1 ===============================
2 Documentation for /proc/sys/vm/
3 ===============================
7 Copyright (c) 1998, 1999, Rik van Riel <riel@nl.linux.org>
9 Copyright (c) 2008 Peter W. Morreale <pmorreale@novell.com>
11 For general info and legal blurb, please look in index.rst.
13 ------------------------------------------------------------------------------
15 This file contains the documentation for the sysctl files in
16 /proc/sys/vm and is valid for Linux kernel version 2.6.29.
18 The files in this directory can be used to tune the operation
19 of the virtual memory (VM) subsystem of the Linux kernel and
20 the writeout of dirty data to disk.
22 Default values and initialization routines for most of these
23 files can be found in mm/swap.c.
25 Currently, these files are in /proc/sys/vm:
27 - admin_reserve_kbytes
29 - compaction_proactiveness
30 - compact_unevictable_allowed
31 - dirty_background_bytes
32 - dirty_background_ratio
34 - dirty_expire_centisecs
36 - dirtytime_expire_seconds
37 - dirty_writeback_centisecs
40 - highmem_is_dirtyable
44 - lowmem_reserve_ratio
46 - memory_failure_early_kill
47 - memory_failure_recovery
53 - mmap_rnd_compat_bits
55 - nr_hugepages_mempolicy
56 - nr_overcommit_hugepages
57 - nr_trim_pages (only if CONFIG_MMU=n)
60 - oom_kill_allocating_task
66 - percpu_pagelist_high_fraction
71 - unprivileged_userfaultfd
74 - watermark_boost_factor
75 - watermark_scale_factor
82 The amount of free memory in the system that should be reserved for users
83 with the capability cap_sys_admin.
85 admin_reserve_kbytes defaults to min(3% of free pages, 8MB)
87 That should provide enough for the admin to log in and kill a process,
88 if necessary, under the default overcommit 'guess' mode.
90 Systems running under overcommit 'never' should increase this to account
91 for the full Virtual Memory Size of programs used to recover. Otherwise,
92 root may not be able to log in to recover the system.
94 How do you calculate a minimum useful reserve?
96 sshd or login + bash (or some other shell) + top (or ps, kill, etc.)
98 For overcommit 'guess', we can sum resident set sizes (RSS).
99 On x86_64 this is about 8MB.
101 For overcommit 'never', we can take the max of their virtual sizes (VSZ)
102 and add the sum of their RSS.
103 On x86_64 this is about 128MB.
105 Changing this takes effect whenever an application requests memory.
111 Available only when CONFIG_COMPACTION is set. When 1 is written to the file,
112 all zones are compacted such that free memory is available in contiguous
113 blocks where possible. This can be important for example in the allocation of
114 huge pages although processes will also directly compact memory as required.
116 compaction_proactiveness
117 ========================
119 This tunable takes a value in the range [0, 100] with a default value of
120 20. This tunable determines how aggressively compaction is done in the
121 background. Setting it to 0 disables proactive compaction.
123 Note that compaction has a non-trivial system-wide impact as pages
124 belonging to different processes are moved around, which could also lead
125 to latency spikes in unsuspecting applications. The kernel employs
126 various heuristics to avoid wasting CPU cycles if it detects that
127 proactive compaction is not being effective.
129 Be careful when setting it to extreme values like 100, as that may
130 cause excessive background compaction activity.
132 compact_unevictable_allowed
133 ===========================
135 Available only when CONFIG_COMPACTION is set. When set to 1, compaction is
136 allowed to examine the unevictable lru (mlocked pages) for pages to compact.
137 This should be used on systems where stalls for minor page faults are an
138 acceptable trade for large contiguous free memory. Set to 0 to prevent
139 compaction from moving pages that are unevictable. Default value is 1.
140 On CONFIG_PREEMPT_RT the default value is 0 in order to avoid a page fault, due
141 to compaction, which would block the task from becoming active until the fault
145 dirty_background_bytes
146 ======================
148 Contains the amount of dirty memory at which the background kernel
149 flusher threads will start writeback.
152 dirty_background_bytes is the counterpart of dirty_background_ratio. Only
153 one of them may be specified at a time. When one sysctl is written it is
154 immediately taken into account to evaluate the dirty memory limits and the
155 other appears as 0 when read.
