]>
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 |
db0fb184 | 24 | - dirty_background_bytes |
1da177e4 | 25 | - dirty_background_ratio |
db0fb184 | 26 | - dirty_bytes |
1da177e4 | 27 | - dirty_expire_centisecs |
db0fb184 | 28 | - dirty_ratio |
1da177e4 | 29 | - dirty_writeback_centisecs |
db0fb184 | 30 | - drop_caches |
5e771905 | 31 | - extfrag_threshold |
db0fb184 PM |
32 | - hugepages_treat_as_movable |
33 | - hugetlb_shm_group | |
34 | - laptop_mode | |
35 | - legacy_va_layout | |
36 | - lowmem_reserve_ratio | |
1da177e4 | 37 | - max_map_count |
6a46079c AK |
38 | - memory_failure_early_kill |
39 | - memory_failure_recovery | |
1da177e4 | 40 | - min_free_kbytes |
0ff38490 | 41 | - min_slab_ratio |
db0fb184 PM |
42 | - min_unmapped_ratio |
43 | - mmap_min_addr | |
d5dbac87 NA |
44 | - nr_hugepages |
45 | - nr_overcommit_hugepages | |
db0fb184 PM |
46 | - nr_trim_pages (only if CONFIG_MMU=n) |
47 | - numa_zonelist_order | |
48 | - oom_dump_tasks | |
49 | - oom_kill_allocating_task | |
50 | - overcommit_memory | |
51 | - overcommit_ratio | |
52 | - page-cluster | |
53 | - panic_on_oom | |
54 | - percpu_pagelist_fraction | |
55 | - stat_interval | |
56 | - swappiness | |
c9b1d098 | 57 | - user_reserve_kbytes |
db0fb184 PM |
58 | - vfs_cache_pressure |
59 | - zone_reclaim_mode | |
60 | ||
1da177e4 LT |
61 | ============================================================== |
62 | ||
4eeab4f5 AS |
63 | admin_reserve_kbytes |
64 | ||
65 | The amount of free memory in the system that should be reserved for users | |
66 | with the capability cap_sys_admin. | |
67 | ||
68 | admin_reserve_kbytes defaults to min(3% of free pages, 8MB) | |
69 | ||
70 | That should provide enough for the admin to log in and kill a process, | |
71 | if necessary, under the default overcommit 'guess' mode. | |
72 | ||
73 | Systems running under overcommit 'never' should increase this to account | |
74 | for the full Virtual Memory Size of programs used to recover. Otherwise, | |
75 | root may not be able to log in to recover the system. | |
76 | ||
77 | How do you calculate a minimum useful reserve? | |
78 | ||
79 | sshd or login + bash (or some other shell) + top (or ps, kill, etc.) | |
80 | ||
81 | For overcommit 'guess', we can sum resident set sizes (RSS). | |
82 | On x86_64 this is about 8MB. | |
83 | ||
84 | For overcommit 'never', we can take the max of their virtual sizes (VSZ) | |
85 | and add the sum of their RSS. | |
86 | On x86_64 this is about 128MB. | |
87 | ||
88 | Changing this takes effect whenever an application requests memory. | |
89 | ||
90 | ============================================================== | |
91 | ||
db0fb184 | 92 | block_dump |
1da177e4 | 93 | |
db0fb184 PM |
94 | block_dump enables block I/O debugging when set to a nonzero value. More |
95 | information on block I/O debugging is in Documentation/laptops/laptop-mode.txt. | |
1da177e4 LT |
96 | |
97 | ============================================================== | |
98 | ||
76ab0f53 MG |
99 | compact_memory |
100 | ||
101 | Available only when CONFIG_COMPACTION is set. When 1 is written to the file, | |
102 | all zones are compacted such that free memory is available in contiguous | |
103 | blocks where possible. This can be important for example in the allocation of | |
104 | huge pages although processes will also directly compact memory as required. | |
105 | ||
106 | ============================================================== | |
107 | ||
db0fb184 | 108 | dirty_background_bytes |
1da177e4 | 109 | |
6601fac8 AB |
110 | Contains the amount of dirty memory at which the background kernel |
111 | flusher threads will start writeback. | |
1da177e4 | 112 | |
abffc020 AR |
113 | Note: dirty_background_bytes is the counterpart of dirty_background_ratio. Only |
114 | one of them may be specified at a time. When one sysctl is written it is | |
115 | immediately taken into account to evaluate the dirty memory limits and the | |
116 | other appears as 0 when read. | |
1da177e4 | 117 | |
db0fb184 | 118 | ============================================================== |
1da177e4 | 119 | |
db0fb184 | 120 | dirty_background_ratio |
1da177e4 | 121 | |
715ea41e ZL |
122 | Contains, as a percentage of total available memory that contains free pages |
123 | and reclaimable pages, the number of pages at which the background kernel | |
124 | flusher threads will start writing out dirty data. | |
125 | ||
126 | The total avaiable memory is not equal to total system memory. | |
