1 config SELECT_MEMORY_MODEL
3 depends on ARCH_SELECT_MEMORY_MODEL
7 depends on SELECT_MEMORY_MODEL
8 default DISCONTIGMEM_MANUAL if ARCH_DISCONTIGMEM_DEFAULT
9 default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT
10 default FLATMEM_MANUAL
14 depends on !(ARCH_DISCONTIGMEM_ENABLE || ARCH_SPARSEMEM_ENABLE) || ARCH_FLATMEM_ENABLE
16 This option allows you to change some of the ways that
17 Linux manages its memory internally. Most users will
18 only have one option here: FLATMEM. This is normal
21 Some users of more advanced features like NUMA and
22 memory hotplug may have different options here.
23 DISCONTIGMEM is a more mature, better tested system,
24 but is incompatible with memory hotplug and may suffer
25 decreased performance over SPARSEMEM. If unsure between
26 "Sparse Memory" and "Discontiguous Memory", choose
27 "Discontiguous Memory".
29 If unsure, choose this option (Flat Memory) over any other.
31 config DISCONTIGMEM_MANUAL
32 bool "Discontiguous Memory"
33 depends on ARCH_DISCONTIGMEM_ENABLE
35 This option provides enhanced support for discontiguous
36 memory systems, over FLATMEM. These systems have holes
37 in their physical address spaces, and this option provides
38 more efficient handling of these holes. However, the vast
39 majority of hardware has quite flat address spaces, and
40 can have degraded performance from the extra overhead that
43 Many NUMA configurations will have this as the only option.
45 If unsure, choose "Flat Memory" over this option.
47 config SPARSEMEM_MANUAL
49 depends on ARCH_SPARSEMEM_ENABLE
51 This will be the only option for some systems, including
52 memory hotplug systems. This is normal.
54 For many other systems, this will be an alternative to
55 "Discontiguous Memory". This option provides some potential
56 performance benefits, along with decreased code complexity,
57 but it is newer, and more experimental.
59 If unsure, choose "Discontiguous Memory" or "Flat Memory"
66 depends on (!SELECT_MEMORY_MODEL && ARCH_DISCONTIGMEM_ENABLE) || DISCONTIGMEM_MANUAL
70 depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL
74 depends on (!DISCONTIGMEM && !SPARSEMEM) || FLATMEM_MANUAL
76 config FLAT_NODE_MEM_MAP
81 # Both the NUMA code and DISCONTIGMEM use arrays of pg_data_t's
82 # to represent different areas of memory. This variable allows
83 # those dependencies to exist individually.
85 config NEED_MULTIPLE_NODES
87 depends on DISCONTIGMEM || NUMA
89 config HAVE_MEMORY_PRESENT
91 depends on ARCH_HAVE_MEMORY_PRESENT || SPARSEMEM
94 # SPARSEMEM_EXTREME (which is the default) does some bootmem
95 # allocations when memory_present() is called. If this cannot
96 # be done on your architecture, select this option. However,
97 # statically allocating the mem_section[] array can potentially
98 # consume vast quantities of .bss, so be careful.
100 # This option will also potentially produce smaller runtime code
101 # with gcc 3.4 and later.
103 config SPARSEMEM_STATIC
107 # Architecture platforms which require a two level mem_section in SPARSEMEM
108 # must select this option. This is usually for architecture platforms with
109 # an extremely sparse physical address space.
111 config SPARSEMEM_EXTREME
113 depends on SPARSEMEM && !SPARSEMEM_STATIC
115 config SPARSEMEM_VMEMMAP_ENABLE
118 config SPARSEMEM_ALLOC_MEM_MAP_TOGETHER
120 depends on SPARSEMEM && X86_64
122 config SPARSEMEM_VMEMMAP
123 bool "Sparse Memory virtual memmap"
124 depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE
127 SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise
128 pfn_to_page and page_to_pfn operations. This is the most
129 efficient option when sufficient kernel resources are available.
134 config HAVE_MEMBLOCK_NODE_MAP
137 config HAVE_MEMBLOCK_PHYS_MAP
140 config HAVE_GENERIC_GUP
143 config ARCH_DISCARD_MEMBLOCK
149 config MEMORY_ISOLATION
153 bool "Enable to assign a node which has only movable memory"
154 depends on HAVE_MEMBLOCK
155 depends on NO_BOOTMEM
156 depends on X86_64 || OF_EARLY_FLATTREE || MEMORY_HOTPLUG
160 Allow a node to have only movable memory. Pages used by the kernel,
161 such as direct mapping pages cannot be migrated. So the corresponding
162 memory device cannot be hotplugged. This option allows the following
164 - When the system is booting, node full of hotpluggable memory can
165 be arranged to have only movable memory so that the whole node can
166 be hot-removed. (need movable_node boot option specified).
