admin-guide/sysctl/vm.rst

11 For general info and legal blurb, please look in index.rst.
13 ------------------------------------------------------------------------------
15 This file contains the documentation for the sysctl files in
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
23 files can be found in mm/swap.c.
25 Currently, these files are in /proc/sys/vm:
27 - admin_reserve_kbytes
28 - block_dump
29 - compact_memory
30 - compaction_proactiveness
31 - compact_unevictable_allowed
32 - dirty_background_bytes
33 - dirty_background_ratio
34 - dirty_bytes
35 - dirty_expire_centisecs
36 - dirty_ratio
37 - dirtytime_expire_seconds
38 - dirty_writeback_centisecs
39 - drop_caches
40 - extfrag_threshold
41 - extra_free_kbytes
42 - highmem_is_dirtyable
43 - hugetlb_shm_group
44 - laptop_mode
45 - legacy_va_layout
46 - lowmem_reserve_ratio
47 - max_map_count
48 - memory_failure_early_kill
49 - memory_failure_recovery
50 - min_free_kbytes
51 - min_slab_ratio
52 - min_unmapped_ratio
53 - mmap_min_addr
54 - mmap_rnd_bits
55 - mmap_rnd_compat_bits
56 - nr_hugepages
57 - nr_hugepages_mempolicy
58 - nr_overcommit_hugepages
59 - nr_trim_pages         (only if CONFIG_MMU=n)
60 - numa_zonelist_order
61 - oom_dump_tasks
62 - oom_kill_allocating_task
63 - overcommit_kbytes
64 - overcommit_memory
65 - overcommit_ratio
66 - page-cluster
67 - panic_on_oom
68 - percpu_pagelist_fraction
69 - stat_interval
70 - stat_refresh
71 - numa_stat
72 - swappiness
73 - unprivileged_userfaultfd
74 - user_reserve_kbytes
75 - vfs_cache_pressure
76 - watermark_boost_factor
77 - watermark_scale_factor
78 - zone_reclaim_mode
84 The amount of free memory in the system that should be reserved for users
89 That should provide enough for the admin to log in and kill a process,
93 for the full Virtual Memory Size of programs used to recover. Otherwise,
94 root may not be able to log in to recover the system.
107 Changing this takes effect whenever an application requests memory.
114 information on block I/O debugging is in Documentation/admin-guide/laptops/laptop-mode.rst.
121 all zones are compacted such that free memory is available in contiguous
122 blocks where possible. This can be important for example in the allocation of
123 huge pages although processes will also directly compact memory as required.
128 This tunable takes a value in the range [0, 100] with a default value of
129 20. This tunable determines how aggressively compaction is done in the
133 Note that compaction has a non-trivial system-wide impact as pages
135 to latency spikes in unsuspecting applications. The kernel employs
148 acceptable trade for large contiguous free memory.  Set to 0 to prevent
150 On CONFIG_PREEMPT_RT the default value is 0 in order to avoid a page fault, due
158 Contains the amount of dirty memory at which the background kernel
164   immediately taken into account to evaluate the dirty memory limits and the
171 Contains, as a percentage of total available memory that contains free pages
175 The total available memory is not equal to total system memory.
181 Contains the amount of dirty memory at which a process generating disk writes
186 account to evaluate the dirty memory limits and the other appears as 0 when
189 Note: the minimum value allowed for dirty_bytes is two pages (in bytes); any
198 for writeout by the kernel flusher threads.  It is expressed in 100'ths
199 of a second.  Data which has been dirty in-memory for longer than this
206 Contains, as a percentage of total available memory that contains free pages
210 The total available memory is not equal to total system memory.
229 out to disk.  This tunable expresses the interval between those wakeups, in
240 memory becomes free.
254 This is a non-destructive operation and will not free any dirty objects.
262 reclaimed by the kernel when memory is needed elsewhere on the system.
