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 49519bcb79SJohannes Weinerchoice 50b3fbd58fSJohannes Weiner prompt "Default compressor" 51519bcb79SJohannes Weiner depends on ZSWAP 52519bcb79SJohannes Weiner default ZSWAP_COMPRESSOR_DEFAULT_LZO 53519bcb79SJohannes Weiner help 54519bcb79SJohannes Weiner Selects the default compression algorithm for the compressed cache 55519bcb79SJohannes Weiner for swap pages. 56519bcb79SJohannes Weiner 57519bcb79SJohannes Weiner For an overview what kind of performance can be expected from 58519bcb79SJohannes Weiner a particular compression algorithm please refer to the benchmarks 59519bcb79SJohannes Weiner available at the following LWN page: 60519bcb79SJohannes Weiner https://lwn.net/Articles/751795/ 61519bcb79SJohannes Weiner 62519bcb79SJohannes Weiner If in doubt, select 'LZO'. 63519bcb79SJohannes Weiner 64519bcb79SJohannes Weiner The selection made here can be overridden by using the kernel 65519bcb79SJohannes Weiner command line 'zswap.compressor=' option. 66519bcb79SJohannes Weiner 67519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_DEFLATE 68519bcb79SJohannes Weiner bool "Deflate" 69519bcb79SJohannes Weiner select CRYPTO_DEFLATE 70519bcb79SJohannes Weiner help 71519bcb79SJohannes Weiner Use the Deflate algorithm as the default compression algorithm. 72519bcb79SJohannes Weiner 73519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_LZO 74519bcb79SJohannes Weiner bool "LZO" 75519bcb79SJohannes Weiner select CRYPTO_LZO 76519bcb79SJohannes Weiner help 77519bcb79SJohannes Weiner Use the LZO algorithm as the default compression algorithm. 78519bcb79SJohannes Weiner 79519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_842 80519bcb79SJohannes Weiner bool "842" 81519bcb79SJohannes Weiner select CRYPTO_842 82519bcb79SJohannes Weiner help 83519bcb79SJohannes Weiner Use the 842 algorithm as the default compression algorithm. 84519bcb79SJohannes Weiner 85519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_LZ4 86519bcb79SJohannes Weiner bool "LZ4" 87519bcb79SJohannes Weiner select CRYPTO_LZ4 88519bcb79SJohannes Weiner help 89519bcb79SJohannes Weiner Use the LZ4 algorithm as the default compression algorithm. 90519bcb79SJohannes Weiner 91519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_LZ4HC 92519bcb79SJohannes Weiner bool "LZ4HC" 93519bcb79SJohannes Weiner select CRYPTO_LZ4HC 94519bcb79SJohannes Weiner help 95519bcb79SJohannes Weiner Use the LZ4HC algorithm as the default compression algorithm. 96519bcb79SJohannes Weiner 97519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT_ZSTD 98519bcb79SJohannes Weiner bool "zstd" 99519bcb79SJohannes Weiner select CRYPTO_ZSTD 100519bcb79SJohannes Weiner help 101519bcb79SJohannes Weiner Use the zstd algorithm as the default compression algorithm. 102519bcb79SJohannes Weinerendchoice 103519bcb79SJohannes Weiner 104519bcb79SJohannes Weinerconfig ZSWAP_COMPRESSOR_DEFAULT 105519bcb79SJohannes Weiner string 106519bcb79SJohannes Weiner depends on ZSWAP 107519bcb79SJohannes Weiner default "deflate" if ZSWAP_COMPRESSOR_DEFAULT_DEFLATE 108519bcb79SJohannes Weiner default "lzo" if ZSWAP_COMPRESSOR_DEFAULT_LZO 109519bcb79SJohannes Weiner default "842" if ZSWAP_COMPRESSOR_DEFAULT_842 110519bcb79SJohannes Weiner default "lz4" if ZSWAP_COMPRESSOR_DEFAULT_LZ4 111519bcb79SJohannes Weiner default "lz4hc" if ZSWAP_COMPRESSOR_DEFAULT_LZ4HC 112519bcb79SJohannes Weiner default "zstd" if ZSWAP_COMPRESSOR_DEFAULT_ZSTD 113519bcb79SJohannes Weiner default "" 114519bcb79SJohannes Weiner 115519bcb79SJohannes Weinerchoice 116b3fbd58fSJohannes Weiner prompt "Default allocator" 117519bcb79SJohannes Weiner depends on ZSWAP 118519bcb79SJohannes Weiner default ZSWAP_ZPOOL_DEFAULT_ZBUD 119519bcb79SJohannes Weiner help 120519bcb79SJohannes Weiner Selects the default allocator for the compressed cache for 121519bcb79SJohannes Weiner swap pages. 122519bcb79SJohannes Weiner The default is 'zbud' for compatibility, however please do 123519bcb79SJohannes Weiner read the description of each of the allocators below before 124519bcb79SJohannes Weiner making a right choice. 125519bcb79SJohannes Weiner 126519bcb79SJohannes Weiner The selection made here can be overridden by using the kernel 127519bcb79SJohannes Weiner command line 'zswap.zpool=' option. 128519bcb79SJohannes Weiner 129519bcb79SJohannes Weinerconfig ZSWAP_ZPOOL_DEFAULT_ZBUD 130519bcb79SJohannes Weiner bool "zbud" 131519bcb79SJohannes Weiner select ZBUD 132519bcb79SJohannes Weiner help 133519bcb79SJohannes Weiner Use the zbud allocator as the default allocator. 134519bcb79SJohannes Weiner 135519bcb79SJohannes Weinerconfig ZSWAP_ZPOOL_DEFAULT_Z3FOLD 136519bcb79SJohannes Weiner bool "z3fold" 137519bcb79SJohannes Weiner select Z3FOLD 138519bcb79SJohannes Weiner help 139519bcb79SJohannes Weiner Use the z3fold allocator as the default allocator. 140519bcb79SJohannes Weiner 141519bcb79SJohannes Weinerconfig ZSWAP_ZPOOL_DEFAULT_ZSMALLOC 142519bcb79SJohannes Weiner bool "zsmalloc" 143519bcb79SJohannes Weiner select ZSMALLOC 144519bcb79SJohannes Weiner help 145519bcb79SJohannes Weiner Use the zsmalloc allocator as the default allocator. 146519bcb79SJohannes Weinerendchoice 147519bcb79SJohannes Weiner 148519bcb79SJohannes Weinerconfig ZSWAP_ZPOOL_DEFAULT 149519bcb79SJohannes Weiner string 150519bcb79SJohannes Weiner depends on ZSWAP 151519bcb79SJohannes Weiner default "zbud" if ZSWAP_ZPOOL_DEFAULT_ZBUD 152519bcb79SJohannes Weiner default "z3fold" if ZSWAP_ZPOOL_DEFAULT_Z3FOLD 153519bcb79SJohannes Weiner default "zsmalloc" if ZSWAP_ZPOOL_DEFAULT_ZSMALLOC 154519bcb79SJohannes Weiner default "" 155519bcb79SJohannes Weiner 156519bcb79SJohannes Weinerconfig ZBUD 157b3fbd58fSJohannes Weiner tristate "2:1 compression allocator (zbud)" 158b3fbd58fSJohannes Weiner depends on ZSWAP 159519bcb79SJohannes Weiner help 160519bcb79SJohannes Weiner A special purpose allocator for storing compressed pages. 161519bcb79SJohannes Weiner It is designed to store up to two compressed pages per physical 162519bcb79SJohannes Weiner page. While this design limits storage density, it has simple and 163519bcb79SJohannes Weiner deterministic reclaim properties that make it preferable to a higher 164519bcb79SJohannes Weiner density approach when reclaim will be used. 165519bcb79SJohannes Weiner 166519bcb79SJohannes Weinerconfig Z3FOLD 167b3fbd58fSJohannes Weiner tristate "3:1 compression allocator (z3fold)" 168b3fbd58fSJohannes Weiner depends on ZSWAP 169519bcb79SJohannes Weiner help 170519bcb79SJohannes Weiner A special purpose allocator for storing compressed pages. 171519bcb79SJohannes Weiner It is designed to store up to three compressed pages per physical 172519bcb79SJohannes Weiner page. It is a ZBUD derivative so the simplicity and determinism are 173519bcb79SJohannes Weiner still there. 174519bcb79SJohannes Weiner 175519bcb79SJohannes Weinerconfig ZSMALLOC 176b3fbd58fSJohannes Weiner tristate 177b3fbd58fSJohannes Weiner prompt "N:1 compression allocator (zsmalloc)" if ZSWAP 178519bcb79SJohannes Weiner depends on MMU 179519bcb79SJohannes Weiner help 180519bcb79SJohannes Weiner zsmalloc is a slab-based memory allocator designed to store 181b3fbd58fSJohannes Weiner pages of various compression levels efficiently. It achieves 182b3fbd58fSJohannes Weiner the highest storage density with the least amount of fragmentation. 183519bcb79SJohannes Weiner 184519bcb79SJohannes Weinerconfig ZSMALLOC_STAT 185519bcb79SJohannes Weiner bool "Export zsmalloc statistics" 186519bcb79SJohannes Weiner depends on ZSMALLOC 187519bcb79SJohannes Weiner select DEBUG_FS 188519bcb79SJohannes Weiner help 189519bcb79SJohannes Weiner This option enables code in the zsmalloc to collect various 190519bcb79SJohannes Weiner statistics about what's happening in zsmalloc and exports that 191519bcb79SJohannes Weiner information to userspace via debugfs. 