158 dirty_background_ratio
159 ======================
161 Contains, as a percentage of total available memory that contains free pages
162 and reclaimable pages, the number of pages at which the background kernel
163 flusher threads will start writing out dirty data.
165 The total available memory is not equal to total system memory.
171 Contains the amount of dirty memory at which a process generating disk writes
172 will itself start writeback.
174 Note: dirty_bytes is the counterpart of dirty_ratio. Only one of them may be
175 specified at a time. When one sysctl is written it is immediately taken into
176 account to evaluate the dirty memory limits and the other appears as 0 when
179 Note: the minimum value allowed for dirty_bytes is two pages (in bytes); any
180 value lower than this limit will be ignored and the old configuration will be
184 dirty_expire_centisecs
185 ======================
187 This tunable is used to define when dirty data is old enough to be eligible
188 for writeout by the kernel flusher threads. It is expressed in 100'ths
189 of a second. Data which has been dirty in-memory for longer than this
190 interval will be written out next time a flusher thread wakes up.
196 Contains, as a percentage of total available memory that contains free pages
197 and reclaimable pages, the number of pages at which a process which is
198 generating disk writes will itself start writing out dirty data.
200 The total available memory is not equal to total system memory.
203 dirtytime_expire_seconds
204 ========================
206 When a lazytime inode is constantly having its pages dirtied, the inode with
207 an updated timestamp will never get chance to be written out. And, if the
208 only thing that has happened on the file system is a dirtytime inode caused
209 by an atime update, a worker will be scheduled to make sure that inode
210 eventually gets pushed out to disk. This tunable is used to define when dirty
211 inode is old enough to be eligible for writeback by the kernel flusher threads.
212 And, it is also used as the interval to wakeup dirtytime_writeback thread.
215 dirty_writeback_centisecs
216 =========================
218 The kernel flusher threads will periodically wake up and write `old` data
219 out to disk. This tunable expresses the interval between those wakeups, in
222 Setting this to zero disables periodic writeback altogether.
228 Writing to this will cause the kernel to drop clean caches, as well as
229 reclaimable slab objects like dentries and inodes. Once dropped, their
234 echo 1 > /proc/sys/vm/drop_caches
236 To free reclaimable slab objects (includes dentries and inodes)::
238 echo 2 > /proc/sys/vm/drop_caches
240 To free slab objects and pagecache::
242 echo 3 > /proc/sys/vm/drop_caches
244 This is a non-destructive operation and will not free any dirty objects.
245 To increase the number of objects freed by this operation, the user may run
246 `sync` prior to writing to /proc/sys/vm/drop_caches. This will minimize the
247 number of dirty objects on the system and create more candidates to be
250 This file is not a means to control the growth of the various kernel caches
251 (inodes, dentries, pagecache, etc...) These objects are automatically
252 reclaimed by the kernel when memory is needed elsewhere on the system.
254 Use of this file can cause performance problems. Since it discards cached
255 objects, it may cost a significant amount of I/O and CPU to recreate the
256 dropped objects, especially if they were under heavy use. Because of this,
257 use outside of a testing or debugging environment is not recommended.
259 You may see informational messages in your kernel log when this file is
262 cat (1234): drop_caches: 3
264 These are informational only. They do not mean that anything is wrong
265 with your system. To disable them, echo 4 (bit 2) into drop_caches.
271 This parameter affects whether the kernel will compact memory or direct
272 reclaim to satisfy a high-order allocation. The extfrag/extfrag_index file in
273 debugfs shows what the fragmentation index for each order is in each zone in
274 the system. Values tending towards 0 imply allocations would fail due to lack
275 of memory, values towards 1000 imply failures are due to fragmentation and -1
276 implies that the allocation will succeed as long as watermarks are met.
278 The kernel will not compact memory in a zone if the
279 fragmentation index is <= extfrag_threshold. The default value is 500.
285 Available only for systems with CONFIG_HIGHMEM enabled (32b systems).
287 This parameter controls whether the high memory is considered for dirty
288 writers throttling. This is not the case by default which means that
289 only the amount of memory directly visible/usable by the kernel can
290 be dirtied. As a result, on systems with a large amount of memory and
291 lowmem basically depleted writers might be throttled too early and
292 streaming writes can get very slow.
