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 (EXPERIMENTAL)"
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 that swap device
37 reads, can also improve workload performance.
39 This is marked experimental because it is a new feature (as of
40 v3.11) that interacts heavily with memory reclaim. While these
41 interactions don't cause any known issues on simple memory setups,
42 they have not be fully explored on the large set of potential
43 configurations and workloads that exist.
45 config ZSWAP_DEFAULT_ON
46 bool "Enable the compressed cache for swap pages by default"
49 If selected, the compressed cache for swap pages will be enabled
50 at boot, otherwise it will be disabled.
52 The selection made here can be overridden by using the kernel
53 command line 'zswap.enabled=' option.
56 prompt "Default compressor"
58 default ZSWAP_COMPRESSOR_DEFAULT_LZO
60 Selects the default compression algorithm for the compressed cache
63 For an overview what kind of performance can be expected from
64 a particular compression algorithm please refer to the benchmarks
65 available at the following LWN page:
66 https://lwn.net/Articles/751795/
68 If in doubt, select 'LZO'.
70 The selection made here can be overridden by using the kernel
71 command line 'zswap.compressor=' option.
73 config ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
77 Use the Deflate algorithm as the default compression algorithm.
79 config ZSWAP_COMPRESSOR_DEFAULT_LZO
83 Use the LZO algorithm as the default compression algorithm.
85 config ZSWAP_COMPRESSOR_DEFAULT_842
89 Use the 842 algorithm as the default compression algorithm.
91 config ZSWAP_COMPRESSOR_DEFAULT_LZ4
95 Use the LZ4 algorithm as the default compression algorithm.
97 config ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
101 Use the LZ4HC algorithm as the default compression algorithm.
103 config ZSWAP_COMPRESSOR_DEFAULT_ZSTD
107 Use the zstd algorithm as the default compression algorithm.
110 config ZSWAP_COMPRESSOR_DEFAULT
113 default "deflate" if ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
114 default "lzo" if ZSWAP_COMPRESSOR_DEFAULT_LZO
115 default "842" if ZSWAP_COMPRESSOR_DEFAULT_842
116 default "lz4" if ZSWAP_COMPRESSOR_DEFAULT_LZ4
117 default "lz4hc" if ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
118 default "zstd" if ZSWAP_COMPRESSOR_DEFAULT_ZSTD
122 prompt "Default allocator"
124 default ZSWAP_ZPOOL_DEFAULT_ZBUD
126 Selects the default allocator for the compressed cache for
128 The default is 'zbud' for compatibility, however please do
129 read the description of each of the allocators below before
130 making a right choice.
132 The selection made here can be overridden by using the kernel
133 command line 'zswap.zpool=' option.
135 config ZSWAP_ZPOOL_DEFAULT_ZBUD
139 Use the zbud allocator as the default allocator.
141 config ZSWAP_ZPOOL_DEFAULT_Z3FOLD
145 Use the z3fold allocator as the default allocator.
147 config ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
151 Use the zsmalloc allocator as the default allocator.
154 config ZSWAP_ZPOOL_DEFAULT
157 default "zbud" if ZSWAP_ZPOOL_DEFAULT_ZBUD
158 default "z3fold" if ZSWAP_ZPOOL_DEFAULT_Z3FOLD
159 default "zsmalloc" if ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
163 tristate "2:1 compression allocator (zbud)"
166 A special purpose allocator for storing compressed pages.
167 It is designed to store up to two compressed pages per physical
168 page. While this design limits storage density, it has simple and
169 deterministic reclaim properties that make it preferable to a higher
170 density approach when reclaim will be used.
173 tristate "3:1 compression allocator (z3fold)"
176 A special purpose allocator for storing compressed pages.
177 It is designed to store up to three compressed pages per physical
178 page. It is a ZBUD derivative so the simplicity and determinism are
183 prompt "N:1 compression allocator (zsmalloc)" if ZSWAP
186 zsmalloc is a slab-based memory allocator designed to store
187 pages of various compression levels efficiently. It achieves
188 the highest storage density with the least amount of fragmentation.
