1 Documentation 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>
5 For general info and legal blurb, please look in README.
7 ==============================================================
9 This file contains the documentation for the sysctl files in
10 /proc/sys/vm and is valid for Linux kernel version 2.6.29.
12 The files in this directory can be used to tune the operation
13 of the virtual memory (VM) subsystem of the Linux kernel and
14 the writeout of dirty data to disk.
16 Default values and initialization routines for most of these
17 files can be found in mm/swap.c.
19 Currently, these files are in /proc/sys/vm:
23 - dirty_background_bytes
24 - dirty_background_ratio
26 - dirty_expire_centisecs
28 - dirty_writeback_centisecs
31 - hugepages_treat_as_movable
35 - lowmem_reserve_ratio
37 - memory_failure_early_kill
38 - memory_failure_recovery
44 - nr_overcommit_hugepages
45 - nr_trim_pages (only if CONFIG_MMU=n)
48 - oom_kill_allocating_task
53 - percpu_pagelist_fraction
60 ==============================================================
64 block_dump enables block I/O debugging when set to a nonzero value. More
65 information on block I/O debugging is in Documentation/laptops/laptop-mode.txt.
67 ==============================================================
71 Available only when CONFIG_COMPACTION is set. When 1 is written to the file,
72 all zones are compacted such that free memory is available in contiguous
73 blocks where possible. This can be important for example in the allocation of
74 huge pages although processes will also directly compact memory as required.
76 ==============================================================
78 dirty_background_bytes
80 Contains the amount of dirty memory at which the background kernel
81 flusher threads will start writeback.
83 Note: dirty_background_bytes is the counterpart of dirty_background_ratio. Only
84 one of them may be specified at a time. When one sysctl is written it is
85 immediately taken into account to evaluate the dirty memory limits and the
86 other appears as 0 when read.
88 ==============================================================
90 dirty_background_ratio
92 Contains, as a percentage of total system memory, the number of pages at which
93 the background kernel flusher threads will start writing out dirty data.
95 ==============================================================
99 Contains the amount of dirty memory at which a process generating disk writes
100 will itself start writeback.
102 Note: dirty_bytes is the counterpart of dirty_ratio. Only one of them may be
103 specified at a time. When one sysctl is written it is immediately taken into
104 account to evaluate the dirty memory limits and the other appears as 0 when
107 Note: the minimum value allowed for dirty_bytes is two pages (in bytes); any
108 value lower than this limit will be ignored and the old configuration will be
111 ==============================================================
113 dirty_expire_centisecs
115 This tunable is used to define when dirty data is old enough to be eligible
116 for writeout by the kernel flusher threads. It is expressed in 100'ths
117 of a second. Data which has been dirty in-memory for longer than this
118 interval will be written out next time a flusher thread wakes up.
120 ==============================================================
124 Contains, as a percentage of total system memory, the number of pages at which
125 a process which is generating disk writes will itself start writing out dirty
128 ==============================================================
130 dirty_writeback_centisecs
132 The kernel flusher threads will periodically wake up and write `old' data
133 out to disk. This tunable expresses the interval between those wakeups, in
136 Setting this to zero disables periodic writeback altogether.
138 ==============================================================
142 Writing to this will cause the kernel to drop clean caches, dentries and
143 inodes from memory, causing that memory to become free.
146 echo 1 > /proc/sys/vm/drop_caches
147 To free dentries and inodes:
148 echo 2 > /proc/sys/vm/drop_caches
149 To free pagecache, dentries and inodes:
150 echo 3 > /proc/sys/vm/drop_caches
152 As this is a non-destructive operation and dirty objects are not freeable, the
153 user should run `sync' first.
155 ==============================================================
159 This parameter affects whether the kernel will compact memory or direct
160 reclaim to satisfy a high-order allocation. /proc/extfrag_index shows what
161 the fragmentation index for each order is in each zone in the system. Values
162 tending towards 0 imply allocations would fail due to lack of memory,
163 values towards 1000 imply failures are due to fragmentation and -1 implies
164 that the allocation will succeed as long as watermarks are met.
166 The kernel will not compact memory in a zone if the
167 fragmentation index is <= extfrag_threshold. The default value is 500.
