]>
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 | 20 | |
4eeab4f5 | 21 | - admin_reserve_kbytes |
db0fb184 | 22 | - block_dump |
76ab0f53 | 23 | - compact_memory |
5bbe3547 | 24 | - compact_unevictable_allowed |
db0fb184 | 25 | - dirty_background_bytes |
1da177e4 | 26 | - dirty_background_ratio |
db0fb184 | 27 | - dirty_bytes |
1da177e4 | 28 | - dirty_expire_centisecs |
db0fb184 | 29 | - dirty_ratio |
1da177e4 | 30 | - dirty_writeback_centisecs |
db0fb184 | 31 | - drop_caches |
5e771905 | 32 | - extfrag_threshold |
db0fb184 PM |
33 | - hugepages_treat_as_movable |
34 | - hugetlb_shm_group | |
35 | - laptop_mode | |
36 | - legacy_va_layout | |
37 | - lowmem_reserve_ratio | |
1da177e4 | 38 | - max_map_count |
6a46079c AK |
39 | - memory_failure_early_kill |
40 | - memory_failure_recovery | |
1da177e4 | 41 | - min_free_kbytes |
0ff38490 | 42 | - min_slab_ratio |
db0fb184 PM |
43 | - min_unmapped_ratio |
44 | - mmap_min_addr | |
d07e2259 DC |
45 | - mmap_rnd_bits |
46 | - mmap_rnd_compat_bits | |
d5dbac87 NA |
47 | - nr_hugepages |
48 | - nr_overcommit_hugepages | |
db0fb184 PM |
49 | - nr_trim_pages (only if CONFIG_MMU=n) |
50 | - numa_zonelist_order | |
51 | - oom_dump_tasks | |
52 | - oom_kill_allocating_task | |
49f0ce5f | 53 | - overcommit_kbytes |
db0fb184 PM |
54 | - overcommit_memory |
55 | - overcommit_ratio | |
56 | - page-cluster | |
57 | - panic_on_oom | |
58 | - percpu_pagelist_fraction | |
59 | - stat_interval | |
60 | - swappiness | |
c9b1d098 | 61 | - user_reserve_kbytes |
db0fb184 PM |
62 | - vfs_cache_pressure |
63 | - zone_reclaim_mode | |
64 | ||
1da177e4 LT |
65 | ============================================================== |
66 | ||
4eeab4f5 AS |
67 | admin_reserve_kbytes |
68 | ||
69 | The amount of free memory in the system that should be reserved for users | |
70 | with the capability cap_sys_admin. | |
71 | ||
72 | admin_reserve_kbytes defaults to min(3% of free pages, 8MB) | |
73 | ||
74 | That should provide enough for the admin to log in and kill a process, | |
75 | if necessary, under the default overcommit 'guess' mode. | |
76 | ||
77 | Systems running under overcommit 'never' should increase this to account | |
78 | for the full Virtual Memory Size of programs used to recover. Otherwise, | |
79 | root may not be able to log in to recover the system. | |
80 | ||
81 | How do you calculate a minimum useful reserve? | |
82 | ||
83 | sshd or login + bash (or some other shell) + top (or ps, kill, etc.) | |
84 | ||
85 | For overcommit 'guess', we can sum resident set sizes (RSS). | |
86 | On x86_64 this is about 8MB. | |
87 | ||
88 | For overcommit 'never', we can take the max of their virtual sizes (VSZ) | |
89 | and add the sum of their RSS. | |
90 | On x86_64 this is about 128MB. | |
91 | ||
92 | Changing this takes effect whenever an application requests memory. | |
93 | ||
94 | ============================================================== | |
95 | ||
db0fb184 | 96 | block_dump |
1da177e4 | 97 | |
db0fb184 PM |
98 | block_dump enables block I/O debugging when set to a nonzero value. More |
99 | information on block I/O debugging is in Documentation/laptops/laptop-mode.txt. | |
1da177e4 LT |
100 | |
101 | ============================================================== | |
102 | ||
76ab0f53 MG |
103 | compact_memory |
104 | ||
105 | Available only when CONFIG_COMPACTION is set. When 1 is written to the file, | |
106 | all zones are compacted such that free memory is available in contiguous | |
107 | blocks where possible. This can be important for example in the allocation of | |
108 | huge pages although processes will also directly compact memory as required. | |
109 | ||
110 | ============================================================== | |
111 | ||
5bbe3547 EM |
112 | compact_unevictable_allowed |
113 | ||
114 | Available only when CONFIG_COMPACTION is set. When set to 1, compaction is | |
115 | allowed to examine the unevictable lru (mlocked pages) for pages to compact. | |
116 | This should be used on systems where stalls for minor page faults are an | |
117 | acceptable trade for large contiguous free memory. Set to 0 to prevent | |
118 | compaction from moving pages that are unevictable. Default value is 1. | |
119 | ||
120 | ============================================================== | |
121 | ||
db0fb184 | 122 | dirty_background_bytes |
1da177e4 | 123 | |
6601fac8 AB |
124 | Contains the amount of dirty memory at which the background kernel |
125 | flusher threads will start writeback. | |
1da177e4 | 126 | |
abffc020 AR |
127 | Note: dirty_background_bytes is the counterpart of dirty_background_ratio. Only |
128 | one of them may be specified at a time. When one sysctl is written it is | |
129 | immediately taken into account to evaluate the dirty memory limits and the | |
130 | other appears as 0 when read. | |
1da177e4 | 131 | |
db0fb184 | 132 | ============================================================== |
1da177e4 | 133 | |
db0fb184 | 134 | dirty_background_ratio |
1da177e4 | 135 | |
715ea41e ZL |
136 | Contains, as a percentage of total available memory that contains free pages |
137 | and reclaimable pages, the number of pages at which the background kernel | |
