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 |
6fa3eb70 | 32 | - extra_free_kbytes |
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 | |
2a938b94 | 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 | |
53 | - overcommit_memory | |
54 | - overcommit_ratio | |
55 | - page-cluster | |
56 | - panic_on_oom | |
57 | - percpu_pagelist_fraction | |
58 | - stat_interval | |
59 | - swappiness | |
c9b1d098 | 60 | - user_reserve_kbytes |
db0fb184 PM |
61 | - vfs_cache_pressure |
62 | - zone_reclaim_mode | |
63 | ||
1da177e4 LT |
64 | ============================================================== |
65 | ||
4eeab4f5 AS |
66 | admin_reserve_kbytes |
67 | ||
68 | The amount of free memory in the system that should be reserved for users | |
69 | with the capability cap_sys_admin. | |
70 | ||
71 | admin_reserve_kbytes defaults to min(3% of free pages, 8MB) | |
72 | ||
73 | That should provide enough for the admin to log in and kill a process, | |
74 | if necessary, under the default overcommit 'guess' mode. | |
75 | ||
76 | Systems running under overcommit 'never' should increase this to account | |
77 | for the full Virtual Memory Size of programs used to recover. Otherwise, | |
78 | root may not be able to log in to recover the system. | |
79 | ||
80 | How do you calculate a minimum useful reserve? | |
81 | ||
82 | sshd or login + bash (or some other shell) + top (or ps, kill, etc.) | |
83 | ||
84 | For overcommit 'guess', we can sum resident set sizes (RSS). | |
85 | On x86_64 this is about 8MB. | |
86 | ||
87 | For overcommit 'never', we can take the max of their virtual sizes (VSZ) | |
88 | and add the sum of their RSS. | |
89 | On x86_64 this is about 128MB. | |
90 | ||
91 | Changing this takes effect whenever an application requests memory. | |
92 | ||
93 | ============================================================== | |
94 | ||
db0fb184 | 95 | block_dump |
1da177e4 | 96 | |
db0fb184 PM |
97 | block_dump enables block I/O debugging when set to a nonzero value. More |
98 | information on block I/O debugging is in Documentation/laptops/laptop-mode.txt. | |
1da177e4 LT |
99 | |
100 | ============================================================== | |
101 | ||
76ab0f53 MG |
102 | compact_memory |
103 | ||
104 | Available only when CONFIG_COMPACTION is set. When 1 is written to the file, | |
105 | all zones are compacted such that free memory is available in contiguous | |
106 | blocks where possible. This can be important for example in the allocation of | |
107 | huge pages although processes will also directly compact memory as required. | |
108 | ||
109 | ============================================================== | |
110 | ||
db0fb184 | 111 | dirty_background_bytes |
1da177e4 | 112 | |
6601fac8 AB |
113 | Contains the amount of dirty memory at which the background kernel |
114 | flusher threads will start writeback. | |
1da177e4 | 115 | |
abffc020 AR |
116 | Note: dirty_background_bytes is the counterpart of dirty_background_ratio. Only |
117 | one of them may be specified at a time. When one sysctl is written it is | |
118 | immediately taken into account to evaluate the dirty memory limits and the | |
119 | other appears as 0 when read. | |
1da177e4 | 120 | |
db0fb184 | 121 | ============================================================== |
1da177e4 | 122 | |
db0fb184 | 123 | dirty_background_ratio |
1da177e4 | 124 | |
db0fb184 | 125 | Contains, as a percentage of total system memory, the number of pages at which |
6601fac8 | 126 | the background kernel flusher threads will start writing out dirty data. |
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 | ||
157 | Contains, as a percentage of total system memory, the number of pages at which | |
158 | a process which is generating disk writes will itself start writing out dirty | |
159 | data. | |
1da177e4 LT |
160 | |
161 | ============================================================== | |
162 | ||
db0fb184 | 163 | dirty_writeback_centisecs |
1da177e4 | 164 | |
6601fac8 | 165 | The kernel flusher threads will periodically wake up and write `old' data |
db0fb184 PM |
166 | out to disk. This tunable expresses the interval between those wakeups, in |
167 | 100'ths of a second. | |
1da177e4 | 168 | |
db0fb184 | 169 | Setting this to zero disables periodic writeback altogether. |
1da177e4 LT |
170 | |
171 | ============================================================== | |
