blob: 17256f2ad919aa99cd2271eb8b05d09d5c8e2b62 [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
Andrew Shewmaker4eeab4f2013-04-29 15:08:11 -070021- admin_reserve_kbytes
Peter W Morrealedb0fb182009-01-15 13:50:42 -080022- block_dump
Mel Gorman76ab0f52010-05-24 14:32:28 -070023- compact_memory
Eric B Munson5bbe3542015-04-15 16:13:20 -070024- compact_unevictable_allowed
Peter W Morrealedb0fb182009-01-15 13:50:42 -080025- dirty_background_bytes
Linus Torvalds1da177e2005-04-16 15:20:36 -070026- dirty_background_ratio
Peter W Morrealedb0fb182009-01-15 13:50:42 -080027- dirty_bytes
Linus Torvalds1da177e2005-04-16 15:20:36 -070028- dirty_expire_centisecs
Peter W Morrealedb0fb182009-01-15 13:50:42 -080029- dirty_ratio
Linus Torvalds1da177e2005-04-16 15:20:36 -070030- dirty_writeback_centisecs
Peter W Morrealedb0fb182009-01-15 13:50:42 -080031- drop_caches
Mel Gorman5e771902010-05-24 14:32:31 -070032- extfrag_threshold
Peter W Morrealedb0fb182009-01-15 13:50:42 -080033- hugetlb_shm_group
34- laptop_mode
35- legacy_va_layout
36- lowmem_reserve_ratio
Linus Torvalds1da177e2005-04-16 15:20:36 -070037- max_map_count
Andi Kleen6a460792009-09-16 11:50:15 +020038- memory_failure_early_kill
39- memory_failure_recovery
Linus Torvalds1da177e2005-04-16 15:20:36 -070040- min_free_kbytes
Christoph Lameter0ff38492006-09-25 23:31:52 -070041- min_slab_ratio
Peter W Morrealedb0fb182009-01-15 13:50:42 -080042- min_unmapped_ratio
43- mmap_min_addr
Daniel Cashmand07e2252016-01-14 15:19:53 -080044- mmap_rnd_bits
45- mmap_rnd_compat_bits
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -080046- nr_hugepages
47- nr_overcommit_hugepages
Peter W Morrealedb0fb182009-01-15 13:50:42 -080048- nr_trim_pages (only if CONFIG_MMU=n)
49- numa_zonelist_order
50- oom_dump_tasks
51- oom_kill_allocating_task
Jerome Marchand49f0ce52014-01-21 15:49:14 -080052- overcommit_kbytes
Peter W Morrealedb0fb182009-01-15 13:50:42 -080053- overcommit_memory
54- overcommit_ratio
55- page-cluster
56- panic_on_oom
57- percpu_pagelist_fraction
58- stat_interval
Hugh Dickins52b6f462016-05-19 17:12:50 -070059- stat_refresh
Kemi Wang45180852017-11-15 17:38:22 -080060- numa_stat
Peter W Morrealedb0fb182009-01-15 13:50:42 -080061- swappiness
Andrew Shewmakerc9b1d092013-04-29 15:08:10 -070062- user_reserve_kbytes
Peter W Morrealedb0fb182009-01-15 13:50:42 -080063- vfs_cache_pressure
Jerome Marchande6507a02016-07-12 12:05:59 +020064- watermark_scale_factor
Peter W Morrealedb0fb182009-01-15 13:50:42 -080065- zone_reclaim_mode
66
Linus Torvalds1da177e2005-04-16 15:20:36 -070067==============================================================
68
Andrew Shewmaker4eeab4f2013-04-29 15:08:11 -070069admin_reserve_kbytes
70
71The amount of free memory in the system that should be reserved for users
72with the capability cap_sys_admin.
73
74admin_reserve_kbytes defaults to min(3% of free pages, 8MB)
75
76That should provide enough for the admin to log in and kill a process,
77if necessary, under the default overcommit 'guess' mode.
78
79Systems running under overcommit 'never' should increase this to account
80for the full Virtual Memory Size of programs used to recover. Otherwise,
81root may not be able to log in to recover the system.
82
83How do you calculate a minimum useful reserve?
84
85sshd or login + bash (or some other shell) + top (or ps, kill, etc.)
86
87For overcommit 'guess', we can sum resident set sizes (RSS).
88On x86_64 this is about 8MB.
89
90For overcommit 'never', we can take the max of their virtual sizes (VSZ)
91and add the sum of their RSS.
92On x86_64 this is about 128MB.
93
94Changing this takes effect whenever an application requests memory.