1da177e4 | 127 | |
db0fb184 | 128 | ============================================================== |
1da177e4 | 129 | |
db0fb184 PM |
130 | dirty_bytes |
131 | ||
132 | Contains the amount of dirty memory at which a process generating disk writes | |
133 | will itself start writeback. | |
134 | ||
abffc020 AR |
135 | Note: dirty_bytes is the counterpart of dirty_ratio. Only one of them may be |
136 | specified at a time. When one sysctl is written it is immediately taken into | |
137 | account to evaluate the dirty memory limits and the other appears as 0 when | |
138 | read. | |
1da177e4 | 139 | |
9e4a5bda AR |
140 | Note: the minimum value allowed for dirty_bytes is two pages (in bytes); any |
141 | value lower than this limit will be ignored and the old configuration will be | |
142 | retained. | |
143 | ||
1da177e4 LT |
144 | ============================================================== |
145 | ||
db0fb184 | 146 | dirty_expire_centisecs |
1da177e4 | 147 | |
db0fb184 | 148 | This tunable is used to define when dirty data is old enough to be eligible |
6601fac8 AB |
149 | for writeout by the kernel flusher threads. It is expressed in 100'ths |
150 | of a second. Data which has been dirty in-memory for longer than this | |
151 | interval will be written out next time a flusher thread wakes up. | |
db0fb184 PM |
152 | |
153 | ============================================================== | |
154 | ||
155 | dirty_ratio | |
156 | ||
715ea41e ZL |
157 | Contains, as a percentage of total available memory that contains free pages |
158 | and reclaimable pages, the number of pages at which a process which is | |
159 | generating disk writes will itself start writing out dirty data. | |
160 | ||
161 | The total avaiable memory is not equal to total system memory. | |
1da177e4 LT |
162 | |
163 | ============================================================== | |
164 | ||
db0fb184 | 165 | dirty_writeback_centisecs |
1da177e4 | 166 | |
6601fac8 | 167 | The kernel flusher threads will periodically wake up and write `old' data |
db0fb184 PM |
168 | out to disk. This tunable expresses the interval between those wakeups, in |
169 | 100'ths of a second. | |
1da177e4 | 170 | |
db0fb184 | 171 | Setting this to zero disables periodic writeback altogether. |
1da177e4 LT |
172 | |
173 | ============================================================== | |
174 | ||
db0fb184 | 175 | drop_caches |
1da177e4 | 176 | |
db0fb184 PM |
177 | Writing to this will cause the kernel to drop clean caches, dentries and |
178 | inodes from memory, causing that memory to become free. | |
1da177e4 | 179 | |
db0fb184 PM |
180 | To free pagecache: |
181 | echo 1 > /proc/sys/vm/drop_caches | |
182 | To free dentries and inodes: | |
183 | echo 2 > /proc/sys/vm/drop_caches | |
184 | To free pagecache, dentries and inodes: | |
185 | echo 3 > /proc/sys/vm/drop_caches | |
1da177e4 | 186 | |
db0fb184 PM |
187 | As this is a non-destructive operation and dirty objects are not freeable, the |
188 | user should run `sync' first. | |
1da177e4 LT |
189 | |
190 | ============================================================== | |
191 | ||
5e771905 MG |
192 | extfrag_threshold |
193 | ||
194 | This parameter affects whether the kernel will compact memory or direct | |
195 | reclaim to satisfy a high-order allocation. /proc/extfrag_index shows what | |
196 | the fragmentation index for each order is in each zone in the system. Values | |
197 | tending towards 0 imply allocations would fail due to lack of memory, | |
198 | values towards 1000 imply failures are due to fragmentation and -1 implies | |
199 | that the allocation will succeed as long as watermarks are met. | |
200 | ||
201 | The kernel will not compact memory in a zone if the | |
202 | fragmentation index is <= extfrag_threshold. The default value is 500. | |
203 | ||
204 | ============================================================== | |
205 | ||
db0fb184 | 206 | hugepages_treat_as_movable |
1da177e4 | 207 | |
86cdb465 NH |
208 | This parameter controls whether we can allocate hugepages from ZONE_MOVABLE |
209 | or not. If set to non-zero, hugepages can be allocated from ZONE_MOVABLE. | |
210 | ZONE_MOVABLE is created when kernel boot parameter kernelcore= is specified, | |
211 | so this parameter has no effect if used without kernelcore=. | |
212 | ||
213 | Hugepage migration is now available in some situations which depend on the | |
214 | architecture and/or the hugepage size. If a hugepage supports migration, | |
