blob: 078701fdbd4dd936dd485737abe7b5b60e554604 [file] [log] [blame]
Peter W Morrealedb0fb182009-01-15 13:50:42 -08001Documentation for /proc/sys/vm/* kernel version 2.6.29
Linus Torvalds1da177e2005-04-16 15:20:36 -07002 (c) 1998, 1999, Rik van Riel <riel@nl.linux.org>
Peter W Morrealedb0fb182009-01-15 13:50:42 -08003 (c) 2008 Peter W. Morreale <pmorreale@novell.com>
Linus Torvalds1da177e2005-04-16 15:20:36 -07004
5For general info and legal blurb, please look in README.
6
7==============================================================
8
9This file contains the documentation for the sysctl files in
Peter W Morrealedb0fb182009-01-15 13:50:42 -080010/proc/sys/vm and is valid for Linux kernel version 2.6.29.
Linus Torvalds1da177e2005-04-16 15:20:36 -070011
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:
Peter W Morrealedb0fb182009-01-15 13:50:42 -080020
21- block_dump
Mel Gorman76ab0f52010-05-24 14:32:28 -070022- compact_memory
Peter W Morrealedb0fb182009-01-15 13:50:42 -080023- dirty_background_bytes
Linus Torvalds1da177e2005-04-16 15:20:36 -070024- dirty_background_ratio
Peter W Morrealedb0fb182009-01-15 13:50:42 -080025- dirty_bytes
Linus Torvalds1da177e2005-04-16 15:20:36 -070026- dirty_expire_centisecs
Peter W Morrealedb0fb182009-01-15 13:50:42 -080027- dirty_ratio
Linus Torvalds1da177e2005-04-16 15:20:36 -070028- dirty_writeback_centisecs
Peter W Morrealedb0fb182009-01-15 13:50:42 -080029- drop_caches
Mel Gorman5e771902010-05-24 14:32:31 -070030- extfrag_threshold
Peter W Morrealedb0fb182009-01-15 13:50:42 -080031- hugepages_treat_as_movable
32- hugetlb_shm_group
33- laptop_mode
34- legacy_va_layout
35- lowmem_reserve_ratio
Linus Torvalds1da177e2005-04-16 15:20:36 -070036- max_map_count
Andi Kleen6a460792009-09-16 11:50:15 +020037- memory_failure_early_kill
38- memory_failure_recovery
Linus Torvalds1da177e2005-04-16 15:20:36 -070039- min_free_kbytes
Christoph Lameter0ff38492006-09-25 23:31:52 -070040- min_slab_ratio
Peter W Morrealedb0fb182009-01-15 13:50:42 -080041- min_unmapped_ratio
42- mmap_min_addr
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -080043- nr_hugepages
44- nr_overcommit_hugepages
Peter W Morrealedb0fb182009-01-15 13:50:42 -080045- nr_trim_pages (only if CONFIG_MMU=n)
46- numa_zonelist_order
47- oom_dump_tasks
48- oom_kill_allocating_task
49- overcommit_memory
50- overcommit_ratio
51- page-cluster
52- panic_on_oom
53- percpu_pagelist_fraction
54- stat_interval
55- swappiness
56- vfs_cache_pressure
57- zone_reclaim_mode
58
Linus Torvalds1da177e2005-04-16 15:20:36 -070059==============================================================
60
Peter W Morrealedb0fb182009-01-15 13:50:42 -080061block_dump
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
Peter W Morrealedb0fb182009-01-15 13:50:42 -080063block_dump enables block I/O debugging when set to a nonzero value. More
64information on block I/O debugging is in Documentation/laptops/laptop-mode.txt.
Linus Torvalds1da177e2005-04-16 15:20:36 -070065
66==============================================================
67
Mel Gorman76ab0f52010-05-24 14:32:28 -070068compact_memory
69
70Available only when CONFIG_COMPACTION is set. When 1 is written to the file,
71all zones are compacted such that free memory is available in contiguous
72blocks where possible. This can be important for example in the allocation of
73huge pages although processes will also directly compact memory as required.
