]>
Commit | Line | Data |
---|---|---|
314093c9 | 1 | .\" Copyright 2003,2004 Andi Kleen, SuSE Labs. |
73ae0a09 | 2 | .\" and Copyright 2007 Lee Schermerhorn, Hewlett Packard |
314093c9 MK |
3 | .\" |
4 | .\" Permission is granted to make and distribute verbatim copies of this | |
5 | .\" manual provided the copyright notice and this permission notice are | |
6 | .\" preserved on all copies. | |
7 | .\" | |
8 | .\" Permission is granted to copy and distribute modified versions of this | |
9 | .\" manual under the conditions for verbatim copying, provided that the | |
10 | .\" entire resulting derived work is distributed under the terms of a | |
11 | .\" permission notice identical to this one. | |
c13182ef | 12 | .\" |
314093c9 MK |
13 | .\" Since the Linux kernel and libraries are constantly changing, this |
14 | .\" manual page may be incorrect or out-of-date. The author(s) assume no | |
15 | .\" responsibility for errors or omissions, or for damages resulting from | |
c13182ef MK |
16 | .\" the use of the information contained herein. |
17 | .\" | |
314093c9 MK |
18 | .\" Formatted or processed versions of this manual, if unaccompanied by |
19 | .\" the source, must acknowledge the copyright and authors of this work. | |
c13182ef | 20 | .\" |
314093c9 | 21 | .\" 2006-02-03, mtk, substantial wording changes and other improvements |
00045cbb MK |
22 | .\" 2007-08-27, Lee Schermerhorn <Lee.Schermerhorn@hp.com> |
23 | .\" more precise specification of behavior. | |
314093c9 | 24 | .\" |
1313d297 | 25 | .TH SET_MEMPOLICY 2 2008-08-11 Linux "Linux Programmer's Manual" |
314093c9 | 26 | .SH NAME |
73ae0a09 | 27 | set_mempolicy \- set default NUMA memory policy for a process and its children |
314093c9 | 28 | .SH SYNOPSIS |
521bf584 | 29 | .nf |
c13182ef | 30 | .B "#include <numaif.h>" |
314093c9 | 31 | .sp |
73ae0a09 | 32 | .BI "int set_mempolicy(int " mode ", unsigned long *" nodemask , |
521bf584 | 33 | .BI " unsigned long " maxnode ); |
73ae0a09 | 34 | .sp |
00045cbb | 35 | Link with \fI\-lnuma\fP |
521bf584 | 36 | .fi |
314093c9 MK |
37 | .SH DESCRIPTION |
38 | .BR set_mempolicy () | |
73ae0a09 MK |
39 | sets the NUMA memory policy of the calling process, |
40 | which consists of a policy mode and zero or more nodes, | |
41 | to the values specified by the | |
42 | .IR mode , | |
43 | .I nodemask | |
44 | and | |
0daa9e92 | 45 | .I maxnode |
73ae0a09 | 46 | arguments. |
314093c9 MK |
47 | |
48 | A NUMA machine has different | |
49 | memory controllers with different distances to specific CPUs. | |
73ae0a09 | 50 | The memory policy defines from which node memory is allocated for |
c13182ef | 51 | the process. |
314093c9 | 52 | |
73ae0a09 | 53 | This system call defines the default policy for the process. |
ecccf7c2 | 54 | The process policy governs allocation of pages in the process's |
73ae0a09 MK |
55 | address space outside of memory ranges |
56 | controlled by a more specific policy set by | |
314093c9 | 57 | .BR mbind (2). |
73ae0a09 MK |
58 | The process default policy also controls allocation of any pages for |
59 | memory mapped files mapped using the | |
60 | .BR mmap (2) | |
61 | call with the | |
62 | .B MAP_PRIVATE | |
cdba9253 | 63 | flag and that are only read [loaded] from by the process |
73ae0a09 MK |
64 | and of memory mapped files mapped using the |
65 | .BR mmap (2) | |
66 | call with the | |
67 | .B MAP_SHARED | |
68 | flag, regardless of the access type. | |
314093c9 | 69 | The policy is only applied when a new page is allocated |
c13182ef MK |
70 | for the process. |
71 | For anonymous memory this is when the page is first | |
314093c9 MK |
72 | touched by the application. |
73 | ||
73ae0a09 MK |
74 | The |
75 | .I mode | |
76 | argument must specify one of | |
314093c9 MK |
77 | .BR MPOL_DEFAULT , |
78 | .BR MPOL_BIND , | |
73ae0a09 MK |
79 | .B MPOL_INTERLEAVE |
80 | or | |
314093c9 | 81 | .BR MPOL_PREFERRED . |
73ae0a09 | 82 | All modes except |
314093c9 | 83 | .B MPOL_DEFAULT |