215 | allocation from ZONE_MOVABLE is always enabled for the hugepage regardless | |
216 | of the value of this parameter. | |
217 | IOW, this parameter affects only non-migratable hugepages. | |
218 | ||
219 | Assuming that hugepages are not migratable in your system, one usecase of | |
220 | this parameter is that users can make hugepage pool more extensible by | |
221 | enabling the allocation from ZONE_MOVABLE. This is because on ZONE_MOVABLE | |
222 | page reclaim/migration/compaction work more and you can get contiguous | |
223 | memory more likely. Note that using ZONE_MOVABLE for non-migratable | |
224 | hugepages can do harm to other features like memory hotremove (because | |
225 | memory hotremove expects that memory blocks on ZONE_MOVABLE are always | |
226 | removable,) so it's a trade-off responsible for the users. | |
24950898 | 227 | |
8ad4b1fb RS |
228 | ============================================================== |
229 | ||
db0fb184 | 230 | hugetlb_shm_group |
8ad4b1fb | 231 | |
db0fb184 PM |
232 | hugetlb_shm_group contains group id that is allowed to create SysV |
233 | shared memory segment using hugetlb page. | |
8ad4b1fb | 234 | |
db0fb184 | 235 | ============================================================== |
8ad4b1fb | 236 | |
db0fb184 | 237 | laptop_mode |
1743660b | 238 | |
db0fb184 PM |
239 | laptop_mode is a knob that controls "laptop mode". All the things that are |
240 | controlled by this knob are discussed in Documentation/laptops/laptop-mode.txt. | |
1743660b | 241 | |
db0fb184 | 242 | ============================================================== |
1743660b | 243 | |
db0fb184 | 244 | legacy_va_layout |
1b2ffb78 | 245 | |
2174efb6 | 246 | If non-zero, this sysctl disables the new 32-bit mmap layout - the kernel |
db0fb184 | 247 | will use the legacy (2.4) layout for all processes. |
1b2ffb78 | 248 | |
db0fb184 | 249 | ============================================================== |
1b2ffb78 | 250 | |
db0fb184 PM |
251 | lowmem_reserve_ratio |
252 | ||
253 | For some specialised workloads on highmem machines it is dangerous for | |
254 | the kernel to allow process memory to be allocated from the "lowmem" | |
255 | zone. This is because that memory could then be pinned via the mlock() | |
256 | system call, or by unavailability of swapspace. | |
257 | ||
258 | And on large highmem machines this lack of reclaimable lowmem memory | |
259 | can be fatal. | |
260 | ||
261 | So the Linux page allocator has a mechanism which prevents allocations | |
262 | which _could_ use highmem from using too much lowmem. This means that | |
263 | a certain amount of lowmem is defended from the possibility of being | |
264 | captured into pinned user memory. | |
265 | ||
266 | (The same argument applies to the old 16 megabyte ISA DMA region. This | |
267 | mechanism will also defend that region from allocations which could use | |
268 | highmem or lowmem). | |
269 | ||
270 | The `lowmem_reserve_ratio' tunable determines how aggressive the kernel is | |
271 | in defending these lower zones. | |
272 | ||
273 | If you have a machine which uses highmem or ISA DMA and your | |
274 | applications are using mlock(), or if you are running with no swap then | |
275 | you probably should change the lowmem_reserve_ratio setting. | |
276 | ||
277 | The lowmem_reserve_ratio is an array. You can see them by reading this file. | |
278 | - | |
279 | % cat /proc/sys/vm/lowmem_reserve_ratio | |
280 | 256 256 32 | |
281 | - | |
282 | Note: # of this elements is one fewer than number of zones. Because the highest | |
283 | zone's value is not necessary for following calculation. | |
284 | ||
285 | But, these values are not used directly. The kernel calculates # of protection | |
286 | pages for each zones from them. These are shown as array of protection pages | |
287 | in /proc/zoneinfo like followings. (This is an example of x86-64 box). | |
288 | Each zone has an array of protection pages like this. | |
289 | ||
290 | - | |
291 | Node 0, zone DMA | |
292 | pages free 1355 | |
293 | min 3 | |
294 | low 3 | |
295 | high 4 | |
296 | : | |
297 | : | |
298 | numa_other 0 | |
299 | protection: (0, 2004, 2004, 2004) | |
300 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | |
301 | pagesets | |
302 | cpu: 0 pcp: 0 | |
303 | : | |
304 | - | |
305 | These protections are added to score to judge whether this zone should be used | |
306 | for page allocation or should be reclaimed. | |
307 | ||
