1 # SPDX-License-Identifier: GPL-2.0-only
3 menu "Memory Management options"
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.
16 bool "Support for paging of anonymous memory (swap)"
17 depends on MMU && BLOCK && !ARCH_NO_SWAP
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.
26 bool "Compressed cache for swap pages"
32 A lightweight compressed cache for swap pages. It takes
33 pages that are in the process of being swapped out and attempts to
34 compress them into a dynamically allocated RAM-based memory pool.
35 This can result in a significant I/O reduction on swap device and,
36 in the case where decompressing from RAM is faster than swap device
37 reads, can also improve workload performance.
39 config ZSWAP_DEFAULT_ON
40 bool "Enable the compressed cache for swap pages by default"
43 If selected, the compressed cache for swap pages will be enabled
44 at boot, otherwise it will be disabled.
46 The selection made here can be overridden by using the kernel
47 command line 'zswap.enabled=' option.
49 config ZSWAP_EXCLUSIVE_LOADS_DEFAULT_ON
50 bool "Invalidate zswap entries when pages are loaded"
53 If selected, exclusive loads for zswap will be enabled at boot,
54 otherwise it will be disabled.
56 If exclusive loads are enabled, when a page is loaded from zswap,
57 the zswap entry is invalidated at once, as opposed to leaving it
58 in zswap until the swap entry is freed.
60 This avoids having two copies of the same page in memory
61 (compressed and uncompressed) after faulting in a page from zswap.
62 The cost is that if the page was never dirtied and needs to be
63 swapped out again, it will be re-compressed.
66 prompt "Default compressor"
68 default ZSWAP_COMPRESSOR_DEFAULT_LZO
70 Selects the default compression algorithm for the compressed cache
73 For an overview what kind of performance can be expected from
74 a particular compression algorithm please refer to the benchmarks
75 available at the following LWN page:
76 https://lwn.net/Articles/751795/
78 If in doubt, select 'LZO'.
80 The selection made here can be overridden by using the kernel
81 command line 'zswap.compressor=' option.
83 config ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
87 Use the Deflate algorithm as the default compression algorithm.
89 config ZSWAP_COMPRESSOR_DEFAULT_LZO
93 Use the LZO algorithm as the default compression algorithm.
95 config ZSWAP_COMPRESSOR_DEFAULT_842
99 Use the 842 algorithm as the default compression algorithm.
101 config ZSWAP_COMPRESSOR_DEFAULT_LZ4
105 Use the LZ4 algorithm as the default compression algorithm.
107 config ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
111 Use the LZ4HC algorithm as the default compression algorithm.
113 config ZSWAP_COMPRESSOR_DEFAULT_ZSTD
117 Use the zstd algorithm as the default compression algorithm.
120 config ZSWAP_COMPRESSOR_DEFAULT
123 default "deflate" if ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
124 default "lzo" if ZSWAP_COMPRESSOR_DEFAULT_LZO
125 default "842" if ZSWAP_COMPRESSOR_DEFAULT_842
126 default "lz4" if ZSWAP_COMPRESSOR_DEFAULT_LZ4
127 default "lz4hc" if ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
128 default "zstd" if ZSWAP_COMPRESSOR_DEFAULT_ZSTD
132 prompt "Default allocator"
134 default ZSWAP_ZPOOL_DEFAULT_ZBUD
136 Selects the default allocator for the compressed cache for
138 The default is 'zbud' for compatibility, however please do
139 read the description of each of the allocators below before
140 making a right choice.
142 The selection made here can be overridden by using the kernel
143 command line 'zswap.zpool=' option.
145 config ZSWAP_ZPOOL_DEFAULT_ZBUD
149 Use the zbud allocator as the default allocator.
151 config ZSWAP_ZPOOL_DEFAULT_Z3FOLD
155 Use the z3fold allocator as the default allocator.
157 config ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
161 Use the zsmalloc allocator as the default allocator.
164 config ZSWAP_ZPOOL_DEFAULT
167 default "zbud" if ZSWAP_ZPOOL_DEFAULT_ZBUD
168 default "z3fold" if ZSWAP_ZPOOL_DEFAULT_Z3FOLD
169 default "zsmalloc" if ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
173 tristate "2:1 compression allocator (zbud)"
176 A special purpose allocator for storing compressed pages.
177 It is designed to store up to two compressed pages per physical
178 page. While this design limits storage density, it has simple and
179 deterministic reclaim properties that make it preferable to a higher
180 density approach when reclaim will be used.
