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
153 Use the z3fold allocator as the default allocator.
155 config ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
159 Use the zsmalloc allocator as the default allocator.
162 config ZSWAP_ZPOOL_DEFAULT
165 default "zbud" if ZSWAP_ZPOOL_DEFAULT_ZBUD
166 default "z3fold" if ZSWAP_ZPOOL_DEFAULT_Z3FOLD
167 default "zsmalloc" if ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
171 tristate "2:1 compression allocator (zbud)"
174 A special purpose allocator for storing compressed pages.
175 It is designed to store up to two compressed pages per physical
176 page. While this design limits storage density, it has simple and
177 deterministic reclaim properties that make it preferable to a higher
178 density approach when reclaim will be used.
181 tristate "3:1 compression allocator (z3fold)"
184 A special purpose allocator for storing compressed pages.
185 It is designed to store up to three compressed pages per physical
186 page. It is a ZBUD derivative so the simplicity and determinism are
191 prompt "N:1 compression allocator (zsmalloc)" if ZSWAP
194 zsmalloc is a slab-based memory allocator designed to store
195 pages of various compression levels efficiently. It achieves
196 the highest storage density with the least amount of fragmentation.
199 bool "Export zsmalloc statistics"
203 This option enables code in the zsmalloc to collect various
204 statistics about what's happening in zsmalloc and exports that
205 information to userspace via debugfs.
208 config ZSMALLOC_CHAIN_SIZE
209 int "Maximum number of physical pages per-zspage"
214 This option sets the upper limit on the number of physical pages
215 that a zmalloc page (zspage) can consist of. The optimal zspage
216 chain size is calculated for each size class during the
217 initialization of the pool.
219 Changing this option can alter the characteristics of size classes,
220 such as the number of pages per zspage and the number of objects
221 per zspage. This can also result in different configurations of
222 the pool, as zsmalloc merges size classes with similar
225 For more information, see zsmalloc documentation.
227 menu "Slab allocator options"
233 bool "Configure for minimal memory footprint"
235 select SLAB_MERGE_DEFAULT
237 Configures the slab allocator in a way to achieve minimal memory
238 footprint, sacrificing scalability, debugging and other features.
239 This is intended only for the smallest system that had used the
240 SLOB allocator and is not recommended for systems with more than
245 config SLAB_MERGE_DEFAULT
246 bool "Allow slab caches to be merged"
249 For reduced kernel memory fragmentation, slab caches can be
250 merged when they share the same size and other characteristics.
251 This carries a risk of kernel heap overflows being able to
252 overwrite objects from merged caches (and more easily control
253 cache layout), which makes such heap attacks easier to exploit
254 by attackers. By keeping caches unmerged, these kinds of exploits
255 can usually only damage objects in the same cache. To disable
256 merging at runtime, "slab_nomerge" can be passed on the kernel
259 config SLAB_FREELIST_RANDOM
260 bool "Randomize slab freelist"
261 depends on !SLUB_TINY
263 Randomizes the freelist order used on creating new pages. This
264 security feature reduces the predictability of the kernel slab
265 allocator against heap overflows.
267 config SLAB_FREELIST_HARDENED
268 bool "Harden slab freelist metadata"
269 depends on !SLUB_TINY
271 Many kernel heap attacks try to target slab cache metadata and
272 other infrastructure. This options makes minor performance
273 sacrifices to harden the kernel slab allocator against common
274 freelist exploit methods.
278 bool "Enable performance statistics"
279 depends on SYSFS && !SLUB_TINY
281 The statistics are useful to debug slab allocation behavior in
282 order find ways to optimize the allocator. This should never be
283 enabled for production use since keeping statistics slows down
284 the allocator by a few percentage points. The slabinfo command
285 supports the determination of the most active slabs to figure
286 out which slabs are relevant to a particular load.
287 Try running: slabinfo -DA
289 config SLUB_CPU_PARTIAL
291 depends on SMP && !SLUB_TINY
292 bool "Enable per cpu partial caches"
294 Per cpu partial caches accelerate objects allocation and freeing
295 that is local to a processor at the price of more indeterminism
296 in the latency of the free. On overflow these caches will be cleared
297 which requires the taking of locks that may cause latency spikes.
