/Documentation/sysctl/vm.txt

https://bitbucket.org/sammyz/iscream_thunderc-2.6.35-rebase · Plain Text · 697 lines · 491 code · 206 blank · 0 comment · 0 complexity · 8f75aabd1621275888779f9b181367f8 MD5 · raw file

  1. Documentation for /proc/sys/vm/* kernel version 2.6.29
  2. (c) 1998, 1999, Rik van Riel <riel@nl.linux.org>
  3. (c) 2008 Peter W. Morreale <pmorreale@novell.com>
  4. For general info and legal blurb, please look in README.
  5. ==============================================================
  6. This file contains the documentation for the sysctl files in
  7. /proc/sys/vm and is valid for Linux kernel version 2.6.29.
  8. The files in this directory can be used to tune the operation
  9. of the virtual memory (VM) subsystem of the Linux kernel and
  10. the writeout of dirty data to disk.
  11. Default values and initialization routines for most of these
  12. files can be found in mm/swap.c.
  13. Currently, these files are in /proc/sys/vm:
  14. - block_dump
  15. - compact_memory
  16. - dirty_background_bytes
  17. - dirty_background_ratio
  18. - dirty_bytes
  19. - dirty_expire_centisecs
  20. - dirty_ratio
  21. - dirty_writeback_centisecs
  22. - drop_caches
  23. - extfrag_threshold
  24. - hugepages_treat_as_movable
  25. - hugetlb_shm_group
  26. - laptop_mode
  27. - legacy_va_layout
  28. - lowmem_reserve_ratio
  29. - max_map_count
  30. - memory_failure_early_kill
  31. - memory_failure_recovery
  32. - min_free_kbytes
  33. - min_slab_ratio
  34. - min_unmapped_ratio
  35. - mmap_min_addr
  36. - nr_hugepages
  37. - nr_overcommit_hugepages
  38. - nr_pdflush_threads
  39. - nr_trim_pages (only if CONFIG_MMU=n)
  40. - numa_zonelist_order
  41. - oom_dump_tasks
  42. - oom_kill_allocating_task
  43. - overcommit_memory
  44. - overcommit_ratio
  45. - page-cluster
  46. - panic_on_oom
  47. - percpu_pagelist_fraction
  48. - stat_interval
  49. - swappiness
  50. - vfs_cache_pressure
  51. - zone_reclaim_mode
  52. ==============================================================
  53. block_dump
  54. block_dump enables block I/O debugging when set to a nonzero value. More
  55. information on block I/O debugging is in Documentation/laptops/laptop-mode.txt.
  56. ==============================================================
  57. compact_memory
  58. Available only when CONFIG_COMPACTION is set. When 1 is written to the file,
  59. all zones are compacted such that free memory is available in contiguous
  60. blocks where possible. This can be important for example in the allocation of
  61. huge pages although processes will also directly compact memory as required.
  62. ==============================================================
  63. dirty_background_bytes
  64. Contains the amount of dirty memory at which the pdflush background writeback
  65. daemon will start writeback.
  66. If dirty_background_bytes is written, dirty_background_ratio becomes a function
  67. of its value (dirty_background_bytes / the amount of dirtyable system memory).
  68. ==============================================================
  69. dirty_background_ratio
  70. Contains, as a percentage of total system memory, the number of pages at which
  71. the pdflush background writeback daemon will start writing out dirty data.
  72. ==============================================================
  73. dirty_bytes
  74. Contains the amount of dirty memory at which a process generating disk writes
  75. will itself start writeback.
  76. If dirty_bytes is written, dirty_ratio becomes a function of its value
  77. (dirty_bytes / the amount of dirtyable system memory).
  78. Note: the minimum value allowed for dirty_bytes is two pages (in bytes); any
  79. value lower than this limit will be ignored and the old configuration will be
  80. retained.
  81. ==============================================================
  82. dirty_expire_centisecs
  83. This tunable is used to define when dirty data is old enough to be eligible
  84. for writeout by the pdflush daemons. It is expressed in 100'ths of a second.
  85. Data which has been dirty in-memory for longer than this interval will be
  86. written out next time a pdflush daemon wakes up.