169 ==============================================================
171 hugepages_treat_as_movable
173 This parameter is only useful when kernelcore= is specified at boot time to
174 create ZONE_MOVABLE for pages that may be reclaimed or migrated. Huge pages
175 are not movable so are not normally allocated from ZONE_MOVABLE. A non-zero
176 value written to hugepages_treat_as_movable allows huge pages to be allocated
179 Once enabled, the ZONE_MOVABLE is treated as an area of memory the huge
180 pages pool can easily grow or shrink within. Assuming that applications are
181 not running that mlock() a lot of memory, it is likely the huge pages pool
182 can grow to the size of ZONE_MOVABLE by repeatedly entering the desired value
183 into nr_hugepages and triggering page reclaim.
185 ==============================================================
189 hugetlb_shm_group contains group id that is allowed to create SysV
190 shared memory segment using hugetlb page.
192 ==============================================================
196 laptop_mode is a knob that controls "laptop mode". All the things that are
197 controlled by this knob are discussed in Documentation/laptops/laptop-mode.txt.
199 ==============================================================
203 If non-zero, this sysctl disables the new 32-bit mmap layout - the kernel
204 will use the legacy (2.4) layout for all processes.
206 ==============================================================
210 For some specialised workloads on highmem machines it is dangerous for
211 the kernel to allow process memory to be allocated from the "lowmem"
212 zone. This is because that memory could then be pinned via the mlock()
213 system call, or by unavailability of swapspace.
215 And on large highmem machines this lack of reclaimable lowmem memory
218 So the Linux page allocator has a mechanism which prevents allocations
219 which _could_ use highmem from using too much lowmem. This means that
220 a certain amount of lowmem is defended from the possibility of being
221 captured into pinned user memory.
223 (The same argument applies to the old 16 megabyte ISA DMA region. This
224 mechanism will also defend that region from allocations which could use
227 The `lowmem_reserve_ratio' tunable determines how aggressive the kernel is
228 in defending these lower zones.
230 If you have a machine which uses highmem or ISA DMA and your
231 applications are using mlock(), or if you are running with no swap then
232 you probably should change the lowmem_reserve_ratio setting.
234 The lowmem_reserve_ratio is an array. You can see them by reading this file.
236 % cat /proc/sys/vm/lowmem_reserve_ratio
239 Note: # of this elements is one fewer than number of zones. Because the highest
240 zone's value is not necessary for following calculation.
242 But, these values are not used directly. The kernel calculates # of protection
243 pages for each zones from them. These are shown as array of protection pages
244 in /proc/zoneinfo like followings. (This is an example of x86-64 box).
245 Each zone has an array of protection pages like this.
256 protection: (0, 2004, 2004, 2004)
257 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
262 These protections are added to score to judge whether this zone should be used
263 for page allocation or should be reclaimed.
265 In this example, if normal pages (index=2) are required to this DMA zone and
266 watermark[WMARK_HIGH] is used for watermark, the kernel judges this zone should
267 not be used because pages_free(1355) is smaller than watermark + protection[2]
268 (4 + 2004 = 2008). If this protection value is 0, this zone would be used for
269 normal page requirement. If requirement is DMA zone(index=0), protection[0]
272 zone[i]'s protection[j] is calculated by following expression.
275 zone[i]->protection[j]
276 = (total sums of present_pages from zone[i+1] to zone[j] on the node)
277 / lowmem_reserve_ratio[i];
279 (should not be protected. = 0;
281 (not necessary, but looks 0)
283 The default values of lowmem_reserve_ratio[i] are
284 256 (if zone[i] means DMA or DMA32 zone)
286 As above expression, they are reciprocal number of ratio.
287 256 means 1/256. # of protection pages becomes about "0.39%" of total present
288 pages of higher zones on the node.
290 If you would like to protect more pages, smaller values are effective.
291 The minimum value is 1 (1/1 -> 100%).
293 ==============================================================
297 This file contains the maximum number of memory map areas a process
298 may have. Memory map areas are used as a side-effect of calling
299 malloc, directly by mmap and mprotect, and also when loading shared
302 While most applications need less than a thousand maps, certain
303 programs, particularly malloc debuggers, may consume lots of them,
304 e.g., up to one or two maps per allocation.
