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 28b3fbd58fSJohannes Weiner select CRYPTO 29519bcb79SJohannes Weiner select ZPOOL 30519bcb79SJohannes Weiner help 31519bcb79SJohannes Weiner A lightweight compressed cache for swap pages. It takes 32519bcb79SJohannes Weiner pages that are in the process of being swapped out and attempts to 33519bcb79SJohannes Weiner compress them into a dynamically allocated RAM-based memory pool. 34519bcb79SJohannes Weiner This can result in a significant I/O reduction on swap device and, 351a44131dSSophia Gabriella in the case where decompressing from RAM is faster than swap device 36519bcb79SJohannes Weiner reads, can also improve workload performance. 37519bcb79SJohannes Weiner 38b3fbd58fSJohannes Weinerconfig ZSWAP_DEFAULT_ON 39b3fbd58fSJohannes Weiner bool "Enable the compressed cache for swap pages by default" 40b3fbd58fSJohannes Weiner depends on ZSWAP 41b3fbd58fSJohannes Weiner help 42b3fbd58fSJohannes Weiner If selected, the compressed cache for swap pages will be enabled 43b3fbd58fSJohannes Weiner at boot, otherwise it will be disabled. 44b3fbd58fSJohannes Weiner 45b3fbd58fSJohannes Weiner The selection made here can be overridden by using the kernel 46b3fbd58fSJohannes Weiner command line 'zswap.enabled=' option. 47b3fbd58fSJohannes Weiner 48b5ba474fSNhat Phamconfig ZSWAP_SHRINKER_DEFAULT_ON 49b5ba474fSNhat Pham bool "Shrink the zswap pool on memory pressure" 50b5ba474fSNhat Pham depends on ZSWAP 51b5ba474fSNhat Pham default n 52b5ba474fSNhat Pham help 53b5ba474fSNhat Pham If selected, the zswap shrinker will be enabled, and the pages 54b5ba474fSNhat Pham stored in the zswap pool will become available for reclaim (i.e 55b5ba474fSNhat Pham written back to the backing swap device) on memory pressure. 56b5ba474fSNhat Pham 57b5ba474fSNhat Pham This means that zswap writeback could happen even if the pool is 58b5ba474fSNhat Pham not yet full, or the cgroup zswap limit has not been reached, 59b5ba474fSNhat Pham reducing the chance that cold pages will reside in the zswap pool 60b5ba474fSNhat Pham and consume memory indefinitely. 61b5ba474fSNhat Pham 62519bcb79SJohannes Weinerchoice 63b3fbd58fSJohannes Weiner prompt "Default compressor" 64519bcb79SJohannes Weiner depends on ZSWAP 65519bcb79SJohannes Weiner default ZSWAP_COMPRESSOR_DEFAULT_LZO 66519bcb79SJohannes Weiner help 67519bcb79SJohannes Weiner Selects the default compression algorithm for the compressed cache 68519bcb79SJohannes Weiner for swap pages. 69519bcb79SJohannes Weiner 70519bcb79SJohannes Weiner For an overview what kind of performance can be expected from 71519bcb79SJohannes Weiner a particular compression algorithm please refer to the benchmarks 72519bcb79SJohannes Weiner available at the following LWN page: 73519bcb79SJohannes Weiner https://lwn.net/Articles/751795/ 74519bcb79SJohannes Weiner 75519bcb79SJohannes Weiner If in doubt, select 'LZO'. 76519bcb79SJohannes Weiner 77519bcb79SJohannes Weiner The selection made here can be overridden by using the kernel 78519bcb79SJohannes Weiner command line 'zswap.compressor=' option. 79519bcb79SJohannes Weiner 80519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_DEFLATE 81519bcb79SJohannes Weiner bool "Deflate" 82519bcb79SJohannes Weiner select CRYPTO_DEFLATE 83519bcb79SJohannes Weiner help 84519bcb79SJohannes Weiner Use the Deflate algorithm as the default compression algorithm. 85519bcb79SJohannes Weiner 86519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_LZO 87519bcb79SJohannes Weiner bool "LZO" 88519bcb79SJohannes Weiner select CRYPTO_LZO 89519bcb79SJohannes Weiner help 90519bcb79SJohannes Weiner Use the LZO algorithm as the default compression algorithm. 91519bcb79SJohannes Weiner 92519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_842 93519bcb79SJohannes Weiner bool "842" 94519bcb79SJohannes Weiner select CRYPTO_842 95519bcb79SJohannes Weiner help 96519bcb79SJohannes Weiner Use the 842 algorithm as the default compression algorithm. 97519bcb79SJohannes Weiner 98519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_LZ4 99519bcb79SJohannes Weiner bool "LZ4" 100519bcb79SJohannes Weiner select CRYPTO_LZ4 101519bcb79SJohannes Weiner help 102519bcb79SJohannes Weiner Use the LZ4 algorithm as the default compression algorithm. 103519bcb79SJohannes Weiner 104519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_LZ4HC 105519bcb79SJohannes Weiner bool "LZ4HC" 106519bcb79SJohannes Weiner select CRYPTO_LZ4HC 107519bcb79SJohannes Weiner help 108519bcb79SJohannes Weiner Use the LZ4HC algorithm as the default compression algorithm. 109519bcb79SJohannes Weiner 110519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_ZSTD 111519bcb79SJohannes Weiner bool "zstd" 112519bcb79SJohannes Weiner select CRYPTO_ZSTD 113519bcb79SJohannes Weiner help 114519bcb79SJohannes Weiner Use the zstd algorithm as the default compression algorithm. 115519bcb79SJohannes Weinerendchoice 116519bcb79SJohannes Weiner 117519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT 118519bcb79SJohannes Weiner string 119519bcb79SJohannes Weiner depends on ZSWAP 120519bcb79SJohannes Weiner default "deflate" if ZSWAP_COMPRESSOR_DEFAULT_DEFLATE 121519bcb79SJohannes Weiner default "lzo" if ZSWAP_COMPRESSOR_DEFAULT_LZO 122519bcb79SJohannes Weiner default "842" if ZSWAP_COMPRESSOR_DEFAULT_842 123519bcb79SJohannes Weiner default "lz4" if ZSWAP_COMPRESSOR_DEFAULT_LZ4 124519bcb79SJohannes Weiner default "lz4hc" if ZSWAP_COMPRESSOR_DEFAULT_LZ4HC 125519bcb79SJohannes Weiner default "zstd" if ZSWAP_COMPRESSOR_DEFAULT_ZSTD 126519bcb79SJohannes Weiner default "" 127519bcb79SJohannes Weiner 128519bcb79SJohannes Weinerchoice 129b3fbd58fSJohannes Weiner prompt "Default allocator" 130519bcb79SJohannes Weiner depends on ZSWAP 13164d4d49cSNhat Pham default ZSWAP_ZPOOL_DEFAULT_ZSMALLOC if MMU 132519bcb79SJohannes Weiner default ZSWAP_ZPOOL_DEFAULT_ZBUD 133519bcb79SJohannes Weiner help 134519bcb79SJohannes Weiner Selects the default allocator for the compressed cache for 135519bcb79SJohannes Weiner swap pages. 136519bcb79SJohannes Weiner The default is 'zbud' for compatibility, however please do 137519bcb79SJohannes Weiner read the description of each of the allocators below before 138519bcb79SJohannes Weiner making a right choice. 139519bcb79SJohannes Weiner 140519bcb79SJohannes Weiner The selection made here can be overridden by using the kernel 141519bcb79SJohannes Weiner command line 'zswap.zpool=' option. 142519bcb79SJohannes Weiner 143519bcb79SJohannes Weinerconfig ZSWAP_ZPOOL_DEFAULT_ZBUD 144519bcb79SJohannes Weiner bool "zbud" 145519bcb79SJohannes Weiner select ZBUD 146519bcb79SJohannes Weiner help 147519bcb79SJohannes Weiner Use the zbud allocator as the default allocator. 148519bcb79SJohannes Weiner 149519bcb79SJohannes Weinerconfig ZSWAP_ZPOOL_DEFAULT_Z3FOLD 150519bcb79SJohannes Weiner bool "z3fold" 151519bcb79SJohannes Weiner select Z3FOLD 152519bcb79SJohannes Weiner help 153519bcb79SJohannes Weiner Use the z3fold allocator as the default allocator. 154519bcb79SJohannes Weiner 155519bcb79SJohannes Weinerconfig ZSWAP_ZPOOL_DEFAULT_ZSMALLOC 156519bcb79SJohannes Weiner bool "zsmalloc" 157519bcb79SJohannes Weiner select ZSMALLOC 158519bcb79SJohannes Weiner help 159519bcb79SJohannes Weiner Use the zsmalloc allocator as the default allocator. 160519bcb79SJohannes Weinerendchoice 161519bcb79SJohannes Weiner 162519bcb79SJohannes Weinerconfig ZSWAP_ZPOOL_DEFAULT 163519bcb79SJohannes Weiner string 164519bcb79SJohannes Weiner depends on ZSWAP 165519bcb79SJohannes Weiner default "zbud" if ZSWAP_ZPOOL_DEFAULT_ZBUD 166519bcb79SJohannes Weiner default "z3fold" if ZSWAP_ZPOOL_DEFAULT_Z3FOLD 167519bcb79SJohannes Weiner default "zsmalloc" if ZSWAP_ZPOOL_DEFAULT_ZSMALLOC 168519bcb79SJohannes Weiner default "" 169519bcb79SJohannes Weiner 170519bcb79SJohannes Weinerconfig ZBUD 171b3fbd58fSJohannes Weiner tristate "2:1 compression allocator (zbud)" 172b3fbd58fSJohannes Weiner depends on ZSWAP 173519bcb79SJohannes Weiner help 174519bcb79SJohannes Weiner A special purpose allocator for storing compressed pages. 175519bcb79SJohannes Weiner It is designed to store up to two compressed pages per physical 176519bcb79SJohannes Weiner page. While this design limits storage density, it has simple and 177519bcb79SJohannes Weiner deterministic reclaim properties that make it preferable to a higher 178519bcb79SJohannes Weiner density approach when reclaim will be used. 179519bcb79SJohannes Weiner 180519bcb79SJohannes Weinerconfig Z3FOLD 181b3fbd58fSJohannes Weiner tristate "3:1 compression allocator (z3fold)" 182b3fbd58fSJohannes Weiner depends on ZSWAP 183519bcb79SJohannes Weiner help 184519bcb79SJohannes Weiner A special purpose allocator for storing compressed pages. 