95
96==============================================================
97
Peter W Morrealedb0fb182009-01-15 13:50:42 -080098block_dump
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800100block_dump enables block I/O debugging when set to a nonzero value. More
101information on block I/O debugging is in Documentation/laptops/laptop-mode.txt.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102
103==============================================================
104
Mel Gorman76ab0f52010-05-24 14:32:28 -0700105compact_memory
106
107Available only when CONFIG_COMPACTION is set. When 1 is written to the file,
108all zones are compacted such that free memory is available in contiguous
109blocks where possible. This can be important for example in the allocation of
110huge pages although processes will also directly compact memory as required.
111
112==============================================================
113
Eric B Munson5bbe3542015-04-15 16:13:20 -0700114compact_unevictable_allowed
115
116Available only when CONFIG_COMPACTION is set. When set to 1, compaction is
117allowed to examine the unevictable lru (mlocked pages) for pages to compact.
118This should be used on systems where stalls for minor page faults are an
119acceptable trade for large contiguous free memory. Set to 0 to prevent
120compaction from moving pages that are unevictable. Default value is 1.
121
122==============================================================
123
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800124dirty_background_bytes
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125
Artem Bityutskiy6601fac2012-07-25 18:12:01 +0300126Contains the amount of dirty memory at which the background kernel
127flusher threads will start writeback.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128
Andrea Righiabffc022010-10-27 15:33:31 -0700129Note: dirty_background_bytes is the counterpart of dirty_background_ratio. Only
130one of them may be specified at a time. When one sysctl is written it is
131immediately taken into account to evaluate the dirty memory limits and the
132other appears as 0 when read.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133
134==============================================================
135
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800136dirty_background_ratio
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137
Zheng Liu715ea412013-11-12 15:08:30 -0800138Contains, as a percentage of total available memory that contains free pages
139and reclaimable pages, the number of pages at which the background kernel
140flusher threads will start writing out dirty data.
141
Chris Dunlopd83e2a42015-09-18 16:10:55 +1000142The total available memory is not equal to total system memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143
144==============================================================
145
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800146dirty_bytes
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800148Contains the amount of dirty memory at which a process generating disk writes
149will itself start writeback.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150
Andrea Righiabffc022010-10-27 15:33:31 -0700151Note: dirty_bytes is the counterpart of dirty_ratio. Only one of them may be
152specified at a time. When one sysctl is written it is immediately taken into
153account to evaluate the dirty memory limits and the other appears as 0 when
154read.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800155
Andrea Righi9e4a5bd2009-04-30 15:08:57 -0700156Note: the minimum value allowed for dirty_bytes is two pages (in bytes); any
157value lower than this limit will be ignored and the old configuration will be
158retained.
159
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800160==============================================================
161
162dirty_expire_centisecs
163
164This tunable is used to define when dirty data is old enough to be eligible
Artem Bityutskiy6601fac2012-07-25 18:12:01 +0300165for writeout by the kernel flusher threads. It is expressed in 100'ths
166of a second. Data which has been dirty in-memory for longer than this
167interval will be written out next time a flusher thread wakes up.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800168
169==============================================================
170
171dirty_ratio
172
Zheng Liu715ea412013-11-12 15:08:30 -0800173Contains, as a percentage of total available memory that contains free pages
174and reclaimable pages, the number of pages at which a process which is
175generating disk writes will itself start writing out dirty data.
176
Chris Dunlopd83e2a42015-09-18 16:10:55 +1000177The total available memory is not equal to total system memory.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800178
179==============================================================
180
181dirty_writeback_centisecs
182
Artem Bityutskiy6601fac2012-07-25 18:12:01 +0300183The kernel flusher threads will periodically wake up and write `old' data
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800184out to disk. This tunable expresses the interval between those wakeups, in
185100'ths of a second.
186
187Setting this to zero disables periodic writeback altogether.
188
189==============================================================
190
191drop_caches
192
Dave Hansen5509a5d2014-04-03 14:48:19 -0700193Writing to this will cause the kernel to drop clean caches, as well as
194reclaimable slab objects like dentries and inodes. Once dropped, their
195memory becomes free.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800196
197To free pagecache:
198 echo 1 > /proc/sys/vm/drop_caches
Dave Hansen5509a5d2014-04-03 14:48:19 -0700199To free reclaimable slab objects (includes dentries and inodes):
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800200 echo 2 > /proc/sys/vm/drop_caches
Dave Hansen5509a5d2014-04-03 14:48:19 -0700201To free slab objects and pagecache:
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800202 echo 3 > /proc/sys/vm/drop_caches
203
Dave Hansen5509a5d2014-04-03 14:48:19 -0700204This is a non-destructive operation and will not free any dirty objects.