172 | ||
db0fb184 | 173 | drop_caches |
1da177e4 | 174 | |
db0fb184 PM |
175 | Writing to this will cause the kernel to drop clean caches, dentries and |
176 | inodes from memory, causing that memory to become free. | |
1da177e4 | 177 | |
db0fb184 PM |
178 | To free pagecache: |
179 | echo 1 > /proc/sys/vm/drop_caches | |
180 | To free dentries and inodes: | |
181 | echo 2 > /proc/sys/vm/drop_caches | |
182 | To free pagecache, dentries and inodes: | |
183 | echo 3 > /proc/sys/vm/drop_caches | |
1da177e4 | 184 | |
db0fb184 PM |
185 | As this is a non-destructive operation and dirty objects are not freeable, the |
186 | user should run `sync' first. | |
1da177e4 LT |
187 | |
188 | ============================================================== | |
189 | ||
5e771905 MG |
190 | extfrag_threshold |
191 | ||
192 | This parameter affects whether the kernel will compact memory or direct | |
193 | reclaim to satisfy a high-order allocation. /proc/extfrag_index shows what | |
194 | the fragmentation index for each order is in each zone in the system. Values | |
195 | tending towards 0 imply allocations would fail due to lack of memory, | |
196 | values towards 1000 imply failures are due to fragmentation and -1 implies | |
197 | that the allocation will succeed as long as watermarks are met. | |
198 | ||
199 | The kernel will not compact memory in a zone if the | |
200 | fragmentation index is <= extfrag_threshold. The default value is 500. | |
201 | ||
202 | ============================================================== | |
203 | ||
6fa3eb70 S |
204 | extra_free_kbytes |
205 | ||
206 | This parameter tells the VM to keep extra free memory between the threshold | |
207 | where background reclaim (kswapd) kicks in, and the threshold where direct | |
208 | reclaim (by allocating processes) kicks in. | |
209 | ||
210 | This is useful for workloads that require low latency memory allocations | |
211 | and have a bounded burstiness in memory allocations, for example a | |
212 | realtime application that receives and transmits network traffic | |
213 | (causing in-kernel memory allocations) with a maximum total message burst | |
214 | size of 200MB may need 200MB of extra free memory to avoid direct reclaim | |
215 | related latencies. | |
216 | ||
217 | ============================================================== | |
218 | ||
db0fb184 | 219 | hugepages_treat_as_movable |
1da177e4 | 220 | |
db0fb184 PM |
221 | This parameter is only useful when kernelcore= is specified at boot time to |
222 | create ZONE_MOVABLE for pages that may be reclaimed or migrated. Huge pages | |
223 | are not movable so are not normally allocated from ZONE_MOVABLE. A non-zero | |
224 | value written to hugepages_treat_as_movable allows huge pages to be allocated | |
225 | from ZONE_MOVABLE. | |
8ad4b1fb | 226 | |
db0fb184 PM |
227 | Once enabled, the ZONE_MOVABLE is treated as an area of memory the huge |
228 | pages pool can easily grow or shrink within. Assuming that applications are | |
229 | not running that mlock() a lot of memory, it is likely the huge pages pool | |
230 | can grow to the size of ZONE_MOVABLE by repeatedly entering the desired value | |
231 | into nr_hugepages and triggering page reclaim. | |
24950898 | 232 | |
8ad4b1fb RS |
233 | ============================================================== |
234 | ||
db0fb184 | 235 | hugetlb_shm_group |
8ad4b1fb | 236 | |
db0fb184 PM |
237 | hugetlb_shm_group contains group id that is allowed to create SysV |
238 | shared memory segment using hugetlb page. | |
8ad4b1fb | 239 | |
db0fb184 | 240 | ============================================================== |
8ad4b1fb | 241 | |
db0fb184 | 242 | laptop_mode |
1743660b | 243 | |
db0fb184 PM |
244 | laptop_mode is a knob that controls "laptop mode". All the things that are |
245 | controlled by this knob are discussed in Documentation/laptops/laptop-mode.txt. | |
1743660b | 246 | |
db0fb184 | 247 | ============================================================== |
1743660b | 248 | |
db0fb184 | 249 | legacy_va_layout |
1b2ffb78 | 250 | |
2174efb6 | 251 | If non-zero, this sysctl disables the new 32-bit mmap layout - the kernel |
db0fb184 | 252 | will use the legacy (2.4) layout for all processes. |
1b2ffb78 | 253 | |
db0fb184 | 254 | ============================================================== |