192519bcb79SJohannes Weiner If unsure, say N. 193519bcb79SJohannes Weiner 1944ff93b29SSergey Senozhatskyconfig ZSMALLOC_CHAIN_SIZE 1954ff93b29SSergey Senozhatsky int "Maximum number of physical pages per-zspage" 196b46402faSSergey Senozhatsky default 8 1974ff93b29SSergey Senozhatsky range 4 16 1984ff93b29SSergey Senozhatsky depends on ZSMALLOC 1994ff93b29SSergey Senozhatsky help 2004ff93b29SSergey Senozhatsky This option sets the upper limit on the number of physical pages 2014ff93b29SSergey Senozhatsky that a zmalloc page (zspage) can consist of. The optimal zspage 2024ff93b29SSergey Senozhatsky chain size is calculated for each size class during the 2034ff93b29SSergey Senozhatsky initialization of the pool. 2044ff93b29SSergey Senozhatsky 2054ff93b29SSergey Senozhatsky Changing this option can alter the characteristics of size classes, 2064ff93b29SSergey Senozhatsky such as the number of pages per zspage and the number of objects 2074ff93b29SSergey Senozhatsky per zspage. This can also result in different configurations of 2084ff93b29SSergey Senozhatsky the pool, as zsmalloc merges size classes with similar 2094ff93b29SSergey Senozhatsky characteristics. 2104ff93b29SSergey Senozhatsky 2114ff93b29SSergey Senozhatsky For more information, see zsmalloc documentation. 2124ff93b29SSergey Senozhatsky 213519bcb79SJohannes Weinermenu "SLAB allocator options" 214519bcb79SJohannes Weiner 2157b42f104SJohannes Weinerchoice 2167b42f104SJohannes Weiner prompt "Choose SLAB allocator" 2177b42f104SJohannes Weiner default SLUB 2187b42f104SJohannes Weiner help 2197b42f104SJohannes Weiner This option allows to select a slab allocator. 2207b42f104SJohannes Weiner 221*eb07c4f3SVlastimil Babkaconfig SLAB_DEPRECATED 222*eb07c4f3SVlastimil Babka bool "SLAB (DEPRECATED)" 2237b42f104SJohannes Weiner depends on !PREEMPT_RT 2247b42f104SJohannes Weiner select HAVE_HARDENED_USERCOPY_ALLOCATOR 2257b42f104SJohannes Weiner help 226*eb07c4f3SVlastimil Babka Deprecated and scheduled for removal in a few cycles. Replaced by 227*eb07c4f3SVlastimil Babka SLUB. 228*eb07c4f3SVlastimil Babka 229*eb07c4f3SVlastimil Babka If you cannot migrate to SLUB, please contact linux-mm@kvack.org 230*eb07c4f3SVlastimil Babka and the people listed in the SLAB ALLOCATOR section of MAINTAINERS 231*eb07c4f3SVlastimil Babka file, explaining why. 232*eb07c4f3SVlastimil Babka 2337b42f104SJohannes Weiner The regular slab allocator that is established and known to work 2347b42f104SJohannes Weiner well in all environments. It organizes cache hot objects in 2357b42f104SJohannes Weiner per cpu and per node queues. 2367b42f104SJohannes Weiner 2377b42f104SJohannes Weinerconfig SLUB 2387b42f104SJohannes Weiner bool "SLUB (Unqueued Allocator)" 2397b42f104SJohannes Weiner select HAVE_HARDENED_USERCOPY_ALLOCATOR 2407b42f104SJohannes Weiner help 2417b42f104SJohannes Weiner SLUB is a slab allocator that minimizes cache line usage 2427b42f104SJohannes Weiner instead of managing queues of cached objects (SLAB approach). 2437b42f104SJohannes Weiner Per cpu caching is realized using slabs of objects instead 2447b42f104SJohannes Weiner of queues of objects. SLUB can use memory efficiently 2457b42f104SJohannes Weiner and has enhanced diagnostics. SLUB is the default choice for 2467b42f104SJohannes Weiner a slab allocator. 2477b42f104SJohannes Weiner 2487b42f104SJohannes Weinerendchoice 2497b42f104SJohannes Weiner 250*eb07c4f3SVlastimil Babkaconfig SLAB 251*eb07c4f3SVlastimil Babka bool 252*eb07c4f3SVlastimil Babka default y 253*eb07c4f3SVlastimil Babka depends on SLAB_DEPRECATED 254*eb07c4f3SVlastimil Babka 255e240e53aSVlastimil Babkaconfig SLUB_TINY 256e240e53aSVlastimil Babka bool "Configure SLUB for minimal memory footprint" 257e240e53aSVlastimil Babka depends on SLUB && EXPERT 258e240e53aSVlastimil Babka select SLAB_MERGE_DEFAULT 259e240e53aSVlastimil Babka help 260e240e53aSVlastimil Babka Configures the SLUB allocator in a way to achieve minimal memory 261e240e53aSVlastimil Babka footprint, sacrificing scalability, debugging and other features. 262e240e53aSVlastimil Babka This is intended only for the smallest system that had used the 263e240e53aSVlastimil Babka SLOB allocator and is not recommended for systems with more than 264e240e53aSVlastimil Babka 16MB RAM. 265e240e53aSVlastimil Babka 266e240e53aSVlastimil Babka If unsure, say N. 267e240e53aSVlastimil Babka 2687b42f104SJohannes Weinerconfig SLAB_MERGE_DEFAULT 2697b42f104SJohannes Weiner bool "Allow slab caches to be merged" 2707b42f104SJohannes Weiner default y 2717b42f104SJohannes Weiner depends on SLAB || SLUB 2727b42f104SJohannes Weiner help 2737b42f104SJohannes Weiner For reduced kernel memory fragmentation, slab caches can be 2747b42f104SJohannes Weiner merged when they share the same size and other characteristics. 2757b42f104SJohannes Weiner This carries a risk of kernel heap overflows being able to 2767b42f104SJohannes Weiner overwrite objects from merged caches (and more easily control 2777b42f104SJohannes Weiner cache layout), which makes such heap attacks easier to exploit 2787b42f104SJohannes Weiner by attackers. By keeping caches unmerged, these kinds of exploits 2797b42f104SJohannes Weiner can usually only damage objects in the same cache. To disable 2807b42f104SJohannes Weiner merging at runtime, "slab_nomerge" can be passed on the kernel 2817b42f104SJohannes Weiner command line. 2827b42f104SJohannes Weiner 2837b42f104SJohannes Weinerconfig SLAB_FREELIST_RANDOM 2847b42f104SJohannes Weiner bool "Randomize slab freelist" 285e240e53aSVlastimil Babka depends on SLAB || (SLUB && !SLUB_TINY) 2867b42f104SJohannes Weiner help 2877b42f104SJohannes Weiner Randomizes the freelist order used on creating new pages. This 2887b42f104SJohannes Weiner security feature reduces the predictability of the kernel slab 2897b42f104SJohannes Weiner allocator against heap overflows. 2907b42f104SJohannes Weiner 2917b42f104SJohannes Weinerconfig SLAB_FREELIST_HARDENED 2927b42f104SJohannes Weiner bool "Harden slab freelist metadata" 293e240e53aSVlastimil Babka depends on SLAB || (SLUB && !SLUB_TINY) 2947b42f104SJohannes Weiner help 2957b42f104SJohannes Weiner Many kernel heap attacks try to target slab cache metadata and 2967b42f104SJohannes Weiner other infrastructure. This options makes minor performance 2977b42f104SJohannes Weiner sacrifices to harden the kernel slab allocator against common 2987b42f104SJohannes Weiner freelist exploit methods. Some slab implementations have more 2997b42f104SJohannes Weiner sanity-checking than others. This option is most effective with 3007b42f104SJohannes Weiner CONFIG_SLUB. 3017b42f104SJohannes Weiner 3020710d012SVlastimil Babkaconfig SLUB_STATS 3030710d012SVlastimil Babka default n 3040710d012SVlastimil Babka bool "Enable SLUB performance statistics" 305e240e53aSVlastimil Babka depends on SLUB && SYSFS && !SLUB_TINY 3060710d012SVlastimil Babka help 3070710d012SVlastimil Babka SLUB statistics are useful to debug SLUBs allocation behavior in 3080710d012SVlastimil Babka order find ways to optimize the allocator. This should never be 3090710d012SVlastimil Babka enabled for production use since keeping statistics slows down 3100710d012SVlastimil Babka the allocator by a few percentage points. The slabinfo command 3110710d012SVlastimil Babka supports the determination of the most active slabs to figure 3120710d012SVlastimil Babka out which slabs are relevant to a particular load. 3130710d012SVlastimil Babka Try running: slabinfo -DA 3140710d012SVlastimil Babka 315519bcb79SJohannes Weinerconfig SLUB_CPU_PARTIAL 316519bcb79SJohannes Weiner default y 317e240e53aSVlastimil Babka depends on SLUB && SMP && !SLUB_TINY 318519bcb79SJohannes Weiner bool "SLUB per cpu partial cache" 319519bcb79SJohannes Weiner help 320519bcb79SJohannes Weiner Per cpu partial caches accelerate objects allocation and freeing 321519bcb79SJohannes Weiner that is local to a processor at the price of more indeterminism 322519bcb79SJohannes Weiner in the latency of the free. On overflow these caches will be cleared 323519bcb79SJohannes Weiner which requires the taking of locks that may cause latency spikes. 324519bcb79SJohannes Weiner Typically one would choose no for a realtime system. 325519bcb79SJohannes Weiner 326519bcb79SJohannes Weinerendmenu # SLAB allocator options 327519bcb79SJohannes Weiner 3287b42f104SJohannes Weinerconfig SHUFFLE_PAGE_ALLOCATOR 3297b42f104SJohannes Weiner bool "Page allocator randomization" 3307b42f104SJohannes Weiner default SLAB_FREELIST_RANDOM && ACPI_NUMA 3317b42f104SJohannes Weiner help 3327b42f104SJohannes Weiner Randomization of the page allocator improves the average 3337b42f104SJohannes Weiner utilization of a direct-mapped memory-side-cache. See section 3347b42f104SJohannes Weiner 5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI 3357b42f104SJohannes Weiner 6.2a specification for an example of how a platform advertises 3367b42f104SJohannes Weiner the presence of a memory-side-cache. There are also incidental 3377b42f104SJohannes Weiner security benefits as it reduces the predictability of page 3387b42f104SJohannes Weiner allocations to compliment SLAB_FREELIST_RANDOM, but the 33923baf831SKirill A. Shutemov default granularity of shuffling on the MAX_ORDER i.e, 10th 34023baf831SKirill A. Shutemov order of pages is selected based on cache utilization benefits 34123baf831SKirill A. Shutemov on x86. 