298 Typically one would choose no for a realtime system.
300 config RANDOM_KMALLOC_CACHES
302 depends on !SLUB_TINY
303 bool "Randomize slab caches for normal kmalloc"
305 A hardening feature that creates multiple copies of slab caches for
306 normal kmalloc allocation and makes kmalloc randomly pick one based
307 on code address, which makes the attackers more difficult to spray
308 vulnerable memory objects on the heap for the purpose of exploiting
309 memory vulnerabilities.
311 Currently the number of copies is set to 16, a reasonably large value
312 that effectively diverges the memory objects allocated for different
313 subsystems or modules into different caches, at the expense of a
314 limited degree of memory and CPU overhead that relates to hardware and
317 endmenu # Slab allocator options
319 config SHUFFLE_PAGE_ALLOCATOR
320 bool "Page allocator randomization"
321 default SLAB_FREELIST_RANDOM && ACPI_NUMA
323 Randomization of the page allocator improves the average
324 utilization of a direct-mapped memory-side-cache. See section
325 5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI
326 6.2a specification for an example of how a platform advertises
327 the presence of a memory-side-cache. There are also incidental
328 security benefits as it reduces the predictability of page
329 allocations to compliment SLAB_FREELIST_RANDOM, but the
330 default granularity of shuffling on the MAX_PAGE_ORDER i.e, 10th
331 order of pages is selected based on cache utilization benefits
334 While the randomization improves cache utilization it may
335 negatively impact workloads on platforms without a cache. For
336 this reason, by default, the randomization is not enabled even
337 if SHUFFLE_PAGE_ALLOCATOR=y. The randomization may be force enabled
338 with the 'page_alloc.shuffle' kernel command line parameter.
343 bool "Disable heap randomization"
346 Randomizing heap placement makes heap exploits harder, but it
347 also breaks ancient binaries (including anything libc5 based).
348 This option changes the bootup default to heap randomization
349 disabled, and can be overridden at runtime by setting
350 /proc/sys/kernel/randomize_va_space to 2.
352 On non-ancient distros (post-2000 ones) N is usually a safe choice.
354 config MMAP_ALLOW_UNINITIALIZED
355 bool "Allow mmapped anonymous memory to be uninitialized"
356 depends on EXPERT && !MMU
359 Normally, and according to the Linux spec, anonymous memory obtained
360 from mmap() has its contents cleared before it is passed to
361 userspace. Enabling this config option allows you to request that
362 mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus
363 providing a huge performance boost. If this option is not enabled,
364 then the flag will be ignored.
366 This is taken advantage of by uClibc's malloc(), and also by
367 ELF-FDPIC binfmt's brk and stack allocator.
369 Because of the obvious security issues, this option should only be
370 enabled on embedded devices where you control what is run in
371 userspace. Since that isn't generally a problem on no-MMU systems,
372 it is normally safe to say Y here.
374 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
376 config SELECT_MEMORY_MODEL
378 depends on ARCH_SELECT_MEMORY_MODEL
381 prompt "Memory model"
382 depends on SELECT_MEMORY_MODEL
383 default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT
384 default FLATMEM_MANUAL
386 This option allows you to change some of the ways that
387 Linux manages its memory internally. Most users will
388 only have one option here selected by the architecture
389 configuration. This is normal.
391 config FLATMEM_MANUAL
393 depends on !ARCH_SPARSEMEM_ENABLE || ARCH_FLATMEM_ENABLE
395 This option is best suited for non-NUMA systems with
396 flat address space. The FLATMEM is the most efficient
397 system in terms of performance and resource consumption
398 and it is the best option for smaller systems.
400 For systems that have holes in their physical address
401 spaces and for features like NUMA and memory hotplug,
402 choose "Sparse Memory".
404 If unsure, choose this option (Flat Memory) over any other.
406 config SPARSEMEM_MANUAL
408 depends on ARCH_SPARSEMEM_ENABLE
410 This will be the only option for some systems, including
411 memory hot-plug systems. This is normal.
