1 # SPDX-License-Identifier: GPL-2.0-only
3 menu "Memory Management options"
5 config SELECT_MEMORY_MODEL
7 depends on ARCH_SELECT_MEMORY_MODEL
11 depends on SELECT_MEMORY_MODEL
12 default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT
13 default FLATMEM_MANUAL
15 This option allows you to change some of the ways that
16 Linux manages its memory internally. Most users will
17 only have one option here selected by the architecture
18 configuration. This is normal.
22 depends on !ARCH_SPARSEMEM_ENABLE || ARCH_FLATMEM_ENABLE
24 This option is best suited for non-NUMA systems with
25 flat address space. The FLATMEM is the most efficient
26 system in terms of performance and resource consumption
27 and it is the best option for smaller systems.
29 For systems that have holes in their physical address
30 spaces and for features like NUMA and memory hotplug,
31 choose "Sparse Memory".
33 If unsure, choose this option (Flat Memory) over any other.
35 config SPARSEMEM_MANUAL
37 depends on ARCH_SPARSEMEM_ENABLE
39 This will be the only option for some systems, including
40 memory hot-plug systems. This is normal.
42 This option provides efficient support for systems with
43 holes is their physical address space and allows memory
44 hot-plug and hot-remove.
46 If unsure, choose "Flat Memory" over this option.
52 depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL
56 depends on !SPARSEMEM || FLATMEM_MANUAL
59 # SPARSEMEM_EXTREME (which is the default) does some bootmem
60 # allocations when sparse_init() is called. If this cannot
61 # be done on your architecture, select this option. However,
62 # statically allocating the mem_section[] array can potentially
63 # consume vast quantities of .bss, so be careful.
65 # This option will also potentially produce smaller runtime code
66 # with gcc 3.4 and later.
68 config SPARSEMEM_STATIC
72 # Architecture platforms which require a two level mem_section in SPARSEMEM
73 # must select this option. This is usually for architecture platforms with
74 # an extremely sparse physical address space.
76 config SPARSEMEM_EXTREME
78 depends on SPARSEMEM && !SPARSEMEM_STATIC
80 config SPARSEMEM_VMEMMAP_ENABLE
83 config SPARSEMEM_VMEMMAP
84 bool "Sparse Memory virtual memmap"
85 depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE
88 SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise
89 pfn_to_page and page_to_pfn operations. This is the most
90 efficient option when sufficient kernel resources are available.
92 config HAVE_MEMBLOCK_PHYS_MAP
99 # Don't discard allocated memory used to track "memory" and "reserved" memblocks
100 # after early boot, so it can still be used to test for validity of memory.
101 # Also, memblocks are updated with memory hot(un)plug.
102 config ARCH_KEEP_MEMBLOCK
105 # Keep arch NUMA mapping infrastructure post-init.
106 config NUMA_KEEP_MEMINFO
109 config MEMORY_ISOLATION
113 # Only be set on architectures that have completely implemented memory hotplug
114 # feature. If you are not sure, don't touch it.
116 config HAVE_BOOTMEM_INFO_NODE
119 config ARCH_ENABLE_MEMORY_HOTPLUG
122 # eventually, we can have this option just 'select SPARSEMEM'
123 config MEMORY_HOTPLUG
124 bool "Allow for memory hot-add"
125 select MEMORY_ISOLATION
127 depends on ARCH_ENABLE_MEMORY_HOTPLUG
129 select NUMA_KEEP_MEMINFO if NUMA
131 config MEMORY_HOTPLUG_DEFAULT_ONLINE
132 bool "Online the newly added memory blocks by default"
133 depends on MEMORY_HOTPLUG
135 This option sets the default policy setting for memory hotplug
136 onlining policy (/sys/devices/system/memory/auto_online_blocks) which
137 determines what happens to newly added memory regions. Policy setting
138 can always be changed at runtime.
139 See Documentation/admin-guide/mm/memory-hotplug.rst for more information.
141 Say Y here if you want all hot-plugged memory blocks to appear in
142 'online' state by default.
143 Say N here if you want the default policy to keep all hot-plugged
144 memory blocks in 'offline' state.
