1ec8f24b7SThomas Gleixner# SPDX-License-Identifier: GPL-2.0-only 259e0b520SChristoph Hellwig 359e0b520SChristoph Hellwigmenu "Memory Management options" 459e0b520SChristoph Hellwig 57b42f104SJohannes Weiner# 67b42f104SJohannes Weiner# For some reason microblaze and nios2 hard code SWAP=n. Hopefully we can 77b42f104SJohannes Weiner# add proper SWAP support to them, in which case this can be remove. 87b42f104SJohannes Weiner# 97b42f104SJohannes Weinerconfig ARCH_NO_SWAP 107b42f104SJohannes Weiner bool 117b42f104SJohannes Weiner 12b3fbd58fSJohannes Weinerconfig ZPOOL 13b3fbd58fSJohannes Weiner bool 14b3fbd58fSJohannes Weiner 15519bcb79SJohannes Weinermenuconfig SWAP 167b42f104SJohannes Weiner bool "Support for paging of anonymous memory (swap)" 177b42f104SJohannes Weiner depends on MMU && BLOCK && !ARCH_NO_SWAP 187b42f104SJohannes Weiner default y 197b42f104SJohannes Weiner help 207b42f104SJohannes Weiner This option allows you to choose whether you want to have support 217b42f104SJohannes Weiner for so called swap devices or swap files in your kernel that are 227b42f104SJohannes Weiner used to provide more virtual memory than the actual RAM present 237b42f104SJohannes Weiner in your computer. If unsure say Y. 247b42f104SJohannes Weiner 25519bcb79SJohannes Weinerconfig ZSWAP 26fcab9b44SDavid Heidelberg bool "Compressed cache for swap pages" 27b3fbd58fSJohannes Weiner depends on SWAP 28519bcb79SJohannes Weiner select FRONTSWAP 29b3fbd58fSJohannes Weiner select CRYPTO 30519bcb79SJohannes Weiner select ZPOOL 31519bcb79SJohannes Weiner help 32519bcb79SJohannes Weiner A lightweight compressed cache for swap pages. It takes 33519bcb79SJohannes Weiner pages that are in the process of being swapped out and attempts to 34519bcb79SJohannes Weiner compress them into a dynamically allocated RAM-based memory pool. 35519bcb79SJohannes Weiner This can result in a significant I/O reduction on swap device and, 361a44131dSSophia Gabriella in the case where decompressing from RAM is faster than swap device 37519bcb79SJohannes Weiner reads, can also improve workload performance. 38519bcb79SJohannes Weiner 39b3fbd58fSJohannes Weinerconfig ZSWAP_DEFAULT_ON 40b3fbd58fSJohannes Weiner bool "Enable the compressed cache for swap pages by default" 41b3fbd58fSJohannes Weiner depends on ZSWAP 42b3fbd58fSJohannes Weiner help 43b3fbd58fSJohannes Weiner If selected, the compressed cache for swap pages will be enabled 44b3fbd58fSJohannes Weiner at boot, otherwise it will be disabled. 45b3fbd58fSJohannes Weiner 46b3fbd58fSJohannes Weiner The selection made here can be overridden by using the kernel 47b3fbd58fSJohannes Weiner command line 'zswap.enabled=' option. 48b3fbd58fSJohannes Weiner 49b9c91c43SYosry Ahmedconfig ZSWAP_EXCLUSIVE_LOADS_DEFAULT_ON 50b9c91c43SYosry Ahmed bool "Invalidate zswap entries when pages are loaded" 51b9c91c43SYosry Ahmed depends on ZSWAP 52b9c91c43SYosry Ahmed help 53b9c91c43SYosry Ahmed If selected, exclusive loads for zswap will be enabled at boot, 54b9c91c43SYosry Ahmed otherwise it will be disabled. 55b9c91c43SYosry Ahmed 56b9c91c43SYosry Ahmed If exclusive loads are enabled, when a page is loaded from zswap, 57b9c91c43SYosry Ahmed the zswap entry is invalidated at once, as opposed to leaving it 58b9c91c43SYosry Ahmed in zswap until the swap entry is freed. 59b9c91c43SYosry Ahmed 60b9c91c43SYosry Ahmed This avoids having two copies of the same page in memory 61b9c91c43SYosry Ahmed (compressed and uncompressed) after faulting in a page from zswap. 62b9c91c43SYosry Ahmed The cost is that if the page was never dirtied and needs to be 63b9c91c43SYosry Ahmed swapped out again, it will be re-compressed. 64b9c91c43SYosry Ahmed 65519bcb79SJohannes Weinerchoice 66b3fbd58fSJohannes Weiner prompt "Default compressor" 67519bcb79SJohannes Weiner depends on ZSWAP 68519bcb79SJohannes Weiner default ZSWAP_COMPRESSOR_DEFAULT_LZO 69519bcb79SJohannes Weiner help 70519bcb79SJohannes Weiner Selects the default compression algorithm for the compressed cache 71519bcb79SJohannes Weiner for swap pages. 72519bcb79SJohannes Weiner 73519bcb79SJohannes Weiner For an overview what kind of performance can be expected from 74519bcb79SJohannes Weiner a particular compression algorithm please refer to the benchmarks 75519bcb79SJohannes Weiner available at the following LWN page: 76519bcb79SJohannes Weiner https://lwn.net/Articles/751795/ 77519bcb79SJohannes Weiner 78519bcb79SJohannes Weiner If in doubt, select 'LZO'. 79519bcb79SJohannes Weiner 80519bcb79SJohannes Weiner The selection made here can be overridden by using the kernel 81519bcb79SJohannes Weiner command line 'zswap.compressor=' option. 82519bcb79SJohannes Weiner 83519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_DEFLATE 84519bcb79SJohannes Weiner bool "Deflate" 85519bcb79SJohannes Weiner select CRYPTO_DEFLATE 86519bcb79SJohannes Weiner help 87519bcb79SJohannes Weiner Use the Deflate algorithm as the default compression algorithm. 88519bcb79SJohannes Weiner 89519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_LZO 90519bcb79SJohannes Weiner bool "LZO" 91519bcb79SJohannes Weiner select CRYPTO_LZO 92519bcb79SJohannes Weiner help 93519bcb79SJohannes Weiner Use the LZO algorithm as the default compression algorithm. 94519bcb79SJohannes Weiner 95519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_842 96519bcb79SJohannes Weiner bool "842" 97519bcb79SJohannes Weiner select CRYPTO_842 98519bcb79SJohannes Weiner help 99519bcb79SJohannes Weiner Use the 842 algorithm as the default compression algorithm. 100519bcb79SJohannes Weiner 101519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_LZ4 102519bcb79SJohannes Weiner bool "LZ4" 103519bcb79SJohannes Weiner select CRYPTO_LZ4 104519bcb79SJohannes Weiner help 105519bcb79SJohannes Weiner Use the LZ4 algorithm as the default compression algorithm. 106519bcb79SJohannes Weiner 107519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_LZ4HC 108519bcb79SJohannes Weiner bool "LZ4HC" 109519bcb79SJohannes Weiner select CRYPTO_LZ4HC 110519bcb79SJohannes Weiner help 111519bcb79SJohannes Weiner Use the LZ4HC algorithm as the default compression algorithm. 112519bcb79SJohannes Weiner 113519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_ZSTD 114519bcb79SJohannes Weiner bool "zstd" 115519bcb79SJohannes Weiner select CRYPTO_ZSTD 116519bcb79SJohannes Weiner help 117519bcb79SJohannes Weiner Use the zstd algorithm as the default compression algorithm. 118519bcb79SJohannes Weinerendchoice 119519bcb79SJohannes Weiner 120519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT 121519bcb79SJohannes Weiner string 122519bcb79SJohannes Weiner depends on ZSWAP 123519bcb79SJohannes Weiner default "deflate" if ZSWAP_COMPRESSOR_DEFAULT_DEFLATE 124519bcb79SJohannes Weiner default "lzo" if ZSWAP_COMPRESSOR_DEFAULT_LZO 125519bcb79SJohannes Weiner default "842" if ZSWAP_COMPRESSOR_DEFAULT_842 126519bcb79SJohannes Weiner default "lz4" if ZSWAP_COMPRESSOR_DEFAULT_LZ4 127519bcb79SJohannes Weiner default "lz4hc" if ZSWAP_COMPRESSOR_DEFAULT_LZ4HC 128519bcb79SJohannes Weiner default "zstd" if ZSWAP_COMPRESSOR_DEFAULT_ZSTD 129519bcb79SJohannes Weiner default "" 130519bcb79SJohannes Weiner 131519bcb79SJohannes Weinerchoice 132b3fbd58fSJohannes Weiner prompt "Default allocator" 133519bcb79SJohannes Weiner depends on ZSWAP 134519bcb79SJohannes Weiner default ZSWAP_ZPOOL_DEFAULT_ZBUD 135519bcb79SJohannes Weiner help 136519bcb79SJohannes Weiner Selects the default allocator for the compressed cache for 137519bcb79SJohannes Weiner swap pages. 138519bcb79SJohannes Weiner The default is 'zbud' for compatibility, however please do 139519bcb79SJohannes Weiner read the description of each of the allocators below before 140519bcb79SJohannes Weiner making a right choice. 141519bcb79SJohannes Weiner 142519bcb79SJohannes Weiner The selection made here can be overridden by using the kernel 143519bcb79SJohannes Weiner command line 'zswap.zpool=' option. 144519bcb79SJohannes Weiner 145519bcb79SJohannes Weinerconfig ZSWAP_ZPOOL_DEFAULT_ZBUD 146519bcb79SJohannes Weiner bool "zbud" 147519bcb79SJohannes Weiner select ZBUD 148519bcb79SJohannes Weiner help 149519bcb79SJohannes Weiner Use the zbud allocator as the default allocator. 150519bcb79SJohannes Weiner 151519bcb79SJohannes Weinerconfig ZSWAP_ZPOOL_DEFAULT_Z3FOLD 152519bcb79SJohannes Weiner bool "z3fold" 153519bcb79SJohannes Weiner select Z3FOLD 154519bcb79SJohannes Weiner help 155519bcb79SJohannes Weiner Use the z3fold allocator as the default allocator. 156519bcb79SJohannes Weiner 157519bcb79SJohannes Weinerconfig ZSWAP_ZPOOL_DEFAULT_ZSMALLOC 158519bcb79SJohannes Weiner bool "zsmalloc" 159519bcb79SJohannes Weiner select ZSMALLOC 160519bcb79SJohannes Weiner help 161519bcb79SJohannes Weiner Use the zsmalloc allocator as the default allocator. 