blob: d79eeda7a699cbeb72ad0a03332388f8d8a68eeb [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
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
42==============================================================
43
44dirty_ratio, dirty_background_ratio, dirty_expire_centisecs,
Bron Gondwana195cf4532008-02-04 22:29:20 -080045dirty_writeback_centisecs, highmem_is_dirtyable,
46vfs_cache_pressure, laptop_mode, block_dump, swap_token_timeout,
47drop-caches, hugepages_treat_as_movable:
Linus Torvalds1da177e2005-04-16 15:20:36 -070048
49See Documentation/filesystems/proc.txt
50
51==============================================================
52
53overcommit_memory:
54
55This value contains a flag that enables memory overcommitment.
56
57When this flag is 0, the kernel attempts to estimate the amount
58of free memory left when userspace requests more memory.
59
60When this flag is 1, the kernel pretends there is always enough
61memory until it actually runs out.
62
63When this flag is 2, the kernel uses a "never overcommit"
64policy that attempts to prevent any overcommit of memory.
65
66This feature can be very useful because there are a lot of
67programs that malloc() huge amounts of memory "just-in-case"
68and don't use much of it.
69
70The default value is 0.
71
72See Documentation/vm/overcommit-accounting and
73security/commoncap.c::cap_vm_enough_memory() for more information.
74
75==============================================================
76
77overcommit_ratio:
78
79When overcommit_memory is set to 2, the committed address
80space is not permitted to exceed swap plus this percentage
81of physical RAM. See above.
82
83==============================================================
84
85page-cluster:
86
87The Linux VM subsystem avoids excessive disk seeks by reading
88multiple pages on a page fault. The number of pages it reads
89is dependent on the amount of memory in your machine.
90
91The number of pages the kernel reads in at once is equal to
922 ^ page-cluster. Values above 2 ^ 5 don't make much sense
93for swap because we only cluster swap data in 32-page groups.
94
95==============================================================
96
97max_map_count:
98
99This file contains the maximum number of memory map areas a process
100may have. Memory map areas are used as a side-effect of calling
101malloc, directly by mmap and mprotect, and also when loading shared
102libraries.
103
104While most applications need less than a thousand maps, certain
105programs, particularly malloc debuggers, may consume lots of them,
106e.g., up to one or two maps per allocation.
107
108The default value is 65536.
109
110==============================================================
111
112min_free_kbytes:
113
114This is used to force the Linux VM to keep a minimum number
115of kilobytes free. The VM uses this number to compute a pages_min
116value for each lowmem zone in the system. Each lowmem zone gets
117a number of reserved free pages based proportionally on its size.
Rohit Seth8ad4b1f2006-01-08 01:00:40 -0800118
Matt LaPlanted9195882008-07-25 19:45:33 -0700119Some minimal amount of memory is needed to satisfy PF_MEMALLOC
Pavel Machek24950892007-10-16 23:31:28 -0700120allocations; if you set this to lower than 1024KB, your system will
121become subtly broken, and prone to deadlock under high loads.
122
123Setting this too high will OOM your machine instantly.
124
Rohit Seth8ad4b1f2006-01-08 01:00:40 -0800125==============================================================
126
127percpu_pagelist_fraction
128
129This is the fraction of pages at most (high mark pcp->high) in each zone that
130are allocated for each per cpu page list. The min value for this is 8. It
131means that we don't allow more than 1/8th of pages in each zone to be
132allocated in any single per_cpu_pagelist. This entry only changes the value
133of hot per cpu pagelists. User can specify a number like 100 to allocate
1341/100th of each zone to each per cpu page list.
135
136The batch value of each per cpu pagelist is also updated as a result. It is
137set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8)
138
139The initial value is zero. Kernel does not use this value at boot time to set
140the high water marks for each per cpu page list.
Christoph Lameter17436602006-01-18 17:42:32 -0800141
142===============================================================
143
144zone_reclaim_mode:
145
Matt LaPlante5d3f0832006-11-30 05:21:10 +0100146Zone_reclaim_mode allows someone to set more or less aggressive approaches to
Christoph Lameter1b2ffb72006-02-01 03:05:34 -0800147reclaim memory when a zone runs out of memory. If it is set to zero then no
148zone reclaim occurs. Allocations will be satisfied from other zones / nodes
149in the system.