191 bool "Export zsmalloc statistics"
195 This option enables code in the zsmalloc to collect various
196 statistics about what's happening in zsmalloc and exports that
197 information to userspace via debugfs.
200 menu "SLAB allocator options"
203 prompt "Choose SLAB allocator"
206 This option allows to select a slab allocator.
210 depends on !PREEMPT_RT
211 select HAVE_HARDENED_USERCOPY_ALLOCATOR
213 The regular slab allocator that is established and known to work
214 well in all environments. It organizes cache hot objects in
215 per cpu and per node queues.
218 bool "SLUB (Unqueued Allocator)"
219 select HAVE_HARDENED_USERCOPY_ALLOCATOR
221 SLUB is a slab allocator that minimizes cache line usage
222 instead of managing queues of cached objects (SLAB approach).
223 Per cpu caching is realized using slabs of objects instead
224 of queues of objects. SLUB can use memory efficiently
225 and has enhanced diagnostics. SLUB is the default choice for
230 bool "SLOB (Simple Allocator)"
231 depends on !PREEMPT_RT
233 SLOB replaces the stock allocator with a drastically simpler
234 allocator. SLOB is generally more space efficient but
235 does not perform as well on large systems.
239 config SLAB_MERGE_DEFAULT
240 bool "Allow slab caches to be merged"
242 depends on SLAB || SLUB
244 For reduced kernel memory fragmentation, slab caches can be
245 merged when they share the same size and other characteristics.
246 This carries a risk of kernel heap overflows being able to
247 overwrite objects from merged caches (and more easily control
248 cache layout), which makes such heap attacks easier to exploit
249 by attackers. By keeping caches unmerged, these kinds of exploits
250 can usually only damage objects in the same cache. To disable
251 merging at runtime, "slab_nomerge" can be passed on the kernel
254 config SLAB_FREELIST_RANDOM
255 bool "Randomize slab freelist"
256 depends on SLAB || SLUB
258 Randomizes the freelist order used on creating new pages. This
259 security feature reduces the predictability of the kernel slab
260 allocator against heap overflows.
262 config SLAB_FREELIST_HARDENED
263 bool "Harden slab freelist metadata"
264 depends on SLAB || SLUB
266 Many kernel heap attacks try to target slab cache metadata and
267 other infrastructure. This options makes minor performance
268 sacrifices to harden the kernel slab allocator against common
269 freelist exploit methods. Some slab implementations have more
270 sanity-checking than others. This option is most effective with
275 bool "Enable SLUB performance statistics"
276 depends on SLUB && SYSFS
278 SLUB statistics are useful to debug SLUBs allocation behavior in
279 order find ways to optimize the allocator. This should never be
280 enabled for production use since keeping statistics slows down
281 the allocator by a few percentage points. The slabinfo command
282 supports the determination of the most active slabs to figure
283 out which slabs are relevant to a particular load.
284 Try running: slabinfo -DA
286 config SLUB_CPU_PARTIAL
288 depends on SLUB && SMP
289 bool "SLUB per cpu partial cache"
291 Per cpu partial caches accelerate objects allocation and freeing
292 that is local to a processor at the price of more indeterminism
293 in the latency of the free. On overflow these caches will be cleared
294 which requires the taking of locks that may cause latency spikes.
295 Typically one would choose no for a realtime system.
297 endmenu # SLAB allocator options
299 config SHUFFLE_PAGE_ALLOCATOR
300 bool "Page allocator randomization"
301 default SLAB_FREELIST_RANDOM && ACPI_NUMA
303 Randomization of the page allocator improves the average
304 utilization of a direct-mapped memory-side-cache. See section
305 5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI
306 6.2a specification for an example of how a platform advertises
307 the presence of a memory-side-cache. There are also incidental
308 security benefits as it reduces the predictability of page
309 allocations to compliment SLAB_FREELIST_RANDOM, but the
310 default granularity of shuffling on the "MAX_ORDER - 1" i.e,
311 10th order of pages is selected based on cache utilization
314 While the randomization improves cache utilization it may
315 negatively impact workloads on platforms without a cache. For
316 this reason, by default, the randomization is enabled only
317 after runtime detection of a direct-mapped memory-side-cache.