294 Changing the value to non zero would allow more memory to be dirtied
295 and thus allow writers to write more data which can be flushed to the
296 storage more effectively. Note this also comes with a risk of pre-mature
297 OOM killer because some writers (e.g. direct block device writes) can
298 only use the low memory and they can fill it up with dirty data without
305 hugetlb_shm_group contains group id that is allowed to create SysV
306 shared memory segment using hugetlb page.
312 laptop_mode is a knob that controls "laptop mode". All the things that are
313 controlled by this knob are discussed in Documentation/admin-guide/laptops/laptop-mode.rst.
319 If non-zero, this sysctl disables the new 32-bit mmap layout - the kernel
320 will use the legacy (2.4) layout for all processes.
326 For some specialised workloads on highmem machines it is dangerous for
327 the kernel to allow process memory to be allocated from the "lowmem"
328 zone. This is because that memory could then be pinned via the mlock()
329 system call, or by unavailability of swapspace.
331 And on large highmem machines this lack of reclaimable lowmem memory
334 So the Linux page allocator has a mechanism which prevents allocations
335 which *could* use highmem from using too much lowmem. This means that
336 a certain amount of lowmem is defended from the possibility of being
337 captured into pinned user memory.
339 (The same argument applies to the old 16 megabyte ISA DMA region. This
340 mechanism will also defend that region from allocations which could use
343 The `lowmem_reserve_ratio` tunable determines how aggressive the kernel is
344 in defending these lower zones.
346 If you have a machine which uses highmem or ISA DMA and your
347 applications are using mlock(), or if you are running with no swap then
348 you probably should change the lowmem_reserve_ratio setting.
350 The lowmem_reserve_ratio is an array. You can see them by reading this file::
352 % cat /proc/sys/vm/lowmem_reserve_ratio
355 But, these values are not used directly. The kernel calculates # of protection
356 pages for each zones from them. These are shown as array of protection pages
357 in /proc/zoneinfo like followings. (This is an example of x86-64 box).
358 Each zone has an array of protection pages like this::
368 protection: (0, 2004, 2004, 2004)
369 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
374 These protections are added to score to judge whether this zone should be used
375 for page allocation or should be reclaimed.
377 In this example, if normal pages (index=2) are required to this DMA zone and
378 watermark[WMARK_HIGH] is used for watermark, the kernel judges this zone should
379 not be used because pages_free(1355) is smaller than watermark + protection[2]
380 (4 + 2004 = 2008). If this protection value is 0, this zone would be used for
381 normal page requirement. If requirement is DMA zone(index=0), protection[0]
384 zone[i]'s protection[j] is calculated by following expression::
387 zone[i]->protection[j]
388 = (total sums of managed_pages from zone[i+1] to zone[j] on the node)
389 / lowmem_reserve_ratio[i];
391 (should not be protected. = 0;
393 (not necessary, but looks 0)
395 The default values of lowmem_reserve_ratio[i] are
397 === ====================================
398 256 (if zone[i] means DMA or DMA32 zone)
400 === ====================================
402 As above expression, they are reciprocal number of ratio.
403 256 means 1/256. # of protection pages becomes about "0.39%" of total managed
404 pages of higher zones on the node.
406 If you would like to protect more pages, smaller values are effective.
407 The minimum value is 1 (1/1 -> 100%). The value less than 1 completely
408 disables protection of the pages.
414 This file contains the maximum number of memory map areas a process
415 may have. Memory map areas are used as a side-effect of calling
416 malloc, directly by mmap, mprotect, and madvise, and also when loading
419 While most applications need less than a thousand maps, certain
420 programs, particularly malloc debuggers, may consume lots of them,
421 e.g., up to one or two maps per allocation.
423 The default value is 65530.
426 memory_failure_early_kill:
427 ==========================
429 Control how to kill processes when uncorrected memory error (typically
430 a 2bit error in a memory module) is detected in the background by hardware
431 that cannot be handled by the kernel. In some cases (like the page
432 still having a valid copy on disk) the kernel will handle the failure
433 transparently without affecting any applications. But if there is
434 no other uptodate copy of the data it will kill to prevent any data
435 corruptions from propagating.
437 1: Kill all processes that have the corrupted and not reloadable page mapped
438 as soon as the corruption is detected. Note this is not supported
439 for a few types of pages, like kernel internally allocated data or
440 the swap cache, but works for the majority of user pages.
442 0: Only unmap the corrupted page from all processes and only kill a process
443 who tries to access it.
445 The kill is done using a catchable SIGBUS with BUS_MCEERR_AO, so processes can
446 handle this if they want to.
