1MOVE_PAGES(2) Linux Programmer's Manual MOVE_PAGES(2)
2
6 move_pages - move individual pages of a process to another node
7
9 #include <numaif.h>
10 long move_pages(int pid, unsigned long count, void **pages,
12 const int *nodes, int *status, int flags);
13 Link with -lnuma.
15 Note: There is no glibc wrapper for this system call; see NOTES.
17
19 move_pages() moves the specified pages of the process pid to the memory
20 nodes specified by nodes. The result of the move is reflected in sta‐
21 tus. The flags indicate constraints on the pages to be moved.
22 pid is the ID of the process in which pages are to be moved. If pid is
24 0, then move_pages() moves pages of the calling process.
25 To move pages in another process requires the following privileges:
27 * In kernels up to and including Linux 4.12: the caller must be privi‐
29 leged (CAP_SYS_NICE) or the real or effective user ID of the calling
30 process must match the real or saved-set user ID of the target
31 process.
32 * The older rules allowed the caller to discover various virtual ad‐
34 dress choices made by the kernel that could lead to the defeat of
35 address-space-layout randomization for a process owned by the same
36 UID as the caller, the rules were changed starting with Linux 4.13.
37 Since Linux 4.13, permission is governed by a ptrace access mode
38 PTRACE_MODE_READ_REALCREDS check with respect to the target process;
39 see ptrace(2).
40 count is the number of pages to move. It defines the size of the three
42 arrays pages, nodes, and status.
43 pages is an array of pointers to the pages that should be moved. These
45 are pointers that should be aligned to page boundaries. Addresses are
46 specified as seen by the process specified by pid.
47 nodes is an array of integers that specify the desired location for
49 each page. Each element in the array is a node number. nodes can also
50 be NULL, in which case move_pages() does not move any pages but instead
51 will return the node where each page currently resides, in the status
52 array. Obtaining the status of each page may be necessary to determine
53 pages that need to be moved.
54 status is an array of integers that return the status of each page.
56 The array contains valid values only if move_pages() did not return an
57 error. Preinitialization of the array to a value which cannot repre‐
58 sent a real numa node or valid error of status array could help to
59 identify pages that have been migrated.
60 flags specify what types of pages to move. MPOL_MF_MOVE means that
62 only pages that are in exclusive use by the process are to be moved.
63 MPOL_MF_MOVE_ALL means that pages shared between multiple processes can
64 also be moved. The process must be privileged (CAP_SYS_NICE) to use
65 MPOL_MF_MOVE_ALL.
66 Page states in the status array
68 The following values can be returned in each element of the status ar‐
69 ray.
70 0..MAX_NUMNODES
72 Identifies the node on which the page resides.
73 -EACCES
75 The page is mapped by multiple processes and can be moved only
76 if MPOL_MF_MOVE_ALL is specified.
77 -EBUSY The page is currently busy and cannot be moved. Try again
79 later. This occurs if a page is undergoing I/O or another ker‐
80 nel subsystem is holding a reference to the page.
81 -EFAULT
83 This is a zero page or the memory area is not mapped by the
84 process.
85 -EIO Unable to write back a page. The page has to be written back in
87 order to move it since the page is dirty and the filesystem does
88 not provide a migration function that would allow the move of
89 dirty pages.
90 -EINVAL
92 A dirty page cannot be moved. The filesystem does not provide a
93 migration function and has no ability to write back pages.
94 -ENOENT
96 The page is not present.
97 -ENOMEM
99 Unable to allocate memory on target node.
100
102 On success move_pages() returns zero. On error, it returns -1, and
103 sets errno to indicate the error. If positive value is returned, it is
104 the number of nonmigrated pages.
105
107 Positive value
108 The number of nonmigrated pages if they were the result of non‐
109 fatal reasons (since Linux 4.17).
110 E2BIG Too many pages to move. Since Linux 2.6.29, the kernel no
112 longer generates this error.
113 EACCES One of the target nodes is not allowed by the current cpuset.
115 EFAULT Parameter array could not be accessed.
117 EINVAL Flags other than MPOL_MF_MOVE and MPOL_MF_MOVE_ALL was specified
119 or an attempt was made to migrate pages of a kernel thread.
120 ENODEV One of the target nodes is not online.
122 EPERM The caller specified MPOL_MF_MOVE_ALL without sufficient privi‐
124 leges (CAP_SYS_NICE). Or, the caller attempted to move pages of
125 a process belonging to another user but did not have privilege
126 to do so (CAP_SYS_NICE).
127 ESRCH Process does not exist.
129
131 move_pages() first appeared on Linux in version 2.6.18.
132
134 This system call is Linux-specific.
135
137 Glibc does not provide a wrapper for this system call. For information
138 on library support, see numa(7).
139 Use get_mempolicy(2) with the MPOL_F_MEMS_ALLOWED flag to obtain the
141 set of nodes that are allowed by the current cpuset. Note that this
142 information is subject to change at any time by manual or automatic re‐
143 configuration of the cpuset.
144 Use of this function may result in pages whose location (node) violates
146 the memory policy established for the specified addresses (See
147 mbind(2)) and/or the specified process (See set_mempolicy(2)). That
148 is, memory policy does not constrain the destination nodes used by
149 move_pages().
150 The <numaif.h> header is not included with glibc, but requires in‐
152 stalling libnuma-devel or a similar package.
153
155 get_mempolicy(2), mbind(2), set_mempolicy(2), numa(3), numa_maps(5),
156 cpuset(7), numa(7), migratepages(8), numastat(8)
157
159 This page is part of release 5.13 of the Linux man-pages project. A
160 description of the project, information about reporting bugs, and the
161 latest version of this page, can be found at
162 https://www.kernel.org/doc/man-pages/.
163Linux 2021-03-22 MOVE_PAGES(2)