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"
31 A lightweight compressed cache for swap pages. It takes
32 pages that are in the process of being swapped out and attempts to
33 compress them into a dynamically allocated RAM-based memory pool.
34 This can result in a significant I/O reduction on swap device and,
35 in the case where decompressing from RAM is faster than swap device
36 reads, can also improve workload performance.
38 config ZSWAP_DEFAULT_ON
39 bool "Enable the compressed cache for swap pages by default"
42 If selected, the compressed cache for swap pages will be enabled
43 at boot, otherwise it will be disabled.
45 The selection made here can be overridden by using the kernel
46 command line 'zswap.enabled=' option.
48 config ZSWAP_SHRINKER_DEFAULT_ON
49 bool "Shrink the zswap pool on memory pressure"
53 If selected, the zswap shrinker will be enabled, and the pages
54 stored in the zswap pool will become available for reclaim (i.e
55 written back to the backing swap device) on memory pressure.
57 This means that zswap writeback could happen even if the pool is
58 not yet full, or the cgroup zswap limit has not been reached,
59 reducing the chance that cold pages will reside in the zswap pool
60 and consume memory indefinitely.
63 prompt "Default compressor"
65 default ZSWAP_COMPRESSOR_DEFAULT_LZO
67 Selects the default compression algorithm for the compressed cache
70 For an overview what kind of performance can be expected from
71 a particular compression algorithm please refer to the benchmarks
72 available at the following LWN page:
73 https://lwn.net/Articles/751795/
75 If in doubt, select 'LZO'.
77 The selection made here can be overridden by using the kernel
78 command line 'zswap.compressor=' option.
80 config ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
84 Use the Deflate algorithm as the default compression algorithm.
86 config ZSWAP_COMPRESSOR_DEFAULT_LZO
90 Use the LZO algorithm as the default compression algorithm.
92 config ZSWAP_COMPRESSOR_DEFAULT_842
96 Use the 842 algorithm as the default compression algorithm.
98 config ZSWAP_COMPRESSOR_DEFAULT_LZ4
102 Use the LZ4 algorithm as the default compression algorithm.
104 config ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
108 Use the LZ4HC algorithm as the default compression algorithm.
110 config ZSWAP_COMPRESSOR_DEFAULT_ZSTD
114 Use the zstd algorithm as the default compression algorithm.
117 config ZSWAP_COMPRESSOR_DEFAULT
120 default "deflate" if ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
121 default "lzo" if ZSWAP_COMPRESSOR_DEFAULT_LZO
122 default "842" if ZSWAP_COMPRESSOR_DEFAULT_842
123 default "lz4" if ZSWAP_COMPRESSOR_DEFAULT_LZ4
124 default "lz4hc" if ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
125 default "zstd" if ZSWAP_COMPRESSOR_DEFAULT_ZSTD
129 prompt "Default allocator"
131 default ZSWAP_ZPOOL_DEFAULT_ZSMALLOC if MMU
132 default ZSWAP_ZPOOL_DEFAULT_ZBUD
134 Selects the default allocator for the compressed cache for
136 The default is 'zbud' for compatibility, however please do
137 read the description of each of the allocators below before
138 making a right choice.
140 The selection made here can be overridden by using the kernel
141 command line 'zswap.zpool=' option.
143 config ZSWAP_ZPOOL_DEFAULT_ZBUD
147 Use the zbud allocator as the default allocator.
149 config ZSWAP_ZPOOL_DEFAULT_Z3FOLD_DEPRECATED
150 bool "z3foldi (DEPRECATED)"
151 select Z3FOLD_DEPRECATED
153 Use the z3fold allocator as the default allocator.
155 Deprecated and scheduled for removal in a few cycles,
156 see CONFIG_Z3FOLD_DEPRECATED.
158 config ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
162 Use the zsmalloc allocator as the default allocator.
165 config ZSWAP_ZPOOL_DEFAULT
168 default "zbud" if ZSWAP_ZPOOL_DEFAULT_ZBUD
169 default "z3fold" if ZSWAP_ZPOOL_DEFAULT_Z3FOLD_DEPRECATED
170 default "zsmalloc" if ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
174 tristate "2:1 compression allocator (zbud)"
177 A special purpose allocator for storing compressed pages.
178 It is designed to store up to two compressed pages per physical
179 page. While this design limits storage density, it has simple and
180 deterministic reclaim properties that make it preferable to a higher
181 density approach when reclaim will be used.
183 config Z3FOLD_DEPRECATED
184 tristate "3:1 compression allocator (z3fold) (DEPRECATED)"
187 Deprecated and scheduled for removal in a few cycles. If you have
188 a good reason for using Z3FOLD over ZSMALLOC, please contact
191 A special purpose allocator for storing compressed pages.
192 It is designed to store up to three compressed pages per physical
193 page. It is a ZBUD derivative so the simplicity and determinism are
198 default y if Z3FOLD_DEPRECATED=y
199 default m if Z3FOLD_DEPRECATED=m
200 depends on Z3FOLD_DEPRECATED
204 prompt "N:1 compression allocator (zsmalloc)" if (ZSWAP || ZRAM)
207 zsmalloc is a slab-based memory allocator designed to store
208 pages of various compression levels efficiently. It achieves
209 the highest storage density with the least amount of fragmentation.