3427b42f104SJohannes Weiner 3437b42f104SJohannes Weiner While the randomization improves cache utilization it may 3447b42f104SJohannes Weiner negatively impact workloads on platforms without a cache. For 3457b42f104SJohannes Weiner this reason, by default, the randomization is enabled only 3467b42f104SJohannes Weiner after runtime detection of a direct-mapped memory-side-cache. 3477b42f104SJohannes Weiner Otherwise, the randomization may be force enabled with the 3487b42f104SJohannes Weiner 'page_alloc.shuffle' kernel command line parameter. 3497b42f104SJohannes Weiner 3507b42f104SJohannes Weiner Say Y if unsure. 3517b42f104SJohannes Weiner 3520710d012SVlastimil Babkaconfig COMPAT_BRK 3530710d012SVlastimil Babka bool "Disable heap randomization" 3540710d012SVlastimil Babka default y 3550710d012SVlastimil Babka help 3560710d012SVlastimil Babka Randomizing heap placement makes heap exploits harder, but it 3570710d012SVlastimil Babka also breaks ancient binaries (including anything libc5 based). 3580710d012SVlastimil Babka This option changes the bootup default to heap randomization 3590710d012SVlastimil Babka disabled, and can be overridden at runtime by setting 3600710d012SVlastimil Babka /proc/sys/kernel/randomize_va_space to 2. 3610710d012SVlastimil Babka 3620710d012SVlastimil Babka On non-ancient distros (post-2000 ones) N is usually a safe choice. 3630710d012SVlastimil Babka 3640710d012SVlastimil Babkaconfig MMAP_ALLOW_UNINITIALIZED 3650710d012SVlastimil Babka bool "Allow mmapped anonymous memory to be uninitialized" 3660710d012SVlastimil Babka depends on EXPERT && !MMU 3670710d012SVlastimil Babka default n 3680710d012SVlastimil Babka help 3690710d012SVlastimil Babka Normally, and according to the Linux spec, anonymous memory obtained 3700710d012SVlastimil Babka from mmap() has its contents cleared before it is passed to 3710710d012SVlastimil Babka userspace. Enabling this config option allows you to request that 3720710d012SVlastimil Babka mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus 3730710d012SVlastimil Babka providing a huge performance boost. If this option is not enabled, 3740710d012SVlastimil Babka then the flag will be ignored. 3750710d012SVlastimil Babka 3760710d012SVlastimil Babka This is taken advantage of by uClibc's malloc(), and also by 3770710d012SVlastimil Babka ELF-FDPIC binfmt's brk and stack allocator. 3780710d012SVlastimil Babka 3790710d012SVlastimil Babka Because of the obvious security issues, this option should only be 3800710d012SVlastimil Babka enabled on embedded devices where you control what is run in 3810710d012SVlastimil Babka userspace. Since that isn't generally a problem on no-MMU systems, 3820710d012SVlastimil Babka it is normally safe to say Y here. 3830710d012SVlastimil Babka 3840710d012SVlastimil Babka See Documentation/admin-guide/mm/nommu-mmap.rst for more information. 3850710d012SVlastimil Babka 386e1785e85SDave Hansenconfig SELECT_MEMORY_MODEL 387e1785e85SDave Hansen def_bool y 388a8826eebSKees Cook depends on ARCH_SELECT_MEMORY_MODEL 389e1785e85SDave Hansen 3903a9da765SDave Hansenchoice 3913a9da765SDave Hansen prompt "Memory model" 392e1785e85SDave Hansen depends on SELECT_MEMORY_MODEL 393d41dee36SAndy Whitcroft default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT 394e1785e85SDave Hansen default FLATMEM_MANUAL 395d66d109dSMike Rapoport help 396d66d109dSMike Rapoport This option allows you to change some of the ways that 397d66d109dSMike Rapoport Linux manages its memory internally. Most users will 398d66d109dSMike Rapoport only have one option here selected by the architecture 399d66d109dSMike Rapoport configuration. This is normal. 4003a9da765SDave Hansen 401e1785e85SDave Hansenconfig FLATMEM_MANUAL 4023a9da765SDave Hansen bool "Flat Memory" 403bb1c50d3SMike Rapoport depends on !ARCH_SPARSEMEM_ENABLE || ARCH_FLATMEM_ENABLE 4043a9da765SDave Hansen help 405d66d109dSMike Rapoport This option is best suited for non-NUMA systems with 406d66d109dSMike Rapoport flat address space. The FLATMEM is the most efficient 407d66d109dSMike Rapoport system in terms of performance and resource consumption 408d66d109dSMike Rapoport and it is the best option for smaller systems. 4093a9da765SDave Hansen 410d66d109dSMike Rapoport For systems that have holes in their physical address 411d66d109dSMike Rapoport spaces and for features like NUMA and memory hotplug, 412dd33d29aSRandy Dunlap choose "Sparse Memory". 413d41dee36SAndy Whitcroft 414d41dee36SAndy Whitcroft If unsure, choose this option (Flat Memory) over any other. 4153a9da765SDave Hansen 416d41dee36SAndy Whitcroftconfig SPARSEMEM_MANUAL 417d41dee36SAndy Whitcroft bool "Sparse Memory" 418d41dee36SAndy Whitcroft depends on ARCH_SPARSEMEM_ENABLE 419d41dee36SAndy Whitcroft help 420d41dee36SAndy Whitcroft This will be the only option for some systems, including 421d66d109dSMike Rapoport memory hot-plug systems. This is normal. 422d41dee36SAndy Whitcroft 423d66d109dSMike Rapoport This option provides efficient support for systems with 424d66d109dSMike Rapoport holes is their physical address space and allows memory 425d66d109dSMike Rapoport hot-plug and hot-remove. 426d41dee36SAndy Whitcroft 427d66d109dSMike Rapoport If unsure, choose "Flat Memory" over this option. 428d41dee36SAndy Whitcroft 4293a9da765SDave Hansenendchoice 4303a9da765SDave Hansen 431d41dee36SAndy Whitcroftconfig SPARSEMEM 432d41dee36SAndy Whitcroft def_bool y 4331a83e175SRussell King depends on (!SELECT_MEMORY_MODEL && ARCH_SPARSEMEM_ENABLE) || SPARSEMEM_MANUAL 434d41dee36SAndy Whitcroft 435e1785e85SDave Hansenconfig FLATMEM 436e1785e85SDave Hansen def_bool y 437bb1c50d3SMike Rapoport depends on !SPARSEMEM || FLATMEM_MANUAL 438d41dee36SAndy Whitcroft 43993b7504eSDave Hansen# 4403e347261SBob Picco# SPARSEMEM_EXTREME (which is the default) does some bootmem 441c89ab04fSMike Rapoport# allocations when sparse_init() is called. If this cannot 4423e347261SBob Picco# be done on your architecture, select this option. However, 4433e347261SBob Picco# statically allocating the mem_section[] array can potentially 4443e347261SBob Picco# consume vast quantities of .bss, so be careful. 4453e347261SBob Picco# 4463e347261SBob Picco# This option will also potentially produce smaller runtime code 4473e347261SBob Picco# with gcc 3.4 and later. 4483e347261SBob Picco# 4493e347261SBob Piccoconfig SPARSEMEM_STATIC 4509ba16087SJan Beulich bool 4513e347261SBob Picco 4523e347261SBob Picco# 45344c09201SMatt LaPlante# Architecture platforms which require a two level mem_section in SPARSEMEM 454802f192eSBob Picco# must select this option. This is usually for architecture platforms with 455802f192eSBob Picco# an extremely sparse physical address space. 456802f192eSBob Picco# 4573e347261SBob Piccoconfig SPARSEMEM_EXTREME 4583e347261SBob Picco def_bool y 4593e347261SBob Picco depends on SPARSEMEM && !SPARSEMEM_STATIC 4604c21e2f2SHugh Dickins 46129c71111SAndy Whitcroftconfig SPARSEMEM_VMEMMAP_ENABLE 4629ba16087SJan Beulich bool 46329c71111SAndy Whitcroft 46429c71111SAndy Whitcroftconfig SPARSEMEM_VMEMMAP 465a5ee6daaSGeoff Levand bool "Sparse Memory virtual memmap" 466a5ee6daaSGeoff Levand depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE 467a5ee6daaSGeoff Levand default y 468a5ee6daaSGeoff Levand help 469a5ee6daaSGeoff Levand SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise 470a5ee6daaSGeoff Levand pfn_to_page and page_to_pfn operations. This is the most 471a5ee6daaSGeoff Levand efficient option when sufficient kernel resources are available. 4720b376f1eSAneesh Kumar K.V# 4730b376f1eSAneesh Kumar K.V# Select this config option from the architecture Kconfig, if it is preferred 4740b376f1eSAneesh Kumar K.V# to enable the feature of HugeTLB/dev_dax vmemmap optimization. 