Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | Documentation for /proc/sys/vm/* kernel version 2.2.10 |
| 2 | (c) 1998, 1999, Rik van Riel <riel@nl.linux.org> |
| 3 | |
| 4 | For general info and legal blurb, please look in README. |
| 5 | |
| 6 | ============================================================== |
| 7 | |
| 8 | This file contains the documentation for the sysctl files in |
| 9 | /proc/sys/vm and is valid for Linux kernel version 2.2. |
| 10 | |
| 11 | The files in this directory can be used to tune the operation |
| 12 | of the virtual memory (VM) subsystem of the Linux kernel and |
| 13 | the writeout of dirty data to disk. |
| 14 | |
| 15 | Default values and initialization routines for most of these |
| 16 | files can be found in mm/swap.c. |
| 17 | |
| 18 | Currently, 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 Gondwana | 195cf45 | 2008-02-04 22:29:20 -0800 | [diff] [blame] | 25 | - highmem_is_dirtyable (only if CONFIG_HIGHMEM set) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 26 | - max_map_count |
| 27 | - min_free_kbytes |
| 28 | - laptop_mode |
| 29 | - block_dump |
Andrew Morton | 9d0243b | 2006-01-08 01:00:39 -0800 | [diff] [blame] | 30 | - drop-caches |
Christoph Lameter | 1743660 | 2006-01-18 17:42:32 -0800 | [diff] [blame] | 31 | - zone_reclaim_mode |
Christoph Lameter | 9614634 | 2006-07-03 00:24:13 -0700 | [diff] [blame] | 32 | - min_unmapped_ratio |
Christoph Lameter | 0ff3849 | 2006-09-25 23:31:52 -0700 | [diff] [blame] | 33 | - min_slab_ratio |
KAMEZAWA Hiroyuki | fadd8fb | 2006-06-23 02:03:13 -0700 | [diff] [blame] | 34 | - panic_on_oom |
David Rientjes | fe071d7 | 2007-10-16 23:25:56 -0700 | [diff] [blame] | 35 | - oom_kill_allocating_task |
Eric Paris | ed03218 | 2007-06-28 15:55:21 -0400 | [diff] [blame] | 36 | - mmap_min_address |
KAMEZAWA Hiroyuki | f0c0b2b | 2007-07-15 23:38:01 -0700 | [diff] [blame] | 37 | - numa_zonelist_order |
Nishanth Aravamudan | d5dbac8 | 2007-12-17 16:20:25 -0800 | [diff] [blame] | 38 | - nr_hugepages |
| 39 | - nr_overcommit_hugepages |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 40 | |
| 41 | ============================================================== |
| 42 | |
| 43 | dirty_ratio, dirty_background_ratio, dirty_expire_centisecs, |
Bron Gondwana | 195cf45 | 2008-02-04 22:29:20 -0800 | [diff] [blame] | 44 | dirty_writeback_centisecs, highmem_is_dirtyable, |
| 45 | vfs_cache_pressure, laptop_mode, block_dump, swap_token_timeout, |
| 46 | drop-caches, hugepages_treat_as_movable: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 47 | |
| 48 | See Documentation/filesystems/proc.txt |
| 49 | |
| 50 | ============================================================== |
| 51 | |
| 52 | overcommit_memory: |
| 53 | |
| 54 | This value contains a flag that enables memory overcommitment. |
| 55 | |
| 56 | When this flag is 0, the kernel attempts to estimate the amount |
| 57 | of free memory left when userspace requests more memory. |
| 58 | |
| 59 | When this flag is 1, the kernel pretends there is always enough |
| 60 | memory until it actually runs out. |
| 61 | |
| 62 | When this flag is 2, the kernel uses a "never overcommit" |
| 63 | policy that attempts to prevent any overcommit of memory. |
| 64 | |
| 65 | This feature can be very useful because there are a lot of |
| 66 | programs that malloc() huge amounts of memory "just-in-case" |
| 67 | and don't use much of it. |
| 68 | |
| 69 | The default value is 0. |
| 70 | |
| 71 | See Documentation/vm/overcommit-accounting and |
| 72 | security/commoncap.c::cap_vm_enough_memory() for more information. |
| 73 | |
| 74 | ============================================================== |
| 75 | |
| 76 | overcommit_ratio: |
| 77 | |
| 78 | When overcommit_memory is set to 2, the committed address |
| 79 | space is not permitted to exceed swap plus this percentage |
| 80 | of physical RAM. See above. |
| 81 | |
| 82 | ============================================================== |
| 83 | |
| 84 | page-cluster: |
| 85 | |
| 86 | The Linux VM subsystem avoids excessive disk seeks by reading |
| 87 | multiple pages on a page fault. The number of pages it reads |
| 88 | is dependent on the amount of memory in your machine. |
| 89 | |
| 90 | The number of pages the kernel reads in at once is equal to |
| 91 | 2 ^ page-cluster. Values above 2 ^ 5 don't make much sense |