413 This option provides efficient support for systems with
414 holes is their physical address space and allows memory
415 hot-plug and hot-remove.
417 If unsure, choose "Flat Memory" over this option.
423 depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL
427 depends on !SPARSEMEM || FLATMEM_MANUAL
430 # SPARSEMEM_EXTREME (which is the default) does some bootmem
431 # allocations when sparse_init() is called. If this cannot
432 # be done on your architecture, select this option. However,
433 # statically allocating the mem_section[] array can potentially
434 # consume vast quantities of .bss, so be careful.
436 # This option will also potentially produce smaller runtime code
437 # with gcc 3.4 and later.
439 config SPARSEMEM_STATIC
443 # Architecture platforms which require a two level mem_section in SPARSEMEM
444 # must select this option. This is usually for architecture platforms with
445 # an extremely sparse physical address space.
447 config SPARSEMEM_EXTREME
449 depends on SPARSEMEM && !SPARSEMEM_STATIC
451 config SPARSEMEM_VMEMMAP_ENABLE
454 config SPARSEMEM_VMEMMAP
455 bool "Sparse Memory virtual memmap"
456 depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE
459 SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise
460 pfn_to_page and page_to_pfn operations. This is the most
461 efficient option when sufficient kernel resources are available.
463 # Select this config option from the architecture Kconfig, if it is preferred
464 # to enable the feature of HugeTLB/dev_dax vmemmap optimization.
466 config ARCH_WANT_OPTIMIZE_DAX_VMEMMAP
469 config ARCH_WANT_OPTIMIZE_HUGETLB_VMEMMAP
472 config HAVE_MEMBLOCK_PHYS_MAP
479 # Don't discard allocated memory used to track "memory" and "reserved" memblocks
480 # after early boot, so it can still be used to test for validity of memory.
481 # Also, memblocks are updated with memory hot(un)plug.
482 config ARCH_KEEP_MEMBLOCK
485 # Keep arch NUMA mapping infrastructure post-init.
486 config NUMA_KEEP_MEMINFO
489 config MEMORY_ISOLATION
492 # IORESOURCE_SYSTEM_RAM regions in the kernel resource tree that are marked
493 # IORESOURCE_EXCLUSIVE cannot be mapped to user space, for example, via
495 config EXCLUSIVE_SYSTEM_RAM
497 depends on !DEVMEM || STRICT_DEVMEM
500 # Only be set on architectures that have completely implemented memory hotplug
501 # feature. If you are not sure, don't touch it.
503 config HAVE_BOOTMEM_INFO_NODE
506 config ARCH_ENABLE_MEMORY_HOTPLUG
509 config ARCH_ENABLE_MEMORY_HOTREMOVE
512 # eventually, we can have this option just 'select SPARSEMEM'
513 menuconfig MEMORY_HOTPLUG
514 bool "Memory hotplug"
515 select MEMORY_ISOLATION
517 depends on ARCH_ENABLE_MEMORY_HOTPLUG
519 select NUMA_KEEP_MEMINFO if NUMA
523 config MEMORY_HOTPLUG_DEFAULT_ONLINE
524 bool "Online the newly added memory blocks by default"
525 depends on MEMORY_HOTPLUG
527 This option sets the default policy setting for memory hotplug
528 onlining policy (/sys/devices/system/memory/auto_online_blocks) which
529 determines what happens to newly added memory regions. Policy setting
530 can always be changed at runtime.
531 See Documentation/admin-guide/mm/memory-hotplug.rst for more information.
533 Say Y here if you want all hot-plugged memory blocks to appear in
534 'online' state by default.
535 Say N here if you want the default policy to keep all hot-plugged
536 memory blocks in 'offline' state.
538 config MEMORY_HOTREMOVE
539 bool "Allow for memory hot remove"
540 select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64)
541 depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE
544 config MHP_MEMMAP_ON_MEMORY
546 depends on MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP
547 depends on ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
549 endif # MEMORY_HOTPLUG
551 config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
554 # Heavily threaded applications may benefit from splitting the mm-wide
555 # page_table_lock, so that faults on different parts of the user address
556 # space can be handled with less contention: split it at this NR_CPUS.
