]>
Commit | Line | Data |
---|---|---|
db0fb184 | 1 | Documentation for /proc/sys/vm/* kernel version 2.6.29 |
1da177e4 | 2 | (c) 1998, 1999, Rik van Riel <riel@nl.linux.org> |
db0fb184 | 3 | (c) 2008 Peter W. Morreale <pmorreale@novell.com> |
1da177e4 LT |
4 | |
5 | For general info and legal blurb, please look in README. | |
6 | ||
7 | ============================================================== | |
8 | ||
9 | This file contains the documentation for the sysctl files in | |
db0fb184 | 10 | /proc/sys/vm and is valid for Linux kernel version 2.6.29. |
1da177e4 LT |
11 | |
12 | The files in this directory can be used to tune the operation | |
13 | of the virtual memory (VM) subsystem of the Linux kernel and | |
14 | the writeout of dirty data to disk. | |
15 | ||
16 | Default values and initialization routines for most of these | |
17 | files can be found in mm/swap.c. | |
18 | ||
19 | Currently, these files are in /proc/sys/vm: | |
db0fb184 PM |
20 | |
21 | - block_dump | |
22 | - dirty_background_bytes | |
1da177e4 | 23 | - dirty_background_ratio |
db0fb184 | 24 | - dirty_bytes |
1da177e4 | 25 | - dirty_expire_centisecs |
db0fb184 | 26 | - dirty_ratio |
1da177e4 | 27 | - dirty_writeback_centisecs |
db0fb184 PM |
28 | - drop_caches |
29 | - hugepages_treat_as_movable | |
30 | - hugetlb_shm_group | |
31 | - laptop_mode | |
32 | - legacy_va_layout | |
33 | - lowmem_reserve_ratio | |
1da177e4 | 34 | - max_map_count |
6a46079c AK |
35 | - memory_failure_early_kill |
36 | - memory_failure_recovery | |
1da177e4 | 37 | - min_free_kbytes |
0ff38490 | 38 | - min_slab_ratio |
db0fb184 PM |
39 | - min_unmapped_ratio |
40 | - mmap_min_addr | |
d5dbac87 NA |
41 | - nr_hugepages |
42 | - nr_overcommit_hugepages | |
db0fb184 PM |
43 | - nr_pdflush_threads |
44 | - nr_trim_pages (only if CONFIG_MMU=n) | |
45 | - numa_zonelist_order | |
46 | - oom_dump_tasks | |
47 | - oom_kill_allocating_task | |
48 | - overcommit_memory | |
49 | - overcommit_ratio | |
50 | - page-cluster | |
51 | - panic_on_oom | |
52 | - percpu_pagelist_fraction | |
53 | - stat_interval | |
54 | - swappiness | |
55 | - vfs_cache_pressure | |
56 | - zone_reclaim_mode | |
57 | ||
1da177e4 LT |
58 | ============================================================== |
59 | ||
db0fb184 | 60 | block_dump |
1da177e4 | 61 | |
db0fb184 PM |
62 | block_dump enables block I/O debugging when set to a nonzero value. More |
63 | information on block I/O debugging is in Documentation/laptops/laptop-mode.txt. | |
1da177e4 LT |
64 | |
65 | ============================================================== | |
66 | ||
db0fb184 | 67 | dirty_background_bytes |
1da177e4 | 68 | |
db0fb184 PM |
69 | Contains the amount of dirty memory at which the pdflush background writeback |
70 | daemon will start writeback. | |
1da177e4 | 71 | |
db0fb184 PM |
72 | If dirty_background_bytes is written, dirty_background_ratio becomes a function |
73 | of its value (dirty_background_bytes / the amount of dirtyable system memory). | |
1da177e4 | 74 | |
db0fb184 | 75 | ============================================================== |
1da177e4 | 76 | |
db0fb184 | 77 | dirty_background_ratio |
1da177e4 | 78 | |
db0fb184 PM |
79 | Contains, as a percentage of total system memory, the number of pages at which |
80 | the pdflush background writeback daemon will start writing out dirty data. | |
1da177e4 | 81 | |
db0fb184 | 82 | ============================================================== |
1da177e4 | 83 | |
db0fb184 PM |
84 | dirty_bytes |
85 | ||
86 | Contains the amount of dirty memory at which a process generating disk writes | |
87 | will itself start writeback. | |
88 | ||
89 | If dirty_bytes is written, dirty_ratio becomes a function of its value | |
90 | (dirty_bytes / the amount of dirtyable system memory). | |
1da177e4 | 91 | |
9e4a5bda AR |
92 | Note: the minimum value allowed for dirty_bytes is two pages (in bytes); any |
93 | value lower than this limit will be ignored and the old configuration will be | |
