1 |
3 |
xianfeng |
Documentation for /proc/sys/vm/* kernel version 2.2.10
|
2 |
|
|
(c) 1998, 1999, Rik van Riel
|
3 |
|
|
|
4 |
|
|
For general info and legal blurb, please look in README.
|
5 |
|
|
|
6 |
|
|
==============================================================
|
7 |
|
|
|
8 |
|
|
This file contains the documentation for the sysctl files in
|
9 |
|
|
/proc/sys/vm and is valid for Linux kernel version 2.2.
|
10 |
|
|
|
11 |
|
|
The files in this directory can be used to tune the operation
|
12 |
|
|
of the virtual memory (VM) subsystem of the Linux kernel and
|
13 |
|
|
the writeout of dirty data to disk.
|
14 |
|
|
|
15 |
|
|
Default values and initialization routines for most of these
|
16 |
|
|
files can be found in mm/swap.c.
|
17 |
|
|
|
18 |
|
|
Currently, these files are in /proc/sys/vm:
|
19 |
|
|
- overcommit_memory
|
20 |
|
|
- page-cluster
|
21 |
|
|
- dirty_ratio
|
22 |
|
|
- dirty_background_ratio
|
23 |
|
|
- dirty_expire_centisecs
|
24 |
|
|
- dirty_writeback_centisecs
|
25 |
|
|
- max_map_count
|
26 |
|
|
- min_free_kbytes
|
27 |
|
|
- laptop_mode
|
28 |
|
|
- block_dump
|
29 |
|
|
- drop-caches
|
30 |
|
|
- zone_reclaim_mode
|
31 |
|
|
- min_unmapped_ratio
|
32 |
|
|
- min_slab_ratio
|
33 |
|
|
- panic_on_oom
|
34 |
|
|
- oom_kill_allocating_task
|
35 |
|
|
- mmap_min_address
|
36 |
|
|
- numa_zonelist_order
|
37 |
|
|
- nr_hugepages
|
38 |
|
|
- nr_overcommit_hugepages
|
39 |
|
|
|
40 |
|
|
==============================================================
|
41 |
|
|
|
42 |
|
|
dirty_ratio, dirty_background_ratio, dirty_expire_centisecs,
|
43 |
|
|
dirty_writeback_centisecs, vfs_cache_pressure, laptop_mode,
|
44 |
|
|
block_dump, swap_token_timeout, drop-caches,
|
45 |
|
|
hugepages_treat_as_movable:
|
46 |
|
|
|
47 |
|
|
See Documentation/filesystems/proc.txt
|
48 |
|
|
|
49 |
|
|
==============================================================
|
50 |
|
|
|
51 |
|
|
overcommit_memory:
|
52 |
|
|
|
53 |
|
|
This value contains a flag that enables memory overcommitment.
|
54 |
|
|
|
55 |
|
|
When this flag is 0, the kernel attempts to estimate the amount
|
56 |
|
|
of free memory left when userspace requests more memory.
|
57 |
|
|
|
58 |
|
|
When this flag is 1, the kernel pretends there is always enough
|
59 |
|
|
memory until it actually runs out.
|
60 |
|
|
|
61 |
|
|
When this flag is 2, the kernel uses a "never overcommit"
|
62 |
|
|
policy that attempts to prevent any overcommit of memory.
|
63 |
|
|
|
64 |
|
|
This feature can be very useful because there are a lot of
|
65 |
|
|
programs that malloc() huge amounts of memory "just-in-case"
|
66 |
|
|
and don't use much of it.
|
67 |
|
|
|
68 |
|
|
The default value is 0.
|
69 |
|
|
|
70 |
|
|
See Documentation/vm/overcommit-accounting and
|
71 |
|
|
security/commoncap.c::cap_vm_enough_memory() for more information.
|
72 |
|
|
|
73 |
|
|
==============================================================
|
74 |
|
|
|
75 |
|
|
overcommit_ratio:
|
76 |
|
|
|
77 |
|
|
When overcommit_memory is set to 2, the committed address
|
78 |
|
|
space is not permitted to exceed swap plus this percentage
|
79 |
|
|
of physical RAM. See above.
|
80 |
|
|
|
81 |
|
|
==============================================================
|
82 |
|
|
|
83 |
|
|
page-cluster:
|
84 |
|
|
|
85 |
|
|
The Linux VM subsystem avoids excessive disk seeks by reading
|
86 |
|
|
multiple pages on a page fault. The number of pages it reads
|
87 |
|
|
is dependent on the amount of memory in your machine.
