linux/Documentation/sysctl/vm.txt
<<
>>
Prefs
   1Documentation for /proc/sys/vm/*        kernel version 2.6.29
   2        (c) 1998, 1999,  Rik van Riel <riel@nl.linux.org>
   3        (c) 2008         Peter W. Morreale <pmorreale@novell.com>
   4
   5For general info and legal blurb, please look in README.
   6
   7==============================================================
   8
   9This file contains the documentation for the sysctl files in
  10/proc/sys/vm and is valid for Linux kernel version 2.6.29.
  11
  12The files in this directory can be used to tune the operation
  13of the virtual memory (VM) subsystem of the Linux kernel and
  14the writeout of dirty data to disk.
  15
  16Default values and initialization routines for most of these
  17files can be found in mm/swap.c.
  18
  19Currently, these files are in /proc/sys/vm:
  20
  21- admin_reserve_kbytes
  22- block_dump
  23- compact_memory
  24- compact_unevictable_allowed
  25- dirty_background_bytes
  26- dirty_background_ratio
  27- dirty_bytes
  28- dirty_expire_centisecs
  29- dirty_ratio
  30- dirty_writeback_centisecs
  31- drop_caches
  32- extfrag_threshold
  33- hugepages_treat_as_movable
  34- hugetlb_shm_group
  35- laptop_mode
  36- legacy_va_layout
  37- lowmem_reserve_ratio
  38- max_map_count
  39- memory_failure_early_kill
  40- memory_failure_recovery
  41- min_free_kbytes
  42- min_slab_ratio
  43- min_unmapped_ratio
  44- mmap_min_addr
  45- mmap_rnd_bits
  46- mmap_rnd_compat_bits
  47- nr_hugepages
  48- nr_overcommit_hugepages
  49- nr_trim_pages         (only if CONFIG_MMU=n)
  50- numa_zonelist_order
  51- oom_dump_tasks
  52- oom_kill_allocating_task
  53- overcommit_kbytes
  54- overcommit_memory
  55- overcommit_ratio
  56- page-cluster
  57- panic_on_oom
  58- percpu_pagelist_fraction
  59- stat_interval
  60- stat_refresh
  61- numa_stat
  62- swappiness
  63- user_reserve_kbytes
  64- vfs_cache_pressure
  65- watermark_scale_factor
  66- zone_reclaim_mode
  67
  68==============================================================
  69
  70admin_reserve_kbytes
  71
  72The amount of free memory in the system that should be reserved for users
  73with the capability cap_sys_admin.
  74
  75admin_reserve_kbytes defaults to min(3% of free pages, 8MB)
  76
  77That should provide enough for the admin to log in and kill a process,
  78if necessary, under the default overcommit 'guess' mode.
  79
  80Systems running under overcommit 'never' should increase this to account
  81for the full Virtual Memory Size of programs used to recover. Otherwise,
  82root may not be able to log in to recover the system.
  83
  84How do you calculate a minimum useful reserve?
  85
  86sshd or login + bash (or some other shell) + top (or ps, kill, etc.)
  87
  88For overcommit 'guess', we can sum resident set sizes (RSS).
  89On x86_64 this is about 8MB.
  90
  91For overcommit 'never', we can take the max of their virtual sizes (VSZ)
  92and add the sum of their RSS.
  93On x86_64 this is about 128MB.
  94
  95Changing this takes effect whenever an application requests memory.
  96
  97==============================================================
  98
  99block_dump
 100
 101block_dump enables block I/O debugging when set to a nonzero value. More
 102information on block I/O debugging is in Documentation/laptops/laptop-mode.txt.
 103
 104==============================================================
 105
 106compact_memory
 107
 108Available only when CONFIG_COMPACTION is set. When 1 is written to the file,
 109all zones are compacted such that free memory is available in contiguous
 110blocks where possible. This can be important for example in the allocation of
 111huge pages although processes will also directly compact memory as required.
 112
 113==============================================================
 114
 115compact_unevictable_allowed
 116
 117Available only when CONFIG_COMPACTION is set. When set to 1, compaction is
 118allowed to examine the unevictable lru (mlocked pages) for pages to compact.