1b2ffb78 | 255 | |
db0fb184 PM |
256 | lowmem_reserve_ratio |
257 | ||
258 | For some specialised workloads on highmem machines it is dangerous for | |
259 | the kernel to allow process memory to be allocated from the "lowmem" | |
260 | zone. This is because that memory could then be pinned via the mlock() | |
261 | system call, or by unavailability of swapspace. | |
262 | ||
263 | And on large highmem machines this lack of reclaimable lowmem memory | |
264 | can be fatal. | |
265 | ||
266 | So the Linux page allocator has a mechanism which prevents allocations | |
267 | which _could_ use highmem from using too much lowmem. This means that | |
268 | a certain amount of lowmem is defended from the possibility of being | |
269 | captured into pinned user memory. | |
270 | ||
271 | (The same argument applies to the old 16 megabyte ISA DMA region. This | |
272 | mechanism will also defend that region from allocations which could use | |
273 | highmem or lowmem). | |
274 | ||
275 | The `lowmem_reserve_ratio' tunable determines how aggressive the kernel is | |
276 | in defending these lower zones. | |
277 | ||
278 | If you have a machine which uses highmem or ISA DMA and your | |
279 | applications are using mlock(), or if you are running with no swap then | |
280 | you probably should change the lowmem_reserve_ratio setting. | |
281 | ||
282 | The lowmem_reserve_ratio is an array. You can see them by reading this file. | |
283 | - | |
284 | % cat /proc/sys/vm/lowmem_reserve_ratio | |
285 | 256 256 32 | |
286 | - | |
287 | Note: # of this elements is one fewer than number of zones. Because the highest | |
288 | zone's value is not necessary for following calculation. | |
289 | ||
290 | But, these values are not used directly. The kernel calculates # of protection | |
291 | pages for each zones from them. These are shown as array of protection pages | |
292 | in /proc/zoneinfo like followings. (This is an example of x86-64 box). | |
293 | Each zone has an array of protection pages like this. | |
294 | ||
295 | - | |
296 | Node 0, zone DMA | |
297 | pages free 1355 | |
298 | min 3 | |
299 | low 3 | |
300 | high 4 | |
301 | : | |
302 | : | |
303 | numa_other 0 | |
304 | protection: (0, 2004, 2004, 2004) | |
305 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | |
306 | pagesets | |
307 | cpu: 0 pcp: 0 | |
308 | : | |
309 | - | |
310 | These protections are added to score to judge whether this zone should be used | |
311 | for page allocation or should be reclaimed. | |
312 | ||
313 | In this example, if normal pages (index=2) are required to this DMA zone and | |
41858966 MG |
314 | watermark[WMARK_HIGH] is used for watermark, the kernel judges this zone should |
315 | not be used because pages_free(1355) is smaller than watermark + protection[2] | |
db0fb184 PM |
316 | (4 + 2004 = 2008). If this protection value is 0, this zone would be used for |
317 | normal page requirement. If requirement is DMA zone(index=0), protection[0] | |
318 | (=0) is used. | |
319 | ||
320 | zone[i]'s protection[j] is calculated by following expression. | |
321 | ||
322 | (i < j): | |
323 | zone[i]->protection[j] | |
324 | = (total sums of present_pages from zone[i+1] to zone[j] on the node) | |
325 | / lowmem_reserve_ratio[i]; | |
326 | (i = j): | |
327 | (should not be protected. = 0; | |
328 | (i > j): | |
329 | (not necessary, but looks 0) | |
330 | ||
331 | The default values of lowmem_reserve_ratio[i] are | |
332 | 256 (if zone[i] means DMA or DMA32 zone) | |
333 | 32 (others). | |
334 | As above expression, they are reciprocal number of ratio. | |
335 | 256 means 1/256. # of protection pages becomes about "0.39%" of total present | |
336 | pages of higher zones on the node. | |
337 | ||
338 | If you would like to protect more pages, smaller values are effective. | |
339 | The minimum value is 1 (1/1 -> 100%). | |
1b2ffb78 | 340 | |
db0fb184 | 341 | ============================================================== |
1b2ffb78 | 342 | |
db0fb184 | 343 | max_map_count: |
1743660b | 344 | |
db0fb184 PM |
345 | This file contains the maximum number of memory map areas a process |
346 | may have. Memory map areas are used as a side-effect of calling | |
347 | malloc, directly by mmap and mprotect, and also when loading shared | |
348 | libraries. | |
1743660b | 349 | |
db0fb184 PM |
350 | While most applications need less than a thousand maps, certain |
