits to 15 bits. The actual reduction in entropy depends
1212	  on the physical address bits, on processor features, kernel config
1213	  (5 level page table) and physical memory present on the system.
1214
1215#
1216# Helpers to mirror range of the CPU page tables of a process into device page
1217# tables.
1218#
1219config HMM_MIRROR
1220	bool
1221	depends on MMU
1222
1223config GET_FREE_REGION
1224	bool
1225
1226config DEVICE_PRIVATE
1227	bool "Unaddressable device memory (GPU memory, ...)"
1228	depends on ZONE_DEVICE
1229	select GET_FREE_REGION
1230
1231	help
1232	  Allows creation of struct pages to represent unaddressable device
1233	  memory; i.e., memory that is only accessible from the device (or
1234	  group of devices). You likely also want to select HMM_MIRROR.
1235
1236config VMAP_PFN
1237	bool
1238
1239config ARCH_USES_HIGH_VMA_FLAGS
1240	bool
1241config ARCH_HAS_PKEYS
1242	bool
1243
1244config ARCH_USES_PG_ARCH_2
1245	bool
1246config ARCH_USES_PG_ARCH_3
1247	bool
1248
1249config VM_EVENT_COUNTERS
1250	default y
1251	bool "Enable VM event counters for /proc/vmstat" if EXPERT
1252	help
1253	  VM event counters are needed for event counts to be shown.
1254	  This option allows the disabling of the VM event counters
1255	  on EXPERT systems.  /proc/vmstat will only show page counts
1256	  if VM event counters are disabled.
1257
1258config PERCPU_STATS
1259	bool "Collect percpu memory statistics"
1260	help
1261	  This feature collects and exposes statistics via debugfs. The
1262	  information includes global and per chunk statistics, which can
1263	  be used to help understand percpu memory usage.
1264
1265config GUP_TEST
1266	bool "Enable infrastructure for get_user_pages()-related unit tests"
1267	depends on DEBUG_FS
1268	help
1269	  Provides /sys/kernel/debug/gup_test, which in turn provides a way
1270	  to make ioctl calls that can launch kernel-based unit tests for
1271	  the get_user_pages*() and pin_user_pages*() family of API calls.
1272
1273	  These tests include benchmark testing of the _fast variants of
1274	  get_user_pages*() and pin_user_pages*(), as well as smoke tests of
1275	  the non-_fast variants.
1276
1277	  There is also a sub-test that allows running dump_page() on any
1278	  of up to eight pages (selected by command line args) within the
1279	  range of user-space addresses. These pages are either pinned via
1280	  pin_user_pages*(), or pinned via get_user_pages*(), as specified
1281	  by other command line arguments.
1282
1283	  See tools/testing/selftests/mm/gup_test.c
1284
1285comment "GUP_TEST needs to have DEBUG_FS enabled"
1286	depends on !GUP_TEST && !DEBUG_FS
1287
1288config GUP_GET_PXX_LOW_HIGH
1289	bool
1290
1291config DMAPOOL_TEST
1292	tristate "Enable a module to run time tests on dma_pool"
1293	depends on HAS_DMA
1294	help
1295	  Provides a test module that will allocate and free many blocks of
1296	  various sizes and report how long it takes. This is intended to
1297	  provide a consistent way to measure how changes to the
1298	  dma_pool_alloc/free routines affect performance.
1299
1300config ARCH_HAS_PTE_SPECIAL
1301	bool
1302
1303config MAPPING_DIRTY_HELPERS
1304        bool
1305
1306config KMAP_LOCAL
1307	bool
1308
1309config KMAP_LOCAL_NON_LINEAR_PTE_ARRAY
1310	bool
1311
1312config MEMFD_CREATE
1313	bool "Enable memfd_create() system call" if EXPERT
1314
1315config SECRETMEM
1316	default y
1317	bool "Enable memfd_secret() system call" if EXPERT
1318	depends on ARCH_HAS_SET_DIRECT_MAP
1319	help
1320	  Enable the memfd_secret() system call with the ability to create
1321	  memory areas visible only in the context of the owning process and
1322	  not mapped to other processes and other kernel page tables.
1323
1324config ANON_VMA_NAME
1325	bool "Anonymous VMA name support"
1326	depends on PROC_FS && ADVISE_SYSCALLS && MMU
1327
1328	help
1329	  Allow naming anonymous virtual memory areas.
1330
1331	  This feature allows assigning names to virtual memory areas. Assigned
1332	  names can be later retrieved from /proc/pid/maps and /proc/pid/smaps
1333	  and help identifying individual anonymous memory areas.
1334	  Assigning a name to anonymous virtual memory area might prevent that
1335	  area from being merged with adjacent virtual memory areas due to the
1336	  difference in their name.