185519bcb79SJohannes Weiner It is designed to store up to three compressed pages per physical 186519bcb79SJohannes Weiner page. It is a ZBUD derivative so the simplicity and determinism are 187519bcb79SJohannes Weiner still there. 188519bcb79SJohannes Weiner 189519bcb79SJohannes Weinerconfig ZSMALLOC 190b3fbd58fSJohannes Weiner tristate 191b3fbd58fSJohannes Weiner prompt "N:1 compression allocator (zsmalloc)" if ZSWAP 192519bcb79SJohannes Weiner depends on MMU 193519bcb79SJohannes Weiner help 194519bcb79SJohannes Weiner zsmalloc is a slab-based memory allocator designed to store 195b3fbd58fSJohannes Weiner pages of various compression levels efficiently. It achieves 196b3fbd58fSJohannes Weiner the highest storage density with the least amount of fragmentation. 197519bcb79SJohannes Weiner 198519bcb79SJohannes Weinerconfig ZSMALLOC_STAT 199519bcb79SJohannes Weiner bool "Export zsmalloc statistics" 200519bcb79SJohannes Weiner depends on ZSMALLOC 201519bcb79SJohannes Weiner select DEBUG_FS 202519bcb79SJohannes Weiner help 203519bcb79SJohannes Weiner This option enables code in the zsmalloc to collect various 204519bcb79SJohannes Weiner statistics about what's happening in zsmalloc and exports that 205519bcb79SJohannes Weiner information to userspace via debugfs. 206519bcb79SJohannes Weiner If unsure, say N. 207519bcb79SJohannes Weiner 2084ff93b29SSergey Senozhatskyconfig ZSMALLOC_CHAIN_SIZE 2094ff93b29SSergey Senozhatsky int "Maximum number of physical pages per-zspage" 210b46402faSSergey Senozhatsky default 8 2114ff93b29SSergey Senozhatsky range 4 16 2124ff93b29SSergey Senozhatsky depends on ZSMALLOC 2134ff93b29SSergey Senozhatsky help 2144ff93b29SSergey Senozhatsky This option sets the upper limit on the number of physical pages 2154ff93b29SSergey Senozhatsky that a zmalloc page (zspage) can consist of. The optimal zspage 2164ff93b29SSergey Senozhatsky chain size is calculated for each size class during the 2174ff93b29SSergey Senozhatsky initialization of the pool. 2184ff93b29SSergey Senozhatsky 2194ff93b29SSergey Senozhatsky Changing this option can alter the characteristics of size classes, 2204ff93b29SSergey Senozhatsky such as the number of pages per zspage and the number of objects 2214ff93b29SSergey Senozhatsky per zspage. This can also result in different configurations of 2224ff93b29SSergey Senozhatsky the pool, as zsmalloc merges size classes with similar 2234ff93b29SSergey Senozhatsky characteristics. 2244ff93b29SSergey Senozhatsky 2254ff93b29SSergey Senozhatsky For more information, see zsmalloc documentation. 2264ff93b29SSergey Senozhatsky 2272a19be61SVlastimil Babkamenu "Slab allocator options" 2287b42f104SJohannes Weiner 2297b42f104SJohannes Weinerconfig SLUB 2302a19be61SVlastimil Babka def_bool y 231eb07c4f3SVlastimil Babka 232e240e53aSVlastimil Babkaconfig SLUB_TINY 2332a19be61SVlastimil Babka bool "Configure for minimal memory footprint" 2342a19be61SVlastimil Babka depends on EXPERT 235e240e53aSVlastimil Babka select SLAB_MERGE_DEFAULT 236e240e53aSVlastimil Babka help 2372a19be61SVlastimil Babka Configures the slab allocator in a way to achieve minimal memory 238e240e53aSVlastimil Babka footprint, sacrificing scalability, debugging and other features. 239e240e53aSVlastimil Babka This is intended only for the smallest system that had used the 240e240e53aSVlastimil Babka SLOB allocator and is not recommended for systems with more than 241e240e53aSVlastimil Babka 16MB RAM. 242e240e53aSVlastimil Babka 243e240e53aSVlastimil Babka If unsure, say N. 244e240e53aSVlastimil Babka 2457b42f104SJohannes Weinerconfig SLAB_MERGE_DEFAULT 2467b42f104SJohannes Weiner bool "Allow slab caches to be merged" 2477b42f104SJohannes Weiner default y 2487b42f104SJohannes Weiner help 2497b42f104SJohannes Weiner For reduced kernel memory fragmentation, slab caches can be 2507b42f104SJohannes Weiner merged when they share the same size and other characteristics. 2517b42f104SJohannes Weiner This carries a risk of kernel heap overflows being able to 2527b42f104SJohannes Weiner overwrite objects from merged caches (and more easily control 2537b42f104SJohannes Weiner cache layout), which makes such heap attacks easier to exploit 2547b42f104SJohannes Weiner by attackers. By keeping caches unmerged, these kinds of exploits 2557b42f104SJohannes Weiner can usually only damage objects in the same cache. To disable 2567b42f104SJohannes Weiner merging at runtime, "slab_nomerge" can be passed on the kernel 2577b42f104SJohannes Weiner command line. 2587b42f104SJohannes Weiner 2597b42f104SJohannes Weinerconfig SLAB_FREELIST_RANDOM 2607b42f104SJohannes Weiner bool "Randomize slab freelist" 2612a19be61SVlastimil Babka depends on !SLUB_TINY 2627b42f104SJohannes Weiner help 2637b42f104SJohannes Weiner Randomizes the freelist order used on creating new pages. This 2647b42f104SJohannes Weiner security feature reduces the predictability of the kernel slab 2657b42f104SJohannes Weiner allocator against heap overflows. 2667b42f104SJohannes Weiner 2677b42f104SJohannes Weinerconfig SLAB_FREELIST_HARDENED 2687b42f104SJohannes Weiner bool "Harden slab freelist metadata" 2692a19be61SVlastimil Babka depends on !SLUB_TINY 2707b42f104SJohannes Weiner help 2717b42f104SJohannes Weiner Many kernel heap attacks try to target slab cache metadata and 2727b42f104SJohannes Weiner other infrastructure. This options makes minor performance 2737b42f104SJohannes Weiner sacrifices to harden the kernel slab allocator against common 2742a19be61SVlastimil Babka freelist exploit methods. 2757b42f104SJohannes Weiner 2760710d012SVlastimil Babkaconfig SLUB_STATS 2770710d012SVlastimil Babka default n 2782a19be61SVlastimil Babka bool "Enable performance statistics" 2792a19be61SVlastimil Babka depends on SYSFS && !SLUB_TINY 2800710d012SVlastimil Babka help 2812a19be61SVlastimil Babka The statistics are useful to debug slab allocation behavior in 2820710d012SVlastimil Babka order find ways to optimize the allocator. This should never be 2830710d012SVlastimil Babka enabled for production use since keeping statistics slows down 2840710d012SVlastimil Babka the allocator by a few percentage points. The slabinfo command 2850710d012SVlastimil Babka supports the determination of the most active slabs to figure 2860710d012SVlastimil Babka out which slabs are relevant to a particular load. 2870710d012SVlastimil Babka Try running: slabinfo -DA 2880710d012SVlastimil Babka 289519bcb79SJohannes Weinerconfig SLUB_CPU_PARTIAL 290519bcb79SJohannes Weiner default y 2912a19be61SVlastimil Babka depends on SMP && !SLUB_TINY 2922a19be61SVlastimil Babka bool "Enable per cpu partial caches" 293519bcb79SJohannes Weiner help 294519bcb79SJohannes Weiner Per cpu partial caches accelerate objects allocation and freeing 295519bcb79SJohannes Weiner that is local to a processor at the price of more indeterminism 296519bcb79SJohannes Weiner in the latency of the free. On overflow these caches will be cleared 297519bcb79SJohannes Weiner which requires the taking of locks that may cause latency spikes. 298519bcb79SJohannes Weiner Typically one would choose no for a realtime system. 299519bcb79SJohannes Weiner 3003c615294SGONG, Ruiqiconfig RANDOM_KMALLOC_CACHES 3013c615294SGONG, Ruiqi default n 3022a19be61SVlastimil Babka depends on !SLUB_TINY 3033c615294SGONG, Ruiqi bool "Randomize slab caches for normal kmalloc" 3043c615294SGONG, Ruiqi help 3053c615294SGONG, Ruiqi A hardening feature that creates multiple copies of slab caches for 3063c615294SGONG, Ruiqi normal kmalloc allocation and makes kmalloc randomly pick one based 3073c615294SGONG, Ruiqi on code address, which makes the attackers more difficult to spray 3083c615294SGONG, Ruiqi vulnerable memory objects on the heap for the purpose of exploiting 3093c615294SGONG, Ruiqi memory vulnerabilities. 3103c615294SGONG, Ruiqi 3113c615294SGONG, Ruiqi Currently the number of copies is set to 16, a reasonably large value 3123c615294SGONG, Ruiqi that effectively diverges the memory objects allocated for different 3133c615294SGONG, Ruiqi subsystems or modules into different caches, at the expense of a 3143c615294SGONG, Ruiqi limited degree of memory and CPU overhead that relates to hardware and 3153c615294SGONG, Ruiqi system workload. 