4750b376f1eSAneesh Kumar K.V# 4760b376f1eSAneesh Kumar K.Vconfig ARCH_WANT_OPTIMIZE_VMEMMAP 4770b376f1eSAneesh Kumar K.V bool 47829c71111SAndy Whitcroft 47970210ed9SPhilipp Hachtmannconfig HAVE_MEMBLOCK_PHYS_MAP 4806341e62bSChristoph Jaeger bool 48170210ed9SPhilipp Hachtmann 48267a929e0SChristoph Hellwigconfig HAVE_FAST_GUP 483050a9adcSChristoph Hellwig depends on MMU 4846341e62bSChristoph Jaeger bool 4852667f50eSSteve Capper 48652219aeaSDavid Hildenbrand# Don't discard allocated memory used to track "memory" and "reserved" memblocks 48752219aeaSDavid Hildenbrand# after early boot, so it can still be used to test for validity of memory. 48852219aeaSDavid Hildenbrand# Also, memblocks are updated with memory hot(un)plug. 489350e88baSMike Rapoportconfig ARCH_KEEP_MEMBLOCK 4906341e62bSChristoph Jaeger bool 491c378ddd5STejun Heo 4921e5d8e1eSDan Williams# Keep arch NUMA mapping infrastructure post-init. 4931e5d8e1eSDan Williamsconfig NUMA_KEEP_MEMINFO 4941e5d8e1eSDan Williams bool 4951e5d8e1eSDan Williams 496ee6f509cSMinchan Kimconfig MEMORY_ISOLATION 4976341e62bSChristoph Jaeger bool 498ee6f509cSMinchan Kim 499a9e7b8d4SDavid Hildenbrand# IORESOURCE_SYSTEM_RAM regions in the kernel resource tree that are marked 500a9e7b8d4SDavid Hildenbrand# IORESOURCE_EXCLUSIVE cannot be mapped to user space, for example, via 501a9e7b8d4SDavid Hildenbrand# /dev/mem. 502a9e7b8d4SDavid Hildenbrandconfig EXCLUSIVE_SYSTEM_RAM 503a9e7b8d4SDavid Hildenbrand def_bool y 504a9e7b8d4SDavid Hildenbrand depends on !DEVMEM || STRICT_DEVMEM 505a9e7b8d4SDavid Hildenbrand 50646723bfaSYasuaki Ishimatsu# 50746723bfaSYasuaki Ishimatsu# Only be set on architectures that have completely implemented memory hotplug 50846723bfaSYasuaki Ishimatsu# feature. If you are not sure, don't touch it. 50946723bfaSYasuaki Ishimatsu# 51046723bfaSYasuaki Ishimatsuconfig HAVE_BOOTMEM_INFO_NODE 51146723bfaSYasuaki Ishimatsu def_bool n 51246723bfaSYasuaki Ishimatsu 51391024b3cSAnshuman Khandualconfig ARCH_ENABLE_MEMORY_HOTPLUG 51491024b3cSAnshuman Khandual bool 51591024b3cSAnshuman Khandual 516519bcb79SJohannes Weinerconfig ARCH_ENABLE_MEMORY_HOTREMOVE 517519bcb79SJohannes Weiner bool 518519bcb79SJohannes Weiner 5193947be19SDave Hansen# eventually, we can have this option just 'select SPARSEMEM' 520519bcb79SJohannes Weinermenuconfig MEMORY_HOTPLUG 521519bcb79SJohannes Weiner bool "Memory hotplug" 522b30c5927SDavid Hildenbrand select MEMORY_ISOLATION 52371b6f2ddSDavid Hildenbrand depends on SPARSEMEM 52440b31360SStephen Rothwell depends on ARCH_ENABLE_MEMORY_HOTPLUG 5257ec58a2bSDavid Hildenbrand depends on 64BIT 5261e5d8e1eSDan Williams select NUMA_KEEP_MEMINFO if NUMA 5273947be19SDave Hansen 528519bcb79SJohannes Weinerif MEMORY_HOTPLUG 529519bcb79SJohannes Weiner 5308604d9e5SVitaly Kuznetsovconfig MEMORY_HOTPLUG_DEFAULT_ONLINE 5318604d9e5SVitaly Kuznetsov bool "Online the newly added memory blocks by default" 5328604d9e5SVitaly Kuznetsov depends on MEMORY_HOTPLUG 5338604d9e5SVitaly Kuznetsov help 5348604d9e5SVitaly Kuznetsov This option sets the default policy setting for memory hotplug 5358604d9e5SVitaly Kuznetsov onlining policy (/sys/devices/system/memory/auto_online_blocks) which 5368604d9e5SVitaly Kuznetsov determines what happens to newly added memory regions. Policy setting 5378604d9e5SVitaly Kuznetsov can always be changed at runtime. 538cb1aaebeSMauro Carvalho Chehab See Documentation/admin-guide/mm/memory-hotplug.rst for more information. 5398604d9e5SVitaly Kuznetsov 5408604d9e5SVitaly Kuznetsov Say Y here if you want all hot-plugged memory blocks to appear in 5418604d9e5SVitaly Kuznetsov 'online' state by default. 5428604d9e5SVitaly Kuznetsov Say N here if you want the default policy to keep all hot-plugged 5438604d9e5SVitaly Kuznetsov memory blocks in 'offline' state. 5448604d9e5SVitaly Kuznetsov 5450c0e6195SKAMEZAWA Hiroyukiconfig MEMORY_HOTREMOVE 5460c0e6195SKAMEZAWA Hiroyuki bool "Allow for memory hot remove" 547f7e3334aSNathan Fontenot select HAVE_BOOTMEM_INFO_NODE if (X86_64 || PPC64) 5480c0e6195SKAMEZAWA Hiroyuki depends on MEMORY_HOTPLUG && ARCH_ENABLE_MEMORY_HOTREMOVE 5490c0e6195SKAMEZAWA Hiroyuki depends on MIGRATION 5500c0e6195SKAMEZAWA Hiroyuki 551a08a2ae3SOscar Salvadorconfig MHP_MEMMAP_ON_MEMORY 552a08a2ae3SOscar Salvador def_bool y 553a08a2ae3SOscar Salvador depends on MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP 554a08a2ae3SOscar Salvador depends on ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE 555a08a2ae3SOscar Salvador 556519bcb79SJohannes Weinerendif # MEMORY_HOTPLUG 557519bcb79SJohannes Weiner 5584c21e2f2SHugh Dickins# Heavily threaded applications may benefit from splitting the mm-wide 5594c21e2f2SHugh Dickins# page_table_lock, so that faults on different parts of the user address 5604c21e2f2SHugh Dickins# space can be handled with less contention: split it at this NR_CPUS. 5614c21e2f2SHugh Dickins# Default to 4 for wider testing, though 8 might be more appropriate. 5624c21e2f2SHugh Dickins# ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock. 5637b6ac9dfSHugh Dickins# PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes. 56460bccaa6SWill Deacon# SPARC32 allocates multiple pte tables within a single page, and therefore 56560bccaa6SWill Deacon# a per-page lock leads to problems when multiple tables need to be locked 56660bccaa6SWill Deacon# at the same time (e.g. copy_page_range()). 567a70caa8bSHugh Dickins# DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page. 5684c21e2f2SHugh Dickins# 5694c21e2f2SHugh Dickinsconfig SPLIT_PTLOCK_CPUS 5704c21e2f2SHugh Dickins int 5719164550eSKirill A. Shutemov default "999999" if !MMU 572a70caa8bSHugh Dickins default "999999" if ARM && !CPU_CACHE_VIPT 573a70caa8bSHugh Dickins default "999999" if PARISC && !PA20 57460bccaa6SWill Deacon default "999999" if SPARC32 5754c21e2f2SHugh Dickins default "4" 5767cbe34cfSChristoph Lameter 577e009bb30SKirill A. Shutemovconfig ARCH_ENABLE_SPLIT_PMD_PTLOCK 5786341e62bSChristoph Jaeger bool 579e009bb30SKirill A. Shutemov 5807cbe34cfSChristoph Lameter# 58109316c09SKonstantin Khlebnikov# support for memory balloon 58209316c09SKonstantin Khlebnikovconfig MEMORY_BALLOON 5836341e62bSChristoph Jaeger bool 58409316c09SKonstantin Khlebnikov 58509316c09SKonstantin Khlebnikov# 58618468d93SRafael Aquini# support for memory balloon compaction 58718468d93SRafael Aquiniconfig BALLOON_COMPACTION 58818468d93SRafael Aquini bool "Allow for balloon memory compaction/migration" 58918468d93SRafael Aquini def_bool y 59009316c09SKonstantin Khlebnikov depends on COMPACTION && MEMORY_BALLOON 59118468d93SRafael Aquini help 59218468d93SRafael Aquini Memory fragmentation introduced by ballooning might reduce 59318468d93SRafael Aquini significantly the number of 2MB contiguous memory blocks that can be 59418468d93SRafael Aquini used within a guest, thus imposing performance penalties associated 59518468d93SRafael Aquini with the reduced number of transparent huge pages that could be used 59618468d93SRafael Aquini by the guest workload. Allowing the compaction & migration for memory 59718468d93SRafael Aquini pages enlisted as being part of memory balloon devices avoids the 59818468d93SRafael Aquini scenario aforementioned and helps improving memory defragmentation. 