138 | flusher threads will start writing out dirty data. | |
139 | ||
d83e2a4e | 140 | The total available memory is not equal to total system memory. |
1da177e4 | 141 | |
db0fb184 | 142 | ============================================================== |
1da177e4 | 143 | |
db0fb184 PM |
144 | dirty_bytes |
145 | ||
146 | Contains the amount of dirty memory at which a process generating disk writes | |
147 | will itself start writeback. | |
148 | ||
abffc020 AR |
149 | Note: dirty_bytes is the counterpart of dirty_ratio. Only one of them may be |
150 | specified at a time. When one sysctl is written it is immediately taken into | |
151 | account to evaluate the dirty memory limits and the other appears as 0 when | |
152 | read. | |
1da177e4 | 153 | |
9e4a5bda AR |
154 | Note: the minimum value allowed for dirty_bytes is two pages (in bytes); any |
155 | value lower than this limit will be ignored and the old configuration will be | |
156 | retained. | |
157 | ||
1da177e4 LT |
158 | ============================================================== |
159 | ||
db0fb184 | 160 | dirty_expire_centisecs |
1da177e4 | 161 | |
db0fb184 | 162 | This tunable is used to define when dirty data is old enough to be eligible |
6601fac8 AB |
163 | for writeout by the kernel flusher threads. It is expressed in 100'ths |
164 | of a second. Data which has been dirty in-memory for longer than this | |
165 | interval will be written out next time a flusher thread wakes up. | |
db0fb184 PM |
166 | |
167 | ============================================================== | |
168 | ||
169 | dirty_ratio | |
170 | ||
715ea41e ZL |
171 | Contains, as a percentage of total available memory that contains free pages |
172 | and reclaimable pages, the number of pages at which a process which is | |
173 | generating disk writes will itself start writing out dirty data. | |
174 | ||
d83e2a4e | 175 | The total available memory is not equal to total system memory. |
1da177e4 LT |
176 | |
177 | ============================================================== | |
178 | ||
db0fb184 | 179 | dirty_writeback_centisecs |
1da177e4 | 180 | |
6601fac8 | 181 | The kernel flusher threads will periodically wake up and write `old' data |
db0fb184 PM |
182 | out to disk. This tunable expresses the interval between those wakeups, in |
183 | 100'ths of a second. | |
1da177e4 | 184 | |
db0fb184 | 185 | Setting this to zero disables periodic writeback altogether. |
1da177e4 LT |
186 | |
187 | ============================================================== | |
188 | ||
db0fb184 | 189 | drop_caches |
1da177e4 | 190 | |
5509a5d2 DH |
191 | Writing to this will cause the kernel to drop clean caches, as well as |
192 | reclaimable slab objects like dentries and inodes. Once dropped, their | |
193 | memory becomes free. | |
1da177e4 | 194 | |
db0fb184 PM |
195 | To free pagecache: |
196 | echo 1 > /proc/sys/vm/drop_caches | |
5509a5d2 | 197 | To free reclaimable slab objects (includes dentries and inodes): |
db0fb184 | 198 | echo 2 > /proc/sys/vm/drop_caches |
5509a5d2 | 199 | To free slab objects and pagecache: |
db0fb184 | 200 | echo 3 > /proc/sys/vm/drop_caches |
1da177e4 | 201 | |
5509a5d2 DH |
202 | This is a non-destructive operation and will not free any dirty objects. |
203 | To increase the number of objects freed by this operation, the user may run | |
204 | `sync' prior to writing to /proc/sys/vm/drop_caches. This will minimize the | |
205 | number of dirty objects on the system and create more candidates to be | |
206 | dropped. | |
207 | ||
208 | This file is not a means to control the growth of the various kernel caches | |
209 | (inodes, dentries, pagecache, etc...) These objects are automatically | |
210 | reclaimed by the kernel when memory is needed elsewhere on the system. | |
211 | ||
212 | Use of this file can cause performance problems. Since it discards cached | |
213 | objects, it may cost a significant amount of I/O and CPU to recreate the | |
214 | dropped objects, especially if they were under heavy use. Because of this, | |
215 | use outside of a testing or debugging environment is not recommended. | |
216 | ||
217 | You may see informational messages in your kernel log when this file is | |
218 | used: | |
219 | ||
220 | cat (1234): drop_caches: 3 | |
221 | ||
222 | These are informational only. They do not mean that anything is wrong | |
223 | with your system. To disable them, echo 4 (bit 3) into drop_caches. | |
1da177e4 LT |
224 | |
225 | ============================================================== | |
226 | ||
5e771905 MG |
227 | extfrag_threshold |
228 | ||
229 | This parameter affects whether the kernel will compact memory or direct | |
a10726bb RV |
230 | reclaim to satisfy a high-order allocation. The extfrag/extfrag_index file in |
231 | debugfs shows what the fragmentation index for each order is in each zone in | |
232 | the system. Values tending towards 0 imply allocations would fail due to lack | |
233 | of memory, values towards 1000 imply failures are due to fragmentation and -1 | |
234 | implies that the allocation will succeed as long as watermarks are met. | |
5e771905 MG |
235 | |
236 | The kernel will not compact memory in a zone if the | |