162519bcb79SJohannes Weinerendchoice 163519bcb79SJohannes Weiner 164519bcb79SJohannes Weinerconfig ZSWAP_ZPOOL_DEFAULT 165519bcb79SJohannes Weiner string 166519bcb79SJohannes Weiner depends on ZSWAP 167519bcb79SJohannes Weiner default "zbud" if ZSWAP_ZPOOL_DEFAULT_ZBUD 168519bcb79SJohannes Weiner default "z3fold" if ZSWAP_ZPOOL_DEFAULT_Z3FOLD 169519bcb79SJohannes Weiner default "zsmalloc" if ZSWAP_ZPOOL_DEFAULT_ZSMALLOC 170519bcb79SJohannes Weiner default "" 171519bcb79SJohannes Weiner 172519bcb79SJohannes Weinerconfig ZBUD 173b3fbd58fSJohannes Weiner tristate "2:1 compression allocator (zbud)" 174b3fbd58fSJohannes Weiner depends on ZSWAP 175519bcb79SJohannes Weiner help 176519bcb79SJohannes Weiner A special purpose allocator for storing compressed pages. 177519bcb79SJohannes Weiner It is designed to store up to two compressed pages per physical 178519bcb79SJohannes Weiner page. While this design limits storage density, it has simple and 179519bcb79SJohannes Weiner deterministic reclaim properties that make it preferable to a higher 180519bcb79SJohannes Weiner density approach when reclaim will be used. 181519bcb79SJohannes Weiner 182519bcb79SJohannes Weinerconfig Z3FOLD 183b3fbd58fSJohannes Weiner tristate "3:1 compression allocator (z3fold)" 184b3fbd58fSJohannes Weiner depends on ZSWAP 185519bcb79SJohannes Weiner help 186519bcb79SJohannes Weiner A special purpose allocator for storing compressed pages. 187519bcb79SJohannes Weiner It is designed to store up to three compressed pages per physical 188519bcb79SJohannes Weiner page. It is a ZBUD derivative so the simplicity and determinism are 189519bcb79SJohannes Weiner still there. 190519bcb79SJohannes Weiner 191519bcb79SJohannes Weinerconfig ZSMALLOC 192b3fbd58fSJohannes Weiner tristate 193b3fbd58fSJohannes Weiner prompt "N:1 compression allocator (zsmalloc)" if ZSWAP 194519bcb79SJohannes Weiner depends on MMU 195519bcb79SJohannes Weiner help 196519bcb79SJohannes Weiner zsmalloc is a slab-based memory allocator designed to store 197b3fbd58fSJohannes Weiner pages of various compression levels efficiently. It achieves 198b3fbd58fSJohannes Weiner the highest storage density with the least amount of fragmentation. 199519bcb79SJohannes Weiner 200519bcb79SJohannes Weinerconfig ZSMALLOC_STAT 201519bcb79SJohannes Weiner bool "Export zsmalloc statistics" 202519bcb79SJohannes Weiner depends on ZSMALLOC 203519bcb79SJohannes Weiner select DEBUG_FS 204519bcb79SJohannes Weiner help 205519bcb79SJohannes Weiner This option enables code in the zsmalloc to collect various 206519bcb79SJohannes Weiner statistics about what's happening in zsmalloc and exports that 207519bcb79SJohannes Weiner information to userspace via debugfs. 208519bcb79SJohannes Weiner If unsure, say N. 209519bcb79SJohannes Weiner 2104ff93b29SSergey Senozhatskyconfig ZSMALLOC_CHAIN_SIZE 2114ff93b29SSergey Senozhatsky int "Maximum number of physical pages per-zspage" 212b46402faSSergey Senozhatsky default 8 2134ff93b29SSergey Senozhatsky range 4 16 2144ff93b29SSergey Senozhatsky depends on ZSMALLOC 2154ff93b29SSergey Senozhatsky help 2164ff93b29SSergey Senozhatsky This option sets the upper limit on the number of physical pages 2174ff93b29SSergey Senozhatsky that a zmalloc page (zspage) can consist of. The optimal zspage 2184ff93b29SSergey Senozhatsky chain size is calculated for each size class during the 2194ff93b29SSergey Senozhatsky initialization of the pool. 2204ff93b29SSergey Senozhatsky 2214ff93b29SSergey Senozhatsky Changing this option can alter the characteristics of size classes, 2224ff93b29SSergey Senozhatsky such as the number of pages per zspage and the number of objects 2234ff93b29SSergey Senozhatsky per zspage. This can also result in different configurations of 2244ff93b29SSergey Senozhatsky the pool, as zsmalloc merges size classes with similar 2254ff93b29SSergey Senozhatsky characteristics. 2264ff93b29SSergey Senozhatsky 2274ff93b29SSergey Senozhatsky For more information, see zsmalloc documentation. 2284ff93b29SSergey Senozhatsky 229519bcb79SJohannes Weinermenu "SLAB allocator options" 230519bcb79SJohannes Weiner 2317b42f104SJohannes Weinerchoice 2327b42f104SJohannes Weiner prompt "Choose SLAB allocator" 2337b42f104SJohannes Weiner default SLUB 2347b42f104SJohannes Weiner help 2357b42f104SJohannes Weiner This option allows to select a slab allocator. 2367b42f104SJohannes Weiner 237eb07c4f3SVlastimil Babkaconfig SLAB_DEPRECATED 238eb07c4f3SVlastimil Babka bool "SLAB (DEPRECATED)" 2397b42f104SJohannes Weiner depends on !PREEMPT_RT 2407b42f104SJohannes Weiner help 241eb07c4f3SVlastimil Babka Deprecated and scheduled for removal in a few cycles. Replaced by 242eb07c4f3SVlastimil Babka SLUB. 243eb07c4f3SVlastimil Babka 244eb07c4f3SVlastimil Babka If you cannot migrate to SLUB, please contact linux-mm@kvack.org 245eb07c4f3SVlastimil Babka and the people listed in the SLAB ALLOCATOR section of MAINTAINERS 246eb07c4f3SVlastimil Babka file, explaining why. 247eb07c4f3SVlastimil Babka 2487b42f104SJohannes Weiner The regular slab allocator that is established and known to work 2497b42f104SJohannes Weiner well in all environments. It organizes cache hot objects in 2507b42f104SJohannes Weiner per cpu and per node queues. 2517b42f104SJohannes Weiner 2527b42f104SJohannes Weinerconfig SLUB 2537b42f104SJohannes Weiner bool "SLUB (Unqueued Allocator)" 2547b42f104SJohannes Weiner help 2557b42f104SJohannes Weiner SLUB is a slab allocator that minimizes cache line usage 2567b42f104SJohannes Weiner instead of managing queues of cached objects (SLAB approach). 2577b42f104SJohannes Weiner Per cpu caching is realized using slabs of objects instead 2587b42f104SJohannes Weiner of queues of objects. SLUB can use memory efficiently 2597b42f104SJohannes Weiner and has enhanced diagnostics. SLUB is the default choice for 2607b42f104SJohannes Weiner a slab allocator. 2617b42f104SJohannes Weiner 2627b42f104SJohannes Weinerendchoice 2637b42f104SJohannes Weiner 264eb07c4f3SVlastimil Babkaconfig SLAB 265eb07c4f3SVlastimil Babka bool 266eb07c4f3SVlastimil Babka default y 267eb07c4f3SVlastimil Babka depends on SLAB_DEPRECATED 268eb07c4f3SVlastimil Babka 269e240e53aSVlastimil Babkaconfig SLUB_TINY 270e240e53aSVlastimil Babka bool "Configure SLUB for minimal memory footprint" 271e240e53aSVlastimil Babka depends on SLUB && EXPERT 272e240e53aSVlastimil Babka select SLAB_MERGE_DEFAULT 273e240e53aSVlastimil Babka help 274e240e53aSVlastimil Babka Configures the SLUB allocator in a way to achieve minimal memory 275e240e53aSVlastimil Babka footprint, sacrificing scalability, debugging and other features. 276e240e53aSVlastimil Babka This is intended only for the smallest system that had used the 277e240e53aSVlastimil Babka SLOB allocator and is not recommended for systems with more than 278e240e53aSVlastimil Babka 16MB RAM. 279e240e53aSVlastimil Babka 280e240e53aSVlastimil Babka If unsure, say N. 281e240e53aSVlastimil Babka 2827b42f104SJohannes Weinerconfig SLAB_MERGE_DEFAULT 2837b42f104SJohannes Weiner bool "Allow slab caches to be merged" 2847b42f104SJohannes Weiner default y 2857b42f104SJohannes Weiner depends on SLAB || SLUB 2867b42f104SJohannes Weiner help 2877b42f104SJohannes Weiner For reduced kernel memory fragmentation, slab caches can be 2887b42f104SJohannes Weiner merged when they share the same size and other characteristics. 2897b42f104SJohannes Weiner This carries a risk of kernel heap overflows being able to 2907b42f104SJohannes Weiner overwrite objects from merged caches (and more easily control 2917b42f104SJohannes Weiner cache layout), which makes such heap attacks easier to exploit 2927b42f104SJohannes Weiner by attackers. By keeping caches unmerged, these kinds of exploits 2937b42f104SJohannes Weiner can usually only damage objects in the same cache. To disable 2947b42f104SJohannes Weiner merging at runtime, "slab_nomerge" can be passed on the kernel 2957b42f104SJohannes Weiner command line. 2967b42f104SJohannes Weiner 2977b42f104SJohannes Weinerconfig SLAB_FREELIST_RANDOM 2987b42f104SJohannes Weiner bool "Randomize slab freelist" 299e240e53aSVlastimil Babka depends on SLAB || (SLUB && !SLUB_TINY) 3007b42f104SJohannes Weiner help 3017b42f104SJohannes Weiner Randomizes the freelist order used on creating new pages. This 3027b42f104SJohannes Weiner security feature reduces the predictability of the kernel slab 3037b42f104SJohannes Weiner allocator against heap overflows. 