1ec8f24b7SThomas Gleixner# SPDX-License-Identifier: GPL-2.0-only 259e0b520SChristoph Hellwig 359e0b520SChristoph Hellwigmenu "Memory Management options" 459e0b520SChristoph Hellwig 5e1785e85SDave Hansenconfig SELECT_MEMORY_MODEL 6e1785e85SDave Hansen def_bool y 7a8826eebSKees Cook depends on ARCH_SELECT_MEMORY_MODEL 8e1785e85SDave Hansen 93a9da765SDave Hansenchoice 103a9da765SDave Hansen prompt "Memory model" 11e1785e85SDave Hansen depends on SELECT_MEMORY_MODEL 12e1785e85SDave Hansen default DISCONTIGMEM_MANUAL if ARCH_DISCONTIGMEM_DEFAULT 13d41dee36SAndy Whitcroft default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT 14e1785e85SDave Hansen default FLATMEM_MANUAL 15d66d109dSMike Rapoport help 16d66d109dSMike Rapoport This option allows you to change some of the ways that 17d66d109dSMike Rapoport Linux manages its memory internally. Most users will 18d66d109dSMike Rapoport only have one option here selected by the architecture 19d66d109dSMike Rapoport configuration. This is normal. 203a9da765SDave Hansen 21e1785e85SDave Hansenconfig FLATMEM_MANUAL 223a9da765SDave Hansen bool "Flat Memory" 23c898ec16SAnton Blanchard depends on !(ARCH_DISCONTIGMEM_ENABLE || ARCH_SPARSEMEM_ENABLE) || ARCH_FLATMEM_ENABLE 243a9da765SDave Hansen help 25d66d109dSMike Rapoport This option is best suited for non-NUMA systems with 26d66d109dSMike Rapoport flat address space. The FLATMEM is the most efficient 27d66d109dSMike Rapoport system in terms of performance and resource consumption 28d66d109dSMike Rapoport and it is the best option for smaller systems. 293a9da765SDave Hansen 30d66d109dSMike Rapoport For systems that have holes in their physical address 31d66d109dSMike Rapoport spaces and for features like NUMA and memory hotplug, 32d66d109dSMike Rapoport choose "Sparse Memory" 33d41dee36SAndy Whitcroft 34d41dee36SAndy Whitcroft If unsure, choose this option (Flat Memory) over any other. 353a9da765SDave Hansen 36e1785e85SDave Hansenconfig DISCONTIGMEM_MANUAL 37f3519f91SDave Hansen bool "Discontiguous Memory" 383a9da765SDave Hansen depends on ARCH_DISCONTIGMEM_ENABLE 393a9da765SDave Hansen help 40785dcd44SDave Hansen This option provides enhanced support for discontiguous 41785dcd44SDave Hansen memory systems, over FLATMEM. These systems have holes 42785dcd44SDave Hansen in their physical address spaces, and this option provides 43d66d109dSMike Rapoport more efficient handling of these holes. 44785dcd44SDave Hansen 45d66d109dSMike Rapoport Although "Discontiguous Memory" is still used by several 46d66d109dSMike Rapoport architectures, it is considered deprecated in favor of 47d66d109dSMike Rapoport "Sparse Memory". 48785dcd44SDave Hansen 49d66d109dSMike Rapoport If unsure, choose "Sparse Memory" over this option. 503a9da765SDave Hansen 51d41dee36SAndy Whitcroftconfig SPARSEMEM_MANUAL 52d41dee36SAndy Whitcroft bool "Sparse Memory" 53d41dee36SAndy Whitcroft depends on ARCH_SPARSEMEM_ENABLE 54d41dee36SAndy Whitcroft help 55d41dee36SAndy Whitcroft This will be the only option for some systems, including 56d66d109dSMike Rapoport memory hot-plug systems. This is normal. 57d41dee36SAndy Whitcroft 58d66d109dSMike Rapoport This option provides efficient support for systems with 59d66d109dSMike Rapoport holes is their physical address space and allows memory 60d66d109dSMike Rapoport hot-plug and hot-remove. 61d41dee36SAndy Whitcroft 62d66d109dSMike Rapoport If unsure, choose "Flat Memory" over this option. 63d41dee36SAndy Whitcroft 643a9da765SDave Hansenendchoice 653a9da765SDave Hansen 66e1785e85SDave Hansenconfig DISCONTIGMEM 67e1785e85SDave Hansen def_bool y 68e1785e85SDave Hansen depends on (!SELECT_MEMORY_MODEL && ARCH_DISCONTIGMEM_ENABLE) || DISCONTIGMEM_MANUAL 69e1785e85SDave Hansen 70d41dee36SAndy Whitcroftconfig SPARSEMEM 71d41dee36SAndy Whitcroft def_bool y 721a83e175SRussell King depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL 73d41dee36SAndy Whitcroft 74e1785e85SDave Hansenconfig FLATMEM 75e1785e85SDave Hansen def_bool y 76d41dee36SAndy Whitcroft depends on (!DISCONTIGMEM && !SPARSEMEM) || FLATMEM_MANUAL 77d41dee36SAndy Whitcroft 78d41dee36SAndy Whitcroftconfig FLAT_NODE_MEM_MAP 79d41dee36SAndy Whitcroft def_bool y 80d41dee36SAndy Whitcroft depends on !SPARSEMEM 81e1785e85SDave Hansen 8293b7504eSDave Hansen# 8393b7504eSDave Hansen# Both the NUMA code and DISCONTIGMEM use arrays of pg_data_t's 8493b7504eSDave Hansen# to represent different areas of memory. This variable allows 8593b7504eSDave Hansen# those dependencies to exist individually. 8693b7504eSDave Hansen# 8793b7504eSDave Hansenconfig NEED_MULTIPLE_NODES 8893b7504eSDave Hansen def_bool y 8993b7504eSDave Hansen depends on DISCONTIGMEM || NUMA 90af705362SAndy Whitcroft 91af705362SAndy Whitcroftconfig HAVE_MEMORY_PRESENT 92af705362SAndy Whitcroft def_bool y 93d41dee36SAndy Whitcroft depends on ARCH_HAVE_MEMORY_PRESENT || SPARSEMEM 94802f192eSBob Picco 95802f192eSBob Picco# 963e347261SBob Picco# SPARSEMEM_EXTREME (which is the default) does some bootmem 973e347261SBob Picco# allocations when memory_present() is called. If this cannot 983e347261SBob Picco# be done on your architecture, select this option. However, 993e347261SBob Picco# statically allocating the mem_section[] array can potentially 1003e347261SBob Picco# consume vast quantities of .bss, so be careful. 