3163c615294SGONG, Ruiqi 3172a19be61SVlastimil Babkaendmenu # Slab allocator options 318519bcb79SJohannes Weiner 3197b42f104SJohannes Weinerconfig SHUFFLE_PAGE_ALLOCATOR 3207b42f104SJohannes Weiner bool "Page allocator randomization" 3217b42f104SJohannes Weiner default SLAB_FREELIST_RANDOM && ACPI_NUMA 3227b42f104SJohannes Weiner help 3237b42f104SJohannes Weiner Randomization of the page allocator improves the average 3247b42f104SJohannes Weiner utilization of a direct-mapped memory-side-cache. See section 3257b42f104SJohannes Weiner 5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI 3267b42f104SJohannes Weiner 6.2a specification for an example of how a platform advertises 3277b42f104SJohannes Weiner the presence of a memory-side-cache. There are also incidental 3287b42f104SJohannes Weiner security benefits as it reduces the predictability of page 3297b42f104SJohannes Weiner allocations to compliment SLAB_FREELIST_RANDOM, but the 3305e0a760bSKirill A. Shutemov default granularity of shuffling on the MAX_PAGE_ORDER i.e, 10th 33123baf831SKirill A. Shutemov order of pages is selected based on cache utilization benefits 33223baf831SKirill A. Shutemov on x86. 3337b42f104SJohannes Weiner 3347b42f104SJohannes Weiner While the randomization improves cache utilization it may 3357b42f104SJohannes Weiner negatively impact workloads on platforms without a cache. For 3367b42f104SJohannes Weiner this reason, by default, the randomization is enabled only 3377b42f104SJohannes Weiner after runtime detection of a direct-mapped memory-side-cache. 3387b42f104SJohannes Weiner Otherwise, the randomization may be force enabled with the 3397b42f104SJohannes Weiner 'page_alloc.shuffle' kernel command line parameter. 3407b42f104SJohannes Weiner 3417b42f104SJohannes Weiner Say Y if unsure. 3427b42f104SJohannes Weiner 3430710d012SVlastimil Babkaconfig COMPAT_BRK 3440710d012SVlastimil Babka bool "Disable heap randomization" 3450710d012SVlastimil Babka default y 3460710d012SVlastimil Babka help 3470710d012SVlastimil Babka Randomizing heap placement makes heap exploits harder, but it 3480710d012SVlastimil Babka also breaks ancient binaries (including anything libc5 based). 3490710d012SVlastimil Babka This option changes the bootup default to heap randomization 3500710d012SVlastimil Babka disabled, and can be overridden at runtime by setting 3510710d012SVlastimil Babka /proc/sys/kernel/randomize_va_space to 2. 3520710d012SVlastimil Babka 3530710d012SVlastimil Babka On non-ancient distros (post-2000 ones) N is usually a safe choice. 3540710d012SVlastimil Babka 3550710d012SVlastimil Babkaconfig MMAP_ALLOW_UNINITIALIZED 3560710d012SVlastimil Babka bool "Allow mmapped anonymous memory to be uninitialized" 3570710d012SVlastimil Babka depends on EXPERT && !MMU 3580710d012SVlastimil Babka default n 3590710d012SVlastimil Babka help 3600710d012SVlastimil Babka Normally, and according to the Linux spec, anonymous memory obtained 3610710d012SVlastimil Babka from mmap() has its contents cleared before it is passed to 3620710d012SVlastimil Babka userspace. Enabling this config option allows you to request that 3630710d012SVlastimil Babka mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus 3640710d012SVlastimil Babka providing a huge performance boost. If this option is not enabled, 3650710d012SVlastimil Babka then the flag will be ignored. 3660710d012SVlastimil Babka 3670710d012SVlastimil Babka This is taken advantage of by uClibc's malloc(), and also by 3680710d012SVlastimil Babka ELF-FDPIC binfmt's brk and stack allocator. 3690710d012SVlastimil Babka 3700710d012SVlastimil Babka Because of the obvious security issues, this option should only be 3710710d012SVlastimil Babka enabled on embedded devices where you control what is run in 3720710d012SVlastimil Babka userspace. Since that isn't generally a problem on no-MMU systems, 3730710d012SVlastimil Babka it is normally safe to say Y here. 3740710d012SVlastimil Babka 3750710d012SVlastimil Babka See Documentation/admin-guide/mm/nommu-mmap.rst for more information. 3760710d012SVlastimil Babka 377e1785e85SDave Hansenconfig SELECT_MEMORY_MODEL 378e1785e85SDave Hansen def_bool y 379a8826eebSKees Cook depends on ARCH_SELECT_MEMORY_MODEL 380e1785e85SDave Hansen 3813a9da765SDave Hansenchoice 3823a9da765SDave Hansen prompt "Memory model" 383e1785e85SDave Hansen depends on SELECT_MEMORY_MODEL 384d41dee36SAndy Whitcroft default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT 385e1785e85SDave Hansen default FLATMEM_MANUAL 386d66d109dSMike Rapoport help 387d66d109dSMike Rapoport This option allows you to change some of the ways that 388d66d109dSMike Rapoport Linux manages its memory internally. Most users will 389d66d109dSMike Rapoport only have one option here selected by the architecture 390d66d109dSMike Rapoport configuration. This is normal. 3913a9da765SDave Hansen 392e1785e85SDave Hansenconfig FLATMEM_MANUAL 3933a9da765SDave Hansen bool "Flat Memory" 394bb1c50d3SMike Rapoport depends on !ARCH_SPARSEMEM_ENABLE || ARCH_FLATMEM_ENABLE 3953a9da765SDave Hansen help 396d66d109dSMike Rapoport This option is best suited for non-NUMA systems with 397d66d109dSMike Rapoport flat address space. The FLATMEM is the most efficient 398d66d109dSMike Rapoport system in terms of performance and resource consumption 399d66d109dSMike Rapoport and it is the best option for smaller systems. 4003a9da765SDave Hansen 401d66d109dSMike Rapoport For systems that have holes in their physical address 402d66d109dSMike Rapoport spaces and for features like NUMA and memory hotplug, 403dd33d29aSRandy Dunlap choose "Sparse Memory". 404d41dee36SAndy Whitcroft 405d41dee36SAndy Whitcroft If unsure, choose this option (Flat Memory) over any other. 4063a9da765SDave Hansen 407d41dee36SAndy Whitcroftconfig SPARSEMEM_MANUAL 408d41dee36SAndy Whitcroft bool "Sparse Memory" 409d41dee36SAndy Whitcroft depends on ARCH_SPARSEMEM_ENABLE 410d41dee36SAndy Whitcroft help 411d41dee36SAndy Whitcroft This will be the only option for some systems, including 412d66d109dSMike Rapoport memory hot-plug systems. This is normal. 413d41dee36SAndy Whitcroft 414d66d109dSMike Rapoport This option provides efficient support for systems with 415d66d109dSMike Rapoport holes is their physical address space and allows memory 416d66d109dSMike Rapoport hot-plug and hot-remove. 417d41dee36SAndy Whitcroft 418d66d109dSMike Rapoport If unsure, choose "Flat Memory" over this option. 419d41dee36SAndy Whitcroft 4203a9da765SDave Hansenendchoice 4213a9da765SDave Hansen 422d41dee36SAndy Whitcroftconfig SPARSEMEM 423d41dee36SAndy Whitcroft def_bool y 4241a83e175SRussell King depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL 425d41dee36SAndy Whitcroft 426e1785e85SDave Hansenconfig FLATMEM 427e1785e85SDave Hansen def_bool y 428bb1c50d3SMike Rapoport depends on !SPARSEMEM || FLATMEM_MANUAL 429d41dee36SAndy Whitcroft 43093b7504eSDave Hansen# 4313e347261SBob Picco# SPARSEMEM_EXTREME (which is the default) does some bootmem 432c89ab04fSMike Rapoport# allocations when sparse_init() is called. If this cannot 4333e347261SBob Picco# be done on your architecture, select this option. However, 4343e347261SBob Picco# statically allocating the mem_section[] array can potentially 4353e347261SBob Picco# consume vast quantities of .bss, so be careful. 4363e347261SBob Picco# 4373e347261SBob Picco# This option will also potentially produce smaller runtime code 4383e347261SBob Picco# with gcc 3.4 and later. 4393e347261SBob Picco# 4403e347261SBob Piccoconfig SPARSEMEM_STATIC 4419ba16087SJan Beulich bool 4423e347261SBob Picco 4433e347261SBob Picco# 44444c09201SMatt LaPlante# Architecture platforms which require a two level mem_section in SPARSEMEM 445802f192eSBob Picco# must select this option. This is usually for architecture platforms with 446802f192eSBob Picco# an extremely sparse physical address space. 447802f192eSBob Picco# 4483e347261SBob Piccoconfig SPARSEMEM_EXTREME 4493e347261SBob Picco def_bool y 4503e347261SBob Picco depends on SPARSEMEM && !SPARSEMEM_STATIC 4514c21e2f2SHugh Dickins 45229c71111SAndy Whitcroftconfig SPARSEMEM_VMEMMAP_ENABLE 4539ba16087SJan Beulich bool 45429c71111SAndy Whitcroft 45529c71111SAndy Whitcroftconfig SPARSEMEM_VMEMMAP 456a5ee6daaSGeoff Levand bool "Sparse Memory virtual memmap" 457a5ee6daaSGeoff Levand depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE 458a5ee6daaSGeoff Levand default y 459a5ee6daaSGeoff Levand help 460a5ee6daaSGeoff Levand SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise 461a5ee6daaSGeoff Levand pfn_to_page and page_to_pfn operations. This is the most 462a5ee6daaSGeoff Levand efficient option when sufficient kernel resources are available. 4630b376f1eSAneesh Kumar K.V# 4640b376f1eSAneesh Kumar K.V# Select this config option from the architecture Kconfig, if it is preferred 4650b376f1eSAneesh Kumar K.V# to enable the feature of HugeTLB/dev_dax vmemmap optimization. 