Debian Codegraph @

upstream/debian-trixie/man5/proc_sys_vm.5

raw
1 .\" Copyright (C) 1994, 1995, Daniel Quinlan <[email protected]>
2 .\" Copyright (C) 2002-2008, 2017, Michael Kerrisk <[email protected]>
3 .\" Copyright (C) , Andries Brouwer <[email protected]>
4 .\" Copyright (C) 2023, Alejandro Colomar <[email protected]>
5 .\"
6 .\" SPDX-License-Identifier: GPL-3.0-or-later
7 .\"
8 .TH proc_sys_vm 5 2024-05-02 "Linux man-pages 6.9.1"
9 .SH NAME
10 /proc/sys/vm/ \- virtual memory subsystem
11 .SH DESCRIPTION
12 .TP
13 .I /proc/sys/vm/
14 This directory contains files for memory management tuning, buffer, and
15 cache management.
16 .TP
17 .IR /proc/sys/vm/admin_reserve_kbytes " (since Linux 3.10)"
18 .\" commit 4eeab4f5580d11bffedc697684b91b0bca0d5009
19 This file defines the amount of free memory (in KiB) on the system that
20 should be reserved for users with the capability
21 .BR CAP_SYS_ADMIN .
22 .IP
23 The default value in this file is the minimum of [3% of free pages, 8MiB]
24 expressed as KiB.
25 The default is intended to provide enough for the superuser
26 to log in and kill a process, if necessary,
27 under the default overcommit 'guess' mode (i.e., 0 in
28 .IR /proc/sys/vm/overcommit_memory ).
29 .IP
30 Systems running in "overcommit never" mode (i.e., 2 in
31 .IR /proc/sys/vm/overcommit_memory )
32 should increase the value in this file to account
33 for the full virtual memory size of the programs used to recover (e.g.,
34 .BR login (1)
35 .BR ssh (1),
36 and
37 .BR top (1))
38 Otherwise, the superuser may not be able to log in to recover the system.
39 For example, on x86-64 a suitable value is 131072 (128MiB reserved).
40 .IP
41 Changing the value in this file takes effect whenever
42 an application requests memory.
43 .TP
44 .IR /proc/sys/vm/compact_memory " (since Linux 2.6.35)"
45 When 1 is written to this file, all zones are compacted such that free
46 memory is available in contiguous blocks where possible.
47 The effect of this action can be seen by examining
48 .IR /proc/buddyinfo .
49 .IP
50 Present only if the kernel was configured with
51 .BR CONFIG_COMPACTION .
52 .TP
53 .IR /proc/sys/vm/drop_caches " (since Linux 2.6.16)"
54 Writing to this file causes the kernel to drop clean caches, dentries, and
55 inodes from memory, causing that memory to become free.
56 This can be useful for memory management testing and
57 performing reproducible filesystem benchmarks.
58 Because writing to this file causes the benefits of caching to be lost,
59 it can degrade overall system performance.
60 .IP
61 To free pagecache, use:
62 .IP
63 .in +4n
64 .EX
65 echo 1 > /proc/sys/vm/drop_caches
66 .EE
67 .in
68 .IP
69 To free dentries and inodes, use:
70 .IP
71 .in +4n
72 .EX
73 echo 2 > /proc/sys/vm/drop_caches
74 .EE
75 .in
76 .IP
77 To free pagecache, dentries, and inodes, use:
78 .IP
79 .in +4n
80 .EX
81 echo 3 > /proc/sys/vm/drop_caches
82 .EE
83 .in
84 .IP
85 Because writing to this file is a nondestructive operation and dirty objects
86 are not freeable, the
87 user should run
88 .BR sync (1)
89 first.
90 .TP
91 .IR /proc/sys/vm/sysctl_hugetlb_shm_group " (since Linux 2.6.7)"
92 This writable file contains a group ID that is allowed
93 to allocate memory using huge pages.