1013e347261SBob Picco# 1023e347261SBob Picco# This option will also potentially produce smaller runtime code 1033e347261SBob Picco# with gcc 3.4 and later. 1043e347261SBob Picco# 1053e347261SBob Piccoconfig SPARSEMEM_STATIC 1069ba16087SJan Beulich bool 1073e347261SBob Picco 1083e347261SBob Picco# 10944c09201SMatt LaPlante# Architecture platforms which require a two level mem_section in SPARSEMEM 110802f192eSBob Picco# must select this option. This is usually for architecture platforms with 111802f192eSBob Picco# an extremely sparse physical address space. 112802f192eSBob Picco# 1133e347261SBob Piccoconfig SPARSEMEM_EXTREME 1143e347261SBob Picco def_bool y 1153e347261SBob Picco depends on SPARSEMEM && !SPARSEMEM_STATIC 1164c21e2f2SHugh Dickins 11729c71111SAndy Whitcroftconfig SPARSEMEM_VMEMMAP_ENABLE 1189ba16087SJan Beulich bool 11929c71111SAndy Whitcroft 12029c71111SAndy Whitcroftconfig SPARSEMEM_VMEMMAP 121a5ee6daaSGeoff Levand bool "Sparse Memory virtual memmap" 122a5ee6daaSGeoff Levand depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE 123a5ee6daaSGeoff Levand default y 124a5ee6daaSGeoff Levand help 125a5ee6daaSGeoff Levand SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise 126a5ee6daaSGeoff Levand pfn_to_page and page_to_pfn operations. This is the most 127a5ee6daaSGeoff Levand efficient option when sufficient kernel resources are available. 12829c71111SAndy Whitcroft 1297c0caeb8STejun Heoconfig HAVE_MEMBLOCK_NODE_MAP 1306341e62bSChristoph Jaeger bool 1317c0caeb8STejun Heo 13270210ed9SPhilipp Hachtmannconfig HAVE_MEMBLOCK_PHYS_MAP 1336341e62bSChristoph Jaeger bool 13470210ed9SPhilipp Hachtmann 13567a929e0SChristoph Hellwigconfig HAVE_FAST_GUP 136050a9adcSChristoph Hellwig depends on MMU 1376341e62bSChristoph Jaeger bool 1382667f50eSSteve Capper 139350e88baSMike Rapoportconfig ARCH_KEEP_MEMBLOCK 1406341e62bSChristoph Jaeger bool 141c378ddd5STejun Heo 142ee6f509cSMinchan Kimconfig MEMORY_ISOLATION 1436341e62bSChristoph Jaeger bool 144ee6f509cSMinchan Kim 14546723bfaSYasuaki Ishimatsu# 14646723bfaSYasuaki Ishimatsu# Only be set on architectures that have completely implemented memory hotplug 14746723bfaSYasuaki Ishimatsu# feature. If you are not sure, don't touch it. 14846723bfaSYasuaki Ishimatsu# 14946723bfaSYasuaki Ishimatsuconfig HAVE_BOOTMEM_INFO_NODE 15046723bfaSYasuaki Ishimatsu def_bool n 15146723bfaSYasuaki Ishimatsu 1523947be19SDave Hansen# eventually, we can have this option just 'select SPARSEMEM' 1533947be19SDave Hansenconfig MEMORY_HOTPLUG 1543947be19SDave Hansen bool "Allow for memory hot-add" 155ec69acbbSKeith Mannthey depends on SPARSEMEM || X86_64_ACPI_NUMA 15640b31360SStephen Rothwell depends on ARCH_ENABLE_MEMORY_HOTPLUG 1573947be19SDave Hansen 158ec69acbbSKeith Manntheyconfig MEMORY_HOTPLUG_SPARSE 159ec69acbbSKeith Mannthey def_bool y 160ec69acbbSKeith Mannthey depends on SPARSEMEM && MEMORY_HOTPLUG 161ec69acbbSKeith Mannthey 1628604d9e5SVitaly Kuznetsovconfig MEMORY_HOTPLUG_DEFAULT_ONLINE 1638604d9e5SVitaly Kuznetsov bool "Online the newly added memory blocks by default" 1648604d9e5SVitaly Kuznetsov depends on MEMORY_HOTPLUG 1658604d9e5SVitaly Kuznetsov help 1668604d9e5SVitaly Kuznetsov This option sets the default policy setting for memory hotplug 1678604d9e5SVitaly Kuznetsov onlining policy (/sys/devices/system/memory/auto_online_blocks) which 1688604d9e5SVitaly Kuznetsov determines what happens to newly added memory regions. Policy setting 1698604d9e5SVitaly Kuznetsov can always be changed at runtime. 170cb1aaebeSMauro Carvalho Chehab See Documentation/admin-guide/mm/memory-hotplug.rst for more information. 1718604d9e5SVitaly Kuznetsov 1728604d9e5SVitaly Kuznetsov Say Y here if you want all hot-plugged memory blocks to appear in 1738604d9e5SVitaly Kuznetsov 'online' state by default. 1748604d9e5SVitaly Kuznetsov Say N here if you want the default policy to keep all hot-plugged 1758604d9e5SVitaly Kuznetsov memory blocks in 'offline' state. 1768604d9e5SVitaly Kuznetsov 1770c0e6195SKAMEZAWA Hiroyukiconfig MEMORY_HOTREMOVE 1780c0e6195SKAMEZAWA Hiroyuki bool "Allow for memory hot remove" 17946723bfaSYasuaki Ishimatsu select MEMORY_ISOLATION 180f7e3334aSNathan Fontenot select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64) 1810c0e6195SKAMEZAWA Hiroyuki depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE 1820c0e6195SKAMEZAWA Hiroyuki depends on MIGRATION 1830c0e6195SKAMEZAWA Hiroyuki 1844c21e2f2SHugh Dickins# Heavily threaded applications may benefit from splitting the mm-wide 1854c21e2f2SHugh Dickins# page_table_lock, so that faults on different parts of the user address 1864c21e2f2SHugh Dickins# space can be handled with less contention: split it at this NR_CPUS. 1874c21e2f2SHugh Dickins# Default to 4 for wider testing, though 8 might be more appropriate. 