205To increase the number of objects freed by this operation, the user may run
206`sync' prior to writing to /proc/sys/vm/drop_caches. This will minimize the
207number of dirty objects on the system and create more candidates to be
208dropped.
209
210This file is not a means to control the growth of the various kernel caches
211(inodes, dentries, pagecache, etc...) These objects are automatically
212reclaimed by the kernel when memory is needed elsewhere on the system.
213
214Use of this file can cause performance problems. Since it discards cached
215objects, it may cost a significant amount of I/O and CPU to recreate the
216dropped objects, especially if they were under heavy use. Because of this,
217use outside of a testing or debugging environment is not recommended.
218
219You may see informational messages in your kernel log when this file is
220used:
221
222 cat (1234): drop_caches: 3
223
224These are informational only. They do not mean that anything is wrong
225with your system. To disable them, echo 4 (bit 3) into drop_caches.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800226
227==============================================================
228
Mel Gorman5e771902010-05-24 14:32:31 -0700229extfrag_threshold
230
231This parameter affects whether the kernel will compact memory or direct
Rabin Vincenta10726b2015-07-14 07:35:11 +0200232reclaim to satisfy a high-order allocation. The extfrag/extfrag_index file in
233debugfs shows what the fragmentation index for each order is in each zone in
234the system. Values tending towards 0 imply allocations would fail due to lack
235of memory, values towards 1000 imply failures are due to fragmentation and -1
236implies that the allocation will succeed as long as watermarks are met.
Mel Gorman5e771902010-05-24 14:32:31 -0700237
238The kernel will not compact memory in a zone if the
239fragmentation index is <= extfrag_threshold. The default value is 500.
240
241==============================================================
242
Michal Hockod09b6462017-07-10 15:49:38 -0700243highmem_is_dirtyable
244
245Available only for systems with CONFIG_HIGHMEM enabled (32b systems).
246
247This parameter controls whether the high memory is considered for dirty
248writers throttling. This is not the case by default which means that
249only the amount of memory directly visible/usable by the kernel can
250be dirtied. As a result, on systems with a large amount of memory and
251lowmem basically depleted writers might be throttled too early and
252streaming writes can get very slow.
253
254Changing the value to non zero would allow more memory to be dirtied
255and thus allow writers to write more data which can be flushed to the
256storage more effectively. Note this also comes with a risk of pre-mature
257OOM killer because some writers (e.g. direct block device writes) can
258only use the low memory and they can fill it up with dirty data without
259any throttling.
260
261==============================================================
262
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800263hugetlb_shm_group
264
265hugetlb_shm_group contains group id that is allowed to create SysV
266shared memory segment using hugetlb page.
267
268==============================================================
269
270laptop_mode
271
272laptop_mode is a knob that controls "laptop mode". All the things that are
273controlled by this knob are discussed in Documentation/laptops/laptop-mode.txt.
274
275==============================================================
276
277legacy_va_layout
278
Kulikov Vasiliy2174efb2010-06-28 13:59:28 +0200279If non-zero, this sysctl disables the new 32-bit mmap layout - the kernel
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800280will use the legacy (2.4) layout for all processes.
281
282==============================================================
283
284lowmem_reserve_ratio
285
286For some specialised workloads on highmem machines it is dangerous for
287the kernel to allow process memory to be allocated from the "lowmem"
288zone. This is because that memory could then be pinned via the mlock()
289system call, or by unavailability of swapspace.
290
291And on large highmem machines this lack of reclaimable lowmem memory
292can be fatal.
293
294So the Linux page allocator has a mechanism which prevents allocations
295which _could_ use highmem from using too much lowmem. This means that
296a certain amount of lowmem is defended from the possibility of being
297captured into pinned user memory.
298
299(The same argument applies to the old 16 megabyte ISA DMA region. This
300mechanism will also defend that region from allocations which could use
301highmem or lowmem).
302
303The `lowmem_reserve_ratio' tunable determines how aggressive the kernel is
304in defending these lower zones.
305
306If you have a machine which uses highmem or ISA DMA and your
307applications are using mlock(), or if you are running with no swap then
308you probably should change the lowmem_reserve_ratio setting.
309
310The lowmem_reserve_ratio is an array. You can see them by reading this file.
311-
312% cat /proc/sys/vm/lowmem_reserve_ratio
313256 256 32
314-
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800315
316But, these values are not used directly. The kernel calculates # of protection
317pages for each zones from them. These are shown as array of protection pages
318in /proc/zoneinfo like followings. (This is an example of x86-64 box).
319Each zone has an array of protection pages like this.