146 config ARCH_ENABLE_MEMORY_HOTREMOVE
149 config MEMORY_HOTREMOVE
150 bool "Allow for memory hot remove"
151 select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64)
152 depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE
155 config MHP_MEMMAP_ON_MEMORY
157 depends on MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP
158 depends on ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
160 # Heavily threaded applications may benefit from splitting the mm-wide
161 # page_table_lock, so that faults on different parts of the user address
162 # space can be handled with less contention: split it at this NR_CPUS.
163 # Default to 4 for wider testing, though 8 might be more appropriate.
164 # ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
165 # PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
166 # SPARC32 allocates multiple pte tables within a single page, and therefore
167 # a per-page lock leads to problems when multiple tables need to be locked
168 # at the same time (e.g. copy_page_range()).
169 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
171 config SPLIT_PTLOCK_CPUS
173 default "999999" if !MMU
174 default "999999" if ARM && !CPU_CACHE_VIPT
175 default "999999" if PARISC && !PA20
176 default "999999" if SPARC32
179 config ARCH_ENABLE_SPLIT_PMD_PTLOCK
183 # support for memory balloon
184 config MEMORY_BALLOON
188 # support for memory balloon compaction
189 config BALLOON_COMPACTION
190 bool "Allow for balloon memory compaction/migration"
192 depends on COMPACTION && MEMORY_BALLOON
194 Memory fragmentation introduced by ballooning might reduce
195 significantly the number of 2MB contiguous memory blocks that can be
196 used within a guest, thus imposing performance penalties associated
197 with the reduced number of transparent huge pages that could be used
198 by the guest workload. Allowing the compaction & migration for memory
199 pages enlisted as being part of memory balloon devices avoids the
200 scenario aforementioned and helps improving memory defragmentation.
203 # support for memory compaction
205 bool "Allow for memory compaction"
210 Compaction is the only memory management component to form
211 high order (larger physically contiguous) memory blocks
212 reliably. The page allocator relies on compaction heavily and
213 the lack of the feature can lead to unexpected OOM killer
214 invocations for high order memory requests. You shouldn't
215 disable this option unless there really is a strong reason for
216 it and then we would be really interested to hear about that at
220 # support for free page reporting
221 config PAGE_REPORTING
222 bool "Free page reporting"
225 Free page reporting allows for the incremental acquisition of
226 free pages from the buddy allocator for the purpose of reporting
227 those pages to another entity, such as a hypervisor, so that the
228 memory can be freed within the host for other uses.
231 # support for page migration
234 bool "Page migration"
236 depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU
238 Allows the migration of the physical location of pages of processes
239 while the virtual addresses are not changed. This is useful in
240 two situations. The first is on NUMA systems to put pages nearer
241 to the processors accessing. The second is when allocating huge
242 pages as migration can relocate pages to satisfy a huge page
243 allocation instead of reclaiming.
245 config ARCH_ENABLE_HUGEPAGE_MIGRATION
248 config ARCH_ENABLE_THP_MIGRATION
251 config HUGETLB_PAGE_SIZE_VARIABLE
254 Allows the pageblock_order value to be dynamic instead of just standard
255 HUGETLB_PAGE_ORDER when there are multiple HugeTLB page sizes available
259 def_bool (MEMORY_ISOLATION && COMPACTION) || CMA
261 config PHYS_ADDR_T_64BIT
265 bool "Enable bounce buffers"
267 depends on BLOCK && MMU && HIGHMEM
269 Enable bounce buffers for devices that cannot access the full range of
270 memory available to the CPU. Enabled by default when HIGHMEM is
271 selected, but you may say n to override this.
276 An architecture should select this if it implements the
277 deprecated interface virt_to_bus(). All new architectures
278 should probably not select this.
287 bool "Enable KSM for page merging"
291 Enable Kernel Samepage Merging: KSM periodically scans those areas
292 of an application's address space that an app has advised may be
293 mergeable. When it finds pages of identical content, it replaces
294 the many instances by a single page with that content, so
295 saving memory until one or another app needs to modify the content.
296 Recommended for use with KVM, or with other duplicative applications.
297 See Documentation/vm/ksm.rst for more information: KSM is inactive
298 until a program has madvised that an area is MADV_MERGEABLE, and
299 root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set).
301 config DEFAULT_MMAP_MIN_ADDR
302 int "Low address space to protect from user allocation"
306 This is the portion of low virtual memory which should be protected
307 from userspace allocation. Keeping a user from writing to low pages
308 can help reduce the impact of kernel NULL pointer bugs.