308 | In this example, if normal pages (index=2) are required to this DMA zone and | |
41858966 MG |
309 | watermark[WMARK_HIGH] is used for watermark, the kernel judges this zone should |
310 | not be used because pages_free(1355) is smaller than watermark + protection[2] | |
db0fb184 PM |
311 | (4 + 2004 = 2008). If this protection value is 0, this zone would be used for |
312 | normal page requirement. If requirement is DMA zone(index=0), protection[0] | |
313 | (=0) is used. | |
314 | ||
315 | zone[i]'s protection[j] is calculated by following expression. | |
316 | ||
317 | (i < j): | |
318 | zone[i]->protection[j] | |
319 | = (total sums of present_pages from zone[i+1] to zone[j] on the node) | |
320 | / lowmem_reserve_ratio[i]; | |
321 | (i = j): | |
322 | (should not be protected. = 0; | |
323 | (i > j): | |
324 | (not necessary, but looks 0) | |
325 | ||
326 | The default values of lowmem_reserve_ratio[i] are | |
327 | 256 (if zone[i] means DMA or DMA32 zone) | |
328 | 32 (others). | |
329 | As above expression, they are reciprocal number of ratio. | |
330 | 256 means 1/256. # of protection pages becomes about "0.39%" of total present | |
331 | pages of higher zones on the node. | |
332 | ||
333 | If you would like to protect more pages, smaller values are effective. | |
334 | The minimum value is 1 (1/1 -> 100%). | |
1b2ffb78 | 335 | |
db0fb184 | 336 | ============================================================== |
1b2ffb78 | 337 | |
db0fb184 | 338 | max_map_count: |
1743660b | 339 | |
db0fb184 PM |
340 | This file contains the maximum number of memory map areas a process |
341 | may have. Memory map areas are used as a side-effect of calling | |
342 | malloc, directly by mmap and mprotect, and also when loading shared | |
343 | libraries. | |
1743660b | 344 | |
db0fb184 PM |
345 | While most applications need less than a thousand maps, certain |
346 | programs, particularly malloc debuggers, may consume lots of them, | |
347 | e.g., up to one or two maps per allocation. | |
fadd8fbd | 348 | |
db0fb184 | 349 | The default value is 65536. |
9614634f | 350 | |
6a46079c AK |
351 | ============================================================= |
352 | ||
353 | memory_failure_early_kill: | |
354 | ||
355 | Control how to kill processes when uncorrected memory error (typically | |
356 | a 2bit error in a memory module) is detected in the background by hardware | |
357 | that cannot be handled by the kernel. In some cases (like the page | |
358 | still having a valid copy on disk) the kernel will handle the failure | |
359 | transparently without affecting any applications. But if there is | |
360 | no other uptodate copy of the data it will kill to prevent any data | |
361 | corruptions from propagating. | |
362 | ||
363 | 1: Kill all processes that have the corrupted and not reloadable page mapped | |
364 | as soon as the corruption is detected. Note this is not supported | |
365 | for a few types of pages, like kernel internally allocated data or | |
366 | the swap cache, but works for the majority of user pages. | |
367 | ||
368 | 0: Only unmap the corrupted page from all processes and only kill a process | |
369 | who tries to access it. | |
370 | ||
371 | The kill is done using a catchable SIGBUS with BUS_MCEERR_AO, so processes can | |
372 | handle this if they want to. | |
373 | ||
374 | This is only active on architectures/platforms with advanced machine | |
375 | check handling and depends on the hardware capabilities. | |
376 | ||
377 | Applications can override this setting individually with the PR_MCE_KILL prctl | |
378 | ||
379 | ============================================================== | |
380 | ||
381 | memory_failure_recovery | |
382 | ||
383 | Enable memory failure recovery (when supported by the platform) | |
384 | ||
385 | 1: Attempt recovery. | |
386 | ||
387 | 0: Always panic on a memory failure. | |
388 | ||
db0fb184 | 389 | ============================================================== |
9614634f | 390 | |
db0fb184 | 391 | min_free_kbytes: |
9614634f | 392 | |
db0fb184 | 393 | This is used to force the Linux VM to keep a minimum number |
41858966 MG |
394 | of kilobytes free. The VM uses this number to compute a |
395 | watermark[WMARK_MIN] value for each lowmem zone in the system. | |
396 | Each lowmem zone gets a number of reserved free pages based | |
397 | proportionally on its size. | |
db0fb184 PM |
398 | |
399 | Some minimal amount of memory is needed to satisfy PF_MEMALLOC | |