212 bool "Export zsmalloc statistics"
216 This option enables code in the zsmalloc to collect various
217 statistics about what's happening in zsmalloc and exports that
218 information to userspace via debugfs.
221 config ZSMALLOC_CHAIN_SIZE
222 int "Maximum number of physical pages per-zspage"
227 This option sets the upper limit on the number of physical pages
228 that a zmalloc page (zspage) can consist of. The optimal zspage
229 chain size is calculated for each size class during the
230 initialization of the pool.
232 Changing this option can alter the characteristics of size classes,
233 such as the number of pages per zspage and the number of objects
234 per zspage. This can also result in different configurations of
235 the pool, as zsmalloc merges size classes with similar
238 For more information, see zsmalloc documentation.
240 menu "Slab allocator options"
246 bool "Configure for minimal memory footprint"
248 select SLAB_MERGE_DEFAULT
250 Configures the slab allocator in a way to achieve minimal memory
251 footprint, sacrificing scalability, debugging and other features.
252 This is intended only for the smallest system that had used the
253 SLOB allocator and is not recommended for systems with more than
258 config SLAB_MERGE_DEFAULT
259 bool "Allow slab caches to be merged"
262 For reduced kernel memory fragmentation, slab caches can be
263 merged when they share the same size and other characteristics.
264 This carries a risk of kernel heap overflows being able to
265 overwrite objects from merged caches (and more easily control
266 cache layout), which makes such heap attacks easier to exploit
267 by attackers. By keeping caches unmerged, these kinds of exploits
268 can usually only damage objects in the same cache. To disable
269 merging at runtime, "slab_nomerge" can be passed on the kernel
272 config SLAB_FREELIST_RANDOM
273 bool "Randomize slab freelist"
274 depends on !SLUB_TINY
276 Randomizes the freelist order used on creating new pages. This
277 security feature reduces the predictability of the kernel slab
278 allocator against heap overflows.
280 config SLAB_FREELIST_HARDENED
281 bool "Harden slab freelist metadata"
282 depends on !SLUB_TINY
284 Many kernel heap attacks try to target slab cache metadata and
285 other infrastructure. This options makes minor performance
286 sacrifices to harden the kernel slab allocator against common
287 freelist exploit methods.
290 bool "Support allocation from separate kmalloc buckets"
291 depends on !SLUB_TINY
292 default SLAB_FREELIST_HARDENED
294 Kernel heap attacks frequently depend on being able to create
295 specifically-sized allocations with user-controlled contents
296 that will be allocated into the same kmalloc bucket as a
297 target object. To avoid sharing these allocation buckets,
298 provide an explicitly separated set of buckets to be used for
299 user-controlled allocations. This may very slightly increase
300 memory fragmentation, though in practice it's only a handful
301 of extra pages since the bulk of user-controlled allocations
302 are relatively long-lived.
308 bool "Enable performance statistics"
309 depends on SYSFS && !SLUB_TINY
311 The statistics are useful to debug slab 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 SMP && !SLUB_TINY
322 bool "Enable per cpu partial caches"
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 config RANDOM_KMALLOC_CACHES
332 depends on !SLUB_TINY
333 bool "Randomize slab caches for normal kmalloc"
335 A hardening feature that creates multiple copies of slab caches for
336 normal kmalloc allocation and makes kmalloc randomly pick one based
337 on code address, which makes the attackers more difficult to spray
338 vulnerable memory objects on the heap for the purpose of exploiting
339 memory vulnerabilities.
341 Currently the number of copies is set to 16, a reasonably large value
342 that effectively diverges the memory objects allocated for different
343 subsystems or modules into different caches, at the expense of a
344 limited degree of memory and CPU overhead that relates to hardware and
347 endmenu # Slab allocator options
349 config SHUFFLE_PAGE_ALLOCATOR
350 bool "Page allocator randomization"
351 default SLAB_FREELIST_RANDOM && ACPI_NUMA
353 Randomization of the page allocator improves the average
354 utilization of a direct-mapped memory-side-cache. See section
355 5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI
356 6.2a specification for an example of how a platform advertises
357 the presence of a memory-side-cache. There are also incidental
358 security benefits as it reduces the predictability of page
359 allocations to compliment SLAB_FREELIST_RANDOM, but the
360 default granularity of shuffling on the MAX_PAGE_ORDER i.e, 10th
361 order of pages is selected based on cache utilization benefits
364 While the randomization improves cache utilization it may
365 negatively impact workloads on platforms without a cache. For
366 this reason, by default, the randomization is not enabled even
367 if SHUFFLE_PAGE_ALLOCATOR=y. The randomization may be force enabled
368 with the 'page_alloc.shuffle' kernel command line parameter.