1884c21e2f2SHugh Dickins# ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock. 1897b6ac9dfSHugh Dickins# PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes. 190a70caa8bSHugh Dickins# DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page. 1914c21e2f2SHugh Dickins# 1924c21e2f2SHugh Dickinsconfig SPLIT_PTLOCK_CPUS 1934c21e2f2SHugh Dickins int 1949164550eSKirill A. Shutemov default "999999" if !MMU 195a70caa8bSHugh Dickins default "999999" if ARM && !CPU_CACHE_VIPT 196a70caa8bSHugh Dickins default "999999" if PARISC && !PA20 1974c21e2f2SHugh Dickins default "4" 1987cbe34cfSChristoph Lameter 199e009bb30SKirill A. Shutemovconfig ARCH_ENABLE_SPLIT_PMD_PTLOCK 2006341e62bSChristoph Jaeger bool 201e009bb30SKirill A. Shutemov 2027cbe34cfSChristoph Lameter# 20309316c09SKonstantin Khlebnikov# support for memory balloon 20409316c09SKonstantin Khlebnikovconfig MEMORY_BALLOON 2056341e62bSChristoph Jaeger bool 20609316c09SKonstantin Khlebnikov 20709316c09SKonstantin Khlebnikov# 20818468d93SRafael Aquini# support for memory balloon compaction 20918468d93SRafael Aquiniconfig BALLOON_COMPACTION 21018468d93SRafael Aquini bool "Allow for balloon memory compaction/migration" 21118468d93SRafael Aquini def_bool y 21209316c09SKonstantin Khlebnikov depends on COMPACTION && MEMORY_BALLOON 21318468d93SRafael Aquini help 21418468d93SRafael Aquini Memory fragmentation introduced by ballooning might reduce 21518468d93SRafael Aquini significantly the number of 2MB contiguous memory blocks that can be 21618468d93SRafael Aquini used within a guest, thus imposing performance penalties associated 21718468d93SRafael Aquini with the reduced number of transparent huge pages that could be used 21818468d93SRafael Aquini by the guest workload. Allowing the compaction & migration for memory 21918468d93SRafael Aquini pages enlisted as being part of memory balloon devices avoids the 22018468d93SRafael Aquini scenario aforementioned and helps improving memory defragmentation. 22118468d93SRafael Aquini 22218468d93SRafael Aquini# 223e9e96b39SMel Gorman# support for memory compaction 224e9e96b39SMel Gormanconfig COMPACTION 225e9e96b39SMel Gorman bool "Allow for memory compaction" 22605106e6aSRik van Riel def_bool y 227e9e96b39SMel Gorman select MIGRATION 22833a93877SAndrea Arcangeli depends on MMU 229e9e96b39SMel Gorman help 230b32eaf71SMichal Hocko Compaction is the only memory management component to form 231b32eaf71SMichal Hocko high order (larger physically contiguous) memory blocks 232b32eaf71SMichal Hocko reliably. The page allocator relies on compaction heavily and 233b32eaf71SMichal Hocko the lack of the feature can lead to unexpected OOM killer 234b32eaf71SMichal Hocko invocations for high order memory requests. You shouldn't 235b32eaf71SMichal Hocko disable this option unless there really is a strong reason for 236b32eaf71SMichal Hocko it and then we would be really interested to hear about that at 237b32eaf71SMichal Hocko linux-mm@kvack.org. 238e9e96b39SMel Gorman 239e9e96b39SMel Gorman# 2407cbe34cfSChristoph Lameter# support for page migration 2417cbe34cfSChristoph Lameter# 2427cbe34cfSChristoph Lameterconfig MIGRATION 243b20a3503SChristoph Lameter bool "Page migration" 2446c5240aeSChristoph Lameter def_bool y 245de32a817SChen Gang depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU 246b20a3503SChristoph Lameter help 247b20a3503SChristoph Lameter Allows the migration of the physical location of pages of processes 248e9e96b39SMel Gorman while the virtual addresses are not changed. This is useful in 249e9e96b39SMel Gorman two situations. The first is on NUMA systems to put pages nearer 250e9e96b39SMel Gorman to the processors accessing. The second is when allocating huge 251e9e96b39SMel Gorman pages as migration can relocate pages to satisfy a huge page 252e9e96b39SMel Gorman allocation instead of reclaiming. 2536550e07fSGreg Kroah-Hartman 254c177c81eSNaoya Horiguchiconfig ARCH_ENABLE_HUGEPAGE_MIGRATION 2556341e62bSChristoph Jaeger bool 256c177c81eSNaoya Horiguchi 2579c670ea3SNaoya Horiguchiconfig ARCH_ENABLE_THP_MIGRATION 2589c670ea3SNaoya Horiguchi bool 2599c670ea3SNaoya Horiguchi 2608df995f6SAlexandre Ghiticonfig CONTIG_ALLOC 2618df995f6SAlexandre Ghiti def_bool (MEMORY_ISOLATION && COMPACTION) || CMA 2628df995f6SAlexandre Ghiti 263600715dcSJeremy Fitzhardingeconfig PHYS_ADDR_T_64BIT 264d4a451d5SChristoph Hellwig def_bool 64BIT 265600715dcSJeremy Fitzhardinge 2662a7326b5SChristoph Lameterconfig BOUNCE 2679ca24e2eSVinayak Menon bool "Enable bounce buffers" 2689ca24e2eSVinayak Menon default y 2692a7326b5SChristoph Lameter depends on BLOCK && MMU && (ZONE_DMA || HIGHMEM) 2709ca24e2eSVinayak Menon help 2719ca24e2eSVinayak Menon Enable bounce buffers for devices that cannot access 2729ca24e2eSVinayak Menon the full range of memory available to the CPU. Enabled 2739ca24e2eSVinayak Menon by default when ZONE_DMA or HIGHMEM is selected, but you 2749ca24e2eSVinayak Menon may say n to override this. 2752a7326b5SChristoph Lameter 2766225e937SChristoph Lameterconfig NR_QUICK 2776225e937SChristoph Lameter int 2786225e937SChristoph Lameter depends on QUICKLIST 2796225e937SChristoph Lameter default "1" 280f057eac0SStephen Rothwell 281f057eac0SStephen Rothwellconfig VIRT_TO_BUS 2824febd95aSStephen Rothwell bool 2834febd95aSStephen Rothwell help 2844febd95aSStephen Rothwell An architecture should select this if it implements the 2854febd95aSStephen Rothwell deprecated interface virt_to_bus(). All new architectures 2864febd95aSStephen Rothwell should probably not select this. 2874febd95aSStephen Rothwell 288cddb8a5cSAndrea Arcangeli 289cddb8a5cSAndrea Arcangeliconfig MMU_NOTIFIER 290cddb8a5cSAndrea Arcangeli bool 29183fe27eaSPranith Kumar select SRCU 292fc4d5c29SDavid Howells 293f8af4da3SHugh Dickinsconfig KSM 294f8af4da3SHugh Dickins bool "Enable KSM for page merging" 295f8af4da3SHugh Dickins depends on MMU 29659e1a2f4STimofey Titovets select XXHASH 297f8af4da3SHugh Dickins help 298f8af4da3SHugh Dickins Enable Kernel Samepage Merging: KSM periodically scans those areas 299f8af4da3SHugh Dickins of an application's address space that an app has advised may be 300f8af4da3SHugh Dickins mergeable. When it finds pages of identical content, it replaces 301d0f209f6SHugh Dickins the many instances by a single page with that content, so 302f8af4da3SHugh Dickins saving memory until one or another app needs to modify the content. 