4660b376f1eSAneesh Kumar K.V# 4670b6f1582SAneesh Kumar K.Vconfig ARCH_WANT_OPTIMIZE_DAX_VMEMMAP 4680b6f1582SAneesh Kumar K.V bool 4690b6f1582SAneesh Kumar K.V 4700b6f1582SAneesh Kumar K.Vconfig ARCH_WANT_OPTIMIZE_HUGETLB_VMEMMAP 4710b376f1eSAneesh Kumar K.V bool 47229c71111SAndy Whitcroft 47370210ed9SPhilipp Hachtmannconfig HAVE_MEMBLOCK_PHYS_MAP 4746341e62bSChristoph Jaeger bool 47570210ed9SPhilipp Hachtmann 47667a929e0SChristoph Hellwigconfig HAVE_FAST_GUP 477050a9adcSChristoph Hellwig depends on MMU 4786341e62bSChristoph Jaeger bool 4792667f50eSSteve Capper 48052219aeaSDavid Hildenbrand# Don't discard allocated memory used to track "memory" and "reserved" memblocks 48152219aeaSDavid Hildenbrand# after early boot, so it can still be used to test for validity of memory. 48252219aeaSDavid Hildenbrand# Also, memblocks are updated with memory hot(un)plug. 483350e88baSMike Rapoportconfig ARCH_KEEP_MEMBLOCK 4846341e62bSChristoph Jaeger bool 485c378ddd5STejun Heo 4861e5d8e1eSDan Williams# Keep arch NUMA mapping infrastructure post-init. 4871e5d8e1eSDan Williamsconfig NUMA_KEEP_MEMINFO 4881e5d8e1eSDan Williams bool 4891e5d8e1eSDan Williams 490ee6f509cSMinchan Kimconfig MEMORY_ISOLATION 4916341e62bSChristoph Jaeger bool 492ee6f509cSMinchan Kim 493a9e7b8d4SDavid Hildenbrand# IORESOURCE_SYSTEM_RAM regions in the kernel resource tree that are marked 494a9e7b8d4SDavid Hildenbrand# IORESOURCE_EXCLUSIVE cannot be mapped to user space, for example, via 495a9e7b8d4SDavid Hildenbrand# /dev/mem. 496a9e7b8d4SDavid Hildenbrandconfig EXCLUSIVE_SYSTEM_RAM 497a9e7b8d4SDavid Hildenbrand def_bool y 498a9e7b8d4SDavid Hildenbrand depends on !DEVMEM || STRICT_DEVMEM 499a9e7b8d4SDavid Hildenbrand 50046723bfaSYasuaki Ishimatsu# 50146723bfaSYasuaki Ishimatsu# Only be set on architectures that have completely implemented memory hotplug 50246723bfaSYasuaki Ishimatsu# feature. If you are not sure, don't touch it. 50346723bfaSYasuaki Ishimatsu# 50446723bfaSYasuaki Ishimatsuconfig HAVE_BOOTMEM_INFO_NODE 50546723bfaSYasuaki Ishimatsu def_bool n 50646723bfaSYasuaki Ishimatsu 50791024b3cSAnshuman Khandualconfig ARCH_ENABLE_MEMORY_HOTPLUG 50891024b3cSAnshuman Khandual bool 50991024b3cSAnshuman Khandual 510519bcb79SJohannes Weinerconfig ARCH_ENABLE_MEMORY_HOTREMOVE 511519bcb79SJohannes Weiner bool 512519bcb79SJohannes Weiner 5133947be19SDave Hansen# eventually, we can have this option just 'select SPARSEMEM' 514519bcb79SJohannes Weinermenuconfig MEMORY_HOTPLUG 515519bcb79SJohannes Weiner bool "Memory hotplug" 516b30c5927SDavid Hildenbrand select MEMORY_ISOLATION 51771b6f2ddSDavid Hildenbrand depends on SPARSEMEM 51840b31360SStephen Rothwell depends on ARCH_ENABLE_MEMORY_HOTPLUG 5197ec58a2bSDavid Hildenbrand depends on 64BIT 5201e5d8e1eSDan Williams select NUMA_KEEP_MEMINFO if NUMA 5213947be19SDave Hansen 522519bcb79SJohannes Weinerif MEMORY_HOTPLUG 523519bcb79SJohannes Weiner 5248604d9e5SVitaly Kuznetsovconfig MEMORY_HOTPLUG_DEFAULT_ONLINE 5258604d9e5SVitaly Kuznetsov bool "Online the newly added memory blocks by default" 5268604d9e5SVitaly Kuznetsov depends on MEMORY_HOTPLUG 5278604d9e5SVitaly Kuznetsov help 5288604d9e5SVitaly Kuznetsov This option sets the default policy setting for memory hotplug 5298604d9e5SVitaly Kuznetsov onlining policy (/sys/devices/system/memory/auto_online_blocks) which 5308604d9e5SVitaly Kuznetsov determines what happens to newly added memory regions. Policy setting 5318604d9e5SVitaly Kuznetsov can always be changed at runtime. 532cb1aaebeSMauro Carvalho Chehab See Documentation/admin-guide/mm/memory-hotplug.rst for more information. 5338604d9e5SVitaly Kuznetsov 5348604d9e5SVitaly Kuznetsov Say Y here if you want all hot-plugged memory blocks to appear in 5358604d9e5SVitaly Kuznetsov 'online' state by default. 5368604d9e5SVitaly Kuznetsov Say N here if you want the default policy to keep all hot-plugged 5378604d9e5SVitaly Kuznetsov memory blocks in 'offline' state. 5388604d9e5SVitaly Kuznetsov 5390c0e6195SKAMEZAWA Hiroyukiconfig MEMORY_HOTREMOVE 5400c0e6195SKAMEZAWA Hiroyuki bool "Allow for memory hot remove" 541f7e3334aSNathan Fontenot select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64) 5420c0e6195SKAMEZAWA Hiroyuki depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE 5430c0e6195SKAMEZAWA Hiroyuki depends on MIGRATION 5440c0e6195SKAMEZAWA Hiroyuki 545a08a2ae3SOscar Salvadorconfig MHP_MEMMAP_ON_MEMORY 546a08a2ae3SOscar Salvador def_bool y 547a08a2ae3SOscar Salvador depends on MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP 548a08a2ae3SOscar Salvador depends on ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE 549a08a2ae3SOscar Salvador 550519bcb79SJohannes Weinerendif # MEMORY_HOTPLUG 551519bcb79SJohannes Weiner 55204d5ea46SAneesh Kumar K.Vconfig ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE 55304d5ea46SAneesh Kumar K.V bool 55404d5ea46SAneesh Kumar K.V 5554c21e2f2SHugh Dickins# Heavily threaded applications may benefit from splitting the mm-wide 5564c21e2f2SHugh Dickins# page_table_lock, so that faults on different parts of the user address 5574c21e2f2SHugh Dickins# space can be handled with less contention: split it at this NR_CPUS. 5584c21e2f2SHugh Dickins# Default to 4 for wider testing, though 8 might be more appropriate. 5594c21e2f2SHugh Dickins# ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock. 5607b6ac9dfSHugh Dickins# PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes. 56160bccaa6SWill Deacon# SPARC32 allocates multiple pte tables within a single page, and therefore 56260bccaa6SWill Deacon# a per-page lock leads to problems when multiple tables need to be locked 56360bccaa6SWill Deacon# at the same time (e.g. copy_page_range()). 564a70caa8bSHugh Dickins# DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page. 5654c21e2f2SHugh Dickins# 5664c21e2f2SHugh Dickinsconfig SPLIT_PTLOCK_CPUS 5674c21e2f2SHugh Dickins int 5689164550eSKirill A. Shutemov default "999999" if !MMU 569a70caa8bSHugh Dickins default "999999" if ARM && !CPU_CACHE_VIPT 570a70caa8bSHugh Dickins default "999999" if PARISC && !PA20 57160bccaa6SWill Deacon default "999999" if SPARC32 5724c21e2f2SHugh Dickins default "4" 5737cbe34cfSChristoph Lameter 574e009bb30SKirill A. Shutemovconfig ARCH_ENABLE_SPLIT_PMD_PTLOCK 5756341e62bSChristoph Jaeger bool 576e009bb30SKirill A. Shutemov 5777cbe34cfSChristoph Lameter# 57809316c09SKonstantin Khlebnikov# support for memory balloon 57909316c09SKonstantin Khlebnikovconfig MEMORY_BALLOON 5806341e62bSChristoph Jaeger bool 58109316c09SKonstantin Khlebnikov 58209316c09SKonstantin Khlebnikov# 58318468d93SRafael Aquini# support for memory balloon compaction 58418468d93SRafael Aquiniconfig BALLOON_COMPACTION 58518468d93SRafael Aquini bool "Allow for balloon memory compaction/migration" 586*cd14b018SMasahiro Yamada default y 58709316c09SKonstantin Khlebnikov depends on COMPACTION && MEMORY_BALLOON 58818468d93SRafael Aquini help 58918468d93SRafael Aquini Memory fragmentation introduced by ballooning might reduce 59018468d93SRafael Aquini significantly the number of 2MB contiguous memory blocks that can be 59118468d93SRafael Aquini used within a guest, thus imposing performance penalties associated 59218468d93SRafael Aquini with the reduced number of transparent huge pages that could be used 59318468d93SRafael Aquini by the guest workload. Allowing the compaction & migration for memory 59418468d93SRafael Aquini pages enlisted as being part of memory balloon devices avoids the 59518468d93SRafael Aquini scenario aforementioned and helps improving memory defragmentation. 