blob: a3415070bcac1fe6e16c8eafebccea30e3c55cb7 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001Documentation for /proc/sys/vm/* kernel version 2.2.10
2 (c) 1998, 1999, Rik van Riel <riel@nl.linux.org>
3
4For general info and legal blurb, please look in README.
5
6==============================================================
7
8This file contains the documentation for the sysctl files in
9/proc/sys/vm and is valid for Linux kernel version 2.2.
10
11The files in this directory can be used to tune the operation
12of the virtual memory (VM) subsystem of the Linux kernel and
13the writeout of dirty data to disk.
14
15Default values and initialization routines for most of these
16files can be found in mm/swap.c.
17
18Currently, these files are in /proc/sys/vm:
19- overcommit_memory
20- page-cluster
21- dirty_ratio
22- dirty_background_ratio
23- dirty_expire_centisecs
24- dirty_writeback_centisecs
Bron Gondwana195cf4532008-02-04 22:29:20 -080025- highmem_is_dirtyable (only if CONFIG_HIGHMEM set)
Linus Torvalds1da177e2005-04-16 15:20:36 -070026- max_map_count
27- min_free_kbytes
28- laptop_mode
29- block_dump
Andrew Morton9d0243b2006-01-08 01:00:39 -080030- drop-caches
Christoph Lameter17436602006-01-18 17:42:32 -080031- zone_reclaim_mode
Christoph Lameter96146342006-07-03 00:24:13 -070032- min_unmapped_ratio
Christoph Lameter0ff38492006-09-25 23:31:52 -070033- min_slab_ratio
KAMEZAWA Hiroyukifadd8fb2006-06-23 02:03:13 -070034- panic_on_oom
David Rientjesfef1bdd2008-02-07 00:14:07 -080035- oom_dump_tasks
David Rientjesfe071d72007-10-16 23:25:56 -070036- oom_kill_allocating_task
Eric Parised032182007-06-28 15:55:21 -040037- mmap_min_address
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -070038- numa_zonelist_order
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -080039- nr_hugepages
40- nr_overcommit_hugepages
Paul Mundtdd8632a2009-01-08 12:04:47 +000041- nr_trim_pages (only if CONFIG_MMU=n)
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
43==============================================================
44
David Rientjes2da02992009-01-06 14:39:31 -080045dirty_bytes, dirty_ratio, dirty_background_bytes,
46dirty_background_ratio, dirty_expire_centisecs,
Bron Gondwana195cf4532008-02-04 22:29:20 -080047dirty_writeback_centisecs, highmem_is_dirtyable,
48vfs_cache_pressure, laptop_mode, block_dump, swap_token_timeout,
49drop-caches, hugepages_treat_as_movable:
Linus Torvalds1da177e2005-04-16 15:20:36 -070050
51See Documentation/filesystems/proc.txt
52
53==============================================================
54
55overcommit_memory:
56
57This value contains a flag that enables memory overcommitment.
58
59When this flag is 0, the kernel attempts to estimate the amount
60of free memory left when userspace requests more memory.
61
62When this flag is 1, the kernel pretends there is always enough
63memory until it actually runs out.
64
65When this flag is 2, the kernel uses a "never overcommit"
66policy that attempts to prevent any overcommit of memory.
67
68This feature can be very useful because there are a lot of
69programs that malloc() huge amounts of memory "just-in-case"
70and don't use much of it.
71
72The default value is 0.
73
74See Documentation/vm/overcommit-accounting and
75security/commoncap.c::cap_vm_enough_memory() for more information.
76
77==============================================================
78
79overcommit_ratio:
80
81When overcommit_memory is set to 2, the committed address
82space is not permitted to exceed swap plus this percentage
83of physical RAM. See above.
84
85==============================================================
86
87page-cluster:
88
89The Linux VM subsystem avoids excessive disk seeks by reading
90multiple pages on a page fault. The number of pages it reads
91is dependent on the amount of memory in your machine.
92
93The number of pages the kernel reads in at once is equal to
942 ^ page-cluster. Values above 2 ^ 5 don't make much sense
95for swap because we only cluster swap data in 32-page groups.
96
97==============================================================
98
99max_map_count:
100
101This file contains the maximum number of memory map areas a process
102may have. Memory map areas are used as a side-effect of calling
103malloc, directly by mmap and mprotect, and also when loading shared
104libraries.
105
106While most applications need less than a thousand maps, certain
107programs, particularly malloc debuggers, may consume lots of them,
108e.g., up to one or two maps per allocation.
109
110The default value is 65536.
111
112==============================================================
113
114min_free_kbytes:
115
116This is used to force the Linux VM to keep a minimum number
117of kilobytes free. The VM uses this number to compute a pages_min
118value for each lowmem zone in the system. Each lowmem zone gets
119a number of reserved free pages based proportionally on its size.
Rohit Seth8ad4b1f2006-01-08 01:00:40 -0800120
Matt LaPlanted9195882008-07-25 19:45:33 -0700121Some minimal amount of memory is needed to satisfy PF_MEMALLOC
Pavel Machek24950892007-10-16 23:31:28 -0700122allocations; if you set this to lower than 1024KB, your system will
123become subtly broken, and prone to deadlock under high loads.