94 If a process has a filesystem group ID or any supplementary group ID that
95 matches this group ID,
96 then it can make huge-page allocations without holding the
97 .B CAP_IPC_LOCK
98 capability; see
99 .BR memfd_create (2),
100 .BR mmap (2),
101 and
102 .BR shmget (2).
103 .TP
104 .IR /proc/sys/vm/legacy_va_layout " (since Linux 2.6.9)"
105 .\" The following is from Documentation/filesystems/proc.txt
106 If nonzero, this disables the new 32-bit memory-mapping layout;
107 the kernel will use the legacy (2.4) layout for all processes.
108 .TP
109 .IR /proc/sys/vm/memory_failure_early_kill " (since Linux 2.6.32)"
110 .\" The following is based on the text in Documentation/sysctl/vm.txt
111 Control how to kill processes when an uncorrected memory error
112 (typically a 2-bit error in a memory module)
113 that cannot be handled by the kernel
114 is detected in the background by hardware.
115 In some cases (like the page still having a valid copy on disk),
116 the kernel will handle the failure
117 transparently without affecting any applications.
118 But if there is no other up-to-date copy of the data,
119 it will kill processes to prevent any data corruptions from propagating.
120 .IP
121 The file has one of the following values:
122 .RS
123 .TP
124 .B 1
125 Kill all processes that have the corrupted-and-not-reloadable page mapped
126 as soon as the corruption is detected.
127 Note that this is not supported for a few types of pages,
128 such as kernel internally
129 allocated data or the swap cache, but works for the majority of user pages.
130 .TP
131 .B 0
132 Unmap the corrupted page from all processes and kill a process
133 only if it tries to access the page.
134 .RE
135 .IP
136 The kill is performed using a
137 .B SIGBUS
138 signal with
139 .I si_code
140 set to
141 .BR BUS_MCEERR_AO .
142 Processes can handle this if they want to; see
143 .BR sigaction (2)
144 for more details.
145 .IP
146 This feature is active only on architectures/platforms with advanced machine
147 check handling and depends on the hardware capabilities.
148 .IP
149 Applications can override the
150 .I memory_failure_early_kill
151 setting individually with the
152 .BR prctl (2)
153 .B PR_MCE_KILL
154 operation.
155 .IP
156 Present only if the kernel was configured with
157 .BR CONFIG_MEMORY_FAILURE .
158 .TP
159 .IR /proc/sys/vm/memory_failure_recovery " (since Linux 2.6.32)"
160 .\" The following is based on the text in Documentation/sysctl/vm.txt
161 Enable memory failure recovery (when supported by the platform).
162 .RS
163 .TP
164 .B 1
165 Attempt recovery.
166 .TP
167 .B 0
168 Always panic on a memory failure.
169 .RE
170 .IP
171 Present only if the kernel was configured with
172 .BR CONFIG_MEMORY_FAILURE .
173 .TP
174 .IR /proc/sys/vm/oom_dump_tasks " (since Linux 2.6.25)"
175 .\" The following is from Documentation/sysctl/vm.txt
176 Enables a system-wide task dump (excluding kernel threads) to be
177 produced when the kernel performs an OOM-killing.
178 The dump includes the following information
179 for each task (thread, process):
180 thread ID, real user ID, thread group ID (process ID),
181 virtual memory size, resident set size,
182 the CPU that the task is scheduled on,
183 oom_adj score (see the description of
184 .IR /proc/ pid /oom_adj ),
185 and command name.
186 This is helpful to determine why the OOM-killer was invoked
187 and to identify the rogue task that caused it.
188 .IP
189 If this contains the value zero, this information is suppressed.
190 On very large systems with thousands of tasks,
191 it may not be feasible to dump the memory state information for each one.
192 Such systems should not be forced to incur a performance penalty in
193 OOM situations when the information may not be desired.
194 .IP
195 If this is set to nonzero, this information is shown whenever the
196 OOM-killer actually kills a memory-hogging task.
197 .IP
198 The default value is 0.
199 .TP
200 .IR /proc/sys/vm/oom_kill_allocating_task " (since Linux 2.6.24)"
201 .\" The following is from Documentation/sysctl/vm.txt
202 This enables or disables killing the OOM-triggering task in
203 out-of-memory situations.