74
75==============================================================
76
Peter W Morrealedb0fb182009-01-15 13:50:42 -080077dirty_background_bytes
Linus Torvalds1da177e2005-04-16 15:20:36 -070078
Artem Bityutskiy6601fac2012-07-25 18:12:01 +030079Contains the amount of dirty memory at which the background kernel
80flusher threads will start writeback.
Linus Torvalds1da177e2005-04-16 15:20:36 -070081
Andrea Righiabffc022010-10-27 15:33:31 -070082Note: dirty_background_bytes is the counterpart of dirty_background_ratio. Only
83one of them may be specified at a time. When one sysctl is written it is
84immediately taken into account to evaluate the dirty memory limits and the
85other appears as 0 when read.
Linus Torvalds1da177e2005-04-16 15:20:36 -070086
87==============================================================
88
Peter W Morrealedb0fb182009-01-15 13:50:42 -080089dirty_background_ratio
Linus Torvalds1da177e2005-04-16 15:20:36 -070090
Peter W Morrealedb0fb182009-01-15 13:50:42 -080091Contains, as a percentage of total system memory, the number of pages at which
Artem Bityutskiy6601fac2012-07-25 18:12:01 +030092the background kernel flusher threads will start writing out dirty data.
Linus Torvalds1da177e2005-04-16 15:20:36 -070093
94==============================================================
95
Peter W Morrealedb0fb182009-01-15 13:50:42 -080096dirty_bytes
Linus Torvalds1da177e2005-04-16 15:20:36 -070097
Peter W Morrealedb0fb182009-01-15 13:50:42 -080098Contains the amount of dirty memory at which a process generating disk writes
99will itself start writeback.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
Andrea Righiabffc022010-10-27 15:33:31 -0700101Note: dirty_bytes is the counterpart of dirty_ratio. Only one of them may be
102specified at a time. When one sysctl is written it is immediately taken into
103account to evaluate the dirty memory limits and the other appears as 0 when
104read.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800105
Andrea Righi9e4a5bd2009-04-30 15:08:57 -0700106Note: the minimum value allowed for dirty_bytes is two pages (in bytes); any
107value lower than this limit will be ignored and the old configuration will be
108retained.
109
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800110==============================================================
111
112dirty_expire_centisecs
113
114This tunable is used to define when dirty data is old enough to be eligible
Artem Bityutskiy6601fac2012-07-25 18:12:01 +0300115for writeout by the kernel flusher threads. It is expressed in 100'ths
116of a second. Data which has been dirty in-memory for longer than this
117interval will be written out next time a flusher thread wakes up.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800118
119==============================================================
120
121dirty_ratio
122
123Contains, as a percentage of total system memory, the number of pages at which
124a process which is generating disk writes will itself start writing out dirty
125data.
126
127==============================================================
128
129dirty_writeback_centisecs
130
Artem Bityutskiy6601fac2012-07-25 18:12:01 +0300131The kernel flusher threads will periodically wake up and write `old' data
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800132out to disk. This tunable expresses the interval between those wakeups, in
133100'ths of a second.
134
135Setting this to zero disables periodic writeback altogether.
136
137==============================================================
138
139drop_caches
140
141Writing to this will cause the kernel to drop clean caches, dentries and
142inodes from memory, causing that memory to become free.
143
144To free pagecache:
145 echo 1 > /proc/sys/vm/drop_caches
146To free dentries and inodes:
147 echo 2 > /proc/sys/vm/drop_caches
148To free pagecache, dentries and inodes:
149 echo 3 > /proc/sys/vm/drop_caches
150
151As this is a non-destructive operation and dirty objects are not freeable, the
152user should run `sync' first.
153
154==============================================================
155
Mel Gorman5e771902010-05-24 14:32:31 -0700156extfrag_threshold
157
158This parameter affects whether the kernel will compact memory or direct
159reclaim to satisfy a high-order allocation. /proc/extfrag_index shows what
160the fragmentation index for each order is in each zone in the system. Values
161tending towards 0 imply allocations would fail due to lack of memory,
162values towards 1000 imply failures are due to fragmentation and -1 implies
163that the allocation will succeed as long as watermarks are met.