400 | allocations; if you set this to lower than 1024KB, your system will | |
401 | become subtly broken, and prone to deadlock under high loads. | |
402 | ||
403 | Setting this too high will OOM your machine instantly. | |
9614634f CL |
404 | |
405 | ============================================================= | |
406 | ||
0ff38490 CL |
407 | min_slab_ratio: |
408 | ||
409 | This is available only on NUMA kernels. | |
410 | ||
411 | A percentage of the total pages in each zone. On Zone reclaim | |
412 | (fallback from the local zone occurs) slabs will be reclaimed if more | |
413 | than this percentage of pages in a zone are reclaimable slab pages. | |
414 | This insures that the slab growth stays under control even in NUMA | |
415 | systems that rarely perform global reclaim. | |
416 | ||
417 | The default is 5 percent. | |
418 | ||
419 | Note that slab reclaim is triggered in a per zone / node fashion. | |
420 | The process of reclaiming slab memory is currently not node specific | |
421 | and may not be fast. | |
422 | ||
423 | ============================================================= | |
424 | ||
db0fb184 | 425 | min_unmapped_ratio: |
fadd8fbd | 426 | |
db0fb184 | 427 | This is available only on NUMA kernels. |
fadd8fbd | 428 | |
90afa5de MG |
429 | This is a percentage of the total pages in each zone. Zone reclaim will |
430 | only occur if more than this percentage of pages are in a state that | |
431 | zone_reclaim_mode allows to be reclaimed. | |
432 | ||
433 | If zone_reclaim_mode has the value 4 OR'd, then the percentage is compared | |
434 | against all file-backed unmapped pages including swapcache pages and tmpfs | |
435 | files. Otherwise, only unmapped pages backed by normal files but not tmpfs | |
436 | files and similar are considered. | |
2b744c01 | 437 | |
db0fb184 | 438 | The default is 1 percent. |
fadd8fbd | 439 | |
db0fb184 | 440 | ============================================================== |
2b744c01 | 441 | |
db0fb184 | 442 | mmap_min_addr |
ed032189 | 443 | |
db0fb184 | 444 | This file indicates the amount of address space which a user process will |
af901ca1 | 445 | be restricted from mmapping. Since kernel null dereference bugs could |
db0fb184 PM |
446 | accidentally operate based on the information in the first couple of pages |
447 | of memory userspace processes should not be allowed to write to them. By | |
448 | default this value is set to 0 and no protections will be enforced by the | |
449 | security module. Setting this value to something like 64k will allow the | |
450 | vast majority of applications to work correctly and provide defense in depth | |
451 | against future potential kernel bugs. | |
fe071d7e | 452 | |
db0fb184 | 453 | ============================================================== |
fef1bdd6 | 454 | |
db0fb184 | 455 | nr_hugepages |
fef1bdd6 | 456 | |
db0fb184 | 457 | Change the minimum size of the hugepage pool. |
fef1bdd6 | 458 | |
db0fb184 | 459 | See Documentation/vm/hugetlbpage.txt |
fef1bdd6 | 460 | |
db0fb184 | 461 | ============================================================== |
fef1bdd6 | 462 | |
db0fb184 | 463 | nr_overcommit_hugepages |
fef1bdd6 | 464 | |
db0fb184 PM |
465 | Change the maximum size of the hugepage pool. The maximum is |
466 | nr_hugepages + nr_overcommit_hugepages. | |
fe071d7e | 467 | |
db0fb184 | 468 | See Documentation/vm/hugetlbpage.txt |
fe071d7e | 469 | |
db0fb184 | 470 | ============================================================== |
fe071d7e | 471 | |
db0fb184 | 472 | nr_trim_pages |
ed032189 | 473 | |
db0fb184 PM |
474 | This is available only on NOMMU kernels. |
475 | ||
476 | This value adjusts the excess page trimming behaviour of power-of-2 aligned | |
477 | NOMMU mmap allocations. | |
478 | ||
479 | A value of 0 disables trimming of allocations entirely, while a value of 1 | |
480 | trims excess pages aggressively. Any value >= 1 acts as the watermark where | |
481 | trimming of allocations is initiated. | |
482 | ||
483 | The default value is 1. | |
484 | ||
485 | See Documentation/nommu-mmap.txt for more information. | |
ed032189 | 486 | |
f0c0b2b8 KH |
487 | ============================================================== |
488 | ||
489 | numa_zonelist_order | |
490 | ||
491 | This sysctl is only for NUMA. | |
492 | 'where the memory is allocated from' is controlled by zonelists. | |
493 | (This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation. | |
494 | you may be able to read ZONE_DMA as ZONE_DMA32...) | |
495 | ||