204 .IP
205 If this is set to zero, the OOM-killer will scan through the entire
206 tasklist and select a task based on heuristics to kill.
207 This normally selects a rogue memory-hogging task that
208 frees up a large amount of memory when killed.
209 .IP
210 If this is set to nonzero, the OOM-killer simply kills the task that
211 triggered the out-of-memory condition.
212 This avoids a possibly expensive tasklist scan.
213 .IP
214 If
215 .I /proc/sys/vm/panic_on_oom
216 is nonzero, it takes precedence over whatever value is used in
217 .IR /proc/sys/vm/oom_kill_allocating_task .
218 .IP
219 The default value is 0.
220 .TP
221 .IR /proc/sys/vm/overcommit_kbytes " (since Linux 3.14)"
222 .\" commit 49f0ce5f92321cdcf741e35f385669a421013cb7
223 This writable file provides an alternative to
224 .I /proc/sys/vm/overcommit_ratio
225 for controlling the
226 .I CommitLimit
227 when
228 .I /proc/sys/vm/overcommit_memory
229 has the value 2.
230 It allows the amount of memory overcommitting to be specified as
231 an absolute value (in kB),
232 rather than as a percentage, as is done with
233 .IR overcommit_ratio .
234 This allows for finer-grained control of
235 .I CommitLimit
236 on systems with extremely large memory sizes.
237 .IP
238 Only one of
239 .I overcommit_kbytes
240 or
241 .I overcommit_ratio
242 can have an effect:
243 if
244 .I overcommit_kbytes
245 has a nonzero value, then it is used to calculate
246 .IR CommitLimit ,
247 otherwise
248 .I overcommit_ratio
249 is used.
250 Writing a value to either of these files causes the
251 value in the other file to be set to zero.
252 .TP
253 .I /proc/sys/vm/overcommit_memory
254 This file contains the kernel virtual memory accounting mode.
255 Values are:
256 .RS
257 .IP
258 0: heuristic overcommit (this is the default)
259 .br
260 1: always overcommit, never check
261 .br
262 2: always check, never overcommit
263 .RE
264 .IP
265 In mode 0, calls of
266 .BR mmap (2)
267 with
268 .B MAP_NORESERVE
269 are not checked, and the default check is very weak,
270 leading to the risk of getting a process "OOM-killed".
271 .IP
272 In mode 1, the kernel pretends there is always enough memory,
273 until memory actually runs out.
274 One use case for this mode is scientific computing applications
275 that employ large sparse arrays.
276 Before Linux 2.6.0, any nonzero value implies mode 1.
277 .IP
278 In mode 2 (available since Linux 2.6), the total virtual address space
279 that can be allocated
280 .RI ( CommitLimit
281 in
282 .IR /proc/meminfo )
283 is calculated as
284 .IP
285 .in +4n
286 .EX
287 CommitLimit = (total_RAM \- total_huge_TLB) *
288 overcommit_ratio / 100 + total_swap
289 .EE
290 .in
291 .IP
292 where:
293 .RS
294 .IP \[bu] 3
295 .I total_RAM
296 is the total amount of RAM on the system;
297 .IP \[bu]
298 .I total_huge_TLB
299 is the amount of memory set aside for huge pages;
300 .IP \[bu]
301 .I overcommit_ratio
302 is the value in
303 .IR /proc/sys/vm/overcommit_ratio ;
304 and
305 .IP \[bu]
306 .I total_swap
307 is the amount of swap space.
308 .RE
309 .IP
310 For example, on a system with 16 GB of physical RAM, 16 GB
311 of swap, no space dedicated to huge pages, and an
312 .I overcommit_ratio
313 of 50, this formula yields a
314 .I CommitLimit
315 of 24 GB.