320
321-
322Node 0, zone DMA
323 pages free 1355
324 min 3
325 low 3
326 high 4
327 :
328 :
329 numa_other 0
330 protection: (0, 2004, 2004, 2004)
331 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
332 pagesets
333 cpu: 0 pcp: 0
334 :
335-
336These protections are added to score to judge whether this zone should be used
337for page allocation or should be reclaimed.
338
339In this example, if normal pages (index=2) are required to this DMA zone and
Mel Gorman41858962009-06-16 15:32:12 -0700340watermark[WMARK_HIGH] is used for watermark, the kernel judges this zone should
341not be used because pages_free(1355) is smaller than watermark + protection[2]
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800342(4 + 2004 = 2008). If this protection value is 0, this zone would be used for
343normal page requirement. If requirement is DMA zone(index=0), protection[0]
344(=0) is used.
345
346zone[i]'s protection[j] is calculated by following expression.
347
348(i < j):
349 zone[i]->protection[j]
Yaowei Bai013110a2015-09-08 15:04:10 -0700350 = (total sums of managed_pages from zone[i+1] to zone[j] on the node)
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800351 / lowmem_reserve_ratio[i];
352(i = j):
353 (should not be protected. = 0;
354(i > j):
355 (not necessary, but looks 0)
356
357The default values of lowmem_reserve_ratio[i] are
358 256 (if zone[i] means DMA or DMA32 zone)
359 32 (others).
360As above expression, they are reciprocal number of ratio.
Yaowei Bai013110a2015-09-08 15:04:10 -0700361256 means 1/256. # of protection pages becomes about "0.39%" of total managed
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800362pages of higher zones on the node.
363
364If you would like to protect more pages, smaller values are effective.
Joonsoo Kimd3cda232018-04-10 16:30:11 -0700365The minimum value is 1 (1/1 -> 100%). The value less than 1 completely
366disables protection of the pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367
368==============================================================
369
370max_map_count:
371
372This file contains the maximum number of memory map areas a process
373may have. Memory map areas are used as a side-effect of calling
David Rientjesdef5efe2017-02-24 14:58:47 -0800374malloc, directly by mmap, mprotect, and madvise, and also when loading
375shared libraries.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376
377While most applications need less than a thousand maps, certain
378programs, particularly malloc debuggers, may consume lots of them,
379e.g., up to one or two maps per allocation.
380
381The default value is 65536.
382
Andi Kleen6a460792009-09-16 11:50:15 +0200383=============================================================
384
385memory_failure_early_kill:
386
387Control how to kill processes when uncorrected memory error (typically
388a 2bit error in a memory module) is detected in the background by hardware
389that cannot be handled by the kernel. In some cases (like the page
390still having a valid copy on disk) the kernel will handle the failure
391transparently without affecting any applications. But if there is
392no other uptodate copy of the data it will kill to prevent any data
393corruptions from propagating.
394
3951: Kill all processes that have the corrupted and not reloadable page mapped
396as soon as the corruption is detected. Note this is not supported
397for a few types of pages, like kernel internally allocated data or
398the swap cache, but works for the majority of user pages.
399
4000: Only unmap the corrupted page from all processes and only kill a process
401who tries to access it.
402
403The kill is done using a catchable SIGBUS with BUS_MCEERR_AO, so processes can
404handle this if they want to.
405
406This is only active on architectures/platforms with advanced machine
407check handling and depends on the hardware capabilities.
408
409Applications can override this setting individually with the PR_MCE_KILL prctl
410
411==============================================================
412
413memory_failure_recovery
414
415Enable memory failure recovery (when supported by the platform)
416
4171: Attempt recovery.
418
4190: Always panic on a memory failure.
420
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421==============================================================
422
423min_free_kbytes:
424
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800425This is used to force the Linux VM to keep a minimum number
Mel Gorman41858962009-06-16 15:32:12 -0700426of kilobytes free. The VM uses this number to compute a
427watermark[WMARK_MIN] value for each lowmem zone in the system.
428Each lowmem zone gets a number of reserved free pages based
429proportionally on its size.
Rohit Seth8ad4b1f2006-01-08 01:00:40 -0800430
Matt LaPlanted9195882008-07-25 19:45:33 -0700431Some minimal amount of memory is needed to satisfy PF_MEMALLOC
Pavel Machek24950892007-10-16 23:31:28 -0700432allocations; if you set this to lower than 1024KB, your system will
433become subtly broken, and prone to deadlock under high loads.
434
435Setting this too high will OOM your machine instantly.
436
Christoph Lameter96146342006-07-03 00:24:13 -0700437=============================================================
438
Christoph Lameter0ff38492006-09-25 23:31:52 -0700439min_slab_ratio:
440
441This is available only on NUMA kernels.