448 This is only active on architectures/platforms with advanced machine
449 check handling and depends on the hardware capabilities.
451 Applications can override this setting individually with the PR_MCE_KILL prctl
454 memory_failure_recovery
455 =======================
457 Enable memory failure recovery (when supported by the platform)
461 0: Always panic on a memory failure.
467 This is used to force the Linux VM to keep a minimum number
468 of kilobytes free. The VM uses this number to compute a
469 watermark[WMARK_MIN] value for each lowmem zone in the system.
470 Each lowmem zone gets a number of reserved free pages based
471 proportionally on its size.
473 Some minimal amount of memory is needed to satisfy PF_MEMALLOC
474 allocations; if you set this to lower than 1024KB, your system will
475 become subtly broken, and prone to deadlock under high loads.
477 Setting this too high will OOM your machine instantly.
483 This is available only on NUMA kernels.
485 A percentage of the total pages in each zone. On Zone reclaim
486 (fallback from the local zone occurs) slabs will be reclaimed if more
487 than this percentage of pages in a zone are reclaimable slab pages.
488 This insures that the slab growth stays under control even in NUMA
489 systems that rarely perform global reclaim.
491 The default is 5 percent.
493 Note that slab reclaim is triggered in a per zone / node fashion.
494 The process of reclaiming slab memory is currently not node specific
501 This is available only on NUMA kernels.
503 This is a percentage of the total pages in each zone. Zone reclaim will
504 only occur if more than this percentage of pages are in a state that
505 zone_reclaim_mode allows to be reclaimed.
507 If zone_reclaim_mode has the value 4 OR'd, then the percentage is compared
508 against all file-backed unmapped pages including swapcache pages and tmpfs
509 files. Otherwise, only unmapped pages backed by normal files but not tmpfs
510 files and similar are considered.
512 The default is 1 percent.
518 This file indicates the amount of address space which a user process will
519 be restricted from mmapping. Since kernel null dereference bugs could
520 accidentally operate based on the information in the first couple of pages
521 of memory userspace processes should not be allowed to write to them. By
522 default this value is set to 0 and no protections will be enforced by the
523 security module. Setting this value to something like 64k will allow the
524 vast majority of applications to work correctly and provide defense in depth
525 against future potential kernel bugs.
531 This value can be used to select the number of bits to use to
532 determine the random offset to the base address of vma regions
533 resulting from mmap allocations on architectures which support
534 tuning address space randomization. This value will be bounded
535 by the architecture's minimum and maximum supported values.
537 This value can be changed after boot using the
538 /proc/sys/vm/mmap_rnd_bits tunable
544 This value can be used to select the number of bits to use to
545 determine the random offset to the base address of vma regions
546 resulting from mmap allocations for applications run in
547 compatibility mode on architectures which support tuning address
548 space randomization. This value will be bounded by the
549 architecture's minimum and maximum supported values.
551 This value can be changed after boot using the
552 /proc/sys/vm/mmap_rnd_compat_bits tunable
558 Change the minimum size of the hugepage pool.
560 See Documentation/admin-guide/mm/hugetlbpage.rst
563 nr_hugepages_mempolicy
564 ======================
566 Change the size of the hugepage pool at run-time on a specific
569 See Documentation/admin-guide/mm/hugetlbpage.rst
572 nr_overcommit_hugepages
573 =======================
575 Change the maximum size of the hugepage pool. The maximum is
576 nr_hugepages + nr_overcommit_hugepages.
578 See Documentation/admin-guide/mm/hugetlbpage.rst
584 This is available only on NOMMU kernels.
586 This value adjusts the excess page trimming behaviour of power-of-2 aligned
587 NOMMU mmap allocations.
589 A value of 0 disables trimming of allocations entirely, while a value of 1
590 trims excess pages aggressively. Any value >= 1 acts as the watermark where
591 trimming of allocations is initiated.
593 The default value is 1.
595 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
601 This sysctl is only for NUMA and it is deprecated. Anything but
602 Node order will fail!
604 'where the memory is allocated from' is controlled by zonelists.
606 (This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation.
607 you may be able to read ZONE_DMA as ZONE_DMA32...)
609 In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following.
610 ZONE_NORMAL -> ZONE_DMA
611 This means that a memory allocation request for GFP_KERNEL will
612 get memory from ZONE_DMA only when ZONE_NORMAL is not available.