167 - After the system is up, the option allows users to online all the
168 memory of a node as movable memory so that the whole node can be
171 Users who don't use the memory hotplug feature are fine with this
172 option on since they don't specify movable_node boot option or they
173 don't online memory as movable.
175 Say Y here if you want to hotplug a whole node.
176 Say N here if you want kernel to use memory on all nodes evenly.
179 # Only be set on architectures that have completely implemented memory hotplug
180 # feature. If you are not sure, don't touch it.
182 config HAVE_BOOTMEM_INFO_NODE
185 # eventually, we can have this option just 'select SPARSEMEM'
186 config MEMORY_HOTPLUG
187 bool "Allow for memory hot-add"
188 depends on SPARSEMEM || X86_64_ACPI_NUMA
189 depends on ARCH_ENABLE_MEMORY_HOTPLUG
190 depends on COMPILE_TEST || !KASAN
192 config MEMORY_HOTPLUG_SPARSE
194 depends on SPARSEMEM && MEMORY_HOTPLUG
196 config MEMORY_HOTPLUG_DEFAULT_ONLINE
197 bool "Online the newly added memory blocks by default"
199 depends on MEMORY_HOTPLUG
201 This option sets the default policy setting for memory hotplug
202 onlining policy (/sys/devices/system/memory/auto_online_blocks) which
203 determines what happens to newly added memory regions. Policy setting
204 can always be changed at runtime.
205 See Documentation/memory-hotplug.txt for more information.
207 Say Y here if you want all hot-plugged memory blocks to appear in
208 'online' state by default.
209 Say N here if you want the default policy to keep all hot-plugged
210 memory blocks in 'offline' state.
212 config MEMORY_HOTREMOVE
213 bool "Allow for memory hot remove"
214 select MEMORY_ISOLATION
215 select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64)
216 depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE
219 # Heavily threaded applications may benefit from splitting the mm-wide
220 # page_table_lock, so that faults on different parts of the user address
221 # space can be handled with less contention: split it at this NR_CPUS.
222 # Default to 4 for wider testing, though 8 might be more appropriate.
223 # ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
224 # PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
225 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
227 config SPLIT_PTLOCK_CPUS
229 default "999999" if !MMU
230 default "999999" if ARM && !CPU_CACHE_VIPT
231 default "999999" if PARISC && !PA20
234 config ARCH_ENABLE_SPLIT_PMD_PTLOCK
238 # support for memory balloon
239 config MEMORY_BALLOON
243 # support for memory balloon compaction
244 config BALLOON_COMPACTION
245 bool "Allow for balloon memory compaction/migration"
247 depends on COMPACTION && MEMORY_BALLOON
249 Memory fragmentation introduced by ballooning might reduce
250 significantly the number of 2MB contiguous memory blocks that can be
251 used within a guest, thus imposing performance penalties associated
252 with the reduced number of transparent huge pages that could be used
253 by the guest workload. Allowing the compaction & migration for memory
254 pages enlisted as being part of memory balloon devices avoids the
255 scenario aforementioned and helps improving memory defragmentation.
258 # support for memory compaction
260 bool "Allow for memory compaction"
265 Compaction is the only memory management component to form
266 high order (larger physically contiguous) memory blocks
267 reliably. The page allocator relies on compaction heavily and
268 the lack of the feature can lead to unexpected OOM killer
269 invocations for high order memory requests. You shouldn't
270 disable this option unless there really is a strong reason for
271 it and then we would be really interested to hear about that at
275 # support for page migration
278 bool "Page migration"
280 depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU
282 Allows the migration of the physical location of pages of processes
283 while the virtual addresses are not changed. This is useful in
284 two situations. The first is on NUMA systems to put pages nearer
285 to the processors accessing. The second is when allocating huge
286 pages as migration can relocate pages to satisfy a huge page
287 allocation instead of reclaiming.
289 config ARCH_ENABLE_HUGEPAGE_MIGRATION
292 config PHYS_ADDR_T_64BIT
293 def_bool 64BIT || ARCH_PHYS_ADDR_T_64BIT
296 bool "Enable bounce buffers"
298 depends on BLOCK && MMU && (ZONE_DMA || HIGHMEM)
300 Enable bounce buffers for devices that cannot access
301 the full range of memory available to the CPU. Enabled
302 by default when ZONE_DMA or HIGHMEM is selected, but you
303 may say n to override this.