183 tristate "3:1 compression allocator (z3fold)"
186 A special purpose allocator for storing compressed pages.
187 It is designed to store up to three compressed pages per physical
188 page. It is a ZBUD derivative so the simplicity and determinism are
193 prompt "N:1 compression allocator (zsmalloc)" if ZSWAP
196 zsmalloc is a slab-based memory allocator designed to store
197 pages of various compression levels efficiently. It achieves
198 the highest storage density with the least amount of fragmentation.
201 bool "Export zsmalloc statistics"
205 This option enables code in the zsmalloc to collect various
206 statistics about what's happening in zsmalloc and exports that
207 information to userspace via debugfs.
210 config ZSMALLOC_CHAIN_SIZE
211 int "Maximum number of physical pages per-zspage"
216 This option sets the upper limit on the number of physical pages
217 that a zmalloc page (zspage) can consist of. The optimal zspage
218 chain size is calculated for each size class during the
219 initialization of the pool.
221 Changing this option can alter the characteristics of size classes,
222 such as the number of pages per zspage and the number of objects
223 per zspage. This can also result in different configurations of
224 the pool, as zsmalloc merges size classes with similar
227 For more information, see zsmalloc documentation.
229 menu "SLAB allocator options"
232 prompt "Choose SLAB allocator"
235 This option allows to select a slab allocator.
239 depends on !PREEMPT_RT
240 select HAVE_HARDENED_USERCOPY_ALLOCATOR
242 The regular slab allocator that is established and known to work
243 well in all environments. It organizes cache hot objects in
244 per cpu and per node queues.
247 bool "SLUB (Unqueued Allocator)"
248 select HAVE_HARDENED_USERCOPY_ALLOCATOR
250 SLUB is a slab allocator that minimizes cache line usage
251 instead of managing queues of cached objects (SLAB approach).
252 Per cpu caching is realized using slabs of objects instead
253 of queues of objects. SLUB can use memory efficiently
254 and has enhanced diagnostics. SLUB is the default choice for
260 bool "Configure SLUB for minimal memory footprint"
261 depends on SLUB && EXPERT
262 select SLAB_MERGE_DEFAULT
264 Configures the SLUB allocator in a way to achieve minimal memory
265 footprint, sacrificing scalability, debugging and other features.
266 This is intended only for the smallest system that had used the
267 SLOB allocator and is not recommended for systems with more than
272 config SLAB_MERGE_DEFAULT
273 bool "Allow slab caches to be merged"
275 depends on SLAB || SLUB
277 For reduced kernel memory fragmentation, slab caches can be
278 merged when they share the same size and other characteristics.
279 This carries a risk of kernel heap overflows being able to
280 overwrite objects from merged caches (and more easily control
281 cache layout), which makes such heap attacks easier to exploit
282 by attackers. By keeping caches unmerged, these kinds of exploits
283 can usually only damage objects in the same cache. To disable
284 merging at runtime, "slab_nomerge" can be passed on the kernel
287 config SLAB_FREELIST_RANDOM
288 bool "Randomize slab freelist"
289 depends on SLAB || (SLUB && !SLUB_TINY)
291 Randomizes the freelist order used on creating new pages. This
292 security feature reduces the predictability of the kernel slab
293 allocator against heap overflows.
295 config SLAB_FREELIST_HARDENED
296 bool "Harden slab freelist metadata"
297 depends on SLAB || (SLUB && !SLUB_TINY)
299 Many kernel heap attacks try to target slab cache metadata and
300 other infrastructure. This options makes minor performance
301 sacrifices to harden the kernel slab allocator against common
302 freelist exploit methods. Some slab implementations have more
303 sanity-checking than others. This option is most effective with
308 bool "Enable SLUB performance statistics"
309 depends on SLUB && SYSFS && !SLUB_TINY
311 SLUB statistics are useful to debug SLUBs allocation behavior in
312 order find ways to optimize the allocator. This should never be
313 enabled for production use since keeping statistics slows down
314 the allocator by a few percentage points. The slabinfo command
315 supports the determination of the most active slabs to figure
316 out which slabs are relevant to a particular load.
317 Try running: slabinfo -DA
319 config SLUB_CPU_PARTIAL
321 depends on SLUB && SMP && !SLUB_TINY
322 bool "SLUB per cpu partial cache"
324 Per cpu partial caches accelerate objects allocation and freeing
325 that is local to a processor at the price of more indeterminism
326 in the latency of the free. On overflow these caches will be cleared
327 which requires the taking of locks that may cause latency spikes.