373 bool "Disable heap randomization"
376 Randomizing heap placement makes heap exploits harder, but it
377 also breaks ancient binaries (including anything libc5 based).
378 This option changes the bootup default to heap randomization
379 disabled, and can be overridden at runtime by setting
380 /proc/sys/kernel/randomize_va_space to 2.
382 On non-ancient distros (post-2000 ones) N is usually a safe choice.
384 config MMAP_ALLOW_UNINITIALIZED
385 bool "Allow mmapped anonymous memory to be uninitialized"
386 depends on EXPERT && !MMU
389 Normally, and according to the Linux spec, anonymous memory obtained
390 from mmap() has its contents cleared before it is passed to
391 userspace. Enabling this config option allows you to request that
392 mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus
393 providing a huge performance boost. If this option is not enabled,
394 then the flag will be ignored.
396 This is taken advantage of by uClibc's malloc(), and also by
397 ELF-FDPIC binfmt's brk and stack allocator.
399 Because of the obvious security issues, this option should only be
400 enabled on embedded devices where you control what is run in
401 userspace. Since that isn't generally a problem on no-MMU systems,
402 it is normally safe to say Y here.
404 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
406 config SELECT_MEMORY_MODEL
408 depends on ARCH_SELECT_MEMORY_MODEL
411 prompt "Memory model"
412 depends on SELECT_MEMORY_MODEL
413 default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT
414 default FLATMEM_MANUAL
416 This option allows you to change some of the ways that
417 Linux manages its memory internally. Most users will
418 only have one option here selected by the architecture
419 configuration. This is normal.
421 config FLATMEM_MANUAL
423 depends on !ARCH_SPARSEMEM_ENABLE || ARCH_FLATMEM_ENABLE
425 This option is best suited for non-NUMA systems with
426 flat address space. The FLATMEM is the most efficient
427 system in terms of performance and resource consumption
428 and it is the best option for smaller systems.
430 For systems that have holes in their physical address
431 spaces and for features like NUMA and memory hotplug,
432 choose "Sparse Memory".
434 If unsure, choose this option (Flat Memory) over any other.
436 config SPARSEMEM_MANUAL
438 depends on ARCH_SPARSEMEM_ENABLE
440 This will be the only option for some systems, including
441 memory hot-plug systems. This is normal.
443 This option provides efficient support for systems with
444 holes is their physical address space and allows memory
445 hot-plug and hot-remove.
447 If unsure, choose "Flat Memory" over this option.
453 depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL
457 depends on !SPARSEMEM || FLATMEM_MANUAL
460 # SPARSEMEM_EXTREME (which is the default) does some bootmem
461 # allocations when sparse_init() is called. If this cannot
462 # be done on your architecture, select this option. However,
463 # statically allocating the mem_section[] array can potentially
464 # consume vast quantities of .bss, so be careful.
466 # This option will also potentially produce smaller runtime code
467 # with gcc 3.4 and later.
469 config SPARSEMEM_STATIC
473 # Architecture platforms which require a two level mem_section in SPARSEMEM
474 # must select this option. This is usually for architecture platforms with
475 # an extremely sparse physical address space.
477 config SPARSEMEM_EXTREME
479 depends on SPARSEMEM && !SPARSEMEM_STATIC
481 config SPARSEMEM_VMEMMAP_ENABLE
484 config SPARSEMEM_VMEMMAP
485 bool "Sparse Memory virtual memmap"
486 depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE
489 SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise
490 pfn_to_page and page_to_pfn operations. This is the most
491 efficient option when sufficient kernel resources are available.
493 # Select this config option from the architecture Kconfig, if it is preferred
494 # to enable the feature of HugeTLB/dev_dax vmemmap optimization.
496 config ARCH_WANT_OPTIMIZE_DAX_VMEMMAP
499 config ARCH_WANT_OPTIMIZE_HUGETLB_VMEMMAP
502 config HAVE_MEMBLOCK_PHYS_MAP
509 # Don't discard allocated memory used to track "memory" and "reserved" memblocks
510 # after early boot, so it can still be used to test for validity of memory.
511 # Also, memblocks are updated with memory hot(un)plug.
512 config ARCH_KEEP_MEMBLOCK
515 # Keep arch NUMA mapping infrastructure post-init.
516 config NUMA_KEEP_MEMINFO
519 config MEMORY_ISOLATION
522 # IORESOURCE_SYSTEM_RAM regions in the kernel resource tree that are marked
523 # IORESOURCE_EXCLUSIVE cannot be mapped to user space, for example, via
525 config EXCLUSIVE_SYSTEM_RAM
527 depends on !DEVMEM || STRICT_DEVMEM
530 # Only be set on architectures that have completely implemented memory hotplug
531 # feature. If you are not sure, don't touch it.
