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:
22 - dirty_background_bytes
23 - dirty_background_ratio
25 - dirty_expire_centisecs
27 - dirty_writeback_centisecs
29 - hugepages_treat_as_movable
33 - lowmem_reserve_ratio
40 - nr_overcommit_hugepages
42 - nr_trim_pages (only if CONFIG_MMU=n)
45 - oom_kill_allocating_task
50 - percpu_pagelist_fraction
57 ==============================================================
61 block_dump enables block I/O debugging when set to a nonzero value. More
62 information on block I/O debugging is in Documentation/laptops/laptop-mode.txt.
64 ==============================================================
66 dirty_background_bytes
68 Contains the amount of dirty memory at which the pdflush background writeback
69 daemon will start writeback.
71 If dirty_background_bytes is written, dirty_background_ratio becomes a function
72 of its value (dirty_background_bytes / the amount of dirtyable system memory).
74 ==============================================================
76 dirty_background_ratio
78 Contains, as a percentage of total system memory, the number of pages at which
79 the pdflush background writeback daemon will start writing out dirty data.
81 ==============================================================
85 Contains the amount of dirty memory at which a process generating disk writes
86 will itself start writeback.
88 If dirty_bytes is written, dirty_ratio becomes a function of its value
89 (dirty_bytes / the amount of dirtyable system memory).
91 Note: the minimum value allowed for dirty_bytes is two pages (in bytes); any
92 value lower than this limit will be ignored and the old configuration will be
95 ==============================================================
97 dirty_expire_centisecs
99 This tunable is used to define when dirty data is old enough to be eligible
100 for writeout by the pdflush daemons. It is expressed in 100'ths of a second.
101 Data which has been dirty in-memory for longer than this interval will be
102 written out next time a pdflush daemon wakes up.
104 ==============================================================
108 Contains, as a percentage of total system memory, the number of pages at which
109 a process which is generating disk writes will itself start writing out dirty
112 ==============================================================
114 dirty_writeback_centisecs
116 The pdflush writeback daemons will periodically wake up and write `old' data
117 out to disk. This tunable expresses the interval between those wakeups, in
120 Setting this to zero disables periodic writeback altogether.
122 ==============================================================
126 Writing to this will cause the kernel to drop clean caches, dentries and
127 inodes from memory, causing that memory to become free.
130 echo 1 > /proc/sys/vm/drop_caches
131 To free dentries and inodes:
132 echo 2 > /proc/sys/vm/drop_caches
133 To free pagecache, dentries and inodes:
134 echo 3 > /proc/sys/vm/drop_caches
136 As this is a non-destructive operation and dirty objects are not freeable, the
137 user should run `sync' first.
139 ==============================================================
141 hugepages_treat_as_movable
143 This parameter is only useful when kernelcore= is specified at boot time to
144 create ZONE_MOVABLE for pages that may be reclaimed or migrated. Huge pages
145 are not movable so are not normally allocated from ZONE_MOVABLE. A non-zero
146 value written to hugepages_treat_as_movable allows huge pages to be allocated
149 Once enabled, the ZONE_MOVABLE is treated as an area of memory the huge
150 pages pool can easily grow or shrink within. Assuming that applications are
151 not running that mlock() a lot of memory, it is likely the huge pages pool
152 can grow to the size of ZONE_MOVABLE by repeatedly entering the desired value
153 into nr_hugepages and triggering page reclaim.
155 ==============================================================
159 hugetlb_shm_group contains group id that is allowed to create SysV
160 shared memory segment using hugetlb page.
162 ==============================================================
166 laptop_mode is a knob that controls "laptop mode". All the things that are
167 controlled by this knob are discussed in Documentation/laptops/laptop-mode.txt.
169 ==============================================================
173 If non-zero, this sysctl disables the new 32-bit mmap mmap layout - the kernel
174 will use the legacy (2.4) layout for all processes.
176 ==============================================================
180 For some specialised workloads on highmem machines it is dangerous for
181 the kernel to allow process memory to be allocated from the "lowmem"
182 zone. This is because that memory could then be pinned via the mlock()
183 system call, or by unavailability of swapspace.
185 And on large highmem machines this lack of reclaimable lowmem memory
188 So the Linux page allocator has a mechanism which prevents allocations
189 which _could_ use highmem from using too much lowmem. This means that
190 a certain amount of lowmem is defended from the possibility of being
191 captured into pinned user memory.