59918468d93SRafael Aquini 60018468d93SRafael Aquini# 601e9e96b39SMel Gorman# support for memory compaction 602e9e96b39SMel Gormanconfig COMPACTION 603e9e96b39SMel Gorman bool "Allow for memory compaction" 60405106e6aSRik van Riel def_bool y 605e9e96b39SMel Gorman select MIGRATION 60633a93877SAndrea Arcangeli depends on MMU 607e9e96b39SMel Gorman help 608b32eaf71SMichal Hocko Compaction is the only memory management component to form 609b32eaf71SMichal Hocko high order (larger physically contiguous) memory blocks 610b32eaf71SMichal Hocko reliably. The page allocator relies on compaction heavily and 611b32eaf71SMichal Hocko the lack of the feature can lead to unexpected OOM killer 612b32eaf71SMichal Hocko invocations for high order memory requests. You shouldn't 613b32eaf71SMichal Hocko disable this option unless there really is a strong reason for 614b32eaf71SMichal Hocko it and then we would be really interested to hear about that at 615b32eaf71SMichal Hocko linux-mm@kvack.org. 616e9e96b39SMel Gorman 617c7e0b3d0SThomas Gleixnerconfig COMPACT_UNEVICTABLE_DEFAULT 618c7e0b3d0SThomas Gleixner int 619c7e0b3d0SThomas Gleixner depends on COMPACTION 620c7e0b3d0SThomas Gleixner default 0 if PREEMPT_RT 621c7e0b3d0SThomas Gleixner default 1 622c7e0b3d0SThomas Gleixner 623e9e96b39SMel Gorman# 62436e66c55SAlexander Duyck# support for free page reporting 62536e66c55SAlexander Duyckconfig PAGE_REPORTING 62636e66c55SAlexander Duyck bool "Free page reporting" 62736e66c55SAlexander Duyck def_bool n 62836e66c55SAlexander Duyck help 62936e66c55SAlexander Duyck Free page reporting allows for the incremental acquisition of 63036e66c55SAlexander Duyck free pages from the buddy allocator for the purpose of reporting 63136e66c55SAlexander Duyck those pages to another entity, such as a hypervisor, so that the 63236e66c55SAlexander Duyck memory can be freed within the host for other uses. 63336e66c55SAlexander Duyck 63436e66c55SAlexander Duyck# 6357cbe34cfSChristoph Lameter# support for page migration 6367cbe34cfSChristoph Lameter# 6377cbe34cfSChristoph Lameterconfig MIGRATION 638b20a3503SChristoph Lameter bool "Page migration" 6396c5240aeSChristoph Lameter def_bool y 640de32a817SChen Gang depends on (NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE || COMPACTION || CMA) && MMU 641b20a3503SChristoph Lameter help 642b20a3503SChristoph Lameter Allows the migration of the physical location of pages of processes 643e9e96b39SMel Gorman while the virtual addresses are not changed. This is useful in 644e9e96b39SMel Gorman two situations. The first is on NUMA systems to put pages nearer 645e9e96b39SMel Gorman to the processors accessing. The second is when allocating huge 646e9e96b39SMel Gorman pages as migration can relocate pages to satisfy a huge page 647e9e96b39SMel Gorman allocation instead of reclaiming. 6486550e07fSGreg Kroah-Hartman 64976cbbeadSChristoph Hellwigconfig DEVICE_MIGRATION 650d90a25f8SChristoph Hellwig def_bool MIGRATION && ZONE_DEVICE 65176cbbeadSChristoph Hellwig 652c177c81eSNaoya Horiguchiconfig ARCH_ENABLE_HUGEPAGE_MIGRATION 6536341e62bSChristoph Jaeger bool 654c177c81eSNaoya Horiguchi 6559c670ea3SNaoya Horiguchiconfig ARCH_ENABLE_THP_MIGRATION 6569c670ea3SNaoya Horiguchi bool 6579c670ea3SNaoya Horiguchi 6584bfb68a0SAnshuman Khandualconfig HUGETLB_PAGE_SIZE_VARIABLE 6594bfb68a0SAnshuman Khandual def_bool n 6604bfb68a0SAnshuman Khandual help 6614bfb68a0SAnshuman Khandual Allows the pageblock_order value to be dynamic instead of just standard 6624bfb68a0SAnshuman Khandual HUGETLB_PAGE_ORDER when there are multiple HugeTLB page sizes available 6634bfb68a0SAnshuman Khandual on a platform. 6644bfb68a0SAnshuman Khandual 66523baf831SKirill A. Shutemov Note that the pageblock_order cannot exceed MAX_ORDER and will be 66623baf831SKirill A. Shutemov clamped down to MAX_ORDER. 667b3d40a2bSDavid Hildenbrand 6688df995f6SAlexandre Ghiticonfig CONTIG_ALLOC 6698df995f6SAlexandre Ghiti def_bool (MEMORY_ISOLATION && COMPACTION) || CMA 6708df995f6SAlexandre Ghiti 671600715dcSJeremy Fitzhardingeconfig PHYS_ADDR_T_64BIT 672d4a451d5SChristoph Hellwig def_bool 64BIT 673600715dcSJeremy Fitzhardinge 6742a7326b5SChristoph Lameterconfig BOUNCE 6759ca24e2eSVinayak Menon bool "Enable bounce buffers" 6769ca24e2eSVinayak Menon default y 677ce288e05SChristoph Hellwig depends on BLOCK && MMU && HIGHMEM 6789ca24e2eSVinayak Menon help 679ce288e05SChristoph Hellwig Enable bounce buffers for devices that cannot access the full range of 680ce288e05SChristoph Hellwig memory available to the CPU. Enabled by default when HIGHMEM is 681ce288e05SChristoph Hellwig selected, but you may say n to override this. 6822a7326b5SChristoph Lameter 683cddb8a5cSAndrea Arcangeliconfig MMU_NOTIFIER 684cddb8a5cSAndrea Arcangeli bool 68599cb252fSJason Gunthorpe select INTERVAL_TREE 686fc4d5c29SDavid Howells 687f8af4da3SHugh Dickinsconfig KSM 688f8af4da3SHugh Dickins bool "Enable KSM for page merging" 689f8af4da3SHugh Dickins depends on MMU 69059e1a2f4STimofey Titovets select XXHASH 691f8af4da3SHugh Dickins help 692f8af4da3SHugh Dickins Enable Kernel Samepage Merging: KSM periodically scans those areas 693f8af4da3SHugh Dickins of an application's address space that an app has advised may be 694f8af4da3SHugh Dickins mergeable. When it finds pages of identical content, it replaces 695d0f209f6SHugh Dickins the many instances by a single page with that content, so 696f8af4da3SHugh Dickins saving memory until one or another app needs to modify the content. 697f8af4da3SHugh Dickins Recommended for use with KVM, or with other duplicative applications. 698ee65728eSMike Rapoport See Documentation/mm/ksm.rst for more information: KSM is inactive 699c73602adSHugh Dickins until a program has madvised that an area is MADV_MERGEABLE, and 700c73602adSHugh Dickins root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set). 701f8af4da3SHugh Dickins 702e0a94c2aSChristoph Lameterconfig DEFAULT_MMAP_MIN_ADDR 703e0a94c2aSChristoph Lameter int "Low address space to protect from user allocation" 7046e141546SDavid Howells depends on MMU 705e0a94c2aSChristoph Lameter default 4096 706e0a94c2aSChristoph Lameter help 707e0a94c2aSChristoph Lameter This is the portion of low virtual memory which should be protected 708e0a94c2aSChristoph Lameter from userspace allocation. Keeping a user from writing to low pages 709e0a94c2aSChristoph Lameter can help reduce the impact of kernel NULL pointer bugs. 710e0a94c2aSChristoph Lameter 711e0a94c2aSChristoph Lameter For most ia64, ppc64 and x86 users with lots of address space 712e0a94c2aSChristoph Lameter a value of 65536 is reasonable and should cause no problems. 713e0a94c2aSChristoph Lameter On arm and other archs it should not be higher than 32768. 714788084abSEric Paris Programs which use vm86 functionality or have some need to map 715788084abSEric Paris this low address space will need CAP_SYS_RAWIO or disable this 716788084abSEric Paris protection by setting the value to 0. 717e0a94c2aSChristoph Lameter 718e0a94c2aSChristoph Lameter This value can be changed after boot using the 719e0a94c2aSChristoph Lameter /proc/sys/vm/mmap_min_addr tunable. 720e0a94c2aSChristoph Lameter 721d949f36fSLinus Torvaldsconfig ARCH_SUPPORTS_MEMORY_FAILURE 722d949f36fSLinus Torvalds bool 723e0a94c2aSChristoph Lameter 7246a46079cSAndi Kleenconfig MEMORY_FAILURE 7256a46079cSAndi Kleen depends on MMU 726d949f36fSLinus Torvalds depends on ARCH_SUPPORTS_MEMORY_FAILURE 7276a46079cSAndi Kleen bool "Enable recovery from hardware memory errors" 728ee6f509cSMinchan Kim select MEMORY_ISOLATION 72997f0b134SXie XiuQi select RAS 7306a46079cSAndi Kleen help 7316a46079cSAndi Kleen Enables code to recover from some memory failures on systems 7326a46079cSAndi Kleen with MCA recovery. This allows a system to continue running 7336a46079cSAndi Kleen even when some of its memory has uncorrected errors. This requires 7346a46079cSAndi Kleen special hardware support and typically ECC memory. 7356a46079cSAndi Kleen 736cae681fcSAndi Kleenconfig HWPOISON_INJECT 737413f9efbSAndi Kleen tristate "HWPoison pages injector" 73827df5068SAndi Kleen depends on MEMORY_FAILURE && DEBUG_KERNEL && PROC_FS 739478c5ffcSWu Fengguang select PROC_PAGE_MONITOR 740cae681fcSAndi Kleen 741fc4d5c29SDavid Howellsconfig NOMMU_INITIAL_TRIM_EXCESS 742fc4d5c29SDavid Howells int "Turn on mmap() excess space trimming before booting" 743fc4d5c29SDavid Howells depends on !MMU 744fc4d5c29SDavid Howells default 1 745fc4d5c29SDavid Howells help 746fc4d5c29SDavid Howells The NOMMU mmap() frequently needs to allocate large contiguous chunks 747fc4d5c29SDavid Howells of memory on which to store mappings, but it can only ask the system 748fc4d5c29SDavid Howells allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently 749fc4d5c29SDavid Howells more than it requires. To deal with this, mmap() is able to trim off 750fc4d5c29SDavid Howells the excess and return it to the allocator. 751fc4d5c29SDavid Howells 752fc4d5c29SDavid Howells If trimming is enabled, the excess is trimmed off and returned to the 753fc4d5c29SDavid Howells system allocator, which can cause extra fragmentation, particularly 754fc4d5c29SDavid Howells if there are a lot of transient processes. 755fc4d5c29SDavid Howells 756fc4d5c29SDavid Howells If trimming is disabled, the excess is kept, but not used, which for 757fc4d5c29SDavid Howells long-term mappings means that the space is wasted. 758fc4d5c29SDavid Howells 759fc4d5c29SDavid Howells Trimming can be dynamically controlled through a sysctl option 760fc4d5c29SDavid Howells (/proc/sys/vm/nr_trim_pages) which specifies the minimum number of 761fc4d5c29SDavid Howells excess pages there must be before trimming should occur, or zero if 762fc4d5c29SDavid Howells no trimming is to occur. 