  87. ==============================================================
  88. dirty_ratio
  89. Contains, as a percentage of total system memory, the number of pages at which
  90. a process which is generating disk writes will itself start writing out dirty
  91. data.
  92. ==============================================================
  93. dirty_writeback_centisecs
  94. The pdflush writeback daemons will periodically wake up and write `old' data
  95. out to disk. This tunable expresses the interval between those wakeups, in
  96. 100'ths of a second.
  97. Setting this to zero disables periodic writeback altogether.
  98. ==============================================================
  99. drop_caches
  100. Writing to this will cause the kernel to drop clean caches, dentries and
  101. inodes from memory, causing that memory to become free.
  102. To free pagecache:
  103. echo 1 > /proc/sys/vm/drop_caches
  104. To free dentries and inodes:
  105. echo 2 > /proc/sys/vm/drop_caches
  106. To free pagecache, dentries and inodes:
  107. echo 3 > /proc/sys/vm/drop_caches
  108. As this is a non-destructive operation and dirty objects are not freeable, the
  109. user should run `sync' first.
  110. ==============================================================
  111. extfrag_threshold
  112. This parameter affects whether the kernel will compact memory or direct
  113. reclaim to satisfy a high-order allocation. /proc/extfrag_index shows what
  114. the fragmentation index for each order is in each zone in the system. Values
  115. tending towards 0 imply allocations would fail due to lack of memory,
  116. values towards 1000 imply failures are due to fragmentation and -1 implies
  117. that the allocation will succeed as long as watermarks are met.
  118. The kernel will not compact memory in a zone if the
  119. fragmentation index is <= extfrag_threshold. The default value is 500.
  120. ==============================================================
  121. hugepages_treat_as_movable
  122. This parameter is only useful when kernelcore= is specified at boot time to
  123. create ZONE_MOVABLE for pages that may be reclaimed or migrated. Huge pages
  124. are not movable so are not normally allocated from ZONE_MOVABLE. A non-zero
  125. value written to hugepages_treat_as_movable allows huge pages to be allocated
  126. from ZONE_MOVABLE.
  127. Once enabled, the ZONE_MOVABLE is treated as an area of memory the huge
  128. pages pool can easily grow or shrink within. Assuming that applications are
  129. not running that mlock() a lot of memory, it is likely the huge pages pool
  130. can grow to the size of ZONE_MOVABLE by repeatedly entering the desired value
  131. into nr_hugepages and triggering page reclaim.
  132. ==============================================================
  133. hugetlb_shm_group
  134. hugetlb_shm_group contains group id that is allowed to create SysV
  135. shared memory segment using hugetlb page.
  136. ==============================================================
  137. laptop_mode
  138. laptop_mode is a knob that controls "laptop mode". All the things that are
  139. controlled by this knob are discussed in Documentation/laptops/laptop-mode.txt.
  140. ==============================================================
  141. legacy_va_layout
  142. If non-zero, this sysctl disables the new 32-bit mmap mmap layout - the kernel
  143. will use the legacy (2.4) layout for all processes.
  144. ==============================================================
  145. lowmem_reserve_ratio
  146. For some specialised workloads on highmem machines it is dangerous for
  147. the kernel to allow process memory to be allocated from the "lowmem"
  148. zone. This is because that memory could then be pinned via the mlock()
  149. system call, or by unavailability of swapspace.
  150. And on large highmem machines this lack of reclaimable lowmem memory
  151. can be fatal.
  152. So the Linux page allocator has a mechanism which prevents allocations
  153. which _could_ use highmem from using too much lowmem. This means that
  154. a certain amount of lowmem is defended from the possibility of being
  155. captured into pinned user memory.
  156. (The same argument applies to the old 16 megabyte ISA DMA region. This
  157. mechanism will also defend that region from allocations which could use
  158. highmem or lowmem).
  159. The `lowmem_reserve_ratio' tunable determines how aggressive the kernel is
  160. in defending these lower zones.
  161. If you have a machine which uses highmem or ISA DMA and your
  162. applications are using mlock(), or if you are running with no swap then
  163. you probably should change the lowmem_reserve_ratio setting.