164
165The kernel will not compact memory in a zone if the
166fragmentation index is <= extfrag_threshold. The default value is 500.
167
168==============================================================
169
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800170hugepages_treat_as_movable
171
172This parameter is only useful when kernelcore= is specified at boot time to
173create ZONE_MOVABLE for pages that may be reclaimed or migrated. Huge pages
174are not movable so are not normally allocated from ZONE_MOVABLE. A non-zero
175value written to hugepages_treat_as_movable allows huge pages to be allocated
176from ZONE_MOVABLE.
177
178Once enabled, the ZONE_MOVABLE is treated as an area of memory the huge
179pages pool can easily grow or shrink within. Assuming that applications are
180not running that mlock() a lot of memory, it is likely the huge pages pool
181can grow to the size of ZONE_MOVABLE by repeatedly entering the desired value
182into nr_hugepages and triggering page reclaim.
183
184==============================================================
185
186hugetlb_shm_group
187
188hugetlb_shm_group contains group id that is allowed to create SysV
189shared memory segment using hugetlb page.
190
191==============================================================
192
193laptop_mode
194
195laptop_mode is a knob that controls "laptop mode". All the things that are
196controlled by this knob are discussed in Documentation/laptops/laptop-mode.txt.
197
198==============================================================
199
200legacy_va_layout
201
Kulikov Vasiliy2174efb2010-06-28 13:59:28 +0200202If non-zero, this sysctl disables the new 32-bit mmap layout - the kernel
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800203will use the legacy (2.4) layout for all processes.
204
205==============================================================
206
207lowmem_reserve_ratio
208
209For some specialised workloads on highmem machines it is dangerous for
210the kernel to allow process memory to be allocated from the "lowmem"
211zone. This is because that memory could then be pinned via the mlock()
212system call, or by unavailability of swapspace.
213
214And on large highmem machines this lack of reclaimable lowmem memory
215can be fatal.
216
217So the Linux page allocator has a mechanism which prevents allocations
218which _could_ use highmem from using too much lowmem. This means that
219a certain amount of lowmem is defended from the possibility of being
220captured into pinned user memory.
221
222(The same argument applies to the old 16 megabyte ISA DMA region. This
223mechanism will also defend that region from allocations which could use
224highmem or lowmem).
225
226The `lowmem_reserve_ratio' tunable determines how aggressive the kernel is
227in defending these lower zones.
228
229If you have a machine which uses highmem or ISA DMA and your
230applications are using mlock(), or if you are running with no swap then
231you probably should change the lowmem_reserve_ratio setting.
232
233The lowmem_reserve_ratio is an array. You can see them by reading this file.
234-
235% cat /proc/sys/vm/lowmem_reserve_ratio
236256 256 32
237-
238Note: # of this elements is one fewer than number of zones. Because the highest
239 zone's value is not necessary for following calculation.
240
241But, these values are not used directly. The kernel calculates # of protection
242pages for each zones from them. These are shown as array of protection pages
243in /proc/zoneinfo like followings. (This is an example of x86-64 box).
244Each zone has an array of protection pages like this.
245
246-
247Node 0, zone DMA
248 pages free 1355
249 min 3
250 low 3
251 high 4
252 :
253 :
254 numa_other 0
255 protection: (0, 2004, 2004, 2004)
256 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
257 pagesets
258 cpu: 0 pcp: 0
259 :
260-
261These protections are added to score to judge whether this zone should be used
262for page allocation or should be reclaimed.
263
264In this example, if normal pages (index=2) are required to this DMA zone and
Mel Gorman41858962009-06-16 15:32:12 -0700265watermark[WMARK_HIGH] is used for watermark, the kernel judges this zone should
266not be used because pages_free(1355) is smaller than watermark + protection[2]
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800267(4 + 2004 = 2008). If this protection value is 0, this zone would be used for
268normal page requirement. If requirement is DMA zone(index=0), protection[0]
269(=0) is used.
270
271zone[i]'s protection[j] is calculated by following expression.