305 # On the 'tile' arch, USB OHCI needs the bounce pool since tilegx will often
306 # have more than 4GB of memory, but we don't currently use the IOTLB to present
307 # a 32-bit address to OHCI. So we need to use a bounce pool instead.
308 config NEED_BOUNCE_POOL
310 default y if TILE && USB_OHCI_HCD
320 An architecture should select this if it implements the
321 deprecated interface virt_to_bus(). All new architectures
322 should probably not select this.
330 bool "Enable KSM for page merging"
333 Enable Kernel Samepage Merging: KSM periodically scans those areas
334 of an application's address space that an app has advised may be
335 mergeable. When it finds pages of identical content, it replaces
336 the many instances by a single page with that content, so
337 saving memory until one or another app needs to modify the content.
338 Recommended for use with KVM, or with other duplicative applications.
339 See Documentation/vm/ksm.txt for more information: KSM is inactive
340 until a program has madvised that an area is MADV_MERGEABLE, and
341 root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set).
343 config DEFAULT_MMAP_MIN_ADDR
344 int "Low address space to protect from user allocation"
348 This is the portion of low virtual memory which should be protected
349 from userspace allocation. Keeping a user from writing to low pages
350 can help reduce the impact of kernel NULL pointer bugs.
352 For most ia64, ppc64 and x86 users with lots of address space
353 a value of 65536 is reasonable and should cause no problems.
354 On arm and other archs it should not be higher than 32768.
355 Programs which use vm86 functionality or have some need to map
356 this low address space will need CAP_SYS_RAWIO or disable this
357 protection by setting the value to 0.
359 This value can be changed after boot using the
360 /proc/sys/vm/mmap_min_addr tunable.
362 config ARCH_SUPPORTS_MEMORY_FAILURE
365 config MEMORY_FAILURE
367 depends on ARCH_SUPPORTS_MEMORY_FAILURE
368 bool "Enable recovery from hardware memory errors"
369 select MEMORY_ISOLATION
372 Enables code to recover from some memory failures on systems
373 with MCA recovery. This allows a system to continue running
374 even when some of its memory has uncorrected errors. This requires
375 special hardware support and typically ECC memory.
377 config HWPOISON_INJECT
378 tristate "HWPoison pages injector"
379 depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS
380 select PROC_PAGE_MONITOR
382 config NOMMU_INITIAL_TRIM_EXCESS
383 int "Turn on mmap() excess space trimming before booting"
387 The NOMMU mmap() frequently needs to allocate large contiguous chunks
388 of memory on which to store mappings, but it can only ask the system
389 allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently
390 more than it requires. To deal with this, mmap() is able to trim off
391 the excess and return it to the allocator.
393 If trimming is enabled, the excess is trimmed off and returned to the
394 system allocator, which can cause extra fragmentation, particularly
395 if there are a lot of transient processes.
397 If trimming is disabled, the excess is kept, but not used, which for
398 long-term mappings means that the space is wasted.
400 Trimming can be dynamically controlled through a sysctl option
401 (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of
402 excess pages there must be before trimming should occur, or zero if
403 no trimming is to occur.
405 This option specifies the initial value of this option. The default
406 of 1 says that all excess pages should be trimmed.
408 See Documentation/nommu-mmap.txt for more information.
410 config TRANSPARENT_HUGEPAGE
411 bool "Transparent Hugepage Support"
412 depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE
414 select RADIX_TREE_MULTIORDER
416 Transparent Hugepages allows the kernel to use huge pages and
417 huge tlb transparently to the applications whenever possible.
418 This feature can improve computing performance to certain
419 applications by speeding up page faults during memory
420 allocation, by reducing the number of tlb misses and by speeding
421 up the pagetable walking.
423 If memory constrained on embedded, you may want to say N.
426 prompt "Transparent Hugepage Support sysfs defaults"
427 depends on TRANSPARENT_HUGEPAGE
428 default TRANSPARENT_HUGEPAGE_ALWAYS
430 Selects the sysfs defaults for Transparent Hugepage Support.
432 config TRANSPARENT_HUGEPAGE_ALWAYS
435 Enabling Transparent Hugepage always, can increase the
436 memory footprint of applications without a guaranteed
437 benefit but it will work automatically for all applications.