328 Typically one would choose no for a realtime system.
330 endmenu # SLAB allocator options
332 config SHUFFLE_PAGE_ALLOCATOR
333 bool "Page allocator randomization"
334 default SLAB_FREELIST_RANDOM && ACPI_NUMA
336 Randomization of the page allocator improves the average
337 utilization of a direct-mapped memory-side-cache. See section
338 5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI
339 6.2a specification for an example of how a platform advertises
340 the presence of a memory-side-cache. There are also incidental
341 security benefits as it reduces the predictability of page
342 allocations to compliment SLAB_FREELIST_RANDOM, but the
343 default granularity of shuffling on the MAX_ORDER i.e, 10th
344 order of pages is selected based on cache utilization benefits
347 While the randomization improves cache utilization it may
348 negatively impact workloads on platforms without a cache. For
349 this reason, by default, the randomization is enabled only
350 after runtime detection of a direct-mapped memory-side-cache.
351 Otherwise, the randomization may be force enabled with the
352 'page_alloc.shuffle' kernel command line parameter.
357 bool "Disable heap randomization"
360 Randomizing heap placement makes heap exploits harder, but it
361 also breaks ancient binaries (including anything libc5 based).
362 This option changes the bootup default to heap randomization
363 disabled, and can be overridden at runtime by setting
364 /proc/sys/kernel/randomize_va_space to 2.
366 On non-ancient distros (post-2000 ones) N is usually a safe choice.
368 config MMAP_ALLOW_UNINITIALIZED
369 bool "Allow mmapped anonymous memory to be uninitialized"
370 depends on EXPERT && !MMU
373 Normally, and according to the Linux spec, anonymous memory obtained
374 from mmap() has its contents cleared before it is passed to
375 userspace. Enabling this config option allows you to request that
376 mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus
377 providing a huge performance boost. If this option is not enabled,
378 then the flag will be ignored.
380 This is taken advantage of by uClibc's malloc(), and also by
381 ELF-FDPIC binfmt's brk and stack allocator.
383 Because of the obvious security issues, this option should only be
384 enabled on embedded devices where you control what is run in
385 userspace. Since that isn't generally a problem on no-MMU systems,
386 it is normally safe to say Y here.
388 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
390 config SELECT_MEMORY_MODEL
392 depends on ARCH_SELECT_MEMORY_MODEL
395 prompt "Memory model"
396 depends on SELECT_MEMORY_MODEL
397 default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT
398 default FLATMEM_MANUAL
400 This option allows you to change some of the ways that
401 Linux manages its memory internally. Most users will
402 only have one option here selected by the architecture
403 configuration. This is normal.
405 config FLATMEM_MANUAL
407 depends on !ARCH_SPARSEMEM_ENABLE || ARCH_FLATMEM_ENABLE
409 This option is best suited for non-NUMA systems with
410 flat address space. The FLATMEM is the most efficient
411 system in terms of performance and resource consumption
412 and it is the best option for smaller systems.
414 For systems that have holes in their physical address
415 spaces and for features like NUMA and memory hotplug,
416 choose "Sparse Memory".
418 If unsure, choose this option (Flat Memory) over any other.
420 config SPARSEMEM_MANUAL
422 depends on ARCH_SPARSEMEM_ENABLE
424 This will be the only option for some systems, including
425 memory hot-plug systems. This is normal.
427 This option provides efficient support for systems with
428 holes is their physical address space and allows memory
429 hot-plug and hot-remove.
431 If unsure, choose "Flat Memory" over this option.
437 depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL
441 depends on !SPARSEMEM || FLATMEM_MANUAL
444 # SPARSEMEM_EXTREME (which is the default) does some bootmem
445 # allocations when sparse_init() is called. If this cannot
446 # be done on your architecture, select this option. However,
447 # statically allocating the mem_section[] array can potentially
448 # consume vast quantities of .bss, so be careful.
450 # This option will also potentially produce smaller runtime code
451 # with gcc 3.4 and later.
453 config SPARSEMEM_STATIC
457 # Architecture platforms which require a two level mem_section in SPARSEMEM
458 # must select this option. This is usually for architecture platforms with
459 # an extremely sparse physical address space.
461 config SPARSEMEM_EXTREME
463 depends on SPARSEMEM && !SPARSEMEM_STATIC
465 config SPARSEMEM_VMEMMAP_ENABLE
468 config SPARSEMEM_VMEMMAP
469 bool "Sparse Memory virtual memmap"
470 depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE
473 SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise
474 pfn_to_page and page_to_pfn operations. This is the most
475 efficient option when sufficient kernel resources are available.