269 You may see informational messages in your kernel log when this file is
281 This parameter affects whether the kernel will compact memory or direct
282 reclaim to satisfy a high-order allocation. The extfrag/extfrag_index file in
283 debugfs shows what the fragmentation index for each order is in each zone in
285 of memory, values towards 1000 imply failures are due to fragmentation and -1
288 The kernel will not compact memory in a zone if the
297 This parameter controls whether the high memory is considered for dirty
299 only the amount of memory directly visible/usable by the kernel can
300 be dirtied. As a result, on systems with a large amount of memory and
304 Changing the value to non zero would allow more memory to be dirtied
306 storage more effectively. Note this also comes with a risk of pre-mature
308 only use the low memory and they can fill it up with dirty data without
314 This parameter tells the VM to keep extra free memory between the threshold
315 where background reclaim (kswapd) kicks in, and the threshold where direct
316 reclaim (by allocating processes) kicks in.
318 This is useful for workloads that require low latency memory allocations
319 and have a bounded burstiness in memory allocations, for example a
321 (causing in-kernel memory allocations) with a maximum total message burst
322 size of 200MB may need 200MB of extra free memory to avoid direct reclaim
331 shared memory segment using hugetlb page.
338 controlled by this knob are discussed in Documentation/admin-guide/laptops/laptop-mode.rst.
344 If non-zero, this sysctl disables the new 32-bit mmap layout - the kernel
352 the kernel to allow process memory to be allocated from the "lowmem"
353 zone.  This is because that memory could then be pinned via the mlock()
356 And on large highmem machines this lack of reclaimable lowmem memory
362 captured into pinned user memory.
369 in defending these lower zones.
382 in /proc/zoneinfo like followings. (This is an example of x86-64 box).
402 In this example, if normal pages (index=2) are required to this DMA zone and
412     zone[i]->protection[j]
432 The minimum value is 1 (1/1 -> 100%). The value less than 1 completely
439 This file contains the maximum number of memory map areas a process
440 may have. Memory map areas are used as a side-effect of calling
454 Control how to kill processes when uncorrected memory error (typically
455 a 2bit error in a memory module) is detected in the background by hardware
456 that cannot be handled by the kernel. In some cases (like the page
482 Enable memory failure recovery (when supported by the platform)
486 0: Always panic on a memory failure.
494 watermark[WMARK_MIN] value for each lowmem zone in the system.
498 Some minimal amount of memory is needed to satisfy PF_MEMALLOC
510 A percentage of the total pages in each zone.  On Zone reclaim
512 than this percentage of pages in a zone are reclaimable slab pages.
513 This insures that the slab growth stays under control even in NUMA
518 Note that slab reclaim is triggered in a per zone / node fashion.
519 The process of reclaiming slab memory is currently not node specific
528 This is a percentage of the total pages in each zone. Zone reclaim will
529 only occur if more than this percentage of pages are in a state that
533 against all file-backed unmapped pages including swapcache pages and tmpfs
545 accidentally operate based on the information in the first couple of pages
546 of memory userspace processes should not be allowed to write to them.  By
549 vast majority of applications to work correctly and provide defense in depth
571 resulting from mmap allocations for applications run in
585 See Documentation/admin-guide/mm/hugetlbpage.rst
591 Change the size of the hugepage pool at run-time on a specific
594 See Documentation/admin-guide/mm/hugetlbpage.rst
603 See Documentation/admin-guide/mm/hugetlbpage.rst
611 This value adjusts the excess page trimming behaviour of power-of-2 aligned
620 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
629 'where the memory is allocated from' is controlled by zonelists.
634 In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following.
635 ZONE_NORMAL -> ZONE_DMA
636 This means that a memory allocation request for GFP_KERNEL will
637 get memory from ZONE_DMA only when ZONE_NORMAL is not available.
639 In NUMA case, you can think of following 2 types of order.
642   (A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL
643   (B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA.
647 out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small.
662 On 32-bit, the Normal zone needs to be preserved for allocations accessible
665 On 64-bit, devices that require DMA32/DMA are relatively rare, so "node"
675 Enables a system-wide task dump (excluding kernel threads) to be produced
676 when the kernel performs an OOM-killing and includes such information as
684 the memory state information for each one.  Such systems should not
685 be forced to incur a performance penalty in OOM conditions when the
688 If this is set to non-zero, this information is shown whenever the
689 OOM killer actually kills a memory-hogging task.
697 This enables or disables killing the OOM-triggering task in
698 out-of-memory situations.