303f8af4da3SHugh Dickins Recommended for use with KVM, or with other duplicative applications. 304ad56b738SMike Rapoport See Documentation/vm/ksm.rst for more information: KSM is inactive 305c73602adSHugh Dickins until a program has madvised that an area is MADV_MERGEABLE, and 306c73602adSHugh Dickins root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set). 307f8af4da3SHugh Dickins 308e0a94c2aSChristoph Lameterconfig DEFAULT_MMAP_MIN_ADDR 309e0a94c2aSChristoph Lameter int "Low address space to protect from user allocation" 3106e141546SDavid Howells depends on MMU 311e0a94c2aSChristoph Lameter default 4096 312e0a94c2aSChristoph Lameter help 313e0a94c2aSChristoph Lameter This is the portion of low virtual memory which should be protected 314e0a94c2aSChristoph Lameter from userspace allocation. Keeping a user from writing to low pages 315e0a94c2aSChristoph Lameter can help reduce the impact of kernel NULL pointer bugs. 316e0a94c2aSChristoph Lameter 317e0a94c2aSChristoph Lameter For most ia64, ppc64 and x86 users with lots of address space 318e0a94c2aSChristoph Lameter a value of 65536 is reasonable and should cause no problems. 319e0a94c2aSChristoph Lameter On arm and other archs it should not be higher than 32768. 320788084abSEric Paris Programs which use vm86 functionality or have some need to map 321788084abSEric Paris this low address space will need CAP_SYS_RAWIO or disable this 322788084abSEric Paris protection by setting the value to 0. 323e0a94c2aSChristoph Lameter 324e0a94c2aSChristoph Lameter This value can be changed after boot using the 325e0a94c2aSChristoph Lameter /proc/sys/vm/mmap_min_addr tunable. 326e0a94c2aSChristoph Lameter 327d949f36fSLinus Torvaldsconfig ARCH_SUPPORTS_MEMORY_FAILURE 328d949f36fSLinus Torvalds bool 329e0a94c2aSChristoph Lameter 3306a46079cSAndi Kleenconfig MEMORY_FAILURE 3316a46079cSAndi Kleen depends on MMU 332d949f36fSLinus Torvalds depends on ARCH_SUPPORTS_MEMORY_FAILURE 3336a46079cSAndi Kleen bool "Enable recovery from hardware memory errors" 334ee6f509cSMinchan Kim select MEMORY_ISOLATION 33597f0b134SXie XiuQi select RAS 3366a46079cSAndi Kleen help 3376a46079cSAndi Kleen Enables code to recover from some memory failures on systems 3386a46079cSAndi Kleen with MCA recovery. This allows a system to continue running 3396a46079cSAndi Kleen even when some of its memory has uncorrected errors. This requires 3406a46079cSAndi Kleen special hardware support and typically ECC memory. 3416a46079cSAndi Kleen 342cae681fcSAndi Kleenconfig HWPOISON_INJECT 343413f9efbSAndi Kleen tristate "HWPoison pages injector" 34427df5068SAndi Kleen depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS 345478c5ffcSWu Fengguang select PROC_PAGE_MONITOR 346cae681fcSAndi Kleen 347fc4d5c29SDavid Howellsconfig NOMMU_INITIAL_TRIM_EXCESS 348fc4d5c29SDavid Howells int "Turn on mmap() excess space trimming before booting" 349fc4d5c29SDavid Howells depends on !MMU 350fc4d5c29SDavid Howells default 1 351fc4d5c29SDavid Howells help 352fc4d5c29SDavid Howells The NOMMU mmap() frequently needs to allocate large contiguous chunks 353fc4d5c29SDavid Howells of memory on which to store mappings, but it can only ask the system 354fc4d5c29SDavid Howells allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently 355fc4d5c29SDavid Howells more than it requires. To deal with this, mmap() is able to trim off 356fc4d5c29SDavid Howells the excess and return it to the allocator. 357fc4d5c29SDavid Howells 358fc4d5c29SDavid Howells If trimming is enabled, the excess is trimmed off and returned to the 359fc4d5c29SDavid Howells system allocator, which can cause extra fragmentation, particularly 360fc4d5c29SDavid Howells if there are a lot of transient processes. 361fc4d5c29SDavid Howells 362fc4d5c29SDavid Howells If trimming is disabled, the excess is kept, but not used, which for 363fc4d5c29SDavid Howells long-term mappings means that the space is wasted. 364fc4d5c29SDavid Howells 365fc4d5c29SDavid Howells Trimming can be dynamically controlled through a sysctl option 366fc4d5c29SDavid Howells (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of 367fc4d5c29SDavid Howells excess pages there must be before trimming should occur, or zero if 368fc4d5c29SDavid Howells no trimming is to occur. 369fc4d5c29SDavid Howells 370fc4d5c29SDavid Howells This option specifies the initial value of this option. The default 371fc4d5c29SDavid Howells of 1 says that all excess pages should be trimmed. 372fc4d5c29SDavid Howells 373fc4d5c29SDavid Howells See Documentation/nommu-mmap.txt for more information. 374bbddff05STejun Heo 3754c76d9d1SAndrea Arcangeliconfig TRANSPARENT_HUGEPAGE 37613ece886SAndrea Arcangeli bool "Transparent Hugepage Support" 37715626062SGerald Schaefer depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE 3785d689240SAndrea Arcangeli select COMPACTION 3793a08cd52SMatthew Wilcox select XARRAY_MULTI 3804c76d9d1SAndrea Arcangeli help 3814c76d9d1SAndrea Arcangeli Transparent Hugepages allows the kernel to use huge pages and 3824c76d9d1SAndrea Arcangeli huge tlb transparently to the applications whenever possible. 