 119This should be used on systems where stalls for minor page faults are an
 120acceptable trade for large contiguous free memory.  Set to 0 to prevent
 121compaction from moving pages that are unevictable.  Default value is 1.
 122
 123==============================================================
 124
 125dirty_background_bytes
 126
 127Contains the amount of dirty memory at which the background kernel
 128flusher threads will start writeback.
 129
 130Note: dirty_background_bytes is the counterpart of dirty_background_ratio. Only
 131one of them may be specified at a time. When one sysctl is written it is
 132immediately taken into account to evaluate the dirty memory limits and the
 133other appears as 0 when read.
 134
 135==============================================================
 136
 137dirty_background_ratio
 138
 139Contains, as a percentage of total available memory that contains free pages
 140and reclaimable pages, the number of pages at which the background kernel
 141flusher threads will start writing out dirty data.
 142
 143The total available memory is not equal to total system memory.
 144
 145==============================================================
 146
 147dirty_bytes
 148
 149Contains the amount of dirty memory at which a process generating disk writes
 150will itself start writeback.
 151
 152Note: dirty_bytes is the counterpart of dirty_ratio. Only one of them may be
 153specified at a time. When one sysctl is written it is immediately taken into
 154account to evaluate the dirty memory limits and the other appears as 0 when
 155read.
 156
 157Note: the minimum value allowed for dirty_bytes is two pages (in bytes); any
 158value lower than this limit will be ignored and the old configuration will be
 159retained.
 160
 161==============================================================
 162
 163dirty_expire_centisecs
 164
 165This tunable is used to define when dirty data is old enough to be eligible
 166for writeout by the kernel flusher threads.  It is expressed in 100'ths
 167of a second.  Data which has been dirty in-memory for longer than this
 168interval will be written out next time a flusher thread wakes up.
 169
 170==============================================================
 171
 172dirty_ratio
 173
 174Contains, as a percentage of total available memory that contains free pages
 175and reclaimable pages, the number of pages at which a process which is
 176generating disk writes will itself start writing out dirty data.
 177
 178The total available memory is not equal to total system memory.
 179
 180==============================================================
 181
 182dirty_writeback_centisecs
 183
 184The kernel flusher threads will periodically wake up and write `old' data
 185out to disk.  This tunable expresses the interval between those wakeups, in
 186100'ths of a second.
 187
 188Setting this to zero disables periodic writeback altogether.
 189
 190==============================================================
 191
 192drop_caches
 193
 194Writing to this will cause the kernel to drop clean caches, as well as
 195reclaimable slab objects like dentries and inodes.  Once dropped, their
 196memory becomes free.
 197
 198To free pagecache:
 199        echo 1 > /proc/sys/vm/drop_caches
 200To free reclaimable slab objects (includes dentries and inodes):
 201        echo 2 > /proc/sys/vm/drop_caches
 202To free slab objects and pagecache:
 203        echo 3 > /proc/sys/vm/drop_caches
 204
 205This is a non-destructive operation and will not free any dirty objects.
 206To increase the number of objects freed by this operation, the user may run
 207`sync' prior to writing to /proc/sys/vm/drop_caches.  This will minimize the
 208number of dirty objects on the system and create more candidates to be
 209dropped.
 210
 211This file is not a means to control the growth of the various kernel caches
 212(inodes, dentries, pagecache, etc...)  These objects are automatically
 213reclaimed by the kernel when memory is needed elsewhere on the system.
 214
 215Use of this file can cause performance problems.  Since it discards cached
 216objects, it may cost a significant amount of I/O and CPU to recreate the
 217dropped objects, especially if they were under heavy use.  Because of this,
 218use outside of a testing or debugging environment is not recommended.
 219
 220You may see informational messages in your kernel log when this file is
 221used:
 222
 223        cat (1234): drop_caches: 3
 224
 225These are informational only.  They do not mean that anything is wrong
 226with your system.  To disable them, echo 4 (bit 3) into drop_caches.
 227
 228==============================================================
 229
 230extfrag_threshold
 231
 232This parameter affects whether the kernel will compact memory or direct
 233reclaim to satisfy a high-order allocation. The extfrag/extfrag_index file in
 234debugfs shows what the fragmentation index for each order is in each zone in
 235the system. Values tending towards 0 imply allocations would fail due to lack
 236of memory, values towards 1000 imply failures are due to fragmentation and -1
 237implies that the allocation will succeed as long as watermarks are met.