59618468d93SRafael Aquini 59718468d93SRafael Aquini# 598e9e96b39SMel Gorman# support for memory compaction 599e9e96b39SMel Gormanconfig COMPACTION 600e9e96b39SMel Gorman bool "Allow for memory compaction" 601*cd14b018SMasahiro Yamada default y 602e9e96b39SMel Gorman select MIGRATION 60333a93877SAndrea Arcangeli depends on MMU 604e9e96b39SMel Gorman help 605b32eaf71SMichal Hocko Compaction is the only memory management component to form 606b32eaf71SMichal Hocko high order (larger physically contiguous) memory blocks 607b32eaf71SMichal Hocko reliably. The page allocator relies on compaction heavily and 608b32eaf71SMichal Hocko the lack of the feature can lead to unexpected OOM killer 609b32eaf71SMichal Hocko invocations for high order memory requests. You shouldn't 610b32eaf71SMichal Hocko disable this option unless there really is a strong reason for 611b32eaf71SMichal Hocko it and then we would be really interested to hear about that at 612b32eaf71SMichal Hocko linux-mm@kvack.org. 613e9e96b39SMel Gorman 614c7e0b3d0SThomas Gleixnerconfig COMPACT_UNEVICTABLE_DEFAULT 615c7e0b3d0SThomas Gleixner int 616c7e0b3d0SThomas Gleixner depends on COMPACTION 617c7e0b3d0SThomas Gleixner default 0 if PREEMPT_RT 618c7e0b3d0SThomas Gleixner default 1 619c7e0b3d0SThomas Gleixner 620e9e96b39SMel Gorman# 62136e66c55SAlexander Duyck# support for free page reporting 62236e66c55SAlexander Duyckconfig PAGE_REPORTING 62336e66c55SAlexander Duyck bool "Free page reporting" 62436e66c55SAlexander Duyck help 62536e66c55SAlexander Duyck Free page reporting allows for the incremental acquisition of 62636e66c55SAlexander Duyck free pages from the buddy allocator for the purpose of reporting 62736e66c55SAlexander Duyck those pages to another entity, such as a hypervisor, so that the 62836e66c55SAlexander Duyck memory can be freed within the host for other uses. 62936e66c55SAlexander Duyck 63036e66c55SAlexander Duyck# 6317cbe34cfSChristoph Lameter# support for page migration 6327cbe34cfSChristoph Lameter# 6337cbe34cfSChristoph Lameterconfig MIGRATION 634b20a3503SChristoph Lameter bool "Page migration" 635*cd14b018SMasahiro Yamada default y 636de32a817SChen Gang depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU 637b20a3503SChristoph Lameter help 638b20a3503SChristoph Lameter Allows the migration of the physical location of pages of processes 639e9e96b39SMel Gorman while the virtual addresses are not changed. This is useful in 640e9e96b39SMel Gorman two situations. The first is on NUMA systems to put pages nearer 641e9e96b39SMel Gorman to the processors accessing. The second is when allocating huge 642e9e96b39SMel Gorman pages as migration can relocate pages to satisfy a huge page 643e9e96b39SMel Gorman allocation instead of reclaiming. 6446550e07fSGreg Kroah-Hartman 64576cbbeadSChristoph Hellwigconfig DEVICE_MIGRATION 646d90a25f8SChristoph Hellwig def_bool MIGRATION && ZONE_DEVICE 64776cbbeadSChristoph Hellwig 648c177c81eSNaoya Horiguchiconfig ARCH_ENABLE_HUGEPAGE_MIGRATION 6496341e62bSChristoph Jaeger bool 650c177c81eSNaoya Horiguchi 6519c670ea3SNaoya Horiguchiconfig ARCH_ENABLE_THP_MIGRATION 6529c670ea3SNaoya Horiguchi bool 6539c670ea3SNaoya Horiguchi 6544bfb68a0SAnshuman Khandualconfig HUGETLB_PAGE_SIZE_VARIABLE 6554bfb68a0SAnshuman Khandual def_bool n 6564bfb68a0SAnshuman Khandual help 6574bfb68a0SAnshuman Khandual Allows the pageblock_order value to be dynamic instead of just standard 6584bfb68a0SAnshuman Khandual HUGETLB_PAGE_ORDER when there are multiple HugeTLB page sizes available 6594bfb68a0SAnshuman Khandual on a platform. 6604bfb68a0SAnshuman Khandual 6615e0a760bSKirill A. Shutemov Note that the pageblock_order cannot exceed MAX_PAGE_ORDER and will be 6625e0a760bSKirill A. Shutemov clamped down to MAX_PAGE_ORDER. 663b3d40a2bSDavid Hildenbrand 6648df995f6SAlexandre Ghiticonfig CONTIG_ALLOC 6658df995f6SAlexandre Ghiti def_bool (MEMORY_ISOLATION && COMPACTION) || CMA 6668df995f6SAlexandre Ghiti 66752166607SHuang Yingconfig PCP_BATCH_SCALE_MAX 66852166607SHuang Ying int "Maximum scale factor of PCP (Per-CPU pageset) batch allocate/free" 66952166607SHuang Ying default 5 67052166607SHuang Ying range 0 6 67152166607SHuang Ying help 67252166607SHuang Ying In page allocator, PCP (Per-CPU pageset) is refilled and drained in 67352166607SHuang Ying batches. The batch number is scaled automatically to improve page 67452166607SHuang Ying allocation/free throughput. But too large scale factor may hurt 67552166607SHuang Ying latency. This option sets the upper limit of scale factor to limit 67652166607SHuang Ying the maximum latency. 67752166607SHuang Ying 678600715dcSJeremy Fitzhardingeconfig PHYS_ADDR_T_64BIT 679d4a451d5SChristoph Hellwig def_bool 64BIT 680600715dcSJeremy Fitzhardinge 6812a7326b5SChristoph Lameterconfig BOUNCE 6829ca24e2eSVinayak Menon bool "Enable bounce buffers" 6839ca24e2eSVinayak Menon default y 684ce288e05SChristoph Hellwig depends on BLOCK && MMU && HIGHMEM 6859ca24e2eSVinayak Menon help 686ce288e05SChristoph Hellwig Enable bounce buffers for devices that cannot access the full range of 687ce288e05SChristoph Hellwig memory available to the CPU. Enabled by default when HIGHMEM is 688ce288e05SChristoph Hellwig selected, but you may say n to override this. 6892a7326b5SChristoph Lameter 690cddb8a5cSAndrea Arcangeliconfig MMU_NOTIFIER 691cddb8a5cSAndrea Arcangeli bool 69299cb252fSJason Gunthorpe select INTERVAL_TREE 693fc4d5c29SDavid Howells 694f8af4da3SHugh Dickinsconfig KSM 695f8af4da3SHugh Dickins bool "Enable KSM for page merging" 696f8af4da3SHugh Dickins depends on MMU 69759e1a2f4STimofey Titovets select XXHASH 698f8af4da3SHugh Dickins help 699f8af4da3SHugh Dickins Enable Kernel Samepage Merging: KSM periodically scans those areas 700f8af4da3SHugh Dickins of an application's address space that an app has advised may be 701f8af4da3SHugh Dickins mergeable. When it finds pages of identical content, it replaces 702d0f209f6SHugh Dickins the many instances by a single page with that content, so 703f8af4da3SHugh Dickins saving memory until one or another app needs to modify the content. 704f8af4da3SHugh Dickins Recommended for use with KVM, or with other duplicative applications. 705ee65728eSMike Rapoport See Documentation/mm/ksm.rst for more information: KSM is inactive 706c73602adSHugh Dickins until a program has madvised that an area is MADV_MERGEABLE, and 707c73602adSHugh Dickins root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set). 708f8af4da3SHugh Dickins 709e0a94c2aSChristoph Lameterconfig DEFAULT_MMAP_MIN_ADDR 710e0a94c2aSChristoph Lameter int "Low address space to protect from user allocation" 7116e141546SDavid Howells depends on MMU 712e0a94c2aSChristoph Lameter default 4096 713e0a94c2aSChristoph Lameter help 714e0a94c2aSChristoph Lameter This is the portion of low virtual memory which should be protected 715e0a94c2aSChristoph Lameter from userspace allocation. Keeping a user from writing to low pages 716e0a94c2aSChristoph Lameter can help reduce the impact of kernel NULL pointer bugs. 717e0a94c2aSChristoph Lameter 718e99fb98dSKefeng Wang For most ppc64 and x86 users with lots of address space 719e0a94c2aSChristoph Lameter a value of 65536 is reasonable and should cause no problems. 720e0a94c2aSChristoph Lameter On arm and other archs it should not be higher than 32768. 721788084abSEric Paris Programs which use vm86 functionality or have some need to map 722788084abSEric Paris this low address space will need CAP_SYS_RAWIO or disable this 723788084abSEric Paris protection by setting the value to 0. 724e0a94c2aSChristoph Lameter 725e0a94c2aSChristoph Lameter This value can be changed after boot using the 726e0a94c2aSChristoph Lameter /proc/sys/vm/mmap_min_addr tunable. 727e0a94c2aSChristoph Lameter 728d949f36fSLinus Torvaldsconfig ARCH_SUPPORTS_MEMORY_FAILURE 729d949f36fSLinus Torvalds bool 730e0a94c2aSChristoph Lameter 7316a46079cSAndi Kleenconfig MEMORY_FAILURE 7326a46079cSAndi Kleen depends on MMU 733d949f36fSLinus Torvalds depends on ARCH_SUPPORTS_MEMORY_FAILURE 7346a46079cSAndi Kleen bool "Enable recovery from hardware memory errors" 735ee6f509cSMinchan Kim select MEMORY_ISOLATION 73697f0b134SXie XiuQi select RAS 7376a46079cSAndi Kleen help 7386a46079cSAndi Kleen Enables code to recover from some memory failures on systems 7396a46079cSAndi Kleen with MCA recovery. This allows a system to continue running 7406a46079cSAndi Kleen even when some of its memory has uncorrected errors. This requires 7416a46079cSAndi Kleen special hardware support and typically ECC memory. 7426a46079cSAndi Kleen 743cae681fcSAndi Kleenconfig HWPOISON_INJECT 744413f9efbSAndi Kleen tristate "HWPoison pages injector" 74527df5068SAndi Kleen depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS 746478c5ffcSWu Fengguang select PROC_PAGE_MONITOR 747cae681fcSAndi Kleen 748fc4d5c29SDavid Howellsconfig NOMMU_INITIAL_TRIM_EXCESS 749fc4d5c29SDavid Howells int "Turn on mmap() excess space trimming before booting" 750fc4d5c29SDavid Howells depends on !MMU 751fc4d5c29SDavid Howells default 1 752fc4d5c29SDavid Howells help 753fc4d5c29SDavid Howells The NOMMU mmap() frequently needs to allocate large contiguous chunks 754fc4d5c29SDavid Howells of memory on which to store mappings, but it can only ask the system 755fc4d5c29SDavid Howells allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently 756fc4d5c29SDavid Howells more than it requires. To deal with this, mmap() is able to trim off 757fc4d5c29SDavid Howells the excess and return it to the allocator. 758fc4d5c29SDavid Howells 759fc4d5c29SDavid Howells If trimming is enabled, the excess is trimmed off and returned to the 760fc4d5c29SDavid Howells system allocator, which can cause extra fragmentation, particularly 761fc4d5c29SDavid Howells if there are a lot of transient processes. 762fc4d5c29SDavid Howells 763fc4d5c29SDavid Howells If trimming is disabled, the excess is kept, but not used, which for 764fc4d5c29SDavid Howells long-term mappings means that the space is wasted. 