318 Otherwise, the randomization may be force enabled with the
319 'page_alloc.shuffle' kernel command line parameter.
324 bool "Disable heap randomization"
327 Randomizing heap placement makes heap exploits harder, but it
328 also breaks ancient binaries (including anything libc5 based).
329 This option changes the bootup default to heap randomization
330 disabled, and can be overridden at runtime by setting
331 /proc/sys/kernel/randomize_va_space to 2.
333 On non-ancient distros (post-2000 ones) N is usually a safe choice.
335 config MMAP_ALLOW_UNINITIALIZED
336 bool "Allow mmapped anonymous memory to be uninitialized"
337 depends on EXPERT && !MMU
340 Normally, and according to the Linux spec, anonymous memory obtained
341 from mmap() has its contents cleared before it is passed to
342 userspace. Enabling this config option allows you to request that
343 mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus
344 providing a huge performance boost. If this option is not enabled,
345 then the flag will be ignored.
347 This is taken advantage of by uClibc's malloc(), and also by
348 ELF-FDPIC binfmt's brk and stack allocator.
350 Because of the obvious security issues, this option should only be
351 enabled on embedded devices where you control what is run in
352 userspace. Since that isn't generally a problem on no-MMU systems,
353 it is normally safe to say Y here.
355 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
357 config SELECT_MEMORY_MODEL
359 depends on ARCH_SELECT_MEMORY_MODEL
362 prompt "Memory model"
363 depends on SELECT_MEMORY_MODEL
364 default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT
365 default FLATMEM_MANUAL
367 This option allows you to change some of the ways that
368 Linux manages its memory internally. Most users will
369 only have one option here selected by the architecture
370 configuration. This is normal.
372 config FLATMEM_MANUAL
374 depends on !ARCH_SPARSEMEM_ENABLE || ARCH_FLATMEM_ENABLE
376 This option is best suited for non-NUMA systems with
377 flat address space. The FLATMEM is the most efficient
378 system in terms of performance and resource consumption
379 and it is the best option for smaller systems.
381 For systems that have holes in their physical address
382 spaces and for features like NUMA and memory hotplug,
383 choose "Sparse Memory".
385 If unsure, choose this option (Flat Memory) over any other.
387 config SPARSEMEM_MANUAL
389 depends on ARCH_SPARSEMEM_ENABLE
391 This will be the only option for some systems, including
392 memory hot-plug systems. This is normal.
394 This option provides efficient support for systems with
395 holes is their physical address space and allows memory
396 hot-plug and hot-remove.
398 If unsure, choose "Flat Memory" over this option.
404 depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL
408 depends on !SPARSEMEM || FLATMEM_MANUAL
411 # SPARSEMEM_EXTREME (which is the default) does some bootmem
412 # allocations when sparse_init() is called. If this cannot
413 # be done on your architecture, select this option. However,
414 # statically allocating the mem_section[] array can potentially
415 # consume vast quantities of .bss, so be careful.
417 # This option will also potentially produce smaller runtime code
418 # with gcc 3.4 and later.
420 config SPARSEMEM_STATIC
424 # Architecture platforms which require a two level mem_section in SPARSEMEM
425 # must select this option. This is usually for architecture platforms with
426 # an extremely sparse physical address space.
428 config SPARSEMEM_EXTREME
430 depends on SPARSEMEM && !SPARSEMEM_STATIC
432 config SPARSEMEM_VMEMMAP_ENABLE
435 config SPARSEMEM_VMEMMAP
436 bool "Sparse Memory virtual memmap"
437 depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE
440 SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise
441 pfn_to_page and page_to_pfn operations. This is the most
442 efficient option when sufficient kernel resources are available.
444 config HAVE_MEMBLOCK_PHYS_MAP
451 # Don't discard allocated memory used to track "memory" and "reserved" memblocks
452 # after early boot, so it can still be used to test for validity of memory.