439 config TRANSPARENT_HUGEPAGE_MADVISE
442 Enabling Transparent Hugepage madvise, will only provide a
443 performance improvement benefit to the applications using
444 madvise(MADV_HUGEPAGE) but it won't risk to increase the
445 memory footprint of applications without a guaranteed
449 config ARCH_WANTS_THP_SWAP
454 depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP
456 Swap transparent huge pages in one piece, without splitting.
457 XXX: For now this only does clustered swap space allocation.
459 For selection by architectures with reasonable THP sizes.
461 config TRANSPARENT_HUGE_PAGECACHE
463 depends on TRANSPARENT_HUGEPAGE
466 # UP and nommu archs use km based percpu allocator
468 config NEED_PER_CPU_KM
474 bool "Enable cleancache driver to cache clean pages if tmem is present"
477 Cleancache can be thought of as a page-granularity victim cache
478 for clean pages that the kernel's pageframe replacement algorithm
479 (PFRA) would like to keep around, but can't since there isn't enough
480 memory. So when the PFRA "evicts" a page, it first attempts to use
481 cleancache code to put the data contained in that page into
482 "transcendent memory", memory that is not directly accessible or
483 addressable by the kernel and is of unknown and possibly
484 time-varying size. And when a cleancache-enabled
485 filesystem wishes to access a page in a file on disk, it first
486 checks cleancache to see if it already contains it; if it does,
487 the page is copied into the kernel and a disk access is avoided.
488 When a transcendent memory driver is available (such as zcache or
489 Xen transcendent memory), a significant I/O reduction
490 may be achieved. When none is available, all cleancache calls
491 are reduced to a single pointer-compare-against-NULL resulting
492 in a negligible performance hit.
494 If unsure, say Y to enable cleancache
497 bool "Enable frontswap to cache swap pages if tmem is present"
501 Frontswap is so named because it can be thought of as the opposite
502 of a "backing" store for a swap device. The data is stored into
503 "transcendent memory", memory that is not directly accessible or
504 addressable by the kernel and is of unknown and possibly
505 time-varying size. When space in transcendent memory is available,
506 a significant swap I/O reduction may be achieved. When none is
507 available, all frontswap calls are reduced to a single pointer-
508 compare-against-NULL resulting in a negligible performance hit
509 and swap data is stored as normal on the matching swap device.
511 If unsure, say Y to enable frontswap.
514 bool "Contiguous Memory Allocator"
515 depends on HAVE_MEMBLOCK && MMU
517 select MEMORY_ISOLATION
519 This enables the Contiguous Memory Allocator which allows other
520 subsystems to allocate big physically-contiguous blocks of memory.
521 CMA reserves a region of memory and allows only movable pages to
522 be allocated from it. This way, the kernel can use the memory for
523 pagecache and when a subsystem requests for contiguous area, the
524 allocated pages are migrated away to serve the contiguous request.
529 bool "CMA debug messages (DEVELOPMENT)"
530 depends on DEBUG_KERNEL && CMA
532 Turns on debug messages in CMA. This produces KERN_DEBUG
533 messages for every CMA call as well as various messages while
534 processing calls such as dma_alloc_from_contiguous().
535 This option does not affect warning and error messages.
538 bool "CMA debugfs interface"
539 depends on CMA && DEBUG_FS
541 Turns on the DebugFS interface for CMA.
544 int "Maximum count of the CMA areas"
548 CMA allows to create CMA areas for particular purpose, mainly,
549 used as device private area. This parameter sets the maximum
550 number of CMA area in the system.
552 If unsure, leave the default value "7".
554 config MEM_SOFT_DIRTY
555 bool "Track memory changes"
556 depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS
557 select PROC_PAGE_MONITOR
559 This option enables memory changes tracking by introducing a
560 soft-dirty bit on pte-s. This bit it set when someone writes
561 into a page just as regular dirty bit, but unlike the latter
562 it can be cleared by hands.
564 See Documentation/vm/soft-dirty.txt for more details.
567 bool "Compressed cache for swap pages (EXPERIMENTAL)"
568 depends on FRONTSWAP && CRYPTO=y
573 A lightweight compressed cache for swap pages. It takes
574 pages that are in the process of being swapped out and attempts to
575 compress them into a dynamically allocated RAM-based memory pool.
576 This can result in a significant I/O reduction on swap device and,
577 in the case where decompressing from RAM is faster that swap device
578 reads, can also improve workload performance.
580 This is marked experimental because it is a new feature (as of
581 v3.11) that interacts heavily with memory reclaim. While these
582 interactions don't cause any known issues on simple memory setups,
583 they have not be fully explored on the large set of potential
584 configurations and workloads that exist.