557 # Default to 4 for wider testing, though 8 might be more appropriate.
558 # ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
559 # PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
560 # SPARC32 allocates multiple pte tables within a single page, and therefore
561 # a per-page lock leads to problems when multiple tables need to be locked
562 # at the same time (e.g. copy_page_range()).
563 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
565 config SPLIT_PTLOCK_CPUS
567 default "999999" if !MMU
568 default "999999" if ARM && !CPU_CACHE_VIPT
569 default "999999" if PARISC && !PA20
570 default "999999" if SPARC32
573 config ARCH_ENABLE_SPLIT_PMD_PTLOCK
577 # support for memory balloon
578 config MEMORY_BALLOON
582 # support for memory balloon compaction
583 config BALLOON_COMPACTION
584 bool "Allow for balloon memory compaction/migration"
586 depends on COMPACTION && MEMORY_BALLOON
588 Memory fragmentation introduced by ballooning might reduce
589 significantly the number of 2MB contiguous memory blocks that can be
590 used within a guest, thus imposing performance penalties associated
591 with the reduced number of transparent huge pages that could be used
592 by the guest workload. Allowing the compaction & migration for memory
593 pages enlisted as being part of memory balloon devices avoids the
594 scenario aforementioned and helps improving memory defragmentation.
597 # support for memory compaction
599 bool "Allow for memory compaction"
604 Compaction is the only memory management component to form
605 high order (larger physically contiguous) memory blocks
606 reliably. The page allocator relies on compaction heavily and
607 the lack of the feature can lead to unexpected OOM killer
608 invocations for high order memory requests. You shouldn't
609 disable this option unless there really is a strong reason for
610 it and then we would be really interested to hear about that at
613 config COMPACT_UNEVICTABLE_DEFAULT
615 depends on COMPACTION
616 default 0 if PREEMPT_RT
620 # support for free page reporting
621 config PAGE_REPORTING
622 bool "Free page reporting"
624 Free page reporting allows for the incremental acquisition of
625 free pages from the buddy allocator for the purpose of reporting
626 those pages to another entity, such as a hypervisor, so that the
627 memory can be freed within the host for other uses.
630 # support for page migration
633 bool "Page migration"
635 depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU
637 Allows the migration of the physical location of pages of processes
638 while the virtual addresses are not changed. This is useful in
639 two situations. The first is on NUMA systems to put pages nearer
640 to the processors accessing. The second is when allocating huge
641 pages as migration can relocate pages to satisfy a huge page
642 allocation instead of reclaiming.
644 config DEVICE_MIGRATION
645 def_bool MIGRATION && ZONE_DEVICE
647 config ARCH_ENABLE_HUGEPAGE_MIGRATION
650 config ARCH_ENABLE_THP_MIGRATION
653 config HUGETLB_PAGE_SIZE_VARIABLE
656 Allows the pageblock_order value to be dynamic instead of just standard
657 HUGETLB_PAGE_ORDER when there are multiple HugeTLB page sizes available
660 Note that the pageblock_order cannot exceed MAX_PAGE_ORDER and will be
661 clamped down to MAX_PAGE_ORDER.
664 def_bool (MEMORY_ISOLATION && COMPACTION) || CMA
666 config PCP_BATCH_SCALE_MAX
667 int "Maximum scale factor of PCP (Per-CPU pageset) batch allocate/free"
671 In page allocator, PCP (Per-CPU pageset) is refilled and drained in
672 batches. The batch number is scaled automatically to improve page
673 allocation/free throughput. But too large scale factor may hurt
674 latency. This option sets the upper limit of scale factor to limit
677 config PHYS_ADDR_T_64BIT
681 bool "Enable bounce buffers"
683 depends on BLOCK && MMU && HIGHMEM
685 Enable bounce buffers for devices that cannot access the full range of
686 memory available to the CPU. Enabled by default when HIGHMEM is
687 selected, but you may say n to override this.
694 bool "Enable KSM for page merging"
698 Enable Kernel Samepage Merging: KSM periodically scans those areas
699 of an application's address space that an app has advised may be
700 mergeable. When it finds pages of identical content, it replaces
701 the many instances by a single page with that content, so
702 saving memory until one or another app needs to modify the content.