477 # Select this config option from the architecture Kconfig, if it is preferred
478 # to enable the feature of HugeTLB/dev_dax vmemmap optimization.
480 config ARCH_WANT_OPTIMIZE_VMEMMAP
483 config HAVE_MEMBLOCK_PHYS_MAP
490 # Don't discard allocated memory used to track "memory" and "reserved" memblocks
491 # after early boot, so it can still be used to test for validity of memory.
492 # Also, memblocks are updated with memory hot(un)plug.
493 config ARCH_KEEP_MEMBLOCK
496 # Keep arch NUMA mapping infrastructure post-init.
497 config NUMA_KEEP_MEMINFO
500 config MEMORY_ISOLATION
503 # IORESOURCE_SYSTEM_RAM regions in the kernel resource tree that are marked
504 # IORESOURCE_EXCLUSIVE cannot be mapped to user space, for example, via
506 config EXCLUSIVE_SYSTEM_RAM
508 depends on !DEVMEM || STRICT_DEVMEM
511 # Only be set on architectures that have completely implemented memory hotplug
512 # feature. If you are not sure, don't touch it.
514 config HAVE_BOOTMEM_INFO_NODE
517 config ARCH_ENABLE_MEMORY_HOTPLUG
520 config ARCH_ENABLE_MEMORY_HOTREMOVE
523 # eventually, we can have this option just 'select SPARSEMEM'
524 menuconfig MEMORY_HOTPLUG
525 bool "Memory hotplug"
526 select MEMORY_ISOLATION
528 depends on ARCH_ENABLE_MEMORY_HOTPLUG
530 select NUMA_KEEP_MEMINFO if NUMA
534 config MEMORY_HOTPLUG_DEFAULT_ONLINE
535 bool "Online the newly added memory blocks by default"
536 depends on MEMORY_HOTPLUG
538 This option sets the default policy setting for memory hotplug
539 onlining policy (/sys/devices/system/memory/auto_online_blocks) which
540 determines what happens to newly added memory regions. Policy setting
541 can always be changed at runtime.
542 See Documentation/admin-guide/mm/memory-hotplug.rst for more information.
544 Say Y here if you want all hot-plugged memory blocks to appear in
545 'online' state by default.
546 Say N here if you want the default policy to keep all hot-plugged
547 memory blocks in 'offline' state.
549 config MEMORY_HOTREMOVE
550 bool "Allow for memory hot remove"
551 select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64)
552 depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE
555 config MHP_MEMMAP_ON_MEMORY
557 depends on MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP
558 depends on ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
560 endif # MEMORY_HOTPLUG
562 # Heavily threaded applications may benefit from splitting the mm-wide
563 # page_table_lock, so that faults on different parts of the user address
564 # space can be handled with less contention: split it at this NR_CPUS.
565 # Default to 4 for wider testing, though 8 might be more appropriate.
566 # ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
567 # PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
568 # SPARC32 allocates multiple pte tables within a single page, and therefore
569 # a per-page lock leads to problems when multiple tables need to be locked
570 # at the same time (e.g. copy_page_range()).
571 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
573 config SPLIT_PTLOCK_CPUS
575 default "999999" if !MMU
576 default "999999" if ARM && !CPU_CACHE_VIPT
577 default "999999" if PARISC && !PA20
578 default "999999" if SPARC32
581 config ARCH_ENABLE_SPLIT_PMD_PTLOCK
585 # support for memory balloon
586 config MEMORY_BALLOON
590 # support for memory balloon compaction
591 config BALLOON_COMPACTION
592 bool "Allow for balloon memory compaction/migration"
594 depends on COMPACTION && MEMORY_BALLOON
596 Memory fragmentation introduced by ballooning might reduce
597 significantly the number of 2MB contiguous memory blocks that can be
598 used within a guest, thus imposing performance penalties associated
599 with the reduced number of transparent huge pages that could be used
600 by the guest workload. Allowing the compaction & migration for memory
601 pages enlisted as being part of memory balloon devices avoids the
602 scenario aforementioned and helps improving memory defragmentation.