150
151This is value ORed together of
152
1531 = Zone reclaim on
1542 = Zone reclaim writes dirty pages out
1554 = Zone reclaim swaps pages
156
157zone_reclaim_mode is set during bootup to 1 if it is determined that pages
158from remote zones will cause a measurable performance reduction. The
Christoph Lameter17436602006-01-18 17:42:32 -0800159page allocator will then reclaim easily reusable pages (those page
Christoph Lameter1b2ffb72006-02-01 03:05:34 -0800160cache pages that are currently not used) before allocating off node pages.
Christoph Lameter17436602006-01-18 17:42:32 -0800161
Christoph Lameter1b2ffb72006-02-01 03:05:34 -0800162It may be beneficial to switch off zone reclaim if the system is
163used for a file server and all of memory should be used for caching files
164from disk. In that case the caching effect is more important than
165data locality.
Christoph Lameter17436602006-01-18 17:42:32 -0800166
Christoph Lameter1b2ffb72006-02-01 03:05:34 -0800167Allowing zone reclaim to write out pages stops processes that are
168writing large amounts of data from dirtying pages on other nodes. Zone
169reclaim will write out dirty pages if a zone fills up and so effectively
170throttle the process. This may decrease the performance of a single process
171since it cannot use all of system memory to buffer the outgoing writes
172anymore but it preserve the memory on other nodes so that the performance
173of other processes running on other nodes will not be affected.
174
175Allowing regular swap effectively restricts allocations to the local
176node unless explicitly overridden by memory policies or cpuset
177configurations.
178
KAMEZAWA Hiroyukifadd8fb2006-06-23 02:03:13 -0700179=============================================================
180
Christoph Lameter96146342006-07-03 00:24:13 -0700181min_unmapped_ratio:
182
183This is available only on NUMA kernels.
184
Christoph Lameter0ff38492006-09-25 23:31:52 -0700185A percentage of the total pages in each zone. Zone reclaim will only
Christoph Lameter96146342006-07-03 00:24:13 -0700186occur if more than this percentage of pages are file backed and unmapped.
187This is to insure that a minimal amount of local pages is still available for
188file I/O even if the node is overallocated.
189
190The default is 1 percent.
191
192=============================================================
193
Christoph Lameter0ff38492006-09-25 23:31:52 -0700194min_slab_ratio:
195
196This is available only on NUMA kernels.
197
198A percentage of the total pages in each zone. On Zone reclaim
199(fallback from the local zone occurs) slabs will be reclaimed if more
200than this percentage of pages in a zone are reclaimable slab pages.
201This insures that the slab growth stays under control even in NUMA
202systems that rarely perform global reclaim.
203
204The default is 5 percent.
205
206Note that slab reclaim is triggered in a per zone / node fashion.
207The process of reclaiming slab memory is currently not node specific
208and may not be fast.
209
210=============================================================
211
KAMEZAWA Hiroyukifadd8fb2006-06-23 02:03:13 -0700212panic_on_oom
213
Yasunori Goto2b744c02007-05-06 14:49:59 -0700214This enables or disables panic on out-of-memory feature.
215
216If this is set to 0, the kernel will kill some rogue process,
217called oom_killer. Usually, oom_killer can kill rogue processes and
218system will survive.
219
220If this is set to 1, the kernel panics when out-of-memory happens.
221However, if a process limits using nodes by mempolicy/cpusets,
222and those nodes become memory exhaustion status, one process
223may be killed by oom-killer. No panic occurs in this case.
224Because other nodes' memory may be free. This means system total status
225may be not fatal yet.
226
227If this is set to 2, the kernel panics compulsorily even on the
228above-mentioned.
KAMEZAWA Hiroyukifadd8fb2006-06-23 02:03:13 -0700229
230The default value is 0.
Yasunori Goto2b744c02007-05-06 14:49:59 -07002311 and 2 are for failover of clustering. Please select either
232according to your policy of failover.