310 For most ia64, ppc64 and x86 users with lots of address space
311 a value of 65536 is reasonable and should cause no problems.
312 On arm and other archs it should not be higher than 32768.
313 Programs which use vm86 functionality or have some need to map
314 this low address space will need CAP_SYS_RAWIO or disable this
315 protection by setting the value to 0.
317 This value can be changed after boot using the
318 /proc/sys/vm/mmap_min_addr tunable.
320 config ARCH_SUPPORTS_MEMORY_FAILURE
323 config MEMORY_FAILURE
325 depends on ARCH_SUPPORTS_MEMORY_FAILURE
326 bool "Enable recovery from hardware memory errors"
327 select MEMORY_ISOLATION
330 Enables code to recover from some memory failures on systems
331 with MCA recovery. This allows a system to continue running
332 even when some of its memory has uncorrected errors. This requires
333 special hardware support and typically ECC memory.
335 config HWPOISON_INJECT
336 tristate "HWPoison pages injector"
337 depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS
338 select PROC_PAGE_MONITOR
340 config NOMMU_INITIAL_TRIM_EXCESS
341 int "Turn on mmap() excess space trimming before booting"
345 The NOMMU mmap() frequently needs to allocate large contiguous chunks
346 of memory on which to store mappings, but it can only ask the system
347 allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently
348 more than it requires. To deal with this, mmap() is able to trim off
349 the excess and return it to the allocator.
351 If trimming is enabled, the excess is trimmed off and returned to the
352 system allocator, which can cause extra fragmentation, particularly
353 if there are a lot of transient processes.
355 If trimming is disabled, the excess is kept, but not used, which for
356 long-term mappings means that the space is wasted.
358 Trimming can be dynamically controlled through a sysctl option
359 (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of
360 excess pages there must be before trimming should occur, or zero if
361 no trimming is to occur.
363 This option specifies the initial value of this option. The default
364 of 1 says that all excess pages should be trimmed.
366 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
368 config TRANSPARENT_HUGEPAGE
369 bool "Transparent Hugepage Support"
370 depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT
374 Transparent Hugepages allows the kernel to use huge pages and
375 huge tlb transparently to the applications whenever possible.
376 This feature can improve computing performance to certain
377 applications by speeding up page faults during memory
378 allocation, by reducing the number of tlb misses and by speeding
379 up the pagetable walking.
381 If memory constrained on embedded, you may want to say N.
384 prompt "Transparent Hugepage Support sysfs defaults"
385 depends on TRANSPARENT_HUGEPAGE
386 default TRANSPARENT_HUGEPAGE_ALWAYS
388 Selects the sysfs defaults for Transparent Hugepage Support.
390 config TRANSPARENT_HUGEPAGE_ALWAYS
393 Enabling Transparent Hugepage always, can increase the
394 memory footprint of applications without a guaranteed
395 benefit but it will work automatically for all applications.
397 config TRANSPARENT_HUGEPAGE_MADVISE
400 Enabling Transparent Hugepage madvise, will only provide a
401 performance improvement benefit to the applications using
402 madvise(MADV_HUGEPAGE) but it won't risk to increase the
403 memory footprint of applications without a guaranteed
407 config ARCH_WANTS_THP_SWAP
412 depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP
414 Swap transparent huge pages in one piece, without splitting.
415 XXX: For now, swap cluster backing transparent huge page
416 will be split after swapout.
418 For selection by architectures with reasonable THP sizes.
421 # UP and nommu archs use km based percpu allocator
423 config NEED_PER_CPU_KM
429 bool "Enable cleancache driver to cache clean pages if tmem is present"
431 Cleancache can be thought of as a page-granularity victim cache
432 for clean pages that the kernel's pageframe replacement algorithm
433 (PFRA) would like to keep around, but can't since there isn't enough
434 memory. So when the PFRA "evicts" a page, it first attempts to use
435 cleancache code to put the data contained in that page into
436 "transcendent memory", memory that is not directly accessible or
437 addressable by the kernel and is of unknown and possibly
438 time-varying size. And when a cleancache-enabled
439 filesystem wishes to access a page in a file on disk, it first
440 checks cleancache to see if it already contains it; if it does,
441 the page is copied into the kernel and a disk access is avoided.
442 When a transcendent memory driver is available (such as zcache or
443 Xen transcendent memory), a significant I/O reduction
444 may be achieved. When none is available, all cleancache calls
445 are reduced to a single pointer-compare-against-NULL resulting
446 in a negligible performance hit.
