man2/set_mempolicy.2

   1 .\" Copyright 2003,2004 Andi Kleen, SuSE Labs.
   2 .\" and Copyright 2007 Lee Schermerhorn, Hewlett Packard
   3 .\"
   4 .\" %%%LICENSE_START(VERBATIM_PROF)
   5 .\" Permission is granted to make and distribute verbatim copies of this
   6 .\" manual provided the copyright notice and this permission notice are
   7 .\" preserved on all copies.
   8 .\"
   9 .\" Permission is granted to copy and distribute modified versions of this
  10 .\" manual under the conditions for verbatim copying, provided that the
  11 .\" entire resulting derived work is distributed under the terms of a
  12 .\" permission notice identical to this one.
  13 .\"
  14 .\" Since the Linux kernel and libraries are constantly changing, this
  15 .\" manual page may be incorrect or out-of-date.  The author(s) assume no
  16 .\" responsibility for errors or omissions, or for damages resulting from
  17 .\" the use of the information contained herein.
  18 .\"
  19 .\" Formatted or processed versions of this manual, if unaccompanied by
  20 .\" the source, must acknowledge the copyright and authors of this work.
  21 .\" %%%LICENSE_END
  22 .\"
  23 .\" 2006-02-03, mtk, substantial wording changes and other improvements
  24 .\" 2007-08-27, Lee Schermerhorn <Lee.Schermerhorn@hp.com>
  25 .\"     more precise specification of behavior.
  26 .\"
  27 .TH SET_MEMPOLICY 2 2015-05-07 Linux "Linux Programmer's Manual"
  28 .SH NAME
  29 set_mempolicy \- set default NUMA memory policy for a thread and its children
  30 .SH SYNOPSIS
  31 .nf
  32 .B "#include <numaif.h>"
  33 .sp
  34 .BI "long set_mempolicy(int " mode ", const unsigned long *" nodemask ,
  35 .BI "                   unsigned long " maxnode );
  36 .sp
  37 Link with \fI\-lnuma\fP.
  38 .fi
  39 .SH DESCRIPTION
  40 .BR set_mempolicy ()
  41 sets the NUMA memory policy of the calling thread,
  42 which consists of a policy mode and zero or more nodes,
  43 to the values specified by the
  44 .IR mode ,
  45 .I nodemask
  46 and
  47 .I maxnode
  48 arguments.
  49
  50 A NUMA machine has different
  51 memory controllers with different distances to specific CPUs.
  52 The memory policy defines from which node memory is allocated for
  53 the thread.
  54
  55 This system call defines the default policy for the thread.
  56 The thread policy governs allocation of pages in the process's
  57 address space outside of memory ranges
  58 controlled by a more specific policy set by
  59 .BR mbind (2).
  60 The thread default policy also controls allocation of any pages for
  61 memory-mapped files mapped using the
  62 .BR mmap (2)
  63 call with the
  64 .B MAP_PRIVATE
  65 flag and that are only read [loaded] from by the thread
  66 and of memory-mapped files mapped using the
  67 .BR mmap (2)
  68 call with the
  69 .B MAP_SHARED
  70 flag, regardless of the access type.
  71 The policy is applied only when a new page is allocated
  72 for the thread.
  73 For anonymous memory this is when the page is first
  74 touched by the thread.
  75
  76 The
  77 .I mode
  78 argument must specify one of
  79 .BR MPOL_DEFAULT ,
  80 .BR MPOL_BIND ,
  81 .BR MPOL_INTERLEAVE ,
  82 .BR MPOL_PREFERRED ,
  83 or
  84 .BR MPOL_LOCAL
  85 (which are described in detail below).
  86 All modes except
  87 .B MPOL_DEFAULT
  88 require the caller to specify via the
  89 .I nodemask
  90 argument one or more nodes.
  91
  92 The
  93 .I mode
  94 argument may also include an optional
  95 .IR "mode flag" .
  96 The supported
  97 .I "mode flags"
  98 are:
  99 .TP
 100 .BR MPOL_F_STATIC_NODES " (since Linux 2.6.26)"
 101 A nonempty
 102 .I nodemask
 103 specifies physical node ids.
 104 Linux will not remap the
 105 .I nodemask
 106 when the process moves to a different cpuset context,
 107 nor when the set of nodes allowed by the process's
 108 current cpuset context changes.
 109 .TP
 110 .BR MPOL_F_RELATIVE_NODES " (since Linux 2.6.26)"
 111 A nonempty
 112 .I nodemask
 113 specifies node ids that are relative to the set of
 114 node ids allowed by the process's current cpuset.
 115 .PP
 116 .I nodemask
 117 points to a bit mask of node IDs that contains up to
 118 .I maxnode
 119 bits.
 120 The bit mask size is rounded to the next multiple of
 121 .IR "sizeof(unsigned long)" ,
 122 but the kernel will use bits only up to
 123 .IR maxnode .
 124 A NULL value of
 125 .I nodemask
 126 or a
 127 .I maxnode
 128 value of zero specifies the empty set of nodes.
 129 If the value of
 130 .I maxnode
 131 is zero,
 132 the
 133 .I nodemask
 134 argument is ignored.
 135
 136 Where a
 137 .I nodemask
 138 is required, it must contain at least one node that is on-line,
 139 allowed by the process's current cpuset context,
 140 [unless the
 141 .B MPOL_F_STATIC_NODES
 142 mode flag is specified],
 143 and contains memory.
 144 If the
 145 .B MPOL_F_STATIC_NODES
 146 is set in
 147 .I mode
 148 and a required
 149 .I nodemask
 150 contains no nodes that are allowed by the process's current cpuset context,
 151 the memory policy reverts to
 152 .IR "local allocation" .