  164. The lowmem_reserve_ratio is an array. You can see them by reading this file.
  165. -
  166. % cat /proc/sys/vm/lowmem_reserve_ratio
  167. 256 256 32
  168. -
  169. Note: # of this elements is one fewer than number of zones. Because the highest
  170. zone's value is not necessary for following calculation.
  171. But, these values are not used directly. The kernel calculates # of protection
  172. pages for each zones from them. These are shown as array of protection pages
  173. in /proc/zoneinfo like followings. (This is an example of x86-64 box).
  174. Each zone has an array of protection pages like this.
  175. -
  176. Node 0, zone DMA
  177. pages free 1355
  178. min 3
  179. low 3
  180. high 4
  181. :
  182. :
  183. numa_other 0
  184. protection: (0, 2004, 2004, 2004)
  185. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  186. pagesets
  187. cpu: 0 pcp: 0
  188. :
  189. -
  190. These protections are added to score to judge whether this zone should be used
  191. for page allocation or should be reclaimed.
  192. In this example, if normal pages (index=2) are required to this DMA zone and
  193. watermark[WMARK_HIGH] is used for watermark, the kernel judges this zone should
  194. not be used because pages_free(1355) is smaller than watermark + protection[2]
  195. (4 + 2004 = 2008). If this protection value is 0, this zone would be used for
  196. normal page requirement. If requirement is DMA zone(index=0), protection[0]
  197. (=0) is used.
  198. zone[i]'s protection[j] is calculated by following expression.
  199. (i < j):
  200. zone[i]->protection[j]
  201. = (total sums of present_pages from zone[i+1] to zone[j] on the node)
  202. / lowmem_reserve_ratio[i];
  203. (i = j):
  204. (should not be protected. = 0;
  205. (i > j):
  206. (not necessary, but looks 0)
  207. The default values of lowmem_reserve_ratio[i] are
  208. 256 (if zone[i] means DMA or DMA32 zone)
  209. 32 (others).
  210. As above expression, they are reciprocal number of ratio.
  211. 256 means 1/256. # of protection pages becomes about "0.39%" of total present
  212. pages of higher zones on the node.
  213. If you would like to protect more pages, smaller values are effective.
  214. The minimum value is 1 (1/1 -> 100%).
  215. ==============================================================
  216. max_map_count:
  217. This file contains the maximum number of memory map areas a process
  218. may have. Memory map areas are used as a side-effect of calling
  219. malloc, directly by mmap and mprotect, and also when loading shared
  220. libraries.
  221. While most applications need less than a thousand maps, certain
  222. programs, particularly malloc debuggers, may consume lots of them,
  223. e.g., up to one or two maps per allocation.
  224. The default value is 65536.
  225. =============================================================
  226. memory_failure_early_kill:
  227. Control how to kill processes when uncorrected memory error (typically
  228. a 2bit error in a memory module) is detected in the background by hardware
  229. that cannot be handled by the kernel. In some cases (like the page
  230. still having a valid copy on disk) the kernel will handle the failure
  231. transparently without affecting any applications. But if there is
  232. no other uptodate copy of the data it will kill to prevent any data
  233. corruptions from propagating.
  234. 1: Kill all processes that have the corrupted and not reloadable page mapped
  235. as soon as the corruption is detected. Note this is not supported
  236. for a few types of pages, like kernel internally allocated data or
  237. the swap cache, but works for the majority of user pages.
  238. 0: Only unmap the corrupted page from all processes and only kill a process
  239. who tries to access it.
  240. The kill is done using a catchable SIGBUS with BUS_MCEERR_AO, so processes can
  241. handle this if they want to.
  242. This is only active on architectures/platforms with advanced machine
  243. check handling and depends on the hardware capabilities.
  244. Applications can override this setting individually with the PR_MCE_KILL prctl
  245. ==============================================================
  246. memory_failure_recovery
  247. Enable memory failure recovery (when supported by the platform)
  248. 1: Attempt recovery.
  249. 0: Always panic on a memory failure.