94 | retained. | |
95 | ||
1da177e4 LT |
96 | ============================================================== |
97 | ||
db0fb184 | 98 | dirty_expire_centisecs |
1da177e4 | 99 | |
db0fb184 PM |
100 | This tunable is used to define when dirty data is old enough to be eligible |
101 | for writeout by the pdflush daemons. It is expressed in 100'ths of a second. | |
102 | Data which has been dirty in-memory for longer than this interval will be | |
103 | written out next time a pdflush daemon wakes up. | |
104 | ||
105 | ============================================================== | |
106 | ||
107 | dirty_ratio | |
108 | ||
109 | Contains, as a percentage of total system memory, the number of pages at which | |
110 | a process which is generating disk writes will itself start writing out dirty | |
111 | data. | |
1da177e4 LT |
112 | |
113 | ============================================================== | |
114 | ||
db0fb184 | 115 | dirty_writeback_centisecs |
1da177e4 | 116 | |
db0fb184 PM |
117 | The pdflush writeback daemons will periodically wake up and write `old' data |
118 | out to disk. This tunable expresses the interval between those wakeups, in | |
119 | 100'ths of a second. | |
1da177e4 | 120 | |
db0fb184 | 121 | Setting this to zero disables periodic writeback altogether. |
1da177e4 LT |
122 | |
123 | ============================================================== | |
124 | ||
db0fb184 | 125 | drop_caches |
1da177e4 | 126 | |
db0fb184 PM |
127 | Writing to this will cause the kernel to drop clean caches, dentries and |
128 | inodes from memory, causing that memory to become free. | |
1da177e4 | 129 | |
db0fb184 PM |
130 | To free pagecache: |
131 | echo 1 > /proc/sys/vm/drop_caches | |
132 | To free dentries and inodes: | |
133 | echo 2 > /proc/sys/vm/drop_caches | |
134 | To free pagecache, dentries and inodes: | |
135 | echo 3 > /proc/sys/vm/drop_caches | |
1da177e4 | 136 | |
db0fb184 PM |
137 | As this is a non-destructive operation and dirty objects are not freeable, the |
138 | user should run `sync' first. | |
1da177e4 LT |
139 | |
140 | ============================================================== | |
141 | ||
db0fb184 | 142 | hugepages_treat_as_movable |
1da177e4 | 143 | |
db0fb184 PM |
144 | This parameter is only useful when kernelcore= is specified at boot time to |
145 | create ZONE_MOVABLE for pages that may be reclaimed or migrated. Huge pages | |
146 | are not movable so are not normally allocated from ZONE_MOVABLE. A non-zero | |
147 | value written to hugepages_treat_as_movable allows huge pages to be allocated | |
148 | from ZONE_MOVABLE. | |
8ad4b1fb | 149 | |
db0fb184 PM |
150 | Once enabled, the ZONE_MOVABLE is treated as an area of memory the huge |
151 | pages pool can easily grow or shrink within. Assuming that applications are | |
152 | not running that mlock() a lot of memory, it is likely the huge pages pool | |
153 | can grow to the size of ZONE_MOVABLE by repeatedly entering the desired value | |
154 | into nr_hugepages and triggering page reclaim. | |
24950898 | 155 | |
8ad4b1fb RS |
156 | ============================================================== |
157 | ||
db0fb184 | 158 | hugetlb_shm_group |
8ad4b1fb | 159 | |
db0fb184 PM |
160 | hugetlb_shm_group contains group id that is allowed to create SysV |
161 | shared memory segment using hugetlb page. | |
8ad4b1fb | 162 | |
db0fb184 | 163 | ============================================================== |
8ad4b1fb | 164 | |
db0fb184 | 165 | laptop_mode |
1743660b | 166 | |
db0fb184 PM |
167 | laptop_mode is a knob that controls "laptop mode". All the things that are |
168 | controlled by this knob are discussed in Documentation/laptops/laptop-mode.txt. | |
1743660b | 169 | |
db0fb184 | 170 | ============================================================== |
1743660b | 171 | |
db0fb184 | 172 | legacy_va_layout |
1b2ffb78 | 173 | |
db0fb184 PM |
