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- dirty_background_bytes
  25- dirty_background_ratio
  26- dirty_bytes
  27- dirty_expire_centisecs
  28- dirty_ratio
  29- dirty_writeback_centisecs
  30- drop_caches
  31- extfrag_threshold
  32- hugepages_treat_as_movable
  33- hugetlb_shm_group
  34- laptop_mode
  35- legacy_va_layout
  36- lowmem_reserve_ratio
  37- max_map_count
  38- memory_failure_early_kill
  39- memory_failure_recovery
  40- min_free_kbytes
  41- min_slab_ratio
  42- min_unmapped_ratio
  43- mmap_min_addr
  44- nr_hugepages
  45- nr_overcommit_hugepages
  46- nr_trim_pages         (only if CONFIG_MMU=n)
  47- numa_zonelist_order
  48- oom_dump_tasks
  49- oom_kill_allocating_task
  50- overcommit_kbytes
  51- overcommit_memory
  52- overcommit_ratio
  53- page-cluster
  54- panic_on_oom
  55- percpu_pagelist_fraction
  56- stat_interval
  57- swappiness
  58- user_reserve_kbytes
  59- vfs_cache_pressure
  60- zone_reclaim_mode
  61
  62==============================================================
  63
  64admin_reserve_kbytes
  65
  66The amount of free memory in the system that should be reserved for users
  67with the capability cap_sys_admin.
  68
  69admin_reserve_kbytes defaults to min(3% of free pages, 8MB)
  70
  71That should provide enough for the admin to log in and kill a process,
  72if necessary, under the default overcommit 'guess' mode.
  73
  74Systems running under overcommit 'never' should increase this to account
  75for the full Virtual Memory Size of programs used to recover. Otherwise,
  76root may not be able to log in to recover the system.
  77
  78How do you calculate a minimum useful reserve?
  79
  80sshd or login + bash (or some other shell) + top (or ps, kill, etc.)
  81
  82For overcommit 'guess', we can sum resident set sizes (RSS).
  83On x86_64 this is about 8MB.
  84
  85For overcommit 'never', we can take the max of their virtual sizes (VSZ)
  86and add the sum of their RSS.
  87On x86_64 this is about 128MB.
  88
  89Changing this takes effect whenever an application requests memory.
  90
  91==============================================================
  92
  93block_dump
  94
  95block_dump enables block I/O debugging when set to a nonzero value. More
  96information on block I/O debugging is in Documentation/laptops/laptop-mode.txt.
  97
  98==============================================================
  99
 100compact_memory
 101
 102Available only when CONFIG_COMPACTION is set. When 1 is written to the file,
 103all zones are compacted such that free memory is available in contiguous
 104blocks where possible. This can be important for example in the allocation of
 105huge pages although processes will also directly compact memory as required.
 106
 107==============================================================
 108
 109dirty_background_bytes
 110
 111Contains the amount of dirty memory at which the background kernel
 112flusher threads will start writeback.
 113
 114Note: dirty_background_bytes is the counterpart of dirty_background_ratio. Only
 115one of them may be specified at a time. When one sysctl is written it is
 116immediately taken into account to evaluate the dirty memory limits and the
 117other appears as 0 when read.
 118
 119==============================================================
 120
 121dirty_background_ratio
 122
 123Contains, as a percentage of total available memory that contains free pages
 124and reclaimable pages, the number of pages at which the background kernel
 125flusher threads will start writing out dirty data.
 126
 127The total avaiable memory is not equal to total system memory.
 128
 129==============================================================
 130
 131dirty_bytes
 132
 133Contains the amount of dirty memory at which a process generating disk writes
 134will itself start writeback.
 135
 136Note: dirty_bytes is the counterpart of dirty_ratio. Only one of them may be
 137specified at a time. When one sysctl is written it is immediately taken into
 138account to evaluate the dirty memory limits and the other appears as 0 when
 139read.