3047b42f104SJohannes Weiner 3057b42f104SJohannes Weinerconfig SLAB_FREELIST_HARDENED 3067b42f104SJohannes Weiner bool "Harden slab freelist metadata" 307e240e53aSVlastimil Babka depends on SLAB || (SLUB && !SLUB_TINY) 3087b42f104SJohannes Weiner help 3097b42f104SJohannes Weiner Many kernel heap attacks try to target slab cache metadata and 3107b42f104SJohannes Weiner other infrastructure. This options makes minor performance 3117b42f104SJohannes Weiner sacrifices to harden the kernel slab allocator against common 3127b42f104SJohannes Weiner freelist exploit methods. Some slab implementations have more 3137b42f104SJohannes Weiner sanity-checking than others. This option is most effective with 3147b42f104SJohannes Weiner CONFIG_SLUB. 3157b42f104SJohannes Weiner 3160710d012SVlastimil Babkaconfig SLUB_STATS 3170710d012SVlastimil Babka default n 3180710d012SVlastimil Babka bool "Enable SLUB performance statistics" 319e240e53aSVlastimil Babka depends on SLUB && SYSFS && !SLUB_TINY 3200710d012SVlastimil Babka help 3210710d012SVlastimil Babka SLUB statistics are useful to debug SLUBs allocation behavior in 3220710d012SVlastimil Babka order find ways to optimize the allocator. This should never be 3230710d012SVlastimil Babka enabled for production use since keeping statistics slows down 3240710d012SVlastimil Babka the allocator by a few percentage points. The slabinfo command 3250710d012SVlastimil Babka supports the determination of the most active slabs to figure 3260710d012SVlastimil Babka out which slabs are relevant to a particular load. 3270710d012SVlastimil Babka Try running: slabinfo -DA 3280710d012SVlastimil Babka 329519bcb79SJohannes Weinerconfig SLUB_CPU_PARTIAL 330519bcb79SJohannes Weiner default y 331e240e53aSVlastimil Babka depends on SLUB && SMP && !SLUB_TINY 332519bcb79SJohannes Weiner bool "SLUB per cpu partial cache" 333519bcb79SJohannes Weiner help 334519bcb79SJohannes Weiner Per cpu partial caches accelerate objects allocation and freeing 335519bcb79SJohannes Weiner that is local to a processor at the price of more indeterminism 336519bcb79SJohannes Weiner in the latency of the free. On overflow these caches will be cleared 337519bcb79SJohannes Weiner which requires the taking of locks that may cause latency spikes. 338519bcb79SJohannes Weiner Typically one would choose no for a realtime system. 339519bcb79SJohannes Weiner 340*3c615294SGONG, Ruiqiconfig RANDOM_KMALLOC_CACHES 341*3c615294SGONG, Ruiqi default n 342*3c615294SGONG, Ruiqi depends on SLUB && !SLUB_TINY 343*3c615294SGONG, Ruiqi bool "Randomize slab caches for normal kmalloc" 344*3c615294SGONG, Ruiqi help 345*3c615294SGONG, Ruiqi A hardening feature that creates multiple copies of slab caches for 346*3c615294SGONG, Ruiqi normal kmalloc allocation and makes kmalloc randomly pick one based 347*3c615294SGONG, Ruiqi on code address, which makes the attackers more difficult to spray 348*3c615294SGONG, Ruiqi vulnerable memory objects on the heap for the purpose of exploiting 349*3c615294SGONG, Ruiqi memory vulnerabilities. 350*3c615294SGONG, Ruiqi 351*3c615294SGONG, Ruiqi Currently the number of copies is set to 16, a reasonably large value 352*3c615294SGONG, Ruiqi that effectively diverges the memory objects allocated for different 353*3c615294SGONG, Ruiqi subsystems or modules into different caches, at the expense of a 354*3c615294SGONG, Ruiqi limited degree of memory and CPU overhead that relates to hardware and 355*3c615294SGONG, Ruiqi system workload. 356*3c615294SGONG, Ruiqi 357519bcb79SJohannes Weinerendmenu # SLAB allocator options 358519bcb79SJohannes Weiner 3597b42f104SJohannes Weinerconfig SHUFFLE_PAGE_ALLOCATOR 3607b42f104SJohannes Weiner bool "Page allocator randomization" 3617b42f104SJohannes Weiner default SLAB_FREELIST_RANDOM && ACPI_NUMA 3627b42f104SJohannes Weiner help 3637b42f104SJohannes Weiner Randomization of the page allocator improves the average 3647b42f104SJohannes Weiner utilization of a direct-mapped memory-side-cache. See section 3657b42f104SJohannes Weiner 5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI 3667b42f104SJohannes Weiner 6.2a specification for an example of how a platform advertises 3677b42f104SJohannes Weiner the presence of a memory-side-cache. There are also incidental 3687b42f104SJohannes Weiner security benefits as it reduces the predictability of page 3697b42f104SJohannes Weiner allocations to compliment SLAB_FREELIST_RANDOM, but the 37023baf831SKirill A. Shutemov default granularity of shuffling on the MAX_ORDER i.e, 10th 37123baf831SKirill A. Shutemov order of pages is selected based on cache utilization benefits 37223baf831SKirill A. Shutemov on x86. 3737b42f104SJohannes Weiner 3747b42f104SJohannes Weiner While the randomization improves cache utilization it may 3757b42f104SJohannes Weiner negatively impact workloads on platforms without a cache. For 3767b42f104SJohannes Weiner this reason, by default, the randomization is enabled only 3777b42f104SJohannes Weiner after runtime detection of a direct-mapped memory-side-cache. 3787b42f104SJohannes Weiner Otherwise, the randomization may be force enabled with the 3797b42f104SJohannes Weiner 'page_alloc.shuffle' kernel command line parameter. 3807b42f104SJohannes Weiner 3817b42f104SJohannes Weiner Say Y if unsure. 3827b42f104SJohannes Weiner 3830710d012SVlastimil Babkaconfig COMPAT_BRK 3840710d012SVlastimil Babka bool "Disable heap randomization" 3850710d012SVlastimil Babka default y 3860710d012SVlastimil Babka help 3870710d012SVlastimil Babka Randomizing heap placement makes heap exploits harder, but it 3880710d012SVlastimil Babka also breaks ancient binaries (including anything libc5 based). 3890710d012SVlastimil Babka This option changes the bootup default to heap randomization 3900710d012SVlastimil Babka disabled, and can be overridden at runtime by setting 3910710d012SVlastimil Babka /proc/sys/kernel/randomize_va_space to 2. 3920710d012SVlastimil Babka 3930710d012SVlastimil Babka On non-ancient distros (post-2000 ones) N is usually a safe choice. 3940710d012SVlastimil Babka 3950710d012SVlastimil Babkaconfig MMAP_ALLOW_UNINITIALIZED 3960710d012SVlastimil Babka bool "Allow mmapped anonymous memory to be uninitialized" 3970710d012SVlastimil Babka depends on EXPERT && !MMU 3980710d012SVlastimil Babka default n 3990710d012SVlastimil Babka help 4000710d012SVlastimil Babka Normally, and according to the Linux spec, anonymous memory obtained 4010710d012SVlastimil Babka from mmap() has its contents cleared before it is passed to 4020710d012SVlastimil Babka userspace. Enabling this config option allows you to request that 4030710d012SVlastimil Babka mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus 4040710d012SVlastimil Babka providing a huge performance boost. If this option is not enabled, 4050710d012SVlastimil Babka then the flag will be ignored. 4060710d012SVlastimil Babka 4070710d012SVlastimil Babka This is taken advantage of by uClibc's malloc(), and also by 4080710d012SVlastimil Babka ELF-FDPIC binfmt's brk and stack allocator. 4090710d012SVlastimil Babka 4100710d012SVlastimil Babka Because of the obvious security issues, this option should only be 4110710d012SVlastimil Babka enabled on embedded devices where you control what is run in 4120710d012SVlastimil Babka userspace. Since that isn't generally a problem on no-MMU systems, 4130710d012SVlastimil Babka it is normally safe to say Y here. 4140710d012SVlastimil Babka 4150710d012SVlastimil Babka See Documentation/admin-guide/mm/nommu-mmap.rst for more information. 4160710d012SVlastimil Babka 417e1785e85SDave Hansenconfig SELECT_MEMORY_MODEL 418e1785e85SDave Hansen def_bool y 419a8826eebSKees Cook depends on ARCH_SELECT_MEMORY_MODEL 420e1785e85SDave Hansen 4213a9da765SDave Hansenchoice 4223a9da765SDave Hansen prompt "Memory model" 423e1785e85SDave Hansen depends on SELECT_MEMORY_MODEL 424d41dee36SAndy Whitcroft default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT 425e1785e85SDave Hansen default FLATMEM_MANUAL 426d66d109dSMike Rapoport help 427d66d109dSMike Rapoport This option allows you to change some of the ways that 428d66d109dSMike Rapoport Linux manages its memory internally. Most users will 429d66d109dSMike Rapoport only have one option here selected by the architecture 430d66d109dSMike Rapoport configuration. This is normal. 4313a9da765SDave Hansen 432e1785e85SDave Hansenconfig FLATMEM_MANUAL 4333a9da765SDave Hansen bool "Flat Memory" 434bb1c50d3SMike Rapoport depends on !ARCH_SPARSEMEM_ENABLE || ARCH_FLATMEM_ENABLE 4353a9da765SDave Hansen help 436d66d109dSMike Rapoport This option is best suited for non-NUMA systems with 437d66d109dSMike Rapoport flat address space. The FLATMEM is the most efficient 438d66d109dSMike Rapoport system in terms of performance and resource consumption 439d66d109dSMike Rapoport and it is the best option for smaller systems. 4403a9da765SDave Hansen 441d66d109dSMike Rapoport For systems that have holes in their physical address 442d66d109dSMike Rapoport spaces and for features like NUMA and memory hotplug, 443dd33d29aSRandy Dunlap choose "Sparse Memory". 444d41dee36SAndy Whitcroft 445d41dee36SAndy Whitcroft If unsure, choose this option (Flat Memory) over any other. 4463a9da765SDave Hansen 447d41dee36SAndy Whitcroftconfig SPARSEMEM_MANUAL 448d41dee36SAndy Whitcroft bool "Sparse Memory" 449d41dee36SAndy Whitcroft depends on ARCH_SPARSEMEM_ENABLE 450d41dee36SAndy Whitcroft help 451d41dee36SAndy Whitcroft This will be the only option for some systems, including 452d66d109dSMike Rapoport memory hot-plug systems. This is normal. 