272
273(i < j):
274 zone[i]->protection[j]
275 = (total sums of present_pages from zone[i+1] to zone[j] on the node)
276 / lowmem_reserve_ratio[i];
277(i = j):
278 (should not be protected. = 0;
279(i > j):
280 (not necessary, but looks 0)
281
282The default values of lowmem_reserve_ratio[i] are
283 256 (if zone[i] means DMA or DMA32 zone)
284 32 (others).
285As above expression, they are reciprocal number of ratio.
286256 means 1/256. # of protection pages becomes about "0.39%" of total present
287pages of higher zones on the node.
288
289If you would like to protect more pages, smaller values are effective.
290The minimum value is 1 (1/1 -> 100%).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291
292==============================================================
293
294max_map_count:
295
296This file contains the maximum number of memory map areas a process
297may have. Memory map areas are used as a side-effect of calling
298malloc, directly by mmap and mprotect, and also when loading shared
299libraries.
300
301While most applications need less than a thousand maps, certain
302programs, particularly malloc debuggers, may consume lots of them,
303e.g., up to one or two maps per allocation.
304
305The default value is 65536.
306
Andi Kleen6a460792009-09-16 11:50:15 +0200307=============================================================
308
309memory_failure_early_kill:
310
311Control how to kill processes when uncorrected memory error (typically
312a 2bit error in a memory module) is detected in the background by hardware
313that cannot be handled by the kernel. In some cases (like the page
314still having a valid copy on disk) the kernel will handle the failure
315transparently without affecting any applications. But if there is
316no other uptodate copy of the data it will kill to prevent any data
317corruptions from propagating.
318
3191: Kill all processes that have the corrupted and not reloadable page mapped
320as soon as the corruption is detected. Note this is not supported
321for a few types of pages, like kernel internally allocated data or
322the swap cache, but works for the majority of user pages.
323
3240: Only unmap the corrupted page from all processes and only kill a process
325who tries to access it.
326
327The kill is done using a catchable SIGBUS with BUS_MCEERR_AO, so processes can
328handle this if they want to.
329
330This is only active on architectures/platforms with advanced machine
331check handling and depends on the hardware capabilities.
332
333Applications can override this setting individually with the PR_MCE_KILL prctl
334
335==============================================================
336
337memory_failure_recovery
338
339Enable memory failure recovery (when supported by the platform)
340
3411: Attempt recovery.
342
3430: Always panic on a memory failure.
344
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345==============================================================
346
347min_free_kbytes:
348
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800349This is used to force the Linux VM to keep a minimum number
Mel Gorman41858962009-06-16 15:32:12 -0700350of kilobytes free. The VM uses this number to compute a
351watermark[WMARK_MIN] value for each lowmem zone in the system.
352Each lowmem zone gets a number of reserved free pages based
353proportionally on its size.
Rohit Seth8ad4b1f2006-01-08 01:00:40 -0800354
Matt LaPlanted9195882008-07-25 19:45:33 -0700355Some minimal amount of memory is needed to satisfy PF_MEMALLOC
Pavel Machek24950892007-10-16 23:31:28 -0700356allocations; if you set this to lower than 1024KB, your system will
357become subtly broken, and prone to deadlock under high loads.
358
359Setting this too high will OOM your machine instantly.
360
Christoph Lameter96146342006-07-03 00:24:13 -0700361=============================================================
362
Christoph Lameter0ff38492006-09-25 23:31:52 -0700363min_slab_ratio:
364
365This is available only on NUMA kernels.
366
367A percentage of the total pages in each zone. On Zone reclaim
368(fallback from the local zone occurs) slabs will be reclaimed if more
369than this percentage of pages in a zone are reclaimable slab pages.
370This insures that the slab growth stays under control even in NUMA
371systems that rarely perform global reclaim.
372
373The default is 5 percent.
374
375Note that slab reclaim is triggered in a per zone / node fashion.
376The process of reclaiming slab memory is currently not node specific
377and may not be fast.
378
379=============================================================
380
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800381min_unmapped_ratio:
KAMEZAWA Hiroyukifadd8fb2006-06-23 02:03:13 -0700382
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800383This is available only on NUMA kernels.