| 92 | for swap because we only cluster swap data in 32-page groups. |
| 93 | |
| 94 | ============================================================== |
| 95 | |
| 96 | max_map_count: |
| 97 | |
| 98 | This file contains the maximum number of memory map areas a process |
| 99 | may have. Memory map areas are used as a side-effect of calling |
| 100 | malloc, directly by mmap and mprotect, and also when loading shared |
| 101 | libraries. |
| 102 | |
| 103 | While most applications need less than a thousand maps, certain |
| 104 | programs, particularly malloc debuggers, may consume lots of them, |
| 105 | e.g., up to one or two maps per allocation. |
| 106 | |
| 107 | The default value is 65536. |
| 108 | |
| 109 | ============================================================== |
| 110 | |
| 111 | min_free_kbytes: |
| 112 | |
| 113 | This is used to force the Linux VM to keep a minimum number |
| 114 | of kilobytes free. The VM uses this number to compute a pages_min |
| 115 | value for each lowmem zone in the system. Each lowmem zone gets |
| 116 | a number of reserved free pages based proportionally on its size. |
Rohit Seth | 8ad4b1f | 2006-01-08 01:00:40 -0800 | [diff] [blame] | 117 | |
Pavel Machek | 2495089 | 2007-10-16 23:31:28 -0700 | [diff] [blame] | 118 | Some minimal ammount of memory is needed to satisfy PF_MEMALLOC |
| 119 | allocations; if you set this to lower than 1024KB, your system will |
| 120 | become subtly broken, and prone to deadlock under high loads. |
| 121 | |
| 122 | Setting this too high will OOM your machine instantly. |
| 123 | |
Rohit Seth | 8ad4b1f | 2006-01-08 01:00:40 -0800 | [diff] [blame] | 124 | ============================================================== |
| 125 | |
| 126 | percpu_pagelist_fraction |
| 127 | |
| 128 | This is the fraction of pages at most (high mark pcp->high) in each zone that |
| 129 | are allocated for each per cpu page list. The min value for this is 8. It |
| 130 | means that we don't allow more than 1/8th of pages in each zone to be |
| 131 | allocated in any single per_cpu_pagelist. This entry only changes the value |
| 132 | of hot per cpu pagelists. User can specify a number like 100 to allocate |
| 133 | 1/100th of each zone to each per cpu page list. |
| 134 | |
| 135 | The batch value of each per cpu pagelist is also updated as a result. It is |
| 136 | set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8) |
| 137 | |
| 138 | The initial value is zero. Kernel does not use this value at boot time to set |
| 139 | the high water marks for each per cpu page list. |
Christoph Lameter | 1743660 | 2006-01-18 17:42:32 -0800 | [diff] [blame] | 140 | |
| 141 | =============================================================== |
| 142 | |
| 143 | zone_reclaim_mode: |
| 144 | |
Matt LaPlante | 5d3f083 | 2006-11-30 05:21:10 +0100 | [diff] [blame] | 145 | Zone_reclaim_mode allows someone to set more or less aggressive approaches to |
Christoph Lameter | 1b2ffb7 | 2006-02-01 03:05:34 -0800 | [diff] [blame] | 146 | reclaim memory when a zone runs out of memory. If it is set to zero then no |
| 147 | zone reclaim occurs. Allocations will be satisfied from other zones / nodes |
| 148 | in the system. |
| 149 | |
| 150 | This is value ORed together of |
| 151 | |
| 152 | 1 = Zone reclaim on |
| 153 | 2 = Zone reclaim writes dirty pages out |
| 154 | 4 = Zone reclaim swaps pages |
| 155 | |
| 156 | zone_reclaim_mode is set during bootup to 1 if it is determined that pages |
| 157 | from remote zones will cause a measurable performance reduction. The |
Christoph Lameter | 1743660 | 2006-01-18 17:42:32 -0800 | [diff] [blame] | 158 | page allocator will then reclaim easily reusable pages (those page |
Christoph Lameter | 1b2ffb7 | 2006-02-01 03:05:34 -0800 | [diff] [blame] | 159 | cache pages that are currently not used) before allocating off node pages. |
Christoph Lameter | 1743660 | 2006-01-18 17:42:32 -0800 | [diff] [blame] | 160 | |
Christoph Lameter | 1b2ffb7 | 2006-02-01 03:05:34 -0800 | [diff] [blame] | 161 | It may be beneficial to switch off zone reclaim if the system is |
| 162 | used for a file server and all of memory should be used for caching files |
| 163 | from disk. In that case the caching effect is more important than |
| 164 | data locality. |
Christoph Lameter | 1743660 | 2006-01-18 17:42:32 -0800 | [diff] [blame] | 165 | |