765fc4d5c29SDavid Howells 766fc4d5c29SDavid Howells Trimming can be dynamically controlled through a sysctl option 767fc4d5c29SDavid Howells (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of 768fc4d5c29SDavid Howells excess pages there must be before trimming should occur, or zero if 769fc4d5c29SDavid Howells no trimming is to occur. 770fc4d5c29SDavid Howells 771fc4d5c29SDavid Howells This option specifies the initial value of this option. The default 772fc4d5c29SDavid Howells of 1 says that all excess pages should be trimmed. 773fc4d5c29SDavid Howells 774dd19d293SStephen Kitt See Documentation/admin-guide/mm/nommu-mmap.rst for more information. 775bbddff05STejun Heo 776519bcb79SJohannes Weinerconfig ARCH_WANT_GENERAL_HUGETLB 777519bcb79SJohannes Weiner bool 778519bcb79SJohannes Weiner 779519bcb79SJohannes Weinerconfig ARCH_WANTS_THP_SWAP 780519bcb79SJohannes Weiner def_bool n 781519bcb79SJohannes Weiner 782519bcb79SJohannes Weinermenuconfig TRANSPARENT_HUGEPAGE 78313ece886SAndrea Arcangeli bool "Transparent Hugepage Support" 784554b0f3cSSebastian Andrzej Siewior depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT 7855d689240SAndrea Arcangeli select COMPACTION 7863a08cd52SMatthew Wilcox select XARRAY_MULTI 7874c76d9d1SAndrea Arcangeli help 7884c76d9d1SAndrea Arcangeli Transparent Hugepages allows the kernel to use huge pages and 7894c76d9d1SAndrea Arcangeli huge tlb transparently to the applications whenever possible. 7904c76d9d1SAndrea Arcangeli This feature can improve computing performance to certain 7914c76d9d1SAndrea Arcangeli applications by speeding up page faults during memory 7924c76d9d1SAndrea Arcangeli allocation, by reducing the number of tlb misses and by speeding 7934c76d9d1SAndrea Arcangeli up the pagetable walking. 7944c76d9d1SAndrea Arcangeli 7954c76d9d1SAndrea Arcangeli If memory constrained on embedded, you may want to say N. 7964c76d9d1SAndrea Arcangeli 797519bcb79SJohannes Weinerif TRANSPARENT_HUGEPAGE 798519bcb79SJohannes Weiner 79913ece886SAndrea Arcangelichoice 80013ece886SAndrea Arcangeli prompt "Transparent Hugepage Support sysfs defaults" 80113ece886SAndrea Arcangeli depends on TRANSPARENT_HUGEPAGE 80213ece886SAndrea Arcangeli default TRANSPARENT_HUGEPAGE_ALWAYS 80313ece886SAndrea Arcangeli help 80413ece886SAndrea Arcangeli Selects the sysfs defaults for Transparent Hugepage Support. 80513ece886SAndrea Arcangeli 80613ece886SAndrea Arcangeli config TRANSPARENT_HUGEPAGE_ALWAYS 80713ece886SAndrea Arcangeli bool "always" 80813ece886SAndrea Arcangeli help 80913ece886SAndrea Arcangeli Enabling Transparent Hugepage always, can increase the 81013ece886SAndrea Arcangeli memory footprint of applications without a guaranteed 81113ece886SAndrea Arcangeli benefit but it will work automatically for all applications. 81213ece886SAndrea Arcangeli 81313ece886SAndrea Arcangeli config TRANSPARENT_HUGEPAGE_MADVISE 81413ece886SAndrea Arcangeli bool "madvise" 81513ece886SAndrea Arcangeli help 81613ece886SAndrea Arcangeli Enabling Transparent Hugepage madvise, will only provide a 81713ece886SAndrea Arcangeli performance improvement benefit to the applications using 81813ece886SAndrea Arcangeli madvise(MADV_HUGEPAGE) but it won't risk to increase the 81913ece886SAndrea Arcangeli memory footprint of applications without a guaranteed 82013ece886SAndrea Arcangeli benefit. 821683ec99fSDmytro Maluka 822683ec99fSDmytro Maluka config TRANSPARENT_HUGEPAGE_NEVER 823683ec99fSDmytro Maluka bool "never" 824683ec99fSDmytro Maluka help 825683ec99fSDmytro Maluka Disable Transparent Hugepage by default. It can still be 826683ec99fSDmytro Maluka enabled at runtime via sysfs. 82713ece886SAndrea Arcangeliendchoice 82813ece886SAndrea Arcangeli 82938d8b4e6SHuang Yingconfig THP_SWAP 83038d8b4e6SHuang Ying def_bool y 831dad6a5ebSHugh Dickins depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP && 64BIT 83238d8b4e6SHuang Ying help 83338d8b4e6SHuang Ying Swap transparent huge pages in one piece, without splitting. 83414fef284SHuang Ying XXX: For now, swap cluster backing transparent huge page 83514fef284SHuang Ying will be split after swapout. 83638d8b4e6SHuang Ying 83738d8b4e6SHuang Ying For selection by architectures with reasonable THP sizes. 83838d8b4e6SHuang Ying 839519bcb79SJohannes Weinerconfig READ_ONLY_THP_FOR_FS 840519bcb79SJohannes Weiner bool "Read-only THP for filesystems (EXPERIMENTAL)" 841519bcb79SJohannes Weiner depends on TRANSPARENT_HUGEPAGE && SHMEM 842519bcb79SJohannes Weiner 843519bcb79SJohannes Weiner help 844519bcb79SJohannes Weiner Allow khugepaged to put read-only file-backed pages in THP. 845519bcb79SJohannes Weiner 846519bcb79SJohannes Weiner This is marked experimental because it is a new feature. Write 847519bcb79SJohannes Weiner support of file THPs will be developed in the next few release 848519bcb79SJohannes Weiner cycles. 849519bcb79SJohannes Weiner 850519bcb79SJohannes Weinerendif # TRANSPARENT_HUGEPAGE 851519bcb79SJohannes Weiner 852e496cf3dSKirill A. Shutemov# 853bbddff05STejun Heo# UP and nommu archs use km based percpu allocator 854bbddff05STejun Heo# 855bbddff05STejun Heoconfig NEED_PER_CPU_KM 8563583521aSVladimir Murzin depends on !SMP || !MMU 857bbddff05STejun Heo bool 858bbddff05STejun Heo default y 859077b1f83SDan Magenheimer 8607ecd19cfSKefeng Wangconfig NEED_PER_CPU_EMBED_FIRST_CHUNK 8617ecd19cfSKefeng Wang bool 8627ecd19cfSKefeng Wang 8637ecd19cfSKefeng Wangconfig NEED_PER_CPU_PAGE_FIRST_CHUNK 8647ecd19cfSKefeng Wang bool 8657ecd19cfSKefeng Wang 8667ecd19cfSKefeng Wangconfig USE_PERCPU_NUMA_NODE_ID 8677ecd19cfSKefeng Wang bool 8687ecd19cfSKefeng Wang 8697ecd19cfSKefeng Wangconfig HAVE_SETUP_PER_CPU_AREA 8707ecd19cfSKefeng Wang bool 8717ecd19cfSKefeng Wang 872f825c736SAneesh Kumar K.Vconfig CMA 873f825c736SAneesh Kumar K.V bool "Contiguous Memory Allocator" 874aca52c39SMike Rapoport depends on MMU 875f825c736SAneesh Kumar K.V select MIGRATION 876f825c736SAneesh Kumar K.V select MEMORY_ISOLATION 877f825c736SAneesh Kumar K.V help 878f825c736SAneesh Kumar K.V This enables the Contiguous Memory Allocator which allows other 879f825c736SAneesh Kumar K.V subsystems to allocate big physically-contiguous blocks of memory. 880f825c736SAneesh Kumar K.V CMA reserves a region of memory and allows only movable pages to 881f825c736SAneesh Kumar K.V be allocated from it. This way, the kernel can use the memory for 882f825c736SAneesh Kumar K.V pagecache and when a subsystem requests for contiguous area, the 883f825c736SAneesh Kumar K.V allocated pages are migrated away to serve the contiguous request. 884f825c736SAneesh Kumar K.V 885f825c736SAneesh Kumar K.V If unsure, say "n". 886f825c736SAneesh Kumar K.V 88728b24c1fSSasha Levinconfig CMA_DEBUGFS 88828b24c1fSSasha Levin bool "CMA debugfs interface" 88928b24c1fSSasha Levin depends on CMA && DEBUG_FS 89028b24c1fSSasha Levin help 89128b24c1fSSasha Levin Turns on the DebugFS interface for CMA. 89228b24c1fSSasha Levin 89343ca106fSMinchan Kimconfig CMA_SYSFS 89443ca106fSMinchan Kim bool "CMA information through sysfs interface" 89543ca106fSMinchan Kim depends on CMA && SYSFS 89643ca106fSMinchan Kim help 89743ca106fSMinchan Kim This option exposes some sysfs attributes to get information 89843ca106fSMinchan Kim from CMA. 89943ca106fSMinchan Kim 900a254129eSJoonsoo Kimconfig CMA_AREAS 901a254129eSJoonsoo Kim int "Maximum count of the CMA areas" 902a254129eSJoonsoo Kim depends on CMA 90373307523SAnshuman Khandual default 20 if NUMA 90473307523SAnshuman Khandual default 8 905a254129eSJoonsoo Kim help 906a254129eSJoonsoo Kim CMA allows to create CMA areas for particular purpose, mainly, 907a254129eSJoonsoo Kim used as device private area. This parameter sets the maximum 908a254129eSJoonsoo Kim number of CMA area in the system. 909a254129eSJoonsoo Kim 91073307523SAnshuman Khandual If unsure, leave the default value "8" in UMA and "20" in NUMA. 911a254129eSJoonsoo Kim 912af8d417aSDan Streetmanconfig MEM_SOFT_DIRTY 913af8d417aSDan Streetman bool "Track memory changes" 914af8d417aSDan Streetman depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS 915af8d417aSDan Streetman select PROC_PAGE_MONITOR 9164e2e2770SSeth Jennings help 917af8d417aSDan Streetman This option enables memory changes tracking by introducing a 918af8d417aSDan Streetman soft-dirty bit on pte-s. This bit it set when someone writes 919af8d417aSDan Streetman into a page just as regular dirty bit, but unlike the latter 920af8d417aSDan Streetman it can be cleared by hands. 921af8d417aSDan Streetman 9221ad1335dSMike Rapoport See Documentation/admin-guide/mm/soft-dirty.rst for more details. 