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