237 | fragmentation index is <= extfrag_threshold. The default value is 500. | |
238 | ||
239 | ============================================================== | |
240 | ||
db0fb184 | 241 | hugepages_treat_as_movable |
1da177e4 | 242 | |
86cdb465 NH |
243 | This parameter controls whether we can allocate hugepages from ZONE_MOVABLE |
244 | or not. If set to non-zero, hugepages can be allocated from ZONE_MOVABLE. | |
245 | ZONE_MOVABLE is created when kernel boot parameter kernelcore= is specified, | |
246 | so this parameter has no effect if used without kernelcore=. | |
247 | ||
248 | Hugepage migration is now available in some situations which depend on the | |
249 | architecture and/or the hugepage size. If a hugepage supports migration, | |
250 | allocation from ZONE_MOVABLE is always enabled for the hugepage regardless | |
251 | of the value of this parameter. | |
252 | IOW, this parameter affects only non-migratable hugepages. | |
253 | ||
254 | Assuming that hugepages are not migratable in your system, one usecase of | |
255 | this parameter is that users can make hugepage pool more extensible by | |
256 | enabling the allocation from ZONE_MOVABLE. This is because on ZONE_MOVABLE | |
257 | page reclaim/migration/compaction work more and you can get contiguous | |
258 | memory more likely. Note that using ZONE_MOVABLE for non-migratable | |
259 | hugepages can do harm to other features like memory hotremove (because | |
260 | memory hotremove expects that memory blocks on ZONE_MOVABLE are always | |
261 | removable,) so it's a trade-off responsible for the users. | |
24950898 | 262 | |
8ad4b1fb RS |
263 | ============================================================== |
264 | ||
db0fb184 | 265 | hugetlb_shm_group |
8ad4b1fb | 266 | |
db0fb184 PM |
267 | hugetlb_shm_group contains group id that is allowed to create SysV |
268 | shared memory segment using hugetlb page. | |
8ad4b1fb | 269 | |
db0fb184 | 270 | ============================================================== |
8ad4b1fb | 271 | |
db0fb184 | 272 | laptop_mode |
1743660b | 273 | |
db0fb184 PM |
274 | laptop_mode is a knob that controls "laptop mode". All the things that are |
275 | controlled by this knob are discussed in Documentation/laptops/laptop-mode.txt. | |
1743660b | 276 | |
db0fb184 | 277 | ============================================================== |
1743660b | 278 | |
db0fb184 | 279 | legacy_va_layout |
1b2ffb78 | 280 | |
2174efb6 | 281 | If non-zero, this sysctl disables the new 32-bit mmap layout - the kernel |
db0fb184 | 282 | will use the legacy (2.4) layout for all processes. |
1b2ffb78 | 283 | |
db0fb184 | 284 | ============================================================== |
1b2ffb78 | 285 | |
db0fb184 PM |
286 | lowmem_reserve_ratio |
287 | ||
288 | For some specialised workloads on highmem machines it is dangerous for | |
289 | the kernel to allow process memory to be allocated from the "lowmem" | |
290 | zone. This is because that memory could then be pinned via the mlock() | |
291 | system call, or by unavailability of swapspace. | |
292 | ||
293 | And on large highmem machines this lack of reclaimable lowmem memory | |
294 | can be fatal. | |
295 | ||
296 | So the Linux page allocator has a mechanism which prevents allocations | |
297 | which _could_ use highmem from using too much lowmem. This means that | |
298 | a certain amount of lowmem is defended from the possibility of being | |
299 | captured into pinned user memory. | |
300 | ||
301 | (The same argument applies to the old 16 megabyte ISA DMA region. This | |
302 | mechanism will also defend that region from allocations which could use | |
303 | highmem or lowmem). | |
304 | ||
305 | The `lowmem_reserve_ratio' tunable determines how aggressive the kernel is | |
306 | in defending these lower zones. | |
307 | ||
308 | If you have a machine which uses highmem or ISA DMA and your | |
309 | applications are using mlock(), or if you are running with no swap then | |
310 | you probably should change the lowmem_reserve_ratio setting. | |
311 | ||
312 | The lowmem_reserve_ratio is an array. You can see them by reading this file. | |
313 | - | |
314 | % cat /proc/sys/vm/lowmem_reserve_ratio | |
315 | 256 256 32 | |
316 | - | |
317 | Note: # of this elements is one fewer than number of zones. Because the highest | |
318 | zone's value is not necessary for following calculation. | |
319 | ||
320 | But, these values are not used directly. The kernel calculates # of protection | |
321 | pages for each zones from them. These are shown as array of protection pages | |
322 | in /proc/zoneinfo like followings. (This is an example of x86-64 box). | |
323 | Each zone has an array of protection pages like this. | |
324 | ||
325 | - | |
326 | Node 0, zone DMA | |
327 | pages free 1355 | |
328 | min 3 | |
329 | low 3 | |
330 | high 4 | |
331 | : | |
332 | : | |
333 | numa_other 0 | |
334 | protection: (0, 2004, 2004, 2004) | |
335 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | |
336 | pagesets | |
337 | cpu: 0 pcp: 0 | |
338 | : | |
339 | - | |
340 | These protections are added to score to judge whether this zone should be used | |
341 | for page allocation or should be reclaimed. | |
342 | ||
343 | In this example, if normal pages (index=2) are required to this DMA zone and | |