3834c76d9d1SAndrea Arcangeli This feature can improve computing performance to certain 3844c76d9d1SAndrea Arcangeli applications by speeding up page faults during memory 3854c76d9d1SAndrea Arcangeli allocation, by reducing the number of tlb misses and by speeding 3864c76d9d1SAndrea Arcangeli up the pagetable walking. 3874c76d9d1SAndrea Arcangeli 3884c76d9d1SAndrea Arcangeli If memory constrained on embedded, you may want to say N. 3894c76d9d1SAndrea Arcangeli 39013ece886SAndrea Arcangelichoice 39113ece886SAndrea Arcangeli prompt "Transparent Hugepage Support sysfs defaults" 39213ece886SAndrea Arcangeli depends on TRANSPARENT_HUGEPAGE 39313ece886SAndrea Arcangeli default TRANSPARENT_HUGEPAGE_ALWAYS 39413ece886SAndrea Arcangeli help 39513ece886SAndrea Arcangeli Selects the sysfs defaults for Transparent Hugepage Support. 39613ece886SAndrea Arcangeli 39713ece886SAndrea Arcangeli config TRANSPARENT_HUGEPAGE_ALWAYS 39813ece886SAndrea Arcangeli bool "always" 39913ece886SAndrea Arcangeli help 40013ece886SAndrea Arcangeli Enabling Transparent Hugepage always, can increase the 40113ece886SAndrea Arcangeli memory footprint of applications without a guaranteed 40213ece886SAndrea Arcangeli benefit but it will work automatically for all applications. 40313ece886SAndrea Arcangeli 40413ece886SAndrea Arcangeli config TRANSPARENT_HUGEPAGE_MADVISE 40513ece886SAndrea Arcangeli bool "madvise" 40613ece886SAndrea Arcangeli help 40713ece886SAndrea Arcangeli Enabling Transparent Hugepage madvise, will only provide a 40813ece886SAndrea Arcangeli performance improvement benefit to the applications using 40913ece886SAndrea Arcangeli madvise(MADV_HUGEPAGE) but it won't risk to increase the 41013ece886SAndrea Arcangeli memory footprint of applications without a guaranteed 41113ece886SAndrea Arcangeli benefit. 41213ece886SAndrea Arcangeliendchoice 41313ece886SAndrea Arcangeli 41438d8b4e6SHuang Yingconfig ARCH_WANTS_THP_SWAP 41538d8b4e6SHuang Ying def_bool n 41638d8b4e6SHuang Ying 41738d8b4e6SHuang Yingconfig THP_SWAP 41838d8b4e6SHuang Ying def_bool y 41914fef284SHuang Ying depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP 42038d8b4e6SHuang Ying help 42138d8b4e6SHuang Ying Swap transparent huge pages in one piece, without splitting. 42214fef284SHuang Ying XXX: For now, swap cluster backing transparent huge page 42314fef284SHuang Ying will be split after swapout. 42438d8b4e6SHuang Ying 42538d8b4e6SHuang Ying For selection by architectures with reasonable THP sizes. 42638d8b4e6SHuang Ying 427e496cf3dSKirill A. Shutemovconfig TRANSPARENT_HUGE_PAGECACHE 428e496cf3dSKirill A. Shutemov def_bool y 429953c66c2SAneesh Kumar K.V depends on TRANSPARENT_HUGEPAGE 430e496cf3dSKirill A. Shutemov 431e496cf3dSKirill A. Shutemov# 432bbddff05STejun Heo# UP and nommu archs use km based percpu allocator 433bbddff05STejun Heo# 434bbddff05STejun Heoconfig NEED_PER_CPU_KM 435bbddff05STejun Heo depends on !SMP 436bbddff05STejun Heo bool 437bbddff05STejun Heo default y 438077b1f83SDan Magenheimer 439077b1f83SDan Magenheimerconfig CLEANCACHE 440077b1f83SDan Magenheimer bool "Enable cleancache driver to cache clean pages if tmem is present" 441077b1f83SDan Magenheimer help 442077b1f83SDan Magenheimer Cleancache can be thought of as a page-granularity victim cache 443077b1f83SDan Magenheimer for clean pages that the kernel's pageframe replacement algorithm 444077b1f83SDan Magenheimer (PFRA) would like to keep around, but can't since there isn't enough 445077b1f83SDan Magenheimer memory. So when the PFRA "evicts" a page, it first attempts to use 446140a1ef2SMichael Witten cleancache code to put the data contained in that page into 447077b1f83SDan Magenheimer "transcendent memory", memory that is not directly accessible or 448077b1f83SDan Magenheimer addressable by the kernel and is of unknown and possibly 449077b1f83SDan Magenheimer time-varying size. And when a cleancache-enabled 450077b1f83SDan Magenheimer filesystem wishes to access a page in a file on disk, it first 451077b1f83SDan Magenheimer checks cleancache to see if it already contains it; if it does, 452077b1f83SDan Magenheimer the page is copied into the kernel and a disk access is avoided. 453077b1f83SDan Magenheimer When a transcendent memory driver is available (such as zcache or 454077b1f83SDan Magenheimer Xen transcendent memory), a significant I/O reduction 455077b1f83SDan Magenheimer may be achieved. When none is available, all cleancache calls 456077b1f83SDan Magenheimer are reduced to a single pointer-compare-against-NULL resulting 457077b1f83SDan Magenheimer in a negligible performance hit. 458077b1f83SDan Magenheimer 459077b1f83SDan Magenheimer If unsure, say Y to enable cleancache 46027c6aec2SDan Magenheimer 46127c6aec2SDan Magenheimerconfig FRONTSWAP 46227c6aec2SDan Magenheimer bool "Enable frontswap to cache swap pages if tmem is present" 46327c6aec2SDan Magenheimer depends on SWAP 46427c6aec2SDan Magenheimer help 46527c6aec2SDan Magenheimer Frontswap is so named because it can be thought of as the opposite 46627c6aec2SDan Magenheimer of a "backing" store for a swap device. The data is stored into 46727c6aec2SDan Magenheimer "transcendent memory", memory that is not directly accessible or 46827c6aec2SDan Magenheimer addressable by the kernel and is of unknown and possibly 46927c6aec2SDan Magenheimer time-varying size. When space in transcendent memory is available, 47027c6aec2SDan Magenheimer a significant swap I/O reduction may be achieved. When none is 47127c6aec2SDan Magenheimer available, all frontswap calls are reduced to a single pointer- 47227c6aec2SDan Magenheimer compare-against-NULL resulting in a negligible performance hit 47327c6aec2SDan Magenheimer and swap data is stored as normal on the matching swap device. 