 140
 141Note: the minimum value allowed for dirty_bytes is two pages (in bytes); any
 142value lower than this limit will be ignored and the old configuration will be
 143retained.
 144
 145==============================================================
 146
 147dirty_expire_centisecs
 148
 149This tunable is used to define when dirty data is old enough to be eligible
 150for writeout by the kernel flusher threads.  It is expressed in 100'ths
 151of a second.  Data which has been dirty in-memory for longer than this
 152interval will be written out next time a flusher thread wakes up.
 153
 154==============================================================
 155
 156dirty_ratio
 157
 158Contains, as a percentage of total available memory that contains free pages
 159and reclaimable pages, the number of pages at which a process which is
 160generating disk writes will itself start writing out dirty data.
 161
 162The total avaiable memory is not equal to total system memory.
 163
 164==============================================================
 165
 166dirty_writeback_centisecs
 167
 168The kernel flusher threads will periodically wake up and write `old' data
 169out to disk.  This tunable expresses the interval between those wakeups, in
 170100'ths of a second.
 171
 172Setting this to zero disables periodic writeback altogether.
 173
 174==============================================================
 175
 176drop_caches
 177
 178Writing to this will cause the kernel to drop clean caches, dentries and
 179inodes from memory, causing that memory to become free.
 180
 181To free pagecache:
 182        echo 1 > /proc/sys/vm/drop_caches
 183To free dentries and inodes:
 184        echo 2 > /proc/sys/vm/drop_caches
 185To free pagecache, dentries and inodes:
 186        echo 3 > /proc/sys/vm/drop_caches
 187
 188As this is a non-destructive operation and dirty objects are not freeable, the
 189user should run `sync' first.
 190
 191==============================================================
 192
 193extfrag_threshold
 194
 195This parameter affects whether the kernel will compact memory or direct
 196reclaim to satisfy a high-order allocation. /proc/extfrag_index shows what
 197the fragmentation index for each order is in each zone in the system. Values
 198tending towards 0 imply allocations would fail due to lack of memory,
 199values towards 1000 imply failures are due to fragmentation and -1 implies
 200that the allocation will succeed as long as watermarks are met.
 201
 202The kernel will not compact memory in a zone if the
 203fragmentation index is <= extfrag_threshold. The default value is 500.
 204
 205==============================================================
 206
 207hugepages_treat_as_movable
 208
 209This parameter controls whether we can allocate hugepages from ZONE_MOVABLE
 210or not. If set to non-zero, hugepages can be allocated from ZONE_MOVABLE.
 211ZONE_MOVABLE is created when kernel boot parameter kernelcore= is specified,
 212so this parameter has no effect if used without kernelcore=.
 213
 214Hugepage migration is now available in some situations which depend on the
 215architecture and/or the hugepage size. If a hugepage supports migration,
 216allocation from ZONE_MOVABLE is always enabled for the hugepage regardless
 217of the value of this parameter.
 218IOW, this parameter affects only non-migratable hugepages.
 219
 220Assuming that hugepages are not migratable in your system, one usecase of
 221this parameter is that users can make hugepage pool more extensible by
 222enabling the allocation from ZONE_MOVABLE. This is because on ZONE_MOVABLE
 223page reclaim/migration/compaction work more and you can get contiguous
 224memory more likely. Note that using ZONE_MOVABLE for non-migratable
 225hugepages can do harm to other features like memory hotremove (because
 226memory hotremove expects that memory blocks on ZONE_MOVABLE are always
 227removable,) so it's a trade-off responsible for the users.
 228
 229==============================================================
 230
 231hugetlb_shm_group
 232
 233hugetlb_shm_group contains group id that is allowed to create SysV
 234shared memory segment using hugetlb page.
 235
 236==============================================================
 237
 238laptop_mode
 239
 240laptop_mode is a knob that controls "laptop mode". All the things that are
 241controlled by this knob are discussed in Documentation/laptops/laptop-mode.txt.