533 config HAVE_BOOTMEM_INFO_NODE
536 config ARCH_ENABLE_MEMORY_HOTPLUG
539 config ARCH_ENABLE_MEMORY_HOTREMOVE
542 # eventually, we can have this option just 'select SPARSEMEM'
543 menuconfig MEMORY_HOTPLUG
544 bool "Memory hotplug"
545 select MEMORY_ISOLATION
547 depends on ARCH_ENABLE_MEMORY_HOTPLUG
549 select NUMA_KEEP_MEMINFO if NUMA
554 prompt "Memory Hotplug Default Online Type"
555 default MHP_DEFAULT_ONLINE_TYPE_OFFLINE
557 Default memory type for hotplugged memory.
559 This option sets the default policy setting for memory hotplug
560 onlining policy (/sys/devices/system/memory/auto_online_blocks) which
561 determines what happens to newly added memory regions. Policy setting
562 can always be changed at runtime.
564 The default is 'offline'.
566 Select offline to defer onlining to drivers and user policy.
567 Select auto to let the kernel choose what zones to utilize.
568 Select online_kernel to generally allow kernel usage of this memory.
569 Select online_movable to generally disallow kernel usage of this memory.
571 Example kernel usage would be page structs and page tables.
573 See Documentation/admin-guide/mm/memory-hotplug.rst for more information.
575 config MHP_DEFAULT_ONLINE_TYPE_OFFLINE
578 Hotplugged memory will not be onlined by default.
579 Choose this for systems with drivers and user policy that
580 handle onlining of hotplug memory policy.
582 config MHP_DEFAULT_ONLINE_TYPE_ONLINE_AUTO
585 Select this if you want the kernel to automatically online
586 hotplugged memory into the zone it thinks is reasonable.
587 This memory may be utilized for kernel data.
589 config MHP_DEFAULT_ONLINE_TYPE_ONLINE_KERNEL
592 Select this if you want the kernel to automatically online
593 hotplugged memory into a zone capable of being used for kernel
594 data. This typically means ZONE_NORMAL.
596 config MHP_DEFAULT_ONLINE_TYPE_ONLINE_MOVABLE
599 Select this if you want the kernel to automatically online
600 hotplug memory into ZONE_MOVABLE. This memory will generally
601 not be utilized for kernel data.
603 This should only be used when the admin knows sufficient
604 ZONE_NORMAL memory is available to describe hotplug memory,
605 otherwise hotplug memory may fail to online. For example,
606 sufficient kernel-capable memory (ZONE_NORMAL) must be
607 available to allocate page structs to describe ZONE_MOVABLE.
611 config MEMORY_HOTREMOVE
612 bool "Allow for memory hot remove"
613 select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64)
614 depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE
617 config MHP_MEMMAP_ON_MEMORY
619 depends on MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP
620 depends on ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
622 endif # MEMORY_HOTPLUG
624 config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
627 # Heavily threaded applications may benefit from splitting the mm-wide
628 # page_table_lock, so that faults on different parts of the user address
629 # space can be handled with less contention: split it at this NR_CPUS.
630 # Default to 4 for wider testing, though 8 might be more appropriate.
631 # ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
632 # PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
633 # SPARC32 allocates multiple pte tables within a single page, and therefore
634 # a per-page lock leads to problems when multiple tables need to be locked
635 # at the same time (e.g. copy_page_range()).
636 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
638 config SPLIT_PTE_PTLOCKS
642 depends on NR_CPUS >= 4
643 depends on !ARM || CPU_CACHE_VIPT
644 depends on !PARISC || PA20
647 config ARCH_ENABLE_SPLIT_PMD_PTLOCK
650 config SPLIT_PMD_PTLOCKS
652 depends on SPLIT_PTE_PTLOCKS && ARCH_ENABLE_SPLIT_PMD_PTLOCK
655 # support for memory balloon
656 config MEMORY_BALLOON
660 # support for memory balloon compaction
661 config BALLOON_COMPACTION
662 bool "Allow for balloon memory compaction/migration"
664 depends on COMPACTION && MEMORY_BALLOON
666 Memory fragmentation introduced by ballooning might reduce
667 significantly the number of 2MB contiguous memory blocks that can be
668 used within a guest, thus imposing performance penalties associated
669 with the reduced number of transparent huge pages that could be used
670 by the guest workload. Allowing the compaction & migration for memory
671 pages enlisted as being part of memory balloon devices avoids the
672 scenario aforementioned and helps improving memory defragmentation.
675 # support for memory compaction
677 bool "Allow for memory compaction"
682 Compaction is the only memory management component to form
683 high order (larger physically contiguous) memory blocks
684 reliably. The page allocator relies on compaction heavily and
685 the lack of the feature can lead to unexpected OOM killer
686 invocations for high order memory requests. You shouldn't
687 disable this option unless there really is a strong reason for
688 it and then we would be really interested to hear about that at
691 config COMPACT_UNEVICTABLE_DEFAULT
693 depends on COMPACTION
694 default 0 if PREEMPT_RT
698 # support for free page reporting
699 config PAGE_REPORTING
700 bool "Free page reporting"
702 Free page reporting allows for the incremental acquisition of
703 free pages from the buddy allocator for the purpose of reporting
704 those pages to another entity, such as a hypervisor, so that the
705 memory can be freed within the host for other uses.