Yasunori Goto2b744c02007-05-06 14:49:59 -0700384
Mel Gorman90afa5d2009-06-16 15:33:20 -0700385This is a percentage of the total pages in each zone. Zone reclaim will
386only occur if more than this percentage of pages are in a state that
387zone_reclaim_mode allows to be reclaimed.
388
389If zone_reclaim_mode has the value 4 OR'd, then the percentage is compared
390against all file-backed unmapped pages including swapcache pages and tmpfs
391files. Otherwise, only unmapped pages backed by normal files but not tmpfs
392files and similar are considered.
Yasunori Goto2b744c02007-05-06 14:49:59 -0700393
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800394The default is 1 percent.
David Rientjesfe071d72007-10-16 23:25:56 -0700395
Eric Parised032182007-06-28 15:55:21 -0400396==============================================================
397
398mmap_min_addr
399
400This file indicates the amount of address space which a user process will
André Goddard Rosaaf901ca2009-11-14 13:09:05 -0200401be restricted from mmapping. Since kernel null dereference bugs could
Eric Parised032182007-06-28 15:55:21 -0400402accidentally operate based on the information in the first couple of pages
403of memory userspace processes should not be allowed to write to them. By
404default this value is set to 0 and no protections will be enforced by the
405security module. Setting this value to something like 64k will allow the
406vast majority of applications to work correctly and provide defense in depth
407against future potential kernel bugs.
408
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700409==============================================================
410
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800411nr_hugepages
412
413Change the minimum size of the hugepage pool.
414
415See Documentation/vm/hugetlbpage.txt
416
417==============================================================
418
419nr_overcommit_hugepages
420
421Change the maximum size of the hugepage pool. The maximum is
422nr_hugepages + nr_overcommit_hugepages.
423
424See Documentation/vm/hugetlbpage.txt
425
426==============================================================
427
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800428nr_trim_pages
429
430This is available only on NOMMU kernels.
431
432This value adjusts the excess page trimming behaviour of power-of-2 aligned
433NOMMU mmap allocations.
434
435A value of 0 disables trimming of allocations entirely, while a value of 1
436trims excess pages aggressively. Any value >= 1 acts as the watermark where
437trimming of allocations is initiated.
438
439The default value is 1.
440
441See Documentation/nommu-mmap.txt for more information.
442
443==============================================================
444
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700445numa_zonelist_order
446
447This sysctl is only for NUMA.
448'where the memory is allocated from' is controlled by zonelists.
449(This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation.
450 you may be able to read ZONE_DMA as ZONE_DMA32...)
451
452In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following.
453ZONE_NORMAL -> ZONE_DMA
454This means that a memory allocation request for GFP_KERNEL will
455get memory from ZONE_DMA only when ZONE_NORMAL is not available.
456
457In NUMA case, you can think of following 2 types of order.
458Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL
459
460(A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL
461(B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA.
462
463Type(A) offers the best locality for processes on Node(0), but ZONE_DMA
464will be used before ZONE_NORMAL exhaustion. This increases possibility of
465out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small.
466
467Type(B) cannot offer the best locality but is more robust against OOM of
468the DMA zone.
469
470Type(A) is called as "Node" order. Type (B) is "Zone" order.
471
472"Node order" orders the zonelists by node, then by zone within each node.
Paul Bolle5a3016a2011-04-06 11:09:55 +0200473Specify "[Nn]ode" for node order
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700474
475"Zone Order" orders the zonelists by zone type, then by node within each
Paul Bolle5a3016a2011-04-06 11:09:55 +0200476zone. Specify "[Zz]one" for zone order.
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700477
478Specify "[Dd]efault" to request automatic configuration. Autoconfiguration
479will select "node" order in following case.
480(1) if the DMA zone does not exist or
481(2) if the DMA zone comprises greater than 50% of the available memory or
482(3) if any node's DMA zone comprises greater than 60% of its local memory and
483 the amount of local memory is big enough.
484
485Otherwise, "zone" order will be selected. Default order is recommended unless
486this is causing problems for your system/application.