174 | If non-zero, this sysctl disables the new 32-bit mmap mmap layout - the kernel |
175 | will use the legacy (2.4) layout for all processes. | |
1b2ffb78 | 176 | |
db0fb184 | 177 | ============================================================== |
1b2ffb78 | 178 | |
db0fb184 PM |
179 | lowmem_reserve_ratio |
180 | ||
181 | For some specialised workloads on highmem machines it is dangerous for | |
182 | the kernel to allow process memory to be allocated from the "lowmem" | |
183 | zone. This is because that memory could then be pinned via the mlock() | |
184 | system call, or by unavailability of swapspace. | |
185 | ||
186 | And on large highmem machines this lack of reclaimable lowmem memory | |
187 | can be fatal. | |
188 | ||
189 | So the Linux page allocator has a mechanism which prevents allocations | |
190 | which _could_ use highmem from using too much lowmem. This means that | |
191 | a certain amount of lowmem is defended from the possibility of being | |
192 | captured into pinned user memory. | |
193 | ||
194 | (The same argument applies to the old 16 megabyte ISA DMA region. This | |
195 | mechanism will also defend that region from allocations which could use | |
196 | highmem or lowmem). | |
197 | ||
198 | The `lowmem_reserve_ratio' tunable determines how aggressive the kernel is | |
199 | in defending these lower zones. | |
200 | ||
201 | If you have a machine which uses highmem or ISA DMA and your | |
202 | applications are using mlock(), or if you are running with no swap then | |
203 | you probably should change the lowmem_reserve_ratio setting. | |
204 | ||
205 | The lowmem_reserve_ratio is an array. You can see them by reading this file. | |
206 | - | |
207 | % cat /proc/sys/vm/lowmem_reserve_ratio | |
208 | 256 256 32 | |
209 | - | |
210 | Note: # of this elements is one fewer than number of zones. Because the highest | |
211 | zone's value is not necessary for following calculation. | |
212 | ||
213 | But, these values are not used directly. The kernel calculates # of protection | |
214 | pages for each zones from them. These are shown as array of protection pages | |
215 | in /proc/zoneinfo like followings. (This is an example of x86-64 box). | |
216 | Each zone has an array of protection pages like this. | |
217 | ||
218 | - | |
219 | Node 0, zone DMA | |
220 | pages free 1355 | |
221 | min 3 | |
222 | low 3 | |
223 | high 4 | |
224 | : | |
225 | : | |
226 | numa_other 0 | |
227 | protection: (0, 2004, 2004, 2004) | |
228 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | |
229 | pagesets | |
230 | cpu: 0 pcp: 0 | |
231 | : | |
232 | - | |
233 | These protections are added to score to judge whether this zone should be used | |
234 | for page allocation or should be reclaimed. | |
235 | ||
236 | In this example, if normal pages (index=2) are required to this DMA zone and | |
41858966 MG |
237 | watermark[WMARK_HIGH] is used for watermark, the kernel judges this zone should |
238 | not be used because pages_free(1355) is smaller than watermark + protection[2] | |
db0fb184 PM |
239 | (4 + 2004 = 2008). If this protection value is 0, this zone would be used for |
240 | normal page requirement. If requirement is DMA zone(index=0), protection[0] | |
241 | (=0) is used. | |
242 | ||
243 | zone[i]'s protection[j] is calculated by following expression. | |
244 | ||
245 | (i < j): | |
246 | zone[i]->protection[j] | |
247 | = (total sums of present_pages from zone[i+1] to zone[j] on the node) | |
248 | / lowmem_reserve_ratio[i]; | |
249 | (i = j): | |
250 | (should not be protected. = 0; | |
251 | (i > j): | |
252 | (not necessary, but looks 0) | |
253 | ||
254 | The default values of lowmem_reserve_ratio[i] are | |
255 | 256 (if zone[i] means DMA or DMA32 zone) | |
256 | 32 (others). | |
257 | As above expression, they are reciprocal number of ratio. | |
258 | 256 means 1/256. # of protection pages becomes about "0.39%" of total present | |
259 | pages of higher zones on the node. | |