448 If unsure, say Y to enable cleancache
451 bool "Enable frontswap to cache swap pages if tmem is present"
454 Frontswap is so named because it can be thought of as the opposite
455 of a "backing" store for a swap device. The data is stored into
456 "transcendent memory", memory that is not directly accessible or
457 addressable by the kernel and is of unknown and possibly
458 time-varying size. When space in transcendent memory is available,
459 a significant swap I/O reduction may be achieved. When none is
460 available, all frontswap calls are reduced to a single pointer-
461 compare-against-NULL resulting in a negligible performance hit
462 and swap data is stored as normal on the matching swap device.
464 If unsure, say Y to enable frontswap.
467 bool "Contiguous Memory Allocator"
470 select MEMORY_ISOLATION
472 This enables the Contiguous Memory Allocator which allows other
473 subsystems to allocate big physically-contiguous blocks of memory.
474 CMA reserves a region of memory and allows only movable pages to
475 be allocated from it. This way, the kernel can use the memory for
476 pagecache and when a subsystem requests for contiguous area, the
477 allocated pages are migrated away to serve the contiguous request.
482 bool "CMA debug messages (DEVELOPMENT)"
483 depends on DEBUG_KERNEL && CMA
485 Turns on debug messages in CMA. This produces KERN_DEBUG
486 messages for every CMA call as well as various messages while
487 processing calls such as dma_alloc_from_contiguous().
488 This option does not affect warning and error messages.
491 bool "CMA debugfs interface"
492 depends on CMA && DEBUG_FS
494 Turns on the DebugFS interface for CMA.
497 bool "CMA information through sysfs interface"
498 depends on CMA && SYSFS
500 This option exposes some sysfs attributes to get information
504 int "Maximum count of the CMA areas"
509 CMA allows to create CMA areas for particular purpose, mainly,
510 used as device private area. This parameter sets the maximum
511 number of CMA area in the system.
513 If unsure, leave the default value "7" in UMA and "19" in NUMA.
515 config MEM_SOFT_DIRTY
516 bool "Track memory changes"
517 depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS
518 select PROC_PAGE_MONITOR
520 This option enables memory changes tracking by introducing a
521 soft-dirty bit on pte-s. This bit it set when someone writes
522 into a page just as regular dirty bit, but unlike the latter
523 it can be cleared by hands.
525 See Documentation/admin-guide/mm/soft-dirty.rst for more details.
528 bool "Compressed cache for swap pages (EXPERIMENTAL)"
529 depends on FRONTSWAP && CRYPTO=y
532 A lightweight compressed cache for swap pages. It takes
533 pages that are in the process of being swapped out and attempts to
534 compress them into a dynamically allocated RAM-based memory pool.
535 This can result in a significant I/O reduction on swap device and,
536 in the case where decompressing from RAM is faster that swap device
537 reads, can also improve workload performance.
539 This is marked experimental because it is a new feature (as of
540 v3.11) that interacts heavily with memory reclaim. While these
541 interactions don't cause any known issues on simple memory setups,
542 they have not be fully explored on the large set of potential
543 configurations and workloads that exist.
546 prompt "Compressed cache for swap pages default compressor"
548 default ZSWAP_COMPRESSOR_DEFAULT_LZO
550 Selects the default compression algorithm for the compressed cache
553 For an overview what kind of performance can be expected from
554 a particular compression algorithm please refer to the benchmarks
555 available at the following LWN page:
556 https://lwn.net/Articles/751795/
558 If in doubt, select 'LZO'.
560 The selection made here can be overridden by using the kernel
561 command line 'zswap.compressor=' option.
