Based on kernel version 2.6.26. Page generated on 2008-07-16 21:13 EST.
1 Documentation for /proc/sys/vm/* kernel version 2.2.10 2 (c) 1998, 1999, Rik van Riel <riel[AT]nl.linux[DOT]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 25 - highmem_is_dirtyable (only if CONFIG_HIGHMEM set) 26 - max_map_count 27 - min_free_kbytes 28 - laptop_mode 29 - block_dump 30 - drop-caches 31 - zone_reclaim_mode 32 - min_unmapped_ratio 33 - min_slab_ratio 34 - panic_on_oom 35 - oom_dump_tasks 36 - oom_kill_allocating_task 37 - mmap_min_address 38 - numa_zonelist_order 39 - nr_hugepages 40 - nr_overcommit_hugepages 41 42 ============================================================== 43 44 dirty_ratio, dirty_background_ratio, dirty_expire_centisecs, 45 dirty_writeback_centisecs, highmem_is_dirtyable, 46 vfs_cache_pressure, laptop_mode, block_dump, swap_token_timeout, 47 drop-caches, hugepages_treat_as_movable: 48 49 See Documentation/filesystems/proc.txt 50 51 ============================================================== 52 53 overcommit_memory: 54 55 This value contains a flag that enables memory overcommitment. 56 57 When this flag is 0, the kernel attempts to estimate the amount 58 of free memory left when userspace requests more memory. 59 60 When this flag is 1, the kernel pretends there is always enough 61 memory until it actually runs out. 62 63 When this flag is 2, the kernel uses a "never overcommit" 64 policy that attempts to prevent any overcommit of memory. 65 66 This feature can be very useful because there are a lot of 67 programs that malloc() huge amounts of memory "just-in-case" 68 and don't use much of it. 69 70 The default value is 0. 71 72 See Documentation/vm/overcommit-accounting and 73 security/commoncap.c::cap_vm_enough_memory() for more information. 74 75 ============================================================== 76 77 overcommit_ratio: 78 79 When overcommit_memory is set to 2, the committed address 80 space is not permitted to exceed swap plus this percentage 81 of physical RAM. See above. 82 83 ============================================================== 84 85 page-cluster: 86 87 The Linux VM subsystem avoids excessive disk seeks by reading 88 multiple pages on a page fault. The number of pages it reads 89 is dependent on the amount of memory in your machine. 90 91 The number of pages the kernel reads in at once is equal to 92 2 ^ page-cluster. Values above 2 ^ 5 don't make much sense 93 for swap because we only cluster swap data in 32-page groups. 94 95 ============================================================== 96 97 max_map_count: 98 99 This file contains the maximum number of memory map areas a process 100 may have. Memory map areas are used as a side-effect of calling 101 malloc, directly by mmap and mprotect, and also when loading shared 102 libraries. 103 104 While most applications need less than a thousand maps, certain 105 programs, particularly malloc debuggers, may consume lots of them, 106 e.g., up to one or two maps per allocation. 107 108 The default value is 65536. 109 110 ============================================================== 111 112 min_free_kbytes: 113 114 This is used to force the Linux VM to keep a minimum number 115 of kilobytes free. The VM uses this number to compute a pages_min 116 value for each lowmem zone in the system. Each lowmem zone gets 117 a number of reserved free pages based proportionally on its size. 118 119 Some minimal ammount of memory is needed to satisfy PF_MEMALLOC 120 allocations; if you set this to lower than 1024KB, your system will 121 become subtly broken, and prone to deadlock under high loads. 122 123 Setting this too high will OOM your machine instantly. 124 125 ============================================================== 126 127 percpu_pagelist_fraction 128 129 This is the fraction of pages at most (high mark pcp->high) in each zone that 130 are allocated for each per cpu page list. The min value for this is 8. It 131 means that we don't allow more than 1/8th of pages in each zone to be 132 allocated in any single per_cpu_pagelist. This entry only changes the value 133 of hot per cpu pagelists. User can specify a number like 100 to allocate 134 1/100th of each zone to each per cpu page list. 135 136 The batch value of each per cpu pagelist is also updated as a result. It is 137 set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8) 138 139 The initial value is zero. Kernel does not use this value at boot time to set 140 the high water marks for each per cpu page list. 141 142 =============================================================== 143 144 zone_reclaim_mode: 145 146 Zone_reclaim_mode allows someone to set more or less aggressive approaches to 147 reclaim memory when a zone runs out of memory. If it is set to zero then no 148 zone reclaim occurs. Allocations will be satisfied from other zones / nodes 149 in the system. 150 151 This is value ORed together of 152 153 1 = Zone reclaim on 154 2 = Zone reclaim writes dirty pages out 155 4 = Zone reclaim swaps pages 156 157 zone_reclaim_mode is set during bootup to 1 if it is determined that pages 158 from remote zones will cause a measurable performance reduction. The 159 page allocator will then reclaim easily reusable pages (those page 160 cache pages that are currently not used) before allocating off node pages. 161 162 It may be beneficial to switch off zone reclaim if the system is 163 used for a file server and all of memory should be used for caching files 164 from disk. In that case the caching effect is more important than 165 data locality. 166 167 Allowing zone reclaim to write out pages stops processes that are 168 writing large amounts of data from dirtying pages on other nodes. Zone 169 reclaim will write out dirty pages if a zone fills up and so effectively 170 throttle the process. This may decrease the performance of a single process 171 since it cannot use all of system memory to buffer the outgoing writes 172 anymore but it preserve the memory on other nodes so that the performance 173 of other processes running on other nodes will not be affected. 174 175 Allowing regular swap effectively restricts allocations to the local 176 node unless explicitly overridden by memory policies or cpuset 177 configurations. 