708 # support for page migration
711 bool "Page migration"
713 depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU
715 Allows the migration of the physical location of pages of processes
716 while the virtual addresses are not changed. This is useful in
717 two situations. The first is on NUMA systems to put pages nearer
718 to the processors accessing. The second is when allocating huge
719 pages as migration can relocate pages to satisfy a huge page
720 allocation instead of reclaiming.
722 config DEVICE_MIGRATION
723 def_bool MIGRATION && ZONE_DEVICE
725 config ARCH_ENABLE_HUGEPAGE_MIGRATION
728 config ARCH_ENABLE_THP_MIGRATION
731 config HUGETLB_PAGE_SIZE_VARIABLE
734 Allows the pageblock_order value to be dynamic instead of just standard
735 HUGETLB_PAGE_ORDER when there are multiple HugeTLB page sizes available
738 Note that the pageblock_order cannot exceed MAX_PAGE_ORDER and will be
739 clamped down to MAX_PAGE_ORDER.
742 def_bool (MEMORY_ISOLATION && COMPACTION) || CMA
744 config PCP_BATCH_SCALE_MAX
745 int "Maximum scale factor of PCP (Per-CPU pageset) batch allocate/free"
749 In page allocator, PCP (Per-CPU pageset) is refilled and drained in
750 batches. The batch number is scaled automatically to improve page
751 allocation/free throughput. But too large scale factor may hurt
752 latency. This option sets the upper limit of scale factor to limit
755 config PHYS_ADDR_T_64BIT
759 bool "Enable bounce buffers"
761 depends on BLOCK && MMU && HIGHMEM
763 Enable bounce buffers for devices that cannot access the full range of
764 memory available to the CPU. Enabled by default when HIGHMEM is
765 selected, but you may say n to override this.
772 bool "Enable KSM for page merging"
776 Enable Kernel Samepage Merging: KSM periodically scans those areas
777 of an application's address space that an app has advised may be
778 mergeable. When it finds pages of identical content, it replaces
779 the many instances by a single page with that content, so
780 saving memory until one or another app needs to modify the content.
781 Recommended for use with KVM, or with other duplicative applications.
782 See Documentation/mm/ksm.rst for more information: KSM is inactive
783 until a program has madvised that an area is MADV_MERGEABLE, and
784 root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set).
786 config DEFAULT_MMAP_MIN_ADDR
787 int "Low address space to protect from user allocation"
791 This is the portion of low virtual memory which should be protected
792 from userspace allocation. Keeping a user from writing to low pages
793 can help reduce the impact of kernel NULL pointer bugs.
795 For most arm64, ppc64 and x86 users with lots of address space
796 a value of 65536 is reasonable and should cause no problems.
797 On arm and other archs it should not be higher than 32768.
798 Programs which use vm86 functionality or have some need to map
799 this low address space will need CAP_SYS_RAWIO or disable this
800 protection by setting the value to 0.
802 This value can be changed after boot using the
803 /proc/sys/vm/mmap_min_addr tunable.
805 config ARCH_SUPPORTS_MEMORY_FAILURE
808 config MEMORY_FAILURE
810 depends on ARCH_SUPPORTS_MEMORY_FAILURE
811 bool "Enable recovery from hardware memory errors"
812 select MEMORY_ISOLATION
815 Enables code to recover from some memory failures on systems
816 with MCA recovery. This allows a system to continue running
817 even when some of its memory has uncorrected errors. This requires
818 special hardware support and typically ECC memory.
820 config HWPOISON_INJECT
821 tristate "HWPoison pages injector"
822 depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS
823 select PROC_PAGE_MONITOR
825 config NOMMU_INITIAL_TRIM_EXCESS
826 int "Turn on mmap() excess space trimming before booting"
830 The NOMMU mmap() frequently needs to allocate large contiguous chunks
831 of memory on which to store mappings, but it can only ask the system
832 allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently
833 more than it requires. To deal with this, mmap() is able to trim off
834 the excess and return it to the allocator.
836 If trimming is enabled, the excess is trimmed off and returned to the
837 system allocator, which can cause extra fragmentation, particularly
838 if there are a lot of transient processes.
840 If trimming is disabled, the excess is kept, but not used, which for
841 long-term mappings means that the space is wasted.
843 Trimming can be dynamically controlled through a sysctl option
844 (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of
845 excess pages there must be before trimming should occur, or zero if
846 no trimming is to occur.
848 This option specifies the initial value of this option. The default
849 of 1 says that all excess pages should be trimmed.
851 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
853 config ARCH_WANT_GENERAL_HUGETLB
856 config ARCH_WANTS_THP_SWAP
859 menuconfig TRANSPARENT_HUGEPAGE
860 bool "Transparent Hugepage Support"
861 depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT
865 Transparent Hugepages allows the kernel to use huge pages and
866 huge tlb transparently to the applications whenever possible.