 242
 243==============================================================
 244
 245legacy_va_layout
 246
 247If non-zero, this sysctl disables the new 32-bit mmap layout - the kernel
 248will use the legacy (2.4) layout for all processes.
 249
 250==============================================================
 251
 252lowmem_reserve_ratio
 253
 254For some specialised workloads on highmem machines it is dangerous for
 255the kernel to allow process memory to be allocated from the "lowmem"
 256zone.  This is because that memory could then be pinned via the mlock()
 257system call, or by unavailability of swapspace.
 258
 259And on large highmem machines this lack of reclaimable lowmem memory
 260can be fatal.
 261
 262So the Linux page allocator has a mechanism which prevents allocations
 263which _could_ use highmem from using too much lowmem.  This means that
 264a certain amount of lowmem is defended from the possibility of being
 265captured into pinned user memory.
 266
 267(The same argument applies to the old 16 megabyte ISA DMA region.  This
 268mechanism will also defend that region from allocations which could use
 269highmem or lowmem).
 270
 271The `lowmem_reserve_ratio' tunable determines how aggressive the kernel is
 272in defending these lower zones.
 273
 274If you have a machine which uses highmem or ISA DMA and your
 275applications are using mlock(), or if you are running with no swap then
 276you probably should change the lowmem_reserve_ratio setting.
 277
 278The lowmem_reserve_ratio is an array. You can see them by reading this file.
 279-
 280% cat /proc/sys/vm/lowmem_reserve_ratio
 281256     256     32
 282-
 283Note: # of this elements is one fewer than number of zones. Because the highest
 284      zone's value is not necessary for following calculation.
 285
 286But, these values are not used directly. The kernel calculates # of protection
 287pages for each zones from them. These are shown as array of protection pages
 288in /proc/zoneinfo like followings. (This is an example of x86-64 box).
 289Each zone has an array of protection pages like this.
 290
 291-
 292Node 0, zone      DMA
 293  pages free     1355
 294        min      3
 295        low      3
 296        high     4
 297        :
 298        :
 299    numa_other   0
 300        protection: (0, 2004, 2004, 2004)
 301        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 302  pagesets
 303    cpu: 0 pcp: 0
 304        :
 305-
 306These protections are added to score to judge whether this zone should be used
 307for page allocation or should be reclaimed.
 308
 309In this example, if normal pages (index=2) are required to this DMA zone and
 310watermark[WMARK_HIGH] is used for watermark, the kernel judges this zone should
 311not be used because pages_free(1355) is smaller than watermark + protection[2]
 312(4 + 2004 = 2008). If this protection value is 0, this zone would be used for
 313normal page requirement. If requirement is DMA zone(index=0), protection[0]
 314(=0) is used.
 315
 316zone[i]'s protection[j] is calculated by following expression.
 317
 318(i < j):
 319  zone[i]->protection[j]
 320  = (total sums of present_pages from zone[i+1] to zone[j] on the node)
 321    / lowmem_reserve_ratio[i];
 322(i = j):
 323   (should not be protected. = 0;
 324(i > j):
 325   (not necessary, but looks 0)
 326
 327The default values of lowmem_reserve_ratio[i] are
 328    256 (if zone[i] means DMA or DMA32 zone)
 329    32  (others).
 330As above expression, they are reciprocal number of ratio.
 331256 means 1/256. # of protection pages becomes about "0.39%" of total present
 332pages of higher zones on the node.
 333
 334If you would like to protect more pages, smaller values are effective.
 335The minimum value is 1 (1/1 -> 100%).
 336
 337==============================================================
 338
 339max_map_count:
 340
 341This file contains the maximum number of memory map areas a process
 342may have. Memory map areas are used as a side-effect of calling
 343malloc, directly by mmap and mprotect, and also when loading shared
 344libraries.
 345
 346While most applications need less than a thousand maps, certain
 347programs, particularly malloc debuggers, may consume lots of them,
 348e.g., up to one or two maps per allocation.