306 The default value is 65536.
308 =============================================================
310 memory_failure_early_kill:
312 Control how to kill processes when uncorrected memory error (typically
313 a 2bit error in a memory module) is detected in the background by hardware
314 that cannot be handled by the kernel. In some cases (like the page
315 still having a valid copy on disk) the kernel will handle the failure
316 transparently without affecting any applications. But if there is
317 no other uptodate copy of the data it will kill to prevent any data
318 corruptions from propagating.
320 1: Kill all processes that have the corrupted and not reloadable page mapped
321 as soon as the corruption is detected. Note this is not supported
322 for a few types of pages, like kernel internally allocated data or
323 the swap cache, but works for the majority of user pages.
325 0: Only unmap the corrupted page from all processes and only kill a process
326 who tries to access it.
328 The kill is done using a catchable SIGBUS with BUS_MCEERR_AO, so processes can
329 handle this if they want to.
331 This is only active on architectures/platforms with advanced machine
332 check handling and depends on the hardware capabilities.
334 Applications can override this setting individually with the PR_MCE_KILL prctl
336 ==============================================================
338 memory_failure_recovery
340 Enable memory failure recovery (when supported by the platform)
344 0: Always panic on a memory failure.
346 ==============================================================
350 This is used to force the Linux VM to keep a minimum number
351 of kilobytes free. The VM uses this number to compute a
352 watermark[WMARK_MIN] value for each lowmem zone in the system.
353 Each lowmem zone gets a number of reserved free pages based
354 proportionally on its size.
356 Some minimal amount of memory is needed to satisfy PF_MEMALLOC
357 allocations; if you set this to lower than 1024KB, your system will
358 become subtly broken, and prone to deadlock under high loads.
360 Setting this too high will OOM your machine instantly.
362 =============================================================
366 This is available only on NUMA kernels.
368 A percentage of the total pages in each zone. On Zone reclaim
369 (fallback from the local zone occurs) slabs will be reclaimed if more
370 than this percentage of pages in a zone are reclaimable slab pages.
371 This insures that the slab growth stays under control even in NUMA
372 systems that rarely perform global reclaim.
374 The default is 5 percent.
376 Note that slab reclaim is triggered in a per zone / node fashion.
377 The process of reclaiming slab memory is currently not node specific
380 =============================================================
384 This is available only on NUMA kernels.
386 This is a percentage of the total pages in each zone. Zone reclaim will
387 only occur if more than this percentage of pages are in a state that
388 zone_reclaim_mode allows to be reclaimed.
390 If zone_reclaim_mode has the value 4 OR'd, then the percentage is compared
391 against all file-backed unmapped pages including swapcache pages and tmpfs
392 files. Otherwise, only unmapped pages backed by normal files but not tmpfs
393 files and similar are considered.
395 The default is 1 percent.
397 ==============================================================
401 This file indicates the amount of address space which a user process will
402 be restricted from mmapping. Since kernel null dereference bugs could
403 accidentally operate based on the information in the first couple of pages
404 of memory userspace processes should not be allowed to write to them. By
405 default this value is set to 0 and no protections will be enforced by the
406 security module. Setting this value to something like 64k will allow the
407 vast majority of applications to work correctly and provide defense in depth
408 against future potential kernel bugs.
410 ==============================================================
414 Change the minimum size of the hugepage pool.
416 See Documentation/vm/hugetlbpage.txt
418 ==============================================================
420 nr_overcommit_hugepages
422 Change the maximum size of the hugepage pool. The maximum is
423 nr_hugepages + nr_overcommit_hugepages.
425 See Documentation/vm/hugetlbpage.txt
427 ==============================================================
431 This is available only on NOMMU kernels.
433 This value adjusts the excess page trimming behaviour of power-of-2 aligned
434 NOMMU mmap allocations.
436 A value of 0 disables trimming of allocations entirely, while a value of 1
437 trims excess pages aggressively. Any value >= 1 acts as the watermark where
438 trimming of allocations is initiated.
440 The default value is 1.
442 See Documentation/nommu-mmap.txt for more information.
444 ==============================================================
448 This sysctl is only for NUMA.
449 'where the memory is allocated from' is controlled by zonelists.
450 (This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation.
451 you may be able to read ZONE_DMA as ZONE_DMA32...)
453 In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following.
454 ZONE_NORMAL -> ZONE_DMA
455 This means that a memory allocation request for GFP_KERNEL will
456 get memory from ZONE_DMA only when ZONE_NORMAL is not available.