442
443A percentage of the total pages in each zone. On Zone reclaim
444(fallback from the local zone occurs) slabs will be reclaimed if more
445than this percentage of pages in a zone are reclaimable slab pages.
446This insures that the slab growth stays under control even in NUMA
447systems that rarely perform global reclaim.
448
449The default is 5 percent.
450
451Note that slab reclaim is triggered in a per zone / node fashion.
452The process of reclaiming slab memory is currently not node specific
453and may not be fast.
454
455=============================================================
456
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800457min_unmapped_ratio:
KAMEZAWA Hiroyukifadd8fb2006-06-23 02:03:13 -0700458
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800459This is available only on NUMA kernels.
Yasunori Goto2b744c02007-05-06 14:49:59 -0700460
Mel Gorman90afa5d2009-06-16 15:33:20 -0700461This is a percentage of the total pages in each zone. Zone reclaim will
462only occur if more than this percentage of pages are in a state that
463zone_reclaim_mode allows to be reclaimed.
464
465If zone_reclaim_mode has the value 4 OR'd, then the percentage is compared
466against all file-backed unmapped pages including swapcache pages and tmpfs
467files. Otherwise, only unmapped pages backed by normal files but not tmpfs
468files and similar are considered.
Yasunori Goto2b744c02007-05-06 14:49:59 -0700469
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800470The default is 1 percent.
David Rientjesfe071d72007-10-16 23:25:56 -0700471
Eric Parised032182007-06-28 15:55:21 -0400472==============================================================
473
474mmap_min_addr
475
476This file indicates the amount of address space which a user process will
André Goddard Rosaaf901ca2009-11-14 13:09:05 -0200477be restricted from mmapping. Since kernel null dereference bugs could
Eric Parised032182007-06-28 15:55:21 -0400478accidentally operate based on the information in the first couple of pages
479of memory userspace processes should not be allowed to write to them. By
480default this value is set to 0 and no protections will be enforced by the
481security module. Setting this value to something like 64k will allow the
482vast majority of applications to work correctly and provide defense in depth
483against future potential kernel bugs.
484
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700485==============================================================
486
Daniel Cashmand07e2252016-01-14 15:19:53 -0800487mmap_rnd_bits:
488
489This value can be used to select the number of bits to use to
490determine the random offset to the base address of vma regions
491resulting from mmap allocations on architectures which support
492tuning address space randomization. This value will be bounded
493by the architecture's minimum and maximum supported values.
494
495This value can be changed after boot using the
496/proc/sys/vm/mmap_rnd_bits tunable
497
498==============================================================
499
500mmap_rnd_compat_bits:
501
502This value can be used to select the number of bits to use to
503determine the random offset to the base address of vma regions
504resulting from mmap allocations for applications run in
505compatibility mode on architectures which support tuning address
506space randomization. This value will be bounded by the
507architecture's minimum and maximum supported values.
508
509This value can be changed after boot using the
510/proc/sys/vm/mmap_rnd_compat_bits tunable
511
512==============================================================
513
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800514nr_hugepages
515
516Change the minimum size of the hugepage pool.
517
518See Documentation/vm/hugetlbpage.txt
519
520==============================================================
521
522nr_overcommit_hugepages
523
524Change the maximum size of the hugepage pool. The maximum is
525nr_hugepages + nr_overcommit_hugepages.
526
527See Documentation/vm/hugetlbpage.txt
528
529==============================================================
530
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800531nr_trim_pages
532
533This is available only on NOMMU kernels.
534
535This value adjusts the excess page trimming behaviour of power-of-2 aligned
536NOMMU mmap allocations.
537
538A value of 0 disables trimming of allocations entirely, while a value of 1
539trims excess pages aggressively. Any value >= 1 acts as the watermark where
540trimming of allocations is initiated.
541
542The default value is 1.
543
544See Documentation/nommu-mmap.txt for more information.
545
546==============================================================
547
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700548numa_zonelist_order
549
Michal Hockoc9bff3e2017-09-06 16:20:13 -0700550This sysctl is only for NUMA and it is deprecated. Anything but
551Node order will fail!
552
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700553'where the memory is allocated from' is controlled by zonelists.
554(This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation.
555 you may be able to read ZONE_DMA as ZONE_DMA32...)
556
557In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following.
558ZONE_NORMAL -> ZONE_DMA
559This means that a memory allocation request for GFP_KERNEL will
560get memory from ZONE_DMA only when ZONE_NORMAL is not available.
561
562In NUMA case, you can think of following 2 types of order.
563Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL
564
565(A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL
566(B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA.
567
568Type(A) offers the best locality for processes on Node(0), but ZONE_DMA
569will be used before ZONE_NORMAL exhaustion. This increases possibility of
570out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small.