 349
 350The default value is 65536.
 351
 352=============================================================
 353
 354memory_failure_early_kill:
 355
 356Control how to kill processes when uncorrected memory error (typically
 357a 2bit error in a memory module) is detected in the background by hardware
 358that cannot be handled by the kernel. In some cases (like the page
 359still having a valid copy on disk) the kernel will handle the failure
 360transparently without affecting any applications. But if there is
 361no other uptodate copy of the data it will kill to prevent any data
 362corruptions from propagating.
 363
 3641: Kill all processes that have the corrupted and not reloadable page mapped
 365as soon as the corruption is detected.  Note this is not supported
 366for a few types of pages, like kernel internally allocated data or
 367the swap cache, but works for the majority of user pages.
 368
 3690: Only unmap the corrupted page from all processes and only kill a process
 370who tries to access it.
 371
 372The kill is done using a catchable SIGBUS with BUS_MCEERR_AO, so processes can
 373handle this if they want to.
 374
 375This is only active on architectures/platforms with advanced machine
 376check handling and depends on the hardware capabilities.
 377
 378Applications can override this setting individually with the PR_MCE_KILL prctl
 379
 380==============================================================
 381
 382memory_failure_recovery
 383
 384Enable memory failure recovery (when supported by the platform)
 385
 3861: Attempt recovery.
 387
 3880: Always panic on a memory failure.
 389
 390==============================================================
 391
 392min_free_kbytes:
 393
 394This is used to force the Linux VM to keep a minimum number
 395of kilobytes free.  The VM uses this number to compute a
 396watermark[WMARK_MIN] value for each lowmem zone in the system.
 397Each lowmem zone gets a number of reserved free pages based
 398proportionally on its size.
 399
 400Some minimal amount of memory is needed to satisfy PF_MEMALLOC
 401allocations; if you set this to lower than 1024KB, your system will
 402become subtly broken, and prone to deadlock under high loads.
 403
 404Setting this too high will OOM your machine instantly.
 405
 406=============================================================
 407
 408min_slab_ratio:
 409
 410This is available only on NUMA kernels.
 411
 412A percentage of the total pages in each zone.  On Zone reclaim
 413(fallback from the local zone occurs) slabs will be reclaimed if more
 414than this percentage of pages in a zone are reclaimable slab pages.
 415This insures that the slab growth stays under control even in NUMA
 416systems that rarely perform global reclaim.
 417
 418The default is 5 percent.
 419
 420Note that slab reclaim is triggered in a per zone / node fashion.
 421The process of reclaiming slab memory is currently not node specific
 422and may not be fast.
 423
 424=============================================================
 425
 426min_unmapped_ratio:
 427
 428This is available only on NUMA kernels.
 429
 430This is a percentage of the total pages in each zone. Zone reclaim will
 431only occur if more than this percentage of pages are in a state that
 432zone_reclaim_mode allows to be reclaimed.
 433
 434If zone_reclaim_mode has the value 4 OR'd, then the percentage is compared
 435against all file-backed unmapped pages including swapcache pages and tmpfs
 436files. Otherwise, only unmapped pages backed by normal files but not tmpfs
 437files and similar are considered.
 438
 439The default is 1 percent.
 440
 441==============================================================
 442
 443mmap_min_addr
 444
 445This file indicates the amount of address space  which a user process will
 446be restricted from mmapping.  Since kernel null dereference bugs could
 447accidentally operate based on the information in the first couple of pages
 448of memory userspace processes should not be allowed to write to them.  By
 449default this value is set to 0 and no protections will be enforced by the
 450security module.  Setting this value to something like 64k will allow the
 451vast majority of applications to work correctly and provide defense in depth
 452against future potential kernel bugs.
 453
 454==============================================================
 455
 456nr_hugepages
 457
 458Change the minimum size of the hugepage pool.