41858966 MG |
344 | watermark[WMARK_HIGH] is used for watermark, the kernel judges this zone should |
345 | not be used because pages_free(1355) is smaller than watermark + protection[2] | |
db0fb184 PM |
346 | (4 + 2004 = 2008). If this protection value is 0, this zone would be used for |
347 | normal page requirement. If requirement is DMA zone(index=0), protection[0] | |
348 | (=0) is used. | |
349 | ||
350 | zone[i]'s protection[j] is calculated by following expression. | |
351 | ||
352 | (i < j): | |
353 | zone[i]->protection[j] | |
013110a7 | 354 | = (total sums of managed_pages from zone[i+1] to zone[j] on the node) |
db0fb184 PM |
355 | / lowmem_reserve_ratio[i]; |
356 | (i = j): | |
357 | (should not be protected. = 0; | |
358 | (i > j): | |
359 | (not necessary, but looks 0) | |
360 | ||
361 | The default values of lowmem_reserve_ratio[i] are | |
362 | 256 (if zone[i] means DMA or DMA32 zone) | |
363 | 32 (others). | |
364 | As above expression, they are reciprocal number of ratio. | |
013110a7 | 365 | 256 means 1/256. # of protection pages becomes about "0.39%" of total managed |
db0fb184 PM |
366 | pages of higher zones on the node. |
367 | ||
368 | If you would like to protect more pages, smaller values are effective. | |
369 | The minimum value is 1 (1/1 -> 100%). | |
1b2ffb78 | 370 | |
db0fb184 | 371 | ============================================================== |
1b2ffb78 | 372 | |
db0fb184 | 373 | max_map_count: |
1743660b | 374 | |
db0fb184 PM |
375 | This file contains the maximum number of memory map areas a process |
376 | may have. Memory map areas are used as a side-effect of calling | |
377 | malloc, directly by mmap and mprotect, and also when loading shared | |
378 | libraries. | |
1743660b | 379 | |
db0fb184 PM |
380 | While most applications need less than a thousand maps, certain |
381 | programs, particularly malloc debuggers, may consume lots of them, | |
382 | e.g., up to one or two maps per allocation. | |
fadd8fbd | 383 | |
db0fb184 | 384 | The default value is 65536. |
9614634f | 385 | |
6a46079c AK |
386 | ============================================================= |
387 | ||
388 | memory_failure_early_kill: | |
389 | ||
390 | Control how to kill processes when uncorrected memory error (typically | |
391 | a 2bit error in a memory module) is detected in the background by hardware | |
392 | that cannot be handled by the kernel. In some cases (like the page | |
393 | still having a valid copy on disk) the kernel will handle the failure | |
394 | transparently without affecting any applications. But if there is | |
395 | no other uptodate copy of the data it will kill to prevent any data | |
396 | corruptions from propagating. | |
397 | ||
398 | 1: Kill all processes that have the corrupted and not reloadable page mapped | |
399 | as soon as the corruption is detected. Note this is not supported | |
400 | for a few types of pages, like kernel internally allocated data or | |
401 | the swap cache, but works for the majority of user pages. | |
402 | ||
403 | 0: Only unmap the corrupted page from all processes and only kill a process | |
404 | who tries to access it. | |
405 | ||
406 | The kill is done using a catchable SIGBUS with BUS_MCEERR_AO, so processes can | |
407 | handle this if they want to. | |
408 | ||
409 | This is only active on architectures/platforms with advanced machine | |
410 | check handling and depends on the hardware capabilities. | |
411 | ||
412 | Applications can override this setting individually with the PR_MCE_KILL prctl | |
413 | ||
414 | ============================================================== | |
415 | ||
416 | memory_failure_recovery | |
417 | ||
418 | Enable memory failure recovery (when supported by the platform) | |
419 | ||
420 | 1: Attempt recovery. | |
421 | ||
422 | 0: Always panic on a memory failure. | |
423 | ||
db0fb184 | 424 | ============================================================== |
9614634f | 425 | |
db0fb184 | 426 | min_free_kbytes: |
9614634f | 427 | |
db0fb184 | 428 | This is used to force the Linux VM to keep a minimum number |
41858966 MG |
429 | of kilobytes free. The VM uses this number to compute a |
430 | watermark[WMARK_MIN] value for each lowmem zone in the system. | |
431 | Each lowmem zone gets a number of reserved free pages based | |
432 | proportionally on its size. | |
db0fb184 PM |
433 | |
434 | Some minimal amount of memory is needed to satisfy PF_MEMALLOC | |
435 | allocations; if you set this to lower than 1024KB, your system will | |
436 | become subtly broken, and prone to deadlock under high loads. | |
437 | ||
438 | Setting this too high will OOM your machine instantly. | |
9614634f CL |
439 | |
440 | ============================================================= | |
441 | ||
0ff38490 CL |
442 | min_slab_ratio: |
443 | ||
444 | This is available only on NUMA kernels. | |
445 | ||
446 | A percentage of the total pages in each zone. On Zone reclaim | |
447 | (fallback from the local zone occurs) slabs will be reclaimed if more | |
448 | than this percentage of pages in a zone are reclaimable slab pages. | |
449 | This insures that the slab growth stays under control even in NUMA | |