|
88 |
|
|
|
89 |
|
|
The number of pages the kernel reads in at once is equal to
|
90 |
|
|
2 ^ page-cluster. Values above 2 ^ 5 don't make much sense
|
91 |
|
|
for swap because we only cluster swap data in 32-page groups.
|
92 |
|
|
|
93 |
|
|
==============================================================
|
94 |
|
|
|
95 |
|
|
max_map_count:
|
96 |
|
|
|
97 |
|
|
This file contains the maximum number of memory map areas a process
|
98 |
|
|
may have. Memory map areas are used as a side-effect of calling
|
99 |
|
|
malloc, directly by mmap and mprotect, and also when loading shared
|
100 |
|
|
libraries.
|
101 |
|
|
|
102 |
|
|
While most applications need less than a thousand maps, certain
|
103 |
|
|
programs, particularly malloc debuggers, may consume lots of them,
|
104 |
|
|
e.g., up to one or two maps per allocation.
|
105 |
|
|
|
106 |
|
|
The default value is 65536.
|
107 |
|
|
|
108 |
|
|
==============================================================
|
109 |
|
|
|
110 |
|
|
min_free_kbytes:
|
111 |
|
|
|
112 |
|
|
This is used to force the Linux VM to keep a minimum number
|
113 |
|
|
of kilobytes free. The VM uses this number to compute a pages_min
|
114 |
|
|
value for each lowmem zone in the system. Each lowmem zone gets
|
115 |
|
|
a number of reserved free pages based proportionally on its size.
|
116 |
|
|
|
117 |
|
|
Some minimal ammount of memory is needed to satisfy PF_MEMALLOC
|
118 |
|
|
allocations; if you set this to lower than 1024KB, your system will
|
119 |
|
|
become subtly broken, and prone to deadlock under high loads.
|
120 |
|
|
|
121 |
|
|
Setting this too high will OOM your machine instantly.
|
122 |
|
|
|
123 |
|
|
==============================================================
|
124 |
|
|
|
125 |
|
|
percpu_pagelist_fraction
|
126 |
|
|
|
127 |
|
|
This is the fraction of pages at most (high mark pcp->high) in each zone that
|
128 |
|
|
are allocated for each per cpu page list. The min value for this is 8. It
|
129 |
|
|
means that we don't allow more than 1/8th of pages in each zone to be
|
130 |
|
|
allocated in any single per_cpu_pagelist. This entry only changes the value
|
131 |
|
|
of hot per cpu pagelists. User can specify a number like 100 to allocate
|
132 |
|
|
1/100th of each zone to each per cpu page list.
|
133 |
|
|
|
134 |
|
|
The batch value of each per cpu pagelist is also updated as a result. It is
|
135 |
|
|
set to pcp->high/4. The upper limit of batch is (PAGE_SHIFT * 8)
|
136 |
|
|
|
137 |
|
|
The initial value is zero. Kernel does not use this value at boot time to set
|
138 |
|
|
the high water marks for each per cpu page list.
|
139 |
|
|
|
140 |
|
|
===============================================================
|
141 |
|
|
|
142 |
|
|
zone_reclaim_mode:
|
143 |
|
|
|
144 |
|
|
Zone_reclaim_mode allows someone to set more or less aggressive approaches to
|
145 |
|
|
reclaim memory when a zone runs out of memory. If it is set to zero then no
|
146 |
|
|
zone reclaim occurs. Allocations will be satisfied from other zones / nodes
|
147 |
|
|
in the system.
|
148 |
|
|
|
149 |
|
|
This is value ORed together of
|
150 |
|
|
|
151 |
|
|
1 = Zone reclaim on
|
152 |
|
|
2 = Zone reclaim writes dirty pages out
|
153 |
|
|
4 = Zone reclaim swaps pages
|
154 |
|
|
|
155 |
|
|
zone_reclaim_mode is set during bootup to 1 if it is determined that pages
|
156 |
|
|
from remote zones will cause a measurable performance reduction. The
|
157 |
|
|
page allocator will then reclaim easily reusable pages (those page
|
158 |
|
|
cache pages that are currently not used) before allocating off node pages.
|
159 |
|
|
|
160 |
|
|
It may be beneficial to switch off zone reclaim if the system is
|
161 |
|
|
used for a file server and all of memory should be used for caching files
|
162 |
|
|
from disk. In that case the caching effect is more important than
|
163 |
|
|
data locality.