 153 This effectively overrides the specified policy until the process's
 154 cpuset context includes one or more of the nodes specified by
 155 .IR nodemask .
 156
 157 The
 158 .B MPOL_DEFAULT
 159 mode specifies that any nondefault thread memory policy be removed,
 160 so that the memory policy "falls back" to the system default policy.
 161 The system default policy is "local allocation"\(emthat is,
 162 allocate memory on the node of the CPU that triggered the allocation.
 163 .I nodemask
 164 must be specified as NULL.
 165 If the "local node" contains no free memory, the system will
 166 attempt to allocate memory from a "near by" node.
 167
 168 The
 169 .B MPOL_BIND
 170 mode defines a strict policy that restricts memory allocation to the
 171 nodes specified in
 172 .IR nodemask .
 173 If
 174 .I nodemask
 175 specifies more than one node, page allocations will come from
 176 the node with the lowest numeric node ID first, until that node
 177 contains no free memory.
 178 Allocations will then come from the node with the next highest
 179 node ID specified in
 180 .I nodemask
 181 and so forth, until none of the specified nodes contain free memory.
 182 Pages will not be allocated from any node not specified in the
 183 .IR nodemask .
 184
 185 .B MPOL_INTERLEAVE
 186 interleaves page allocations across the nodes specified in
 187 .I nodemask
 188 in numeric node ID order.
 189 This optimizes for bandwidth instead of latency
 190 by spreading out pages and memory accesses to those pages across
 191 multiple nodes.
 192 However, accesses to a single page will still be limited to
 193 the memory bandwidth of a single node.
 194 .\" NOTE:  the following sentence doesn't make sense in the context
 195 .\" of set_mempolicy() -- no memory area specified.
 196 .\" To be effective the memory area should be fairly large,
 197 .\" at least 1MB or bigger.
 198
 199 .B MPOL_PREFERRED
 200 sets the preferred node for allocation.
 201 The kernel will try to allocate pages from this node first
 202 and fall back to "near by" nodes if the preferred node is low on free
 203 memory.
 204 If
 205 .I nodemask
 206 specifies more than one node ID, the first node in the
 207 mask will be selected as the preferred node.
 208 If the
 209 .I nodemask
 210 and
 211 .I maxnode
 212 arguments specify the empty set, then the policy
 213 specifies "local allocation"
 214 (like the system default policy discussed above).
 215
 216 .BR MPOL_LOCAL " (since Linux 3.8)"
 217 .\" commit 479e2802d09f1e18a97262c4c6f8f17ae5884bd8
 218 .\" commit f2a07f40dbc603c15f8b06e6ec7f768af67b424f
 219 specifies "local allocation"; the memory is allocated on
 220 the node of the CPU that triggered the allocation (the "local node").
 221 The
 222 .I nodemask
 223 and
 224 .I maxnode
 225 arguments must specify the empty set.
 226 If the "local node" is low on free memory,
 227 the kernel will try to allocate memory from other nodes.
 228 The kernel will allocate memory from the "local node"
 229 whenever memory for this node is available.
 230 If the "local node" is not allowed by the process's current cpuset context,
 231 the kernel will try to allocate memory from other nodes.
 232 The kernel will allocate memory from the "local node" whenever
 233 it becomes allowed by the process's current cpuset context.
 234
 235 The thread memory policy is preserved across an
 236 .BR execve (2),
 237 and is inherited by child threads created using
 238 .BR fork (2)
 239 or
 240 .BR clone (2).
 241 .SH RETURN VALUE
 242 On success,
 243 .BR set_mempolicy ()
 244 returns 0;
 245 on error, \-1 is returned and
 246 .I errno
 247 is set to indicate the error.
 248 .SH ERRORS
 249 .TP
 250 .B EFAULT
 251 Part of all of the memory range specified by
 252 .I nodemask
 253 and
 254 .I maxnode
 255 points outside your accessible address space.
 256 .TP
 257 .B EINVAL
 258 .I mode
 259 is invalid.
 260 Or,
 261 .I mode
 262 is
 263 .B MPOL_DEFAULT
 264 and
 265 .I nodemask
 266 is nonempty,
 267 or
 268 .I mode
 269 is
 270 .B MPOL_BIND
 271 or
 272 .B MPOL_INTERLEAVE
 273 and
 274 .I nodemask
 275 is empty.
 276 Or,
 277 .I maxnode
 278 specifies more than a page worth of bits.
 279 Or,
 280 .I nodemask
 281 specifies one or more node IDs that are
 282 greater than the maximum supported node ID.
 283 Or, none of the node IDs specified by
 284 .I nodemask
 285 are on-line and allowed by the process's current cpuset context,
 286 or none of the specified nodes contain memory.
 287 Or, the
 288 .I mode
 289 argument specified both
 290 .B MPOL_F_STATIC_NODES
 291 and
 292 .BR MPOL_F_RELATIVE_NODES .
 293 .TP
 294 .B ENOMEM
 295 Insufficient kernel memory was available.
 296 .SH VERSIONS
 297 The
 298 .BR set_mempolicy (),
 299 system call was added to the Linux kernel in version 2.6.7.
 300 .SH CONFORMING TO
 301 This system call is Linux-specific.
 302 .SH NOTES
 303 Memory policy is not remembered if the page is swapped out.
 304 When such a page is paged back in, it will use the policy of
 305 the thread or memory range that is in effect at the time the
 306 page is allocated.
 307
 308 For information on library support, see
 309 .BR numa (7).
 310 .SH SEE ALSO
 311 .BR get_mempolicy (2),
 312 .BR getcpu (2),
 313 .BR mbind (2),
 314 .BR mmap (2),
 315 .BR numa (3),
 316 .BR cpuset (7),
 317 .BR numa (7),
 318 .BR numactl (8)