453d41dee36SAndy Whitcroft 454d66d109dSMike Rapoport This option provides efficient support for systems with 455d66d109dSMike Rapoport holes is their physical address space and allows memory 456d66d109dSMike Rapoport hot-plug and hot-remove. 457d41dee36SAndy Whitcroft 458d66d109dSMike Rapoport If unsure, choose "Flat Memory" over this option. 459d41dee36SAndy Whitcroft 4603a9da765SDave Hansenendchoice 4613a9da765SDave Hansen 462d41dee36SAndy Whitcroftconfig SPARSEMEM 463d41dee36SAndy Whitcroft def_bool y 4641a83e175SRussell King depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL 465d41dee36SAndy Whitcroft 466e1785e85SDave Hansenconfig FLATMEM 467e1785e85SDave Hansen def_bool y 468bb1c50d3SMike Rapoport depends on !SPARSEMEM || FLATMEM_MANUAL 469d41dee36SAndy Whitcroft 47093b7504eSDave Hansen# 4713e347261SBob Picco# SPARSEMEM_EXTREME (which is the default) does some bootmem 472c89ab04fSMike Rapoport# allocations when sparse_init() is called. If this cannot 4733e347261SBob Picco# be done on your architecture, select this option. However, 4743e347261SBob Picco# statically allocating the mem_section[] array can potentially 4753e347261SBob Picco# consume vast quantities of .bss, so be careful. 4763e347261SBob Picco# 4773e347261SBob Picco# This option will also potentially produce smaller runtime code 4783e347261SBob Picco# with gcc 3.4 and later. 4793e347261SBob Picco# 4803e347261SBob Piccoconfig SPARSEMEM_STATIC 4819ba16087SJan Beulich bool 4823e347261SBob Picco 4833e347261SBob Picco# 48444c09201SMatt LaPlante# Architecture platforms which require a two level mem_section in SPARSEMEM 485802f192eSBob Picco# must select this option. This is usually for architecture platforms with 486802f192eSBob Picco# an extremely sparse physical address space. 487802f192eSBob Picco# 4883e347261SBob Piccoconfig SPARSEMEM_EXTREME 4893e347261SBob Picco def_bool y 4903e347261SBob Picco depends on SPARSEMEM && !SPARSEMEM_STATIC 4914c21e2f2SHugh Dickins 49229c71111SAndy Whitcroftconfig SPARSEMEM_VMEMMAP_ENABLE 4939ba16087SJan Beulich bool 49429c71111SAndy Whitcroft 49529c71111SAndy Whitcroftconfig SPARSEMEM_VMEMMAP 496a5ee6daaSGeoff Levand bool "Sparse Memory virtual memmap" 497a5ee6daaSGeoff Levand depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE 498a5ee6daaSGeoff Levand default y 499a5ee6daaSGeoff Levand help 500a5ee6daaSGeoff Levand SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise 501a5ee6daaSGeoff Levand pfn_to_page and page_to_pfn operations. This is the most 502a5ee6daaSGeoff Levand efficient option when sufficient kernel resources are available. 5030b376f1eSAneesh Kumar K.V# 5040b376f1eSAneesh Kumar K.V# Select this config option from the architecture Kconfig, if it is preferred 5050b376f1eSAneesh Kumar K.V# to enable the feature of HugeTLB/dev_dax vmemmap optimization. 5060b376f1eSAneesh Kumar K.V# 5070b376f1eSAneesh Kumar K.Vconfig ARCH_WANT_OPTIMIZE_VMEMMAP 5080b376f1eSAneesh Kumar K.V bool 50929c71111SAndy Whitcroft 51070210ed9SPhilipp Hachtmannconfig HAVE_MEMBLOCK_PHYS_MAP 5116341e62bSChristoph Jaeger bool 51270210ed9SPhilipp Hachtmann 51367a929e0SChristoph Hellwigconfig HAVE_FAST_GUP 514050a9adcSChristoph Hellwig depends on MMU 5156341e62bSChristoph Jaeger bool 5162667f50eSSteve Capper 51752219aeaSDavid Hildenbrand# Don't discard allocated memory used to track "memory" and "reserved" memblocks 51852219aeaSDavid Hildenbrand# after early boot, so it can still be used to test for validity of memory. 51952219aeaSDavid Hildenbrand# Also, memblocks are updated with memory hot(un)plug. 520350e88baSMike Rapoportconfig ARCH_KEEP_MEMBLOCK 5216341e62bSChristoph Jaeger bool 522c378ddd5STejun Heo 5231e5d8e1eSDan Williams# Keep arch NUMA mapping infrastructure post-init. 5241e5d8e1eSDan Williamsconfig NUMA_KEEP_MEMINFO 5251e5d8e1eSDan Williams bool 5261e5d8e1eSDan Williams 527ee6f509cSMinchan Kimconfig MEMORY_ISOLATION 5286341e62bSChristoph Jaeger bool 529ee6f509cSMinchan Kim 530a9e7b8d4SDavid Hildenbrand# IORESOURCE_SYSTEM_RAM regions in the kernel resource tree that are marked 531a9e7b8d4SDavid Hildenbrand# IORESOURCE_EXCLUSIVE cannot be mapped to user space, for example, via 532a9e7b8d4SDavid Hildenbrand# /dev/mem. 533a9e7b8d4SDavid Hildenbrandconfig EXCLUSIVE_SYSTEM_RAM 534a9e7b8d4SDavid Hildenbrand def_bool y 535a9e7b8d4SDavid Hildenbrand depends on !DEVMEM || STRICT_DEVMEM 536a9e7b8d4SDavid Hildenbrand 53746723bfaSYasuaki Ishimatsu# 53846723bfaSYasuaki Ishimatsu# Only be set on architectures that have completely implemented memory hotplug 53946723bfaSYasuaki Ishimatsu# feature. If you are not sure, don't touch it. 54046723bfaSYasuaki Ishimatsu# 54146723bfaSYasuaki Ishimatsuconfig HAVE_BOOTMEM_INFO_NODE 54246723bfaSYasuaki Ishimatsu def_bool n 54346723bfaSYasuaki Ishimatsu 54491024b3cSAnshuman Khandualconfig ARCH_ENABLE_MEMORY_HOTPLUG 54591024b3cSAnshuman Khandual bool 54691024b3cSAnshuman Khandual 547519bcb79SJohannes Weinerconfig ARCH_ENABLE_MEMORY_HOTREMOVE 548519bcb79SJohannes Weiner bool 549519bcb79SJohannes Weiner 5503947be19SDave Hansen# eventually, we can have this option just 'select SPARSEMEM' 551519bcb79SJohannes Weinermenuconfig MEMORY_HOTPLUG 552519bcb79SJohannes Weiner bool "Memory hotplug" 553b30c5927SDavid Hildenbrand select MEMORY_ISOLATION 55471b6f2ddSDavid Hildenbrand depends on SPARSEMEM 55540b31360SStephen Rothwell depends on ARCH_ENABLE_MEMORY_HOTPLUG 5567ec58a2bSDavid Hildenbrand depends on 64BIT 5571e5d8e1eSDan Williams select NUMA_KEEP_MEMINFO if NUMA 5583947be19SDave Hansen 559519bcb79SJohannes Weinerif MEMORY_HOTPLUG 560519bcb79SJohannes Weiner 5618604d9e5SVitaly Kuznetsovconfig MEMORY_HOTPLUG_DEFAULT_ONLINE 5628604d9e5SVitaly Kuznetsov bool "Online the newly added memory blocks by default" 5638604d9e5SVitaly Kuznetsov depends on MEMORY_HOTPLUG 5648604d9e5SVitaly Kuznetsov help 5658604d9e5SVitaly Kuznetsov This option sets the default policy setting for memory hotplug 5668604d9e5SVitaly Kuznetsov onlining policy (/sys/devices/system/memory/auto_online_blocks) which 5678604d9e5SVitaly Kuznetsov determines what happens to newly added memory regions. Policy setting 5688604d9e5SVitaly Kuznetsov can always be changed at runtime. 569cb1aaebeSMauro Carvalho Chehab See Documentation/admin-guide/mm/memory-hotplug.rst for more information. 5708604d9e5SVitaly Kuznetsov 5718604d9e5SVitaly Kuznetsov Say Y here if you want all hot-plugged memory blocks to appear in 5728604d9e5SVitaly Kuznetsov 'online' state by default. 5738604d9e5SVitaly Kuznetsov Say N here if you want the default policy to keep all hot-plugged 5748604d9e5SVitaly Kuznetsov memory blocks in 'offline' state. 5758604d9e5SVitaly Kuznetsov 5760c0e6195SKAMEZAWA Hiroyukiconfig MEMORY_HOTREMOVE 5770c0e6195SKAMEZAWA Hiroyuki bool "Allow for memory hot remove" 578f7e3334aSNathan Fontenot select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64) 5790c0e6195SKAMEZAWA Hiroyuki depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE 5800c0e6195SKAMEZAWA Hiroyuki depends on MIGRATION 5810c0e6195SKAMEZAWA Hiroyuki 582a08a2ae3SOscar Salvadorconfig MHP_MEMMAP_ON_MEMORY 583a08a2ae3SOscar Salvador def_bool y 584a08a2ae3SOscar Salvador depends on MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP 585a08a2ae3SOscar Salvador depends on ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE 586a08a2ae3SOscar Salvador 587519bcb79SJohannes Weinerendif # MEMORY_HOTPLUG 588519bcb79SJohannes Weiner 5894c21e2f2SHugh Dickins# Heavily threaded applications may benefit from splitting the mm-wide 5904c21e2f2SHugh Dickins# page_table_lock, so that faults on different parts of the user address 5914c21e2f2SHugh Dickins# space can be handled with less contention: split it at this NR_CPUS. 5924c21e2f2SHugh Dickins# Default to 4 for wider testing, though 8 might be more appropriate. 5934c21e2f2SHugh Dickins# ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock. 5947b6ac9dfSHugh Dickins# PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes. 59560bccaa6SWill Deacon# SPARC32 allocates multiple pte tables within a single page, and therefore 59660bccaa6SWill Deacon# a per-page lock leads to problems when multiple tables need to be locked 59760bccaa6SWill Deacon# at the same time (e.g. copy_page_range()). 