351 | programs, particularly malloc debuggers, may consume lots of them, | |
352 | e.g., up to one or two maps per allocation. | |
fadd8fbd | 353 | |
db0fb184 | 354 | The default value is 65536. |
9614634f | 355 | |
6a46079c AK |
356 | ============================================================= |
357 | ||
358 | memory_failure_early_kill: | |
359 | ||
360 | Control how to kill processes when uncorrected memory error (typically | |
361 | a 2bit error in a memory module) is detected in the background by hardware | |
362 | that cannot be handled by the kernel. In some cases (like the page | |
363 | still having a valid copy on disk) the kernel will handle the failure | |
364 | transparently without affecting any applications. But if there is | |
365 | no other uptodate copy of the data it will kill to prevent any data | |
366 | corruptions from propagating. | |
367 | ||
368 | 1: Kill all processes that have the corrupted and not reloadable page mapped | |
369 | as soon as the corruption is detected. Note this is not supported | |
370 | for a few types of pages, like kernel internally allocated data or | |
371 | the swap cache, but works for the majority of user pages. | |
372 | ||
373 | 0: Only unmap the corrupted page from all processes and only kill a process | |
374 | who tries to access it. | |
375 | ||
376 | The kill is done using a catchable SIGBUS with BUS_MCEERR_AO, so processes can | |
377 | handle this if they want to. | |
378 | ||
379 | This is only active on architectures/platforms with advanced machine | |
380 | check handling and depends on the hardware capabilities. | |
381 | ||
382 | Applications can override this setting individually with the PR_MCE_KILL prctl | |
383 | ||
384 | ============================================================== | |
385 | ||
386 | memory_failure_recovery | |
387 | ||
388 | Enable memory failure recovery (when supported by the platform) | |
389 | ||
390 | 1: Attempt recovery. | |
391 | ||
392 | 0: Always panic on a memory failure. | |
393 | ||
db0fb184 | 394 | ============================================================== |
9614634f | 395 | |
db0fb184 | 396 | min_free_kbytes: |
9614634f | 397 | |
db0fb184 | 398 | This is used to force the Linux VM to keep a minimum number |
41858966 MG |
399 | of kilobytes free. The VM uses this number to compute a |
400 | watermark[WMARK_MIN] value for each lowmem zone in the system. | |
401 | Each lowmem zone gets a number of reserved free pages based | |
402 | proportionally on its size. | |
db0fb184 PM |
403 | |
404 | Some minimal amount of memory is needed to satisfy PF_MEMALLOC | |
405 | allocations; if you set this to lower than 1024KB, your system will | |
406 | become subtly broken, and prone to deadlock under high loads. | |
407 | ||
408 | Setting this too high will OOM your machine instantly. | |
9614634f CL |
409 | |
410 | ============================================================= | |
411 | ||
0ff38490 CL |
412 | min_slab_ratio: |
413 | ||
414 | This is available only on NUMA kernels. | |
415 | ||
416 | A percentage of the total pages in each zone. On Zone reclaim | |
417 | (fallback from the local zone occurs) slabs will be reclaimed if more | |
418 | than this percentage of pages in a zone are reclaimable slab pages. | |
419 | This insures that the slab growth stays under control even in NUMA | |
420 | systems that rarely perform global reclaim. | |
421 | ||
422 | The default is 5 percent. | |
423 | ||
424 | Note that slab reclaim is triggered in a per zone / node fashion. | |
425 | The process of reclaiming slab memory is currently not node specific | |
426 | and may not be fast. | |
427 | ||
428 | ============================================================= | |
429 | ||
db0fb184 | 430 | min_unmapped_ratio: |
fadd8fbd | 431 | |
db0fb184 | 432 | This is available only on NUMA kernels. |
fadd8fbd | 433 | |
90afa5de MG |
434 | This is a percentage of the total pages in each zone. Zone reclaim will |
435 | only occur if more than this percentage of pages are in a state that | |
436 | zone_reclaim_mode allows to be reclaimed. | |
437 | ||
438 | If zone_reclaim_mode has the value 4 OR'd, then the percentage is compared | |
439 | against all file-backed unmapped pages including swapcache pages and tmpfs | |
440 | files. Otherwise, only unmapped pages backed by normal files but not tmpfs | |