Christoph Lameter | 1b2ffb7 | 2006-02-01 03:05:34 -0800 | [diff] [blame] | 166 | Allowing zone reclaim to write out pages stops processes that are |
| 167 | writing large amounts of data from dirtying pages on other nodes. Zone |
| 168 | reclaim will write out dirty pages if a zone fills up and so effectively |
| 169 | throttle the process. This may decrease the performance of a single process |
| 170 | since it cannot use all of system memory to buffer the outgoing writes |
| 171 | anymore but it preserve the memory on other nodes so that the performance |
| 172 | of other processes running on other nodes will not be affected. |
| 173 | |
| 174 | Allowing regular swap effectively restricts allocations to the local |
| 175 | node unless explicitly overridden by memory policies or cpuset |
| 176 | configurations. |
| 177 | |
KAMEZAWA Hiroyuki | fadd8fb | 2006-06-23 02:03:13 -0700 | [diff] [blame] | 178 | ============================================================= |
| 179 | |
Christoph Lameter | 9614634 | 2006-07-03 00:24:13 -0700 | [diff] [blame] | 180 | min_unmapped_ratio: |
| 181 | |
| 182 | This is available only on NUMA kernels. |
| 183 | |
Christoph Lameter | 0ff3849 | 2006-09-25 23:31:52 -0700 | [diff] [blame] | 184 | A percentage of the total pages in each zone. Zone reclaim will only |
Christoph Lameter | 9614634 | 2006-07-03 00:24:13 -0700 | [diff] [blame] | 185 | occur if more than this percentage of pages are file backed and unmapped. |
| 186 | This is to insure that a minimal amount of local pages is still available for |
| 187 | file I/O even if the node is overallocated. |
| 188 | |
| 189 | The default is 1 percent. |
| 190 | |
| 191 | ============================================================= |
| 192 | |
Christoph Lameter | 0ff3849 | 2006-09-25 23:31:52 -0700 | [diff] [blame] | 193 | min_slab_ratio: |
| 194 | |
| 195 | This is available only on NUMA kernels. |
| 196 | |
| 197 | A percentage of the total pages in each zone. On Zone reclaim |
| 198 | (fallback from the local zone occurs) slabs will be reclaimed if more |
| 199 | than this percentage of pages in a zone are reclaimable slab pages. |
| 200 | This insures that the slab growth stays under control even in NUMA |
| 201 | systems that rarely perform global reclaim. |
| 202 | |
| 203 | The default is 5 percent. |
| 204 | |
| 205 | Note that slab reclaim is triggered in a per zone / node fashion. |
| 206 | The process of reclaiming slab memory is currently not node specific |
| 207 | and may not be fast. |
| 208 | |
| 209 | ============================================================= |
| 210 | |
KAMEZAWA Hiroyuki | fadd8fb | 2006-06-23 02:03:13 -0700 | [diff] [blame] | 211 | panic_on_oom |
| 212 | |
Yasunori Goto | 2b744c0 | 2007-05-06 14:49:59 -0700 | [diff] [blame] | 213 | This enables or disables panic on out-of-memory feature. |
| 214 | |
| 215 | If this is set to 0, the kernel will kill some rogue process, |
| 216 | called oom_killer. Usually, oom_killer can kill rogue processes and |
| 217 | system will survive. |
| 218 | |
| 219 | If this is set to 1, the kernel panics when out-of-memory happens. |
| 220 | However, if a process limits using nodes by mempolicy/cpusets, |
| 221 | and those nodes become memory exhaustion status, one process |
| 222 | may be killed by oom-killer. No panic occurs in this case. |
| 223 | Because other nodes' memory may be free. This means system total status |
| 224 | may be not fatal yet. |
| 225 | |
| 226 | If this is set to 2, the kernel panics compulsorily even on the |
| 227 | above-mentioned. |
KAMEZAWA Hiroyuki | fadd8fb | 2006-06-23 02:03:13 -0700 | [diff] [blame] | 228 | |
| 229 | The default value is 0. |
Yasunori Goto | 2b744c0 | 2007-05-06 14:49:59 -0700 | [diff] [blame] | 230 | 1 and 2 are for failover of clustering. Please select either |
| 231 | according to your policy of failover. |
Eric Paris | ed03218 | 2007-06-28 15:55:21 -0400 | [diff] [blame] | 232 | |
David Rientjes | fe071d7 | 2007-10-16 23:25:56 -0700 | [diff] [blame] | 233 | ============================================================= |
| 234 | |
| 235 | oom_kill_allocating_task |
| 236 | |
| 237 | This enables or disables killing the OOM-triggering task in |
| 238 | out-of-memory situations. |
| 239 | |
| 240 | If this is set to zero, the OOM killer will scan through the entire |