47427c6aec2SDan Magenheimer 47527c6aec2SDan Magenheimer If unsure, say Y to enable frontswap. 476f825c736SAneesh Kumar K.V 477f825c736SAneesh Kumar K.Vconfig CMA 478f825c736SAneesh Kumar K.V bool "Contiguous Memory Allocator" 479aca52c39SMike Rapoport depends on MMU 480f825c736SAneesh Kumar K.V select MIGRATION 481f825c736SAneesh Kumar K.V select MEMORY_ISOLATION 482f825c736SAneesh Kumar K.V help 483f825c736SAneesh Kumar K.V This enables the Contiguous Memory Allocator which allows other 484f825c736SAneesh Kumar K.V subsystems to allocate big physically-contiguous blocks of memory. 485f825c736SAneesh Kumar K.V CMA reserves a region of memory and allows only movable pages to 486f825c736SAneesh Kumar K.V be allocated from it. This way, the kernel can use the memory for 487f825c736SAneesh Kumar K.V pagecache and when a subsystem requests for contiguous area, the 488f825c736SAneesh Kumar K.V allocated pages are migrated away to serve the contiguous request. 489f825c736SAneesh Kumar K.V 490f825c736SAneesh Kumar K.V If unsure, say "n". 491f825c736SAneesh Kumar K.V 492f825c736SAneesh Kumar K.Vconfig CMA_DEBUG 493f825c736SAneesh Kumar K.V bool "CMA debug messages (DEVELOPMENT)" 494f825c736SAneesh Kumar K.V depends on DEBUG_KERNEL && CMA 495f825c736SAneesh Kumar K.V help 496f825c736SAneesh Kumar K.V Turns on debug messages in CMA. This produces KERN_DEBUG 497f825c736SAneesh Kumar K.V messages for every CMA call as well as various messages while 498f825c736SAneesh Kumar K.V processing calls such as dma_alloc_from_contiguous(). 499f825c736SAneesh Kumar K.V This option does not affect warning and error messages. 500bf550fc9SAlexander Graf 50128b24c1fSSasha Levinconfig CMA_DEBUGFS 50228b24c1fSSasha Levin bool "CMA debugfs interface" 50328b24c1fSSasha Levin depends on CMA && DEBUG_FS 50428b24c1fSSasha Levin help 50528b24c1fSSasha Levin Turns on the DebugFS interface for CMA. 50628b24c1fSSasha Levin 507a254129eSJoonsoo Kimconfig CMA_AREAS 508a254129eSJoonsoo Kim int "Maximum count of the CMA areas" 509a254129eSJoonsoo Kim depends on CMA 510a254129eSJoonsoo Kim default 7 511a254129eSJoonsoo Kim help 512a254129eSJoonsoo Kim CMA allows to create CMA areas for particular purpose, mainly, 513a254129eSJoonsoo Kim used as device private area. This parameter sets the maximum 514a254129eSJoonsoo Kim number of CMA area in the system. 515a254129eSJoonsoo Kim 516a254129eSJoonsoo Kim If unsure, leave the default value "7". 517a254129eSJoonsoo Kim 518af8d417aSDan Streetmanconfig MEM_SOFT_DIRTY 519af8d417aSDan Streetman bool "Track memory changes" 520af8d417aSDan Streetman depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS 521af8d417aSDan Streetman select PROC_PAGE_MONITOR 5224e2e2770SSeth Jennings help 523af8d417aSDan Streetman This option enables memory changes tracking by introducing a 524af8d417aSDan Streetman soft-dirty bit on pte-s. This bit it set when someone writes 525af8d417aSDan Streetman into a page just as regular dirty bit, but unlike the latter 526af8d417aSDan Streetman it can be cleared by hands. 527af8d417aSDan Streetman 5281ad1335dSMike Rapoport See Documentation/admin-guide/mm/soft-dirty.rst for more details. 5294e2e2770SSeth Jennings 5302b281117SSeth Jenningsconfig ZSWAP 5312b281117SSeth Jennings bool "Compressed cache for swap pages (EXPERIMENTAL)" 5322b281117SSeth Jennings depends on FRONTSWAP && CRYPTO=y 5332b281117SSeth Jennings select CRYPTO_LZO 53412d79d64SDan Streetman select ZPOOL 5352b281117SSeth Jennings help 5362b281117SSeth Jennings A lightweight compressed cache for swap pages. It takes 5372b281117SSeth Jennings pages that are in the process of being swapped out and attempts to 5382b281117SSeth Jennings compress them into a dynamically allocated RAM-based memory pool. 5392b281117SSeth Jennings This can result in a significant I/O reduction on swap device and, 5402b281117SSeth Jennings in the case where decompressing from RAM is faster that swap device 5412b281117SSeth Jennings reads, can also improve workload performance. 5422b281117SSeth Jennings 5432b281117SSeth Jennings This is marked experimental because it is a new feature (as of 5442b281117SSeth Jennings v3.11) that interacts heavily with memory reclaim. While these 5452b281117SSeth Jennings interactions don't cause any known issues on simple memory setups, 5462b281117SSeth Jennings they have not be fully explored on the large set of potential 5472b281117SSeth Jennings configurations and workloads that exist. 5482b281117SSeth Jennings 549af8d417aSDan Streetmanconfig ZPOOL 550af8d417aSDan Streetman tristate "Common API for compressed memory storage" 5510f8975ecSPavel Emelyanov help 552af8d417aSDan Streetman Compressed memory storage API. This allows using either zbud or 553af8d417aSDan Streetman zsmalloc. 