458 In NUMA case, you can think of following 2 types of order.
459 Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL
461 (A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL
462 (B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA.
464 Type(A) offers the best locality for processes on Node(0), but ZONE_DMA
465 will be used before ZONE_NORMAL exhaustion. This increases possibility of
466 out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small.
468 Type(B) cannot offer the best locality but is more robust against OOM of
471 Type(A) is called as "Node" order. Type (B) is "Zone" order.
473 "Node order" orders the zonelists by node, then by zone within each node.
474 Specify "[Nn]ode" for node order
476 "Zone Order" orders the zonelists by zone type, then by node within each
477 zone. Specify "[Zz]one" for zone order.
479 Specify "[Dd]efault" to request automatic configuration. Autoconfiguration
480 will select "node" order in following case.
481 (1) if the DMA zone does not exist or
482 (2) if the DMA zone comprises greater than 50% of the available memory or
483 (3) if any node's DMA zone comprises greater than 60% of its local memory and
484 the amount of local memory is big enough.
486 Otherwise, "zone" order will be selected. Default order is recommended unless
487 this is causing problems for your system/application.
489 ==============================================================
493 Enables a system-wide task dump (excluding kernel threads) to be
494 produced when the kernel performs an OOM-killing and includes such
495 information as pid, uid, tgid, vm size, rss, nr_ptes, swapents,
496 oom_score_adj score, and name. This is helpful to determine why the
497 OOM killer was invoked, to identify the rogue task that caused it,
498 and to determine why the OOM killer chose the task it did to kill.
500 If this is set to zero, this information is suppressed. On very
501 large systems with thousands of tasks it may not be feasible to dump
502 the memory state information for each one. Such systems should not
503 be forced to incur a performance penalty in OOM conditions when the
504 information may not be desired.
506 If this is set to non-zero, this information is shown whenever the
507 OOM killer actually kills a memory-hogging task.
509 The default value is 1 (enabled).
511 ==============================================================
513 oom_kill_allocating_task
515 This enables or disables killing the OOM-triggering task in
516 out-of-memory situations.
518 If this is set to zero, the OOM killer will scan through the entire
519 tasklist and select a task based on heuristics to kill. This normally
520 selects a rogue memory-hogging task that frees up a large amount of
523 If this is set to non-zero, the OOM killer simply kills the task that
524 triggered the out-of-memory condition. This avoids the expensive
527 If panic_on_oom is selected, it takes precedence over whatever value
528 is used in oom_kill_allocating_task.
530 The default value is 0.
532 ==============================================================
536 This value contains a flag that enables memory overcommitment.
538 When this flag is 0, the kernel attempts to estimate the amount
539 of free memory left when userspace requests more memory.
541 When this flag is 1, the kernel pretends there is always enough
542 memory until it actually runs out.
544 When this flag is 2, the kernel uses a "never overcommit"
545 policy that attempts to prevent any overcommit of memory.
546 Note that user_reserve_kbytes affects this policy.
548 This feature can be very useful because there are a lot of
549 programs that malloc() huge amounts of memory "just-in-case"
550 and don't use much of it.
552 The default value is 0.
554 See Documentation/vm/overcommit-accounting and
555 security/commoncap.c::cap_vm_enough_memory() for more information.
557 ==============================================================
561 When overcommit_memory is set to 2, the committed address
562 space is not permitted to exceed swap plus this percentage
563 of physical RAM. See above.
565 ==============================================================
569 page-cluster controls the number of pages up to which consecutive pages
570 are read in from swap in a single attempt. This is the swap counterpart
571 to page cache readahead.
572 The mentioned consecutivity is not in terms of virtual/physical addresses,
573 but consecutive on swap space - that means they were swapped out together.
575 It is a logarithmic value - setting it to zero means "1 page", setting
576 it to 1 means "2 pages", setting it to 2 means "4 pages", etc.
577 Zero disables swap readahead completely.
579 The default value is three (eight pages at a time). There may be some
580 small benefits in tuning this to a different value if your workload is
583 Lower values mean lower latencies for initial faults, but at the same time
584 extra faults and I/O delays for following faults if they would have been part of
585 that consecutive pages readahead would have brought in.