 238
 239The kernel will not compact memory in a zone if the
 240fragmentation index is <= extfrag_threshold. The default value is 500.
 241
 242==============================================================
 243
 244highmem_is_dirtyable
 245
 246Available only for systems with CONFIG_HIGHMEM enabled (32b systems).
 247
 248This parameter controls whether the high memory is considered for dirty
 249writers throttling.  This is not the case by default which means that
 250only the amount of memory directly visible/usable by the kernel can
 251be dirtied. As a result, on systems with a large amount of memory and
 252lowmem basically depleted writers might be throttled too early and
 253streaming writes can get very slow.
 254
 255Changing the value to non zero would allow more memory to be dirtied
 256and thus allow writers to write more data which can be flushed to the
 257storage more effectively. Note this also comes with a risk of pre-mature
 258OOM killer because some writers (e.g. direct block device writes) can
 259only use the low memory and they can fill it up with dirty data without
 260any throttling.
 261
 262==============================================================
 263
 264hugepages_treat_as_movable
 265
 266This parameter controls whether we can allocate hugepages from ZONE_MOVABLE
 267or not. If set to non-zero, hugepages can be allocated from ZONE_MOVABLE.
 268ZONE_MOVABLE is created when kernel boot parameter kernelcore= is specified,
 269so this parameter has no effect if used without kernelcore=.
 270
 271Hugepage migration is now available in some situations which depend on the
 272architecture and/or the hugepage size. If a hugepage supports migration,
 273allocation from ZONE_MOVABLE is always enabled for the hugepage regardless
 274of the value of this parameter.
 275IOW, this parameter affects only non-migratable hugepages.
 276
 277Assuming that hugepages are not migratable in your system, one usecase of
 278this parameter is that users can make hugepage pool more extensible by
 279enabling the allocation from ZONE_MOVABLE. This is because on ZONE_MOVABLE
 280page reclaim/migration/compaction work more and you can get contiguous
 281memory more likely. Note that using ZONE_MOVABLE for non-migratable
 282hugepages can do harm to other features like memory hotremove (because
 283memory hotremove expects that memory blocks on ZONE_MOVABLE are always
 284removable,) so it's a trade-off responsible for the users.
 285
 286==============================================================
 287
 288hugetlb_shm_group
 289
 290hugetlb_shm_group contains group id that is allowed to create SysV
 291shared memory segment using hugetlb page.
 292
 293==============================================================
 294
 295laptop_mode
 296
 297laptop_mode is a knob that controls "laptop mode". All the things that are
 298controlled by this knob are discussed in Documentation/laptops/laptop-mode.txt.
 299
 300==============================================================
 301
 302legacy_va_layout
 303
 304If non-zero, this sysctl disables the new 32-bit mmap layout - the kernel
 305will use the legacy (2.4) layout for all processes.
 306
 307==============================================================
 308
 309lowmem_reserve_ratio
 310
 311For some specialised workloads on highmem machines it is dangerous for
 312the kernel to allow process memory to be allocated from the "lowmem"
 313zone.  This is because that memory could then be pinned via the mlock()
 314system call, or by unavailability of swapspace.
 315
 316And on large highmem machines this lack of reclaimable lowmem memory
 317can be fatal.
 318
 319So the Linux page allocator has a mechanism which prevents allocations
 320which _could_ use highmem from using too much lowmem.  This means that
 321a certain amount of lowmem is defended from the possibility of being
 322captured into pinned user memory.
 323
 324(The same argument applies to the old 16 megabyte ISA DMA region.  This
 325mechanism will also defend that region from allocations which could use
 326highmem or lowmem).
 327
 328The `lowmem_reserve_ratio' tunable determines how aggressive the kernel is
 329in defending these lower zones.
 330
 331If you have a machine which uses highmem or ISA DMA and your
 332applications are using mlock(), or if you are running with no swap then
 333you probably should change the lowmem_reserve_ratio setting.
 334
 335The lowmem_reserve_ratio is an array. You can see them by reading this file.
 336-
 337% cat /proc/sys/vm/lowmem_reserve_ratio
 338256     256     32
 339-
 340Note: # of this elements is one fewer than number of zones. Because the highest
 341      zone's value is not necessary for following calculation.