453 # Also, memblocks are updated with memory hot(un)plug.
454 config ARCH_KEEP_MEMBLOCK
457 # Keep arch NUMA mapping infrastructure post-init.
458 config NUMA_KEEP_MEMINFO
461 config MEMORY_ISOLATION
464 # IORESOURCE_SYSTEM_RAM regions in the kernel resource tree that are marked
465 # IORESOURCE_EXCLUSIVE cannot be mapped to user space, for example, via
467 config EXCLUSIVE_SYSTEM_RAM
469 depends on !DEVMEM || STRICT_DEVMEM
472 # Only be set on architectures that have completely implemented memory hotplug
473 # feature. If you are not sure, don't touch it.
475 config HAVE_BOOTMEM_INFO_NODE
478 config ARCH_ENABLE_MEMORY_HOTPLUG
481 config ARCH_ENABLE_MEMORY_HOTREMOVE
484 # eventually, we can have this option just 'select SPARSEMEM'
485 menuconfig MEMORY_HOTPLUG
486 bool "Memory hotplug"
487 select MEMORY_ISOLATION
489 depends on ARCH_ENABLE_MEMORY_HOTPLUG
491 select NUMA_KEEP_MEMINFO if NUMA
495 config MEMORY_HOTPLUG_DEFAULT_ONLINE
496 bool "Online the newly added memory blocks by default"
497 depends on MEMORY_HOTPLUG
499 This option sets the default policy setting for memory hotplug
500 onlining policy (/sys/devices/system/memory/auto_online_blocks) which
501 determines what happens to newly added memory regions. Policy setting
502 can always be changed at runtime.
503 See Documentation/admin-guide/mm/memory-hotplug.rst for more information.
505 Say Y here if you want all hot-plugged memory blocks to appear in
506 'online' state by default.
507 Say N here if you want the default policy to keep all hot-plugged
508 memory blocks in 'offline' state.
510 config MEMORY_HOTREMOVE
511 bool "Allow for memory hot remove"
512 select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64)
513 depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE
516 config MHP_MEMMAP_ON_MEMORY
518 depends on MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP
519 depends on ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
521 endif # MEMORY_HOTPLUG
523 # Heavily threaded applications may benefit from splitting the mm-wide
524 # page_table_lock, so that faults on different parts of the user address
525 # space can be handled with less contention: split it at this NR_CPUS.
526 # Default to 4 for wider testing, though 8 might be more appropriate.
527 # ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
528 # PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
529 # SPARC32 allocates multiple pte tables within a single page, and therefore
530 # a per-page lock leads to problems when multiple tables need to be locked
531 # at the same time (e.g. copy_page_range()).
532 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
534 config SPLIT_PTLOCK_CPUS
536 default "999999" if !MMU
537 default "999999" if ARM && !CPU_CACHE_VIPT
538 default "999999" if PARISC && !PA20
539 default "999999" if SPARC32
542 config ARCH_ENABLE_SPLIT_PMD_PTLOCK
546 # support for memory balloon
547 config MEMORY_BALLOON
551 # support for memory balloon compaction
552 config BALLOON_COMPACTION
553 bool "Allow for balloon memory compaction/migration"
555 depends on COMPACTION && MEMORY_BALLOON
557 Memory fragmentation introduced by ballooning might reduce
558 significantly the number of 2MB contiguous memory blocks that can be
559 used within a guest, thus imposing performance penalties associated
560 with the reduced number of transparent huge pages that could be used
561 by the guest workload. Allowing the compaction & migration for memory
562 pages enlisted as being part of memory balloon devices avoids the
563 scenario aforementioned and helps improving memory defragmentation.