867 This feature can improve computing performance to certain
868 applications by speeding up page faults during memory
869 allocation, by reducing the number of tlb misses and by speeding
870 up the pagetable walking.
872 If memory constrained on embedded, you may want to say N.
874 if TRANSPARENT_HUGEPAGE
877 prompt "Transparent Hugepage Support sysfs defaults"
878 depends on TRANSPARENT_HUGEPAGE
879 default TRANSPARENT_HUGEPAGE_ALWAYS
881 Selects the sysfs defaults for Transparent Hugepage Support.
883 config TRANSPARENT_HUGEPAGE_ALWAYS
886 Enabling Transparent Hugepage always, can increase the
887 memory footprint of applications without a guaranteed
888 benefit but it will work automatically for all applications.
890 config TRANSPARENT_HUGEPAGE_MADVISE
893 Enabling Transparent Hugepage madvise, will only provide a
894 performance improvement benefit to the applications using
895 madvise(MADV_HUGEPAGE) but it won't risk to increase the
896 memory footprint of applications without a guaranteed
899 config TRANSPARENT_HUGEPAGE_NEVER
902 Disable Transparent Hugepage by default. It can still be
903 enabled at runtime via sysfs.
908 depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP && 64BIT
910 Swap transparent huge pages in one piece, without splitting.
911 XXX: For now, swap cluster backing transparent huge page
912 will be split after swapout.
914 For selection by architectures with reasonable THP sizes.
916 config READ_ONLY_THP_FOR_FS
917 bool "Read-only THP for filesystems (EXPERIMENTAL)"
918 depends on TRANSPARENT_HUGEPAGE && SHMEM
921 Allow khugepaged to put read-only file-backed pages in THP.
923 This is marked experimental because it is a new feature. Write
924 support of file THPs will be developed in the next few release
927 endif # TRANSPARENT_HUGEPAGE
930 # The architecture supports pgtable leaves that is larger than PAGE_SIZE
932 config PGTABLE_HAS_HUGE_LEAVES
933 def_bool TRANSPARENT_HUGEPAGE || HUGETLB_PAGE
935 # TODO: Allow to be enabled without THP
936 config ARCH_SUPPORTS_HUGE_PFNMAP
938 depends on TRANSPARENT_HUGEPAGE
940 config ARCH_SUPPORTS_PMD_PFNMAP
942 depends on ARCH_SUPPORTS_HUGE_PFNMAP && HAVE_ARCH_TRANSPARENT_HUGEPAGE
944 config ARCH_SUPPORTS_PUD_PFNMAP
946 depends on ARCH_SUPPORTS_HUGE_PFNMAP && HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
949 # UP and nommu archs use km based percpu allocator
951 config NEED_PER_CPU_KM
952 depends on !SMP || !MMU
956 config NEED_PER_CPU_EMBED_FIRST_CHUNK
959 config NEED_PER_CPU_PAGE_FIRST_CHUNK
962 config USE_PERCPU_NUMA_NODE_ID
965 config HAVE_SETUP_PER_CPU_AREA
969 bool "Contiguous Memory Allocator"
972 select MEMORY_ISOLATION
974 This enables the Contiguous Memory Allocator which allows other
975 subsystems to allocate big physically-contiguous blocks of memory.
976 CMA reserves a region of memory and allows only movable pages to
977 be allocated from it. This way, the kernel can use the memory for
978 pagecache and when a subsystem requests for contiguous area, the
979 allocated pages are migrated away to serve the contiguous request.
984 bool "CMA debugfs interface"
985 depends on CMA && DEBUG_FS
987 Turns on the DebugFS interface for CMA.
990 bool "CMA information through sysfs interface"
991 depends on CMA && SYSFS
993 This option exposes some sysfs attributes to get information
997 int "Maximum count of the CMA areas"
1002 CMA allows to create CMA areas for particular purpose, mainly,
1003 used as device private area. This parameter sets the maximum
1004 number of CMA area in the system.
1006 If unsure, leave the default value "8" in UMA and "20" in NUMA.
1008 config MEM_SOFT_DIRTY
1009 bool "Track memory changes"
1010 depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS
1011 select PROC_PAGE_MONITOR
1013 This option enables memory changes tracking by introducing a
1014 soft-dirty bit on pte-s. This bit it set when someone writes
1015 into a page just as regular dirty bit, but unlike the latter
1016 it can be cleared by hands.
1018 See Documentation/admin-guide/mm/soft-dirty.rst for more details.
1020 config GENERIC_EARLY_IOREMAP
1023 config STACK_MAX_DEFAULT_SIZE_MB
1024 int "Default maximum user stack size for 32-bit processes (MB)"
1027 depends on STACK_GROWSUP && (!64BIT || COMPAT)
1029 This is the maximum stack size in Megabytes in the VM layout of 32-bit
1030 user processes when the stack grows upwards (currently only on parisc
1031 arch) when the RLIMIT_STACK hard limit is unlimited.