703 Recommended for use with KVM, or with other duplicative applications.
704 See Documentation/mm/ksm.rst for more information: KSM is inactive
705 until a program has madvised that an area is MADV_MERGEABLE, and
706 root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set).
708 config DEFAULT_MMAP_MIN_ADDR
709 int "Low address space to protect from user allocation"
713 This is the portion of low virtual memory which should be protected
714 from userspace allocation. Keeping a user from writing to low pages
715 can help reduce the impact of kernel NULL pointer bugs.
717 For most ppc64 and x86 users with lots of address space
718 a value of 65536 is reasonable and should cause no problems.
719 On arm and other archs it should not be higher than 32768.
720 Programs which use vm86 functionality or have some need to map
721 this low address space will need CAP_SYS_RAWIO or disable this
722 protection by setting the value to 0.
724 This value can be changed after boot using the
725 /proc/sys/vm/mmap_min_addr tunable.
727 config ARCH_SUPPORTS_MEMORY_FAILURE
730 config MEMORY_FAILURE
732 depends on ARCH_SUPPORTS_MEMORY_FAILURE
733 bool "Enable recovery from hardware memory errors"
734 select MEMORY_ISOLATION
737 Enables code to recover from some memory failures on systems
738 with MCA recovery. This allows a system to continue running
739 even when some of its memory has uncorrected errors. This requires
740 special hardware support and typically ECC memory.
742 config HWPOISON_INJECT
743 tristate "HWPoison pages injector"
744 depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS
745 select PROC_PAGE_MONITOR
747 config NOMMU_INITIAL_TRIM_EXCESS
748 int "Turn on mmap() excess space trimming before booting"
752 The NOMMU mmap() frequently needs to allocate large contiguous chunks
753 of memory on which to store mappings, but it can only ask the system
754 allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently
755 more than it requires. To deal with this, mmap() is able to trim off
756 the excess and return it to the allocator.
758 If trimming is enabled, the excess is trimmed off and returned to the
759 system allocator, which can cause extra fragmentation, particularly
760 if there are a lot of transient processes.
762 If trimming is disabled, the excess is kept, but not used, which for
763 long-term mappings means that the space is wasted.
765 Trimming can be dynamically controlled through a sysctl option
766 (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of
767 excess pages there must be before trimming should occur, or zero if
768 no trimming is to occur.
770 This option specifies the initial value of this option. The default
771 of 1 says that all excess pages should be trimmed.
773 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
775 config ARCH_WANT_GENERAL_HUGETLB
778 config ARCH_WANTS_THP_SWAP
781 menuconfig TRANSPARENT_HUGEPAGE
782 bool "Transparent Hugepage Support"
783 depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT
787 Transparent Hugepages allows the kernel to use huge pages and
788 huge tlb transparently to the applications whenever possible.
789 This feature can improve computing performance to certain
790 applications by speeding up page faults during memory
791 allocation, by reducing the number of tlb misses and by speeding
792 up the pagetable walking.
794 If memory constrained on embedded, you may want to say N.
796 if TRANSPARENT_HUGEPAGE
799 prompt "Transparent Hugepage Support sysfs defaults"
800 depends on TRANSPARENT_HUGEPAGE
801 default TRANSPARENT_HUGEPAGE_ALWAYS
803 Selects the sysfs defaults for Transparent Hugepage Support.
805 config TRANSPARENT_HUGEPAGE_ALWAYS
808 Enabling Transparent Hugepage always, can increase the
809 memory footprint of applications without a guaranteed
810 benefit but it will work automatically for all applications.
812 config TRANSPARENT_HUGEPAGE_MADVISE
815 Enabling Transparent Hugepage madvise, will only provide a
816 performance improvement benefit to the applications using
817 madvise(MADV_HUGEPAGE) but it won't risk to increase the
818 memory footprint of applications without a guaranteed
821 config TRANSPARENT_HUGEPAGE_NEVER
824 Disable Transparent Hugepage by default. It can still be
825 enabled at runtime via sysfs.