|
164 |
|
|
|
165 |
|
|
Allowing zone reclaim to write out pages stops processes that are
|
166 |
|
|
writing large amounts of data from dirtying pages on other nodes. Zone
|
167 |
|
|
reclaim will write out dirty pages if a zone fills up and so effectively
|
168 |
|
|
throttle the process. This may decrease the performance of a single process
|
169 |
|
|
since it cannot use all of system memory to buffer the outgoing writes
|
170 |
|
|
anymore but it preserve the memory on other nodes so that the performance
|
171 |
|
|
of other processes running on other nodes will not be affected.
|
172 |
|
|
|
173 |
|
|
Allowing regular swap effectively restricts allocations to the local
|
174 |
|
|
node unless explicitly overridden by memory policies or cpuset
|
175 |
|
|
configurations.
|
176 |
|
|
|
177 |
|
|
=============================================================
|
178 |
|
|
|
179 |
|
|
min_unmapped_ratio:
|
180 |
|
|
|
181 |
|
|
This is available only on NUMA kernels.
|
182 |
|
|
|
183 |
|
|
A percentage of the total pages in each zone. Zone reclaim will only
|
184 |
|
|
occur if more than this percentage of pages are file backed and unmapped.
|
185 |
|
|
This is to insure that a minimal amount of local pages is still available for
|
186 |
|
|
file I/O even if the node is overallocated.
|
187 |
|
|
|
188 |
|
|
The default is 1 percent.
|
189 |
|
|
|
190 |
|
|
=============================================================
|
191 |
|
|
|
192 |
|
|
min_slab_ratio:
|
193 |
|
|
|
194 |
|
|
This is available only on NUMA kernels.
|
195 |
|
|
|
196 |
|
|
A percentage of the total pages in each zone. On Zone reclaim
|
197 |
|
|
(fallback from the local zone occurs) slabs will be reclaimed if more
|
198 |
|
|
than this percentage of pages in a zone are reclaimable slab pages.
|
199 |
|
|
This insures that the slab growth stays under control even in NUMA
|
200 |
|
|
systems that rarely perform global reclaim.
|
201 |
|
|
|
202 |
|
|
The default is 5 percent.
|
203 |
|
|
|
204 |
|
|
Note that slab reclaim is triggered in a per zone / node fashion.
|
205 |
|
|
The process of reclaiming slab memory is currently not node specific
|
206 |
|
|
and may not be fast.
|
207 |
|
|
|
208 |
|
|
=============================================================
|
209 |
|
|
|
210 |
|
|
panic_on_oom
|
211 |
|
|
|
212 |
|
|
This enables or disables panic on out-of-memory feature.
|
213 |
|
|
|
214 |
|
|
If this is set to 0, the kernel will kill some rogue process,
|
215 |
|
|
called oom_killer. Usually, oom_killer can kill rogue processes and
|
216 |
|
|
system will survive.
|
217 |
|
|
|
218 |
|
|
If this is set to 1, the kernel panics when out-of-memory happens.
|
219 |
|
|
However, if a process limits using nodes by mempolicy/cpusets,
|
220 |
|
|
and those nodes become memory exhaustion status, one process
|
221 |
|
|
may be killed by oom-killer. No panic occurs in this case.
|
222 |
|
|
Because other nodes' memory may be free. This means system total status
|
223 |
|
|
may be not fatal yet.
|
224 |
|
|
|
225 |
|
|
If this is set to 2, the kernel panics compulsorily even on the
|
226 |
|
|
above-mentioned.
|
227 |
|
|
|
228 |
|
|
The default value is 0.
|
229 |
|
|
1 and 2 are for failover of clustering. Please select either
|
230 |
|
|
according to your policy of failover.
|
231 |
|
|
|
232 |
|
|
=============================================================
|
233 |
|
|
|
234 |
|
|
oom_kill_allocating_task
|
235 |
|
|
|
236 |
|
|
This enables or disables killing the OOM-triggering task in
|
237 |
|
|
out-of-memory situations.
|
238 |
|
|
|
239 |
|
|
If this is set to zero, the OOM killer will scan through the entire
|
240 |
|
|
tasklist and select a task based on heuristics to kill. This normally
|
241 |
|
|
selects a rogue memory-hogging task that frees up a large amount of
|
242 |
|
|
memory when killed.
|
243 |
|
|
|
244 |
|
|
If this is set to non-zero, the OOM killer simply kills the task that
|
245 |
|
|
triggered the out-of-memory condition. This avoids the expensive
|
246 |
|
|
tasklist scan.
|
247 |
|
|
|
248 |
|
|
If panic_on_oom is selected, it takes precedence over whatever value
|
249 |
|
|
is used in oom_kill_allocating_task.
|
250 |
|
|
|
251 |
|
|
The default value is 0.