73ae0a09 | 84 | require the caller to specify via the |
c13182ef | 85 | .I nodemask |
c4bb193f | 86 | argument one or more nodes. |
73ae0a09 | 87 | |
f98b728e MK |
88 | The |
89 | .I mode | |
90 | argument may also include an optional | |
91 | .IR "mode flag ". | |
92 | The supported | |
93 | .I "mode flags" | |
94 | are: | |
95 | .TP | |
96 | .BR MPOL_F_STATIC_NODES " (since Linux 2.6.26)" | |
97 | A non-empty | |
98 | .I nodemask | |
99 | specifies physical node ids. | |
100 | Linux does will not remap the | |
101 | .I nodemask | |
102 | when the process moves to a different cpuset context, | |
103 | nor when the set of nodes allowed by the process's | |
104 | current cpuset context changes. | |
105 | .TP | |
106 | .BR MPOL_F_RELATIVE_NODES " (since Linux 2.6.26)" | |
107 | A non-empty | |
108 | .I nodemask | |
109 | specifies node ids that are relative to the set of | |
110 | node ids allowed by the process's current cpuset. | |
111 | .PP | |
c13182ef | 112 | .I nodemask |
00045cbb | 113 | points to a bit mask of node IDs that contains up to |
314093c9 | 114 | .I maxnode |
c13182ef | 115 | bits. |
73ae0a09 | 116 | The bit mask size is rounded to the next multiple of |
c13182ef MK |
117 | .IR "sizeof(unsigned long)" , |
118 | but the kernel will only use bits up to | |
314093c9 | 119 | .IR maxnode . |
73ae0a09 MK |
120 | A NULL value of |
121 | .I nodemask | |
122 | or a | |
123 | .I maxnode | |
124 | value of zero specifies the empty set of nodes. | |
125 | If the value of | |
126 | .I maxnode | |
127 | is zero, | |
128 | the | |
129 | .I nodemask | |
130 | argument is ignored. | |
f98b728e | 131 | |
cdba9253 MK |
132 | Where a |
133 | .I nodemask | |
134 | is required, it must contain at least one node that is on-line, | |
135 | allowed by the process's current cpuset context, | |
f98b728e MK |
136 | [unless the |
137 | .B MPOL_F_STATIC_NODES | |
138 | mode flag is specified], | |
cdba9253 | 139 | and contains memory. |
f98b728e MK |
140 | If the |
141 | .B MPOL_F_STATIC_NODES | |
142 | is set in | |
143 | .I mode | |
144 | and a required | |
145 | .I nodemask | |
146 | contains no nodes that are allowed by the process's current cpuset context, | |
147 | the memory policy reverts to | |
148 | .IR "local allocation" . | |
149 | This effectively overrides the specified policy until the process's | |
150 | cpuset context includes one or more of the nodes specified by | |
151 | .IR nodemask. | |
314093c9 | 152 | |
c13182ef | 153 | The |
314093c9 | 154 | .B MPOL_DEFAULT |
f98b728e MK |
155 | mode specifies that any non-default process memory policy be removed, |
156 | so that the memory policy "falls back" to the system default policy. | |
157 | The system default policy is "local allocation"-- | |
158 | i.e., allocate memory on the node of the CPU that triggered the allocation. | |
c13182ef | 159 | .I nodemask |
73ae0a09 MK |
160 | must be specified as NULL. |
161 | If the "local node" contains no free memory, the system will | |
162 | attempt to allocate memory from a "near by" node. | |
314093c9 MK |
163 | |
164 | The | |
165 | .B MPOL_BIND | |
73ae0a09 | 166 | mode defines a strict policy that restricts memory allocation to the |
c13182ef | 167 | nodes specified in |
314093c9 | 168 | .IR nodemask . |
73ae0a09 MK |
169 | If |
170 | .I nodemask | |
171 | specifies more than one node, page allocations will come from | |
00045cbb | 172 | the node with the lowest numeric node ID first, until that node |
73ae0a09 MK |
173 | contains no free memory. |
174 | Allocations will then come from the node with the next highest | |
00045cbb | 175 | node ID specified in |
73ae0a09 MK |
176 | .I nodemask |
177 | and so forth, until none of the specified nodes contain free memory. | |
178 | Pages will not be allocated from any node not specified in the | |
179 | .IR nodemask . | |
314093c9 MK |
180 | |
181 | .B MPOL_INTERLEAVE | |
73ae0a09 MK |
182 | interleaves page allocations across the nodes specified in |
183 | .I nodemask | |
00045cbb | 184 | in numeric node ID order. |