566 # support for memory compaction
568 bool "Allow for memory compaction"
573 Compaction is the only memory management component to form
574 high order (larger physically contiguous) memory blocks
575 reliably. The page allocator relies on compaction heavily and
576 the lack of the feature can lead to unexpected OOM killer
577 invocations for high order memory requests. You shouldn't
578 disable this option unless there really is a strong reason for
579 it and then we would be really interested to hear about that at
583 # support for free page reporting
584 config PAGE_REPORTING
585 bool "Free page reporting"
588 Free page reporting allows for the incremental acquisition of
589 free pages from the buddy allocator for the purpose of reporting
590 those pages to another entity, such as a hypervisor, so that the
591 memory can be freed within the host for other uses.
594 # support for page migration
597 bool "Page migration"
599 depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU
601 Allows the migration of the physical location of pages of processes
602 while the virtual addresses are not changed. This is useful in
603 two situations. The first is on NUMA systems to put pages nearer
604 to the processors accessing. The second is when allocating huge
605 pages as migration can relocate pages to satisfy a huge page
606 allocation instead of reclaiming.
608 config DEVICE_MIGRATION
609 def_bool MIGRATION && ZONE_DEVICE
611 config ARCH_ENABLE_HUGEPAGE_MIGRATION
614 config ARCH_ENABLE_THP_MIGRATION
617 config HUGETLB_PAGE_SIZE_VARIABLE
620 Allows the pageblock_order value to be dynamic instead of just standard
621 HUGETLB_PAGE_ORDER when there are multiple HugeTLB page sizes available
624 Note that the pageblock_order cannot exceed MAX_ORDER - 1 and will be
625 clamped down to MAX_ORDER - 1.
628 def_bool (MEMORY_ISOLATION && COMPACTION) || CMA
630 config PHYS_ADDR_T_64BIT
634 bool "Enable bounce buffers"
636 depends on BLOCK && MMU && HIGHMEM
638 Enable bounce buffers for devices that cannot access the full range of
639 memory available to the CPU. Enabled by default when HIGHMEM is
640 selected, but you may say n to override this.
645 An architecture should select this if it implements the
646 deprecated interface virt_to_bus(). All new architectures
647 should probably not select this.
656 bool "Enable KSM for page merging"
660 Enable Kernel Samepage Merging: KSM periodically scans those areas
661 of an application's address space that an app has advised may be
662 mergeable. When it finds pages of identical content, it replaces
663 the many instances by a single page with that content, so
664 saving memory until one or another app needs to modify the content.
665 Recommended for use with KVM, or with other duplicative applications.
666 See Documentation/vm/ksm.rst for more information: KSM is inactive
667 until a program has madvised that an area is MADV_MERGEABLE, and
668 root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set).
670 config DEFAULT_MMAP_MIN_ADDR
671 int "Low address space to protect from user allocation"
675 This is the portion of low virtual memory which should be protected
676 from userspace allocation. Keeping a user from writing to low pages
677 can help reduce the impact of kernel NULL pointer bugs.
679 For most ia64, ppc64 and x86 users with lots of address space
680 a value of 65536 is reasonable and should cause no problems.
681 On arm and other archs it should not be higher than 32768.
682 Programs which use vm86 functionality or have some need to map
683 this low address space will need CAP_SYS_RAWIO or disable this
684 protection by setting the value to 0.
686 This value can be changed after boot using the
687 /proc/sys/vm/mmap_min_addr tunable.
689 config ARCH_SUPPORTS_MEMORY_FAILURE
692 config MEMORY_FAILURE
694 depends on ARCH_SUPPORTS_MEMORY_FAILURE
695 bool "Enable recovery from hardware memory errors"
696 select MEMORY_ISOLATION
699 Enables code to recover from some memory failures on systems
700 with MCA recovery. This allows a system to continue running
701 even when some of its memory has uncorrected errors. This requires
702 special hardware support and typically ECC memory.
704 config HWPOISON_INJECT
705 tristate "HWPoison pages injector"
706 depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS
707 select PROC_PAGE_MONITOR
709 config NOMMU_INITIAL_TRIM_EXCESS
710 int "Turn on mmap() excess space trimming before booting"
714 The NOMMU mmap() frequently needs to allocate large contiguous chunks
715 of memory on which to store mappings, but it can only ask the system
716 allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently
717 more than it requires. To deal with this, mmap() is able to trim off
718 the excess and return it to the allocator.