5540f8975ecSPavel Emelyanov 555af8d417aSDan Streetmanconfig ZBUD 5569a001fc1SVitaly Wool tristate "Low (Up to 2x) density storage for compressed pages" 557af8d417aSDan Streetman help 558af8d417aSDan Streetman A special purpose allocator for storing compressed pages. 559af8d417aSDan Streetman It is designed to store up to two compressed pages per physical 560af8d417aSDan Streetman page. While this design limits storage density, it has simple and 561af8d417aSDan Streetman deterministic reclaim properties that make it preferable to a higher 562af8d417aSDan Streetman density approach when reclaim will be used. 563bcf1647dSMinchan Kim 5649a001fc1SVitaly Woolconfig Z3FOLD 5659a001fc1SVitaly Wool tristate "Up to 3x density storage for compressed pages" 5669a001fc1SVitaly Wool depends on ZPOOL 5679a001fc1SVitaly Wool help 5689a001fc1SVitaly Wool A special purpose allocator for storing compressed pages. 5699a001fc1SVitaly Wool It is designed to store up to three compressed pages per physical 5709a001fc1SVitaly Wool page. It is a ZBUD derivative so the simplicity and determinism are 5719a001fc1SVitaly Wool still there. 5729a001fc1SVitaly Wool 573bcf1647dSMinchan Kimconfig ZSMALLOC 574d867f203SMinchan Kim tristate "Memory allocator for compressed pages" 575bcf1647dSMinchan Kim depends on MMU 576bcf1647dSMinchan Kim help 577bcf1647dSMinchan Kim zsmalloc is a slab-based memory allocator designed to store 578bcf1647dSMinchan Kim compressed RAM pages. zsmalloc uses virtual memory mapping 579bcf1647dSMinchan Kim in order to reduce fragmentation. However, this results in a 580bcf1647dSMinchan Kim non-standard allocator interface where a handle, not a pointer, is 581bcf1647dSMinchan Kim returned by an alloc(). This handle must be mapped in order to 582bcf1647dSMinchan Kim access the allocated space. 583bcf1647dSMinchan Kim 584bcf1647dSMinchan Kimconfig PGTABLE_MAPPING 585bcf1647dSMinchan Kim bool "Use page table mapping to access object in zsmalloc" 586bcf1647dSMinchan Kim depends on ZSMALLOC 587bcf1647dSMinchan Kim help 588bcf1647dSMinchan Kim By default, zsmalloc uses a copy-based object mapping method to 589bcf1647dSMinchan Kim access allocations that span two pages. However, if a particular 590bcf1647dSMinchan Kim architecture (ex, ARM) performs VM mapping faster than copying, 591bcf1647dSMinchan Kim then you should select this. This causes zsmalloc to use page table 592bcf1647dSMinchan Kim mapping rather than copying for object mapping. 593bcf1647dSMinchan Kim 5942216ee85SBen Hutchings You can check speed with zsmalloc benchmark: 5952216ee85SBen Hutchings https://github.com/spartacus06/zsmapbench 5969e5c33d7SMark Salter 5970f050d99SGanesh Mahendranconfig ZSMALLOC_STAT 5980f050d99SGanesh Mahendran bool "Export zsmalloc statistics" 5990f050d99SGanesh Mahendran depends on ZSMALLOC 6000f050d99SGanesh Mahendran select DEBUG_FS 6010f050d99SGanesh Mahendran help 6020f050d99SGanesh Mahendran This option enables code in the zsmalloc to collect various 6030f050d99SGanesh Mahendran statistics about whats happening in zsmalloc and exports that 6040f050d99SGanesh Mahendran information to userspace via debugfs. 6050f050d99SGanesh Mahendran If unsure, say N. 6060f050d99SGanesh Mahendran 6079e5c33d7SMark Salterconfig GENERIC_EARLY_IOREMAP 6089e5c33d7SMark Salter bool 609042d27acSHelge Deller 610042d27acSHelge Dellerconfig MAX_STACK_SIZE_MB 611042d27acSHelge Deller int "Maximum user stack size for 32-bit processes (MB)" 612042d27acSHelge Deller default 80 613042d27acSHelge Deller range 8 2048 614042d27acSHelge Deller depends on STACK_GROWSUP && (!64BIT || COMPAT) 615042d27acSHelge Deller help 616042d27acSHelge Deller This is the maximum stack size in Megabytes in the VM layout of 32-bit 617042d27acSHelge Deller user processes when the stack grows upwards (currently only on parisc 6185f171577SJames Hogan arch). The stack will be located at the highest memory address minus 6195f171577SJames Hogan the given value, unless the RLIMIT_STACK hard limit is changed to a 6205f171577SJames Hogan smaller value in which case that is used. 621042d27acSHelge Deller 622042d27acSHelge Deller A sane initial value is 80 MB. 6233a80a7faSMel Gorman 6243a80a7faSMel Gormanconfig DEFERRED_STRUCT_PAGE_INIT 6251ce22103SVlastimil Babka bool "Defer initialisation of struct pages to kthreads" 626d39f8fb4SMike Rapoport depends on SPARSEMEM 627ab1e8d89SPavel Tatashin depends on !NEED_PER_CPU_KM 628889c695dSPasha Tatashin depends on 64BIT 6293a80a7faSMel Gorman help 6303a80a7faSMel Gorman Ordinarily all struct pages are initialised during early boot in a 6313a80a7faSMel Gorman single thread. On very large machines this can take a considerable 6323a80a7faSMel Gorman amount of time. If this option is set, large machines will bring up 6333a80a7faSMel Gorman a subset of memmap at boot and then initialise the rest in parallel 6341ce22103SVlastimil Babka by starting one-off "pgdatinitX" kernel thread for each node X. This 6351ce22103SVlastimil Babka has a potential performance impact on processes running early in the 6361ce22103SVlastimil Babka lifetime of the system until these kthreads finish the 6371ce22103SVlastimil Babka initialisation. 