763fc4d5c29SDavid Howells 764fc4d5c29SDavid Howells This option specifies the initial value of this option. The default 765fc4d5c29SDavid Howells of 1 says that all excess pages should be trimmed. 766fc4d5c29SDavid Howells 767dd19d293SStephen Kitt See Documentation/admin-guide/mm/nommu-mmap.rst for more information. 768bbddff05STejun Heo 769519bcb79SJohannes Weinerconfig ARCH_WANT_GENERAL_HUGETLB 770519bcb79SJohannes Weiner bool 771519bcb79SJohannes Weiner 772519bcb79SJohannes Weinerconfig ARCH_WANTS_THP_SWAP 773519bcb79SJohannes Weiner def_bool n 774519bcb79SJohannes Weiner 775519bcb79SJohannes Weinermenuconfig TRANSPARENT_HUGEPAGE 77613ece886SAndrea Arcangeli bool "Transparent Hugepage Support" 777554b0f3cSSebastian Andrzej Siewior depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT 7785d689240SAndrea Arcangeli select COMPACTION 7793a08cd52SMatthew Wilcox select XARRAY_MULTI 7804c76d9d1SAndrea Arcangeli help 7814c76d9d1SAndrea Arcangeli Transparent Hugepages allows the kernel to use huge pages and 7824c76d9d1SAndrea Arcangeli huge tlb transparently to the applications whenever possible. 7834c76d9d1SAndrea Arcangeli This feature can improve computing performance to certain 7844c76d9d1SAndrea Arcangeli applications by speeding up page faults during memory 7854c76d9d1SAndrea Arcangeli allocation, by reducing the number of tlb misses and by speeding 7864c76d9d1SAndrea Arcangeli up the pagetable walking. 7874c76d9d1SAndrea Arcangeli 7884c76d9d1SAndrea Arcangeli If memory constrained on embedded, you may want to say N. 7894c76d9d1SAndrea Arcangeli 790519bcb79SJohannes Weinerif TRANSPARENT_HUGEPAGE 791519bcb79SJohannes Weiner 79213ece886SAndrea Arcangelichoice 79313ece886SAndrea Arcangeli prompt "Transparent Hugepage Support sysfs defaults" 79413ece886SAndrea Arcangeli depends on TRANSPARENT_HUGEPAGE 79513ece886SAndrea Arcangeli default TRANSPARENT_HUGEPAGE_ALWAYS 79613ece886SAndrea Arcangeli help 79713ece886SAndrea Arcangeli Selects the sysfs defaults for Transparent Hugepage Support. 79813ece886SAndrea Arcangeli 79913ece886SAndrea Arcangeli config TRANSPARENT_HUGEPAGE_ALWAYS 80013ece886SAndrea Arcangeli bool "always" 80113ece886SAndrea Arcangeli help 80213ece886SAndrea Arcangeli Enabling Transparent Hugepage always, can increase the 80313ece886SAndrea Arcangeli memory footprint of applications without a guaranteed 80413ece886SAndrea Arcangeli benefit but it will work automatically for all applications. 80513ece886SAndrea Arcangeli 80613ece886SAndrea Arcangeli config TRANSPARENT_HUGEPAGE_MADVISE 80713ece886SAndrea Arcangeli bool "madvise" 80813ece886SAndrea Arcangeli help 80913ece886SAndrea Arcangeli Enabling Transparent Hugepage madvise, will only provide a 81013ece886SAndrea Arcangeli performance improvement benefit to the applications using 81113ece886SAndrea Arcangeli madvise(MADV_HUGEPAGE) but it won't risk to increase the 81213ece886SAndrea Arcangeli memory footprint of applications without a guaranteed 81313ece886SAndrea Arcangeli benefit. 81413ece886SAndrea Arcangeliendchoice 81513ece886SAndrea Arcangeli 81638d8b4e6SHuang Yingconfig THP_SWAP 81738d8b4e6SHuang Ying def_bool y 818dad6a5ebSHugh Dickins depends on TRANSPARENT_HUGEPAGE && ARCH_WANTS_THP_SWAP && SWAP && 64BIT 81938d8b4e6SHuang Ying help 82038d8b4e6SHuang Ying Swap transparent huge pages in one piece, without splitting. 82114fef284SHuang Ying XXX: For now, swap cluster backing transparent huge page 82214fef284SHuang Ying will be split after swapout. 82338d8b4e6SHuang Ying 82438d8b4e6SHuang Ying For selection by architectures with reasonable THP sizes. 82538d8b4e6SHuang Ying 826519bcb79SJohannes Weinerconfig READ_ONLY_THP_FOR_FS 827519bcb79SJohannes Weiner bool "Read-only THP for filesystems (EXPERIMENTAL)" 828519bcb79SJohannes Weiner depends on TRANSPARENT_HUGEPAGE && SHMEM 829519bcb79SJohannes Weiner 830519bcb79SJohannes Weiner help 831519bcb79SJohannes Weiner Allow khugepaged to put read-only file-backed pages in THP. 832519bcb79SJohannes Weiner 833519bcb79SJohannes Weiner This is marked experimental because it is a new feature. Write 834519bcb79SJohannes Weiner support of file THPs will be developed in the next few release 835519bcb79SJohannes Weiner cycles. 836519bcb79SJohannes Weiner 837519bcb79SJohannes Weinerendif # TRANSPARENT_HUGEPAGE 838519bcb79SJohannes Weiner 839e496cf3dSKirill A. Shutemov# 840bbddff05STejun Heo# UP and nommu archs use km based percpu allocator 841bbddff05STejun Heo# 842bbddff05STejun Heoconfig NEED_PER_CPU_KM 8433583521aSVladimir Murzin depends on !SMP || !MMU 844bbddff05STejun Heo bool 845bbddff05STejun Heo default y 846077b1f83SDan Magenheimer 8477ecd19cfSKefeng Wangconfig NEED_PER_CPU_EMBED_FIRST_CHUNK 8487ecd19cfSKefeng Wang bool 8497ecd19cfSKefeng Wang 8507ecd19cfSKefeng Wangconfig NEED_PER_CPU_PAGE_FIRST_CHUNK 8517ecd19cfSKefeng Wang bool 8527ecd19cfSKefeng Wang 8537ecd19cfSKefeng Wangconfig USE_PERCPU_NUMA_NODE_ID 8547ecd19cfSKefeng Wang bool 8557ecd19cfSKefeng Wang 8567ecd19cfSKefeng Wangconfig HAVE_SETUP_PER_CPU_AREA 8577ecd19cfSKefeng Wang bool 8587ecd19cfSKefeng Wang 85927c6aec2SDan Magenheimerconfig FRONTSWAP 8606e61dde8SChristoph Hellwig bool 861f825c736SAneesh Kumar K.V 862f825c736SAneesh Kumar K.Vconfig CMA 863f825c736SAneesh Kumar K.V bool "Contiguous Memory Allocator" 864aca52c39SMike Rapoport depends on MMU 865f825c736SAneesh Kumar K.V select MIGRATION 866f825c736SAneesh Kumar K.V select MEMORY_ISOLATION 867f825c736SAneesh Kumar K.V help 868f825c736SAneesh Kumar K.V This enables the Contiguous Memory Allocator which allows other 869f825c736SAneesh Kumar K.V subsystems to allocate big physically-contiguous blocks of memory. 870f825c736SAneesh Kumar K.V CMA reserves a region of memory and allows only movable pages to 871f825c736SAneesh Kumar K.V be allocated from it. This way, the kernel can use the memory for 872f825c736SAneesh Kumar K.V pagecache and when a subsystem requests for contiguous area, the 873f825c736SAneesh Kumar K.V allocated pages are migrated away to serve the contiguous request. 874f825c736SAneesh Kumar K.V 875f825c736SAneesh Kumar K.V If unsure, say "n". 876f825c736SAneesh Kumar K.V 877f825c736SAneesh Kumar K.Vconfig CMA_DEBUG 878f825c736SAneesh Kumar K.V bool "CMA debug messages (DEVELOPMENT)" 879f825c736SAneesh Kumar K.V depends on DEBUG_KERNEL && CMA 880f825c736SAneesh Kumar K.V help 881f825c736SAneesh Kumar K.V Turns on debug messages in CMA. This produces KERN_DEBUG 882f825c736SAneesh Kumar K.V messages for every CMA call as well as various messages while 883f825c736SAneesh Kumar K.V processing calls such as dma_alloc_from_contiguous(). 884f825c736SAneesh Kumar K.V This option does not affect warning and error messages. 885bf550fc9SAlexander Graf 88628b24c1fSSasha Levinconfig CMA_DEBUGFS 88728b24c1fSSasha Levin bool "CMA debugfs interface" 88828b24c1fSSasha Levin depends on CMA && DEBUG_FS 88928b24c1fSSasha Levin help 89028b24c1fSSasha Levin Turns on the DebugFS interface for CMA. 89128b24c1fSSasha Levin 89243ca106fSMinchan Kimconfig CMA_SYSFS 89343ca106fSMinchan Kim bool "CMA information through sysfs interface" 89443ca106fSMinchan Kim depends on CMA && SYSFS 89543ca106fSMinchan Kim help 89643ca106fSMinchan Kim This option exposes some sysfs attributes to get information 89743ca106fSMinchan Kim from CMA. 89843ca106fSMinchan Kim 899a254129eSJoonsoo Kimconfig CMA_AREAS 900a254129eSJoonsoo Kim int "Maximum count of the CMA areas" 901a254129eSJoonsoo Kim depends on CMA 902b7176c26SBarry Song default 19 if NUMA 903a254129eSJoonsoo Kim default 7 904a254129eSJoonsoo Kim help 905a254129eSJoonsoo Kim CMA allows to create CMA areas for particular purpose, mainly, 906a254129eSJoonsoo Kim used as device private area. This parameter sets the maximum 907a254129eSJoonsoo Kim number of CMA area in the system. 908a254129eSJoonsoo Kim 909b7176c26SBarry Song If unsure, leave the default value "7" in UMA and "19" in NUMA. 910a254129eSJoonsoo Kim 911af8d417aSDan Streetmanconfig MEM_SOFT_DIRTY 912af8d417aSDan Streetman bool "Track memory changes" 913af8d417aSDan Streetman depends on CHECKPOINT_RESTORE && HAVE_ARCH_SOFT_DIRTY && PROC_FS 914af8d417aSDan Streetman select PROC_PAGE_MONITOR 9154e2e2770SSeth Jennings help 916af8d417aSDan Streetman This option enables memory changes tracking by introducing a 917af8d417aSDan Streetman soft-dirty bit on pte-s. This bit it set when someone writes 918af8d417aSDan Streetman into a page just as regular dirty bit, but unlike the latter 919af8d417aSDan Streetman it can be cleared by hands. 