614 In NUMA case, you can think of following 2 types of order.
615 Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL::
617 (A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL
618 (B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA.
620 Type(A) offers the best locality for processes on Node(0), but ZONE_DMA
621 will be used before ZONE_NORMAL exhaustion. This increases possibility of
622 out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small.
624 Type(B) cannot offer the best locality but is more robust against OOM of
627 Type(A) is called as "Node" order. Type (B) is "Zone" order.
629 "Node order" orders the zonelists by node, then by zone within each node.
630 Specify "[Nn]ode" for node order
632 "Zone Order" orders the zonelists by zone type, then by node within each
633 zone. Specify "[Zz]one" for zone order.
635 Specify "[Dd]efault" to request automatic configuration.
637 On 32-bit, the Normal zone needs to be preserved for allocations accessible
638 by the kernel, so "zone" order will be selected.
640 On 64-bit, devices that require DMA32/DMA are relatively rare, so "node"
641 order will be selected.
643 Default order is recommended unless this is causing problems for your
650 Enables a system-wide task dump (excluding kernel threads) to be produced
651 when the kernel performs an OOM-killing and includes such information as
652 pid, uid, tgid, vm size, rss, pgtables_bytes, swapents, oom_score_adj
653 score, and name. This is helpful to determine why the OOM killer was
654 invoked, to identify the rogue task that caused it, and to determine why
655 the OOM killer chose the task it did to kill.
657 If this is set to zero, this information is suppressed. On very
658 large systems with thousands of tasks it may not be feasible to dump
659 the memory state information for each one. Such systems should not
660 be forced to incur a performance penalty in OOM conditions when the
661 information may not be desired.
663 If this is set to non-zero, this information is shown whenever the
664 OOM killer actually kills a memory-hogging task.
666 The default value is 1 (enabled).
669 oom_kill_allocating_task
670 ========================
672 This enables or disables killing the OOM-triggering task in
673 out-of-memory situations.
675 If this is set to zero, the OOM killer will scan through the entire
676 tasklist and select a task based on heuristics to kill. This normally
677 selects a rogue memory-hogging task that frees up a large amount of
680 If this is set to non-zero, the OOM killer simply kills the task that
681 triggered the out-of-memory condition. This avoids the expensive
684 If panic_on_oom is selected, it takes precedence over whatever value
685 is used in oom_kill_allocating_task.
687 The default value is 0.
693 When overcommit_memory is set to 2, the committed address space is not
694 permitted to exceed swap plus this amount of physical RAM. See below.
696 Note: overcommit_kbytes is the counterpart of overcommit_ratio. Only one
697 of them may be specified at a time. Setting one disables the other (which
698 then appears as 0 when read).
704 This value contains a flag that enables memory overcommitment.
706 When this flag is 0, the kernel attempts to estimate the amount
707 of free memory left when userspace requests more memory.
709 When this flag is 1, the kernel pretends there is always enough
710 memory until it actually runs out.
712 When this flag is 2, the kernel uses a "never overcommit"
713 policy that attempts to prevent any overcommit of memory.
714 Note that user_reserve_kbytes affects this policy.
716 This feature can be very useful because there are a lot of
717 programs that malloc() huge amounts of memory "just-in-case"
718 and don't use much of it.
720 The default value is 0.
722 See Documentation/vm/overcommit-accounting.rst and
723 mm/util.c::__vm_enough_memory() for more information.
729 When overcommit_memory is set to 2, the committed address
730 space is not permitted to exceed swap plus this percentage
731 of physical RAM. See above.
737 page-cluster controls the number of pages up to which consecutive pages
738 are read in from swap in a single attempt. This is the swap counterpart
739 to page cache readahead.
740 The mentioned consecutivity is not in terms of virtual/physical addresses,
741 but consecutive on swap space - that means they were swapped out together.
743 It is a logarithmic value - setting it to zero means "1 page", setting
744 it to 1 means "2 pages", setting it to 2 means "4 pages", etc.
745 Zero disables swap readahead completely.
747 The default value is three (eight pages at a time). There may be some
748 small benefits in tuning this to a different value if your workload is
751 Lower values mean lower latencies for initial faults, but at the same time
752 extra faults and I/O delays for following faults if they would have been part of
753 that consecutive pages readahead would have brought in.