 459
 460See Documentation/vm/hugetlbpage.txt
 461
 462==============================================================
 463
 464nr_overcommit_hugepages
 465
 466Change the maximum size of the hugepage pool. The maximum is
 467nr_hugepages + nr_overcommit_hugepages.
 468
 469See Documentation/vm/hugetlbpage.txt
 470
 471==============================================================
 472
 473nr_trim_pages
 474
 475This is available only on NOMMU kernels.
 476
 477This value adjusts the excess page trimming behaviour of power-of-2 aligned
 478NOMMU mmap allocations.
 479
 480A value of 0 disables trimming of allocations entirely, while a value of 1
 481trims excess pages aggressively. Any value >= 1 acts as the watermark where
 482trimming of allocations is initiated.
 483
 484The default value is 1.
 485
 486See Documentation/nommu-mmap.txt for more information.
 487
 488==============================================================
 489
 490numa_zonelist_order
 491
 492This sysctl is only for NUMA.
 493'where the memory is allocated from' is controlled by zonelists.
 494(This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation.
 495 you may be able to read ZONE_DMA as ZONE_DMA32...)
 496
 497In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following.
 498ZONE_NORMAL -> ZONE_DMA
 499This means that a memory allocation request for GFP_KERNEL will
 500get memory from ZONE_DMA only when ZONE_NORMAL is not available.
 501
 502In NUMA case, you can think of following 2 types of order.
 503Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL
 504
 505(A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL
 506(B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA.
 507
 508Type(A) offers the best locality for processes on Node(0), but ZONE_DMA
 509will be used before ZONE_NORMAL exhaustion. This increases possibility of
 510out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small.
 511
 512Type(B) cannot offer the best locality but is more robust against OOM of
 513the DMA zone.
 514
 515Type(A) is called as "Node" order. Type (B) is "Zone" order.
 516
 517"Node order" orders the zonelists by node, then by zone within each node.
 518Specify "[Nn]ode" for node order
 519
 520"Zone Order" orders the zonelists by zone type, then by node within each
 521zone.  Specify "[Zz]one" for zone order.
 522
 523Specify "[Dd]efault" to request automatic configuration.  Autoconfiguration
 524will select "node" order in following case.
 525(1) if the DMA zone does not exist or
 526(2) if the DMA zone comprises greater than 50% of the available memory or
 527(3) if any node's DMA zone comprises greater than 70% of its local memory and
 528    the amount of local memory is big enough.
 529
 530Otherwise, "zone" order will be selected. Default order is recommended unless
 531this is causing problems for your system/application.
 532
 533==============================================================
 534
 535oom_dump_tasks
 536
 537Enables a system-wide task dump (excluding kernel threads) to be
 538produced when the kernel performs an OOM-killing and includes such
 539information as pid, uid, tgid, vm size, rss, nr_ptes, swapents,
 540oom_score_adj score, and name.  This is helpful to determine why the
 541OOM killer was invoked, to identify the rogue task that caused it,
 542and to determine why the OOM killer chose the task it did to kill.
 543
 544If this is set to zero, this information is suppressed.  On very
 545large systems with thousands of tasks it may not be feasible to dump
 546the memory state information for each one.  Such systems should not
 547be forced to incur a performance penalty in OOM conditions when the
 548information may not be desired.
 549
 550If this is set to non-zero, this information is shown whenever the
 551OOM killer actually kills a memory-hogging task.
 552
 553The default value is 1 (enabled).
 554
 555==============================================================
 556
 557oom_kill_allocating_task
 558
 559This enables or disables killing the OOM-triggering task in
 560out-of-memory situations.
 561
 562If this is set to zero, the OOM killer will scan through the entire
 563tasklist and select a task based on heuristics to kill.  This normally
 564selects a rogue memory-hogging task that frees up a large amount of
 565memory when killed.
 566
 567If this is set to non-zero, the OOM killer simply kills the task that
 568triggered the out-of-memory condition.  This avoids the expensive
 569tasklist scan.