 342
 343But, these values are not used directly. The kernel calculates # of protection
 344pages for each zones from them. These are shown as array of protection pages
 345in /proc/zoneinfo like followings. (This is an example of x86-64 box).
 346Each zone has an array of protection pages like this.
 347
 348-
 349Node 0, zone      DMA
 350  pages free     1355
 351        min      3
 352        low      3
 353        high     4
 354        :
 355        :
 356    numa_other   0
 357        protection: (0, 2004, 2004, 2004)
 358        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 359  pagesets
 360    cpu: 0 pcp: 0
 361        :
 362-
 363These protections are added to score to judge whether this zone should be used
 364for page allocation or should be reclaimed.
 365
 366In this example, if normal pages (index=2) are required to this DMA zone and
 367watermark[WMARK_HIGH] is used for watermark, the kernel judges this zone should
 368not be used because pages_free(1355) is smaller than watermark + protection[2]
 369(4 + 2004 = 2008). If this protection value is 0, this zone would be used for
 370normal page requirement. If requirement is DMA zone(index=0), protection[0]
 371(=0) is used.
 372
 373zone[i]'s protection[j] is calculated by following expression.
 374
 375(i < j):
 376  zone[i]->protection[j]
 377  = (total sums of managed_pages from zone[i+1] to zone[j] on the node)
 378    / lowmem_reserve_ratio[i];
 379(i = j):
 380   (should not be protected. = 0;
 381(i > j):
 382   (not necessary, but looks 0)
 383
 384The default values of lowmem_reserve_ratio[i] are
 385    256 (if zone[i] means DMA or DMA32 zone)
 386    32  (others).
 387As above expression, they are reciprocal number of ratio.
 388256 means 1/256. # of protection pages becomes about "0.39%" of total managed
 389pages of higher zones on the node.
 390
 391If you would like to protect more pages, smaller values are effective.
 392The minimum value is 1 (1/1 -> 100%).
 393
 394==============================================================
 395
 396max_map_count:
 397
 398This file contains the maximum number of memory map areas a process
 399may have. Memory map areas are used as a side-effect of calling
 400malloc, directly by mmap, mprotect, and madvise, and also when loading
 401shared libraries.
 402
 403While most applications need less than a thousand maps, certain
 404programs, particularly malloc debuggers, may consume lots of them,
 405e.g., up to one or two maps per allocation.
 406
 407The default value is 65536.
 408
 409=============================================================
 410
 411memory_failure_early_kill:
 412
 413Control how to kill processes when uncorrected memory error (typically
 414a 2bit error in a memory module) is detected in the background by hardware
 415that cannot be handled by the kernel. In some cases (like the page
 416still having a valid copy on disk) the kernel will handle the failure
 417transparently without affecting any applications. But if there is
 418no other uptodate copy of the data it will kill to prevent any data
 419corruptions from propagating.
 420
 4211: Kill all processes that have the corrupted and not reloadable page mapped
 422as soon as the corruption is detected.  Note this is not supported
 423for a few types of pages, like kernel internally allocated data or
 424the swap cache, but works for the majority of user pages.
 425
 4260: Only unmap the corrupted page from all processes and only kill a process
 427who tries to access it.
 428
 429The kill is done using a catchable SIGBUS with BUS_MCEERR_AO, so processes can
 430handle this if they want to.
 431
 432This is only active on architectures/platforms with advanced machine
 433check handling and depends on the hardware capabilities.
 434
 435Applications can override this setting individually with the PR_MCE_KILL prctl
 436
 437==============================================================
 438
 439memory_failure_recovery
 440
 441Enable memory failure recovery (when supported by the platform)
 442
 4431: Attempt recovery.
 444
 4450: Always panic on a memory failure.
 446
 447==============================================================
 448
 449min_free_kbytes:
 450
 451This is used to force the Linux VM to keep a minimum number
 452of kilobytes free.  The VM uses this number to compute a
 453watermark[WMARK_MIN] value for each lowmem zone in the system.
 454Each lowmem zone gets a number of reserved free pages based
 455proportionally on its size.