759 This enables or disables panic on out-of-memory feature.
761 If this is set to 0, the kernel will kill some rogue process,
762 called oom_killer. Usually, oom_killer can kill rogue processes and
765 If this is set to 1, the kernel panics when out-of-memory happens.
766 However, if a process limits using nodes by mempolicy/cpusets,
767 and those nodes become memory exhaustion status, one process
768 may be killed by oom-killer. No panic occurs in this case.
769 Because other nodes' memory may be free. This means system total status
770 may be not fatal yet.
772 If this is set to 2, the kernel panics compulsorily even on the
773 above-mentioned. Even oom happens under memory cgroup, the whole
776 The default value is 0.
778 1 and 2 are for failover of clustering. Please select either
779 according to your policy of failover.
781 panic_on_oom=2+kdump gives you very strong tool to investigate
782 why oom happens. You can get snapshot.
785 percpu_pagelist_high_fraction
786 =============================
788 This is the fraction of pages in each zone that are can be stored to
789 per-cpu page lists. It is an upper boundary that is divided depending
790 on the number of online CPUs. The min value for this is 8 which means
791 that we do not allow more than 1/8th of pages in each zone to be stored
792 on per-cpu page lists. This entry only changes the value of hot per-cpu
793 page lists. A user can specify a number like 100 to allocate 1/100th of
794 each zone between per-cpu lists.
796 The batch value of each per-cpu page list remains the same regardless of
797 the value of the high fraction so allocation latencies are unaffected.
799 The initial value is zero. Kernel uses this value to set the high pcp->high
800 mark based on the low watermark for the zone and the number of local
801 online CPUs. If the user writes '0' to this sysctl, it will revert to
802 this default behavior.
808 The time interval between which vm statistics are updated. The default
815 Any read or write (by root only) flushes all the per-cpu vm statistics
816 into their global totals, for more accurate reports when testing
817 e.g. cat /proc/sys/vm/stat_refresh /proc/meminfo
819 As a side-effect, it also checks for negative totals (elsewhere reported
820 as 0) and "fails" with EINVAL if any are found, with a warning in dmesg.
821 (At time of writing, a few stats are known sometimes to be found negative,
822 with no ill effects: errors and warnings on these stats are suppressed.)
828 This interface allows runtime configuration of numa statistics.
830 When page allocation performance becomes a bottleneck and you can tolerate
831 some possible tool breakage and decreased numa counter precision, you can
834 echo 0 > /proc/sys/vm/numa_stat
836 When page allocation performance is not a bottleneck and you want all
837 tooling to work, you can do::
839 echo 1 > /proc/sys/vm/numa_stat
845 This control is used to define the rough relative IO cost of swapping
846 and filesystem paging, as a value between 0 and 200. At 100, the VM
847 assumes equal IO cost and will thus apply memory pressure to the page
848 cache and swap-backed pages equally; lower values signify more
849 expensive swap IO, higher values indicates cheaper.
851 Keep in mind that filesystem IO patterns under memory pressure tend to
852 be more efficient than swap's random IO. An optimal value will require
853 experimentation and will also be workload-dependent.
855 The default value is 60.
857 For in-memory swap, like zram or zswap, as well as hybrid setups that
858 have swap on faster devices than the filesystem, values beyond 100 can
859 be considered. For example, if the random IO against the swap device
860 is on average 2x faster than IO from the filesystem, swappiness should
861 be 133 (x + 2x = 200, 2x = 133.33).
863 At 0, the kernel will not initiate swap until the amount of free and
864 file-backed pages is less than the high watermark in a zone.
867 unprivileged_userfaultfd
868 ========================
870 This flag controls the mode in which unprivileged users can use the
871 userfaultfd system calls. Set this to 0 to restrict unprivileged users
872 to handle page faults in user mode only. In this case, users without
873 SYS_CAP_PTRACE must pass UFFD_USER_MODE_ONLY in order for userfaultfd to
874 succeed. Prohibiting use of userfaultfd for handling faults from kernel
875 mode may make certain vulnerabilities more difficult to exploit.
877 Set this to 1 to allow unprivileged users to use the userfaultfd system
878 calls without any restrictions.
880 The default value is 0.
886 When overcommit_memory is set to 2, "never overcommit" mode, reserve
887 min(3% of current process size, user_reserve_kbytes) of free memory.
888 This is intended to prevent a user from starting a single memory hogging
889 process, such that they cannot recover (kill the hog).