73ae0a09 MK |
185 | This optimizes for bandwidth instead of latency |
186 | by spreading out pages and memory accesses to those pages across | |
187 | multiple nodes. | |
188 | However, accesses to a single page will still be limited to | |
189 | the memory bandwidth of a single node. | |
190 | .\" NOTE: the following sentence doesn't make sense in the context | |
191 | .\" of set_mempolicy() -- no memory area specified. | |
192 | .\" To be effective the memory area should be fairly large, | |
193 | .\" at least 1MB or bigger. | |
314093c9 MK |
194 | |
195 | .B MPOL_PREFERRED | |
c13182ef | 196 | sets the preferred node for allocation. |
73ae0a09 MK |
197 | The kernel will try to allocate pages from this node first |
198 | and fall back to "near by" nodes if the preferred node is low on free | |
c13182ef | 199 | memory. |
73ae0a09 MK |
200 | If |
201 | .I nodemask | |
00045cbb | 202 | specifies more than one node ID, the first node in the |
73ae0a09 MK |
203 | mask will be selected as the preferred node. |
204 | If the | |
c13182ef | 205 | .I nodemask |
73ae0a09 MK |
206 | and |
207 | .I maxnode | |
1313d297 MK |
208 | arguments specify the empty set, then the policy |
209 | specifies "local allocation" | |
210 | (like the system default policy discussed above). | |
314093c9 | 211 | |
73ae0a09 | 212 | The process memory policy is preserved across an |
3bd6a9b1 MK |
213 | .BR execve (2), |
214 | and is inherited by child processes created using | |
c13182ef MK |
215 | .BR fork (2) |
216 | or | |
314093c9 | 217 | .BR clone (2). |
314093c9 MK |
218 | .SH RETURN VALUE |
219 | On success, | |
220 | .BR set_mempolicy () | |
221 | returns 0; | |
222 | on error, \-1 is returned and | |
c13182ef | 223 | .I errno |
314093c9 | 224 | is set to indicate the error. |
73ae0a09 MK |
225 | .SH ERRORS |
226 | .TP | |
b3a7b55e MK |
227 | .B EFAULT |
228 | Part of all of the memory range specified by | |
229 | .I nodemask | |
230 | and | |
231 | .I maxnode | |
232 | points outside your accessible address space. | |
233 | .TP | |
73ae0a09 | 234 | .B EINVAL |
4d2be0ee MK |
235 | .I mode |
236 | is invalid. | |
73ae0a09 MK |
237 | Or, |
238 | .I mode | |
239 | is | |
00045cbb | 240 | .B MPOL_DEFAULT |
73ae0a09 MK |
241 | and |
242 | .I nodemask | |
1f04cc97 | 243 | is non-empty, |
73ae0a09 MK |
244 | or |
245 | .I mode | |
246 | is | |
00045cbb | 247 | .B MPOL_BIND |
73ae0a09 | 248 | or |
00045cbb | 249 | .B MPOL_INTERLEAVE |
73ae0a09 MK |
250 | and |
251 | .I nodemask | |
252 | is empty. | |
253 | Or, | |
254 | .I maxnode | |
255 | specifies more than a page worth of bits. | |
256 | Or, | |
257 | .I nodemask | |
00045cbb | 258 | specifies one or more node IDs that are |
cdba9253 | 259 | greater than the maximum supported node ID. |
00045cbb | 260 | Or, none of the node IDs specified by |
73ae0a09 | 261 | .I nodemask |
cdba9253 MK |
262 | are on-line and allowed by the process's current cpuset context, |
263 | or none of the specified nodes contain memory. | |
f98b728e MK |
264 | Or, the |
265 | .I mode | |
266 | argument specified both | |
267 | .B MPOL_F_STATIC_NODES | |
268 | and | |
269 | .BR MPOL_F_RELATIVE_NODES . | |
73ae0a09 | 270 | .TP |
73ae0a09 MK |
271 | .B ENOMEM |
272 | Insufficient kernel memory was available. | |
9d9dc1e8 | 273 | .SH CONFORMING TO |
8382f16d | 274 | This system call is Linux-specific. |
a1d5f77c MK |
275 | .SH NOTES |
276 | Process policy is not remembered if the page is swapped out. | |
73ae0a09 MK |
277 | When such a page is paged back in, it will use the policy of |
278 | the process or memory range that is in effect at the time the | |
279 | page is allocated. | |
0e99f2a5 MK |
280 | .SS "Versions and Library Support" |
281 | See | |
282 | .BR mbind (2). | |
314093c9 | 283 | .SH SEE ALSO |
fa23e023 | 284 | .BR get_mempolicy (2), |
f0c34053 | 285 | .BR getcpu (2), |
314093c9 | 286 | .BR mbind (2), |
73ae0a09 | 287 | .BR mmap (2), |
a18e2edb MK |
288 | .BR numa (3), |
289 | .BR cpuset (7), | |
290 | .BR numactl (8) |