 456
 457Some minimal amount of memory is needed to satisfy PF_MEMALLOC
 458allocations; if you set this to lower than 1024KB, your system will
 459become subtly broken, and prone to deadlock under high loads.
 460
 461Setting this too high will OOM your machine instantly.
 462
 463=============================================================
 464
 465min_slab_ratio:
 466
 467This is available only on NUMA kernels.
 468
 469A percentage of the total pages in each zone.  On Zone reclaim
 470(fallback from the local zone occurs) slabs will be reclaimed if more
 471than this percentage of pages in a zone are reclaimable slab pages.
 472This insures that the slab growth stays under control even in NUMA
 473systems that rarely perform global reclaim.
 474
 475The default is 5 percent.
 476
 477Note that slab reclaim is triggered in a per zone / node fashion.
 478The process of reclaiming slab memory is currently not node specific
 479and may not be fast.
 480
 481=============================================================
 482
 483min_unmapped_ratio:
 484
 485This is available only on NUMA kernels.
 486
 487This is a percentage of the total pages in each zone. Zone reclaim will
 488only occur if more than this percentage of pages are in a state that
 489zone_reclaim_mode allows to be reclaimed.
 490
 491If zone_reclaim_mode has the value 4 OR'd, then the percentage is compared
 492against all file-backed unmapped pages including swapcache pages and tmpfs
 493files. Otherwise, only unmapped pages backed by normal files but not tmpfs
 494files and similar are considered.
 495
 496The default is 1 percent.
 497
 498==============================================================
 499
 500mmap_min_addr
 501
 502This file indicates the amount of address space  which a user process will
 503be restricted from mmapping.  Since kernel null dereference bugs could
 504accidentally operate based on the information in the first couple of pages
 505of memory userspace processes should not be allowed to write to them.  By
 506default this value is set to 0 and no protections will be enforced by the
 507security module.  Setting this value to something like 64k will allow the
 508vast majority of applications to work correctly and provide defense in depth
 509against future potential kernel bugs.
 510
 511==============================================================
 512
 513mmap_rnd_bits:
 514
 515This value can be used to select the number of bits to use to
 516determine the random offset to the base address of vma regions
 517resulting from mmap allocations on architectures which support
 518tuning address space randomization.  This value will be bounded
 519by the architecture's minimum and maximum supported values.
 520
 521This value can be changed after boot using the
 522/proc/sys/vm/mmap_rnd_bits tunable
 523
 524==============================================================
 525
 526mmap_rnd_compat_bits:
 527
 528This value can be used to select the number of bits to use to
 529determine the random offset to the base address of vma regions
 530resulting from mmap allocations for applications run in
 531compatibility mode on architectures which support tuning address
 532space randomization.  This value will be bounded by the
 533architecture's minimum and maximum supported values.
 534
 535This value can be changed after boot using the
 536/proc/sys/vm/mmap_rnd_compat_bits tunable
 537
 538==============================================================
 539
 540nr_hugepages
 541
 542Change the minimum size of the hugepage pool.
 543
 544See Documentation/vm/hugetlbpage.txt
 545
 546==============================================================
 547
 548nr_overcommit_hugepages
 549
 550Change the maximum size of the hugepage pool. The maximum is
 551nr_hugepages + nr_overcommit_hugepages.
 552
 553See Documentation/vm/hugetlbpage.txt
 554
 555==============================================================
 556
 557nr_trim_pages
 558
 559This is available only on NOMMU kernels.
 560
 561This value adjusts the excess page trimming behaviour of power-of-2 aligned
 562NOMMU mmap allocations.
 563
 564A value of 0 disables trimming of allocations entirely, while a value of 1
 565trims excess pages aggressively. Any value >= 1 acts as the watermark where
 566trimming of allocations is initiated.
 567
 568The default value is 1.
 569
 570See Documentation/nommu-mmap.txt for more information.
 571
 572==============================================================
 573
 574numa_zonelist_order
 575
 576This sysctl is only for NUMA and it is deprecated. Anything but
 577Node order will fail!
 578
 579'where the memory is allocated from' is controlled by zonelists.
 580(This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation.
 581 you may be able to read ZONE_DMA as ZONE_DMA32...)
 582
 583In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following.
 584ZONE_NORMAL -> ZONE_DMA
 585This means that a memory allocation request for GFP_KERNEL will
 586get memory from ZONE_DMA only when ZONE_NORMAL is not available.