702 selects a rogue memory-hogging task that frees up a large amount of
703 memory when killed.
705 If this is set to non-zero, the OOM killer simply kills the task that
706 triggered the out-of-memory condition.  This avoids the expensive
710 is used in oom_kill_allocating_task.
729 This value contains a flag that enables memory overcommitment.
732 of free memory left when userspace requests more memory.
735 memory until it actually runs out.
738 policy that attempts to prevent any overcommit of memory.
742 programs that malloc() huge amounts of memory "just-in-case"
747 See Documentation/vm/overcommit-accounting.rst and
759 page-cluster
762 page-cluster controls the number of pages up to which consecutive pages
763 are read in from swap in a single attempt. This is the swap counterpart
765 The mentioned consecutivity is not in terms of virtual/physical addresses,
766 but consecutive on swap space - that means they were swapped out together.
768 It is a logarithmic value - setting it to zero means "1 page", setting
773 small benefits in tuning this to a different value if your workload is
774 swap-intensive.
778 that consecutive pages readahead would have brought in.
784 This enables or disables panic on out-of-memory feature.
790 If this is set to 1, the kernel panics when out-of-memory happens.
792 and those nodes become memory exhaustion status, one process
793 may be killed by oom-killer. No panic occurs in this case.
794 Because other nodes' memory may be free. This means system total status
798 above-mentioned. Even oom happens under memory cgroup, the whole
813 This is the fraction of pages at most (high mark pcp->high) in each zone that
815 means that we don't allow more than 1/8th of pages in each zone to be
816 allocated in any single per_cpu_pagelist.  This entry only changes the value
821 set to pcp->high/4.  The upper limit of batch is (PAGE_SHIFT * 8)
838 Any read or write (by root only) flushes all the per-cpu vm statistics
842 As a side-effect, it also checks for negative totals (elsewhere reported
843 as 0) and "fails" with EINVAL if any are found, with a warning in dmesg.
870 assumes equal IO cost and will thus apply memory pressure to the page
871 cache and swap-backed pages equally; lower values signify more
874 Keep in mind that filesystem IO patterns under memory pressure tend to
876 experimentation and will also be workload-dependent.
880 For in-memory swap, like zram or zswap, as well as hybrid setups that
887 file-backed pages is less than the high watermark in a zone.
893 This flag controls the mode in which unprivileged users can use the
895 to handle page faults in user mode only. In this case, users without
896 SYS_CAP_PTRACE must pass UFFD_USER_MODE_ONLY in order for userfaultfd to
910 min(3% of current process size, user_reserve_kbytes) of free memory.
911 This is intended to prevent a user from starting a single memory hogging
917 all free memory with a single process, minus admin_reserve_kbytes.
918 Any subsequent attempts to execute a command will result in
919 "fork: Cannot allocate memory".
921 Changing this takes effect whenever an application requests memory.
928 the memory which is used for caching of directory and inode objects.
934 never reclaim dentries and inodes due to memory pressure and this can easily
935 lead to out-of-memory conditions. Increasing vfs_cache_pressure beyond 100
947 This factor controls the level of reclaim when memory is being fragmented.
950 The intent is that compaction has less work to do in the future and to
951 increase the success rate of future high-order allocations such as SLUB
955 parameter, the unit is in fractions of 10,000. The default value of
957 watermark will be reclaimed in the event of a pageblock being mixed due
959 fragmentation events that occurred in the recent past. If this value is
961 (e.g.  2MB on 64-bit x86). A boost factor of 0 will disable the feature.
968 amount of memory left in a node/system before kswapd is woken up and
969 how much memory needs to be free before kswapd goes back to sleep.
971 The unit is in fractions of 10,000. The default value of 10 means the
972 distances between watermarks are 0.1% of the available memory in the
973 node/system. The maximum value is 1000, or 10% of memory.
978 too small for the allocation bursts occurring in the system. This knob
986 reclaim memory when a zone runs out of memory. If it is set to zero then no
988 in the system.
1005 and that accessing remote memory would cause a measurable performance
1013 since it cannot use all of system memory to buffer the outgoing writes
1014 anymore but it preserve the memory on other nodes so that the performance
1018 node unless explicitly overridden by memory policies or cpuset