605 # support for memory compaction
607 bool "Allow for memory compaction"
612 Compaction is the only memory management component to form
613 high order (larger physically contiguous) memory blocks
614 reliably. The page allocator relies on compaction heavily and
615 the lack of the feature can lead to unexpected OOM killer
616 invocations for high order memory requests. You shouldn't
617 disable this option unless there really is a strong reason for
618 it and then we would be really interested to hear about that at
621 config COMPACT_UNEVICTABLE_DEFAULT
623 depends on COMPACTION
624 default 0 if PREEMPT_RT
628 # support for free page reporting
629 config PAGE_REPORTING
630 bool "Free page reporting"
633 Free page reporting allows for the incremental acquisition of
634 free pages from the buddy allocator for the purpose of reporting
635 those pages to another entity, such as a hypervisor, so that the
636 memory can be freed within the host for other uses.
639 # support for page migration
642 bool "Page migration"
644 depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU
646 Allows the migration of the physical location of pages of processes
647 while the virtual addresses are not changed. This is useful in
648 two situations. The first is on NUMA systems to put pages nearer
649 to the processors accessing. The second is when allocating huge
650 pages as migration can relocate pages to satisfy a huge page
651 allocation instead of reclaiming.
653 config DEVICE_MIGRATION
654 def_bool MIGRATION && ZONE_DEVICE
656 config ARCH_ENABLE_HUGEPAGE_MIGRATION
659 config ARCH_ENABLE_THP_MIGRATION
662 config HUGETLB_PAGE_SIZE_VARIABLE
665 Allows the pageblock_order value to be dynamic instead of just standard
666 HUGETLB_PAGE_ORDER when there are multiple HugeTLB page sizes available
669 Note that the pageblock_order cannot exceed MAX_ORDER and will be
670 clamped down to MAX_ORDER.
673 def_bool (MEMORY_ISOLATION && COMPACTION) || CMA
675 config PHYS_ADDR_T_64BIT
679 bool "Enable bounce buffers"
681 depends on BLOCK && MMU && HIGHMEM
683 Enable bounce buffers for devices that cannot access the full range of
684 memory available to the CPU. Enabled by default when HIGHMEM is
685 selected, but you may say n to override this.
692 bool "Enable KSM for page merging"
696 Enable Kernel Samepage Merging: KSM periodically scans those areas
697 of an application's address space that an app has advised may be
698 mergeable. When it finds pages of identical content, it replaces
699 the many instances by a single page with that content, so
700 saving memory until one or another app needs to modify the content.
701 Recommended for use with KVM, or with other duplicative applications.
702 See Documentation/mm/ksm.rst for more information: KSM is inactive
703 until a program has madvised that an area is MADV_MERGEABLE, and
704 root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set).
706 config DEFAULT_MMAP_MIN_ADDR
707 int "Low address space to protect from user allocation"
711 This is the portion of low virtual memory which should be protected
712 from userspace allocation. Keeping a user from writing to low pages
713 can help reduce the impact of kernel NULL pointer bugs.
715 For most ia64, ppc64 and x86 users with lots of address space
716 a value of 65536 is reasonable and should cause no problems.
717 On arm and other archs it should not be higher than 32768.
718 Programs which use vm86 functionality or have some need to map
719 this low address space will need CAP_SYS_RAWIO or disable this
720 protection by setting the value to 0.
722 This value can be changed after boot using the
723 /proc/sys/vm/mmap_min_addr tunable.
725 config ARCH_SUPPORTS_MEMORY_FAILURE
728 config MEMORY_FAILURE
730 depends on ARCH_SUPPORTS_MEMORY_FAILURE
731 bool "Enable recovery from hardware memory errors"
732 select MEMORY_ISOLATION
735 Enables code to recover from some memory failures on systems
736 with MCA recovery. This allows a system to continue running
737 even when some of its memory has uncorrected errors. This requires
738 special hardware support and typically ECC memory.
740 config HWPOISON_INJECT
741 tristate "HWPoison pages injector"
742 depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS
743 select PROC_PAGE_MONITOR
745 config NOMMU_INITIAL_TRIM_EXCESS
746 int "Turn on mmap() excess space trimming before booting"
750 The NOMMU mmap() frequently needs to allocate large contiguous chunks
751 of memory on which to store mappings, but it can only ask the system
752 allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently
753 more than it requires. To deal with this, mmap() is able to trim off
754 the excess and return it to the allocator.
756 If trimming is enabled, the excess is trimmed off and returned to the
757 system allocator, which can cause extra fragmentation, particularly
758 if there are a lot of transient processes.
760 If trimming is disabled, the excess is kept, but not used, which for
761 long-term mappings means that the space is wasted.
763 Trimming can be dynamically controlled through a sysctl option
764 (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of
765 excess pages there must be before trimming should occur, or zero if
766 no trimming is to occur.