1337
1338config HAVE_ARCH_USERFAULTFD_WP
1339	bool
1340	help
1341	  Arch has userfaultfd write protection support
1342
1343config HAVE_ARCH_USERFAULTFD_MINOR
1344	bool
1345	help
1346	  Arch has userfaultfd minor fault support
1347
1348menuconfig USERFAULTFD
1349	bool "Enable userfaultfd() system call"
1350	depends on MMU
1351	help
1352	  Enable the userfaultfd() system call that allows to intercept and
1353	  handle page faults in userland.
1354
1355if USERFAULTFD
1356config PTE_MARKER_UFFD_WP
1357	bool "Userfaultfd write protection support for shmem/hugetlbfs"
1358	default y
1359	depends on HAVE_ARCH_USERFAULTFD_WP
1360
1361	help
1362	  Allows to create marker PTEs for userfaultfd write protection
1363	  purposes.  It is required to enable userfaultfd write protection on
1364	  file-backed memory types like shmem and hugetlbfs.
1365endif # USERFAULTFD
1366
1367# multi-gen LRU {
1368config LRU_GEN
1369	bool "Multi-Gen LRU"
1370	depends on MMU
1371	# make sure folio->flags has enough spare bits
1372	depends on 64BIT || !SPARSEMEM || SPARSEMEM_VMEMMAP
1373	help
1374	  A high performance LRU implementation to overcommit memory. See
1375	  Documentation/admin-guide/mm/multigen_lru.rst for details.
1376
1377config LRU_GEN_ENABLED
1378	bool "Enable by default"
1379	depends on LRU_GEN
1380	help
1381	  This option enables the multi-gen LRU by default.
1382
1383config LRU_GEN_STATS
1384	bool "Full stats for debugging"
1385	depends on LRU_GEN
1386	help
1387	  Do not enable this option unless you plan to look at historical stats
1388	  from evicted generations for debugging purpose.
1389
1390	  This option has a per-memcg and per-node memory overhead.
1391
1392config LRU_GEN_WALKS_MMU
1393	def_bool y
1394	depends on LRU_GEN && ARCH_HAS_HW_PTE_YOUNG
1395# }
1396
1397config ARCH_SUPPORTS_PER_VMA_LOCK
1398       def_bool n
1399
1400config PER_VMA_LOCK
1401	def_bool y
1402	depends on ARCH_SUPPORTS_PER_VMA_LOCK && MMU && SMP
1403	help
1404	  Allow per-vma locking during page fault handling.
1405
1406	  This feature allows locking each virtual memory area separately when
1407	  handling page faults instead of taking mmap_lock.
1408
1409config LOCK_MM_AND_FIND_VMA
1410	bool
1411	depends on !STACK_GROWSUP
1412
1413config IOMMU_MM_DATA
1414	bool
1415
1416config EXECMEM
1417	bool
1418
1419config NUMA_MEMBLKS
1420	bool
1421
1422config NUMA_EMU
1423	bool "NUMA emulation"
1424	depends on NUMA_MEMBLKS
1425	depends on X86 || GENERIC_ARCH_NUMA
1426	help
1427	  Enable NUMA emulation. A flat machine will be split
1428	  into virtual nodes when booted with "numa=fake=N", where N is the
1429	  number of nodes. This is only useful for debugging.
1430
1431config ARCH_HAS_USER_SHADOW_STACK
1432	bool
1433	help
1434	  The architecture has hardware support for userspace shadow call
1435          stacks (eg, x86 CET, arm64 GCS or RISC-V Zicfiss).
1436
1437config HAVE_ARCH_TLB_REMOVE_TABLE
1438	def_bool n
1439
1440config PT_RECLAIM
1441	def_bool y
1442	depends on MMU_GATHER_RCU_TABLE_FREE && !HAVE_ARCH_TLB_REMOVE_TABLE
1443	help
1444	  Try to reclaim empty user page table pages in paths other than munmap
1445	  and exit_mmap path.
1446
1447	  Note: now only empty user PTE page table pages will be reclaimed.
1448
1449config FIND_NORMAL_PAGE
1450	def_bool n
1451
1452config ARCH_HAS_LAZY_MMU_MODE
1453	bool
1454	help
1455	  The architecture uses the lazy MMU mode. This allows changes to
1456	  MMU-related architectural state to be deferred until the mode is
1457	  exited. See <linux/pgtable.h> for details.
1458
1459config LAZY_MMU_MODE_KUNIT_TEST
1460	tristate "KUnit tests for the lazy MMU mode" if !KUNIT_ALL_TESTS
1461	depends on ARCH_HAS_LAZY_MMU_MODE
1462	depends on KUNIT
1463	default KUNIT_ALL_TESTS
1464	help
1465	  Enable this option to check that the lazy MMU mode interface behaves
1466	  as expected. Only tests for the generic interface are included (not
1467	  architecture-specific behaviours).
1468
1469	  If unsure, say N.
1470
1471source "mm/damon/Kconfig"
1472
1473endmenu
1474