920af8d417aSDan Streetman 9211ad1335dSMike Rapoport See Documentation/admin-guide/mm/soft-dirty.rst for more details. 9224e2e2770SSeth Jennings 9239e5c33d7SMark Salterconfig GENERIC_EARLY_IOREMAP 9249e5c33d7SMark Salter bool 925042d27acSHelge Deller 92622ee3ea5SHelge Dellerconfig STACK_MAX_DEFAULT_SIZE_MB 92722ee3ea5SHelge Deller int "Default maximum user stack size for 32-bit processes (MB)" 92822ee3ea5SHelge Deller default 100 929042d27acSHelge Deller range 8 2048 930042d27acSHelge Deller depends on STACK_GROWSUP && (!64BIT || COMPAT) 931042d27acSHelge Deller help 932042d27acSHelge Deller This is the maximum stack size in Megabytes in the VM layout of 32-bit 933042d27acSHelge Deller user processes when the stack grows upwards (currently only on parisc 93422ee3ea5SHelge Deller arch) when the RLIMIT_STACK hard limit is unlimited. 935042d27acSHelge Deller 93622ee3ea5SHelge Deller A sane initial value is 100 MB. 9373a80a7faSMel Gorman 9383a80a7faSMel Gormanconfig DEFERRED_STRUCT_PAGE_INIT 9391ce22103SVlastimil Babka bool "Defer initialisation of struct pages to kthreads" 940d39f8fb4SMike Rapoport depends on SPARSEMEM 941ab1e8d89SPavel Tatashin depends on !NEED_PER_CPU_KM 942889c695dSPasha Tatashin depends on 64BIT 943e4443149SDaniel Jordan select PADATA 9443a80a7faSMel Gorman help 9453a80a7faSMel Gorman Ordinarily all struct pages are initialised during early boot in a 9463a80a7faSMel Gorman single thread. On very large machines this can take a considerable 9473a80a7faSMel Gorman amount of time. If this option is set, large machines will bring up 948e4443149SDaniel Jordan a subset of memmap at boot and then initialise the rest in parallel. 949e4443149SDaniel Jordan This has a potential performance impact on tasks running early in the 9501ce22103SVlastimil Babka lifetime of the system until these kthreads finish the 9511ce22103SVlastimil Babka initialisation. 952033fbae9SDan Williams 9531c676e0dSSeongJae Parkconfig PAGE_IDLE_FLAG 9541c676e0dSSeongJae Park bool 9551c676e0dSSeongJae Park select PAGE_EXTENSION if !64BIT 9561c676e0dSSeongJae Park help 9571c676e0dSSeongJae Park This adds PG_idle and PG_young flags to 'struct page'. PTE Accessed 9581c676e0dSSeongJae Park bit writers can set the state of the bit in the flags so that PTE 9591c676e0dSSeongJae Park Accessed bit readers may avoid disturbance. 9601c676e0dSSeongJae Park 96133c3fc71SVladimir Davydovconfig IDLE_PAGE_TRACKING 96233c3fc71SVladimir Davydov bool "Enable idle page tracking" 96333c3fc71SVladimir Davydov depends on SYSFS && MMU 9641c676e0dSSeongJae Park select PAGE_IDLE_FLAG 96533c3fc71SVladimir Davydov help 96633c3fc71SVladimir Davydov This feature allows to estimate the amount of user pages that have 96733c3fc71SVladimir Davydov not been touched during a given period of time. This information can 96833c3fc71SVladimir Davydov be useful to tune memory cgroup limits and/or for job placement 96933c3fc71SVladimir Davydov within a compute cluster. 97033c3fc71SVladimir Davydov 9711ad1335dSMike Rapoport See Documentation/admin-guide/mm/idle_page_tracking.rst for 9721ad1335dSMike Rapoport more details. 97333c3fc71SVladimir Davydov 974c2280be8SAnshuman Khandualconfig ARCH_HAS_CACHE_LINE_SIZE 975c2280be8SAnshuman Khandual bool 976c2280be8SAnshuman Khandual 9772792d84eSKees Cookconfig ARCH_HAS_CURRENT_STACK_POINTER 9782792d84eSKees Cook bool 9792792d84eSKees Cook help 9802792d84eSKees Cook In support of HARDENED_USERCOPY performing stack variable lifetime 9812792d84eSKees Cook checking, an architecture-agnostic way to find the stack pointer 9822792d84eSKees Cook is needed. Once an architecture defines an unsigned long global 9832792d84eSKees Cook register alias named "current_stack_pointer", this config can be 9842792d84eSKees Cook selected. 9852792d84eSKees Cook 98617596731SRobin Murphyconfig ARCH_HAS_PTE_DEVMAP 98765f7d049SOliver O'Halloran bool 98865f7d049SOliver O'Halloran 98963703f37SKefeng Wangconfig ARCH_HAS_ZONE_DMA_SET 99063703f37SKefeng Wang bool 99163703f37SKefeng Wang 99263703f37SKefeng Wangconfig ZONE_DMA 99363703f37SKefeng Wang bool "Support DMA zone" if ARCH_HAS_ZONE_DMA_SET 99463703f37SKefeng Wang default y if ARM64 || X86 99563703f37SKefeng Wang 99663703f37SKefeng Wangconfig ZONE_DMA32 99763703f37SKefeng Wang bool "Support DMA32 zone" if ARCH_HAS_ZONE_DMA_SET 99863703f37SKefeng Wang depends on !X86_32 99963703f37SKefeng Wang default y if ARM64 100063703f37SKefeng Wang 1001033fbae9SDan Williamsconfig ZONE_DEVICE 10025042db43SJérôme Glisse bool "Device memory (pmem, HMM, etc...) hotplug support" 1003033fbae9SDan Williams depends on MEMORY_HOTPLUG 1004033fbae9SDan Williams depends on MEMORY_HOTREMOVE 100599490f16SDan Williams depends on SPARSEMEM_VMEMMAP 100617596731SRobin Murphy depends on ARCH_HAS_PTE_DEVMAP 10073a08cd52SMatthew Wilcox select XARRAY_MULTI 1008033fbae9SDan Williams 1009033fbae9SDan Williams help 1010033fbae9SDan Williams Device memory hotplug support allows for establishing pmem, 1011033fbae9SDan Williams or other device driver discovered memory regions, in the 1012033fbae9SDan Williams memmap. This allows pfn_to_page() lookups of otherwise 1013033fbae9SDan Williams "device-physical" addresses which is needed for using a DAX 1014033fbae9SDan Williams mapping in an O_DIRECT operation, among other things. 1015033fbae9SDan Williams 1016033fbae9SDan Williams If FS_DAX is enabled, then say Y. 101706a660adSLinus Torvalds 10189c240a7bSChristoph Hellwig# 10199c240a7bSChristoph Hellwig# Helpers to mirror range of the CPU page tables of a process into device page 10209c240a7bSChristoph Hellwig# tables. 10219c240a7bSChristoph Hellwig# 1022c0b12405SJérôme Glisseconfig HMM_MIRROR 10239c240a7bSChristoph Hellwig bool 1024f442c283SChristoph Hellwig depends on MMU 1025c0b12405SJérôme Glisse 102614b80582SDan Williamsconfig GET_FREE_REGION 102714b80582SDan Williams depends on SPARSEMEM 102814b80582SDan Williams bool 102914b80582SDan Williams 10305042db43SJérôme Glisseconfig DEVICE_PRIVATE 10315042db43SJérôme Glisse bool "Unaddressable device memory (GPU memory, ...)" 10327328d9ccSChristoph Hellwig depends on ZONE_DEVICE 103314b80582SDan Williams select GET_FREE_REGION 10345042db43SJérôme Glisse 10355042db43SJérôme Glisse help 10365042db43SJérôme Glisse Allows creation of struct pages to represent unaddressable device 10375042db43SJérôme Glisse memory; i.e., memory that is only accessible from the device (or 10385042db43SJérôme Glisse group of devices). You likely also want to select HMM_MIRROR. 10395042db43SJérôme Glisse 10403e9a9e25SChristoph Hellwigconfig VMAP_PFN 10413e9a9e25SChristoph Hellwig bool 10423e9a9e25SChristoph Hellwig 104363c17fb8SDave Hansenconfig ARCH_USES_HIGH_VMA_FLAGS 104463c17fb8SDave Hansen bool 104566d37570SDave Hansenconfig ARCH_HAS_PKEYS 104666d37570SDave Hansen bool 104730a5b536SDennis Zhou 1048b0284cd2SCatalin Marinasconfig ARCH_USES_PG_ARCH_X 1049b0284cd2SCatalin Marinas bool 1050b0284cd2SCatalin Marinas help 1051b0284cd2SCatalin Marinas Enable the definition of PG_arch_x page flags with x > 1. Only 1052b0284cd2SCatalin Marinas suitable for 64-bit architectures with CONFIG_FLATMEM or 1053b0284cd2SCatalin Marinas CONFIG_SPARSEMEM_VMEMMAP enabled, otherwise there may not be 1054b0284cd2SCatalin Marinas enough room for additional bits in page->flags. 1055b0284cd2SCatalin Marinas 10560710d012SVlastimil Babkaconfig VM_EVENT_COUNTERS 10570710d012SVlastimil Babka default y 10580710d012SVlastimil Babka bool "Enable VM event counters for /proc/vmstat" if EXPERT 10590710d012SVlastimil Babka help 10600710d012SVlastimil Babka VM event counters are needed for event counts to be shown. 10610710d012SVlastimil Babka This option allows the disabling of the VM event counters 10620710d012SVlastimil Babka on EXPERT systems. /proc/vmstat will only show page counts 10630710d012SVlastimil Babka if VM event counters are disabled. 10640710d012SVlastimil Babka 106530a5b536SDennis Zhouconfig PERCPU_STATS 106630a5b536SDennis Zhou bool "Collect percpu memory statistics" 106730a5b536SDennis Zhou help 106830a5b536SDennis Zhou This feature collects and exposes statistics via debugfs. The 106930a5b536SDennis Zhou information includes global and per chunk statistics, which can 107030a5b536SDennis Zhou be used to help understand percpu memory usage. 