830 depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP && 64BIT
832 Swap transparent huge pages in one piece, without splitting.
833 XXX: For now, swap cluster backing transparent huge page
834 will be split after swapout.
836 For selection by architectures with reasonable THP sizes.
838 config READ_ONLY_THP_FOR_FS
839 bool "Read-only THP for filesystems (EXPERIMENTAL)"
840 depends on TRANSPARENT_HUGEPAGE && SHMEM
843 Allow khugepaged to put read-only file-backed pages in THP.
845 This is marked experimental because it is a new feature. Write
846 support of file THPs will be developed in the next few release
849 endif # TRANSPARENT_HUGEPAGE
852 # UP and nommu archs use km based percpu allocator
854 config NEED_PER_CPU_KM
855 depends on !SMP || !MMU
859 config NEED_PER_CPU_EMBED_FIRST_CHUNK
862 config NEED_PER_CPU_PAGE_FIRST_CHUNK
865 config USE_PERCPU_NUMA_NODE_ID
868 config HAVE_SETUP_PER_CPU_AREA
872 bool "Contiguous Memory Allocator"
875 select MEMORY_ISOLATION
877 This enables the Contiguous Memory Allocator which allows other
878 subsystems to allocate big physically-contiguous blocks of memory.
879 CMA reserves a region of memory and allows only movable pages to
880 be allocated from it. This way, the kernel can use the memory for
881 pagecache and when a subsystem requests for contiguous area, the
882 allocated pages are migrated away to serve the contiguous request.
887 bool "CMA debugfs interface"
888 depends on CMA && DEBUG_FS
890 Turns on the DebugFS interface for CMA.
893 bool "CMA information through sysfs interface"
894 depends on CMA && SYSFS
896 This option exposes some sysfs attributes to get information
900 int "Maximum count of the CMA areas"
905 CMA allows to create CMA areas for particular purpose, mainly,
906 used as device private area. This parameter sets the maximum
907 number of CMA area in the system.
909 If unsure, leave the default value "8" in UMA and "20" in NUMA.
911 config MEM_SOFT_DIRTY
912 bool "Track memory changes"
913 depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS
914 select PROC_PAGE_MONITOR
916 This option enables memory changes tracking by introducing a
917 soft-dirty bit on pte-s. This bit it set when someone writes
918 into a page just as regular dirty bit, but unlike the latter
919 it can be cleared by hands.
921 See Documentation/admin-guide/mm/soft-dirty.rst for more details.
923 config GENERIC_EARLY_IOREMAP
926 config STACK_MAX_DEFAULT_SIZE_MB
927 int "Default maximum user stack size for 32-bit processes (MB)"
930 depends on STACK_GROWSUP && (!64BIT || COMPAT)
932 This is the maximum stack size in Megabytes in the VM layout of 32-bit
933 user processes when the stack grows upwards (currently only on parisc
934 arch) when the RLIMIT_STACK hard limit is unlimited.
936 A sane initial value is 100 MB.
938 config DEFERRED_STRUCT_PAGE_INIT
939 bool "Defer initialisation of struct pages to kthreads"
941 depends on !NEED_PER_CPU_KM
945 Ordinarily all struct pages are initialised during early boot in a
946 single thread. On very large machines this can take a considerable
947 amount of time. If this option is set, large machines will bring up
948 a subset of memmap at boot and then initialise the rest in parallel.
949 This has a potential performance impact on tasks running early in the
950 lifetime of the system until these kthreads finish the
953 config PAGE_IDLE_FLAG
955 select PAGE_EXTENSION if !64BIT
957 This adds PG_idle and PG_young flags to 'struct page'. PTE Accessed
958 bit writers can set the state of the bit in the flags so that PTE
959 Accessed bit readers may avoid disturbance.
961 config IDLE_PAGE_TRACKING
962 bool "Enable idle page tracking"
963 depends on SYSFS && MMU
964 select PAGE_IDLE_FLAG
966 This feature allows to estimate the amount of user pages that have
967 not been touched during a given period of time. This information can
968 be useful to tune memory cgroup limits and/or for job placement
969 within a compute cluster.