768 This option specifies the initial value of this option. The default
769 of 1 says that all excess pages should be trimmed.
771 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
773 config ARCH_WANT_GENERAL_HUGETLB
776 config ARCH_WANTS_THP_SWAP
779 menuconfig TRANSPARENT_HUGEPAGE
780 bool "Transparent Hugepage Support"
781 depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT
785 Transparent Hugepages allows the kernel to use huge pages and
786 huge tlb transparently to the applications whenever possible.
787 This feature can improve computing performance to certain
788 applications by speeding up page faults during memory
789 allocation, by reducing the number of tlb misses and by speeding
790 up the pagetable walking.
792 If memory constrained on embedded, you may want to say N.
794 if TRANSPARENT_HUGEPAGE
797 prompt "Transparent Hugepage Support sysfs defaults"
798 depends on TRANSPARENT_HUGEPAGE
799 default TRANSPARENT_HUGEPAGE_ALWAYS
801 Selects the sysfs defaults for Transparent Hugepage Support.
803 config TRANSPARENT_HUGEPAGE_ALWAYS
806 Enabling Transparent Hugepage always, can increase the
807 memory footprint of applications without a guaranteed
808 benefit but it will work automatically for all applications.
810 config TRANSPARENT_HUGEPAGE_MADVISE
813 Enabling Transparent Hugepage madvise, will only provide a
814 performance improvement benefit to the applications using
815 madvise(MADV_HUGEPAGE) but it won't risk to increase the
816 memory footprint of applications without a guaranteed
822 depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP && 64BIT
824 Swap transparent huge pages in one piece, without splitting.
825 XXX: For now, swap cluster backing transparent huge page
826 will be split after swapout.
828 For selection by architectures with reasonable THP sizes.
830 config READ_ONLY_THP_FOR_FS
831 bool "Read-only THP for filesystems (EXPERIMENTAL)"
832 depends on TRANSPARENT_HUGEPAGE && SHMEM
835 Allow khugepaged to put read-only file-backed pages in THP.
837 This is marked experimental because it is a new feature. Write
838 support of file THPs will be developed in the next few release
841 endif # TRANSPARENT_HUGEPAGE
844 # UP and nommu archs use km based percpu allocator
846 config NEED_PER_CPU_KM
847 depends on !SMP || !MMU
851 config NEED_PER_CPU_EMBED_FIRST_CHUNK
854 config NEED_PER_CPU_PAGE_FIRST_CHUNK
857 config USE_PERCPU_NUMA_NODE_ID
860 config HAVE_SETUP_PER_CPU_AREA
867 bool "Contiguous Memory Allocator"
870 select MEMORY_ISOLATION
872 This enables the Contiguous Memory Allocator which allows other
873 subsystems to allocate big physically-contiguous blocks of memory.
874 CMA reserves a region of memory and allows only movable pages to
875 be allocated from it. This way, the kernel can use the memory for
876 pagecache and when a subsystem requests for contiguous area, the
877 allocated pages are migrated away to serve the contiguous request.
882 bool "CMA debug messages (DEVELOPMENT)"
883 depends on DEBUG_KERNEL && CMA
885 Turns on debug messages in CMA. This produces KERN_DEBUG
886 messages for every CMA call as well as various messages while
887 processing calls such as dma_alloc_from_contiguous().
888 This option does not affect warning and error messages.
891 bool "CMA debugfs interface"
892 depends on CMA && DEBUG_FS
894 Turns on the DebugFS interface for CMA.
897 bool "CMA information through sysfs interface"
898 depends on CMA && SYSFS
900 This option exposes some sysfs attributes to get information
904 int "Maximum count of the CMA areas"
909 CMA allows to create CMA areas for particular purpose, mainly,
910 used as device private area. This parameter sets the maximum
911 number of CMA area in the system.
913 If unsure, leave the default value "7" in UMA and "19" in NUMA.
915 config MEM_SOFT_DIRTY
916 bool "Track memory changes"
917 depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS
918 select PROC_PAGE_MONITOR
920 This option enables memory changes tracking by introducing a
921 soft-dirty bit on pte-s. This bit it set when someone writes
922 into a page just as regular dirty bit, but unlike the latter
923 it can be cleared by hands.