 587
 588In NUMA case, you can think of following 2 types of order.
 589Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL
 590
 591(A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL
 592(B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA.
 593
 594Type(A) offers the best locality for processes on Node(0), but ZONE_DMA
 595will be used before ZONE_NORMAL exhaustion. This increases possibility of
 596out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small.
 597
 598Type(B) cannot offer the best locality but is more robust against OOM of
 599the DMA zone.
 600
 601Type(A) is called as "Node" order. Type (B) is "Zone" order.
 602
 603"Node order" orders the zonelists by node, then by zone within each node.
 604Specify "[Nn]ode" for node order
 605
 606"Zone Order" orders the zonelists by zone type, then by node within each
 607zone.  Specify "[Zz]one" for zone order.
 608
 609Specify "[Dd]efault" to request automatic configuration.
 610
 611On 32-bit, the Normal zone needs to be preserved for allocations accessible
 612by the kernel, so "zone" order will be selected.
 613
 614On 64-bit, devices that require DMA32/DMA are relatively rare, so "node"
 615order will be selected.
 616
 617Default order is recommended unless this is causing problems for your
 618system/application.
 619
 620==============================================================
 621
 622oom_dump_tasks
 623
 624Enables a system-wide task dump (excluding kernel threads) to be produced
 625when the kernel performs an OOM-killing and includes such information as
 626pid, uid, tgid, vm size, rss, pgtables_bytes, swapents, oom_score_adj
 627score, and name.  This is helpful to determine why the OOM killer was
 628invoked, to identify the rogue task that caused it, and to determine why
 629the OOM killer chose the task it did to kill.
 630
 631If this is set to zero, this information is suppressed.  On very
 632large systems with thousands of tasks it may not be feasible to dump
 633the memory state information for each one.  Such systems should not
 634be forced to incur a performance penalty in OOM conditions when the
 635information may not be desired.
 636
 637If this is set to non-zero, this information is shown whenever the
 638OOM killer actually kills a memory-hogging task.
 639
 640The default value is 1 (enabled).
 641
 642==============================================================
 643
 644oom_kill_allocating_task
 645
 646This enables or disables killing the OOM-triggering task in
 647out-of-memory situations.
 648
 649If this is set to zero, the OOM killer will scan through the entire
 650tasklist and select a task based on heuristics to kill.  This normally
 651selects a rogue memory-hogging task that frees up a large amount of
 652memory when killed.
 653
 654If this is set to non-zero, the OOM killer simply kills the task that
 655triggered the out-of-memory condition.  This avoids the expensive
 656tasklist scan.
 657
 658If panic_on_oom is selected, it takes precedence over whatever value
 659is used in oom_kill_allocating_task.
 660
 661The default value is 0.
 662
 663==============================================================
 664
 665overcommit_kbytes:
 666
 667When overcommit_memory is set to 2, the committed address space is not
 668permitted to exceed swap plus this amount of physical RAM. See below.
 669
 670Note: overcommit_kbytes is the counterpart of overcommit_ratio. Only one
 671of them may be specified at a time. Setting one disables the other (which
 672then appears as 0 when read).
 673
 674==============================================================
 675
 676overcommit_memory:
 677
 678This value contains a flag that enables memory overcommitment.
 679
 680When this flag is 0, the kernel attempts to estimate the amount
 681of free memory left when userspace requests more memory.
 682
 683When this flag is 1, the kernel pretends there is always enough
 684memory until it actually runs out.
 685
 686When this flag is 2, the kernel uses a "never overcommit"
 687policy that attempts to prevent any overcommit of memory.
 688Note that user_reserve_kbytes affects this policy.
 689
 690This feature can be very useful because there are a lot of
 691programs that malloc() huge amounts of memory "just-in-case"
 692and don't use much of it.
 693
 694The default value is 0.
 695
 696See Documentation/vm/overcommit-accounting and
 697mm/mmap.c::__vm_enough_memory() for more information.
 698
 699==============================================================
 700
 701overcommit_ratio:
 702
 703When overcommit_memory is set to 2, the committed address
 704space is not permitted to exceed swap plus this percentage
 705of physical RAM.  See above.