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800487
488==============================================================
489
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800490oom_dump_tasks
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800491
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800492Enables a system-wide task dump (excluding kernel threads) to be
493produced when the kernel performs an OOM-killing and includes such
David Rientjesde34d962012-07-31 16:42:56 -0700494information as pid, uid, tgid, vm size, rss, nr_ptes, swapents,
495oom_score_adj score, and name. This is helpful to determine why the
496OOM killer was invoked, to identify the rogue task that caused it,
497and to determine why the OOM killer chose the task it did to kill.
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800498
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800499If this is set to zero, this information is suppressed. On very
500large systems with thousands of tasks it may not be feasible to dump
501the memory state information for each one. Such systems should not
502be forced to incur a performance penalty in OOM conditions when the
503information may not be desired.
504
505If this is set to non-zero, this information is shown whenever the
506OOM killer actually kills a memory-hogging task.
507
David Rientjesad915c42010-08-09 17:18:53 -0700508The default value is 1 (enabled).
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800509
510==============================================================
511
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800512oom_kill_allocating_task
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800513
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800514This enables or disables killing the OOM-triggering task in
515out-of-memory situations.
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800516
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800517If this is set to zero, the OOM killer will scan through the entire
518tasklist and select a task based on heuristics to kill. This normally
519selects a rogue memory-hogging task that frees up a large amount of
520memory when killed.
521
522If this is set to non-zero, the OOM killer simply kills the task that
523triggered the out-of-memory condition. This avoids the expensive
524tasklist scan.
525
526If panic_on_oom is selected, it takes precedence over whatever value
527is used in oom_kill_allocating_task.
528
529The default value is 0.
Paul Mundtdd8632a2009-01-08 12:04:47 +0000530
531==============================================================
532
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800533overcommit_memory:
Paul Mundtdd8632a2009-01-08 12:04:47 +0000534
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800535This value contains a flag that enables memory overcommitment.
Paul Mundtdd8632a2009-01-08 12:04:47 +0000536
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800537When this flag is 0, the kernel attempts to estimate the amount
538of free memory left when userspace requests more memory.
Paul Mundtdd8632a2009-01-08 12:04:47 +0000539
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800540When this flag is 1, the kernel pretends there is always enough
541memory until it actually runs out.
Paul Mundtdd8632a2009-01-08 12:04:47 +0000542
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800543When this flag is 2, the kernel uses a "never overcommit"
544policy that attempts to prevent any overcommit of memory.
Paul Mundtdd8632a2009-01-08 12:04:47 +0000545
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800546This feature can be very useful because there are a lot of
547programs that malloc() huge amounts of memory "just-in-case"
548and don't use much of it.
549
550The default value is 0.
551
552See Documentation/vm/overcommit-accounting and
553security/commoncap.c::cap_vm_enough_memory() for more information.
554
555==============================================================
556
557overcommit_ratio:
558
559When overcommit_memory is set to 2, the committed address
560space is not permitted to exceed swap plus this percentage
561of physical RAM. See above.
562
563==============================================================
564
565page-cluster
566
Christian Ehrhardtdf858fa2012-07-31 16:41:46 -0700567page-cluster controls the number of pages up to which consecutive pages
568are read in from swap in a single attempt. This is the swap counterpart
569to page cache readahead.
570The mentioned consecutivity is not in terms of virtual/physical addresses,
571but consecutive on swap space - that means they were swapped out together.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800572
573It is a logarithmic value - setting it to zero means "1 page", setting
574it to 1 means "2 pages", setting it to 2 means "4 pages", etc.
Christian Ehrhardtdf858fa2012-07-31 16:41:46 -0700575Zero disables swap readahead completely.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800576
577The default value is three (eight pages at a time). There may be some
578small benefits in tuning this to a different value if your workload is
579swap-intensive.
580
Christian Ehrhardtdf858fa2012-07-31 16:41:46 -0700581Lower values mean lower latencies for initial faults, but at the same time
582extra faults and I/O delays for following faults if they would have been part of
583that consecutive pages readahead would have brought in.
584
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800585=============================================================
586
587panic_on_oom
588
589This enables or disables panic on out-of-memory feature.
590
591If this is set to 0, the kernel will kill some rogue process,
592called oom_killer. Usually, oom_killer can kill rogue processes and
593system will survive.