598a70caa8bSHugh Dickins# DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page. 5994c21e2f2SHugh Dickins# 6004c21e2f2SHugh Dickinsconfig SPLIT_PTLOCK_CPUS 6014c21e2f2SHugh Dickins int 6029164550eSKirill A. Shutemov default "999999" if !MMU 603a70caa8bSHugh Dickins default "999999" if ARM && !CPU_CACHE_VIPT 604a70caa8bSHugh Dickins default "999999" if PARISC && !PA20 60560bccaa6SWill Deacon default "999999" if SPARC32 6064c21e2f2SHugh Dickins default "4" 6077cbe34cfSChristoph Lameter 608e009bb30SKirill A. Shutemovconfig ARCH_ENABLE_SPLIT_PMD_PTLOCK 6096341e62bSChristoph Jaeger bool 610e009bb30SKirill A. Shutemov 6117cbe34cfSChristoph Lameter# 61209316c09SKonstantin Khlebnikov# support for memory balloon 61309316c09SKonstantin Khlebnikovconfig MEMORY_BALLOON 6146341e62bSChristoph Jaeger bool 61509316c09SKonstantin Khlebnikov 61609316c09SKonstantin Khlebnikov# 61718468d93SRafael Aquini# support for memory balloon compaction 61818468d93SRafael Aquiniconfig BALLOON_COMPACTION 61918468d93SRafael Aquini bool "Allow for balloon memory compaction/migration" 62018468d93SRafael Aquini def_bool y 62109316c09SKonstantin Khlebnikov depends on COMPACTION && MEMORY_BALLOON 62218468d93SRafael Aquini help 62318468d93SRafael Aquini Memory fragmentation introduced by ballooning might reduce 62418468d93SRafael Aquini significantly the number of 2MB contiguous memory blocks that can be 62518468d93SRafael Aquini used within a guest, thus imposing performance penalties associated 62618468d93SRafael Aquini with the reduced number of transparent huge pages that could be used 62718468d93SRafael Aquini by the guest workload. Allowing the compaction & migration for memory 62818468d93SRafael Aquini pages enlisted as being part of memory balloon devices avoids the 62918468d93SRafael Aquini scenario aforementioned and helps improving memory defragmentation. 63018468d93SRafael Aquini 63118468d93SRafael Aquini# 632e9e96b39SMel Gorman# support for memory compaction 633e9e96b39SMel Gormanconfig COMPACTION 634e9e96b39SMel Gorman bool "Allow for memory compaction" 63505106e6aSRik van Riel def_bool y 636e9e96b39SMel Gorman select MIGRATION 63733a93877SAndrea Arcangeli depends on MMU 638e9e96b39SMel Gorman help 639b32eaf71SMichal Hocko Compaction is the only memory management component to form 640b32eaf71SMichal Hocko high order (larger physically contiguous) memory blocks 641b32eaf71SMichal Hocko reliably. The page allocator relies on compaction heavily and 642b32eaf71SMichal Hocko the lack of the feature can lead to unexpected OOM killer 643b32eaf71SMichal Hocko invocations for high order memory requests. You shouldn't 644b32eaf71SMichal Hocko disable this option unless there really is a strong reason for 645b32eaf71SMichal Hocko it and then we would be really interested to hear about that at 646b32eaf71SMichal Hocko linux-mm@kvack.org. 647e9e96b39SMel Gorman 648c7e0b3d0SThomas Gleixnerconfig COMPACT_UNEVICTABLE_DEFAULT 649c7e0b3d0SThomas Gleixner int 650c7e0b3d0SThomas Gleixner depends on COMPACTION 651c7e0b3d0SThomas Gleixner default 0 if PREEMPT_RT 652c7e0b3d0SThomas Gleixner default 1 653c7e0b3d0SThomas Gleixner 654e9e96b39SMel Gorman# 65536e66c55SAlexander Duyck# support for free page reporting 65636e66c55SAlexander Duyckconfig PAGE_REPORTING 65736e66c55SAlexander Duyck bool "Free page reporting" 65836e66c55SAlexander Duyck def_bool n 65936e66c55SAlexander Duyck help 66036e66c55SAlexander Duyck Free page reporting allows for the incremental acquisition of 66136e66c55SAlexander Duyck free pages from the buddy allocator for the purpose of reporting 66236e66c55SAlexander Duyck those pages to another entity, such as a hypervisor, so that the 66336e66c55SAlexander Duyck memory can be freed within the host for other uses. 66436e66c55SAlexander Duyck 66536e66c55SAlexander Duyck# 6667cbe34cfSChristoph Lameter# support for page migration 6677cbe34cfSChristoph Lameter# 6687cbe34cfSChristoph Lameterconfig MIGRATION 669b20a3503SChristoph Lameter bool "Page migration" 6706c5240aeSChristoph Lameter def_bool y 671de32a817SChen Gang depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU 672b20a3503SChristoph Lameter help 673b20a3503SChristoph Lameter Allows the migration of the physical location of pages of processes 674e9e96b39SMel Gorman while the virtual addresses are not changed. This is useful in 675e9e96b39SMel Gorman two situations. The first is on NUMA systems to put pages nearer 676e9e96b39SMel Gorman to the processors accessing. The second is when allocating huge 677e9e96b39SMel Gorman pages as migration can relocate pages to satisfy a huge page 678e9e96b39SMel Gorman allocation instead of reclaiming. 6796550e07fSGreg Kroah-Hartman 68076cbbeadSChristoph Hellwigconfig DEVICE_MIGRATION 681d90a25f8SChristoph Hellwig def_bool MIGRATION && ZONE_DEVICE 68276cbbeadSChristoph Hellwig 683c177c81eSNaoya Horiguchiconfig ARCH_ENABLE_HUGEPAGE_MIGRATION 6846341e62bSChristoph Jaeger bool 685c177c81eSNaoya Horiguchi 6869c670ea3SNaoya Horiguchiconfig ARCH_ENABLE_THP_MIGRATION 6879c670ea3SNaoya Horiguchi bool 6889c670ea3SNaoya Horiguchi 6894bfb68a0SAnshuman Khandualconfig HUGETLB_PAGE_SIZE_VARIABLE 6904bfb68a0SAnshuman Khandual def_bool n 6914bfb68a0SAnshuman Khandual help 6924bfb68a0SAnshuman Khandual Allows the pageblock_order value to be dynamic instead of just standard 6934bfb68a0SAnshuman Khandual HUGETLB_PAGE_ORDER when there are multiple HugeTLB page sizes available 6944bfb68a0SAnshuman Khandual on a platform. 6954bfb68a0SAnshuman Khandual 69623baf831SKirill A. Shutemov Note that the pageblock_order cannot exceed MAX_ORDER and will be 69723baf831SKirill A. Shutemov clamped down to MAX_ORDER. 698b3d40a2bSDavid Hildenbrand 6998df995f6SAlexandre Ghiticonfig CONTIG_ALLOC 7008df995f6SAlexandre Ghiti def_bool (MEMORY_ISOLATION && COMPACTION) || CMA 7018df995f6SAlexandre Ghiti 702600715dcSJeremy Fitzhardingeconfig PHYS_ADDR_T_64BIT 703d4a451d5SChristoph Hellwig def_bool 64BIT 704600715dcSJeremy Fitzhardinge 7052a7326b5SChristoph Lameterconfig BOUNCE 7069ca24e2eSVinayak Menon bool "Enable bounce buffers" 7079ca24e2eSVinayak Menon default y 708ce288e05SChristoph Hellwig depends on BLOCK && MMU && HIGHMEM 7099ca24e2eSVinayak Menon help 710ce288e05SChristoph Hellwig Enable bounce buffers for devices that cannot access the full range of 711ce288e05SChristoph Hellwig memory available to the CPU. Enabled by default when HIGHMEM is 712ce288e05SChristoph Hellwig selected, but you may say n to override this. 7132a7326b5SChristoph Lameter 714cddb8a5cSAndrea Arcangeliconfig MMU_NOTIFIER 715cddb8a5cSAndrea Arcangeli bool 71699cb252fSJason Gunthorpe select INTERVAL_TREE 717fc4d5c29SDavid Howells 718f8af4da3SHugh Dickinsconfig KSM 719f8af4da3SHugh Dickins bool "Enable KSM for page merging" 720f8af4da3SHugh Dickins depends on MMU 72159e1a2f4STimofey Titovets select XXHASH 722f8af4da3SHugh Dickins help 723f8af4da3SHugh Dickins Enable Kernel Samepage Merging: KSM periodically scans those areas 724f8af4da3SHugh Dickins of an application's address space that an app has advised may be 725f8af4da3SHugh Dickins mergeable. When it finds pages of identical content, it replaces 726d0f209f6SHugh Dickins the many instances by a single page with that content, so 727f8af4da3SHugh Dickins saving memory until one or another app needs to modify the content. 728f8af4da3SHugh Dickins Recommended for use with KVM, or with other duplicative applications. 729ee65728eSMike Rapoport See Documentation/mm/ksm.rst for more information: KSM is inactive 730c73602adSHugh Dickins until a program has madvised that an area is MADV_MERGEABLE, and 731c73602adSHugh Dickins root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set). 732f8af4da3SHugh Dickins 733e0a94c2aSChristoph Lameterconfig DEFAULT_MMAP_MIN_ADDR 734e0a94c2aSChristoph Lameter int "Low address space to protect from user allocation" 7356e141546SDavid Howells depends on MMU 736e0a94c2aSChristoph Lameter default 4096 737e0a94c2aSChristoph Lameter help 738e0a94c2aSChristoph Lameter This is the portion of low virtual memory which should be protected 739e0a94c2aSChristoph Lameter from userspace allocation. Keeping a user from writing to low pages 740e0a94c2aSChristoph Lameter can help reduce the impact of kernel NULL pointer bugs. 741e0a94c2aSChristoph Lameter 742e0a94c2aSChristoph Lameter For most ia64, ppc64 and x86 users with lots of address space 743e0a94c2aSChristoph Lameter a value of 65536 is reasonable and should cause no problems. 744e0a94c2aSChristoph Lameter On arm and other archs it should not be higher than 32768. 745788084abSEric Paris Programs which use vm86 functionality or have some need to map 746788084abSEric Paris this low address space will need CAP_SYS_RAWIO or disable this 747788084abSEric Paris protection by setting the value to 0. 748e0a94c2aSChristoph Lameter 749e0a94c2aSChristoph Lameter This value can be changed after boot using the 750e0a94c2aSChristoph Lameter /proc/sys/vm/mmap_min_addr tunable. 751e0a94c2aSChristoph Lameter 752d949f36fSLinus Torvaldsconfig ARCH_SUPPORTS_MEMORY_FAILURE 753d949f36fSLinus Torvalds bool 754e0a94c2aSChristoph Lameter 7556a46079cSAndi Kleenconfig MEMORY_FAILURE 7566a46079cSAndi Kleen depends on MMU 757d949f36fSLinus Torvalds depends on ARCH_SUPPORTS_MEMORY_FAILURE 7586a46079cSAndi Kleen bool "Enable recovery from hardware memory errors" 759ee6f509cSMinchan Kim select MEMORY_ISOLATION 76097f0b134SXie XiuQi select RAS 7616a46079cSAndi Kleen help 7626a46079cSAndi Kleen Enables code to recover from some memory failures on systems 7636a46079cSAndi Kleen with MCA recovery. This allows a system to continue running 7646a46079cSAndi Kleen even when some of its memory has uncorrected errors. This requires 7656a46079cSAndi Kleen special hardware support and typically ECC memory. 