638033fbae9SDan Williams 63933c3fc71SVladimir Davydovconfig IDLE_PAGE_TRACKING 64033c3fc71SVladimir Davydov bool "Enable idle page tracking" 64133c3fc71SVladimir Davydov depends on SYSFS && MMU 64233c3fc71SVladimir Davydov select PAGE_EXTENSION if !64BIT 64333c3fc71SVladimir Davydov help 64433c3fc71SVladimir Davydov This feature allows to estimate the amount of user pages that have 64533c3fc71SVladimir Davydov not been touched during a given period of time. This information can 64633c3fc71SVladimir Davydov be useful to tune memory cgroup limits and/or for job placement 64733c3fc71SVladimir Davydov within a compute cluster. 64833c3fc71SVladimir Davydov 6491ad1335dSMike Rapoport See Documentation/admin-guide/mm/idle_page_tracking.rst for 6501ad1335dSMike Rapoport more details. 65133c3fc71SVladimir Davydov 652*17596731SRobin Murphyconfig ARCH_HAS_PTE_DEVMAP 65365f7d049SOliver O'Halloran bool 65465f7d049SOliver O'Halloran 655033fbae9SDan Williamsconfig ZONE_DEVICE 6565042db43SJérôme Glisse bool "Device memory (pmem, HMM, etc...) hotplug support" 657033fbae9SDan Williams depends on MEMORY_HOTPLUG 658033fbae9SDan Williams depends on MEMORY_HOTREMOVE 65999490f16SDan Williams depends on SPARSEMEM_VMEMMAP 660*17596731SRobin Murphy depends on ARCH_HAS_PTE_DEVMAP 6613a08cd52SMatthew Wilcox select XARRAY_MULTI 662033fbae9SDan Williams 663033fbae9SDan Williams help 664033fbae9SDan Williams Device memory hotplug support allows for establishing pmem, 665033fbae9SDan Williams or other device driver discovered memory regions, in the 666033fbae9SDan Williams memmap. This allows pfn_to_page() lookups of otherwise 667033fbae9SDan Williams "device-physical" addresses which is needed for using a DAX 668033fbae9SDan Williams mapping in an O_DIRECT operation, among other things. 669033fbae9SDan Williams 670033fbae9SDan Williams If FS_DAX is enabled, then say Y. 67106a660adSLinus Torvalds 6726b368cd4SJérôme Glisseconfig MIGRATE_VMA_HELPER 6736b368cd4SJérôme Glisse bool 6746b368cd4SJérôme Glisse 675e7638488SDan Williamsconfig DEV_PAGEMAP_OPS 676e7638488SDan Williams bool 677e7638488SDan Williams 678c0b12405SJérôme Glisseconfig HMM_MIRROR 679c0b12405SJérôme Glisse bool "HMM mirror CPU page table into a device page table" 68043535b0aSChristoph Hellwig depends on (X86_64 || PPC64) 68143535b0aSChristoph Hellwig depends on MMU && 64BIT 68243535b0aSChristoph Hellwig select MMU_NOTIFIER 683c0b12405SJérôme Glisse help 684c0b12405SJérôme Glisse Select HMM_MIRROR if you want to mirror range of the CPU page table of a 685c0b12405SJérôme Glisse process into a device page table. Here, mirror means "keep synchronized". 686c0b12405SJérôme Glisse Prerequisites: the device must provide the ability to write-protect its 687c0b12405SJérôme Glisse page tables (at PAGE_SIZE granularity), and must be able to recover from 688c0b12405SJérôme Glisse the resulting potential page faults. 689c0b12405SJérôme Glisse 6905042db43SJérôme Glisseconfig DEVICE_PRIVATE 6915042db43SJérôme Glisse bool "Unaddressable device memory (GPU memory, ...)" 6927328d9ccSChristoph Hellwig depends on ZONE_DEVICE 693e7638488SDan Williams select DEV_PAGEMAP_OPS 6945042db43SJérôme Glisse 6955042db43SJérôme Glisse help 6965042db43SJérôme Glisse Allows creation of struct pages to represent unaddressable device 6975042db43SJérôme Glisse memory; i.e., memory that is only accessible from the device (or 6985042db43SJérôme Glisse group of devices). You likely also want to select HMM_MIRROR. 6995042db43SJérôme Glisse 7008025e5ddSJan Karaconfig FRAME_VECTOR 7018025e5ddSJan Kara bool 70263c17fb8SDave Hansen 70363c17fb8SDave Hansenconfig ARCH_USES_HIGH_VMA_FLAGS 70463c17fb8SDave Hansen bool 70566d37570SDave Hansenconfig ARCH_HAS_PKEYS 70666d37570SDave Hansen bool 70730a5b536SDennis Zhou 70830a5b536SDennis Zhouconfig PERCPU_STATS 70930a5b536SDennis Zhou bool "Collect percpu memory statistics" 71030a5b536SDennis Zhou help 71130a5b536SDennis Zhou This feature collects and exposes statistics via debugfs. The 71230a5b536SDennis Zhou information includes global and per chunk statistics, which can 71330a5b536SDennis Zhou be used to help understand percpu memory usage. 71464c349f4SKirill A. Shutemov 71564c349f4SKirill A. Shutemovconfig GUP_BENCHMARK 71664c349f4SKirill A. Shutemov bool "Enable infrastructure for get_user_pages_fast() benchmarking" 71764c349f4SKirill A. Shutemov help 71864c349f4SKirill A. Shutemov Provides /sys/kernel/debug/gup_benchmark that helps with testing 71964c349f4SKirill A. Shutemov performance of get_user_pages_fast(). 72064c349f4SKirill A. Shutemov 72164c349f4SKirill A. Shutemov See tools/testing/selftests/vm/gup_benchmark.c 7223010a5eaSLaurent Dufour 72339656e83SChristoph Hellwigconfig GUP_GET_PTE_LOW_HIGH 72439656e83SChristoph Hellwig bool 72539656e83SChristoph Hellwig 7263010a5eaSLaurent Dufourconfig ARCH_HAS_PTE_SPECIAL 7273010a5eaSLaurent Dufour bool 72859e0b520SChristoph Hellwig 729cbd34da7SChristoph Hellwig# 730cbd34da7SChristoph Hellwig# Some architectures require a special hugepage directory format that is 731cbd34da7SChristoph Hellwig# required to support multiple hugepage sizes. For example a4fe3ce76 732cbd34da7SChristoph Hellwig# "powerpc/mm: Allow more flexible layouts for hugepage pagetables" 733cbd34da7SChristoph Hellwig# introduced it on powerpc. This allows for a more flexible hugepage 734cbd34da7SChristoph Hellwig# pagetable layouts. 735cbd34da7SChristoph Hellwig# 736cbd34da7SChristoph Hellwigconfig ARCH_HAS_HUGEPD 737cbd34da7SChristoph Hellwig bool 738cbd34da7SChristoph Hellwig 73959e0b520SChristoph Hellwigendmenu 740