587 tristate "Common API for compressed memory storage"
590 Compressed memory storage API. This allows using either zbud or
594 tristate "Low (Up to 2x) density storage for compressed pages"
597 A special purpose allocator for storing compressed pages.
598 It is designed to store up to two compressed pages per physical
599 page. While this design limits storage density, it has simple and
600 deterministic reclaim properties that make it preferable to a higher
601 density approach when reclaim will be used.
604 tristate "Up to 3x density storage for compressed pages"
608 A special purpose allocator for storing compressed pages.
609 It is designed to store up to three compressed pages per physical
610 page. It is a ZBUD derivative so the simplicity and determinism are
614 tristate "Memory allocator for compressed pages"
618 zsmalloc is a slab-based memory allocator designed to store
619 compressed RAM pages. zsmalloc uses virtual memory mapping
620 in order to reduce fragmentation. However, this results in a
621 non-standard allocator interface where a handle, not a pointer, is
622 returned by an alloc(). This handle must be mapped in order to
623 access the allocated space.
625 config PGTABLE_MAPPING
626 bool "Use page table mapping to access object in zsmalloc"
629 By default, zsmalloc uses a copy-based object mapping method to
630 access allocations that span two pages. However, if a particular
631 architecture (ex, ARM) performs VM mapping faster than copying,
632 then you should select this. This causes zsmalloc to use page table
633 mapping rather than copying for object mapping.
635 You can check speed with zsmalloc benchmark:
636 https://github.com/spartacus06/zsmapbench
639 bool "Export zsmalloc statistics"
643 This option enables code in the zsmalloc to collect various
644 statistics about whats happening in zsmalloc and exports that
645 information to userspace via debugfs.
648 config GENERIC_EARLY_IOREMAP
651 config MAX_STACK_SIZE_MB
652 int "Maximum user stack size for 32-bit processes (MB)"
656 depends on STACK_GROWSUP && (!64BIT || COMPAT)
658 This is the maximum stack size in Megabytes in the VM layout of 32-bit
659 user processes when the stack grows upwards (currently only on parisc
660 and metag arch). The stack will be located at the highest memory
661 address minus the given value, unless the RLIMIT_STACK hard limit is
662 changed to a smaller value in which case that is used.
664 A sane initial value is 80 MB.
666 # For architectures that support deferred memory initialisation
667 config ARCH_SUPPORTS_DEFERRED_STRUCT_PAGE_INIT
670 config DEFERRED_STRUCT_PAGE_INIT
671 bool "Defer initialisation of struct pages to kthreads"
673 depends on ARCH_SUPPORTS_DEFERRED_STRUCT_PAGE_INIT
674 depends on NO_BOOTMEM && MEMORY_HOTPLUG
677 Ordinarily all struct pages are initialised during early boot in a
678 single thread. On very large machines this can take a considerable
679 amount of time. If this option is set, large machines will bring up
680 a subset of memmap at boot and then initialise the rest in parallel
681 by starting one-off "pgdatinitX" kernel thread for each node X. This
682 has a potential performance impact on processes running early in the
683 lifetime of the system until these kthreads finish the
686 config IDLE_PAGE_TRACKING
687 bool "Enable idle page tracking"
688 depends on SYSFS && MMU
689 select PAGE_EXTENSION if !64BIT
691 This feature allows to estimate the amount of user pages that have
692 not been touched during a given period of time. This information can
693 be useful to tune memory cgroup limits and/or for job placement
694 within a compute cluster.
696 See Documentation/vm/idle_page_tracking.txt for more details.
699 bool "Device memory (pmem, etc...) hotplug support"
700 depends on MEMORY_HOTPLUG
701 depends on MEMORY_HOTREMOVE
702 depends on SPARSEMEM_VMEMMAP
703 depends on X86_64 #arch_add_memory() comprehends device memory
706 Device memory hotplug support allows for establishing pmem,
707 or other device driver discovered memory regions, in the
708 memmap. This allows pfn_to_page() lookups of otherwise
709 "device-physical" addresses which is needed for using a DAX
710 mapping in an O_DIRECT operation, among other things.
712 If FS_DAX is enabled, then say Y.
717 config ARCH_USES_HIGH_VMA_FLAGS
719 config ARCH_HAS_PKEYS
723 bool "Collect percpu memory statistics"
726 This feature collects and exposes statistics via debugfs. The
727 information includes global and per chunk statistics, which can
728 be used to help understand percpu memory usage.