925 See Documentation/admin-guide/mm/soft-dirty.rst for more details.
927 config GENERIC_EARLY_IOREMAP
930 config STACK_MAX_DEFAULT_SIZE_MB
931 int "Default maximum user stack size for 32-bit processes (MB)"
934 depends on STACK_GROWSUP && (!64BIT || COMPAT)
936 This is the maximum stack size in Megabytes in the VM layout of 32-bit
937 user processes when the stack grows upwards (currently only on parisc
938 arch) when the RLIMIT_STACK hard limit is unlimited.
940 A sane initial value is 100 MB.
942 config DEFERRED_STRUCT_PAGE_INIT
943 bool "Defer initialisation of struct pages to kthreads"
945 depends on !NEED_PER_CPU_KM
949 Ordinarily all struct pages are initialised during early boot in a
950 single thread. On very large machines this can take a considerable
951 amount of time. If this option is set, large machines will bring up
952 a subset of memmap at boot and then initialise the rest in parallel.
953 This has a potential performance impact on tasks running early in the
954 lifetime of the system until these kthreads finish the
957 config PAGE_IDLE_FLAG
959 select PAGE_EXTENSION if !64BIT
961 This adds PG_idle and PG_young flags to 'struct page'. PTE Accessed
962 bit writers can set the state of the bit in the flags so that PTE
963 Accessed bit readers may avoid disturbance.
965 config IDLE_PAGE_TRACKING
966 bool "Enable idle page tracking"
967 depends on SYSFS && MMU
968 select PAGE_IDLE_FLAG
970 This feature allows to estimate the amount of user pages that have
971 not been touched during a given period of time. This information can
972 be useful to tune memory cgroup limits and/or for job placement
973 within a compute cluster.
975 See Documentation/admin-guide/mm/idle_page_tracking.rst for
978 config ARCH_HAS_CACHE_LINE_SIZE
981 config ARCH_HAS_CURRENT_STACK_POINTER
984 In support of HARDENED_USERCOPY performing stack variable lifetime
985 checking, an architecture-agnostic way to find the stack pointer
986 is needed. Once an architecture defines an unsigned long global
987 register alias named "current_stack_pointer", this config can be
990 config ARCH_HAS_PTE_DEVMAP
993 config ARCH_HAS_ZONE_DMA_SET
997 bool "Support DMA zone" if ARCH_HAS_ZONE_DMA_SET
998 default y if ARM64 || X86
1001 bool "Support DMA32 zone" if ARCH_HAS_ZONE_DMA_SET
1006 bool "Device memory (pmem, HMM, etc...) hotplug support"
1007 depends on MEMORY_HOTPLUG
1008 depends on MEMORY_HOTREMOVE
1009 depends on SPARSEMEM_VMEMMAP
1010 depends on ARCH_HAS_PTE_DEVMAP
1014 Device memory hotplug support allows for establishing pmem,
1015 or other device driver discovered memory regions, in the
1016 memmap. This allows pfn_to_page() lookups of otherwise
1017 "device-physical" addresses which is needed for using a DAX
1018 mapping in an O_DIRECT operation, among other things.
1020 If FS_DAX is enabled, then say Y.
1023 # Helpers to mirror range of the CPU page tables of a process into device page
1030 config GET_FREE_REGION
1031 depends on SPARSEMEM
1034 config DEVICE_PRIVATE
1035 bool "Unaddressable device memory (GPU memory, ...)"
1036 depends on ZONE_DEVICE
1037 select GET_FREE_REGION
1040 Allows creation of struct pages to represent unaddressable device
1041 memory; i.e., memory that is only accessible from the device (or
1042 group of devices). You likely also want to select HMM_MIRROR.
1047 config ARCH_USES_HIGH_VMA_FLAGS
1049 config ARCH_HAS_PKEYS
1052 config ARCH_USES_PG_ARCH_X
1055 Enable the definition of PG_arch_x page flags with x > 1. Only
1056 suitable for 64-bit architectures with CONFIG_FLATMEM or
1057 CONFIG_SPARSEMEM_VMEMMAP enabled, otherwise there may not be
1058 enough room for additional bits in page->flags.
1060 config VM_EVENT_COUNTERS
1062 bool "Enable VM event counters for /proc/vmstat" if EXPERT
1064 VM event counters are needed for event counts to be shown.
1065 This option allows the disabling of the VM event counters
1066 on EXPERT systems. /proc/vmstat will only show page counts
1067 if VM event counters are disabled.
1070 bool "Collect percpu memory statistics"
1072 This feature collects and exposes statistics via debugfs. The
1073 information includes global and per chunk statistics, which can
1074 be used to help understand percpu memory usage.