316 .IP
317 Since Linux 3.14, if the value in
318 .I /proc/sys/vm/overcommit_kbytes
319 is nonzero, then
320 .I CommitLimit
321 is instead calculated as:
322 .IP
323 .in +4n
324 .EX
325 CommitLimit = overcommit_kbytes + total_swap
326 .EE
327 .in
328 .IP
329 See also the description of
330 .I /proc/sys/vm/admin_reserve_kbytes
331 and
332 .IR /proc/sys/vm/user_reserve_kbytes .
333 .TP
334 .IR /proc/sys/vm/overcommit_ratio " (since Linux 2.6.0)"
335 This writable file defines a percentage by which memory
336 can be overcommitted.
337 The default value in the file is 50.
338 See the description of
339 .IR /proc/sys/vm/overcommit_memory .
340 .TP
341 .IR /proc/sys/vm/panic_on_oom " (since Linux 2.6.18)"
342 .\" The following is adapted from Documentation/sysctl/vm.txt
343 This enables or disables a kernel panic in
344 an out-of-memory situation.
345 .IP
346 If this file is set to the value 0,
347 the kernel's OOM-killer will kill some rogue process.
348 Usually, the OOM-killer is able to kill a rogue process and the
349 system will survive.
350 .IP
351 If this file is set to the value 1,
352 then the kernel normally panics when out-of-memory happens.
353 However, if a process limits allocations to certain nodes
354 using memory policies
355 .RB ( mbind (2)
356 .BR MPOL_BIND )
357 or cpusets
358 .RB ( cpuset (7))
359 and those nodes reach memory exhaustion status,
360 one process may be killed by the OOM-killer.
361 No panic occurs in this case:
362 because other nodes' memory may be free,
363 this means the system as a whole may not have reached
364 an out-of-memory situation yet.
365 .IP
366 If this file is set to the value 2,
367 the kernel always panics when an out-of-memory condition occurs.
368 .IP
369 The default value is 0.
370 1 and 2 are for failover of clustering.
371 Select either according to your policy of failover.
372 .TP
373 .I /proc/sys/vm/swappiness
374 .\" The following is from Documentation/sysctl/vm.txt
375 The value in this file controls how aggressively the kernel will swap
376 memory pages.
377 Higher values increase aggressiveness, lower values
378 decrease aggressiveness.
379 The default value is 60.
380 .TP
381 .IR /proc/sys/vm/user_reserve_kbytes " (since Linux 3.10)"
382 .\" commit c9b1d0981fcce3d9976d7b7a56e4e0503bc610dd
383 Specifies an amount of memory (in KiB) to reserve for user processes.
384 This is intended to prevent a user from starting a single memory hogging
385 process, such that they cannot recover (kill the hog).
386 The value in this file has an effect only when
387 .I /proc/sys/vm/overcommit_memory
388 is set to 2 ("overcommit never" mode).
389 In this case, the system reserves an amount of memory that is the minimum
390 of [3% of current process size,
391 .IR user_reserve_kbytes ].
392 .IP
393 The default value in this file is the minimum of [3% of free pages, 128MiB]
394 expressed as KiB.
395 .IP
396 If the value in this file is set to zero,
397 then a user will be allowed to allocate all free memory with a single process
398 (minus the amount reserved by
399 .IR /proc/sys/vm/admin_reserve_kbytes ).
400 Any subsequent attempts to execute a command will result in
401 "fork: Cannot allocate memory".
402 .IP
403 Changing the value in this file takes effect whenever
404 an application requests memory.
405 .TP
406 .IR /proc/sys/vm/unprivileged_userfaultfd " (since Linux 5.2)"
407 .\" cefdca0a86be517bc390fc4541e3674b8e7803b0
408 This (writable) file exposes a flag that controls whether
409 unprivileged processes are allowed to employ
410 .BR userfaultfd (2).
411 If this file has the value 1, then unprivileged processes may use
412 .BR userfaultfd (2).
413 If this file has the value 0, then only processes that have the
414 .B CAP_SYS_PTRACE
415 capability may employ
416 .BR userfaultfd (2).
417 The default value in this file is 1.
418 .SH SEE ALSO
419 .BR proc (5),
420 .BR proc_sys (5)