260 | ||
261 | If you would like to protect more pages, smaller values are effective. | |
262 | The minimum value is 1 (1/1 -> 100%). | |
1b2ffb78 | 263 | |
db0fb184 | 264 | ============================================================== |
1b2ffb78 | 265 | |
db0fb184 | 266 | max_map_count: |
1743660b | 267 | |
db0fb184 PM |
268 | This file contains the maximum number of memory map areas a process |
269 | may have. Memory map areas are used as a side-effect of calling | |
270 | malloc, directly by mmap and mprotect, and also when loading shared | |
271 | libraries. | |
1743660b | 272 | |
db0fb184 PM |
273 | While most applications need less than a thousand maps, certain |
274 | programs, particularly malloc debuggers, may consume lots of them, | |
275 | e.g., up to one or two maps per allocation. | |
fadd8fbd | 276 | |
db0fb184 | 277 | The default value is 65536. |
9614634f | 278 | |
6a46079c AK |
279 | ============================================================= |
280 | ||
281 | memory_failure_early_kill: | |
282 | ||
283 | Control how to kill processes when uncorrected memory error (typically | |
284 | a 2bit error in a memory module) is detected in the background by hardware | |
285 | that cannot be handled by the kernel. In some cases (like the page | |
286 | still having a valid copy on disk) the kernel will handle the failure | |
287 | transparently without affecting any applications. But if there is | |
288 | no other uptodate copy of the data it will kill to prevent any data | |
289 | corruptions from propagating. | |
290 | ||
291 | 1: Kill all processes that have the corrupted and not reloadable page mapped | |
292 | as soon as the corruption is detected. Note this is not supported | |
293 | for a few types of pages, like kernel internally allocated data or | |
294 | the swap cache, but works for the majority of user pages. | |
295 | ||
296 | 0: Only unmap the corrupted page from all processes and only kill a process | |
297 | who tries to access it. | |
298 | ||
299 | The kill is done using a catchable SIGBUS with BUS_MCEERR_AO, so processes can | |
300 | handle this if they want to. | |
301 | ||
302 | This is only active on architectures/platforms with advanced machine | |
303 | check handling and depends on the hardware capabilities. | |
304 | ||
305 | Applications can override this setting individually with the PR_MCE_KILL prctl | |
306 | ||
307 | ============================================================== | |
308 | ||
309 | memory_failure_recovery | |
310 | ||
311 | Enable memory failure recovery (when supported by the platform) | |
312 | ||
313 | 1: Attempt recovery. | |
314 | ||
315 | 0: Always panic on a memory failure. | |
316 | ||
db0fb184 | 317 | ============================================================== |
9614634f | 318 | |
db0fb184 | 319 | min_free_kbytes: |
9614634f | 320 | |
db0fb184 | 321 | This is used to force the Linux VM to keep a minimum number |
41858966 MG |
322 | of kilobytes free. The VM uses this number to compute a |
323 | watermark[WMARK_MIN] value for each lowmem zone in the system. | |
324 | Each lowmem zone gets a number of reserved free pages based | |
325 | proportionally on its size. | |
db0fb184 PM |
326 | |
327 | Some minimal amount of memory is needed to satisfy PF_MEMALLOC | |
328 | allocations; if you set this to lower than 1024KB, your system will | |
329 | become subtly broken, and prone to deadlock under high loads. | |
330 | ||
331 | Setting this too high will OOM your machine instantly. | |
9614634f CL |
332 | |
333 | ============================================================= | |
334 | ||
0ff38490 CL |
335 | min_slab_ratio: |
336 | ||
337 | This is available only on NUMA kernels. | |
338 | ||
339 | A percentage of the total pages in each zone. On Zone reclaim | |
340 | (fallback from the local zone occurs) slabs will be reclaimed if more | |
341 | than this percentage of pages in a zone are reclaimable slab pages. | |