720 If trimming is enabled, the excess is trimmed off and returned to the
721 system allocator, which can cause extra fragmentation, particularly
722 if there are a lot of transient processes.
724 If trimming is disabled, the excess is kept, but not used, which for
725 long-term mappings means that the space is wasted.
727 Trimming can be dynamically controlled through a sysctl option
728 (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of
729 excess pages there must be before trimming should occur, or zero if
730 no trimming is to occur.
732 This option specifies the initial value of this option. The default
733 of 1 says that all excess pages should be trimmed.
735 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
737 config ARCH_WANT_GENERAL_HUGETLB
740 config ARCH_WANTS_THP_SWAP
743 menuconfig TRANSPARENT_HUGEPAGE
744 bool "Transparent Hugepage Support"
745 depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT
749 Transparent Hugepages allows the kernel to use huge pages and
750 huge tlb transparently to the applications whenever possible.
751 This feature can improve computing performance to certain
752 applications by speeding up page faults during memory
753 allocation, by reducing the number of tlb misses and by speeding
754 up the pagetable walking.
756 If memory constrained on embedded, you may want to say N.
758 if TRANSPARENT_HUGEPAGE
761 prompt "Transparent Hugepage Support sysfs defaults"
762 depends on TRANSPARENT_HUGEPAGE
763 default TRANSPARENT_HUGEPAGE_ALWAYS
765 Selects the sysfs defaults for Transparent Hugepage Support.
767 config TRANSPARENT_HUGEPAGE_ALWAYS
770 Enabling Transparent Hugepage always, can increase the
771 memory footprint of applications without a guaranteed
772 benefit but it will work automatically for all applications.
774 config TRANSPARENT_HUGEPAGE_MADVISE
777 Enabling Transparent Hugepage madvise, will only provide a
778 performance improvement benefit to the applications using
779 madvise(MADV_HUGEPAGE) but it won't risk to increase the
780 memory footprint of applications without a guaranteed
786 depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP
788 Swap transparent huge pages in one piece, without splitting.
789 XXX: For now, swap cluster backing transparent huge page
790 will be split after swapout.
792 For selection by architectures with reasonable THP sizes.
794 config READ_ONLY_THP_FOR_FS
795 bool "Read-only THP for filesystems (EXPERIMENTAL)"
796 depends on TRANSPARENT_HUGEPAGE && SHMEM
799 Allow khugepaged to put read-only file-backed pages in THP.
801 This is marked experimental because it is a new feature. Write
802 support of file THPs will be developed in the next few release
805 endif # TRANSPARENT_HUGEPAGE
808 # UP and nommu archs use km based percpu allocator
810 config NEED_PER_CPU_KM
811 depends on !SMP || !MMU
815 config NEED_PER_CPU_EMBED_FIRST_CHUNK
818 config NEED_PER_CPU_PAGE_FIRST_CHUNK
821 config USE_PERCPU_NUMA_NODE_ID
824 config HAVE_SETUP_PER_CPU_AREA
831 bool "Contiguous Memory Allocator"
834 select MEMORY_ISOLATION
836 This enables the Contiguous Memory Allocator which allows other
837 subsystems to allocate big physically-contiguous blocks of memory.
838 CMA reserves a region of memory and allows only movable pages to
839 be allocated from it. This way, the kernel can use the memory for
840 pagecache and when a subsystem requests for contiguous area, the
841 allocated pages are migrated away to serve the contiguous request.
846 bool "CMA debug messages (DEVELOPMENT)"
847 depends on DEBUG_KERNEL && CMA
849 Turns on debug messages in CMA. This produces KERN_DEBUG
850 messages for every CMA call as well as various messages while
851 processing calls such as dma_alloc_from_contiguous().
852 This option does not affect warning and error messages.
855 bool "CMA debugfs interface"
856 depends on CMA && DEBUG_FS
858 Turns on the DebugFS interface for CMA.
861 bool "CMA information through sysfs interface"
862 depends on CMA && SYSFS
864 This option exposes some sysfs attributes to get information
868 int "Maximum count of the CMA areas"
873 CMA allows to create CMA areas for particular purpose, mainly,
874 used as device private area. This parameter sets the maximum
875 number of CMA area in the system.