107164c349f4SKirill A. Shutemov 10729c84f229SJohn Hubbardconfig GUP_TEST 10739c84f229SJohn Hubbard bool "Enable infrastructure for get_user_pages()-related unit tests" 1074d0de8241SBarry Song depends on DEBUG_FS 107564c349f4SKirill A. Shutemov help 10769c84f229SJohn Hubbard Provides /sys/kernel/debug/gup_test, which in turn provides a way 10779c84f229SJohn Hubbard to make ioctl calls that can launch kernel-based unit tests for 10789c84f229SJohn Hubbard the get_user_pages*() and pin_user_pages*() family of API calls. 107964c349f4SKirill A. Shutemov 10809c84f229SJohn Hubbard These tests include benchmark testing of the _fast variants of 10819c84f229SJohn Hubbard get_user_pages*() and pin_user_pages*(), as well as smoke tests of 10829c84f229SJohn Hubbard the non-_fast variants. 10839c84f229SJohn Hubbard 1084f4f9bda4SJohn Hubbard There is also a sub-test that allows running dump_page() on any 1085f4f9bda4SJohn Hubbard of up to eight pages (selected by command line args) within the 1086f4f9bda4SJohn Hubbard range of user-space addresses. These pages are either pinned via 1087f4f9bda4SJohn Hubbard pin_user_pages*(), or pinned via get_user_pages*(), as specified 1088f4f9bda4SJohn Hubbard by other command line arguments. 1089f4f9bda4SJohn Hubbard 1090baa489faSSeongJae Park See tools/testing/selftests/mm/gup_test.c 10913010a5eaSLaurent Dufour 1092d0de8241SBarry Songcomment "GUP_TEST needs to have DEBUG_FS enabled" 1093d0de8241SBarry Song depends on !GUP_TEST && !DEBUG_FS 10943010a5eaSLaurent Dufour 10956ca297d4SPeter Zijlstraconfig GUP_GET_PXX_LOW_HIGH 109639656e83SChristoph Hellwig bool 109739656e83SChristoph Hellwig 1098def85743SKeith Buschconfig DMAPOOL_TEST 1099def85743SKeith Busch tristate "Enable a module to run time tests on dma_pool" 1100def85743SKeith Busch depends on HAS_DMA 1101def85743SKeith Busch help 1102def85743SKeith Busch Provides a test module that will allocate and free many blocks of 1103def85743SKeith Busch various sizes and report how long it takes. This is intended to 1104def85743SKeith Busch provide a consistent way to measure how changes to the 1105def85743SKeith Busch dma_pool_alloc/free routines affect performance. 1106def85743SKeith Busch 11073010a5eaSLaurent Dufourconfig ARCH_HAS_PTE_SPECIAL 11083010a5eaSLaurent Dufour bool 110959e0b520SChristoph Hellwig 1110cbd34da7SChristoph Hellwig# 1111cbd34da7SChristoph Hellwig# Some architectures require a special hugepage directory format that is 1112cbd34da7SChristoph Hellwig# required to support multiple hugepage sizes. For example a4fe3ce76 1113cbd34da7SChristoph Hellwig# "powerpc/mm: Allow more flexible layouts for hugepage pagetables" 1114cbd34da7SChristoph Hellwig# introduced it on powerpc. This allows for a more flexible hugepage 1115cbd34da7SChristoph Hellwig# pagetable layouts. 1116cbd34da7SChristoph Hellwig# 1117cbd34da7SChristoph Hellwigconfig ARCH_HAS_HUGEPD 1118cbd34da7SChristoph Hellwig bool 1119cbd34da7SChristoph Hellwig 1120c5acad84SThomas Hellstromconfig MAPPING_DIRTY_HELPERS 1121c5acad84SThomas Hellstrom bool 1122c5acad84SThomas Hellstrom 1123298fa1adSThomas Gleixnerconfig KMAP_LOCAL 1124298fa1adSThomas Gleixner bool 1125298fa1adSThomas Gleixner 1126825c43f5SArd Biesheuvelconfig KMAP_LOCAL_NON_LINEAR_PTE_ARRAY 1127825c43f5SArd Biesheuvel bool 1128825c43f5SArd Biesheuvel 11291fbaf8fcSChristoph Hellwig# struct io_mapping based helper. Selected by drivers that need them 11301fbaf8fcSChristoph Hellwigconfig IO_MAPPING 11311fbaf8fcSChristoph Hellwig bool 11321507f512SMike Rapoport 11331507f512SMike Rapoportconfig SECRETMEM 113474947724SLukas Bulwahn default y 113574947724SLukas Bulwahn bool "Enable memfd_secret() system call" if EXPERT 113674947724SLukas Bulwahn depends on ARCH_HAS_SET_DIRECT_MAP 113774947724SLukas Bulwahn help 113874947724SLukas Bulwahn Enable the memfd_secret() system call with the ability to create 113974947724SLukas Bulwahn memory areas visible only in the context of the owning process and 114074947724SLukas Bulwahn not mapped to other processes and other kernel page tables. 11411507f512SMike Rapoport 11429a10064fSColin Crossconfig ANON_VMA_NAME 11439a10064fSColin Cross bool "Anonymous VMA name support" 11449a10064fSColin Cross depends on PROC_FS && ADVISE_SYSCALLS && MMU 11459a10064fSColin Cross 11469a10064fSColin Cross help 11479a10064fSColin Cross Allow naming anonymous virtual memory areas. 11489a10064fSColin Cross 11499a10064fSColin Cross This feature allows assigning names to virtual memory areas. Assigned 11509a10064fSColin Cross names can be later retrieved from /proc/pid/maps and /proc/pid/smaps 11519a10064fSColin Cross and help identifying individual anonymous memory areas. 11529a10064fSColin Cross Assigning a name to anonymous virtual memory area might prevent that 11539a10064fSColin Cross area from being merged with adjacent virtual memory areas due to the 11549a10064fSColin Cross difference in their name. 11559a10064fSColin Cross 1156430529b5SPeter Xuconfig USERFAULTFD 1157430529b5SPeter Xu bool "Enable userfaultfd() system call" 1158430529b5SPeter Xu depends on MMU 1159430529b5SPeter Xu help 1160430529b5SPeter Xu Enable the userfaultfd() system call that allows to intercept and 1161430529b5SPeter Xu handle page faults in userland. 1162430529b5SPeter Xu 1163430529b5SPeter Xuconfig HAVE_ARCH_USERFAULTFD_WP 1164430529b5SPeter Xu bool 1165430529b5SPeter Xu help 1166430529b5SPeter Xu Arch has userfaultfd write protection support 1167430529b5SPeter Xu 1168430529b5SPeter Xuconfig HAVE_ARCH_USERFAULTFD_MINOR 1169430529b5SPeter Xu bool 1170430529b5SPeter Xu help 1171430529b5SPeter Xu Arch has userfaultfd minor fault support 1172430529b5SPeter Xu 11731db9dbc2SPeter Xuconfig PTE_MARKER_UFFD_WP 117481e0f15fSPeter Xu bool "Userfaultfd write protection support for shmem/hugetlbfs" 117581e0f15fSPeter Xu default y 117681e0f15fSPeter Xu depends on HAVE_ARCH_USERFAULTFD_WP 11771db9dbc2SPeter Xu 11781db9dbc2SPeter Xu help 11791db9dbc2SPeter Xu Allows to create marker PTEs for userfaultfd write protection 11801db9dbc2SPeter Xu purposes. It is required to enable userfaultfd write protection on 11811db9dbc2SPeter Xu file-backed memory types like shmem and hugetlbfs. 11821db9dbc2SPeter Xu 1183ac35a490SYu Zhao# multi-gen LRU { 1184ec1c86b2SYu Zhaoconfig LRU_GEN 1185ec1c86b2SYu Zhao bool "Multi-Gen LRU" 1186ec1c86b2SYu Zhao depends on MMU 1187ec1c86b2SYu Zhao # make sure folio->flags has enough spare bits 1188ec1c86b2SYu Zhao depends on 64BIT || !SPARSEMEM || SPARSEMEM_VMEMMAP 1189ec1c86b2SYu Zhao help 119007017acbSYu Zhao A high performance LRU implementation to overcommit memory. See 119107017acbSYu Zhao Documentation/admin-guide/mm/multigen_lru.rst for details. 1192ec1c86b2SYu Zhao 1193354ed597SYu Zhaoconfig LRU_GEN_ENABLED 1194354ed597SYu Zhao bool "Enable by default" 1195354ed597SYu Zhao depends on LRU_GEN 1196354ed597SYu Zhao help 1197354ed597SYu Zhao This option enables the multi-gen LRU by default. 1198354ed597SYu Zhao 1199ac35a490SYu Zhaoconfig LRU_GEN_STATS 1200ac35a490SYu Zhao bool "Full stats for debugging" 1201ac35a490SYu Zhao depends on LRU_GEN 1202ac35a490SYu Zhao help 1203ac35a490SYu Zhao Do not enable this option unless you plan to look at historical stats 1204ac35a490SYu Zhao from evicted generations for debugging purpose. 1205ac35a490SYu Zhao 1206ac35a490SYu Zhao This option has a per-memcg and per-node memory overhead. 1207ac35a490SYu Zhao# } 1208ac35a490SYu Zhao 12090b6cc04fSSuren Baghdasaryanconfig ARCH_SUPPORTS_PER_VMA_LOCK 12100b6cc04fSSuren Baghdasaryan def_bool n 12110b6cc04fSSuren Baghdasaryan 12120b6cc04fSSuren Baghdasaryanconfig PER_VMA_LOCK 12130b6cc04fSSuren Baghdasaryan def_bool y 12140b6cc04fSSuren Baghdasaryan depends on ARCH_SUPPORTS_PER_VMA_LOCK && MMU && SMP 12150b6cc04fSSuren Baghdasaryan help 12160b6cc04fSSuren Baghdasaryan Allow per-vma locking during page fault handling. 12170b6cc04fSSuren Baghdasaryan 12180b6cc04fSSuren Baghdasaryan This feature allows locking each virtual memory area separately when 12190b6cc04fSSuren Baghdasaryan handling page faults instead of taking mmap_lock. 12200b6cc04fSSuren Baghdasaryan 12212224d848SSeongJae Parksource "mm/damon/Kconfig" 12222224d848SSeongJae Park 122359e0b520SChristoph Hellwigendmenu 1224