342 | This insures that the slab growth stays under control even in NUMA | |
343 | systems that rarely perform global reclaim. | |
344 | ||
345 | The default is 5 percent. | |
346 | ||
347 | Note that slab reclaim is triggered in a per zone / node fashion. | |
348 | The process of reclaiming slab memory is currently not node specific | |
349 | and may not be fast. | |
350 | ||
351 | ============================================================= | |
352 | ||
db0fb184 | 353 | min_unmapped_ratio: |
fadd8fbd | 354 | |
db0fb184 | 355 | This is available only on NUMA kernels. |
fadd8fbd | 356 | |
90afa5de MG |
357 | This is a percentage of the total pages in each zone. Zone reclaim will |
358 | only occur if more than this percentage of pages are in a state that | |
359 | zone_reclaim_mode allows to be reclaimed. | |
360 | ||
361 | If zone_reclaim_mode has the value 4 OR'd, then the percentage is compared | |
362 | against all file-backed unmapped pages including swapcache pages and tmpfs | |
363 | files. Otherwise, only unmapped pages backed by normal files but not tmpfs | |
364 | files and similar are considered. | |
2b744c01 | 365 | |
db0fb184 | 366 | The default is 1 percent. |
fadd8fbd | 367 | |
db0fb184 | 368 | ============================================================== |
2b744c01 | 369 | |
db0fb184 | 370 | mmap_min_addr |
ed032189 | 371 | |
db0fb184 | 372 | This file indicates the amount of address space which a user process will |
af901ca1 | 373 | be restricted from mmapping. Since kernel null dereference bugs could |
db0fb184 PM |
374 | accidentally operate based on the information in the first couple of pages |
375 | of memory userspace processes should not be allowed to write to them. By | |
376 | default this value is set to 0 and no protections will be enforced by the | |
377 | security module. Setting this value to something like 64k will allow the | |
378 | vast majority of applications to work correctly and provide defense in depth | |
379 | against future potential kernel bugs. | |
fe071d7e | 380 | |
db0fb184 | 381 | ============================================================== |
fef1bdd6 | 382 | |
db0fb184 | 383 | nr_hugepages |
fef1bdd6 | 384 | |
db0fb184 | 385 | Change the minimum size of the hugepage pool. |
fef1bdd6 | 386 | |
db0fb184 | 387 | See Documentation/vm/hugetlbpage.txt |
fef1bdd6 | 388 | |
db0fb184 | 389 | ============================================================== |
fef1bdd6 | 390 | |
db0fb184 | 391 | nr_overcommit_hugepages |
fef1bdd6 | 392 | |
db0fb184 PM |
393 | Change the maximum size of the hugepage pool. The maximum is |
394 | nr_hugepages + nr_overcommit_hugepages. | |
fe071d7e | 395 | |
db0fb184 | 396 | See Documentation/vm/hugetlbpage.txt |
fe071d7e | 397 | |
db0fb184 | 398 | ============================================================== |
fe071d7e | 399 | |
db0fb184 | 400 | nr_pdflush_threads |
fe071d7e | 401 | |
db0fb184 PM |
402 | The current number of pdflush threads. This value is read-only. |
403 | The value changes according to the number of dirty pages in the system. | |
fe071d7e | 404 | |
19f59460 | 405 | When necessary, additional pdflush threads are created, one per second, up to |
db0fb184 | 406 | nr_pdflush_threads_max. |
fe071d7e | 407 | |
ed032189 EP |
408 | ============================================================== |
409 | ||
db0fb184 | 410 | nr_trim_pages |
ed032189 | 411 | |
db0fb184 PM |
412 | This is available only on NOMMU kernels. |
413 | ||
414 | This value adjusts the excess page trimming behaviour of power-of-2 aligned | |
415 | NOMMU mmap allocations. | |
416 | ||
417 | A value of 0 disables trimming of allocations entirely, while a value of 1 | |
418 | trims excess pages aggressively. Any value >= 1 acts as the watermark where | |
419 | trimming of allocations is initiated. | |
420 | ||
421 | The default value is 1. | |
422 | ||
423 | See Documentation/nommu-mmap.txt for more information. | |
ed032189 | 424 | |
f0c0b2b8 KH |