496 | In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following. | |
497 | ZONE_NORMAL -> ZONE_DMA | |
498 | This means that a memory allocation request for GFP_KERNEL will | |
499 | get memory from ZONE_DMA only when ZONE_NORMAL is not available. | |
500 | ||
501 | In NUMA case, you can think of following 2 types of order. | |
502 | Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL | |
503 | ||
504 | (A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL | |
505 | (B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA. | |
506 | ||
507 | Type(A) offers the best locality for processes on Node(0), but ZONE_DMA | |
508 | will be used before ZONE_NORMAL exhaustion. This increases possibility of | |
509 | out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small. | |
510 | ||
511 | Type(B) cannot offer the best locality but is more robust against OOM of | |
512 | the DMA zone. | |
513 | ||
514 | Type(A) is called as "Node" order. Type (B) is "Zone" order. | |
515 | ||
516 | "Node order" orders the zonelists by node, then by zone within each node. | |
5a3016a6 | 517 | Specify "[Nn]ode" for node order |
f0c0b2b8 KH |
518 | |
519 | "Zone Order" orders the zonelists by zone type, then by node within each | |
5a3016a6 | 520 | zone. Specify "[Zz]one" for zone order. |
f0c0b2b8 KH |
521 | |
522 | Specify "[Dd]efault" to request automatic configuration. Autoconfiguration | |
523 | will select "node" order in following case. | |
524 | (1) if the DMA zone does not exist or | |
525 | (2) if the DMA zone comprises greater than 50% of the available memory or | |
f8f191f1 | 526 | (3) if any node's DMA zone comprises greater than 70% of its local memory and |
f0c0b2b8 KH |
527 | the amount of local memory is big enough. |
528 | ||
529 | Otherwise, "zone" order will be selected. Default order is recommended unless | |
530 | this is causing problems for your system/application. | |
d5dbac87 NA |
531 | |
532 | ============================================================== | |
533 | ||
db0fb184 | 534 | oom_dump_tasks |
d5dbac87 | 535 | |
db0fb184 PM |
536 | Enables a system-wide task dump (excluding kernel threads) to be |
537 | produced when the kernel performs an OOM-killing and includes such | |
de34d965 DR |
538 | information as pid, uid, tgid, vm size, rss, nr_ptes, swapents, |
539 | oom_score_adj score, and name. This is helpful to determine why the | |
540 | OOM killer was invoked, to identify the rogue task that caused it, | |
541 | and to determine why the OOM killer chose the task it did to kill. | |
d5dbac87 | 542 | |
db0fb184 PM |
543 | If this is set to zero, this information is suppressed. On very |
544 | large systems with thousands of tasks it may not be feasible to dump | |
545 | the memory state information for each one. Such systems should not | |
546 | be forced to incur a performance penalty in OOM conditions when the | |
547 | information may not be desired. | |
548 | ||
549 | If this is set to non-zero, this information is shown whenever the | |
550 | OOM killer actually kills a memory-hogging task. | |
551 | ||
ad915c43 | 552 | The default value is 1 (enabled). |
d5dbac87 NA |
553 | |
554 | ============================================================== | |
555 | ||
db0fb184 | 556 | oom_kill_allocating_task |
d5dbac87 | 557 | |
db0fb184 PM |
558 | This enables or disables killing the OOM-triggering task in |
559 | out-of-memory situations. | |
d5dbac87 | 560 | |
db0fb184 PM |
561 | If this is set to zero, the OOM killer will scan through the entire |
562 | tasklist and select a task based on heuristics to kill. This normally | |
563 | selects a rogue memory-hogging task that frees up a large amount of | |
564 | memory when killed. | |
565 | ||
566 | If this is set to non-zero, the OOM killer simply kills the task that | |
567 | triggered the out-of-memory condition. This avoids the expensive | |
568 | tasklist scan. | |
569 | ||
570 | If panic_on_oom is selected, it takes precedence over whatever value | |
571 | is used in oom_kill_allocating_task. | |
572 | ||
573 | The default value is 0. | |
dd8632a1 PM |
574 | |
575 | ============================================================== | |
576 | ||
db0fb184 | 577 | overcommit_memory: |
dd8632a1 | 578 | |
db0fb184 | 579 | This value contains a flag that enables memory overcommitment. |
dd8632a1 | 580 | |
db0fb184 PM |
581 | When this flag is 0, the kernel attempts to estimate the amount |
582 | of free memory left when userspace requests more memory. | |
dd8632a1 | 583 | |