450 | systems that rarely perform global reclaim. | |
451 | ||
452 | The default is 5 percent. | |
453 | ||
454 | Note that slab reclaim is triggered in a per zone / node fashion. | |
455 | The process of reclaiming slab memory is currently not node specific | |
456 | and may not be fast. | |
457 | ||
458 | ============================================================= | |
459 | ||
db0fb184 | 460 | min_unmapped_ratio: |
fadd8fbd | 461 | |
db0fb184 | 462 | This is available only on NUMA kernels. |
fadd8fbd | 463 | |
90afa5de MG |
464 | This is a percentage of the total pages in each zone. Zone reclaim will |
465 | only occur if more than this percentage of pages are in a state that | |
466 | zone_reclaim_mode allows to be reclaimed. | |
467 | ||
468 | If zone_reclaim_mode has the value 4 OR'd, then the percentage is compared | |
469 | against all file-backed unmapped pages including swapcache pages and tmpfs | |
470 | files. Otherwise, only unmapped pages backed by normal files but not tmpfs | |
471 | files and similar are considered. | |
2b744c01 | 472 | |
db0fb184 | 473 | The default is 1 percent. |
fadd8fbd | 474 | |
db0fb184 | 475 | ============================================================== |
2b744c01 | 476 | |
db0fb184 | 477 | mmap_min_addr |
ed032189 | 478 | |
db0fb184 | 479 | This file indicates the amount of address space which a user process will |
af901ca1 | 480 | be restricted from mmapping. Since kernel null dereference bugs could |
db0fb184 PM |
481 | accidentally operate based on the information in the first couple of pages |
482 | of memory userspace processes should not be allowed to write to them. By | |
483 | default this value is set to 0 and no protections will be enforced by the | |
484 | security module. Setting this value to something like 64k will allow the | |
485 | vast majority of applications to work correctly and provide defense in depth | |
486 | against future potential kernel bugs. | |
fe071d7e | 487 | |
db0fb184 | 488 | ============================================================== |
fef1bdd6 | 489 | |
d07e2259 DC |
490 | mmap_rnd_bits: |
491 | ||
492 | This value can be used to select the number of bits to use to | |
493 | determine the random offset to the base address of vma regions | |
494 | resulting from mmap allocations on architectures which support | |
495 | tuning address space randomization. This value will be bounded | |
496 | by the architecture's minimum and maximum supported values. | |
497 | ||
498 | This value can be changed after boot using the | |
499 | /proc/sys/vm/mmap_rnd_bits tunable | |
500 | ||
501 | ============================================================== | |
502 | ||
503 | mmap_rnd_compat_bits: | |
504 | ||
505 | This value can be used to select the number of bits to use to | |
506 | determine the random offset to the base address of vma regions | |
507 | resulting from mmap allocations for applications run in | |
508 | compatibility mode on architectures which support tuning address | |
509 | space randomization. This value will be bounded by the | |
510 | architecture's minimum and maximum supported values. | |
511 | ||
512 | This value can be changed after boot using the | |
513 | /proc/sys/vm/mmap_rnd_compat_bits tunable | |
514 | ||
515 | ============================================================== | |
516 | ||
db0fb184 | 517 | nr_hugepages |
fef1bdd6 | 518 | |
db0fb184 | 519 | Change the minimum size of the hugepage pool. |
fef1bdd6 | 520 | |
db0fb184 | 521 | See Documentation/vm/hugetlbpage.txt |
fef1bdd6 | 522 | |
db0fb184 | 523 | ============================================================== |
fef1bdd6 | 524 | |
db0fb184 | 525 | nr_overcommit_hugepages |
fef1bdd6 | 526 | |
db0fb184 PM |
527 | Change the maximum size of the hugepage pool. The maximum is |
528 | nr_hugepages + nr_overcommit_hugepages. | |
fe071d7e | 529 | |
db0fb184 | 530 | See Documentation/vm/hugetlbpage.txt |
fe071d7e | 531 | |
db0fb184 | 532 | ============================================================== |
fe071d7e | 533 | |
db0fb184 | 534 | nr_trim_pages |
ed032189 | 535 | |
db0fb184 PM |
536 | This is available only on NOMMU kernels. |
537 | ||
538 | This value adjusts the excess page trimming behaviour of power-of-2 aligned | |
539 | NOMMU mmap allocations. | |
540 | ||
541 | A value of 0 disables trimming of allocations entirely, while a value of 1 | |
542 | trims excess pages aggressively. Any value >= 1 acts as the watermark where | |
543 | trimming of allocations is initiated. | |
544 | ||
545 | The default value is 1. | |
546 | ||
547 | See Documentation/nommu-mmap.txt for more information. | |
ed032189 | 548 | |
f0c0b2b8 KH |
549 | ============================================================== |
550 | ||
551 | numa_zonelist_order | |
552 | ||
553 | This sysctl is only for NUMA. | |
554 | 'where the memory is allocated from' is controlled by zonelists. | |
555 | (This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation. | |
556 | you may be able to read ZONE_DMA as ZONE_DMA32...) | |
557 | ||
558 | In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following. | |
559 | ZONE_NORMAL -> ZONE_DMA | |
560 | This means that a memory allocation request for GFP_KERNEL will | |