971 See Documentation/admin-guide/mm/idle_page_tracking.rst for
974 # Architectures which implement cpu_dcache_is_aliasing() to query
975 # whether the data caches are aliased (VIVT or VIPT with dcache
976 # aliasing) need to select this.
977 config ARCH_HAS_CPU_CACHE_ALIASING
980 config ARCH_HAS_CACHE_LINE_SIZE
983 config ARCH_HAS_CURRENT_STACK_POINTER
986 In support of HARDENED_USERCOPY performing stack variable lifetime
987 checking, an architecture-agnostic way to find the stack pointer
988 is needed. Once an architecture defines an unsigned long global
989 register alias named "current_stack_pointer", this config can be
992 config ARCH_HAS_PTE_DEVMAP
995 config ARCH_HAS_ZONE_DMA_SET
999 bool "Support DMA zone" if ARCH_HAS_ZONE_DMA_SET
1000 default y if ARM64 || X86
1003 bool "Support DMA32 zone" if ARCH_HAS_ZONE_DMA_SET
1008 bool "Device memory (pmem, HMM, etc...) hotplug support"
1009 depends on MEMORY_HOTPLUG
1010 depends on MEMORY_HOTREMOVE
1011 depends on SPARSEMEM_VMEMMAP
1012 depends on ARCH_HAS_PTE_DEVMAP
1016 Device memory hotplug support allows for establishing pmem,
1017 or other device driver discovered memory regions, in the
1018 memmap. This allows pfn_to_page() lookups of otherwise
1019 "device-physical" addresses which is needed for using a DAX
1020 mapping in an O_DIRECT operation, among other things.
1022 If FS_DAX is enabled, then say Y.
1025 # Helpers to mirror range of the CPU page tables of a process into device page
1032 config GET_FREE_REGION
1033 depends on SPARSEMEM
1036 config DEVICE_PRIVATE
1037 bool "Unaddressable device memory (GPU memory, ...)"
1038 depends on ZONE_DEVICE
1039 select GET_FREE_REGION
1042 Allows creation of struct pages to represent unaddressable device
1043 memory; i.e., memory that is only accessible from the device (or
1044 group of devices). You likely also want to select HMM_MIRROR.
1049 config ARCH_USES_HIGH_VMA_FLAGS
1051 config ARCH_HAS_PKEYS
1054 config ARCH_USES_PG_ARCH_X
1057 Enable the definition of PG_arch_x page flags with x > 1. Only
1058 suitable for 64-bit architectures with CONFIG_FLATMEM or
1059 CONFIG_SPARSEMEM_VMEMMAP enabled, otherwise there may not be
1060 enough room for additional bits in page->flags.
1062 config VM_EVENT_COUNTERS
1064 bool "Enable VM event counters for /proc/vmstat" if EXPERT
1066 VM event counters are needed for event counts to be shown.
1067 This option allows the disabling of the VM event counters
1068 on EXPERT systems. /proc/vmstat will only show page counts
1069 if VM event counters are disabled.
1072 bool "Collect percpu memory statistics"
1074 This feature collects and exposes statistics via debugfs. The
1075 information includes global and per chunk statistics, which can
1076 be used to help understand percpu memory usage.
1079 bool "Enable infrastructure for get_user_pages()-related unit tests"
1082 Provides /sys/kernel/debug/gup_test, which in turn provides a way
1083 to make ioctl calls that can launch kernel-based unit tests for
1084 the get_user_pages*() and pin_user_pages*() family of API calls.
1086 These tests include benchmark testing of the _fast variants of
1087 get_user_pages*() and pin_user_pages*(), as well as smoke tests of
1088 the non-_fast variants.
1090 There is also a sub-test that allows running dump_page() on any
1091 of up to eight pages (selected by command line args) within the
1092 range of user-space addresses. These pages are either pinned via
1093 pin_user_pages*(), or pinned via get_user_pages*(), as specified
1094 by other command line arguments.