877 If unsure, leave the default value "7" in UMA and "19" in NUMA.
879 config MEM_SOFT_DIRTY
880 bool "Track memory changes"
881 depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS
882 select PROC_PAGE_MONITOR
884 This option enables memory changes tracking by introducing a
885 soft-dirty bit on pte-s. This bit it set when someone writes
886 into a page just as regular dirty bit, but unlike the latter
887 it can be cleared by hands.
889 See Documentation/admin-guide/mm/soft-dirty.rst for more details.
891 config GENERIC_EARLY_IOREMAP
894 config STACK_MAX_DEFAULT_SIZE_MB
895 int "Default maximum user stack size for 32-bit processes (MB)"
898 depends on STACK_GROWSUP && (!64BIT || COMPAT)
900 This is the maximum stack size in Megabytes in the VM layout of 32-bit
901 user processes when the stack grows upwards (currently only on parisc
902 arch) when the RLIMIT_STACK hard limit is unlimited.
904 A sane initial value is 100 MB.
906 config DEFERRED_STRUCT_PAGE_INIT
907 bool "Defer initialisation of struct pages to kthreads"
909 depends on !NEED_PER_CPU_KM
913 Ordinarily all struct pages are initialised during early boot in a
914 single thread. On very large machines this can take a considerable
915 amount of time. If this option is set, large machines will bring up
916 a subset of memmap at boot and then initialise the rest in parallel.
917 This has a potential performance impact on tasks running early in the
918 lifetime of the system until these kthreads finish the
921 config PAGE_IDLE_FLAG
923 select PAGE_EXTENSION if !64BIT
925 This adds PG_idle and PG_young flags to 'struct page'. PTE Accessed
926 bit writers can set the state of the bit in the flags so that PTE
927 Accessed bit readers may avoid disturbance.
929 config IDLE_PAGE_TRACKING
930 bool "Enable idle page tracking"
931 depends on SYSFS && MMU
932 select PAGE_IDLE_FLAG
934 This feature allows to estimate the amount of user pages that have
935 not been touched during a given period of time. This information can
936 be useful to tune memory cgroup limits and/or for job placement
937 within a compute cluster.
939 See Documentation/admin-guide/mm/idle_page_tracking.rst for
942 config ARCH_HAS_CACHE_LINE_SIZE
945 config ARCH_HAS_CURRENT_STACK_POINTER
948 In support of HARDENED_USERCOPY performing stack variable lifetime
949 checking, an architecture-agnostic way to find the stack pointer
950 is needed. Once an architecture defines an unsigned long global
951 register alias named "current_stack_pointer", this config can be
954 config ARCH_HAS_VM_GET_PAGE_PROT
957 config ARCH_HAS_PTE_DEVMAP
960 config ARCH_HAS_ZONE_DMA_SET
964 bool "Support DMA zone" if ARCH_HAS_ZONE_DMA_SET
965 default y if ARM64 || X86
968 bool "Support DMA32 zone" if ARCH_HAS_ZONE_DMA_SET
973 bool "Device memory (pmem, HMM, etc...) hotplug support"
974 depends on MEMORY_HOTPLUG
975 depends on MEMORY_HOTREMOVE
976 depends on SPARSEMEM_VMEMMAP
977 depends on ARCH_HAS_PTE_DEVMAP
981 Device memory hotplug support allows for establishing pmem,
982 or other device driver discovered memory regions, in the
983 memmap. This allows pfn_to_page() lookups of otherwise
984 "device-physical" addresses which is needed for using a DAX
985 mapping in an O_DIRECT operation, among other things.
987 If FS_DAX is enabled, then say Y.
990 # Helpers to mirror range of the CPU page tables of a process into device page
997 config DEVICE_PRIVATE
998 bool "Unaddressable device memory (GPU memory, ...)"