561 | get memory from ZONE_DMA only when ZONE_NORMAL is not available. | |
562 | ||
563 | In NUMA case, you can think of following 2 types of order. | |
564 | Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL | |
565 | ||
566 | (A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL | |
567 | (B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA. | |
568 | ||
569 | Type(A) offers the best locality for processes on Node(0), but ZONE_DMA | |
570 | will be used before ZONE_NORMAL exhaustion. This increases possibility of | |
571 | out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small. | |
572 | ||
573 | Type(B) cannot offer the best locality but is more robust against OOM of | |
574 | the DMA zone. | |
575 | ||
576 | Type(A) is called as "Node" order. Type (B) is "Zone" order. | |
577 | ||
578 | "Node order" orders the zonelists by node, then by zone within each node. | |
5a3016a6 | 579 | Specify "[Nn]ode" for node order |
f0c0b2b8 KH |
580 | |
581 | "Zone Order" orders the zonelists by zone type, then by node within each | |
5a3016a6 | 582 | zone. Specify "[Zz]one" for zone order. |
f0c0b2b8 KH |
583 | |
584 | Specify "[Dd]efault" to request automatic configuration. Autoconfiguration | |
585 | will select "node" order in following case. | |
586 | (1) if the DMA zone does not exist or | |
587 | (2) if the DMA zone comprises greater than 50% of the available memory or | |
f8f191f1 | 588 | (3) if any node's DMA zone comprises greater than 70% of its local memory and |
f0c0b2b8 KH |
589 | the amount of local memory is big enough. |
590 | ||
591 | Otherwise, "zone" order will be selected. Default order is recommended unless | |
592 | this is causing problems for your system/application. | |
d5dbac87 NA |
593 | |
594 | ============================================================== | |
595 | ||
db0fb184 | 596 | oom_dump_tasks |
d5dbac87 | 597 | |
dc6c9a35 KS |
598 | Enables a system-wide task dump (excluding kernel threads) to be produced |
599 | when the kernel performs an OOM-killing and includes such information as | |
600 | pid, uid, tgid, vm size, rss, nr_ptes, nr_pmds, swapents, oom_score_adj | |
601 | score, and name. This is helpful to determine why the OOM killer was | |
602 | invoked, to identify the rogue task that caused it, and to determine why | |
603 | the OOM killer chose the task it did to kill. | |
d5dbac87 | 604 | |
db0fb184 PM |
605 | If this is set to zero, this information is suppressed. On very |
606 | large systems with thousands of tasks it may not be feasible to dump | |
607 | the memory state information for each one. Such systems should not | |
608 | be forced to incur a performance penalty in OOM conditions when the | |
609 | information may not be desired. | |
610 | ||
611 | If this is set to non-zero, this information is shown whenever the | |
612 | OOM killer actually kills a memory-hogging task. | |
613 | ||
ad915c43 | 614 | The default value is 1 (enabled). |
d5dbac87 NA |
615 | |
616 | ============================================================== | |
617 | ||
db0fb184 | 618 | oom_kill_allocating_task |
d5dbac87 | 619 | |
db0fb184 PM |
620 | This enables or disables killing the OOM-triggering task in |
621 | out-of-memory situations. | |
d5dbac87 | 622 | |
db0fb184 PM |
623 | If this is set to zero, the OOM killer will scan through the entire |
624 | tasklist and select a task based on heuristics to kill. This normally | |
625 | selects a rogue memory-hogging task that frees up a large amount of | |
626 | memory when killed. | |
627 | ||
628 | If this is set to non-zero, the OOM killer simply kills the task that | |
629 | triggered the out-of-memory condition. This avoids the expensive | |
630 | tasklist scan. | |
631 | ||
632 | If panic_on_oom is selected, it takes precedence over whatever value | |
633 | is used in oom_kill_allocating_task. | |
634 | ||
635 | The default value is 0. | |
dd8632a1 PM |
636 | |
637 | ============================================================== | |
638 | ||
49f0ce5f JM |
639 | overcommit_kbytes: |
640 | ||
641 | When overcommit_memory is set to 2, the committed address space is not | |
642 | permitted to exceed swap plus this amount of physical RAM. See below. | |
643 | ||
644 | Note: overcommit_kbytes is the counterpart of overcommit_ratio. Only one | |
645 | of them may be specified at a time. Setting one disables the other (which | |
646 | then appears as 0 when read). | |
647 | ||
648 | ============================================================== | |
649 | ||
db0fb184 | 650 | overcommit_memory: |
dd8632a1 | 651 | |
db0fb184 | 652 | This value contains a flag that enables memory overcommitment. |
dd8632a1 | 653 | |
db0fb184 PM |
654 | When this flag is 0, the kernel attempts to estimate the amount |
655 | of free memory left when userspace requests more memory. | |
dd8632a1 | 656 | |
db0fb184 PM |
657 | When this flag is 1, the kernel pretends there is always enough |
658 | memory until it actually runs out. | |
dd8632a1 | 659 | |
db0fb184 PM |
660 | When this flag is 2, the kernel uses a "never overcommit" |
661 | policy that attempts to prevent any overcommit of memory. | |
c9b1d098 | 662 | Note that user_reserve_kbytes affects this policy. |
dd8632a1 | 663 | |
db0fb184 PM |