9234e2e2770SSeth Jennings 9249e5c33d7SMark Salterconfig GENERIC_EARLY_IOREMAP 9259e5c33d7SMark Salter bool 926042d27acSHelge Deller 92722ee3ea5SHelge Dellerconfig STACK_MAX_DEFAULT_SIZE_MB 92822ee3ea5SHelge Deller int "Default maximum user stack size for 32-bit processes (MB)" 92922ee3ea5SHelge Deller default 100 930042d27acSHelge Deller range 8 2048 931042d27acSHelge Deller depends on STACK_GROWSUP && (!64BIT || COMPAT) 932042d27acSHelge Deller help 933042d27acSHelge Deller This is the maximum stack size in Megabytes in the VM layout of 32-bit 934042d27acSHelge Deller user processes when the stack grows upwards (currently only on parisc 93522ee3ea5SHelge Deller arch) when the RLIMIT_STACK hard limit is unlimited. 936042d27acSHelge Deller 93722ee3ea5SHelge Deller A sane initial value is 100 MB. 9383a80a7faSMel Gorman 9393a80a7faSMel Gormanconfig DEFERRED_STRUCT_PAGE_INIT 9401ce22103SVlastimil Babka bool "Defer initialisation of struct pages to kthreads" 941d39f8fb4SMike Rapoport depends on SPARSEMEM 942ab1e8d89SPavel Tatashin depends on !NEED_PER_CPU_KM 943889c695dSPasha Tatashin depends on 64BIT 944e4443149SDaniel Jordan select PADATA 9453a80a7faSMel Gorman help 9463a80a7faSMel Gorman Ordinarily all struct pages are initialised during early boot in a 9473a80a7faSMel Gorman single thread. On very large machines this can take a considerable 9483a80a7faSMel Gorman amount of time. If this option is set, large machines will bring up 949e4443149SDaniel Jordan a subset of memmap at boot and then initialise the rest in parallel. 950e4443149SDaniel Jordan This has a potential performance impact on tasks running early in the 9511ce22103SVlastimil Babka lifetime of the system until these kthreads finish the 9521ce22103SVlastimil Babka initialisation. 953033fbae9SDan Williams 9541c676e0dSSeongJae Parkconfig PAGE_IDLE_FLAG 9551c676e0dSSeongJae Park bool 9561c676e0dSSeongJae Park select PAGE_EXTENSION if !64BIT 9571c676e0dSSeongJae Park help 9581c676e0dSSeongJae Park This adds PG_idle and PG_young flags to 'struct page'. PTE Accessed 9591c676e0dSSeongJae Park bit writers can set the state of the bit in the flags so that PTE 9601c676e0dSSeongJae Park Accessed bit readers may avoid disturbance. 9611c676e0dSSeongJae Park 96233c3fc71SVladimir Davydovconfig IDLE_PAGE_TRACKING 96333c3fc71SVladimir Davydov bool "Enable idle page tracking" 96433c3fc71SVladimir Davydov depends on SYSFS && MMU 9651c676e0dSSeongJae Park select PAGE_IDLE_FLAG 96633c3fc71SVladimir Davydov help 96733c3fc71SVladimir Davydov This feature allows to estimate the amount of user pages that have 96833c3fc71SVladimir Davydov not been touched during a given period of time. This information can 96933c3fc71SVladimir Davydov be useful to tune memory cgroup limits and/or for job placement 97033c3fc71SVladimir Davydov within a compute cluster. 97133c3fc71SVladimir Davydov 9721ad1335dSMike Rapoport See Documentation/admin-guide/mm/idle_page_tracking.rst for 9731ad1335dSMike Rapoport more details. 97433c3fc71SVladimir Davydov 9758690bbcfSMathieu Desnoyers# Architectures which implement cpu_dcache_is_aliasing() to query 9768690bbcfSMathieu Desnoyers# whether the data caches are aliased (VIVT or VIPT with dcache 9778690bbcfSMathieu Desnoyers# aliasing) need to select this. 9788690bbcfSMathieu Desnoyersconfig ARCH_HAS_CPU_CACHE_ALIASING 9798690bbcfSMathieu Desnoyers bool 9808690bbcfSMathieu Desnoyers 981c2280be8SAnshuman Khandualconfig ARCH_HAS_CACHE_LINE_SIZE 982c2280be8SAnshuman Khandual bool 983c2280be8SAnshuman Khandual 9842792d84eSKees Cookconfig ARCH_HAS_CURRENT_STACK_POINTER 9852792d84eSKees Cook bool 9862792d84eSKees Cook help 9872792d84eSKees Cook In support of HARDENED_USERCOPY performing stack variable lifetime 9882792d84eSKees Cook checking, an architecture-agnostic way to find the stack pointer 9892792d84eSKees Cook is needed. Once an architecture defines an unsigned long global 9902792d84eSKees Cook register alias named "current_stack_pointer", this config can be 9912792d84eSKees Cook selected. 9922792d84eSKees Cook 99317596731SRobin Murphyconfig ARCH_HAS_PTE_DEVMAP 99465f7d049SOliver O'Halloran bool 99565f7d049SOliver O'Halloran 99663703f37SKefeng Wangconfig ARCH_HAS_ZONE_DMA_SET 99763703f37SKefeng Wang bool 99863703f37SKefeng Wang 99963703f37SKefeng Wangconfig ZONE_DMA 100063703f37SKefeng Wang bool "Support DMA zone" if ARCH_HAS_ZONE_DMA_SET 100163703f37SKefeng Wang default y if ARM64 || X86 100263703f37SKefeng Wang 100363703f37SKefeng Wangconfig ZONE_DMA32 100463703f37SKefeng Wang bool "Support DMA32 zone" if ARCH_HAS_ZONE_DMA_SET 100563703f37SKefeng Wang depends on !X86_32 100663703f37SKefeng Wang default y if ARM64 100763703f37SKefeng Wang 1008033fbae9SDan Williamsconfig ZONE_DEVICE 10095042db43SJérôme Glisse bool "Device memory (pmem, HMM, etc...) hotplug support" 1010033fbae9SDan Williams depends on MEMORY_HOTPLUG 1011033fbae9SDan Williams depends on MEMORY_HOTREMOVE 101299490f16SDan Williams depends on SPARSEMEM_VMEMMAP 101317596731SRobin Murphy depends on ARCH_HAS_PTE_DEVMAP 10143a08cd52SMatthew Wilcox select XARRAY_MULTI 1015033fbae9SDan Williams 1016033fbae9SDan Williams help 1017033fbae9SDan Williams Device memory hotplug support allows for establishing pmem, 1018033fbae9SDan Williams or other device driver discovered memory regions, in the 1019033fbae9SDan Williams memmap. This allows pfn_to_page() lookups of otherwise 1020033fbae9SDan Williams "device-physical" addresses which is needed for using a DAX 1021033fbae9SDan Williams mapping in an O_DIRECT operation, among other things. 1022033fbae9SDan Williams 1023033fbae9SDan Williams If FS_DAX is enabled, then say Y. 102406a660adSLinus Torvalds 10259c240a7bSChristoph Hellwig# 10269c240a7bSChristoph Hellwig# Helpers to mirror range of the CPU page tables of a process into device page 10279c240a7bSChristoph Hellwig# tables. 10289c240a7bSChristoph Hellwig# 1029c0b12405SJérôme Glisseconfig HMM_MIRROR 10309c240a7bSChristoph Hellwig bool 1031f442c283SChristoph Hellwig depends on MMU 1032c0b12405SJérôme Glisse 103314b80582SDan Williamsconfig GET_FREE_REGION 103414b80582SDan Williams depends on SPARSEMEM 103514b80582SDan Williams bool 103614b80582SDan Williams 10375042db43SJérôme Glisseconfig DEVICE_PRIVATE 10385042db43SJérôme Glisse bool "Unaddressable device memory (GPU memory, ...)" 10397328d9ccSChristoph Hellwig depends on ZONE_DEVICE 104014b80582SDan Williams select GET_FREE_REGION 10415042db43SJérôme Glisse 10425042db43SJérôme Glisse help 10435042db43SJérôme Glisse Allows creation of struct pages to represent unaddressable device 10445042db43SJérôme Glisse memory; i.e., memory that is only accessible from the device (or 10455042db43SJérôme Glisse group of devices). You likely also want to select HMM_MIRROR. 10465042db43SJérôme Glisse 10473e9a9e25SChristoph Hellwigconfig VMAP_PFN 10483e9a9e25SChristoph Hellwig bool 10493e9a9e25SChristoph Hellwig 105063c17fb8SDave Hansenconfig ARCH_USES_HIGH_VMA_FLAGS 105163c17fb8SDave Hansen bool 105266d37570SDave Hansenconfig ARCH_HAS_PKEYS 105366d37570SDave Hansen bool 105430a5b536SDennis Zhou 1055b0284cd2SCatalin Marinasconfig ARCH_USES_PG_ARCH_X 1056b0284cd2SCatalin Marinas bool 1057b0284cd2SCatalin Marinas help 1058b0284cd2SCatalin Marinas Enable the definition of PG_arch_x page flags with x > 1. Only 1059b0284cd2SCatalin Marinas suitable for 64-bit architectures with CONFIG_FLATMEM or 1060b0284cd2SCatalin Marinas CONFIG_SPARSEMEM_VMEMMAP enabled, otherwise there may not be 1061b0284cd2SCatalin Marinas enough room for additional bits in page->flags. 1062b0284cd2SCatalin Marinas 10630710d012SVlastimil Babkaconfig VM_EVENT_COUNTERS 10640710d012SVlastimil Babka default y 10650710d012SVlastimil Babka bool "Enable VM event counters for /proc/vmstat" if EXPERT 10660710d012SVlastimil Babka help 10670710d012SVlastimil Babka VM event counters are needed for event counts to be shown. 10680710d012SVlastimil Babka This option allows the disabling of the VM event counters 10690710d012SVlastimil Babka on EXPERT systems. /proc/vmstat will only show page counts 10700710d012SVlastimil Babka if VM event counters are disabled. 10710710d012SVlastimil Babka 107230a5b536SDennis Zhouconfig PERCPU_STATS 107330a5b536SDennis Zhou bool "Collect percpu memory statistics" 107430a5b536SDennis Zhou help 107530a5b536SDennis Zhou This feature collects and exposes statistics via debugfs. The 107630a5b536SDennis Zhou information includes global and per chunk statistics, which can 107730a5b536SDennis Zhou be used to help understand percpu memory usage. 