193 (The same argument applies to the old 16 megabyte ISA DMA region. This
194 mechanism will also defend that region from allocations which could use
197 The `lowmem_reserve_ratio' tunable determines how aggressive the kernel is
198 in defending these lower zones.
200 If you have a machine which uses highmem or ISA DMA and your
201 applications are using mlock(), or if you are running with no swap then
202 you probably should change the lowmem_reserve_ratio setting.
204 The lowmem_reserve_ratio is an array. You can see them by reading this file.
206 % cat /proc/sys/vm/lowmem_reserve_ratio
209 Note: # of this elements is one fewer than number of zones. Because the highest
210 zone's value is not necessary for following calculation.
212 But, these values are not used directly. The kernel calculates # of protection
213 pages for each zones from them. These are shown as array of protection pages
214 in /proc/zoneinfo like followings. (This is an example of x86-64 box).
215 Each zone has an array of protection pages like this.
226 protection: (0, 2004, 2004, 2004)
227 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
232 These protections are added to score to judge whether this zone should be used
233 for page allocation or should be reclaimed.
235 In this example, if normal pages (index=2) are required to this DMA zone and
236 watermark[WMARK_HIGH] is used for watermark, the kernel judges this zone should
237 not be used because pages_free(1355) is smaller than watermark + protection[2]
238 (4 + 2004 = 2008). If this protection value is 0, this zone would be used for
239 normal page requirement. If requirement is DMA zone(index=0), protection[0]
242 zone[i]'s protection[j] is calculated by following expression.
245 zone[i]->protection[j]
246 = (total sums of present_pages from zone[i+1] to zone[j] on the node)
247 / lowmem_reserve_ratio[i];
249 (should not be protected. = 0;
251 (not necessary, but looks 0)
253 The default values of lowmem_reserve_ratio[i] are
254 256 (if zone[i] means DMA or DMA32 zone)
256 As above expression, they are reciprocal number of ratio.
257 256 means 1/256. # of protection pages becomes about "0.39%" of total present
258 pages of higher zones on the node.
260 If you would like to protect more pages, smaller values are effective.
261 The minimum value is 1 (1/1 -> 100%).
263 ==============================================================
267 This file contains the maximum number of memory map areas a process
268 may have. Memory map areas are used as a side-effect of calling
269 malloc, directly by mmap and mprotect, and also when loading shared
272 While most applications need less than a thousand maps, certain
273 programs, particularly malloc debuggers, may consume lots of them,
274 e.g., up to one or two maps per allocation.
276 The default value is 65536.
278 ==============================================================
282 This is used to force the Linux VM to keep a minimum number
283 of kilobytes free. The VM uses this number to compute a
284 watermark[WMARK_MIN] value for each lowmem zone in the system.
285 Each lowmem zone gets a number of reserved free pages based
286 proportionally on its size.
288 Some minimal amount of memory is needed to satisfy PF_MEMALLOC
289 allocations; if you set this to lower than 1024KB, your system will
290 become subtly broken, and prone to deadlock under high loads.
292 Setting this too high will OOM your machine instantly.
294 =============================================================
298 This is available only on NUMA kernels.
300 A percentage of the total pages in each zone. On Zone reclaim
301 (fallback from the local zone occurs) slabs will be reclaimed if more
302 than this percentage of pages in a zone are reclaimable slab pages.
303 This insures that the slab growth stays under control even in NUMA
304 systems that rarely perform global reclaim.
306 The default is 5 percent.
308 Note that slab reclaim is triggered in a per zone / node fashion.
309 The process of reclaiming slab memory is currently not node specific
312 =============================================================
316 This is available only on NUMA kernels.
318 A percentage of the total pages in each zone. Zone reclaim will only
319 occur if more than this percentage of pages are file backed and unmapped.
320 This is to insure that a minimal amount of local pages is still available for
321 file I/O even if the node is overallocated.
323 The default is 1 percent.
325 ==============================================================
329 This file indicates the amount of address space which a user process will
330 be restricted from mmaping. Since kernel null dereference bugs could
331 accidentally operate based on the information in the first couple of pages
332 of memory userspace processes should not be allowed to write to them. By
333 default this value is set to 0 and no protections will be enforced by the
334 security module. Setting this value to something like 64k will allow the
335 vast majority of applications to work correctly and provide defense in depth
336 against future potential kernel bugs.