441 | files and similar are considered. | |
2b744c01 | 442 | |
db0fb184 | 443 | The default is 1 percent. |
fadd8fbd | 444 | |
db0fb184 | 445 | ============================================================== |
2b744c01 | 446 | |
db0fb184 | 447 | mmap_min_addr |
ed032189 | 448 | |
db0fb184 | 449 | This file indicates the amount of address space which a user process will |
af901ca1 | 450 | be restricted from mmapping. Since kernel null dereference bugs could |
db0fb184 PM |
451 | accidentally operate based on the information in the first couple of pages |
452 | of memory userspace processes should not be allowed to write to them. By | |
453 | default this value is set to 0 and no protections will be enforced by the | |
454 | security module. Setting this value to something like 64k will allow the | |
455 | vast majority of applications to work correctly and provide defense in depth | |
456 | against future potential kernel bugs. | |
fe071d7e | 457 | |
db0fb184 | 458 | ============================================================== |
fef1bdd6 | 459 | |
2a938b94 | 460 | mmap_rnd_bits: |
461 | ||
462 | This value can be used to select the number of bits to use to | |
463 | determine the random offset to the base address of vma regions | |
464 | resulting from mmap allocations on architectures which support | |
465 | tuning address space randomization. This value will be bounded | |
466 | by the architecture's minimum and maximum supported values. | |
467 | ||
468 | This value can be changed after boot using the | |
469 | /proc/sys/vm/mmap_rnd_bits tunable | |
470 | ||
471 | ============================================================== | |
472 | ||
473 | mmap_rnd_compat_bits: | |
474 | ||
475 | This value can be used to select the number of bits to use to | |
476 | determine the random offset to the base address of vma regions | |
477 | resulting from mmap allocations for applications run in | |
478 | compatibility mode on architectures which support tuning address | |
479 | space randomization. This value will be bounded by the | |
480 | architecture's minimum and maximum supported values. | |
481 | ||
482 | This value can be changed after boot using the | |
483 | /proc/sys/vm/mmap_rnd_compat_bits tunable | |
484 | ||
485 | ============================================================== | |
486 | ||
db0fb184 | 487 | nr_hugepages |
fef1bdd6 | 488 | |
db0fb184 | 489 | Change the minimum size of the hugepage pool. |
fef1bdd6 | 490 | |
db0fb184 | 491 | See Documentation/vm/hugetlbpage.txt |
fef1bdd6 | 492 | |
db0fb184 | 493 | ============================================================== |
fef1bdd6 | 494 | |
db0fb184 | 495 | nr_overcommit_hugepages |
fef1bdd6 | 496 | |
db0fb184 PM |
497 | Change the maximum size of the hugepage pool. The maximum is |
498 | nr_hugepages + nr_overcommit_hugepages. | |
fe071d7e | 499 | |
db0fb184 | 500 | See Documentation/vm/hugetlbpage.txt |
fe071d7e | 501 | |
db0fb184 | 502 | ============================================================== |
fe071d7e | 503 | |
db0fb184 | 504 | nr_trim_pages |
ed032189 | 505 | |
db0fb184 PM |
506 | This is available only on NOMMU kernels. |
507 | ||
508 | This value adjusts the excess page trimming behaviour of power-of-2 aligned | |
509 | NOMMU mmap allocations. | |
510 | ||
511 | A value of 0 disables trimming of allocations entirely, while a value of 1 | |
512 | trims excess pages aggressively. Any value >= 1 acts as the watermark where | |
513 | trimming of allocations is initiated. | |
514 | ||
515 | The default value is 1. | |
516 | ||
517 | See Documentation/nommu-mmap.txt for more information. | |
ed032189 | 518 | |
f0c0b2b8 KH |
519 | ============================================================== |
520 | ||
521 | numa_zonelist_order | |
522 | ||
523 | This sysctl is only for NUMA. | |
524 | 'where the memory is allocated from' is controlled by zonelists. | |
525 | (This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation. | |
526 | you may be able to read ZONE_DMA as ZONE_DMA32...) | |
527 | ||
528 | In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following. | |
529 | ZONE_NORMAL -> ZONE_DMA | |
530 | This means that a memory allocation request for GFP_KERNEL will | |
531 | get memory from ZONE_DMA only when ZONE_NORMAL is not available. | |
532 | ||
533 | In NUMA case, you can think of following 2 types of order. | |