587 =============================================================
591 This enables or disables panic on out-of-memory feature.
593 If this is set to 0, the kernel will kill some rogue process,
594 called oom_killer. Usually, oom_killer can kill rogue processes and
597 If this is set to 1, the kernel panics when out-of-memory happens.
598 However, if a process limits using nodes by mempolicy/cpusets,
599 and those nodes become memory exhaustion status, one process
600 may be killed by oom-killer. No panic occurs in this case.
601 Because other nodes' memory may be free. This means system total status
602 may be not fatal yet.
604 If this is set to 2, the kernel panics compulsorily even on the
605 above-mentioned. Even oom happens under memory cgroup, the whole
608 The default value is 0.
609 1 and 2 are for failover of clustering. Please select either
610 according to your policy of failover.
611 panic_on_oom=2+kdump gives you very strong tool to investigate
612 why oom happens. You can get snapshot.
614 =============================================================
616 percpu_pagelist_fraction
618 This is the fraction of pages at most (high mark pcp->high) in each zone that
619 are allocated for each per cpu page list. The min value for this is 8. It
620 means that we don't allow more than 1/8th of pages in each zone to be
621 allocated in any single per_cpu_pagelist. This entry only changes the value
622 of hot per cpu pagelists. User can specify a number like 100 to allocate
623 1/100th of each zone to each per cpu page list.
625 The batch value of each per cpu pagelist is also updated as a result. It is
626 set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8)
628 The initial value is zero. Kernel does not use this value at boot time to set
629 the high water marks for each per cpu page list.
631 ==============================================================
635 The time interval between which vm statistics are updated. The default
638 ==============================================================
642 This control is used to define how aggressive the kernel will swap
643 memory pages. Higher values will increase agressiveness, lower values
644 decrease the amount of swap.
646 The default value is 60.
648 ==============================================================
650 - user_reserve_kbytes
652 When overcommit_memory is set to 2, "never overommit" mode, reserve
653 min(3% of current process size, user_reserve_kbytes) of free memory.
654 This is intended to prevent a user from starting a single memory hogging
655 process, such that they cannot recover (kill the hog).
657 user_reserve_kbytes defaults to min(3% of the current process size, 128MB).
659 If this is reduced to zero, then the user will be allowed to allocate
660 all free memory with a single process, minus admin_reserve_kbytes.
661 Any subsequent attempts to execute a command will result in
662 "fork: Cannot allocate memory".
664 Changing this takes effect whenever an application requests memory.
666 ==============================================================
671 Controls the tendency of the kernel to reclaim the memory which is used for
672 caching of directory and inode objects.
674 At the default value of vfs_cache_pressure=100 the kernel will attempt to
675 reclaim dentries and inodes at a "fair" rate with respect to pagecache and
676 swapcache reclaim. Decreasing vfs_cache_pressure causes the kernel to prefer
677 to retain dentry and inode caches. When vfs_cache_pressure=0, the kernel will
678 never reclaim dentries and inodes due to memory pressure and this can easily
679 lead to out-of-memory conditions. Increasing vfs_cache_pressure beyond 100
680 causes the kernel to prefer to reclaim dentries and inodes.
682 ==============================================================
686 Zone_reclaim_mode allows someone to set more or less aggressive approaches to
687 reclaim memory when a zone runs out of memory. If it is set to zero then no
688 zone reclaim occurs. Allocations will be satisfied from other zones / nodes
691 This is value ORed together of
694 2 = Zone reclaim writes dirty pages out
695 4 = Zone reclaim swaps pages
697 zone_reclaim_mode is set during bootup to 1 if it is determined that pages
698 from remote zones will cause a measurable performance reduction. The
699 page allocator will then reclaim easily reusable pages (those page
700 cache pages that are currently not used) before allocating off node pages.
702 It may be beneficial to switch off zone reclaim if the system is
703 used for a file server and all of memory should be used for caching files
704 from disk. In that case the caching effect is more important than
707 Allowing zone reclaim to write out pages stops processes that are
708 writing large amounts of data from dirtying pages on other nodes. Zone
709 reclaim will write out dirty pages if a zone fills up and so effectively
710 throttle the process. This may decrease the performance of a single process
711 since it cannot use all of system memory to buffer the outgoing writes
712 anymore but it preserve the memory on other nodes so that the performance
713 of other processes running on other nodes will not be affected.
715 Allowing regular swap effectively restricts allocations to the local
716 node unless explicitly overridden by memory policies or cpuset
719 ============ End of Document =================================