 570
 571If panic_on_oom is selected, it takes precedence over whatever value
 572is used in oom_kill_allocating_task.
 573
 574The default value is 0.
 575
 576==============================================================
 577
 578overcommit_kbytes:
 579
 580When overcommit_memory is set to 2, the committed address space is not
 581permitted to exceed swap plus this amount of physical RAM. See below.
 582
 583Note: overcommit_kbytes is the counterpart of overcommit_ratio. Only one
 584of them may be specified at a time. Setting one disables the other (which
 585then appears as 0 when read).
 586
 587==============================================================
 588
 589overcommit_memory:
 590
 591This value contains a flag that enables memory overcommitment.
 592
 593When this flag is 0, the kernel attempts to estimate the amount
 594of free memory left when userspace requests more memory.
 595
 596When this flag is 1, the kernel pretends there is always enough
 597memory until it actually runs out.
 598
 599When this flag is 2, the kernel uses a "never overcommit"
 600policy that attempts to prevent any overcommit of memory.
 601Note that user_reserve_kbytes affects this policy.
 602
 603This feature can be very useful because there are a lot of
 604programs that malloc() huge amounts of memory "just-in-case"
 605and don't use much of it.
 606
 607The default value is 0.
 608
 609See Documentation/vm/overcommit-accounting and
 610security/commoncap.c::cap_vm_enough_memory() for more information.
 611
 612==============================================================
 613
 614overcommit_ratio:
 615
 616When overcommit_memory is set to 2, the committed address
 617space is not permitted to exceed swap plus this percentage
 618of physical RAM.  See above.
 619
 620==============================================================
 621
 622page-cluster
 623
 624page-cluster controls the number of pages up to which consecutive pages
 625are read in from swap in a single attempt. This is the swap counterpart
 626to page cache readahead.
 627The mentioned consecutivity is not in terms of virtual/physical addresses,
 628but consecutive on swap space - that means they were swapped out together.
 629
 630It is a logarithmic value - setting it to zero means "1 page", setting
 631it to 1 means "2 pages", setting it to 2 means "4 pages", etc.
 632Zero disables swap readahead completely.
 633
 634The default value is three (eight pages at a time).  There may be some
 635small benefits in tuning this to a different value if your workload is
 636swap-intensive.
 637
 638Lower values mean lower latencies for initial faults, but at the same time
 639extra faults and I/O delays for following faults if they would have been part of
 640that consecutive pages readahead would have brought in.
 641
 642=============================================================
 643
 644panic_on_oom
 645
 646This enables or disables panic on out-of-memory feature.
 647
 648If this is set to 0, the kernel will kill some rogue process,
 649called oom_killer.  Usually, oom_killer can kill rogue processes and
 650system will survive.
 651
 652If this is set to 1, the kernel panics when out-of-memory happens.
 653However, if a process limits using nodes by mempolicy/cpusets,
 654and those nodes become memory exhaustion status, one process
 655may be killed by oom-killer. No panic occurs in this case.
 656Because other nodes' memory may be free. This means system total status
 657may be not fatal yet.
 658
 659If this is set to 2, the kernel panics compulsorily even on the
 660above-mentioned. Even oom happens under memory cgroup, the whole
 661system panics.
 662
 663The default value is 0.
 6641 and 2 are for failover of clustering. Please select either
 665according to your policy of failover.
 666panic_on_oom=2+kdump gives you very strong tool to investigate
 667why oom happens. You can get snapshot.
 668
 669=============================================================
 670
 671percpu_pagelist_fraction
 672
 673This is the fraction of pages at most (high mark pcp->high) in each zone that
 674are allocated for each per cpu page list.  The min value for this is 8.  It
 675means that we don't allow more than 1/8th of pages in each zone to be
 676allocated in any single per_cpu_pagelist.  This entry only changes the value
 677of hot per cpu pagelists.  User can specify a number like 100 to allocate
 6781/100th of each zone to each per cpu page list.