 706
 707==============================================================
 708
 709page-cluster
 710
 711page-cluster controls the number of pages up to which consecutive pages
 712are read in from swap in a single attempt. This is the swap counterpart
 713to page cache readahead.
 714The mentioned consecutivity is not in terms of virtual/physical addresses,
 715but consecutive on swap space - that means they were swapped out together.
 716
 717It is a logarithmic value - setting it to zero means "1 page", setting
 718it to 1 means "2 pages", setting it to 2 means "4 pages", etc.
 719Zero disables swap readahead completely.
 720
 721The default value is three (eight pages at a time).  There may be some
 722small benefits in tuning this to a different value if your workload is
 723swap-intensive.
 724
 725Lower values mean lower latencies for initial faults, but at the same time
 726extra faults and I/O delays for following faults if they would have been part of
 727that consecutive pages readahead would have brought in.
 728
 729=============================================================
 730
 731panic_on_oom
 732
 733This enables or disables panic on out-of-memory feature.
 734
 735If this is set to 0, the kernel will kill some rogue process,
 736called oom_killer.  Usually, oom_killer can kill rogue processes and
 737system will survive.
 738
 739If this is set to 1, the kernel panics when out-of-memory happens.
 740However, if a process limits using nodes by mempolicy/cpusets,
 741and those nodes become memory exhaustion status, one process
 742may be killed by oom-killer. No panic occurs in this case.
 743Because other nodes' memory may be free. This means system total status
 744may be not fatal yet.
 745
 746If this is set to 2, the kernel panics compulsorily even on the
 747above-mentioned. Even oom happens under memory cgroup, the whole
 748system panics.
 749
 750The default value is 0.
 7511 and 2 are for failover of clustering. Please select either
 752according to your policy of failover.
 753panic_on_oom=2+kdump gives you very strong tool to investigate
 754why oom happens. You can get snapshot.
 755
 756=============================================================
 757
 758percpu_pagelist_fraction
 759
 760This is the fraction of pages at most (high mark pcp->high) in each zone that
 761are allocated for each per cpu page list.  The min value for this is 8.  It
 762means that we don't allow more than 1/8th of pages in each zone to be
 763allocated in any single per_cpu_pagelist.  This entry only changes the value
 764of hot per cpu pagelists.  User can specify a number like 100 to allocate
 7651/100th of each zone to each per cpu page list.
 766
 767The batch value of each per cpu pagelist is also updated as a result.  It is
 768set to pcp->high/4.  The upper limit of batch is (PAGE_SHIFT * 8)
 769
 770The initial value is zero.  Kernel does not use this value at boot time to set
 771the high water marks for each per cpu page list.  If the user writes '0' to this
 772sysctl, it will revert to this default behavior.
 773
 774==============================================================
 775
 776stat_interval
 777
 778The time interval between which vm statistics are updated.  The default
 779is 1 second.
 780
 781==============================================================
 782
 783stat_refresh
 784
 785Any read or write (by root only) flushes all the per-cpu vm statistics
 786into their global totals, for more accurate reports when testing
 787e.g. cat /proc/sys/vm/stat_refresh /proc/meminfo
 788
 789As a side-effect, it also checks for negative totals (elsewhere reported
 790as 0) and "fails" with EINVAL if any are found, with a warning in dmesg.
 791(At time of writing, a few stats are known sometimes to be found negative,
 792with no ill effects: errors and warnings on these stats are suppressed.)
 793
 794==============================================================
 795
 796numa_stat
 797
 798This interface allows runtime configuration of numa statistics.
 799
 800When page allocation performance becomes a bottleneck and you can tolerate
 801some possible tool breakage and decreased numa counter precision, you can
 802do:
 803        echo 0 > /proc/sys/vm/numa_stat
 804
 805When page allocation performance is not a bottleneck and you want all
 806tooling to work, you can do:
 807        echo 1 > /proc/sys/vm/numa_stat
 808
 809==============================================================
 810
 811swappiness
 812
 813This control is used to define how aggressive the kernel will swap
 814memory pages.  Higher values will increase aggressiveness, lower values
 815decrease the amount of swap.  A value of 0 instructs the kernel not to
 816initiate swap until the amount of free and file-backed pages is less
 817than the high water mark in a zone.