563 config ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
565 select CRYPTO_DEFLATE
567 Use the Deflate algorithm as the default compression algorithm.
569 config ZSWAP_COMPRESSOR_DEFAULT_LZO
573 Use the LZO algorithm as the default compression algorithm.
575 config ZSWAP_COMPRESSOR_DEFAULT_842
579 Use the 842 algorithm as the default compression algorithm.
581 config ZSWAP_COMPRESSOR_DEFAULT_LZ4
585 Use the LZ4 algorithm as the default compression algorithm.
587 config ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
591 Use the LZ4HC algorithm as the default compression algorithm.
593 config ZSWAP_COMPRESSOR_DEFAULT_ZSTD
597 Use the zstd algorithm as the default compression algorithm.
600 config ZSWAP_COMPRESSOR_DEFAULT
603 default "deflate" if ZSWAP_COMPRESSOR_DEFAULT_DEFLATE
604 default "lzo" if ZSWAP_COMPRESSOR_DEFAULT_LZO
605 default "842" if ZSWAP_COMPRESSOR_DEFAULT_842
606 default "lz4" if ZSWAP_COMPRESSOR_DEFAULT_LZ4
607 default "lz4hc" if ZSWAP_COMPRESSOR_DEFAULT_LZ4HC
608 default "zstd" if ZSWAP_COMPRESSOR_DEFAULT_ZSTD
612 prompt "Compressed cache for swap pages default allocator"
614 default ZSWAP_ZPOOL_DEFAULT_ZBUD
616 Selects the default allocator for the compressed cache for
618 The default is 'zbud' for compatibility, however please do
619 read the description of each of the allocators below before
620 making a right choice.
622 The selection made here can be overridden by using the kernel
623 command line 'zswap.zpool=' option.
625 config ZSWAP_ZPOOL_DEFAULT_ZBUD
629 Use the zbud allocator as the default allocator.
631 config ZSWAP_ZPOOL_DEFAULT_Z3FOLD
635 Use the z3fold allocator as the default allocator.
637 config ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
641 Use the zsmalloc allocator as the default allocator.
644 config ZSWAP_ZPOOL_DEFAULT
647 default "zbud" if ZSWAP_ZPOOL_DEFAULT_ZBUD
648 default "z3fold" if ZSWAP_ZPOOL_DEFAULT_Z3FOLD
649 default "zsmalloc" if ZSWAP_ZPOOL_DEFAULT_ZSMALLOC
652 config ZSWAP_DEFAULT_ON
653 bool "Enable the compressed cache for swap pages by default"
656 If selected, the compressed cache for swap pages will be enabled
657 at boot, otherwise it will be disabled.
659 The selection made here can be overridden by using the kernel
660 command line 'zswap.enabled=' option.
663 tristate "Common API for compressed memory storage"
665 Compressed memory storage API. This allows using either zbud or
669 tristate "Low (Up to 2x) density storage for compressed pages"
672 A special purpose allocator for storing compressed pages.
673 It is designed to store up to two compressed pages per physical
674 page. While this design limits storage density, it has simple and
675 deterministic reclaim properties that make it preferable to a higher
676 density approach when reclaim will be used.
679 tristate "Up to 3x density storage for compressed pages"
682 A special purpose allocator for storing compressed pages.
683 It is designed to store up to three compressed pages per physical
684 page. It is a ZBUD derivative so the simplicity and determinism are
688 tristate "Memory allocator for compressed pages"
691 zsmalloc is a slab-based memory allocator designed to store
692 compressed RAM pages. zsmalloc uses virtual memory mapping
693 in order to reduce fragmentation. However, this results in a
694 non-standard allocator interface where a handle, not a pointer, is
695 returned by an alloc(). This handle must be mapped in order to
696 access the allocated space.
699 bool "Export zsmalloc statistics"
703 This option enables code in the zsmalloc to collect various
704 statistics about what's happening in zsmalloc and exports that
705 information to userspace via debugfs.
708 config GENERIC_EARLY_IOREMAP
711 config STACK_MAX_DEFAULT_SIZE_MB
712 int "Default maximum user stack size for 32-bit processes (MB)"
715 depends on STACK_GROWSUP && (!64BIT || COMPAT)
717 This is the maximum stack size in Megabytes in the VM layout of 32-bit
718 user processes when the stack grows upwards (currently only on parisc
719 arch) when the RLIMIT_STACK hard limit is unlimited.
721 A sane initial value is 100 MB.
723 config DEFERRED_STRUCT_PAGE_INIT
724 bool "Defer initialisation of struct pages to kthreads"
726 depends on !NEED_PER_CPU_KM
730 Ordinarily all struct pages are initialised during early boot in a
731 single thread. On very large machines this can take a considerable
732 amount of time. If this option is set, large machines will bring up
733 a subset of memmap at boot and then initialise the rest in parallel.