 679
 680The batch value of each per cpu pagelist is also updated as a result.  It is
 681set to pcp->high/4.  The upper limit of batch is (PAGE_SHIFT * 8)
 682
 683The initial value is zero.  Kernel does not use this value at boot time to set
 684the high water marks for each per cpu page list.
 685
 686==============================================================
 687
 688stat_interval
 689
 690The time interval between which vm statistics are updated.  The default
 691is 1 second.
 692
 693==============================================================
 694
 695swappiness
 696
 697This control is used to define how aggressive the kernel will swap
 698memory pages.  Higher values will increase agressiveness, lower values
 699decrease the amount of swap.  A value of 0 instructs the kernel not to
 700initiate swap until the amount of free and file-backed pages is less
 701than the high water mark in a zone.
 702
 703The default value is 60.
 704
 705==============================================================
 706
 707- user_reserve_kbytes
 708
 709When overcommit_memory is set to 2, "never overommit" mode, reserve
 710min(3% of current process size, user_reserve_kbytes) of free memory.
 711This is intended to prevent a user from starting a single memory hogging
 712process, such that they cannot recover (kill the hog).
 713
 714user_reserve_kbytes defaults to min(3% of the current process size, 128MB).
 715
 716If this is reduced to zero, then the user will be allowed to allocate
 717all free memory with a single process, minus admin_reserve_kbytes.
 718Any subsequent attempts to execute a command will result in
 719"fork: Cannot allocate memory".
 720
 721Changing this takes effect whenever an application requests memory.
 722
 723==============================================================
 724
 725vfs_cache_pressure
 726------------------
 727
 728Controls the tendency of the kernel to reclaim the memory which is used for
 729caching of directory and inode objects.
 730
 731At the default value of vfs_cache_pressure=100 the kernel will attempt to
 732reclaim dentries and inodes at a "fair" rate with respect to pagecache and
 733swapcache reclaim.  Decreasing vfs_cache_pressure causes the kernel to prefer
 734to retain dentry and inode caches. When vfs_cache_pressure=0, the kernel will
 735never reclaim dentries and inodes due to memory pressure and this can easily
 736lead to out-of-memory conditions. Increasing vfs_cache_pressure beyond 100
 737causes the kernel to prefer to reclaim dentries and inodes.
 738
 739==============================================================
 740
 741zone_reclaim_mode:
 742
 743Zone_reclaim_mode allows someone to set more or less aggressive approaches to
 744reclaim memory when a zone runs out of memory. If it is set to zero then no
 745zone reclaim occurs. Allocations will be satisfied from other zones / nodes
 746in the system.
 747
 748This is value ORed together of
 749
 7501       = Zone reclaim on
 7512       = Zone reclaim writes dirty pages out
 7524       = Zone reclaim swaps pages
 753
 754zone_reclaim_mode is set during bootup to 1 if it is determined that pages
 755from remote zones will cause a measurable performance reduction. The
 756page allocator will then reclaim easily reusable pages (those page
 757cache pages that are currently not used) before allocating off node pages.
 758
 759It may be beneficial to switch off zone reclaim if the system is
 760used for a file server and all of memory should be used for caching files
 761from disk. In that case the caching effect is more important than
 762data locality.
 763
 764Allowing zone reclaim to write out pages stops processes that are
 765writing large amounts of data from dirtying pages on other nodes. Zone
 766reclaim will write out dirty pages if a zone fills up and so effectively
 767throttle the process. This may decrease the performance of a single process
 768since it cannot use all of system memory to buffer the outgoing writes
 769anymore but it preserve the memory on other nodes so that the performance
 770of other processes running on other nodes will not be affected.
 771
 772Allowing regular swap effectively restricts allocations to the local
 773node unless explicitly overridden by memory policies or cpuset
 774configurations.
 775
 776============ End of Document =================================
 777
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.