534 | Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL | |
535 | ||
536 | (A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL | |
537 | (B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA. | |
538 | ||
539 | Type(A) offers the best locality for processes on Node(0), but ZONE_DMA | |
540 | will be used before ZONE_NORMAL exhaustion. This increases possibility of | |
541 | out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small. | |
542 | ||
543 | Type(B) cannot offer the best locality but is more robust against OOM of | |
544 | the DMA zone. | |
545 | ||
546 | Type(A) is called as "Node" order. Type (B) is "Zone" order. | |
547 | ||
548 | "Node order" orders the zonelists by node, then by zone within each node. | |
5a3016a6 | 549 | Specify "[Nn]ode" for node order |
f0c0b2b8 KH |
550 | |
551 | "Zone Order" orders the zonelists by zone type, then by node within each | |
5a3016a6 | 552 | zone. Specify "[Zz]one" for zone order. |
f0c0b2b8 KH |
553 | |
554 | Specify "[Dd]efault" to request automatic configuration. Autoconfiguration | |
555 | will select "node" order in following case. | |
556 | (1) if the DMA zone does not exist or | |
557 | (2) if the DMA zone comprises greater than 50% of the available memory or | |
558 | (3) if any node's DMA zone comprises greater than 60% of its local memory and | |
559 | the amount of local memory is big enough. | |
560 | ||
561 | Otherwise, "zone" order will be selected. Default order is recommended unless | |
562 | this is causing problems for your system/application. | |
d5dbac87 NA |
563 | |
564 | ============================================================== | |
565 | ||
db0fb184 | 566 | oom_dump_tasks |
d5dbac87 | 567 | |
db0fb184 PM |
568 | Enables a system-wide task dump (excluding kernel threads) to be |
569 | produced when the kernel performs an OOM-killing and includes such | |
de34d965 DR |
570 | information as pid, uid, tgid, vm size, rss, nr_ptes, swapents, |
571 | oom_score_adj score, and name. This is helpful to determine why the | |
572 | OOM killer was invoked, to identify the rogue task that caused it, | |
573 | and to determine why the OOM killer chose the task it did to kill. | |
d5dbac87 | 574 | |
db0fb184 PM |
575 | If this is set to zero, this information is suppressed. On very |
576 | large systems with thousands of tasks it may not be feasible to dump | |
577 | the memory state information for each one. Such systems should not | |
578 | be forced to incur a performance penalty in OOM conditions when the | |
579 | information may not be desired. | |
580 | ||
581 | If this is set to non-zero, this information is shown whenever the | |
582 | OOM killer actually kills a memory-hogging task. | |
583 | ||
ad915c43 | 584 | The default value is 1 (enabled). |
d5dbac87 NA |
585 | |
586 | ============================================================== | |
587 | ||
db0fb184 | 588 | oom_kill_allocating_task |
d5dbac87 | 589 | |
db0fb184 PM |
590 | This enables or disables killing the OOM-triggering task in |
591 | out-of-memory situations. | |
d5dbac87 | 592 | |
db0fb184 PM |
593 | If this is set to zero, the OOM killer will scan through the entire |
594 | tasklist and select a task based on heuristics to kill. This normally | |
595 | selects a rogue memory-hogging task that frees up a large amount of | |
596 | memory when killed. | |
597 | ||
598 | If this is set to non-zero, the OOM killer simply kills the task that | |
599 | triggered the out-of-memory condition. This avoids the expensive | |
600 | tasklist scan. | |
601 | ||
602 | If panic_on_oom is selected, it takes precedence over whatever value | |
603 | is used in oom_kill_allocating_task. | |
604 | ||
605 | The default value is 0. | |
dd8632a1 PM |
606 | |
607 | ============================================================== | |
608 | ||
db0fb184 | 609 | overcommit_memory: |
dd8632a1 | 610 | |
db0fb184 | 611 | This value contains a flag that enables memory overcommitment. |
dd8632a1 | 612 | |
db0fb184 PM |
613 | When this flag is 0, the kernel attempts to estimate the amount |
614 | of free memory left when userspace requests more memory. | |
dd8632a1 | 615 | |
db0fb184 PM |
616 | When this flag is 1, the kernel pretends there is always enough |
617 | memory until it actually runs out. | |
dd8632a1 | 618 | |
db0fb184 PM |
619 | When this flag is 2, the kernel uses a "never overcommit" |
