1SET_MEMPOLICY(2) Linux Programmer's Manual SET_MEMPOLICY(2)
2
6 set_mempolicy - set default NUMA memory policy for a process and its
7 children
8
10 #include <numaif.h>
11 int set_mempolicy(int mode, unsigned long *nodemask,
13 unsigned long maxnode);
14 Link with -lnuma.
16
18 set_mempolicy() sets the NUMA memory policy of the calling process,
19 which consists of a policy mode and zero or more nodes, to the values
20 specified by the mode, nodemask and maxnode arguments.
21 A NUMA machine has different memory controllers with different dis‐
23 tances to specific CPUs. The memory policy defines from which node
24 memory is allocated for the process.
25 This system call defines the default policy for the process. The
27 process policy governs allocation of pages in the process's address
28 space outside of memory ranges controlled by a more specific policy set
29 by mbind(2). The process default policy also controls allocation of
30 any pages for memory mapped files mapped using the mmap(2) call with
31 the MAP_PRIVATE flag and that are only read [loaded] from by the
32 process and of memory mapped files mapped using the mmap(2) call with
33 the MAP_SHARED flag, regardless of the access type. The policy is
34 applied only when a new page is allocated for the process. For anony‐
35 mous memory this is when the page is first touched by the application.
36 The mode argument must specify one of MPOL_DEFAULT, MPOL_BIND,
38 MPOL_INTERLEAVE or MPOL_PREFERRED. All modes except MPOL_DEFAULT
39 require the caller to specify via the nodemask argument one or more
40 nodes.
41 The mode argument may also include an optional mode flag. The sup‐
43 ported mode flags are:
44 MPOL_F_STATIC_NODES (since Linux 2.6.26)
46 A nonempty nodemask specifies physical node ids. Linux does
47 will not remap the nodemask when the process moves to a differ‐
48 ent cpuset context, nor when the set of nodes allowed by the
49 process's current cpuset context changes.
50 MPOL_F_RELATIVE_NODES (since Linux 2.6.26)
52 A nonempty nodemask specifies node ids that are relative to the
53 set of node ids allowed by the process's current cpuset.
54 nodemask points to a bit mask of node IDs that contains up to maxnode
56 bits. The bit mask size is rounded to the next multiple of
57 sizeof(unsigned long), but the kernel will use bits only up to maxnode.
58 A NULL value of nodemask or a maxnode value of zero specifies the empty
59 set of nodes. If the value of maxnode is zero, the nodemask argument
60 is ignored.
61 Where a nodemask is required, it must contain at least one node that is
63 on-line, allowed by the process's current cpuset context, [unless the
64 MPOL_F_STATIC_NODES mode flag is specified], and contains memory. If
65 the MPOL_F_STATIC_NODES is set in mode and a required nodemask contains
66 no nodes that are allowed by the process's current cpuset context, the
67 memory policy reverts to local allocation. This effectively overrides
68 the specified policy until the process's cpuset context includes one or
69 more of the nodes specified by nodemask.
70 The MPOL_DEFAULT mode specifies that any nondefault process memory pol‐
72 icy be removed, so that the memory policy "falls back" to the system
73 default policy. The system default policy is "local allocation"--
74 i.e., allocate memory on the node of the CPU that triggered the alloca‐
75 tion. nodemask must be specified as NULL. If the "local node" con‐
76 tains no free memory, the system will attempt to allocate memory from a
77 "near by" node.
78 The MPOL_BIND mode defines a strict policy that restricts memory allo‐
80 cation to the nodes specified in nodemask. If nodemask specifies more
81 than one node, page allocations will come from the node with the lowest
82 numeric node ID first, until that node contains no free memory. Allo‐
83 cations will then come from the node with the next highest node ID
84 specified in nodemask and so forth, until none of the specified nodes
85 contain free memory. Pages will not be allocated from any node not
86 specified in the nodemask.
87 MPOL_INTERLEAVE interleaves page allocations across the nodes specified
89 in nodemask in numeric node ID order. This optimizes for bandwidth
90 instead of latency by spreading out pages and memory accesses to those
91 pages across multiple nodes. However, accesses to a single page will
92 still be limited to the memory bandwidth of a single node.
93 MPOL_PREFERRED sets the preferred node for allocation. The kernel will
95 try to allocate pages from this node first and fall back to "near by"
96 nodes if the preferred node is low on free memory. If nodemask speci‐
97 fies more than one node ID, the first node in the mask will be selected
98 as the preferred node. If the nodemask and maxnode arguments specify
99 the empty set, then the policy specifies "local allocation" (like the
100 system default policy discussed above).
101 The process memory policy is preserved across an execve(2), and is
103 inherited by child processes created using fork(2) or clone(2).
104
106 On success, set_mempolicy() returns 0; on error, -1 is returned and
107 errno is set to indicate the error.
108
110 EFAULT Part of all of the memory range specified by nodemask and maxn‐
111 ode points outside your accessible address space.
112 EINVAL mode is invalid. Or, mode is MPOL_DEFAULT and nodemask is
114 nonempty, or mode is MPOL_BIND or MPOL_INTERLEAVE and nodemask
115 is empty. Or, maxnode specifies more than a page worth of bits.
116 Or, nodemask specifies one or more node IDs that are greater
117 than the maximum supported node ID. Or, none of the node IDs
118 specified by nodemask are on-line and allowed by the process's
119 current cpuset context, or none of the specified nodes contain
120 memory. Or, the mode argument specified both
121 MPOL_F_STATIC_NODES and MPOL_F_RELATIVE_NODES.
122 ENOMEM Insufficient kernel memory was available.
124
126 The set_mempolicy(), system call was added to the Linux kernel in ver‐
127 sion 2.6.7.
128
130 This system call is Linux-specific.
131
133 Process policy is not remembered if the page is swapped out. When such
134 a page is paged back in, it will use the policy of the process or mem‐
135 ory range that is in effect at the time the page is allocated.
136 For information on library support, see numa(7).
138
140 get_mempolicy(2), getcpu(2), mbind(2), mmap(2), numa(3), cpuset(7),
141 numa(7), numactl(8)
142
144 This page is part of release 3.53 of the Linux man-pages project. A
145 description of the project, and information about reporting bugs, can
146 be found at http://www.kernel.org/doc/man-pages/.
147Linux 2008-08-15 SET_MEMPOLICY(2)