594
595If this is set to 1, the kernel panics when out-of-memory happens.
596However, if a process limits using nodes by mempolicy/cpusets,
597and those nodes become memory exhaustion status, one process
598may be killed by oom-killer. No panic occurs in this case.
599Because other nodes' memory may be free. This means system total status
600may be not fatal yet.
601
602If this is set to 2, the kernel panics compulsorily even on the
KAMEZAWA Hiroyukidaaf1e62010-03-10 15:22:32 -0800603above-mentioned. Even oom happens under memory cgroup, the whole
604system panics.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800605
606The default value is 0.
6071 and 2 are for failover of clustering. Please select either
608according to your policy of failover.
KAMEZAWA Hiroyukidaaf1e62010-03-10 15:22:32 -0800609panic_on_oom=2+kdump gives you very strong tool to investigate
610why oom happens. You can get snapshot.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800611
612=============================================================
613
614percpu_pagelist_fraction
615
616This is the fraction of pages at most (high mark pcp->high) in each zone that
617are allocated for each per cpu page list. The min value for this is 8. It
618means that we don't allow more than 1/8th of pages in each zone to be
619allocated in any single per_cpu_pagelist. This entry only changes the value
620of hot per cpu pagelists. User can specify a number like 100 to allocate
6211/100th of each zone to each per cpu page list.
622
623The batch value of each per cpu pagelist is also updated as a result. It is
624set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8)
625
626The initial value is zero. Kernel does not use this value at boot time to set
627the high water marks for each per cpu page list.
628
629==============================================================
630
631stat_interval
632
633The time interval between which vm statistics are updated. The default
634is 1 second.
635
636==============================================================
637
638swappiness
639
640This control is used to define how aggressive the kernel will swap
641memory pages. Higher values will increase agressiveness, lower values
Matt LaPlante19f59462009-04-27 15:06:31 +0200642decrease the amount of swap.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800643
644The default value is 60.
645
646==============================================================
647
648vfs_cache_pressure
649------------------
650
651Controls the tendency of the kernel to reclaim the memory which is used for
652caching of directory and inode objects.
653
654At the default value of vfs_cache_pressure=100 the kernel will attempt to
655reclaim dentries and inodes at a "fair" rate with respect to pagecache and
656swapcache reclaim. Decreasing vfs_cache_pressure causes the kernel to prefer
Jan Kara55c37a82009-09-21 17:01:40 -0700657to retain dentry and inode caches. When vfs_cache_pressure=0, the kernel will
658never reclaim dentries and inodes due to memory pressure and this can easily
659lead to out-of-memory conditions. Increasing vfs_cache_pressure beyond 100
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800660causes the kernel to prefer to reclaim dentries and inodes.
661
662==============================================================
663
664zone_reclaim_mode:
665
666Zone_reclaim_mode allows someone to set more or less aggressive approaches to
667reclaim memory when a zone runs out of memory. If it is set to zero then no
668zone reclaim occurs. Allocations will be satisfied from other zones / nodes
669in the system.
670
671This is value ORed together of
672
6731 = Zone reclaim on
6742 = Zone reclaim writes dirty pages out
6754 = Zone reclaim swaps pages
676
677zone_reclaim_mode is set during bootup to 1 if it is determined that pages
678from remote zones will cause a measurable performance reduction. The
679page allocator will then reclaim easily reusable pages (those page
680cache pages that are currently not used) before allocating off node pages.
681
682It may be beneficial to switch off zone reclaim if the system is
683used for a file server and all of memory should be used for caching files
684from disk. In that case the caching effect is more important than
685data locality.
686
687Allowing zone reclaim to write out pages stops processes that are
688writing large amounts of data from dirtying pages on other nodes. Zone
689reclaim will write out dirty pages if a zone fills up and so effectively
690throttle the process. This may decrease the performance of a single process
691since it cannot use all of system memory to buffer the outgoing writes
692anymore but it preserve the memory on other nodes so that the performance
693of other processes running on other nodes will not be affected.
694
695Allowing regular swap effectively restricts allocations to the local
696node unless explicitly overridden by memory policies or cpuset
697configurations.
698
699============ End of Document =================================