571
572Type(B) cannot offer the best locality but is more robust against OOM of
573the DMA zone.
574
575Type(A) is called as "Node" order. Type (B) is "Zone" order.
576
577"Node order" orders the zonelists by node, then by zone within each node.
Paul Bolle5a3016a2011-04-06 11:09:55 +0200578Specify "[Nn]ode" for node order
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700579
580"Zone Order" orders the zonelists by zone type, then by node within each
Paul Bolle5a3016a2011-04-06 11:09:55 +0200581zone. Specify "[Zz]one" for zone order.
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700582
Xishi Qiu7c88a292016-04-28 16:19:11 -0700583Specify "[Dd]efault" to request automatic configuration.
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700584
Xishi Qiu7c88a292016-04-28 16:19:11 -0700585On 32-bit, the Normal zone needs to be preserved for allocations accessible
586by the kernel, so "zone" order will be selected.
587
588On 64-bit, devices that require DMA32/DMA are relatively rare, so "node"
589order will be selected.
590
591Default order is recommended unless this is causing problems for your
592system/application.
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800593
594==============================================================
595
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800596oom_dump_tasks
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800597
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -0800598Enables a system-wide task dump (excluding kernel threads) to be produced
599when the kernel performs an OOM-killing and includes such information as
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -0800600pid, uid, tgid, vm size, rss, pgtables_bytes, swapents, oom_score_adj
601score, and name. This is helpful to determine why the OOM killer was
602invoked, to identify the rogue task that caused it, and to determine why
603the OOM killer chose the task it did to kill.
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800604
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800605If this is set to zero, this information is suppressed. On very
606large systems with thousands of tasks it may not be feasible to dump
607the memory state information for each one. Such systems should not
608be forced to incur a performance penalty in OOM conditions when the
609information may not be desired.
610
611If this is set to non-zero, this information is shown whenever the
612OOM killer actually kills a memory-hogging task.
613
David Rientjesad915c42010-08-09 17:18:53 -0700614The default value is 1 (enabled).
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800615
616==============================================================
617
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800618oom_kill_allocating_task
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800619
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800620This enables or disables killing the OOM-triggering task in
621out-of-memory situations.
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800622
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800623If this is set to zero, the OOM killer will scan through the entire
624tasklist and select a task based on heuristics to kill. This normally
625selects a rogue memory-hogging task that frees up a large amount of
626memory when killed.
627
628If this is set to non-zero, the OOM killer simply kills the task that
629triggered the out-of-memory condition. This avoids the expensive
630tasklist scan.
631
632If panic_on_oom is selected, it takes precedence over whatever value
633is used in oom_kill_allocating_task.
634
635The default value is 0.
Paul Mundtdd8632a2009-01-08 12:04:47 +0000636
637==============================================================
638
Jerome Marchand49f0ce52014-01-21 15:49:14 -0800639overcommit_kbytes:
640
641When overcommit_memory is set to 2, the committed address space is not
642permitted to exceed swap plus this amount of physical RAM. See below.
643
644Note: overcommit_kbytes is the counterpart of overcommit_ratio. Only one
645of them may be specified at a time. Setting one disables the other (which
646then appears as 0 when read).
647
648==============================================================
649
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800650overcommit_memory:
Paul Mundtdd8632a2009-01-08 12:04:47 +0000651
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800652This value contains a flag that enables memory overcommitment.
Paul Mundtdd8632a2009-01-08 12:04:47 +0000653
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800654When this flag is 0, the kernel attempts to estimate the amount
655of free memory left when userspace requests more memory.
Paul Mundtdd8632a2009-01-08 12:04:47 +0000656
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800657When this flag is 1, the kernel pretends there is always enough
658memory until it actually runs out.
Paul Mundtdd8632a2009-01-08 12:04:47 +0000659
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800660When this flag is 2, the kernel uses a "never overcommit"
661policy that attempts to prevent any overcommit of memory.
Andrew Shewmakerc9b1d092013-04-29 15:08:10 -0700662Note that user_reserve_kbytes affects this policy.
Paul Mundtdd8632a2009-01-08 12:04:47 +0000663
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800664This feature can be very useful because there are a lot of
665programs that malloc() huge amounts of memory "just-in-case"
666and don't use much of it.
667
668The default value is 0.
669
670See Documentation/vm/overcommit-accounting and
Chun Chenc56050c2015-11-09 14:58:15 -0800671mm/mmap.c::__vm_enough_memory() for more information.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800672
673==============================================================
674
675overcommit_ratio:
676
677When overcommit_memory is set to 2, the committed address
678space is not permitted to exceed swap plus this percentage
679of physical RAM. See above.