| 241 | tasklist and select a task based on heuristics to kill. This normally |
| 242 | selects a rogue memory-hogging task that frees up a large amount of |
| 243 | memory when killed. |
| 244 | |
| 245 | If this is set to non-zero, the OOM killer simply kills the task that |
| 246 | triggered the out-of-memory condition. This avoids the expensive |
| 247 | tasklist scan. |
| 248 | |
| 249 | If panic_on_oom is selected, it takes precedence over whatever value |
| 250 | is used in oom_kill_allocating_task. |
| 251 | |
| 252 | The default value is 0. |
| 253 | |
Eric Paris | ed03218 | 2007-06-28 15:55:21 -0400 | [diff] [blame] | 254 | ============================================================== |
| 255 | |
| 256 | mmap_min_addr |
| 257 | |
| 258 | This file indicates the amount of address space which a user process will |
| 259 | be restricted from mmaping. Since kernel null dereference bugs could |
| 260 | accidentally operate based on the information in the first couple of pages |
| 261 | of memory userspace processes should not be allowed to write to them. By |
| 262 | default this value is set to 0 and no protections will be enforced by the |
| 263 | security module. Setting this value to something like 64k will allow the |
| 264 | vast majority of applications to work correctly and provide defense in depth |
| 265 | against future potential kernel bugs. |
| 266 | |
KAMEZAWA Hiroyuki | f0c0b2b | 2007-07-15 23:38:01 -0700 | [diff] [blame] | 267 | ============================================================== |
| 268 | |
| 269 | numa_zonelist_order |
| 270 | |
| 271 | This sysctl is only for NUMA. |
| 272 | 'where the memory is allocated from' is controlled by zonelists. |
| 273 | (This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation. |
| 274 | you may be able to read ZONE_DMA as ZONE_DMA32...) |
| 275 | |
| 276 | In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following. |
| 277 | ZONE_NORMAL -> ZONE_DMA |
| 278 | This means that a memory allocation request for GFP_KERNEL will |
| 279 | get memory from ZONE_DMA only when ZONE_NORMAL is not available. |
| 280 | |
| 281 | In NUMA case, you can think of following 2 types of order. |
| 282 | Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL |
| 283 | |
| 284 | (A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL |
| 285 | (B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA. |
| 286 | |
| 287 | Type(A) offers the best locality for processes on Node(0), but ZONE_DMA |
| 288 | will be used before ZONE_NORMAL exhaustion. This increases possibility of |
| 289 | out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small. |
| 290 | |
| 291 | Type(B) cannot offer the best locality but is more robust against OOM of |
| 292 | the DMA zone. |
| 293 | |
| 294 | Type(A) is called as "Node" order. Type (B) is "Zone" order. |
| 295 | |
| 296 | "Node order" orders the zonelists by node, then by zone within each node. |
| 297 | Specify "[Nn]ode" for zone order |
| 298 | |
| 299 | "Zone Order" orders the zonelists by zone type, then by node within each |
| 300 | zone. Specify "[Zz]one"for zode order. |
| 301 | |
| 302 | Specify "[Dd]efault" to request automatic configuration. Autoconfiguration |
| 303 | will select "node" order in following case. |
| 304 | (1) if the DMA zone does not exist or |
| 305 | (2) if the DMA zone comprises greater than 50% of the available memory or |
| 306 | (3) if any node's DMA zone comprises greater than 60% of its local memory and |
| 307 | the amount of local memory is big enough. |
| 308 | |
| 309 | Otherwise, "zone" order will be selected. Default order is recommended unless |
| 310 | this is causing problems for your system/application. |
Nishanth Aravamudan | d5dbac8 | 2007-12-17 16:20:25 -0800 | [diff] [blame] | 311 | |
| 312 | ============================================================== |
| 313 | |
| 314 | nr_hugepages |
| 315 | |
| 316 | Change the minimum size of the hugepage pool. |
| 317 | |
| 318 | See Documentation/vm/hugetlbpage.txt |
| 319 | |
| 320 | ============================================================== |
| 321 | |
| 322 | nr_overcommit_hugepages |
| 323 | |
| 324 | Change the maximum size of the hugepage pool. The maximum is |
| 325 | nr_hugepages + nr_overcommit_hugepages. |
| 326 | |
| 327 | See Documentation/vm/hugetlbpage.txt |