db0fb184 PM |
584 | When this flag is 1, the kernel pretends there is always enough |
585 | memory until it actually runs out. | |
dd8632a1 | 586 | |
db0fb184 PM |
587 | When this flag is 2, the kernel uses a "never overcommit" |
588 | policy that attempts to prevent any overcommit of memory. | |
c9b1d098 | 589 | Note that user_reserve_kbytes affects this policy. |
dd8632a1 | 590 | |
db0fb184 PM |
591 | This feature can be very useful because there are a lot of |
592 | programs that malloc() huge amounts of memory "just-in-case" | |
593 | and don't use much of it. | |
594 | ||
595 | The default value is 0. | |
596 | ||
597 | See Documentation/vm/overcommit-accounting and | |
598 | security/commoncap.c::cap_vm_enough_memory() for more information. | |
599 | ||
600 | ============================================================== | |
601 | ||
602 | overcommit_ratio: | |
603 | ||
604 | When overcommit_memory is set to 2, the committed address | |
605 | space is not permitted to exceed swap plus this percentage | |
606 | of physical RAM. See above. | |
607 | ||
608 | ============================================================== | |
609 | ||
610 | page-cluster | |
611 | ||
df858fa8 CE |
612 | page-cluster controls the number of pages up to which consecutive pages |
613 | are read in from swap in a single attempt. This is the swap counterpart | |
614 | to page cache readahead. | |
615 | The mentioned consecutivity is not in terms of virtual/physical addresses, | |
616 | but consecutive on swap space - that means they were swapped out together. | |
db0fb184 PM |
617 | |
618 | It is a logarithmic value - setting it to zero means "1 page", setting | |
619 | it to 1 means "2 pages", setting it to 2 means "4 pages", etc. | |
df858fa8 | 620 | Zero disables swap readahead completely. |
db0fb184 PM |
621 | |
622 | The default value is three (eight pages at a time). There may be some | |
623 | small benefits in tuning this to a different value if your workload is | |
624 | swap-intensive. | |
625 | ||
df858fa8 CE |
626 | Lower values mean lower latencies for initial faults, but at the same time |
627 | extra faults and I/O delays for following faults if they would have been part of | |
628 | that consecutive pages readahead would have brought in. | |
629 | ||
db0fb184 PM |
630 | ============================================================= |
631 | ||
632 | panic_on_oom | |
633 | ||
634 | This enables or disables panic on out-of-memory feature. | |
635 | ||
636 | If this is set to 0, the kernel will kill some rogue process, | |
637 | called oom_killer. Usually, oom_killer can kill rogue processes and | |
638 | system will survive. | |
639 | ||
640 | If this is set to 1, the kernel panics when out-of-memory happens. | |
641 | However, if a process limits using nodes by mempolicy/cpusets, | |
642 | and those nodes become memory exhaustion status, one process | |
643 | may be killed by oom-killer. No panic occurs in this case. | |
644 | Because other nodes' memory may be free. This means system total status | |
645 | may be not fatal yet. | |
646 | ||
647 | If this is set to 2, the kernel panics compulsorily even on the | |
daaf1e68 KH |
648 | above-mentioned. Even oom happens under memory cgroup, the whole |
649 | system panics. | |
db0fb184 PM |
650 | |
651 | The default value is 0. | |
652 | 1 and 2 are for failover of clustering. Please select either | |
653 | according to your policy of failover. | |
daaf1e68 KH |
654 | panic_on_oom=2+kdump gives you very strong tool to investigate |
655 | why oom happens. You can get snapshot. | |
db0fb184 PM |
656 | |
657 | ============================================================= | |
658 | ||
659 | percpu_pagelist_fraction | |
660 | ||
661 | This is the fraction of pages at most (high mark pcp->high) in each zone that | |
662 | are allocated for each per cpu page list. The min value for this is 8. It | |
663 | means that we don't allow more than 1/8th of pages in each zone to be | |
664 | allocated in any single per_cpu_pagelist. This entry only changes the value | |
665 | of hot per cpu pagelists. User can specify a number like 100 to allocate | |
666 | 1/100th of each zone to each per cpu page list. | |
667 | ||
668 | The batch value of each per cpu pagelist is also updated as a result. It is | |
669 | set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8) | |
670 | ||
671 | The initial value is zero. Kernel does not use this value at boot time to set | |
672 | the high water marks for each per cpu page list. | |