 818
 819The default value is 60.
 820
 821==============================================================
 822
 823- user_reserve_kbytes
 824
 825When overcommit_memory is set to 2, "never overcommit" mode, reserve
 826min(3% of current process size, user_reserve_kbytes) of free memory.
 827This is intended to prevent a user from starting a single memory hogging
 828process, such that they cannot recover (kill the hog).
 829
 830user_reserve_kbytes defaults to min(3% of the current process size, 128MB).
 831
 832If this is reduced to zero, then the user will be allowed to allocate
 833all free memory with a single process, minus admin_reserve_kbytes.
 834Any subsequent attempts to execute a command will result in
 835"fork: Cannot allocate memory".
 836
 837Changing this takes effect whenever an application requests memory.
 838
 839==============================================================
 840
 841vfs_cache_pressure
 842------------------
 843
 844This percentage value controls the tendency of the kernel to reclaim
 845the memory which is used for caching of directory and inode objects.
 846
 847At the default value of vfs_cache_pressure=100 the kernel will attempt to
 848reclaim dentries and inodes at a "fair" rate with respect to pagecache and
 849swapcache reclaim.  Decreasing vfs_cache_pressure causes the kernel to prefer
 850to retain dentry and inode caches. When vfs_cache_pressure=0, the kernel will
 851never reclaim dentries and inodes due to memory pressure and this can easily
 852lead to out-of-memory conditions. Increasing vfs_cache_pressure beyond 100
 853causes the kernel to prefer to reclaim dentries and inodes.
 854
 855Increasing vfs_cache_pressure significantly beyond 100 may have negative
 856performance impact. Reclaim code needs to take various locks to find freeable
 857directory and inode objects. With vfs_cache_pressure=1000, it will look for
 858ten times more freeable objects than there are.
 859
 860=============================================================
 861
 862watermark_scale_factor:
 863
 864This factor controls the aggressiveness of kswapd. It defines the
 865amount of memory left in a node/system before kswapd is woken up and
 866how much memory needs to be free before kswapd goes back to sleep.
 867
 868The unit is in fractions of 10,000. The default value of 10 means the
 869distances between watermarks are 0.1% of the available memory in the
 870node/system. The maximum value is 1000, or 10% of memory.
 871
 872A high rate of threads entering direct reclaim (allocstall) or kswapd
 873going to sleep prematurely (kswapd_low_wmark_hit_quickly) can indicate
 874that the number of free pages kswapd maintains for latency reasons is
 875too small for the allocation bursts occurring in the system. This knob
 876can then be used to tune kswapd aggressiveness accordingly.
 877
 878==============================================================
 879
 880zone_reclaim_mode:
 881
 882Zone_reclaim_mode allows someone to set more or less aggressive approaches to
 883reclaim memory when a zone runs out of memory. If it is set to zero then no
 884zone reclaim occurs. Allocations will be satisfied from other zones / nodes
 885in the system.
 886
 887This is value ORed together of
 888
 8891       = Zone reclaim on
 8902       = Zone reclaim writes dirty pages out
 8914       = Zone reclaim swaps pages
 892
 893zone_reclaim_mode is disabled by default.  For file servers or workloads
 894that benefit from having their data cached, zone_reclaim_mode should be
 895left disabled as the caching effect is likely to be more important than
 896data locality.
 897
 898zone_reclaim may be enabled if it's known that the workload is partitioned
 899such that each partition fits within a NUMA node and that accessing remote
 900memory would cause a measurable performance reduction.  The page allocator
 901will then reclaim easily reusable pages (those page cache pages that are
 902currently not used) before allocating off node pages.
 903
 904Allowing zone reclaim to write out pages stops processes that are
 905writing large amounts of data from dirtying pages on other nodes. Zone
 906reclaim will write out dirty pages if a zone fills up and so effectively
 907throttle the process. This may decrease the performance of a single process
 908since it cannot use all of system memory to buffer the outgoing writes
 909anymore but it preserve the memory on other nodes so that the performance
 910of other processes running on other nodes will not be affected.
 911
 912Allowing regular swap effectively restricts allocations to the local
 913node unless explicitly overridden by memory policies or cpuset
 914configurations.
 915
 916============ End of Document =================================
 917
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.