425 | ============================================================== |
426 | ||
427 | numa_zonelist_order | |
428 | ||
429 | This sysctl is only for NUMA. | |
430 | 'where the memory is allocated from' is controlled by zonelists. | |
431 | (This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation. | |
432 | you may be able to read ZONE_DMA as ZONE_DMA32...) | |
433 | ||
434 | In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following. | |
435 | ZONE_NORMAL -> ZONE_DMA | |
436 | This means that a memory allocation request for GFP_KERNEL will | |
437 | get memory from ZONE_DMA only when ZONE_NORMAL is not available. | |
438 | ||
439 | In NUMA case, you can think of following 2 types of order. | |
440 | Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL | |
441 | ||
442 | (A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL | |
443 | (B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA. | |
444 | ||
445 | Type(A) offers the best locality for processes on Node(0), but ZONE_DMA | |
446 | will be used before ZONE_NORMAL exhaustion. This increases possibility of | |
447 | out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small. | |
448 | ||
449 | Type(B) cannot offer the best locality but is more robust against OOM of | |
450 | the DMA zone. | |
451 | ||
452 | Type(A) is called as "Node" order. Type (B) is "Zone" order. | |
453 | ||
454 | "Node order" orders the zonelists by node, then by zone within each node. | |
455 | Specify "[Nn]ode" for zone order | |
456 | ||
457 | "Zone Order" orders the zonelists by zone type, then by node within each | |
458 | zone. Specify "[Zz]one"for zode order. | |
459 | ||
460 | Specify "[Dd]efault" to request automatic configuration. Autoconfiguration | |
461 | will select "node" order in following case. | |
462 | (1) if the DMA zone does not exist or | |
463 | (2) if the DMA zone comprises greater than 50% of the available memory or | |
464 | (3) if any node's DMA zone comprises greater than 60% of its local memory and | |
465 | the amount of local memory is big enough. | |
466 | ||
467 | Otherwise, "zone" order will be selected. Default order is recommended unless | |
468 | this is causing problems for your system/application. | |
d5dbac87 NA |
469 | |
470 | ============================================================== | |
471 | ||
db0fb184 | 472 | oom_dump_tasks |
d5dbac87 | 473 | |
db0fb184 PM |
474 | Enables a system-wide task dump (excluding kernel threads) to be |
475 | produced when the kernel performs an OOM-killing and includes such | |
476 | information as pid, uid, tgid, vm size, rss, cpu, oom_adj score, and | |
477 | name. This is helpful to determine why the OOM killer was invoked | |
478 | and to identify the rogue task that caused it. | |
d5dbac87 | 479 | |
db0fb184 PM |
480 | If this is set to zero, this information is suppressed. On very |
481 | large systems with thousands of tasks it may not be feasible to dump | |
482 | the memory state information for each one. Such systems should not | |
483 | be forced to incur a performance penalty in OOM conditions when the | |
484 | information may not be desired. | |
485 | ||
486 | If this is set to non-zero, this information is shown whenever the | |
487 | OOM killer actually kills a memory-hogging task. | |
488 | ||
489 | The default value is 0. | |
d5dbac87 NA |
490 | |
491 | ============================================================== | |
492 | ||
db0fb184 | 493 | oom_kill_allocating_task |
d5dbac87 | 494 | |
db0fb184 PM |
495 | This enables or disables killing the OOM-triggering task in |
496 | out-of-memory situations. | |
d5dbac87 | 497 | |
db0fb184 PM |
498 | If this is set to zero, the OOM killer will scan through the entire |
499 | tasklist and select a task based on heuristics to kill. This normally | |
500 | selects a rogue memory-hogging task that frees up a large amount of | |
501 | memory when killed. | |
502 | ||