  250. ==============================================================
  251. min_free_kbytes:
  252. This is used to force the Linux VM to keep a minimum number
  253. of kilobytes free. The VM uses this number to compute a
  254. watermark[WMARK_MIN] value for each lowmem zone in the system.
  255. Each lowmem zone gets a number of reserved free pages based
  256. proportionally on its size.
  257. Some minimal amount of memory is needed to satisfy PF_MEMALLOC
  258. allocations; if you set this to lower than 1024KB, your system will
  259. become subtly broken, and prone to deadlock under high loads.
  260. Setting this too high will OOM your machine instantly.
  261. =============================================================
  262. min_slab_ratio:
  263. This is available only on NUMA kernels.
  264. A percentage of the total pages in each zone. On Zone reclaim
  265. (fallback from the local zone occurs) slabs will be reclaimed if more
  266. than this percentage of pages in a zone are reclaimable slab pages.
  267. This insures that the slab growth stays under control even in NUMA
  268. systems that rarely perform global reclaim.
  269. The default is 5 percent.
  270. Note that slab reclaim is triggered in a per zone / node fashion.
  271. The process of reclaiming slab memory is currently not node specific
  272. and may not be fast.
  273. =============================================================
  274. min_unmapped_ratio:
  275. This is available only on NUMA kernels.
  276. This is a percentage of the total pages in each zone. Zone reclaim will
  277. only occur if more than this percentage of pages are in a state that
  278. zone_reclaim_mode allows to be reclaimed.
  279. If zone_reclaim_mode has the value 4 OR'd, then the percentage is compared
  280. against all file-backed unmapped pages including swapcache pages and tmpfs
  281. files. Otherwise, only unmapped pages backed by normal files but not tmpfs
  282. files and similar are considered.
  283. The default is 1 percent.
  284. ==============================================================
  285. mmap_min_addr
  286. This file indicates the amount of address space which a user process will
  287. be restricted from mmapping. Since kernel null dereference bugs could
  288. accidentally operate based on the information in the first couple of pages
  289. of memory userspace processes should not be allowed to write to them. By
  290. default this value is set to 0 and no protections will be enforced by the
  291. security module. Setting this value to something like 64k will allow the
  292. vast majority of applications to work correctly and provide defense in depth
  293. against future potential kernel bugs.
  294. ==============================================================
  295. nr_hugepages
  296. Change the minimum size of the hugepage pool.
  297. See Documentation/vm/hugetlbpage.txt
  298. ==============================================================
  299. nr_overcommit_hugepages
  300. Change the maximum size of the hugepage pool. The maximum is
  301. nr_hugepages + nr_overcommit_hugepages.
  302. See Documentation/vm/hugetlbpage.txt
  303. ==============================================================
  304. nr_pdflush_threads
  305. The current number of pdflush threads. This value is read-only.
  306. The value changes according to the number of dirty pages in the system.
  307. When necessary, additional pdflush threads are created, one per second, up to
  308. nr_pdflush_threads_max.
  309. ==============================================================
  310. nr_trim_pages
  311. This is available only on NOMMU kernels.
  312. This value adjusts the excess page trimming behaviour of power-of-2 aligned
  313. NOMMU mmap allocations.
  314. A value of 0 disables trimming of allocations entirely, while a value of 1
  315. trims excess pages aggressively. Any value >= 1 acts as the watermark where
  316. trimming of allocations is initiated.
  317. The default value is 1.
  318. See Documentation/nommu-mmap.txt for more information.
  319. ==============================================================
  320. numa_zonelist_order
  321. This sysctl is only for NUMA.
  322. 'where the memory is allocated from' is controlled by zonelists.
  323. (This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation.
  324. you may be able to read ZONE_DMA as ZONE_DMA32...)
  325. In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following.
  326. ZONE_NORMAL -> ZONE_DMA
  327. This means that a memory allocation request for GFP_KERNEL will
  328. get memory from ZONE_DMA only when ZONE_NORMAL is not available.
  329. In NUMA case, you can think of following 2 types of order.
  330. Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL
  331. (A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL
  332. (B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA.