1033 A sane initial value is 100 MB.
1035 config DEFERRED_STRUCT_PAGE_INIT
1036 bool "Defer initialisation of struct pages to kthreads"
1037 depends on SPARSEMEM
1038 depends on !NEED_PER_CPU_KM
1043 Ordinarily all struct pages are initialised during early boot in a
1044 single thread. On very large machines this can take a considerable
1045 amount of time. If this option is set, large machines will bring up
1046 a subset of memmap at boot and then initialise the rest in parallel.
1047 This has a potential performance impact on tasks running early in the
1048 lifetime of the system until these kthreads finish the
1051 config PAGE_IDLE_FLAG
1053 select PAGE_EXTENSION if !64BIT
1055 This adds PG_idle and PG_young flags to 'struct page'. PTE Accessed
1056 bit writers can set the state of the bit in the flags so that PTE
1057 Accessed bit readers may avoid disturbance.
1059 config IDLE_PAGE_TRACKING
1060 bool "Enable idle page tracking"
1061 depends on SYSFS && MMU
1062 select PAGE_IDLE_FLAG
1064 This feature allows to estimate the amount of user pages that have
1065 not been touched during a given period of time. This information can
1066 be useful to tune memory cgroup limits and/or for job placement
1067 within a compute cluster.
1069 See Documentation/admin-guide/mm/idle_page_tracking.rst for
1072 # Architectures which implement cpu_dcache_is_aliasing() to query
1073 # whether the data caches are aliased (VIVT or VIPT with dcache
1074 # aliasing) need to select this.
1075 config ARCH_HAS_CPU_CACHE_ALIASING
1078 config ARCH_HAS_CACHE_LINE_SIZE
1081 config ARCH_HAS_CURRENT_STACK_POINTER
1084 In support of HARDENED_USERCOPY performing stack variable lifetime
1085 checking, an architecture-agnostic way to find the stack pointer
1086 is needed. Once an architecture defines an unsigned long global
1087 register alias named "current_stack_pointer", this config can be
1090 config ARCH_HAS_PTE_DEVMAP
1093 config ARCH_HAS_ZONE_DMA_SET
1097 bool "Support DMA zone" if ARCH_HAS_ZONE_DMA_SET
1098 default y if ARM64 || X86
1101 bool "Support DMA32 zone" if ARCH_HAS_ZONE_DMA_SET
1106 bool "Device memory (pmem, HMM, etc...) hotplug support"
1107 depends on MEMORY_HOTPLUG
1108 depends on MEMORY_HOTREMOVE
1109 depends on SPARSEMEM_VMEMMAP
1110 depends on ARCH_HAS_PTE_DEVMAP
1114 Device memory hotplug support allows for establishing pmem,
1115 or other device driver discovered memory regions, in the
1116 memmap. This allows pfn_to_page() lookups of otherwise
1117 "device-physical" addresses which is needed for using a DAX
1118 mapping in an O_DIRECT operation, among other things.
1120 If FS_DAX is enabled, then say Y.
1123 # Helpers to mirror range of the CPU page tables of a process into device page
1130 config GET_FREE_REGION
1133 config DEVICE_PRIVATE
1134 bool "Unaddressable device memory (GPU memory, ...)"
1135 depends on ZONE_DEVICE
1136 select GET_FREE_REGION
1139 Allows creation of struct pages to represent unaddressable device
1140 memory; i.e., memory that is only accessible from the device (or
1141 group of devices). You likely also want to select HMM_MIRROR.
1146 config ARCH_USES_HIGH_VMA_FLAGS
1148 config ARCH_HAS_PKEYS
1151 config ARCH_USES_PG_ARCH_2
1153 config ARCH_USES_PG_ARCH_3
1156 config VM_EVENT_COUNTERS
1158 bool "Enable VM event counters for /proc/vmstat" if EXPERT
1160 VM event counters are needed for event counts to be shown.
1161 This option allows the disabling of the VM event counters
1162 on EXPERT systems. /proc/vmstat will only show page counts
1163 if VM event counters are disabled.
1166 bool "Collect percpu memory statistics"
1168 This feature collects and exposes statistics via debugfs. The
1169 information includes global and per chunk statistics, which can
1170 be used to help understand percpu memory usage.
1173 bool "Enable infrastructure for get_user_pages()-related unit tests"
1176 Provides /sys/kernel/debug/gup_test, which in turn provides a way
1177 to make ioctl calls that can launch kernel-based unit tests for
1178 the get_user_pages*() and pin_user_pages*() family of API calls.
1180 These tests include benchmark testing of the _fast variants of
1181 get_user_pages*() and pin_user_pages*(), as well as smoke tests of
1182 the non-_fast variants.