999 depends on ZONE_DEVICE
1002 Allows creation of struct pages to represent unaddressable device
1003 memory; i.e., memory that is only accessible from the device (or
1004 group of devices). You likely also want to select HMM_MIRROR.
1009 config ARCH_USES_HIGH_VMA_FLAGS
1011 config ARCH_HAS_PKEYS
1014 config VM_EVENT_COUNTERS
1016 bool "Enable VM event counters for /proc/vmstat" if EXPERT
1018 VM event counters are needed for event counts to be shown.
1019 This option allows the disabling of the VM event counters
1020 on EXPERT systems. /proc/vmstat will only show page counts
1021 if VM event counters are disabled.
1024 bool "Collect percpu memory statistics"
1026 This feature collects and exposes statistics via debugfs. The
1027 information includes global and per chunk statistics, which can
1028 be used to help understand percpu memory usage.
1031 bool "Enable infrastructure for get_user_pages()-related unit tests"
1034 Provides /sys/kernel/debug/gup_test, which in turn provides a way
1035 to make ioctl calls that can launch kernel-based unit tests for
1036 the get_user_pages*() and pin_user_pages*() family of API calls.
1038 These tests include benchmark testing of the _fast variants of
1039 get_user_pages*() and pin_user_pages*(), as well as smoke tests of
1040 the non-_fast variants.
1042 There is also a sub-test that allows running dump_page() on any
1043 of up to eight pages (selected by command line args) within the
1044 range of user-space addresses. These pages are either pinned via
1045 pin_user_pages*(), or pinned via get_user_pages*(), as specified
1046 by other command line arguments.
1048 See tools/testing/selftests/vm/gup_test.c
1050 comment "GUP_TEST needs to have DEBUG_FS enabled"
1051 depends on !GUP_TEST && !DEBUG_FS
1053 config GUP_GET_PTE_LOW_HIGH
1056 config ARCH_HAS_PTE_SPECIAL
1060 # Some architectures require a special hugepage directory format that is
1061 # required to support multiple hugepage sizes. For example a4fe3ce76
1062 # "powerpc/mm: Allow more flexible layouts for hugepage pagetables"
1063 # introduced it on powerpc. This allows for a more flexible hugepage
1064 # pagetable layouts.
1066 config ARCH_HAS_HUGEPD
1069 config MAPPING_DIRTY_HELPERS
1075 config KMAP_LOCAL_NON_LINEAR_PTE_ARRAY
1078 # struct io_mapping based helper. Selected by drivers that need them
1083 def_bool ARCH_HAS_SET_DIRECT_MAP && !EMBEDDED
1085 config ANON_VMA_NAME
1086 bool "Anonymous VMA name support"
1087 depends on PROC_FS && ADVISE_SYSCALLS && MMU
1090 Allow naming anonymous virtual memory areas.
1092 This feature allows assigning names to virtual memory areas. Assigned
1093 names can be later retrieved from /proc/pid/maps and /proc/pid/smaps
1094 and help identifying individual anonymous memory areas.
1095 Assigning a name to anonymous virtual memory area might prevent that
1096 area from being merged with adjacent virtual memory areas due to the
1097 difference in their name.
1100 bool "Enable userfaultfd() system call"
1103 Enable the userfaultfd() system call that allows to intercept and
1104 handle page faults in userland.
1106 config HAVE_ARCH_USERFAULTFD_WP
1109 Arch has userfaultfd write protection support
1111 config HAVE_ARCH_USERFAULTFD_MINOR
1114 Arch has userfaultfd minor fault support
1120 Allows to create marker PTEs for file-backed memory.
1122 config PTE_MARKER_UFFD_WP
1123 bool "Userfaultfd write protection support for shmem/hugetlbfs"
1125 depends on HAVE_ARCH_USERFAULTFD_WP
1129 Allows to create marker PTEs for userfaultfd write protection
1130 purposes. It is required to enable userfaultfd write protection on
1131 file-backed memory types like shmem and hugetlbfs.
1133 source "mm/damon/Kconfig"