|
252 |
|
|
|
253 |
|
|
==============================================================
|
254 |
|
|
|
255 |
|
|
mmap_min_addr
|
256 |
|
|
|
257 |
|
|
This file indicates the amount of address space which a user process will
|
258 |
|
|
be restricted from mmaping. Since kernel null dereference bugs could
|
259 |
|
|
accidentally operate based on the information in the first couple of pages
|
260 |
|
|
of memory userspace processes should not be allowed to write to them. By
|
261 |
|
|
default this value is set to 0 and no protections will be enforced by the
|
262 |
|
|
security module. Setting this value to something like 64k will allow the
|
263 |
|
|
vast majority of applications to work correctly and provide defense in depth
|
264 |
|
|
against future potential kernel bugs.
|
265 |
|
|
|
266 |
|
|
==============================================================
|
267 |
|
|
|
268 |
|
|
numa_zonelist_order
|
269 |
|
|
|
270 |
|
|
This sysctl is only for NUMA.
|
271 |
|
|
'where the memory is allocated from' is controlled by zonelists.
|
272 |
|
|
(This documentation ignores ZONE_HIGHMEM/ZONE_DMA32 for simple explanation.
|
273 |
|
|
you may be able to read ZONE_DMA as ZONE_DMA32...)
|
274 |
|
|
|
275 |
|
|
In non-NUMA case, a zonelist for GFP_KERNEL is ordered as following.
|
276 |
|
|
ZONE_NORMAL -> ZONE_DMA
|
277 |
|
|
This means that a memory allocation request for GFP_KERNEL will
|
278 |
|
|
get memory from ZONE_DMA only when ZONE_NORMAL is not available.
|
279 |
|
|
|
280 |
|
|
In NUMA case, you can think of following 2 types of order.
|
281 |
|
|
Assume 2 node NUMA and below is zonelist of Node(0)'s GFP_KERNEL
|
282 |
|
|
|
283 |
|
|
(A) Node(0) ZONE_NORMAL -> Node(0) ZONE_DMA -> Node(1) ZONE_NORMAL
|
284 |
|
|
(B) Node(0) ZONE_NORMAL -> Node(1) ZONE_NORMAL -> Node(0) ZONE_DMA.
|
285 |
|
|
|
286 |
|
|
Type(A) offers the best locality for processes on Node(0), but ZONE_DMA
|
287 |
|
|
will be used before ZONE_NORMAL exhaustion. This increases possibility of
|
288 |
|
|
out-of-memory(OOM) of ZONE_DMA because ZONE_DMA is tend to be small.
|
289 |
|
|
|
290 |
|
|
Type(B) cannot offer the best locality but is more robust against OOM of
|
291 |
|
|
the DMA zone.
|
292 |
|
|
|
293 |
|
|
Type(A) is called as "Node" order. Type (B) is "Zone" order.
|
294 |
|
|
|
295 |
|
|
"Node order" orders the zonelists by node, then by zone within each node.
|
296 |
|
|
Specify "[Nn]ode" for zone order
|
297 |
|
|
|
298 |
|
|
"Zone Order" orders the zonelists by zone type, then by node within each
|
299 |
|
|
zone. Specify "[Zz]one"for zode order.
|
300 |
|
|
|
301 |
|
|
Specify "[Dd]efault" to request automatic configuration. Autoconfiguration
|
302 |
|
|
will select "node" order in following case.
|
303 |
|
|
(1) if the DMA zone does not exist or
|
304 |
|
|
(2) if the DMA zone comprises greater than 50% of the available memory or
|
305 |
|
|
(3) if any node's DMA zone comprises greater than 60% of its local memory and
|
306 |
|
|
the amount of local memory is big enough.
|
307 |
|
|
|
308 |
|
|
Otherwise, "zone" order will be selected. Default order is recommended unless
|
309 |
|
|
this is causing problems for your system/application.
|
310 |
|
|
|
311 |
|
|
==============================================================
|
312 |
|
|
|
313 |
|
|
nr_hugepages
|
314 |
|
|
|
315 |
|
|
Change the minimum size of the hugepage pool.
|
316 |
|
|
|
317 |
|
|
See Documentation/vm/hugetlbpage.txt
|
318 |
|
|
|
319 |
|
|
==============================================================
|
320 |
|
|
|
321 |
|
|
nr_overcommit_hugepages
|
322 |
|
|
|
323 |
|
|
Change the maximum size of the hugepage pool. The maximum is
|
324 |
|
|
nr_hugepages + nr_overcommit_hugepages.
|
325 |
|
|
|
326 |
|
|
See Documentation/vm/hugetlbpage.txt
|