503 | If this is set to non-zero, the OOM killer simply kills the task that | |
504 | triggered the out-of-memory condition. This avoids the expensive | |
505 | tasklist scan. | |
506 | ||
507 | If panic_on_oom is selected, it takes precedence over whatever value | |
508 | is used in oom_kill_allocating_task. | |
509 | ||
510 | The default value is 0. | |
dd8632a1 PM |
511 | |
512 | ============================================================== | |
513 | ||
db0fb184 | 514 | overcommit_memory: |
dd8632a1 | 515 | |
db0fb184 | 516 | This value contains a flag that enables memory overcommitment. |
dd8632a1 | 517 | |
db0fb184 PM |
518 | When this flag is 0, the kernel attempts to estimate the amount |
519 | of free memory left when userspace requests more memory. | |
dd8632a1 | 520 | |
db0fb184 PM |
521 | When this flag is 1, the kernel pretends there is always enough |
522 | memory until it actually runs out. | |
dd8632a1 | 523 | |
db0fb184 PM |
524 | When this flag is 2, the kernel uses a "never overcommit" |
525 | policy that attempts to prevent any overcommit of memory. | |
dd8632a1 | 526 | |
db0fb184 PM |
527 | This feature can be very useful because there are a lot of |
528 | programs that malloc() huge amounts of memory "just-in-case" | |
529 | and don't use much of it. | |
530 | ||
531 | The default value is 0. | |
532 | ||
533 | See Documentation/vm/overcommit-accounting and | |
534 | security/commoncap.c::cap_vm_enough_memory() for more information. | |
535 | ||
536 | ============================================================== | |
537 | ||
538 | overcommit_ratio: | |
539 | ||
540 | When overcommit_memory is set to 2, the committed address | |
541 | space is not permitted to exceed swap plus this percentage | |
542 | of physical RAM. See above. | |
543 | ||
544 | ============================================================== | |
545 | ||
546 | page-cluster | |
547 | ||
548 | page-cluster controls the number of pages which are written to swap in | |
549 | a single attempt. The swap I/O size. | |
550 | ||
551 | It is a logarithmic value - setting it to zero means "1 page", setting | |
552 | it to 1 means "2 pages", setting it to 2 means "4 pages", etc. | |
553 | ||
554 | The default value is three (eight pages at a time). There may be some | |
555 | small benefits in tuning this to a different value if your workload is | |
556 | swap-intensive. | |
557 | ||
558 | ============================================================= | |
559 | ||
560 | panic_on_oom | |
561 | ||
562 | This enables or disables panic on out-of-memory feature. | |
563 | ||
564 | If this is set to 0, the kernel will kill some rogue process, | |
565 | called oom_killer. Usually, oom_killer can kill rogue processes and | |
566 | system will survive. | |
567 | ||
568 | If this is set to 1, the kernel panics when out-of-memory happens. | |
569 | However, if a process limits using nodes by mempolicy/cpusets, | |
570 | and those nodes become memory exhaustion status, one process | |
571 | may be killed by oom-killer. No panic occurs in this case. | |
572 | Because other nodes' memory may be free. This means system total status | |
573 | may be not fatal yet. | |
574 | ||
575 | If this is set to 2, the kernel panics compulsorily even on the | |
576 | above-mentioned. | |
577 | ||
578 | The default value is 0. | |
579 | 1 and 2 are for failover of clustering. Please select either | |
580 | according to your policy of failover. | |
581 | ||
582 | ============================================================= | |
583 | ||
584 | percpu_pagelist_fraction | |
585 | ||
586 | This is the fraction of pages at most (high mark pcp->high) in each zone that | |
587 | are allocated for each per cpu page list. The min value for this is 8. It | |
588 | means that we don't allow more than 1/8th of pages in each zone to be | |
589 | allocated in any single per_cpu_pagelist. This entry only changes the value | |
590 | of hot per cpu pagelists. User can specify a number like 100 to allocate | |