1184 There is also a sub-test that allows running dump_page() on any
1185 of up to eight pages (selected by command line args) within the
1186 range of user-space addresses. These pages are either pinned via
1187 pin_user_pages*(), or pinned via get_user_pages*(), as specified
1188 by other command line arguments.
1190 See tools/testing/selftests/mm/gup_test.c
1192 comment "GUP_TEST needs to have DEBUG_FS enabled"
1193 depends on !GUP_TEST && !DEBUG_FS
1195 config GUP_GET_PXX_LOW_HIGH
1199 tristate "Enable a module to run time tests on dma_pool"
1202 Provides a test module that will allocate and free many blocks of
1203 various sizes and report how long it takes. This is intended to
1204 provide a consistent way to measure how changes to the
1205 dma_pool_alloc/free routines affect performance.
1207 config ARCH_HAS_PTE_SPECIAL
1210 config MAPPING_DIRTY_HELPERS
1216 config KMAP_LOCAL_NON_LINEAR_PTE_ARRAY
1219 # struct io_mapping based helper. Selected by drivers that need them
1224 bool "Enable memfd_create() system call" if EXPERT
1228 bool "Enable memfd_secret() system call" if EXPERT
1229 depends on ARCH_HAS_SET_DIRECT_MAP
1231 Enable the memfd_secret() system call with the ability to create
1232 memory areas visible only in the context of the owning process and
1233 not mapped to other processes and other kernel page tables.
1235 config ANON_VMA_NAME
1236 bool "Anonymous VMA name support"
1237 depends on PROC_FS && ADVISE_SYSCALLS && MMU
1240 Allow naming anonymous virtual memory areas.
1242 This feature allows assigning names to virtual memory areas. Assigned
1243 names can be later retrieved from /proc/pid/maps and /proc/pid/smaps
1244 and help identifying individual anonymous memory areas.
1245 Assigning a name to anonymous virtual memory area might prevent that
1246 area from being merged with adjacent virtual memory areas due to the
1247 difference in their name.
1249 config HAVE_ARCH_USERFAULTFD_WP
1252 Arch has userfaultfd write protection support
1254 config HAVE_ARCH_USERFAULTFD_MINOR
1257 Arch has userfaultfd minor fault support
1259 menuconfig USERFAULTFD
1260 bool "Enable userfaultfd() system call"
1263 Enable the userfaultfd() system call that allows to intercept and
1264 handle page faults in userland.
1267 config PTE_MARKER_UFFD_WP
1268 bool "Userfaultfd write protection support for shmem/hugetlbfs"
1270 depends on HAVE_ARCH_USERFAULTFD_WP
1273 Allows to create marker PTEs for userfaultfd write protection
1274 purposes. It is required to enable userfaultfd write protection on
1275 file-backed memory types like shmem and hugetlbfs.
1280 bool "Multi-Gen LRU"
1282 # make sure folio->flags has enough spare bits
1283 depends on 64BIT || !SPARSEMEM || SPARSEMEM_VMEMMAP
1285 A high performance LRU implementation to overcommit memory. See
1286 Documentation/admin-guide/mm/multigen_lru.rst for details.
1288 config LRU_GEN_ENABLED
1289 bool "Enable by default"
1292 This option enables the multi-gen LRU by default.
1294 config LRU_GEN_STATS
1295 bool "Full stats for debugging"
1298 Do not enable this option unless you plan to look at historical stats
1299 from evicted generations for debugging purpose.
1301 This option has a per-memcg and per-node memory overhead.
1303 config LRU_GEN_WALKS_MMU
1305 depends on LRU_GEN && ARCH_HAS_HW_PTE_YOUNG
1308 config ARCH_SUPPORTS_PER_VMA_LOCK
1313 depends on ARCH_SUPPORTS_PER_VMA_LOCK && MMU && SMP
1315 Allow per-vma locking during page fault handling.
1317 This feature allows locking each virtual memory area separately when
1318 handling page faults instead of taking mmap_lock.
1320 config LOCK_MM_AND_FIND_VMA
1322 depends on !STACK_GROWSUP
1324 config IOMMU_MM_DATA
1334 bool "NUMA emulation"
1335 depends on NUMA_MEMBLKS
1337 Enable NUMA emulation. A flat machine will be split
1338 into virtual nodes when booted with "numa=fake=N", where N is the
1339 number of nodes. This is only useful for debugging.
1341 config ARCH_HAS_USER_SHADOW_STACK
1344 The architecture has hardware support for userspace shadow call
1345 stacks (eg, x86 CET, arm64 GCS or RISC-V Zicfiss).
1347 config ARCH_SUPPORTS_PT_RECLAIM
1351 bool "reclaim empty user page table pages"
1353 depends on ARCH_SUPPORTS_PT_RECLAIM && MMU && SMP
1354 select MMU_GATHER_RCU_TABLE_FREE
1356 Try to reclaim empty user page table pages in paths other than munmap
1359 Note: now only empty user PTE page table pages will be reclaimed.
1362 source "mm/damon/Kconfig"
projects / linux.git / blob