124
125Setting this too high will OOM your machine instantly.
126
Rohit Seth8ad4b1f2006-01-08 01:00:40 -0800127==============================================================
128
129percpu_pagelist_fraction
130
131This is the fraction of pages at most (high mark pcp->high) in each zone that
132are allocated for each per cpu page list. The min value for this is 8. It
133means that we don't allow more than 1/8th of pages in each zone to be
134allocated in any single per_cpu_pagelist. This entry only changes the value
135of hot per cpu pagelists. User can specify a number like 100 to allocate
1361/100th of each zone to each per cpu page list.
137
138The batch value of each per cpu pagelist is also updated as a result. It is
139set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8)
140
141The initial value is zero. Kernel does not use this value at boot time to set
142the high water marks for each per cpu page list.
Christoph Lameter17436602006-01-18 17:42:32 -0800143
144===============================================================
145
146zone_reclaim_mode:
147
Matt LaPlante5d3f0832006-11-30 05:21:10 +0100148Zone_reclaim_mode allows someone to set more or less aggressive approaches to
Christoph Lameter1b2ffb72006-02-01 03:05:34 -0800149reclaim memory when a zone runs out of memory. If it is set to zero then no
150zone reclaim occurs. Allocations will be satisfied from other zones / nodes
151in the system.
152
153This is value ORed together of
154
1551 = Zone reclaim on
1562 = Zone reclaim writes dirty pages out
1574 = Zone reclaim swaps pages
158
159zone_reclaim_mode is set during bootup to 1 if it is determined that pages
160from remote zones will cause a measurable performance reduction. The
Christoph Lameter17436602006-01-18 17:42:32 -0800161page allocator will then reclaim easily reusable pages (those page
Christoph Lameter1b2ffb72006-02-01 03:05:34 -0800162cache pages that are currently not used) before allocating off node pages.
Christoph Lameter17436602006-01-18 17:42:32 -0800163
Christoph Lameter1b2ffb72006-02-01 03:05:34 -0800164It may be beneficial to switch off zone reclaim if the system is
165used for a file server and all of memory should be used for caching files
166from disk. In that case the caching effect is more important than
167data locality.
Christoph Lameter17436602006-01-18 17:42:32 -0800168
Christoph Lameter1b2ffb72006-02-01 03:05:34 -0800169Allowing zone reclaim to write out pages stops processes that are
170writing large amounts of data from dirtying pages on other nodes. Zone
171reclaim will write out dirty pages if a zone fills up and so effectively
172throttle the process. This may decrease the performance of a single process
173since it cannot use all of system memory to buffer the outgoing writes
174anymore but it preserve the memory on other nodes so that the performance
175of other processes running on other nodes will not be affected.
176
177Allowing regular swap effectively restricts allocations to the local
178node unless explicitly overridden by memory policies or cpuset
179configurations.
180
KAMEZAWA Hiroyukifadd8fb2006-06-23 02:03:13 -0700181=============================================================
182
Christoph Lameter96146342006-07-03 00:24:13 -0700183min_unmapped_ratio:
184
185This is available only on NUMA kernels.
186
Christoph Lameter0ff38492006-09-25 23:31:52 -0700187A percentage of the total pages in each zone. Zone reclaim will only
Christoph Lameter96146342006-07-03 00:24:13 -0700188occur if more than this percentage of pages are file backed and unmapped.
189This is to insure that a minimal amount of local pages is still available for
190file I/O even if the node is overallocated.
191
192The default is 1 percent.
193
194=============================================================
195
Christoph Lameter0ff38492006-09-25 23:31:52 -0700196min_slab_ratio:
197
198This is available only on NUMA kernels.
199
200A percentage of the total pages in each zone. On Zone reclaim
201(fallback from the local zone occurs) slabs will be reclaimed if more
202than this percentage of pages in a zone are reclaimable slab pages.
203This insures that the slab growth stays under control even in NUMA
204systems that rarely perform global reclaim.
205
206The default is 5 percent.
207
208Note that slab reclaim is triggered in a per zone / node fashion.
209The process of reclaiming slab memory is currently not node specific
210and may not be fast.
211
212=============================================================
213
KAMEZAWA Hiroyukifadd8fb2006-06-23 02:03:13 -0700214panic_on_oom
215
Yasunori Goto2b744c02007-05-06 14:49:59 -0700216This enables or disables panic on out-of-memory feature.
217
218If this is set to 0, the kernel will kill some rogue process,
219called oom_killer. Usually, oom_killer can kill rogue processes and
220system will survive.
221
222If this is set to 1, the kernel panics when out-of-memory happens.
223However, if a process limits using nodes by mempolicy/cpusets,
224and those nodes become memory exhaustion status, one process
225may be killed by oom-killer. No panic occurs in this case.
226Because other nodes' memory may be free. This means system total status
227may be not fatal yet.