  333. Type(A) offers the best locality for processes on Node(0), but ZONE_DMA
  334. will be used before ZONE_NORMAL exhaustion. This increases possibility of
  335. out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small.
  336. Type(B) cannot offer the best locality but is more robust against OOM of
  337. the DMA zone.
  338. Type(A) is called as "Node" order. Type (B) is "Zone" order.
  339. "Node order" orders the zonelists by node, then by zone within each node.
  340. Specify "[Nn]ode" for zone order
  341. "Zone Order" orders the zonelists by zone type, then by node within each
  342. zone. Specify "[Zz]one"for zode order.
  343. Specify "[Dd]efault" to request automatic configuration. Autoconfiguration
  344. will select "node" order in following case.
  345. (1) if the DMA zone does not exist or
  346. (2) if the DMA zone comprises greater than 50% of the available memory or
  347. (3) if any node's DMA zone comprises greater than 60% of its local memory and
  348. the amount of local memory is big enough.
  349. Otherwise, "zone" order will be selected. Default order is recommended unless
  350. this is causing problems for your system/application.
  351. ==============================================================
  352. oom_dump_tasks
  353. Enables a system-wide task dump (excluding kernel threads) to be
  354. produced when the kernel performs an OOM-killing and includes such
  355. information as pid, uid, tgid, vm size, rss, cpu, oom_adj score, and
  356. name. This is helpful to determine why the OOM killer was invoked
  357. and to identify the rogue task that caused it.
  358. If this is set to zero, this information is suppressed. On very
  359. large systems with thousands of tasks it may not be feasible to dump
  360. the memory state information for each one. Such systems should not
  361. be forced to incur a performance penalty in OOM conditions when the
  362. information may not be desired.
  363. If this is set to non-zero, this information is shown whenever the
  364. OOM killer actually kills a memory-hogging task.
  365. The default value is 0.
  366. ==============================================================
  367. oom_kill_allocating_task
  368. This enables or disables killing the OOM-triggering task in
  369. out-of-memory situations.
  370. If this is set to zero, the OOM killer will scan through the entire
  371. tasklist and select a task based on heuristics to kill. This normally
  372. selects a rogue memory-hogging task that frees up a large amount of
  373. memory when killed.
  374. If this is set to non-zero, the OOM killer simply kills the task that
  375. triggered the out-of-memory condition. This avoids the expensive
  376. tasklist scan.
  377. If panic_on_oom is selected, it takes precedence over whatever value
  378. is used in oom_kill_allocating_task.
  379. The default value is 0.
  380. ==============================================================
  381. overcommit_memory:
  382. This value contains a flag that enables memory overcommitment.
  383. When this flag is 0, the kernel attempts to estimate the amount
  384. of free memory left when userspace requests more memory.
  385. When this flag is 1, the kernel pretends there is always enough
  386. memory until it actually runs out.
  387. When this flag is 2, the kernel uses a "never overcommit"
  388. policy that attempts to prevent any overcommit of memory.
  389. This feature can be very useful because there are a lot of
  390. programs that malloc() huge amounts of memory "just-in-case"
  391. and don't use much of it.
  392. The default value is 0.
  393. See Documentation/vm/overcommit-accounting and
  394. security/commoncap.c::cap_vm_enough_memory() for more information.
  395. ==============================================================
  396. overcommit_ratio:
  397. When overcommit_memory is set to 2, the committed address
  398. space is not permitted to exceed swap plus this percentage
  399. of physical RAM. See above.
  400. ==============================================================
  401. page-cluster
  402. page-cluster controls the number of pages which are written to swap in
  403. a single attempt. The swap I/O size.
  404. It is a logarithmic value - setting it to zero means "1 page", setting
  405. it to 1 means "2 pages", setting it to 2 means "4 pages", etc.
  406. The default value is three (eight pages at a time). There may be some
  407. small benefits in tuning this to a different value if your workload is
  408. swap-intensive.
  409. =============================================================
  410. panic_on_oom
  411. This enables or disables panic on out-of-memory feature.
  412. If this is set to 0, the kernel will kill some rogue process,
  413. called oom_killer. Usually, oom_killer can kill rogue processes and
  414. system will survive.