7666a46079cSAndi Kleen 767cae681fcSAndi Kleenconfig HWPOISON_INJECT 768413f9efbSAndi Kleen tristate "HWPoison pages injector" 76927df5068SAndi Kleen depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS 770478c5ffcSWu Fengguang select PROC_PAGE_MONITOR 771cae681fcSAndi Kleen 772fc4d5c29SDavid Howellsconfig NOMMU_INITIAL_TRIM_EXCESS 773fc4d5c29SDavid Howells int "Turn on mmap() excess space trimming before booting" 774fc4d5c29SDavid Howells depends on !MMU 775fc4d5c29SDavid Howells default 1 776fc4d5c29SDavid Howells help 777fc4d5c29SDavid Howells The NOMMU mmap() frequently needs to allocate large contiguous chunks 778fc4d5c29SDavid Howells of memory on which to store mappings, but it can only ask the system 779fc4d5c29SDavid Howells allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently 780fc4d5c29SDavid Howells more than it requires. To deal with this, mmap() is able to trim off 781fc4d5c29SDavid Howells the excess and return it to the allocator. 782fc4d5c29SDavid Howells 783fc4d5c29SDavid Howells If trimming is enabled, the excess is trimmed off and returned to the 784fc4d5c29SDavid Howells system allocator, which can cause extra fragmentation, particularly 785fc4d5c29SDavid Howells if there are a lot of transient processes. 786fc4d5c29SDavid Howells 787fc4d5c29SDavid Howells If trimming is disabled, the excess is kept, but not used, which for 788fc4d5c29SDavid Howells long-term mappings means that the space is wasted. 789fc4d5c29SDavid Howells 790fc4d5c29SDavid Howells Trimming can be dynamically controlled through a sysctl option 791fc4d5c29SDavid Howells (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of 792fc4d5c29SDavid Howells excess pages there must be before trimming should occur, or zero if 793fc4d5c29SDavid Howells no trimming is to occur. 794fc4d5c29SDavid Howells 795fc4d5c29SDavid Howells This option specifies the initial value of this option. The default 796fc4d5c29SDavid Howells of 1 says that all excess pages should be trimmed. 797fc4d5c29SDavid Howells 798dd19d293SStephen Kitt See Documentation/admin-guide/mm/nommu-mmap.rst for more information. 799bbddff05STejun Heo 800519bcb79SJohannes Weinerconfig ARCH_WANT_GENERAL_HUGETLB 801519bcb79SJohannes Weiner bool 802519bcb79SJohannes Weiner 803519bcb79SJohannes Weinerconfig ARCH_WANTS_THP_SWAP 804519bcb79SJohannes Weiner def_bool n 805519bcb79SJohannes Weiner 806519bcb79SJohannes Weinermenuconfig TRANSPARENT_HUGEPAGE 80713ece886SAndrea Arcangeli bool "Transparent Hugepage Support" 808554b0f3cSSebastian Andrzej Siewior depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT 8095d689240SAndrea Arcangeli select COMPACTION 8103a08cd52SMatthew Wilcox select XARRAY_MULTI 8114c76d9d1SAndrea Arcangeli help 8124c76d9d1SAndrea Arcangeli Transparent Hugepages allows the kernel to use huge pages and 8134c76d9d1SAndrea Arcangeli huge tlb transparently to the applications whenever possible. 8144c76d9d1SAndrea Arcangeli This feature can improve computing performance to certain 8154c76d9d1SAndrea Arcangeli applications by speeding up page faults during memory 8164c76d9d1SAndrea Arcangeli allocation, by reducing the number of tlb misses and by speeding 8174c76d9d1SAndrea Arcangeli up the pagetable walking. 8184c76d9d1SAndrea Arcangeli 8194c76d9d1SAndrea Arcangeli If memory constrained on embedded, you may want to say N. 8204c76d9d1SAndrea Arcangeli 821519bcb79SJohannes Weinerif TRANSPARENT_HUGEPAGE 822519bcb79SJohannes Weiner 82313ece886SAndrea Arcangelichoice 82413ece886SAndrea Arcangeli prompt "Transparent Hugepage Support sysfs defaults" 82513ece886SAndrea Arcangeli depends on TRANSPARENT_HUGEPAGE 82613ece886SAndrea Arcangeli default TRANSPARENT_HUGEPAGE_ALWAYS 82713ece886SAndrea Arcangeli help 82813ece886SAndrea Arcangeli Selects the sysfs defaults for Transparent Hugepage Support. 82913ece886SAndrea Arcangeli 83013ece886SAndrea Arcangeli config TRANSPARENT_HUGEPAGE_ALWAYS 83113ece886SAndrea Arcangeli bool "always" 83213ece886SAndrea Arcangeli help 83313ece886SAndrea Arcangeli Enabling Transparent Hugepage always, can increase the 83413ece886SAndrea Arcangeli memory footprint of applications without a guaranteed 83513ece886SAndrea Arcangeli benefit but it will work automatically for all applications. 83613ece886SAndrea Arcangeli 83713ece886SAndrea Arcangeli config TRANSPARENT_HUGEPAGE_MADVISE 83813ece886SAndrea Arcangeli bool "madvise" 83913ece886SAndrea Arcangeli help 84013ece886SAndrea Arcangeli Enabling Transparent Hugepage madvise, will only provide a 84113ece886SAndrea Arcangeli performance improvement benefit to the applications using 84213ece886SAndrea Arcangeli madvise(MADV_HUGEPAGE) but it won't risk to increase the 84313ece886SAndrea Arcangeli memory footprint of applications without a guaranteed 84413ece886SAndrea Arcangeli benefit. 84513ece886SAndrea Arcangeliendchoice 84613ece886SAndrea Arcangeli 84738d8b4e6SHuang Yingconfig THP_SWAP 84838d8b4e6SHuang Ying def_bool y 849dad6a5ebSHugh Dickins depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP && 64BIT 85038d8b4e6SHuang Ying help 85138d8b4e6SHuang Ying Swap transparent huge pages in one piece, without splitting. 85214fef284SHuang Ying XXX: For now, swap cluster backing transparent huge page 85314fef284SHuang Ying will be split after swapout. 85438d8b4e6SHuang Ying 85538d8b4e6SHuang Ying For selection by architectures with reasonable THP sizes. 85638d8b4e6SHuang Ying 857519bcb79SJohannes Weinerconfig READ_ONLY_THP_FOR_FS 858519bcb79SJohannes Weiner bool "Read-only THP for filesystems (EXPERIMENTAL)" 859519bcb79SJohannes Weiner depends on TRANSPARENT_HUGEPAGE && SHMEM 860519bcb79SJohannes Weiner 861519bcb79SJohannes Weiner help 862519bcb79SJohannes Weiner Allow khugepaged to put read-only file-backed pages in THP. 863519bcb79SJohannes Weiner 864519bcb79SJohannes Weiner This is marked experimental because it is a new feature. Write 865519bcb79SJohannes Weiner support of file THPs will be developed in the next few release 866519bcb79SJohannes Weiner cycles. 867519bcb79SJohannes Weiner 868519bcb79SJohannes Weinerendif # TRANSPARENT_HUGEPAGE 869519bcb79SJohannes Weiner 870e496cf3dSKirill A. Shutemov# 871bbddff05STejun Heo# UP and nommu archs use km based percpu allocator 872bbddff05STejun Heo# 873bbddff05STejun Heoconfig NEED_PER_CPU_KM 8743583521aSVladimir Murzin depends on !SMP || !MMU 875bbddff05STejun Heo bool 876bbddff05STejun Heo default y 877077b1f83SDan Magenheimer 8787ecd19cfSKefeng Wangconfig NEED_PER_CPU_EMBED_FIRST_CHUNK 8797ecd19cfSKefeng Wang bool 8807ecd19cfSKefeng Wang 8817ecd19cfSKefeng Wangconfig NEED_PER_CPU_PAGE_FIRST_CHUNK 8827ecd19cfSKefeng Wang bool 8837ecd19cfSKefeng Wang 8847ecd19cfSKefeng Wangconfig USE_PERCPU_NUMA_NODE_ID 8857ecd19cfSKefeng Wang bool 8867ecd19cfSKefeng Wang 8877ecd19cfSKefeng Wangconfig HAVE_SETUP_PER_CPU_AREA 8887ecd19cfSKefeng Wang bool 8897ecd19cfSKefeng Wang 89027c6aec2SDan Magenheimerconfig FRONTSWAP 8916e61dde8SChristoph Hellwig bool 892f825c736SAneesh Kumar K.V 893f825c736SAneesh Kumar K.Vconfig CMA 894f825c736SAneesh Kumar K.V bool "Contiguous Memory Allocator" 895aca52c39SMike Rapoport depends on MMU 896f825c736SAneesh Kumar K.V select MIGRATION 897f825c736SAneesh Kumar K.V select MEMORY_ISOLATION 898f825c736SAneesh Kumar K.V help 899f825c736SAneesh Kumar K.V This enables the Contiguous Memory Allocator which allows other 900f825c736SAneesh Kumar K.V subsystems to allocate big physically-contiguous blocks of memory. 901f825c736SAneesh Kumar K.V CMA reserves a region of memory and allows only movable pages to 902f825c736SAneesh Kumar K.V be allocated from it. This way, the kernel can use the memory for 903f825c736SAneesh Kumar K.V pagecache and when a subsystem requests for contiguous area, the 904f825c736SAneesh Kumar K.V allocated pages are migrated away to serve the contiguous request. 905f825c736SAneesh Kumar K.V 906f825c736SAneesh Kumar K.V If unsure, say "n". 907f825c736SAneesh Kumar K.V 908f825c736SAneesh Kumar K.Vconfig CMA_DEBUG 909f825c736SAneesh Kumar K.V bool "CMA debug messages (DEVELOPMENT)" 910f825c736SAneesh Kumar K.V depends on DEBUG_KERNEL && CMA 911f825c736SAneesh Kumar K.V help 912f825c736SAneesh Kumar K.V Turns on debug messages in CMA. This produces KERN_DEBUG 913f825c736SAneesh Kumar K.V messages for every CMA call as well as various messages while 914f825c736SAneesh Kumar K.V processing calls such as dma_alloc_from_contiguous(). 915f825c736SAneesh Kumar K.V This option does not affect warning and error messages. 916bf550fc9SAlexander Graf 91728b24c1fSSasha Levinconfig CMA_DEBUGFS 91828b24c1fSSasha Levin bool "CMA debugfs interface" 91928b24c1fSSasha Levin depends on CMA && DEBUG_FS 92028b24c1fSSasha Levin help 92128b24c1fSSasha Levin Turns on the DebugFS interface for CMA. 92228b24c1fSSasha Levin 92343ca106fSMinchan Kimconfig CMA_SYSFS 92443ca106fSMinchan Kim bool "CMA information through sysfs interface" 92543ca106fSMinchan Kim depends on CMA && SYSFS 92643ca106fSMinchan Kim help 92743ca106fSMinchan Kim This option exposes some sysfs attributes to get information 92843ca106fSMinchan Kim from CMA. 92943ca106fSMinchan Kim 930a254129eSJoonsoo Kimconfig CMA_AREAS 931a254129eSJoonsoo Kim int "Maximum count of the CMA areas" 932a254129eSJoonsoo Kim depends on CMA 933b7176c26SBarry Song default 19 if NUMA 934a254129eSJoonsoo Kim default 7 935a254129eSJoonsoo Kim help 936a254129eSJoonsoo Kim CMA allows to create CMA areas for particular purpose, mainly, 937a254129eSJoonsoo Kim used as device private area. This parameter sets the maximum 938a254129eSJoonsoo Kim number of CMA area in the system. 