228
229If this is set to 2, the kernel panics compulsorily even on the
230above-mentioned.
KAMEZAWA Hiroyukifadd8fb2006-06-23 02:03:13 -0700231
232The default value is 0.
Yasunori Goto2b744c02007-05-06 14:49:59 -07002331 and 2 are for failover of clustering. Please select either
234according to your policy of failover.
Eric Parised032182007-06-28 15:55:21 -0400235
David Rientjesfe071d72007-10-16 23:25:56 -0700236=============================================================
237
David Rientjesfef1bdd2008-02-07 00:14:07 -0800238oom_dump_tasks
239
240Enables a system-wide task dump (excluding kernel threads) to be
241produced when the kernel performs an OOM-killing and includes such
242information as pid, uid, tgid, vm size, rss, cpu, oom_adj score, and
243name. This is helpful to determine why the OOM killer was invoked
244and to identify the rogue task that caused it.
245
246If this is set to zero, this information is suppressed. On very
247large systems with thousands of tasks it may not be feasible to dump
248the memory state information for each one. Such systems should not
249be forced to incur a performance penalty in OOM conditions when the
250information may not be desired.
251
252If this is set to non-zero, this information is shown whenever the
253OOM killer actually kills a memory-hogging task.
254
255The default value is 0.
256
257=============================================================
258
David Rientjesfe071d72007-10-16 23:25:56 -0700259oom_kill_allocating_task
260
261This enables or disables killing the OOM-triggering task in
262out-of-memory situations.
263
264If this is set to zero, the OOM killer will scan through the entire
265tasklist and select a task based on heuristics to kill. This normally
266selects a rogue memory-hogging task that frees up a large amount of
267memory when killed.
268
269If this is set to non-zero, the OOM killer simply kills the task that
270triggered the out-of-memory condition. This avoids the expensive
271tasklist scan.
272
273If panic_on_oom is selected, it takes precedence over whatever value
274is used in oom_kill_allocating_task.
275
276The default value is 0.
277
Eric Parised032182007-06-28 15:55:21 -0400278==============================================================
279
280mmap_min_addr
281
282This file indicates the amount of address space which a user process will
283be restricted from mmaping. Since kernel null dereference bugs could
284accidentally operate based on the information in the first couple of pages
285of memory userspace processes should not be allowed to write to them. By
286default this value is set to 0 and no protections will be enforced by the
287security module. Setting this value to something like 64k will allow the
288vast majority of applications to work correctly and provide defense in depth
289against future potential kernel bugs.
290
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700291==============================================================
292
293numa_zonelist_order
294
295This sysctl is only for NUMA.
296'where the memory is allocated from' is controlled by zonelists.
297(This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation.
298 you may be able to read ZONE_DMA as ZONE_DMA32...)
299
300In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following.
301ZONE_NORMAL -> ZONE_DMA
302This means that a memory allocation request for GFP_KERNEL will
303get memory from ZONE_DMA only when ZONE_NORMAL is not available.
304
305In NUMA case, you can think of following 2 types of order.
306Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL
307
308(A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL
309(B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA.
310
311Type(A) offers the best locality for processes on Node(0), but ZONE_DMA
312will be used before ZONE_NORMAL exhaustion. This increases possibility of
313out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small.
314
315Type(B) cannot offer the best locality but is more robust against OOM of
316the DMA zone.
317
318Type(A) is called as "Node" order. Type (B) is "Zone" order.
319
320"Node order" orders the zonelists by node, then by zone within each node.
321Specify "[Nn]ode" for zone order
322
323"Zone Order" orders the zonelists by zone type, then by node within each
324zone. Specify "[Zz]one"for zode order.
325
326Specify "[Dd]efault" to request automatic configuration. Autoconfiguration
327will select "node" order in following case.
328(1) if the DMA zone does not exist or
329(2) if the DMA zone comprises greater than 50% of the available memory or
330(3) if any node's DMA zone comprises greater than 60% of its local memory and
331 the amount of local memory is big enough.
332
333Otherwise, "zone" order will be selected. Default order is recommended unless
334this is causing problems for your system/application.
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800335
336==============================================================
337
338nr_hugepages
339
340Change the minimum size of the hugepage pool.
341
342See Documentation/vm/hugetlbpage.txt
343
344==============================================================
345
346nr_overcommit_hugepages
347
348Change the maximum size of the hugepage pool. The maximum is
349nr_hugepages + nr_overcommit_hugepages.
350
351See Documentation/vm/hugetlbpage.txt
Paul Mundtdd8632a2009-01-08 12:04:47 +0000352
353==============================================================
354
355nr_trim_pages
356
357This is available only on NOMMU kernels.
358
359This value adjusts the excess page trimming behaviour of power-of-2 aligned
360NOMMU mmap allocations.
361
362A value of 0 disables trimming of allocations entirely, while a value of 1
363trims excess pages aggressively. Any value >= 1 acts as the watermark where
364trimming of allocations is initiated.
365
366The default value is 1.
367
368See Documentation/nommu-mmap.txt for more information.