338 ==============================================================
342 Change the minimum size of the hugepage pool.
344 See Documentation/vm/hugetlbpage.txt
346 ==============================================================
348 nr_overcommit_hugepages
350 Change the maximum size of the hugepage pool. The maximum is
351 nr_hugepages + nr_overcommit_hugepages.
353 See Documentation/vm/hugetlbpage.txt
355 ==============================================================
359 The current number of pdflush threads. This value is read-only.
360 The value changes according to the number of dirty pages in the system.
362 When necessary, additional pdflush threads are created, one per second, up to
363 nr_pdflush_threads_max.
365 ==============================================================
369 This is available only on NOMMU kernels.
371 This value adjusts the excess page trimming behaviour of power-of-2 aligned
372 NOMMU mmap allocations.
374 A value of 0 disables trimming of allocations entirely, while a value of 1
375 trims excess pages aggressively. Any value >= 1 acts as the watermark where
376 trimming of allocations is initiated.
378 The default value is 1.
380 See Documentation/nommu-mmap.txt for more information.
382 ==============================================================
386 This sysctl is only for NUMA.
387 'where the memory is allocated from' is controlled by zonelists.
388 (This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation.
389 you may be able to read ZONE_DMA as ZONE_DMA32...)
391 In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following.
392 ZONE_NORMAL -> ZONE_DMA
393 This means that a memory allocation request for GFP_KERNEL will
394 get memory from ZONE_DMA only when ZONE_NORMAL is not available.
396 In NUMA case, you can think of following 2 types of order.
397 Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL
399 (A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL
400 (B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA.
402 Type(A) offers the best locality for processes on Node(0), but ZONE_DMA
403 will be used before ZONE_NORMAL exhaustion. This increases possibility of
404 out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small.
406 Type(B) cannot offer the best locality but is more robust against OOM of
409 Type(A) is called as "Node" order. Type (B) is "Zone" order.
411 "Node order" orders the zonelists by node, then by zone within each node.
412 Specify "[Nn]ode" for zone order
414 "Zone Order" orders the zonelists by zone type, then by node within each
415 zone. Specify "[Zz]one"for zode order.
417 Specify "[Dd]efault" to request automatic configuration. Autoconfiguration
418 will select "node" order in following case.
419 (1) if the DMA zone does not exist or
420 (2) if the DMA zone comprises greater than 50% of the available memory or
421 (3) if any node's DMA zone comprises greater than 60% of its local memory and
422 the amount of local memory is big enough.
424 Otherwise, "zone" order will be selected. Default order is recommended unless
425 this is causing problems for your system/application.
427 ==============================================================
431 Enables a system-wide task dump (excluding kernel threads) to be
432 produced when the kernel performs an OOM-killing and includes such
433 information as pid, uid, tgid, vm size, rss, cpu, oom_adj score, and
434 name. This is helpful to determine why the OOM killer was invoked
435 and to identify the rogue task that caused it.
437 If this is set to zero, this information is suppressed. On very
438 large systems with thousands of tasks it may not be feasible to dump
439 the memory state information for each one. Such systems should not
440 be forced to incur a performance penalty in OOM conditions when the
441 information may not be desired.
443 If this is set to non-zero, this information is shown whenever the
444 OOM killer actually kills a memory-hogging task.
446 The default value is 0.
448 ==============================================================
450 oom_kill_allocating_task
452 This enables or disables killing the OOM-triggering task in
453 out-of-memory situations.
455 If this is set to zero, the OOM killer will scan through the entire
456 tasklist and select a task based on heuristics to kill. This normally
457 selects a rogue memory-hogging task that frees up a large amount of
460 If this is set to non-zero, the OOM killer simply kills the task that
461 triggered the out-of-memory condition. This avoids the expensive
464 If panic_on_oom is selected, it takes precedence over whatever value
465 is used in oom_kill_allocating_task.
467 The default value is 0.
469 ==============================================================
473 This value contains a flag that enables memory overcommitment.
475 When this flag is 0, the kernel attempts to estimate the amount
476 of free memory left when userspace requests more memory.
478 When this flag is 1, the kernel pretends there is always enough
479 memory until it actually runs out.
481 When this flag is 2, the kernel uses a "never overcommit"
482 policy that attempts to prevent any overcommit of memory.