734 This has a potential performance impact on tasks running early in the
735 lifetime of the system until these kthreads finish the
738 config PAGE_IDLE_FLAG
740 select PAGE_EXTENSION if !64BIT
742 This adds PG_idle and PG_young flags to 'struct page'. PTE Accessed
743 bit writers can set the state of the bit in the flags so that PTE
744 Accessed bit readers may avoid disturbance.
746 config IDLE_PAGE_TRACKING
747 bool "Enable idle page tracking"
748 depends on SYSFS && MMU
749 select PAGE_IDLE_FLAG
751 This feature allows to estimate the amount of user pages that have
752 not been touched during a given period of time. This information can
753 be useful to tune memory cgroup limits and/or for job placement
754 within a compute cluster.
756 See Documentation/admin-guide/mm/idle_page_tracking.rst for
759 config ARCH_HAS_CACHE_LINE_SIZE
762 config ARCH_HAS_PTE_DEVMAP
765 config ARCH_HAS_ZONE_DMA_SET
769 bool "Support DMA zone" if ARCH_HAS_ZONE_DMA_SET
770 default y if ARM64 || X86
773 bool "Support DMA32 zone" if ARCH_HAS_ZONE_DMA_SET
778 bool "Device memory (pmem, HMM, etc...) hotplug support"
779 depends on MEMORY_HOTPLUG
780 depends on MEMORY_HOTREMOVE
781 depends on SPARSEMEM_VMEMMAP
782 depends on ARCH_HAS_PTE_DEVMAP
786 Device memory hotplug support allows for establishing pmem,
787 or other device driver discovered memory regions, in the
788 memmap. This allows pfn_to_page() lookups of otherwise
789 "device-physical" addresses which is needed for using a DAX
790 mapping in an O_DIRECT operation, among other things.
792 If FS_DAX is enabled, then say Y.
794 config DEV_PAGEMAP_OPS
798 # Helpers to mirror range of the CPU page tables of a process into device page
805 config DEVICE_PRIVATE
806 bool "Unaddressable device memory (GPU memory, ...)"
807 depends on ZONE_DEVICE
808 select DEV_PAGEMAP_OPS
811 Allows creation of struct pages to represent unaddressable device
812 memory; i.e., memory that is only accessible from the device (or
813 group of devices). You likely also want to select HMM_MIRROR.
818 config ARCH_USES_HIGH_VMA_FLAGS
820 config ARCH_HAS_PKEYS
824 bool "Collect percpu memory statistics"
826 This feature collects and exposes statistics via debugfs. The
827 information includes global and per chunk statistics, which can
828 be used to help understand percpu memory usage.
831 bool "Enable infrastructure for get_user_pages()-related unit tests"
834 Provides /sys/kernel/debug/gup_test, which in turn provides a way
835 to make ioctl calls that can launch kernel-based unit tests for
836 the get_user_pages*() and pin_user_pages*() family of API calls.
838 These tests include benchmark testing of the _fast variants of
839 get_user_pages*() and pin_user_pages*(), as well as smoke tests of
840 the non-_fast variants.
842 There is also a sub-test that allows running dump_page() on any
843 of up to eight pages (selected by command line args) within the
844 range of user-space addresses. These pages are either pinned via
845 pin_user_pages*(), or pinned via get_user_pages*(), as specified
846 by other command line arguments.
848 See tools/testing/selftests/vm/gup_test.c
850 comment "GUP_TEST needs to have DEBUG_FS enabled"
851 depends on !GUP_TEST && !DEBUG_FS
853 config GUP_GET_PTE_LOW_HIGH
856 config READ_ONLY_THP_FOR_FS
857 bool "Read-only THP for filesystems (EXPERIMENTAL)"
858 depends on TRANSPARENT_HUGEPAGE && SHMEM
861 Allow khugepaged to put read-only file-backed pages in THP.
863 This is marked experimental because it is a new feature. Write
864 support of file THPs will be developed in the next few release
867 config ARCH_HAS_PTE_SPECIAL
871 # Some architectures require a special hugepage directory format that is
872 # required to support multiple hugepage sizes. For example a4fe3ce76
873 # "powerpc/mm: Allow more flexible layouts for hugepage pagetables"
874 # introduced it on powerpc. This allows for a more flexible hugepage
877 config ARCH_HAS_HUGEPD
880 config MAPPING_DIRTY_HELPERS
886 # struct io_mapping based helper. Selected by drivers that need them
891 def_bool ARCH_HAS_SET_DIRECT_MAP && !EMBEDDED
893 source "mm/damon/Kconfig"