939a254129eSJoonsoo Kim 940b7176c26SBarry Song If unsure, leave the default value "7" in UMA and "19" in NUMA. 941a254129eSJoonsoo Kim 942af8d417aSDan Streetmanconfig MEM_SOFT_DIRTY 943af8d417aSDan Streetman bool "Track memory changes" 944af8d417aSDan Streetman depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS 945af8d417aSDan Streetman select PROC_PAGE_MONITOR 9464e2e2770SSeth Jennings help 947af8d417aSDan Streetman This option enables memory changes tracking by introducing a 948af8d417aSDan Streetman soft-dirty bit on pte-s. This bit it set when someone writes 949af8d417aSDan Streetman into a page just as regular dirty bit, but unlike the latter 950af8d417aSDan Streetman it can be cleared by hands. 951af8d417aSDan Streetman 9521ad1335dSMike Rapoport See Documentation/admin-guide/mm/soft-dirty.rst for more details. 9534e2e2770SSeth Jennings 9549e5c33d7SMark Salterconfig GENERIC_EARLY_IOREMAP 9559e5c33d7SMark Salter bool 956042d27acSHelge Deller 95722ee3ea5SHelge Dellerconfig STACK_MAX_DEFAULT_SIZE_MB 95822ee3ea5SHelge Deller int "Default maximum user stack size for 32-bit processes (MB)" 95922ee3ea5SHelge Deller default 100 960042d27acSHelge Deller range 8 2048 961042d27acSHelge Deller depends on STACK_GROWSUP && (!64BIT || COMPAT) 962042d27acSHelge Deller help 963042d27acSHelge Deller This is the maximum stack size in Megabytes in the VM layout of 32-bit 964042d27acSHelge Deller user processes when the stack grows upwards (currently only on parisc 96522ee3ea5SHelge Deller arch) when the RLIMIT_STACK hard limit is unlimited. 966042d27acSHelge Deller 96722ee3ea5SHelge Deller A sane initial value is 100 MB. 9683a80a7faSMel Gorman 9693a80a7faSMel Gormanconfig DEFERRED_STRUCT_PAGE_INIT 9701ce22103SVlastimil Babka bool "Defer initialisation of struct pages to kthreads" 971d39f8fb4SMike Rapoport depends on SPARSEMEM 972ab1e8d89SPavel Tatashin depends on !NEED_PER_CPU_KM 973889c695dSPasha Tatashin depends on 64BIT 974e4443149SDaniel Jordan select PADATA 9753a80a7faSMel Gorman help 9763a80a7faSMel Gorman Ordinarily all struct pages are initialised during early boot in a 9773a80a7faSMel Gorman single thread. On very large machines this can take a considerable 9783a80a7faSMel Gorman amount of time. If this option is set, large machines will bring up 979e4443149SDaniel Jordan a subset of memmap at boot and then initialise the rest in parallel. 980e4443149SDaniel Jordan This has a potential performance impact on tasks running early in the 9811ce22103SVlastimil Babka lifetime of the system until these kthreads finish the 9821ce22103SVlastimil Babka initialisation. 983033fbae9SDan Williams 9841c676e0dSSeongJae Parkconfig PAGE_IDLE_FLAG 9851c676e0dSSeongJae Park bool 9861c676e0dSSeongJae Park select PAGE_EXTENSION if !64BIT 9871c676e0dSSeongJae Park help 9881c676e0dSSeongJae Park This adds PG_idle and PG_young flags to 'struct page'. PTE Accessed 9891c676e0dSSeongJae Park bit writers can set the state of the bit in the flags so that PTE 9901c676e0dSSeongJae Park Accessed bit readers may avoid disturbance. 9911c676e0dSSeongJae Park 99233c3fc71SVladimir Davydovconfig IDLE_PAGE_TRACKING 99333c3fc71SVladimir Davydov bool "Enable idle page tracking" 99433c3fc71SVladimir Davydov depends on SYSFS && MMU 9951c676e0dSSeongJae Park select PAGE_IDLE_FLAG 99633c3fc71SVladimir Davydov help 99733c3fc71SVladimir Davydov This feature allows to estimate the amount of user pages that have 99833c3fc71SVladimir Davydov not been touched during a given period of time. This information can 99933c3fc71SVladimir Davydov be useful to tune memory cgroup limits and/or for job placement 100033c3fc71SVladimir Davydov within a compute cluster. 100133c3fc71SVladimir Davydov 10021ad1335dSMike Rapoport See Documentation/admin-guide/mm/idle_page_tracking.rst for 10031ad1335dSMike Rapoport more details. 100433c3fc71SVladimir Davydov 1005c2280be8SAnshuman Khandualconfig ARCH_HAS_CACHE_LINE_SIZE 1006c2280be8SAnshuman Khandual bool 1007c2280be8SAnshuman Khandual 10082792d84eSKees Cookconfig ARCH_HAS_CURRENT_STACK_POINTER 10092792d84eSKees Cook bool 10102792d84eSKees Cook help 10112792d84eSKees Cook In support of HARDENED_USERCOPY performing stack variable lifetime 10122792d84eSKees Cook checking, an architecture-agnostic way to find the stack pointer 10132792d84eSKees Cook is needed. Once an architecture defines an unsigned long global 10142792d84eSKees Cook register alias named "current_stack_pointer", this config can be 10152792d84eSKees Cook selected. 10162792d84eSKees Cook 101717596731SRobin Murphyconfig ARCH_HAS_PTE_DEVMAP 101865f7d049SOliver O'Halloran bool 101965f7d049SOliver O'Halloran 102063703f37SKefeng Wangconfig ARCH_HAS_ZONE_DMA_SET 102163703f37SKefeng Wang bool 102263703f37SKefeng Wang 102363703f37SKefeng Wangconfig ZONE_DMA 102463703f37SKefeng Wang bool "Support DMA zone" if ARCH_HAS_ZONE_DMA_SET 102563703f37SKefeng Wang default y if ARM64 || X86 102663703f37SKefeng Wang 102763703f37SKefeng Wangconfig ZONE_DMA32 102863703f37SKefeng Wang bool "Support DMA32 zone" if ARCH_HAS_ZONE_DMA_SET 102963703f37SKefeng Wang depends on !X86_32 103063703f37SKefeng Wang default y if ARM64 103163703f37SKefeng Wang 1032033fbae9SDan Williamsconfig ZONE_DEVICE 10335042db43SJérôme Glisse bool "Device memory (pmem, HMM, etc...) hotplug support" 1034033fbae9SDan Williams depends on MEMORY_HOTPLUG 1035033fbae9SDan Williams depends on MEMORY_HOTREMOVE 103699490f16SDan Williams depends on SPARSEMEM_VMEMMAP 103717596731SRobin Murphy depends on ARCH_HAS_PTE_DEVMAP 10383a08cd52SMatthew Wilcox select XARRAY_MULTI 1039033fbae9SDan Williams 1040033fbae9SDan Williams help 1041033fbae9SDan Williams Device memory hotplug support allows for establishing pmem, 1042033fbae9SDan Williams or other device driver discovered memory regions, in the 1043033fbae9SDan Williams memmap. This allows pfn_to_page() lookups of otherwise 1044033fbae9SDan Williams "device-physical" addresses which is needed for using a DAX 1045033fbae9SDan Williams mapping in an O_DIRECT operation, among other things. 1046033fbae9SDan Williams 1047033fbae9SDan Williams If FS_DAX is enabled, then say Y. 104806a660adSLinus Torvalds 10499c240a7bSChristoph Hellwig# 10509c240a7bSChristoph Hellwig# Helpers to mirror range of the CPU page tables of a process into device page 10519c240a7bSChristoph Hellwig# tables. 10529c240a7bSChristoph Hellwig# 1053c0b12405SJérôme Glisseconfig HMM_MIRROR 10549c240a7bSChristoph Hellwig bool 1055f442c283SChristoph Hellwig depends on MMU 1056c0b12405SJérôme Glisse 105714b80582SDan Williamsconfig GET_FREE_REGION 105814b80582SDan Williams depends on SPARSEMEM 105914b80582SDan Williams bool 106014b80582SDan Williams 10615042db43SJérôme Glisseconfig DEVICE_PRIVATE 10625042db43SJérôme Glisse bool "Unaddressable device memory (GPU memory, ...)" 10637328d9ccSChristoph Hellwig depends on ZONE_DEVICE 106414b80582SDan Williams select GET_FREE_REGION 10655042db43SJérôme Glisse 10665042db43SJérôme Glisse help 10675042db43SJérôme Glisse Allows creation of struct pages to represent unaddressable device 10685042db43SJérôme Glisse memory; i.e., memory that is only accessible from the device (or 10695042db43SJérôme Glisse group of devices). You likely also want to select HMM_MIRROR. 10705042db43SJérôme Glisse 10713e9a9e25SChristoph Hellwigconfig VMAP_PFN 10723e9a9e25SChristoph Hellwig bool 10733e9a9e25SChristoph Hellwig 107463c17fb8SDave Hansenconfig ARCH_USES_HIGH_VMA_FLAGS 107563c17fb8SDave Hansen bool 107666d37570SDave Hansenconfig ARCH_HAS_PKEYS 107766d37570SDave Hansen bool 107830a5b536SDennis Zhou 1079b0284cd2SCatalin Marinasconfig ARCH_USES_PG_ARCH_X 1080b0284cd2SCatalin Marinas bool 1081b0284cd2SCatalin Marinas help 1082b0284cd2SCatalin Marinas Enable the definition of PG_arch_x page flags with x > 1. Only 1083b0284cd2SCatalin Marinas suitable for 64-bit architectures with CONFIG_FLATMEM or 1084b0284cd2SCatalin Marinas CONFIG_SPARSEMEM_VMEMMAP enabled, otherwise there may not be 1085b0284cd2SCatalin Marinas enough room for additional bits in page->flags. 1086b0284cd2SCatalin Marinas 10870710d012SVlastimil Babkaconfig VM_EVENT_COUNTERS 10880710d012SVlastimil Babka default y 10890710d012SVlastimil Babka bool "Enable VM event counters for /proc/vmstat" if EXPERT 10900710d012SVlastimil Babka help 10910710d012SVlastimil Babka VM event counters are needed for event counts to be shown. 