484 This feature can be very useful because there are a lot of
485 programs that malloc() huge amounts of memory "just-in-case"
486 and don't use much of it.
488 The default value is 0.
490 See Documentation/vm/overcommit-accounting and
491 security/commoncap.c::cap_vm_enough_memory() for more information.
493 ==============================================================
497 When overcommit_memory is set to 2, the committed address
498 space is not permitted to exceed swap plus this percentage
499 of physical RAM. See above.
501 ==============================================================
505 page-cluster controls the number of pages which are written to swap in
506 a single attempt. The swap I/O size.
508 It is a logarithmic value - setting it to zero means "1 page", setting
509 it to 1 means "2 pages", setting it to 2 means "4 pages", etc.
511 The default value is three (eight pages at a time). There may be some
512 small benefits in tuning this to a different value if your workload is
515 =============================================================
519 This enables or disables panic on out-of-memory feature.
521 If this is set to 0, the kernel will kill some rogue process,
522 called oom_killer. Usually, oom_killer can kill rogue processes and
525 If this is set to 1, the kernel panics when out-of-memory happens.
526 However, if a process limits using nodes by mempolicy/cpusets,
527 and those nodes become memory exhaustion status, one process
528 may be killed by oom-killer. No panic occurs in this case.
529 Because other nodes' memory may be free. This means system total status
530 may be not fatal yet.
532 If this is set to 2, the kernel panics compulsorily even on the
535 The default value is 0.
536 1 and 2 are for failover of clustering. Please select either
537 according to your policy of failover.
539 =============================================================
541 percpu_pagelist_fraction
543 This is the fraction of pages at most (high mark pcp->high) in each zone that
544 are allocated for each per cpu page list. The min value for this is 8. It
545 means that we don't allow more than 1/8th of pages in each zone to be
546 allocated in any single per_cpu_pagelist. This entry only changes the value
547 of hot per cpu pagelists. User can specify a number like 100 to allocate
548 1/100th of each zone to each per cpu page list.
550 The batch value of each per cpu pagelist is also updated as a result. It is
551 set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8)
553 The initial value is zero. Kernel does not use this value at boot time to set
554 the high water marks for each per cpu page list.
556 ==============================================================
560 The time interval between which vm statistics are updated. The default
563 ==============================================================
567 This control is used to define how aggressive the kernel will swap
568 memory pages. Higher values will increase agressiveness, lower values
569 decrease the amount of swap.
571 The default value is 60.
573 ==============================================================
578 Controls the tendency of the kernel to reclaim the memory which is used for
579 caching of directory and inode objects.
581 At the default value of vfs_cache_pressure=100 the kernel will attempt to
582 reclaim dentries and inodes at a "fair" rate with respect to pagecache and
583 swapcache reclaim. Decreasing vfs_cache_pressure causes the kernel to prefer
584 to retain dentry and inode caches. Increasing vfs_cache_pressure beyond 100
585 causes the kernel to prefer to reclaim dentries and inodes.
587 ==============================================================
591 Zone_reclaim_mode allows someone to set more or less aggressive approaches to
592 reclaim memory when a zone runs out of memory. If it is set to zero then no
593 zone reclaim occurs. Allocations will be satisfied from other zones / nodes
596 This is value ORed together of
599 2 = Zone reclaim writes dirty pages out
600 4 = Zone reclaim swaps pages
602 zone_reclaim_mode is set during bootup to 1 if it is determined that pages
603 from remote zones will cause a measurable performance reduction. The
604 page allocator will then reclaim easily reusable pages (those page
605 cache pages that are currently not used) before allocating off node pages.
607 It may be beneficial to switch off zone reclaim if the system is
608 used for a file server and all of memory should be used for caching files
609 from disk. In that case the caching effect is more important than
612 Allowing zone reclaim to write out pages stops processes that are
613 writing large amounts of data from dirtying pages on other nodes. Zone
614 reclaim will write out dirty pages if a zone fills up and so effectively
615 throttle the process. This may decrease the performance of a single process
616 since it cannot use all of system memory to buffer the outgoing writes
617 anymore but it preserve the memory on other nodes so that the performance
618 of other processes running on other nodes will not be affected.
620 Allowing regular swap effectively restricts allocations to the local
621 node unless explicitly overridden by memory policies or cpuset
624 ============ End of Document =================================