620 | policy that attempts to prevent any overcommit of memory. | |
c9b1d098 | 621 | Note that user_reserve_kbytes affects this policy. |
dd8632a1 | 622 | |
db0fb184 PM |
623 | This feature can be very useful because there are a lot of |
624 | programs that malloc() huge amounts of memory "just-in-case" | |
625 | and don't use much of it. | |
626 | ||
627 | The default value is 0. | |
628 | ||
629 | See Documentation/vm/overcommit-accounting and | |
630 | security/commoncap.c::cap_vm_enough_memory() for more information. | |
631 | ||
632 | ============================================================== | |
633 | ||
634 | overcommit_ratio: | |
635 | ||
636 | When overcommit_memory is set to 2, the committed address | |
637 | space is not permitted to exceed swap plus this percentage | |
638 | of physical RAM. See above. | |
639 | ||
640 | ============================================================== | |
641 | ||
642 | page-cluster | |
643 | ||
df858fa8 CE |
644 | page-cluster controls the number of pages up to which consecutive pages |
645 | are read in from swap in a single attempt. This is the swap counterpart | |
646 | to page cache readahead. | |
647 | The mentioned consecutivity is not in terms of virtual/physical addresses, | |
648 | but consecutive on swap space - that means they were swapped out together. | |
db0fb184 PM |
649 | |
650 | It is a logarithmic value - setting it to zero means "1 page", setting | |
651 | it to 1 means "2 pages", setting it to 2 means "4 pages", etc. | |
df858fa8 | 652 | Zero disables swap readahead completely. |
db0fb184 PM |
653 | |
654 | The default value is three (eight pages at a time). There may be some | |
655 | small benefits in tuning this to a different value if your workload is | |
656 | swap-intensive. | |
657 | ||
df858fa8 CE |
658 | Lower values mean lower latencies for initial faults, but at the same time |
659 | extra faults and I/O delays for following faults if they would have been part of | |
660 | that consecutive pages readahead would have brought in. | |
661 | ||
db0fb184 PM |
662 | ============================================================= |
663 | ||
664 | panic_on_oom | |
665 | ||
666 | This enables or disables panic on out-of-memory feature. | |
667 | ||
668 | If this is set to 0, the kernel will kill some rogue process, | |
669 | called oom_killer. Usually, oom_killer can kill rogue processes and | |
670 | system will survive. | |
671 | ||
672 | If this is set to 1, the kernel panics when out-of-memory happens. | |
673 | However, if a process limits using nodes by mempolicy/cpusets, | |
674 | and those nodes become memory exhaustion status, one process | |
675 | may be killed by oom-killer. No panic occurs in this case. | |
676 | Because other nodes' memory may be free. This means system total status | |
677 | may be not fatal yet. | |
678 | ||
679 | If this is set to 2, the kernel panics compulsorily even on the | |
daaf1e68 KH |
680 | above-mentioned. Even oom happens under memory cgroup, the whole |
681 | system panics. | |
db0fb184 PM |
682 | |
683 | The default value is 0. | |
684 | 1 and 2 are for failover of clustering. Please select either | |
685 | according to your policy of failover. | |
daaf1e68 KH |
686 | panic_on_oom=2+kdump gives you very strong tool to investigate |
687 | why oom happens. You can get snapshot. | |
db0fb184 PM |
688 | |
689 | ============================================================= | |
690 | ||
691 | percpu_pagelist_fraction | |
692 | ||
693 | This is the fraction of pages at most (high mark pcp->high) in each zone that | |
694 | are allocated for each per cpu page list. The min value for this is 8. It | |
695 | means that we don't allow more than 1/8th of pages in each zone to be | |
696 | allocated in any single per_cpu_pagelist. This entry only changes the value | |
697 | of hot per cpu pagelists. User can specify a number like 100 to allocate | |
698 | 1/100th of each zone to each per cpu page list. | |
699 | ||
700 | The batch value of each per cpu pagelist is also updated as a result. It is | |
701 | set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8) | |
702 | ||
703 | The initial value is zero. Kernel does not use this value at boot time to set | |
704 | the high water marks for each per cpu page list. | |
705 | ||
706 | ============================================================== | |
707 | ||