680
681==============================================================
682
683page-cluster
684
Christian Ehrhardtdf858fa2012-07-31 16:41:46 -0700685page-cluster controls the number of pages up to which consecutive pages
686are read in from swap in a single attempt. This is the swap counterpart
687to page cache readahead.
688The mentioned consecutivity is not in terms of virtual/physical addresses,
689but consecutive on swap space - that means they were swapped out together.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800690
691It is a logarithmic value - setting it to zero means "1 page", setting
692it to 1 means "2 pages", setting it to 2 means "4 pages", etc.
Christian Ehrhardtdf858fa2012-07-31 16:41:46 -0700693Zero disables swap readahead completely.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800694
695The default value is three (eight pages at a time). There may be some
696small benefits in tuning this to a different value if your workload is
697swap-intensive.
698
Christian Ehrhardtdf858fa2012-07-31 16:41:46 -0700699Lower values mean lower latencies for initial faults, but at the same time
700extra faults and I/O delays for following faults if they would have been part of
701that consecutive pages readahead would have brought in.
702
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800703=============================================================
704
705panic_on_oom
706
707This enables or disables panic on out-of-memory feature.
708
709If this is set to 0, the kernel will kill some rogue process,
710called oom_killer. Usually, oom_killer can kill rogue processes and
711system will survive.
712
713If this is set to 1, the kernel panics when out-of-memory happens.
714However, if a process limits using nodes by mempolicy/cpusets,
715and those nodes become memory exhaustion status, one process
716may be killed by oom-killer. No panic occurs in this case.
717Because other nodes' memory may be free. This means system total status
718may be not fatal yet.
719
720If this is set to 2, the kernel panics compulsorily even on the
KAMEZAWA Hiroyukidaaf1e62010-03-10 15:22:32 -0800721above-mentioned. Even oom happens under memory cgroup, the whole
722system panics.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800723
724The default value is 0.
7251 and 2 are for failover of clustering. Please select either
726according to your policy of failover.
KAMEZAWA Hiroyukidaaf1e62010-03-10 15:22:32 -0800727panic_on_oom=2+kdump gives you very strong tool to investigate
728why oom happens. You can get snapshot.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800729
730=============================================================
731
732percpu_pagelist_fraction
733
734This is the fraction of pages at most (high mark pcp->high) in each zone that
735are allocated for each per cpu page list. The min value for this is 8. It
736means that we don't allow more than 1/8th of pages in each zone to be
737allocated in any single per_cpu_pagelist. This entry only changes the value
738of hot per cpu pagelists. User can specify a number like 100 to allocate
7391/100th of each zone to each per cpu page list.
740
741The batch value of each per cpu pagelist is also updated as a result. It is
742set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8)
743
744The initial value is zero. Kernel does not use this value at boot time to set
David Rientjes7cd2b0a2014-06-23 13:22:04 -0700745the high water marks for each per cpu page list. If the user writes '0' to this
746sysctl, it will revert to this default behavior.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800747
748==============================================================
749
750stat_interval
751
752The time interval between which vm statistics are updated. The default
753is 1 second.
754
755==============================================================
756
Hugh Dickins52b6f462016-05-19 17:12:50 -0700757stat_refresh
758
759Any read or write (by root only) flushes all the per-cpu vm statistics
760into their global totals, for more accurate reports when testing
761e.g. cat /proc/sys/vm/stat_refresh /proc/meminfo
762
763As a side-effect, it also checks for negative totals (elsewhere reported
764as 0) and "fails" with EINVAL if any are found, with a warning in dmesg.
765(At time of writing, a few stats are known sometimes to be found negative,
766with no ill effects: errors and warnings on these stats are suppressed.)
767
768==============================================================
769
Kemi Wang45180852017-11-15 17:38:22 -0800770numa_stat
771
772This interface allows runtime configuration of numa statistics.
773
774When page allocation performance becomes a bottleneck and you can tolerate
775some possible tool breakage and decreased numa counter precision, you can
776do:
777 echo 0 > /proc/sys/vm/numa_stat
778
779When page allocation performance is not a bottleneck and you want all
780tooling to work, you can do:
781 echo 1 > /proc/sys/vm/numa_stat
782
783==============================================================
784
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800785swappiness
786
787This control is used to define how aggressive the kernel will swap
Kangmin Park27432322017-11-17 15:30:23 -0800788memory pages. Higher values will increase aggressiveness, lower values
Aaron Tomlin8582cb92014-01-29 14:05:38 -0800789decrease the amount of swap. A value of 0 instructs the kernel not to
790initiate swap until the amount of free and file-backed pages is less
791than the high water mark in a zone.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800792
793The default value is 60.