178 179 ============================================================= 180 181 min_unmapped_ratio: 182 183 This is available only on NUMA kernels. 184 185 A percentage of the total pages in each zone. Zone reclaim will only 186 occur if more than this percentage of pages are file backed and unmapped. 187 This is to insure that a minimal amount of local pages is still available for 188 file I/O even if the node is overallocated. 189 190 The default is 1 percent. 191 192 ============================================================= 193 194 min_slab_ratio: 195 196 This is available only on NUMA kernels. 197 198 A percentage of the total pages in each zone. On Zone reclaim 199 (fallback from the local zone occurs) slabs will be reclaimed if more 200 than this percentage of pages in a zone are reclaimable slab pages. 201 This insures that the slab growth stays under control even in NUMA 202 systems that rarely perform global reclaim. 203 204 The default is 5 percent. 205 206 Note that slab reclaim is triggered in a per zone / node fashion. 207 The process of reclaiming slab memory is currently not node specific 208 and may not be fast. 209 210 ============================================================= 211 212 panic_on_oom 213 214 This enables or disables panic on out-of-memory feature. 215 216 If this is set to 0, the kernel will kill some rogue process, 217 called oom_killer. Usually, oom_killer can kill rogue processes and 218 system will survive. 219 220 If this is set to 1, the kernel panics when out-of-memory happens. 221 However, if a process limits using nodes by mempolicy/cpusets, 222 and those nodes become memory exhaustion status, one process 223 may be killed by oom-killer. No panic occurs in this case. 224 Because other nodes' memory may be free. This means system total status 225 may be not fatal yet. 226 227 If this is set to 2, the kernel panics compulsorily even on the 228 above-mentioned. 229 230 The default value is 0. 231 1 and 2 are for failover of clustering. Please select either 232 according to your policy of failover. 233 234 ============================================================= 235 236 oom_dump_tasks 237 238 Enables a system-wide task dump (excluding kernel threads) to be 239 produced when the kernel performs an OOM-killing and includes such 240 information as pid, uid, tgid, vm size, rss, cpu, oom_adj score, and 241 name. This is helpful to determine why the OOM killer was invoked 242 and to identify the rogue task that caused it. 243 244 If this is set to zero, this information is suppressed. On very 245 large systems with thousands of tasks it may not be feasible to dump 246 the memory state information for each one. Such systems should not 247 be forced to incur a performance penalty in OOM conditions when the 248 information may not be desired. 249 250 If this is set to non-zero, this information is shown whenever the 251 OOM killer actually kills a memory-hogging task. 252 253 The default value is 0. 254 255 ============================================================= 256 257 oom_kill_allocating_task 258 259 This enables or disables killing the OOM-triggering task in 260 out-of-memory situations. 261 262 If this is set to zero, the OOM killer will scan through the entire 263 tasklist and select a task based on heuristics to kill. This normally 264 selects a rogue memory-hogging task that frees up a large amount of 265 memory when killed. 266 267 If this is set to non-zero, the OOM killer simply kills the task that 268 triggered the out-of-memory condition. This avoids the expensive 269 tasklist scan. 270 271 If panic_on_oom is selected, it takes precedence over whatever value 272 is used in oom_kill_allocating_task. 273 274 The default value is 0. 275 276 ============================================================== 277 278 mmap_min_addr 279 280 This file indicates the amount of address space which a user process will 281 be restricted from mmaping. Since kernel null dereference bugs could 282 accidentally operate based on the information in the first couple of pages 283 of memory userspace processes should not be allowed to write to them. By 284 default this value is set to 0 and no protections will be enforced by the 285 security module. Setting this value to something like 64k will allow the 286 vast majority of applications to work correctly and provide defense in depth 287 against future potential kernel bugs. 288 289 ============================================================== 290 291 numa_zonelist_order 292 293 This 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...) 297 298 In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following. 299 ZONE_NORMAL -> ZONE_DMA 300 This means that a memory allocation request for GFP_KERNEL will 301 get memory from ZONE_DMA only when ZONE_NORMAL is not available. 302 303 In NUMA case, you can think of following 2 types of order. 304 Assume 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. 308 309 Type(A) offers the best locality for processes on Node(0), but ZONE_DMA 310 will be used before ZONE_NORMAL exhaustion. This increases possibility of 311 out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small. 312 313 Type(B) cannot offer the best locality but is more robust against OOM of 314 the DMA zone. 315 316 Type(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. 319 Specify "[Nn]ode" for zone order 320 321 "Zone Order" orders the zonelists by zone type, then by node within each 322 zone. Specify "[Zz]one"for zode order. 323 324 Specify "[Dd]efault" to request automatic configuration. Autoconfiguration 325 will 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. 330 331 Otherwise, "zone" order will be selected. Default order is recommended unless 332 this is causing problems for your system/application. 333 334 ============================================================== 335 336 nr_hugepages 337 338 Change the minimum size of the hugepage pool. 339 340 See Documentation/vm/hugetlbpage.txt 341 342 ============================================================== 343 344 nr_overcommit_hugepages 345 346 Change the maximum size of the hugepage pool. The maximum is 347 nr_hugepages + nr_overcommit_hugepages. 348 349 See Documentation/vm/hugetlbpage.txt