708 | stat_interval | |
709 | ||
710 | The time interval between which vm statistics are updated. The default | |
711 | is 1 second. | |
712 | ||
713 | ============================================================== | |
714 | ||
715 | swappiness | |
716 | ||
717 | This control is used to define how aggressive the kernel will swap | |
718 | memory pages. Higher values will increase agressiveness, lower values | |
19f59460 | 719 | decrease the amount of swap. |
db0fb184 PM |
720 | |
721 | The default value is 60. | |
722 | ||
723 | ============================================================== | |
724 | ||
c9b1d098 AS |
725 | - user_reserve_kbytes |
726 | ||
727 | When overcommit_memory is set to 2, "never overommit" mode, reserve | |
728 | min(3% of current process size, user_reserve_kbytes) of free memory. | |
729 | This is intended to prevent a user from starting a single memory hogging | |
730 | process, such that they cannot recover (kill the hog). | |
731 | ||
732 | user_reserve_kbytes defaults to min(3% of the current process size, 128MB). | |
733 | ||
734 | If this is reduced to zero, then the user will be allowed to allocate | |
735 | all free memory with a single process, minus admin_reserve_kbytes. | |
736 | Any subsequent attempts to execute a command will result in | |
737 | "fork: Cannot allocate memory". | |
738 | ||
739 | Changing this takes effect whenever an application requests memory. | |
740 | ||
741 | ============================================================== | |
742 | ||
db0fb184 PM |
743 | vfs_cache_pressure |
744 | ------------------ | |
745 | ||
746 | Controls the tendency of the kernel to reclaim the memory which is used for | |
747 | caching of directory and inode objects. | |
748 | ||
749 | At the default value of vfs_cache_pressure=100 the kernel will attempt to | |
750 | reclaim dentries and inodes at a "fair" rate with respect to pagecache and | |
751 | swapcache reclaim. Decreasing vfs_cache_pressure causes the kernel to prefer | |
55c37a84 JK |
752 | to retain dentry and inode caches. When vfs_cache_pressure=0, the kernel will |
753 | never reclaim dentries and inodes due to memory pressure and this can easily | |
754 | lead to out-of-memory conditions. Increasing vfs_cache_pressure beyond 100 | |
db0fb184 PM |
755 | causes the kernel to prefer to reclaim dentries and inodes. |
756 | ||
757 | ============================================================== | |
758 | ||
759 | zone_reclaim_mode: | |
760 | ||
761 | Zone_reclaim_mode allows someone to set more or less aggressive approaches to | |
762 | reclaim memory when a zone runs out of memory. If it is set to zero then no | |
763 | zone reclaim occurs. Allocations will be satisfied from other zones / nodes | |
764 | in the system. | |
765 | ||
766 | This is value ORed together of | |
767 | ||
768 | 1 = Zone reclaim on | |
769 | 2 = Zone reclaim writes dirty pages out | |
770 | 4 = Zone reclaim swaps pages | |
771 | ||
772 | zone_reclaim_mode is set during bootup to 1 if it is determined that pages | |
773 | from remote zones will cause a measurable performance reduction. The | |
774 | page allocator will then reclaim easily reusable pages (those page | |
775 | cache pages that are currently not used) before allocating off node pages. | |
776 | ||
777 | It may be beneficial to switch off zone reclaim if the system is | |
778 | used for a file server and all of memory should be used for caching files | |
779 | from disk. In that case the caching effect is more important than | |
780 | data locality. | |
781 | ||
782 | Allowing zone reclaim to write out pages stops processes that are | |
783 | writing large amounts of data from dirtying pages on other nodes. Zone | |
784 | reclaim will write out dirty pages if a zone fills up and so effectively | |
785 | throttle the process. This may decrease the performance of a single process | |
786 | since it cannot use all of system memory to buffer the outgoing writes | |
787 | anymore but it preserve the memory on other nodes so that the performance | |
788 | of other processes running on other nodes will not be affected. | |
789 | ||
790 | Allowing regular swap effectively restricts allocations to the local | |
791 | node unless explicitly overridden by memory policies or cpuset | |
792 | configurations. | |
793 | ||
794 | ============ End of Document ================================= |