1077 bool "Enable infrastructure for get_user_pages()-related unit tests"
1080 Provides /sys/kernel/debug/gup_test, which in turn provides a way
1081 to make ioctl calls that can launch kernel-based unit tests for
1082 the get_user_pages*() and pin_user_pages*() family of API calls.
1084 These tests include benchmark testing of the _fast variants of
1085 get_user_pages*() and pin_user_pages*(), as well as smoke tests of
1086 the non-_fast variants.
1088 There is also a sub-test that allows running dump_page() on any
1089 of up to eight pages (selected by command line args) within the
1090 range of user-space addresses. These pages are either pinned via
1091 pin_user_pages*(), or pinned via get_user_pages*(), as specified
1092 by other command line arguments.
1094 See tools/testing/selftests/mm/gup_test.c
1096 comment "GUP_TEST needs to have DEBUG_FS enabled"
1097 depends on !GUP_TEST && !DEBUG_FS
1099 config GUP_GET_PXX_LOW_HIGH
1103 tristate "Enable a module to run time tests on dma_pool"
1106 Provides a test module that will allocate and free many blocks of
1107 various sizes and report how long it takes. This is intended to
1108 provide a consistent way to measure how changes to the
1109 dma_pool_alloc/free routines affect performance.
1111 config ARCH_HAS_PTE_SPECIAL
1115 # Some architectures require a special hugepage directory format that is
1116 # required to support multiple hugepage sizes. For example a4fe3ce76
1117 # "powerpc/mm: Allow more flexible layouts for hugepage pagetables"
1118 # introduced it on powerpc. This allows for a more flexible hugepage
1119 # pagetable layouts.
1121 config ARCH_HAS_HUGEPD
1124 config MAPPING_DIRTY_HELPERS
1130 config KMAP_LOCAL_NON_LINEAR_PTE_ARRAY
1133 # struct io_mapping based helper. Selected by drivers that need them
1139 bool "Enable memfd_secret() system call" if EXPERT
1140 depends on ARCH_HAS_SET_DIRECT_MAP
1142 Enable the memfd_secret() system call with the ability to create
1143 memory areas visible only in the context of the owning process and
1144 not mapped to other processes and other kernel page tables.
1146 config ANON_VMA_NAME
1147 bool "Anonymous VMA name support"
1148 depends on PROC_FS && ADVISE_SYSCALLS && MMU
1151 Allow naming anonymous virtual memory areas.
1153 This feature allows assigning names to virtual memory areas. Assigned
1154 names can be later retrieved from /proc/pid/maps and /proc/pid/smaps
1155 and help identifying individual anonymous memory areas.
1156 Assigning a name to anonymous virtual memory area might prevent that
1157 area from being merged with adjacent virtual memory areas due to the
1158 difference in their name.
1161 bool "Enable userfaultfd() system call"
1164 Enable the userfaultfd() system call that allows to intercept and
1165 handle page faults in userland.
1167 config HAVE_ARCH_USERFAULTFD_WP
1170 Arch has userfaultfd write protection support
1172 config HAVE_ARCH_USERFAULTFD_MINOR
1175 Arch has userfaultfd minor fault support
1177 config PTE_MARKER_UFFD_WP
1178 bool "Userfaultfd write protection support for shmem/hugetlbfs"
1180 depends on HAVE_ARCH_USERFAULTFD_WP
1183 Allows to create marker PTEs for userfaultfd write protection
1184 purposes. It is required to enable userfaultfd write protection on
1185 file-backed memory types like shmem and hugetlbfs.
1189 bool "Multi-Gen LRU"
1191 # make sure folio->flags has enough spare bits
1192 depends on 64BIT || !SPARSEMEM || SPARSEMEM_VMEMMAP
1194 A high performance LRU implementation to overcommit memory. See
1195 Documentation/admin-guide/mm/multigen_lru.rst for details.
1197 config LRU_GEN_ENABLED
1198 bool "Enable by default"
1201 This option enables the multi-gen LRU by default.
1203 config LRU_GEN_STATS
1204 bool "Full stats for debugging"
1207 Do not enable this option unless you plan to look at historical stats
1208 from evicted generations for debugging purpose.
1210 This option has a per-memcg and per-node memory overhead.
1213 config ARCH_SUPPORTS_PER_VMA_LOCK
1218 depends on ARCH_SUPPORTS_PER_VMA_LOCK && MMU && SMP
1220 Allow per-vma locking during page fault handling.
1222 This feature allows locking each virtual memory area separately when
1223 handling page faults instead of taking mmap_lock.
1225 source "mm/damon/Kconfig"