Eric Parised032182007-06-28 15:55:21 -0400233
David Rientjesfe071d72007-10-16 23:25:56 -0700234=============================================================
235
David Rientjesfef1bdd2008-02-07 00:14:07 -0800236oom_dump_tasks
237
238Enables a system-wide task dump (excluding kernel threads) to be
239produced when the kernel performs an OOM-killing and includes such
240information as pid, uid, tgid, vm size, rss, cpu, oom_adj score, and
241name. This is helpful to determine why the OOM killer was invoked
242and to identify the rogue task that caused it.
243
244If this is set to zero, this information is suppressed. On very
245large systems with thousands of tasks it may not be feasible to dump
246the memory state information for each one. Such systems should not
247be forced to incur a performance penalty in OOM conditions when the
248information may not be desired.
249
250If this is set to non-zero, this information is shown whenever the
251OOM killer actually kills a memory-hogging task.
252
253The default value is 0.
254
255=============================================================
256
David Rientjesfe071d72007-10-16 23:25:56 -0700257oom_kill_allocating_task
258
259This enables or disables killing the OOM-triggering task in
260out-of-memory situations.
261
262If this is set to zero, the OOM killer will scan through the entire
263tasklist and select a task based on heuristics to kill. This normally
264selects a rogue memory-hogging task that frees up a large amount of
265memory when killed.
266
267If this is set to non-zero, the OOM killer simply kills the task that
268triggered the out-of-memory condition. This avoids the expensive
269tasklist scan.
270
271If panic_on_oom is selected, it takes precedence over whatever value
272is used in oom_kill_allocating_task.
273
274The default value is 0.
275
Eric Parised032182007-06-28 15:55:21 -0400276==============================================================
277
278mmap_min_addr
279
280This file indicates the amount of address space which a user process will
281be restricted from mmaping. Since kernel null dereference bugs could
282accidentally operate based on the information in the first couple of pages
283of memory userspace processes should not be allowed to write to them. By
284default this value is set to 0 and no protections will be enforced by the
285security module. Setting this value to something like 64k will allow the
286vast majority of applications to work correctly and provide defense in depth
287against future potential kernel bugs.
288
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700289==============================================================
290
291numa_zonelist_order
292
293This sysctl is only for NUMA.
294'where the memory is allocated from' is controlled by zonelists.
295(This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation.
296 you may be able to read ZONE_DMA as ZONE_DMA32...)
297
298In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following.
299ZONE_NORMAL -> ZONE_DMA
300This means that a memory allocation request for GFP_KERNEL will
301get memory from ZONE_DMA only when ZONE_NORMAL is not available.
302
303In NUMA case, you can think of following 2 types of order.
304Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL
305
306(A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL
307(B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA.
308
309Type(A) offers the best locality for processes on Node(0), but ZONE_DMA
310will be used before ZONE_NORMAL exhaustion. This increases possibility of
311out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small.
312
313Type(B) cannot offer the best locality but is more robust against OOM of
314the DMA zone.
315
316Type(A) is called as "Node" order. Type (B) is "Zone" order.
317
318"Node order" orders the zonelists by node, then by zone within each node.
319Specify "[Nn]ode" for zone order
320
321"Zone Order" orders the zonelists by zone type, then by node within each
322zone. Specify "[Zz]one"for zode order.
323
324Specify "[Dd]efault" to request automatic configuration. Autoconfiguration
325will select "node" order in following case.
326(1) if the DMA zone does not exist or
327(2) if the DMA zone comprises greater than 50% of the available memory or
328(3) if any node's DMA zone comprises greater than 60% of its local memory and
329 the amount of local memory is big enough.
330
331Otherwise, "zone" order will be selected. Default order is recommended unless
332this is causing problems for your system/application.
Nishanth Aravamudand5dbac82007-12-17 16:20:25 -0800333
334==============================================================
335
336nr_hugepages
337
338Change the minimum size of the hugepage pool.
339
340See Documentation/vm/hugetlbpage.txt
341
342==============================================================
343
344nr_overcommit_hugepages
345
346Change the maximum size of the hugepage pool. The maximum is
347nr_hugepages + nr_overcommit_hugepages.
348
349See Documentation/vm/hugetlbpage.txt