891 user_reserve_kbytes defaults to min(3% of the current process size, 128MB).
893 If this is reduced to zero, then the user will be allowed to allocate
894 all free memory with a single process, minus admin_reserve_kbytes.
895 Any subsequent attempts to execute a command will result in
896 "fork: Cannot allocate memory".
898 Changing this takes effect whenever an application requests memory.
904 This percentage value controls the tendency of the kernel to reclaim
905 the memory which is used for caching of directory and inode objects.
907 At the default value of vfs_cache_pressure=100 the kernel will attempt to
908 reclaim dentries and inodes at a "fair" rate with respect to pagecache and
909 swapcache reclaim. Decreasing vfs_cache_pressure causes the kernel to prefer
910 to retain dentry and inode caches. When vfs_cache_pressure=0, the kernel will
911 never reclaim dentries and inodes due to memory pressure and this can easily
912 lead to out-of-memory conditions. Increasing vfs_cache_pressure beyond 100
913 causes the kernel to prefer to reclaim dentries and inodes.
915 Increasing vfs_cache_pressure significantly beyond 100 may have negative
916 performance impact. Reclaim code needs to take various locks to find freeable
917 directory and inode objects. With vfs_cache_pressure=1000, it will look for
918 ten times more freeable objects than there are.
921 watermark_boost_factor
922 ======================
924 This factor controls the level of reclaim when memory is being fragmented.
925 It defines the percentage of the high watermark of a zone that will be
926 reclaimed if pages of different mobility are being mixed within pageblocks.
927 The intent is that compaction has less work to do in the future and to
928 increase the success rate of future high-order allocations such as SLUB
929 allocations, THP and hugetlbfs pages.
931 To make it sensible with respect to the watermark_scale_factor
932 parameter, the unit is in fractions of 10,000. The default value of
933 15,000 means that up to 150% of the high watermark will be reclaimed in the
934 event of a pageblock being mixed due to fragmentation. The level of reclaim
935 is determined by the number of fragmentation events that occurred in the
936 recent past. If this value is smaller than a pageblock then a pageblocks
937 worth of pages will be reclaimed (e.g. 2MB on 64-bit x86). A boost factor
938 of 0 will disable the feature.
941 watermark_scale_factor
942 ======================
944 This factor controls the aggressiveness of kswapd. It defines the
945 amount of memory left in a node/system before kswapd is woken up and
946 how much memory needs to be free before kswapd goes back to sleep.
948 The unit is in fractions of 10,000. The default value of 10 means the
949 distances between watermarks are 0.1% of the available memory in the
950 node/system. The maximum value is 1000, or 10% of memory.
952 A high rate of threads entering direct reclaim (allocstall) or kswapd
953 going to sleep prematurely (kswapd_low_wmark_hit_quickly) can indicate
954 that the number of free pages kswapd maintains for latency reasons is
955 too small for the allocation bursts occurring in the system. This knob
956 can then be used to tune kswapd aggressiveness accordingly.
962 Zone_reclaim_mode allows someone to set more or less aggressive approaches to
963 reclaim memory when a zone runs out of memory. If it is set to zero then no
964 zone reclaim occurs. Allocations will be satisfied from other zones / nodes
967 This is value OR'ed together of
969 = ===================================
971 2 Zone reclaim writes dirty pages out
972 4 Zone reclaim swaps pages
973 = ===================================
975 zone_reclaim_mode is disabled by default. For file servers or workloads
976 that benefit from having their data cached, zone_reclaim_mode should be
977 left disabled as the caching effect is likely to be more important than
980 Consider enabling one or more zone_reclaim mode bits if it's known that the
981 workload is partitioned such that each partition fits within a NUMA node
982 and that accessing remote memory would cause a measurable performance
983 reduction. The page allocator will take additional actions before
984 allocating off node pages.
986 Allowing zone reclaim to write out pages stops processes that are
987 writing large amounts of data from dirtying pages on other nodes. Zone
988 reclaim will write out dirty pages if a zone fills up and so effectively
989 throttle the process. This may decrease the performance of a single process
990 since it cannot use all of system memory to buffer the outgoing writes
991 anymore but it preserve the memory on other nodes so that the performance
992 of other processes running on other nodes will not be affected.
994 Allowing regular swap effectively restricts allocations to the local
995 node unless explicitly overridden by memory policies or cpuset