591 | 1/100th of each zone to each per cpu page list. | |
592 | ||
593 | The batch value of each per cpu pagelist is also updated as a result. It is | |
594 | set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8) | |
595 | ||
596 | The initial value is zero. Kernel does not use this value at boot time to set | |
597 | the high water marks for each per cpu page list. | |
598 | ||
599 | ============================================================== | |
600 | ||
601 | stat_interval | |
602 | ||
603 | The time interval between which vm statistics are updated. The default | |
604 | is 1 second. | |
605 | ||
606 | ============================================================== | |
607 | ||
608 | swappiness | |
609 | ||
610 | This control is used to define how aggressive the kernel will swap | |
611 | memory pages. Higher values will increase agressiveness, lower values | |
19f59460 | 612 | decrease the amount of swap. |
db0fb184 PM |
613 | |
614 | The default value is 60. | |
615 | ||
616 | ============================================================== | |
617 | ||
618 | vfs_cache_pressure | |
619 | ------------------ | |
620 | ||
621 | Controls the tendency of the kernel to reclaim the memory which is used for | |
622 | caching of directory and inode objects. | |
623 | ||
624 | At the default value of vfs_cache_pressure=100 the kernel will attempt to | |
625 | reclaim dentries and inodes at a "fair" rate with respect to pagecache and | |
626 | swapcache reclaim. Decreasing vfs_cache_pressure causes the kernel to prefer | |
55c37a84 JK |
627 | to retain dentry and inode caches. When vfs_cache_pressure=0, the kernel will |
628 | never reclaim dentries and inodes due to memory pressure and this can easily | |
629 | lead to out-of-memory conditions. Increasing vfs_cache_pressure beyond 100 | |
db0fb184 PM |
630 | causes the kernel to prefer to reclaim dentries and inodes. |
631 | ||
632 | ============================================================== | |
633 | ||
634 | zone_reclaim_mode: | |
635 | ||
636 | Zone_reclaim_mode allows someone to set more or less aggressive approaches to | |
637 | reclaim memory when a zone runs out of memory. If it is set to zero then no | |
638 | zone reclaim occurs. Allocations will be satisfied from other zones / nodes | |
639 | in the system. | |
640 | ||
641 | This is value ORed together of | |
642 | ||
643 | 1 = Zone reclaim on | |
644 | 2 = Zone reclaim writes dirty pages out | |
645 | 4 = Zone reclaim swaps pages | |
646 | ||
647 | zone_reclaim_mode is set during bootup to 1 if it is determined that pages | |
648 | from remote zones will cause a measurable performance reduction. The | |
649 | page allocator will then reclaim easily reusable pages (those page | |
650 | cache pages that are currently not used) before allocating off node pages. | |
651 | ||
652 | It may be beneficial to switch off zone reclaim if the system is | |
653 | used for a file server and all of memory should be used for caching files | |
654 | from disk. In that case the caching effect is more important than | |
655 | data locality. | |
656 | ||
657 | Allowing zone reclaim to write out pages stops processes that are | |
658 | writing large amounts of data from dirtying pages on other nodes. Zone | |
659 | reclaim will write out dirty pages if a zone fills up and so effectively | |
660 | throttle the process. This may decrease the performance of a single process | |
661 | since it cannot use all of system memory to buffer the outgoing writes | |
662 | anymore but it preserve the memory on other nodes so that the performance | |
663 | of other processes running on other nodes will not be affected. | |
664 | ||
665 | Allowing regular swap effectively restricts allocations to the local | |
666 | node unless explicitly overridden by memory policies or cpuset | |
667 | configurations. | |
668 | ||
669 | ============ End of Document ================================= |