664 | This feature can be very useful because there are a lot of |
665 | programs that malloc() huge amounts of memory "just-in-case" | |
666 | and don't use much of it. | |
667 | ||
668 | The default value is 0. | |
669 | ||
670 | See Documentation/vm/overcommit-accounting and | |
c56050c7 | 671 | mm/mmap.c::__vm_enough_memory() for more information. |
db0fb184 PM |
672 | |
673 | ============================================================== | |
674 | ||
675 | overcommit_ratio: | |
676 | ||
677 | When overcommit_memory is set to 2, the committed address | |
678 | space is not permitted to exceed swap plus this percentage | |
679 | of physical RAM. See above. | |
680 | ||
681 | ============================================================== | |
682 | ||
683 | page-cluster | |
684 | ||
df858fa8 CE |
685 | page-cluster controls the number of pages up to which consecutive pages |
686 | are read in from swap in a single attempt. This is the swap counterpart | |
687 | to page cache readahead. | |
688 | The mentioned consecutivity is not in terms of virtual/physical addresses, | |
689 | but consecutive on swap space - that means they were swapped out together. | |
db0fb184 PM |
690 | |
691 | It is a logarithmic value - setting it to zero means "1 page", setting | |
692 | it to 1 means "2 pages", setting it to 2 means "4 pages", etc. | |
df858fa8 | 693 | Zero disables swap readahead completely. |
db0fb184 PM |
694 | |
695 | The default value is three (eight pages at a time). There may be some | |
696 | small benefits in tuning this to a different value if your workload is | |
697 | swap-intensive. | |
698 | ||
df858fa8 CE |
699 | Lower values mean lower latencies for initial faults, but at the same time |
700 | extra faults and I/O delays for following faults if they would have been part of | |
701 | that consecutive pages readahead would have brought in. | |
702 | ||
db0fb184 PM |
703 | ============================================================= |
704 | ||
705 | panic_on_oom | |
706 | ||
707 | This enables or disables panic on out-of-memory feature. | |
708 | ||
709 | If this is set to 0, the kernel will kill some rogue process, | |
710 | called oom_killer. Usually, oom_killer can kill rogue processes and | |
711 | system will survive. | |
712 | ||
713 | If this is set to 1, the kernel panics when out-of-memory happens. | |
714 | However, if a process limits using nodes by mempolicy/cpusets, | |
715 | and those nodes become memory exhaustion status, one process | |
716 | may be killed by oom-killer. No panic occurs in this case. | |
717 | Because other nodes' memory may be free. This means system total status | |
718 | may be not fatal yet. | |
719 | ||
720 | If this is set to 2, the kernel panics compulsorily even on the | |
daaf1e68 KH |
721 | above-mentioned. Even oom happens under memory cgroup, the whole |
722 | system panics. | |
db0fb184 PM |
723 | |
724 | The default value is 0. | |
725 | 1 and 2 are for failover of clustering. Please select either | |
726 | according to your policy of failover. | |
daaf1e68 KH |
727 | panic_on_oom=2+kdump gives you very strong tool to investigate |
728 | why oom happens. You can get snapshot. | |
db0fb184 PM |
729 | |
730 | ============================================================= | |
731 | ||
732 | percpu_pagelist_fraction | |
733 | ||
734 | This is the fraction of pages at most (high mark pcp->high) in each zone that | |
735 | are allocated for each per cpu page list. The min value for this is 8. It | |
736 | means that we don't allow more than 1/8th of pages in each zone to be | |
737 | allocated in any single per_cpu_pagelist. This entry only changes the value | |
738 | of hot per cpu pagelists. User can specify a number like 100 to allocate | |
739 | 1/100th of each zone to each per cpu page list. | |
740 | ||
741 | The batch value of each per cpu pagelist is also updated as a result. It is | |
742 | set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8) | |
743 | ||
744 | The initial value is zero. Kernel does not use this value at boot time to set | |
7cd2b0a3 DR |
745 | the high water marks for each per cpu page list. If the user writes '0' to this |
746 | sysctl, it will revert to this default behavior. | |
db0fb184 PM |
747 | |
748 | ============================================================== | |
749 | ||
750 | stat_interval | |
751 | ||
752 | The time interval between which vm statistics are updated. The default | |
753 | is 1 second. | |
754 | ||
755 | ============================================================== | |
756 | ||
757 | swappiness | |
758 | ||
759 | This control is used to define how aggressive the kernel will swap | |
760 | memory pages. Higher values will increase agressiveness, lower values | |
8582cb96 AT |
761 | decrease the amount of swap. A value of 0 instructs the kernel not to |
762 | initiate swap until the amount of free and file-backed pages is less | |
763 | than the high water mark in a zone. | |
db0fb184 PM |
764 | |
765 | The default value is 60. | |
766 | ||
767 | ============================================================== | |
768 | ||
c9b1d098 AS |
769 | - user_reserve_kbytes |
770 | ||
633708a4 | 771 | When overcommit_memory is set to 2, "never overcommit" mode, reserve |
