1 .\" Copyright (C) 2011, Eric Biederman <ebiederm@xmission.com>
2 .\" and Copyright (C) 2011, 2012, Michael Kerrisk <mtk.manpages@gamil.com>
4 .\" %%%LICENSE_START(GPLv2_ONELINE)
5 .\" Licensed under the GPLv2
8 .TH SETNS 2 2013-01-01 "Linux" "Linux Programmer's Manual"
10 setns \- reassociate thread with a namespace
13 .BR "#define _GNU_SOURCE" " /* See feature_test_macros(7) */"
16 .BI "int setns(int " fd ", int " nstype );
19 Given a file descriptor referring to a namespace,
20 reassociate the calling thread with that namespace.
24 argument is a file descriptor referring to one of the namespace entries in a
28 for further information on
30 The calling thread will be reassociated with the corresponding namespace,
31 subject to any constraints imposed by the
37 argument specifies which type of namespace
38 the calling thread may be reassociated with.
39 This argument can have one of the following values:
42 Allow any type of namespace to be joined.
44 .BR CLONE_NEWIPC " (since Linux 3.0)"
46 must refer to an IPC namespace.
48 .BR CLONE_NEWNET " (since Linux 3.0)"
50 must refer to a network namespace.
52 .BR CLONE_NEWNS " (since Linux 3.8)"
54 must refer to a mount namespace.
56 .BR CLONE_NEWPID " (since Linux 3.8)"
58 must refer to a PID namespace.
60 .BR CLONE_NEWUSER " (since Linux 3.8)"
62 must refer to a user namespace.
64 .BR CLONE_NEWUTS " (since Linux 3.0)"
66 must refer to a UTS namespace.
70 as 0 suffices if the caller knows (or does not care)
71 what type of namespace is referred to by
73 Specifying a nonzero value for
75 is useful if the caller does not know what type of namespace is referred to by
77 and wants to ensure that the namespace is of a particular type.
78 (The caller might not know the type of the namespace referred to by
80 if the file descriptor was opened by another process and, for example,
81 passed to the caller via a UNIX domain socket.)
84 behaves somewhat differently from the other
87 reassociating the calling thread with a PID namespace only changes
88 the PID namespace that child processes of the caller will be created in;
89 it does not change the PID namespace of the caller itself.
90 Reassociating with a PID namespace is only allowed if the
91 PID namespace specified by
93 is a descendant (child, grandchild, etc.)
94 of the PID namespace of the caller.
95 For further details on PID namespaces, see
96 .BR user_namespaces (7).
98 A process reassociating itself with a user namespace must have the
100 .\" See kernel/user_namespace.c:userns_install() [3.8 source]
101 capability in the target user namespace.
102 Upon successfully joining a user namespace,
103 a process is granted all capabilities in that namespace,
104 regardless of its user and group IDs.
105 A multithreaded process may not change user namespace with
107 It is not permitted to use
109 to reenter the caller's current user namespace.
110 This prevents a caller that has dropped capabilities from regaining
111 those capabilities via a call to
113 For security reasons,
114 .\" commit e66eded8309ebf679d3d3c1f5820d1f2ca332c71
115 .\" https://lwn.net/Articles/543273/
116 a process can't join a new user namespace if it is sharing
117 filesystem-related attributes
118 (the attributes whose sharing is controlled by the
121 flag) with another process.
122 For further details on user namespaces, see
123 .BR user_namespaces (7).
125 A process may not be reassociated with a new mount namespace if it is
127 .\" Above check is in fs/namespace.c:mntns_install() [3.8 source]
128 Changing the mount namespace requires that the caller possess both
132 capabilities in its own user namespace and
134 in the target mount namespace.
139 On failure, \-1 is returned and
141 is set to indicate the error.
146 is not a valid file descriptor.
150 refers to a namespace whose type does not match that specified in
154 There is problem with reassociating
155 the thread with the specified namespace.
158 The caller attempted to join the user namespace
159 in which it is already a member.
162 Cannot allocate sufficient memory to change the specified namespace.
165 The calling thread did not have the required capability
170 system call first appeared in Linux in kernel 3.0;
171 library support was added to glibc in version 2.14.
175 system call is Linux-specific.
177 Not all of the attributes that can be shared when
178 a new thread is created using
183 The program below takes two or more arguments.
184 The first argument specifies the pathname of a namespace file in an existing
187 The remaining arguments specify a command and its arguments.
188 The program opens the namespace file, joins that namespace using
190 and executes the specified command inside that namespace.
192 The following shell session demonstrates the use of this program
193 (compiled as a binary named
195 in conjunction with the
197 example program in the
199 man page (complied as a binary named
202 We begin by executing the example program in
205 That program creates a child in a separate UTS namespace.
206 The child changes the hostname in its namespace,
207 and then both processes display the hostnames in their UTS namespaces,
208 so that we can see that they are different.
212 $ \fBsu\fP # Need privilege for namespace operations
214 # \fB./newuts bizarro &\fP
216 clone() returned 3550
217 uts.nodename in child: bizarro
218 uts.nodename in parent: antero
219 # \fBuname \-n\fP # Verify hostname in the shell
224 We then run the program shown below,
225 using it to execute a shell.
226 Inside that shell, we verify that the hostname is the one
227 set by the child created by the first program:
231 # \fB./ns_exec /proc/3550/ns/uts /bin/bash\fP
232 # \fBuname \-n\fP # Executed in shell started by ns_exec
245 #define errExit(msg) do { perror(msg); exit(EXIT_FAILURE); \\
249 main(int argc, char *argv[])
254 fprintf(stderr, "%s /proc/PID/ns/FILE cmd args...\\n", argv[0]);
258 fd = open(argv[1], O_RDONLY); /* Get descriptor for namespace */
262 if (setns(fd, 0) == \-1) /* Join that namespace */
265 execvp(argv[2], &argv[2]); /* Execute a command in namespace */