10920710d012SVlastimil Babka This option allows the disabling of the VM event counters 10930710d012SVlastimil Babka on EXPERT systems. /proc/vmstat will only show page counts 10940710d012SVlastimil Babka if VM event counters are disabled. 10950710d012SVlastimil Babka 109630a5b536SDennis Zhouconfig PERCPU_STATS 109730a5b536SDennis Zhou bool "Collect percpu memory statistics" 109830a5b536SDennis Zhou help 109930a5b536SDennis Zhou This feature collects and exposes statistics via debugfs. The 110030a5b536SDennis Zhou information includes global and per chunk statistics, which can 110130a5b536SDennis Zhou be used to help understand percpu memory usage. 110264c349f4SKirill A. Shutemov 11039c84f229SJohn Hubbardconfig GUP_TEST 11049c84f229SJohn Hubbard bool "Enable infrastructure for get_user_pages()-related unit tests" 1105d0de8241SBarry Song depends on DEBUG_FS 110664c349f4SKirill A. Shutemov help 11079c84f229SJohn Hubbard Provides /sys/kernel/debug/gup_test, which in turn provides a way 11089c84f229SJohn Hubbard to make ioctl calls that can launch kernel-based unit tests for 11099c84f229SJohn Hubbard the get_user_pages*() and pin_user_pages*() family of API calls. 111064c349f4SKirill A. Shutemov 11119c84f229SJohn Hubbard These tests include benchmark testing of the _fast variants of 11129c84f229SJohn Hubbard get_user_pages*() and pin_user_pages*(), as well as smoke tests of 11139c84f229SJohn Hubbard the non-_fast variants. 11149c84f229SJohn Hubbard 1115f4f9bda4SJohn Hubbard There is also a sub-test that allows running dump_page() on any 1116f4f9bda4SJohn Hubbard of up to eight pages (selected by command line args) within the 1117f4f9bda4SJohn Hubbard range of user-space addresses. These pages are either pinned via 1118f4f9bda4SJohn Hubbard pin_user_pages*(), or pinned via get_user_pages*(), as specified 1119f4f9bda4SJohn Hubbard by other command line arguments. 1120f4f9bda4SJohn Hubbard 1121baa489faSSeongJae Park See tools/testing/selftests/mm/gup_test.c 11223010a5eaSLaurent Dufour 1123d0de8241SBarry Songcomment "GUP_TEST needs to have DEBUG_FS enabled" 1124d0de8241SBarry Song depends on !GUP_TEST && !DEBUG_FS 11253010a5eaSLaurent Dufour 11266ca297d4SPeter Zijlstraconfig GUP_GET_PXX_LOW_HIGH 112739656e83SChristoph Hellwig bool 112839656e83SChristoph Hellwig 1129def85743SKeith Buschconfig DMAPOOL_TEST 1130def85743SKeith Busch tristate "Enable a module to run time tests on dma_pool" 1131def85743SKeith Busch depends on HAS_DMA 1132def85743SKeith Busch help 1133def85743SKeith Busch Provides a test module that will allocate and free many blocks of 1134def85743SKeith Busch various sizes and report how long it takes. This is intended to 1135def85743SKeith Busch provide a consistent way to measure how changes to the 1136def85743SKeith Busch dma_pool_alloc/free routines affect performance. 1137def85743SKeith Busch 11383010a5eaSLaurent Dufourconfig ARCH_HAS_PTE_SPECIAL 11393010a5eaSLaurent Dufour bool 114059e0b520SChristoph Hellwig 1141cbd34da7SChristoph Hellwig# 1142cbd34da7SChristoph Hellwig# Some architectures require a special hugepage directory format that is 1143cbd34da7SChristoph Hellwig# required to support multiple hugepage sizes. For example a4fe3ce76 1144cbd34da7SChristoph Hellwig# "powerpc/mm: Allow more flexible layouts for hugepage pagetables" 1145cbd34da7SChristoph Hellwig# introduced it on powerpc. This allows for a more flexible hugepage 1146cbd34da7SChristoph Hellwig# pagetable layouts. 1147cbd34da7SChristoph Hellwig# 1148cbd34da7SChristoph Hellwigconfig ARCH_HAS_HUGEPD 1149cbd34da7SChristoph Hellwig bool 1150cbd34da7SChristoph Hellwig 1151c5acad84SThomas Hellstromconfig MAPPING_DIRTY_HELPERS 1152c5acad84SThomas Hellstrom bool 1153c5acad84SThomas Hellstrom 1154298fa1adSThomas Gleixnerconfig KMAP_LOCAL 1155298fa1adSThomas Gleixner bool 1156298fa1adSThomas Gleixner 1157825c43f5SArd Biesheuvelconfig KMAP_LOCAL_NON_LINEAR_PTE_ARRAY 1158825c43f5SArd Biesheuvel bool 1159825c43f5SArd Biesheuvel 11601fbaf8fcSChristoph Hellwig# struct io_mapping based helper. Selected by drivers that need them 11611fbaf8fcSChristoph Hellwigconfig IO_MAPPING 11621fbaf8fcSChristoph Hellwig bool 11631507f512SMike Rapoport 11641507f512SMike Rapoportconfig SECRETMEM 116574947724SLukas Bulwahn default y 116674947724SLukas Bulwahn bool "Enable memfd_secret() system call" if EXPERT 116774947724SLukas Bulwahn depends on ARCH_HAS_SET_DIRECT_MAP 116874947724SLukas Bulwahn help 116974947724SLukas Bulwahn Enable the memfd_secret() system call with the ability to create 117074947724SLukas Bulwahn memory areas visible only in the context of the owning process and 117174947724SLukas Bulwahn not mapped to other processes and other kernel page tables. 11721507f512SMike Rapoport 11739a10064fSColin Crossconfig ANON_VMA_NAME 11749a10064fSColin Cross bool "Anonymous VMA name support" 11759a10064fSColin Cross depends on PROC_FS && ADVISE_SYSCALLS && MMU 11769a10064fSColin Cross 11779a10064fSColin Cross help 11789a10064fSColin Cross Allow naming anonymous virtual memory areas. 11799a10064fSColin Cross 11809a10064fSColin Cross This feature allows assigning names to virtual memory areas. Assigned 11819a10064fSColin Cross names can be later retrieved from /proc/pid/maps and /proc/pid/smaps 11829a10064fSColin Cross and help identifying individual anonymous memory areas. 11839a10064fSColin Cross Assigning a name to anonymous virtual memory area might prevent that 11849a10064fSColin Cross area from being merged with adjacent virtual memory areas due to the 11859a10064fSColin Cross difference in their name. 11869a10064fSColin Cross 1187430529b5SPeter Xuconfig USERFAULTFD 1188430529b5SPeter Xu bool "Enable userfaultfd() system call" 1189430529b5SPeter Xu depends on MMU 1190430529b5SPeter Xu help 1191430529b5SPeter Xu Enable the userfaultfd() system call that allows to intercept and 1192430529b5SPeter Xu handle page faults in userland. 1193430529b5SPeter Xu 1194430529b5SPeter Xuconfig HAVE_ARCH_USERFAULTFD_WP 1195430529b5SPeter Xu bool 1196430529b5SPeter Xu help 1197430529b5SPeter Xu Arch has userfaultfd write protection support 1198430529b5SPeter Xu 1199430529b5SPeter Xuconfig HAVE_ARCH_USERFAULTFD_MINOR 1200430529b5SPeter Xu bool 1201430529b5SPeter Xu help 1202430529b5SPeter Xu Arch has userfaultfd minor fault support 1203430529b5SPeter Xu 12041db9dbc2SPeter Xuconfig PTE_MARKER_UFFD_WP 120581e0f15fSPeter Xu bool "Userfaultfd write protection support for shmem/hugetlbfs" 120681e0f15fSPeter Xu default y 120781e0f15fSPeter Xu depends on HAVE_ARCH_USERFAULTFD_WP 12081db9dbc2SPeter Xu 12091db9dbc2SPeter Xu help 12101db9dbc2SPeter Xu Allows to create marker PTEs for userfaultfd write protection 12111db9dbc2SPeter Xu purposes. It is required to enable userfaultfd write protection on 12121db9dbc2SPeter Xu file-backed memory types like shmem and hugetlbfs. 12131db9dbc2SPeter Xu 1214ac35a490SYu Zhao# multi-gen LRU { 1215ec1c86b2SYu Zhaoconfig LRU_GEN 1216ec1c86b2SYu Zhao bool "Multi-Gen LRU" 1217ec1c86b2SYu Zhao depends on MMU 1218ec1c86b2SYu Zhao # make sure folio->flags has enough spare bits 1219ec1c86b2SYu Zhao depends on 64BIT || !SPARSEMEM || SPARSEMEM_VMEMMAP 1220ec1c86b2SYu Zhao help 122107017acbSYu Zhao A high performance LRU implementation to overcommit memory. See 122207017acbSYu Zhao Documentation/admin-guide/mm/multigen_lru.rst for details. 1223ec1c86b2SYu Zhao 1224354ed597SYu Zhaoconfig LRU_GEN_ENABLED 1225354ed597SYu Zhao bool "Enable by default" 1226354ed597SYu Zhao depends on LRU_GEN 1227354ed597SYu Zhao help 1228354ed597SYu Zhao This option enables the multi-gen LRU by default. 1229354ed597SYu Zhao 1230ac35a490SYu Zhaoconfig LRU_GEN_STATS 1231ac35a490SYu Zhao bool "Full stats for debugging" 1232ac35a490SYu Zhao depends on LRU_GEN 1233ac35a490SYu Zhao help 1234ac35a490SYu Zhao Do not enable this option unless you plan to look at historical stats 1235ac35a490SYu Zhao from evicted generations for debugging purpose. 1236ac35a490SYu Zhao 1237ac35a490SYu Zhao This option has a per-memcg and per-node memory overhead. 1238ac35a490SYu Zhao# } 1239ac35a490SYu Zhao 12400b6cc04fSSuren Baghdasaryanconfig ARCH_SUPPORTS_PER_VMA_LOCK 12410b6cc04fSSuren Baghdasaryan def_bool n 12420b6cc04fSSuren Baghdasaryan 12430b6cc04fSSuren Baghdasaryanconfig PER_VMA_LOCK 12440b6cc04fSSuren Baghdasaryan def_bool y 12450b6cc04fSSuren Baghdasaryan depends on ARCH_SUPPORTS_PER_VMA_LOCK && MMU && SMP 12460b6cc04fSSuren Baghdasaryan help 12470b6cc04fSSuren Baghdasaryan Allow per-vma locking during page fault handling. 12480b6cc04fSSuren Baghdasaryan 12490b6cc04fSSuren Baghdasaryan This feature allows locking each virtual memory area separately when 12500b6cc04fSSuren Baghdasaryan handling page faults instead of taking mmap_lock. 12510b6cc04fSSuren Baghdasaryan 1252c2508ec5SLinus Torvaldsconfig LOCK_MM_AND_FIND_VMA 1253c2508ec5SLinus Torvalds bool 1254c2508ec5SLinus Torvalds depends on !STACK_GROWSUP 1255c2508ec5SLinus Torvalds 12562224d848SSeongJae Parksource "mm/damon/Kconfig" 12572224d848SSeongJae Park 125859e0b520SChristoph Hellwigendmenu 1259