c9b1d098 AS |
772 | min(3% of current process size, user_reserve_kbytes) of free memory. |
773 | This is intended to prevent a user from starting a single memory hogging | |
774 | process, such that they cannot recover (kill the hog). | |
775 | ||
776 | user_reserve_kbytes defaults to min(3% of the current process size, 128MB). | |
777 | ||
778 | If this is reduced to zero, then the user will be allowed to allocate | |
779 | all free memory with a single process, minus admin_reserve_kbytes. | |
780 | Any subsequent attempts to execute a command will result in | |
781 | "fork: Cannot allocate memory". | |
782 | ||
783 | Changing this takes effect whenever an application requests memory. | |
784 | ||
785 | ============================================================== | |
786 | ||
db0fb184 PM |
787 | vfs_cache_pressure |
788 | ------------------ | |
789 | ||
4a0da71b DV |
790 | This percentage value controls the tendency of the kernel to reclaim |
791 | the memory which is used for caching of directory and inode objects. | |
db0fb184 PM |
792 | |
793 | At the default value of vfs_cache_pressure=100 the kernel will attempt to | |
794 | reclaim dentries and inodes at a "fair" rate with respect to pagecache and | |
795 | swapcache reclaim. Decreasing vfs_cache_pressure causes the kernel to prefer | |
55c37a84 JK |
796 | to retain dentry and inode caches. When vfs_cache_pressure=0, the kernel will |
797 | never reclaim dentries and inodes due to memory pressure and this can easily | |
798 | lead to out-of-memory conditions. Increasing vfs_cache_pressure beyond 100 | |
db0fb184 PM |
799 | causes the kernel to prefer to reclaim dentries and inodes. |
800 | ||
4a0da71b DV |
801 | Increasing vfs_cache_pressure significantly beyond 100 may have negative |
802 | performance impact. Reclaim code needs to take various locks to find freeable | |
803 | directory and inode objects. With vfs_cache_pressure=1000, it will look for | |
804 | ten times more freeable objects than there are. | |
805 | ||
795ae7a0 JW |
806 | ============================================================= |
807 | ||
808 | watermark_scale_factor: | |
809 | ||
810 | This factor controls the aggressiveness of kswapd. It defines the | |
811 | amount of memory left in a node/system before kswapd is woken up and | |
812 | how much memory needs to be free before kswapd goes back to sleep. | |
813 | ||
814 | The unit is in fractions of 10,000. The default value of 10 means the | |
815 | distances between watermarks are 0.1% of the available memory in the | |
816 | node/system. The maximum value is 1000, or 10% of memory. | |
817 | ||
818 | A high rate of threads entering direct reclaim (allocstall) or kswapd | |
819 | going to sleep prematurely (kswapd_low_wmark_hit_quickly) can indicate | |
820 | that the number of free pages kswapd maintains for latency reasons is | |
821 | too small for the allocation bursts occurring in the system. This knob | |
822 | can then be used to tune kswapd aggressiveness accordingly. | |
823 | ||
db0fb184 PM |
824 | ============================================================== |
825 | ||
826 | zone_reclaim_mode: | |
827 | ||
828 | Zone_reclaim_mode allows someone to set more or less aggressive approaches to | |
829 | reclaim memory when a zone runs out of memory. If it is set to zero then no | |
830 | zone reclaim occurs. Allocations will be satisfied from other zones / nodes | |
831 | in the system. | |
832 | ||
833 | This is value ORed together of | |
834 | ||
835 | 1 = Zone reclaim on | |
836 | 2 = Zone reclaim writes dirty pages out | |
837 | 4 = Zone reclaim swaps pages | |
838 | ||
4f9b16a6 MG |
839 | zone_reclaim_mode is disabled by default. For file servers or workloads |
840 | that benefit from having their data cached, zone_reclaim_mode should be | |
841 | left disabled as the caching effect is likely to be more important than | |
db0fb184 PM |
842 | data locality. |
843 | ||
4f9b16a6 MG |
844 | zone_reclaim may be enabled if it's known that the workload is partitioned |
845 | such that each partition fits within a NUMA node and that accessing remote | |
846 | memory would cause a measurable performance reduction. The page allocator | |
847 | will then reclaim easily reusable pages (those page cache pages that are | |
848 | currently not used) before allocating off node pages. | |
849 | ||
db0fb184 PM |
850 | Allowing zone reclaim to write out pages stops processes that are |
851 | writing large amounts of data from dirtying pages on other nodes. Zone | |
852 | reclaim will write out dirty pages if a zone fills up and so effectively | |
853 | throttle the process. This may decrease the performance of a single process | |
854 | since it cannot use all of system memory to buffer the outgoing writes | |
855 | anymore but it preserve the memory on other nodes so that the performance | |
856 | of other processes running on other nodes will not be affected. | |
857 | ||
858 | Allowing regular swap effectively restricts allocations to the local | |
859 | node unless explicitly overridden by memory policies or cpuset | |
860 | configurations. | |
861 | ||
862 | ============ End of Document ================================= |