1096 See tools/testing/selftests/mm/gup_test.c
1098 comment "GUP_TEST needs to have DEBUG_FS enabled"
1099 depends on !GUP_TEST && !DEBUG_FS
1101 config GUP_GET_PXX_LOW_HIGH
1105 tristate "Enable a module to run time tests on dma_pool"
1108 Provides a test module that will allocate and free many blocks of
1109 various sizes and report how long it takes. This is intended to
1110 provide a consistent way to measure how changes to the
1111 dma_pool_alloc/free routines affect performance.
1113 config ARCH_HAS_PTE_SPECIAL
1117 # Some architectures require a special hugepage directory format that is
1118 # required to support multiple hugepage sizes. For example a4fe3ce76
1119 # "powerpc/mm: Allow more flexible layouts for hugepage pagetables"
1120 # introduced it on powerpc. This allows for a more flexible hugepage
1121 # pagetable layouts.
1123 config ARCH_HAS_HUGEPD
1126 config MAPPING_DIRTY_HELPERS
1132 config KMAP_LOCAL_NON_LINEAR_PTE_ARRAY
1135 # struct io_mapping based helper. Selected by drivers that need them
1140 bool "Enable memfd_create() system call" if EXPERT
1144 bool "Enable memfd_secret() system call" if EXPERT
1145 depends on ARCH_HAS_SET_DIRECT_MAP
1147 Enable the memfd_secret() system call with the ability to create
1148 memory areas visible only in the context of the owning process and
1149 not mapped to other processes and other kernel page tables.
1151 config ANON_VMA_NAME
1152 bool "Anonymous VMA name support"
1153 depends on PROC_FS && ADVISE_SYSCALLS && MMU
1156 Allow naming anonymous virtual memory areas.
1158 This feature allows assigning names to virtual memory areas. Assigned
1159 names can be later retrieved from /proc/pid/maps and /proc/pid/smaps
1160 and help identifying individual anonymous memory areas.
1161 Assigning a name to anonymous virtual memory area might prevent that
1162 area from being merged with adjacent virtual memory areas due to the
1163 difference in their name.
1165 config HAVE_ARCH_USERFAULTFD_WP
1168 Arch has userfaultfd write protection support
1170 config HAVE_ARCH_USERFAULTFD_MINOR
1173 Arch has userfaultfd minor fault support
1175 menuconfig USERFAULTFD
1176 bool "Enable userfaultfd() system call"
1179 Enable the userfaultfd() system call that allows to intercept and
1180 handle page faults in userland.
1183 config PTE_MARKER_UFFD_WP
1184 bool "Userfaultfd write protection support for shmem/hugetlbfs"
1186 depends on HAVE_ARCH_USERFAULTFD_WP
1189 Allows to create marker PTEs for userfaultfd write protection
1190 purposes. It is required to enable userfaultfd write protection on
1191 file-backed memory types like shmem and hugetlbfs.
1196 bool "Multi-Gen LRU"
1198 # make sure folio->flags has enough spare bits
1199 depends on 64BIT || !SPARSEMEM || SPARSEMEM_VMEMMAP
1201 A high performance LRU implementation to overcommit memory. See
1202 Documentation/admin-guide/mm/multigen_lru.rst for details.
1204 config LRU_GEN_ENABLED
1205 bool "Enable by default"
1208 This option enables the multi-gen LRU by default.
1210 config LRU_GEN_STATS
1211 bool "Full stats for debugging"
1214 Do not enable this option unless you plan to look at historical stats
1215 from evicted generations for debugging purpose.
1217 This option has a per-memcg and per-node memory overhead.
1219 config LRU_GEN_WALKS_MMU
1221 depends on LRU_GEN && ARCH_HAS_HW_PTE_YOUNG
1224 config ARCH_SUPPORTS_PER_VMA_LOCK
1229 depends on ARCH_SUPPORTS_PER_VMA_LOCK && MMU && SMP
1231 Allow per-vma locking during page fault handling.
1233 This feature allows locking each virtual memory area separately when
1234 handling page faults instead of taking mmap_lock.
1236 config LOCK_MM_AND_FIND_VMA
1238 depends on !STACK_GROWSUP
1240 config IOMMU_MM_DATA
1246 source "mm/damon/Kconfig"