  415. If this is set to 1, the kernel panics when out-of-memory happens.
  416. However, if a process limits using nodes by mempolicy/cpusets,
  417. and those nodes become memory exhaustion status, one process
  418. may be killed by oom-killer. No panic occurs in this case.
  419. Because other nodes' memory may be free. This means system total status
  420. may be not fatal yet.
  421. If this is set to 2, the kernel panics compulsorily even on the
  422. above-mentioned. Even oom happens under memory cgroup, the whole
  423. system panics.
  424. The default value is 0.
  425. 1 and 2 are for failover of clustering. Please select either
  426. according to your policy of failover.
  427. panic_on_oom=2+kdump gives you very strong tool to investigate
  428. why oom happens. You can get snapshot.
  429. =============================================================
  430. percpu_pagelist_fraction
  431. This is the fraction of pages at most (high mark pcp->high) in each zone that
  432. are allocated for each per cpu page list. The min value for this is 8. It
  433. means that we don't allow more than 1/8th of pages in each zone to be
  434. allocated in any single per_cpu_pagelist. This entry only changes the value
  435. of hot per cpu pagelists. User can specify a number like 100 to allocate
  436. 1/100th of each zone to each per cpu page list.
  437. The batch value of each per cpu pagelist is also updated as a result. It is
  438. set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8)
  439. The initial value is zero. Kernel does not use this value at boot time to set
  440. the high water marks for each per cpu page list.
  441. ==============================================================
  442. stat_interval
  443. The time interval between which vm statistics are updated. The default
  444. is 1 second.
  445. ==============================================================
  446. swappiness
  447. This control is used to define how aggressive the kernel will swap
  448. memory pages. Higher values will increase agressiveness, lower values
  449. decrease the amount of swap.
  450. The default value is 60.
  451. ==============================================================
  452. vfs_cache_pressure
  453. ------------------
  454. Controls the tendency of the kernel to reclaim the memory which is used for
  455. caching of directory and inode objects.
  456. At the default value of vfs_cache_pressure=100 the kernel will attempt to
  457. reclaim dentries and inodes at a "fair" rate with respect to pagecache and
  458. swapcache reclaim. Decreasing vfs_cache_pressure causes the kernel to prefer
  459. to retain dentry and inode caches. When vfs_cache_pressure=0, the kernel will
  460. never reclaim dentries and inodes due to memory pressure and this can easily
  461. lead to out-of-memory conditions. Increasing vfs_cache_pressure beyond 100
  462. causes the kernel to prefer to reclaim dentries and inodes.
  463. ==============================================================
  464. zone_reclaim_mode:
  465. Zone_reclaim_mode allows someone to set more or less aggressive approaches to
  466. reclaim memory when a zone runs out of memory. If it is set to zero then no
  467. zone reclaim occurs. Allocations will be satisfied from other zones / nodes
  468. in the system.
  469. This is value ORed together of
  470. 1 = Zone reclaim on
  471. 2 = Zone reclaim writes dirty pages out
  472. 4 = Zone reclaim swaps pages
  473. zone_reclaim_mode is set during bootup to 1 if it is determined that pages
  474. from remote zones will cause a measurable performance reduction. The
  475. page allocator will then reclaim easily reusable pages (those page
  476. cache pages that are currently not used) before allocating off node pages.
  477. It may be beneficial to switch off zone reclaim if the system is
  478. used for a file server and all of memory should be used for caching files
  479. from disk. In that case the caching effect is more important than
  480. data locality.
  481. Allowing zone reclaim to write out pages stops processes that are
  482. writing large amounts of data from dirtying pages on other nodes. Zone
  483. reclaim will write out dirty pages if a zone fills up and so effectively
  484. throttle the process. This may decrease the performance of a single process
  485. since it cannot use all of system memory to buffer the outgoing writes
  486. anymore but it preserve the memory on other nodes so that the performance
  487. of other processes running on other nodes will not be affected.
  488. Allowing regular swap effectively restricts allocations to the local
  489. node unless explicitly overridden by memory policies or cpuset
  490. configurations.
  491. ============ End of Document =================================