794
795==============================================================
796
Andrew Shewmakerc9b1d092013-04-29 15:08:10 -0700797- user_reserve_kbytes
798
Masanari Iida633708a2015-01-02 12:03:19 +0900799When overcommit_memory is set to 2, "never overcommit" mode, reserve
Andrew Shewmakerc9b1d092013-04-29 15:08:10 -0700800min(3% of current process size, user_reserve_kbytes) of free memory.
801This is intended to prevent a user from starting a single memory hogging
802process, such that they cannot recover (kill the hog).
803
804user_reserve_kbytes defaults to min(3% of the current process size, 128MB).
805
806If this is reduced to zero, then the user will be allowed to allocate
807all free memory with a single process, minus admin_reserve_kbytes.
808Any subsequent attempts to execute a command will result in
809"fork: Cannot allocate memory".
810
811Changing this takes effect whenever an application requests memory.
812
813==============================================================
814
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800815vfs_cache_pressure
816------------------
817
Denys Vlasenko4a0da712014-06-04 16:11:03 -0700818This percentage value controls the tendency of the kernel to reclaim
819the memory which is used for caching of directory and inode objects.
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800820
821At the default value of vfs_cache_pressure=100 the kernel will attempt to
822reclaim dentries and inodes at a "fair" rate with respect to pagecache and
823swapcache reclaim. Decreasing vfs_cache_pressure causes the kernel to prefer
Jan Kara55c37a82009-09-21 17:01:40 -0700824to retain dentry and inode caches. When vfs_cache_pressure=0, the kernel will
825never reclaim dentries and inodes due to memory pressure and this can easily
826lead to out-of-memory conditions. Increasing vfs_cache_pressure beyond 100
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800827causes the kernel to prefer to reclaim dentries and inodes.
828
Denys Vlasenko4a0da712014-06-04 16:11:03 -0700829Increasing vfs_cache_pressure significantly beyond 100 may have negative
830performance impact. Reclaim code needs to take various locks to find freeable
831directory and inode objects. With vfs_cache_pressure=1000, it will look for
832ten times more freeable objects than there are.
833
Johannes Weiner795ae7a2016-03-17 14:19:14 -0700834=============================================================
835
836watermark_scale_factor:
837
838This factor controls the aggressiveness of kswapd. It defines the
839amount of memory left in a node/system before kswapd is woken up and
840how much memory needs to be free before kswapd goes back to sleep.
841
842The unit is in fractions of 10,000. The default value of 10 means the
843distances between watermarks are 0.1% of the available memory in the
844node/system. The maximum value is 1000, or 10% of memory.
845
846A high rate of threads entering direct reclaim (allocstall) or kswapd
847going to sleep prematurely (kswapd_low_wmark_hit_quickly) can indicate
848that the number of free pages kswapd maintains for latency reasons is
849too small for the allocation bursts occurring in the system. This knob
850can then be used to tune kswapd aggressiveness accordingly.
851
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800852==============================================================
853
854zone_reclaim_mode:
855
856Zone_reclaim_mode allows someone to set more or less aggressive approaches to
857reclaim memory when a zone runs out of memory. If it is set to zero then no
858zone reclaim occurs. Allocations will be satisfied from other zones / nodes
859in the system.
860
861This is value ORed together of
862
8631 = Zone reclaim on
8642 = Zone reclaim writes dirty pages out
8654 = Zone reclaim swaps pages
866
Mel Gorman4f9b16a2014-06-04 16:07:14 -0700867zone_reclaim_mode is disabled by default. For file servers or workloads
868that benefit from having their data cached, zone_reclaim_mode should be
869left disabled as the caching effect is likely to be more important than
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800870data locality.
871
Mel Gorman4f9b16a2014-06-04 16:07:14 -0700872zone_reclaim may be enabled if it's known that the workload is partitioned
873such that each partition fits within a NUMA node and that accessing remote
874memory would cause a measurable performance reduction. The page allocator
875will then reclaim easily reusable pages (those page cache pages that are
876currently not used) before allocating off node pages.
877
Peter W Morrealedb0fb182009-01-15 13:50:42 -0800878Allowing zone reclaim to write out pages stops processes that are
879writing large amounts of data from dirtying pages on other nodes. Zone
880reclaim will write out dirty pages if a zone fills up and so effectively
881throttle the process. This may decrease the performance of a single process
882since it cannot use all of system memory to buffer the outgoing writes
883anymore but it preserve the memory on other nodes so that the performance
884of other processes running on other nodes will not be affected.
885
886Allowing regular swap effectively restricts allocations to the local
887node unless explicitly overridden by memory policies or cpuset
888configurations.
889
890============ End of Document =================================