107864c349f4SKirill A. Shutemov 10799c84f229SJohn Hubbardconfig GUP_TEST 10809c84f229SJohn Hubbard bool "Enable infrastructure for get_user_pages()-related unit tests" 1081d0de8241SBarry Song depends on DEBUG_FS 108264c349f4SKirill A. Shutemov help 10839c84f229SJohn Hubbard Provides /sys/kernel/debug/gup_test, which in turn provides a way 10849c84f229SJohn Hubbard to make ioctl calls that can launch kernel-based unit tests for 10859c84f229SJohn Hubbard the get_user_pages*() and pin_user_pages*() family of API calls. 108664c349f4SKirill A. Shutemov 10879c84f229SJohn Hubbard These tests include benchmark testing of the _fast variants of 10889c84f229SJohn Hubbard get_user_pages*() and pin_user_pages*(), as well as smoke tests of 10899c84f229SJohn Hubbard the non-_fast variants. 10909c84f229SJohn Hubbard 1091f4f9bda4SJohn Hubbard There is also a sub-test that allows running dump_page() on any 1092f4f9bda4SJohn Hubbard of up to eight pages (selected by command line args) within the 1093f4f9bda4SJohn Hubbard range of user-space addresses. These pages are either pinned via 1094f4f9bda4SJohn Hubbard pin_user_pages*(), or pinned via get_user_pages*(), as specified 1095f4f9bda4SJohn Hubbard by other command line arguments. 1096f4f9bda4SJohn Hubbard 1097baa489faSSeongJae Park See tools/testing/selftests/mm/gup_test.c 10983010a5eaSLaurent Dufour 1099d0de8241SBarry Songcomment "GUP_TEST needs to have DEBUG_FS enabled" 1100d0de8241SBarry Song depends on !GUP_TEST && !DEBUG_FS 11013010a5eaSLaurent Dufour 11026ca297d4SPeter Zijlstraconfig GUP_GET_PXX_LOW_HIGH 110339656e83SChristoph Hellwig bool 110439656e83SChristoph Hellwig 1105def85743SKeith Buschconfig DMAPOOL_TEST 1106def85743SKeith Busch tristate "Enable a module to run time tests on dma_pool" 1107def85743SKeith Busch depends on HAS_DMA 1108def85743SKeith Busch help 1109def85743SKeith Busch Provides a test module that will allocate and free many blocks of 1110def85743SKeith Busch various sizes and report how long it takes. This is intended to 1111def85743SKeith Busch provide a consistent way to measure how changes to the 1112def85743SKeith Busch dma_pool_alloc/free routines affect performance. 1113def85743SKeith Busch 11143010a5eaSLaurent Dufourconfig ARCH_HAS_PTE_SPECIAL 11153010a5eaSLaurent Dufour bool 111659e0b520SChristoph Hellwig 1117cbd34da7SChristoph Hellwig# 1118cbd34da7SChristoph Hellwig# Some architectures require a special hugepage directory format that is 1119cbd34da7SChristoph Hellwig# required to support multiple hugepage sizes. For example a4fe3ce76 1120cbd34da7SChristoph Hellwig# "powerpc/mm: Allow more flexible layouts for hugepage pagetables" 1121cbd34da7SChristoph Hellwig# introduced it on powerpc. This allows for a more flexible hugepage 1122cbd34da7SChristoph Hellwig# pagetable layouts. 1123cbd34da7SChristoph Hellwig# 1124cbd34da7SChristoph Hellwigconfig ARCH_HAS_HUGEPD 1125cbd34da7SChristoph Hellwig bool 1126cbd34da7SChristoph Hellwig 1127c5acad84SThomas Hellstromconfig MAPPING_DIRTY_HELPERS 1128c5acad84SThomas Hellstrom bool 1129c5acad84SThomas Hellstrom 1130298fa1adSThomas Gleixnerconfig KMAP_LOCAL 1131298fa1adSThomas Gleixner bool 1132298fa1adSThomas Gleixner 1133825c43f5SArd Biesheuvelconfig KMAP_LOCAL_NON_LINEAR_PTE_ARRAY 1134825c43f5SArd Biesheuvel bool 1135825c43f5SArd Biesheuvel 11361fbaf8fcSChristoph Hellwig# struct io_mapping based helper. Selected by drivers that need them 11371fbaf8fcSChristoph Hellwigconfig IO_MAPPING 11381fbaf8fcSChristoph Hellwig bool 11391507f512SMike Rapoport 1140626e98cbSThomas Weißschuhconfig MEMFD_CREATE 1141626e98cbSThomas Weißschuh bool "Enable memfd_create() system call" if EXPERT 1142626e98cbSThomas Weißschuh 11431507f512SMike Rapoportconfig SECRETMEM 114474947724SLukas Bulwahn default y 114574947724SLukas Bulwahn bool "Enable memfd_secret() system call" if EXPERT 114674947724SLukas Bulwahn depends on ARCH_HAS_SET_DIRECT_MAP 114774947724SLukas Bulwahn help 114874947724SLukas Bulwahn Enable the memfd_secret() system call with the ability to create 114974947724SLukas Bulwahn memory areas visible only in the context of the owning process and 115074947724SLukas Bulwahn not mapped to other processes and other kernel page tables. 11511507f512SMike Rapoport 11529a10064fSColin Crossconfig ANON_VMA_NAME 11539a10064fSColin Cross bool "Anonymous VMA name support" 11549a10064fSColin Cross depends on PROC_FS && ADVISE_SYSCALLS && MMU 11559a10064fSColin Cross 11569a10064fSColin Cross help 11579a10064fSColin Cross Allow naming anonymous virtual memory areas. 11589a10064fSColin Cross 11599a10064fSColin Cross This feature allows assigning names to virtual memory areas. Assigned 11609a10064fSColin Cross names can be later retrieved from /proc/pid/maps and /proc/pid/smaps 11619a10064fSColin Cross and help identifying individual anonymous memory areas. 11629a10064fSColin Cross Assigning a name to anonymous virtual memory area might prevent that 11639a10064fSColin Cross area from being merged with adjacent virtual memory areas due to the 11649a10064fSColin Cross difference in their name. 11659a10064fSColin Cross 1166430529b5SPeter Xuconfig HAVE_ARCH_USERFAULTFD_WP 1167430529b5SPeter Xu bool 1168430529b5SPeter Xu help 1169430529b5SPeter Xu Arch has userfaultfd write protection support 1170430529b5SPeter Xu 1171430529b5SPeter Xuconfig HAVE_ARCH_USERFAULTFD_MINOR 1172430529b5SPeter Xu bool 1173430529b5SPeter Xu help 1174430529b5SPeter Xu Arch has userfaultfd minor fault support 1175430529b5SPeter Xu 117697219cc3SPeter Xumenuconfig USERFAULTFD 117797219cc3SPeter Xu bool "Enable userfaultfd() system call" 117897219cc3SPeter Xu depends on MMU 117997219cc3SPeter Xu help 118097219cc3SPeter Xu Enable the userfaultfd() system call that allows to intercept and 118197219cc3SPeter Xu handle page faults in userland. 118297219cc3SPeter Xu 118397219cc3SPeter Xuif USERFAULTFD 11841db9dbc2SPeter Xuconfig PTE_MARKER_UFFD_WP 118581e0f15fSPeter Xu bool "Userfaultfd write protection support for shmem/hugetlbfs" 118681e0f15fSPeter Xu default y 118781e0f15fSPeter Xu depends on HAVE_ARCH_USERFAULTFD_WP 11881db9dbc2SPeter Xu 11891db9dbc2SPeter Xu help 11901db9dbc2SPeter Xu Allows to create marker PTEs for userfaultfd write protection 11911db9dbc2SPeter Xu purposes. It is required to enable userfaultfd write protection on 11921db9dbc2SPeter Xu file-backed memory types like shmem and hugetlbfs. 119397219cc3SPeter Xuendif # USERFAULTFD 11941db9dbc2SPeter Xu 1195ac35a490SYu Zhao# multi-gen LRU { 1196ec1c86b2SYu Zhaoconfig LRU_GEN 1197ec1c86b2SYu Zhao bool "Multi-Gen LRU" 1198ec1c86b2SYu Zhao depends on MMU 1199ec1c86b2SYu Zhao # make sure folio->flags has enough spare bits 1200ec1c86b2SYu Zhao depends on 64BIT || !SPARSEMEM || SPARSEMEM_VMEMMAP 1201ec1c86b2SYu Zhao help 120207017acbSYu Zhao A high performance LRU implementation to overcommit memory. See 120307017acbSYu Zhao Documentation/admin-guide/mm/multigen_lru.rst for details. 1204ec1c86b2SYu Zhao 1205354ed597SYu Zhaoconfig LRU_GEN_ENABLED 1206354ed597SYu Zhao bool "Enable by default" 1207354ed597SYu Zhao depends on LRU_GEN 1208354ed597SYu Zhao help 1209354ed597SYu Zhao This option enables the multi-gen LRU by default. 1210354ed597SYu Zhao 1211ac35a490SYu Zhaoconfig LRU_GEN_STATS 1212ac35a490SYu Zhao bool "Full stats for debugging" 1213ac35a490SYu Zhao depends on LRU_GEN 1214ac35a490SYu Zhao help 1215ac35a490SYu Zhao Do not enable this option unless you plan to look at historical stats 1216ac35a490SYu Zhao from evicted generations for debugging purpose. 1217ac35a490SYu Zhao 1218ac35a490SYu Zhao This option has a per-memcg and per-node memory overhead. 121961dd3f24SKinsey Ho 122061dd3f24SKinsey Hoconfig LRU_GEN_WALKS_MMU 122161dd3f24SKinsey Ho def_bool y 122261dd3f24SKinsey Ho depends on LRU_GEN && ARCH_HAS_HW_PTE_YOUNG 1223ac35a490SYu Zhao# } 1224ac35a490SYu Zhao 12250b6cc04fSSuren Baghdasaryanconfig ARCH_SUPPORTS_PER_VMA_LOCK 12260b6cc04fSSuren Baghdasaryan def_bool n 12270b6cc04fSSuren Baghdasaryan 12280b6cc04fSSuren Baghdasaryanconfig PER_VMA_LOCK 12290b6cc04fSSuren Baghdasaryan def_bool y 12300b6cc04fSSuren Baghdasaryan depends on ARCH_SUPPORTS_PER_VMA_LOCK && MMU && SMP 12310b6cc04fSSuren Baghdasaryan help 12320b6cc04fSSuren Baghdasaryan Allow per-vma locking during page fault handling. 12330b6cc04fSSuren Baghdasaryan 12340b6cc04fSSuren Baghdasaryan This feature allows locking each virtual memory area separately when 12350b6cc04fSSuren Baghdasaryan handling page faults instead of taking mmap_lock. 12360b6cc04fSSuren Baghdasaryan 1237c2508ec5SLinus Torvaldsconfig LOCK_MM_AND_FIND_VMA 1238c2508ec5SLinus Torvalds bool 1239c2508ec5SLinus Torvalds depends on !STACK_GROWSUP 1240c2508ec5SLinus Torvalds 12418f23f5dbSJason Gunthorpeconfig IOMMU_MM_DATA 12428f23f5dbSJason Gunthorpe bool 12438f23f5dbSJason Gunthorpe 12442224d848SSeongJae Parksource "mm/damon/Kconfig" 12452224d848SSeongJae Park 124659e0b520SChristoph Hellwigendmenu 1247