673 | ||
674 | ============================================================== | |
675 | ||
676 | stat_interval | |
677 | ||
678 | The time interval between which vm statistics are updated. The default | |
679 | is 1 second. | |
680 | ||
681 | ============================================================== | |
682 | ||
683 | swappiness | |
684 | ||
685 | This control is used to define how aggressive the kernel will swap | |
686 | memory pages. Higher values will increase agressiveness, lower values | |
19f59460 | 687 | decrease the amount of swap. |
db0fb184 PM |
688 | |
689 | The default value is 60. | |
690 | ||
691 | ============================================================== | |
692 | ||
c9b1d098 AS |
693 | - user_reserve_kbytes |
694 | ||
695 | When overcommit_memory is set to 2, "never overommit" mode, reserve | |
696 | min(3% of current process size, user_reserve_kbytes) of free memory. | |
697 | This is intended to prevent a user from starting a single memory hogging | |
698 | process, such that they cannot recover (kill the hog). | |
699 | ||
700 | user_reserve_kbytes defaults to min(3% of the current process size, 128MB). | |
701 | ||
702 | If this is reduced to zero, then the user will be allowed to allocate | |
703 | all free memory with a single process, minus admin_reserve_kbytes. | |
704 | Any subsequent attempts to execute a command will result in | |
705 | "fork: Cannot allocate memory". | |
706 | ||
707 | Changing this takes effect whenever an application requests memory. | |
708 | ||
709 | ============================================================== | |
710 | ||
db0fb184 PM |
711 | vfs_cache_pressure |
712 | ------------------ | |
713 | ||
714 | Controls the tendency of the kernel to reclaim the memory which is used for | |
715 | caching of directory and inode objects. | |
716 | ||
717 | At the default value of vfs_cache_pressure=100 the kernel will attempt to | |
718 | reclaim dentries and inodes at a "fair" rate with respect to pagecache and | |
719 | swapcache reclaim. Decreasing vfs_cache_pressure causes the kernel to prefer | |
55c37a84 JK |
720 | to retain dentry and inode caches. When vfs_cache_pressure=0, the kernel will |
721 | never reclaim dentries and inodes due to memory pressure and this can easily | |
722 | lead to out-of-memory conditions. Increasing vfs_cache_pressure beyond 100 | |
db0fb184 PM |
723 | causes the kernel to prefer to reclaim dentries and inodes. |
724 | ||
725 | ============================================================== | |
726 | ||
727 | zone_reclaim_mode: | |
728 | ||
729 | Zone_reclaim_mode allows someone to set more or less aggressive approaches to | |
730 | reclaim memory when a zone runs out of memory. If it is set to zero then no | |
731 | zone reclaim occurs. Allocations will be satisfied from other zones / nodes | |
732 | in the system. | |
733 | ||
734 | This is value ORed together of | |
735 | ||
736 | 1 = Zone reclaim on | |
737 | 2 = Zone reclaim writes dirty pages out | |
738 | 4 = Zone reclaim swaps pages | |
739 | ||
740 | zone_reclaim_mode is set during bootup to 1 if it is determined that pages | |
741 | from remote zones will cause a measurable performance reduction. The | |
742 | page allocator will then reclaim easily reusable pages (those page | |
743 | cache pages that are currently not used) before allocating off node pages. | |
744 | ||
745 | It may be beneficial to switch off zone reclaim if the system is | |
746 | used for a file server and all of memory should be used for caching files | |
747 | from disk. In that case the caching effect is more important than | |
748 | data locality. | |
749 | ||
750 | Allowing zone reclaim to write out pages stops processes that are | |
751 | writing large amounts of data from dirtying pages on other nodes. Zone | |
752 | reclaim will write out dirty pages if a zone fills up and so effectively | |
753 | throttle the process. This may decrease the performance of a single process | |
754 | since it cannot use all of system memory to buffer the outgoing writes | |
755 | anymore but it preserve the memory on other nodes so that the performance | |
756 | of other processes running on other nodes will not be affected. | |
757 | ||
758 | Allowing regular swap effectively restricts allocations to the local | |
759 | node unless explicitly overridden by memory policies or cpuset | |
760 | configurations. | |
761 | ||
762 | ============ End of Document ================================= |