1 .\" Copyright (C) 2000 by Werner Almesberger
2 .\" and Copyright (C) 2019 Michael Kerrisk <mtk.manpages@gmail.com>
4 .\" %%%LICENSE_START(GPL_NOVERSION_ONELINE)
5 .\" May be distributed under GPL
8 .\" Written 2000-02-23 by Werner Almesberger
9 .\" Modified 2004-06-17 Michael Kerrisk <mtk.manpages@gmail.com>
11 .TH PIVOT_ROOT 2 2019-08-02 "Linux" "Linux Programmer's Manual"
13 pivot_root \- change the root mount
15 .BI "int pivot_root(const char *" new_root ", const char *" put_old );
18 There is no glibc wrapper for this system call; see NOTES.
21 changes the root mount in the mount namespace of the calling process.
22 More precisely, it moves the root mount to the
23 directory \fIput_old\fP and makes \fInew_root\fP the new root mount.
24 The calling process must have the
26 capability in the user namespace that owns the caller's mount namespace.
29 changes the root directory and the current working directory
30 of each process or thread in the same mount namespace to
32 if they point to the old root directory.
36 does not change the caller's current working directory
37 (unless it is on the old root directory),
38 and thus it should be followed by a
39 \fBchdir("/")\fP call.
41 The following restrictions apply:
51 must not be on the same mount as the current root.
53 \fIput_old\fP must be at or underneath \fInew_root\fP;
54 that is, adding a nonnegative
55 number of \fI/..\fP to the string pointed to by \fIput_old\fP must yield
56 the same directory as \fInew_root\fP.
59 must be a path to a mount point, but can't be
61 A path that is not already a mount point can be converted into one by
62 bind mounting the path onto itself.
64 The propagation type of the parent mount of
66 and the parent mount of the current root directory must not be
70 is an existing mount point, its propagation type must not be
72 These restrictions ensure that
74 never propagates any changes to another mount namespace.
76 The current root directory must be a mount point.
78 On success, zero is returned.
79 On error, \-1 is returned, and
80 \fIerrno\fP is set appropriately.
83 may fail with any of the same errors as
85 Additionally, it may fail with the following errors:
88 .\" Reconfirmed that the following error occurs on Linux 5.0 by
89 .\" specifying 'new_root' as "/rootfs" and 'put_old' as
90 .\" "/rootfs/oldrootfs", and *not* bind mounting "/rootfs" on top of
91 .\" itself. Of course, this is an odd situation, since a later check
92 .\" in the kernel code will in any case yield EINVAL if 'new_root' is
93 .\" not a mount point. However, when the system call was first added,
94 .\" 'new_root' was not required to be a mount point. So, this
95 .\" error is nowadays probably just the result of crufty accumulation.
96 .\" This error can also occur if we bind mount "/" on top of itself
97 .\" and try to specify "/" as the 'new' (again, an odd situation). So,
98 .\" the EBUSY check in the kernel does still seem necessary to prevent
99 .\" that case. Furthermore, the "or put_old" piece is probably
100 .\" redundant text (although the check is in the kernel), since,
101 .\" in another check, 'put_old' is required to be under 'new_root'.
105 is on the current root filesystem.
106 (This error covers the pathological case where
113 is not a mount point.
116 \fIput_old\fP is not underneath \fInew_root\fP.
119 The current root directory is not a mount point
120 (because of an earlier
124 The current root is on the rootfs (initial ramfs) filesystem; see NOTES.
127 Either the mount point at
129 or the parent mount of that mount point,
135 is a mount point and has the propagation type
139 \fInew_root\fP or \fIput_old\fP is not a directory.
142 The calling process does not have the
147 was introduced in Linux 2.3.41.
150 is Linux-specific and hence is not portable.
152 Glibc does not provide a wrapper for this system call; call it using
155 A command-line interface for this system call is provided by
159 allows the caller to switch to a new root filesystem while at the same time
160 placing the old root mount at a location under
162 from where it can subsequently be unmounted.
163 (The fact that it moves all processes that have a root directory
164 or current working directory on the old root directory to the
165 new root frees the old root directory of users,
166 allowing the old root mount to be unmounted more easily.)
170 is during system startup, when the
171 system mounts a temporary root filesystem (e.g., an \fBinitrd\fP), then
172 mounts the real root filesystem, and eventually turns the latter into
173 the current root of all relevant processes or threads.
174 A modern use is to set up a root filesystem during
175 the creation of a container.
179 modifies process root and current working directories in the
180 manner noted in DESCRIPTION
181 is necessary in order to prevent kernel threads from keeping the old
182 root directory busy with their root and current working directory,
183 even if they never access
184 the filesystem in any way.
186 The rootfs (initial ramfs) cannot be
188 The recommended method of changing the root filesystem in this case is
189 to delete everything in rootfs, overmount rootfs with the new root, attach
190 .IR stdin / stdout / stderr
195 Helper programs for this process exist; see
198 .SS pivot_root(\(dq.\(dq, \(dq.\(dq)
203 may be the same directory.
204 In particular, the following sequence allows a pivot-root operation
205 without needing to create and remove a temporary directory:
210 pivot_root(".", ".");
211 umount2(".", MNT_DETACH);
215 This sequence succeeds because the
217 call stacks the old root mount point
218 on top of the new root mount point at
220 At that point, the calling process's root directory and current
221 working directory refer to the new root mount point
223 During the subsequent
229 and then moves up the list of mounts stacked at
231 with the result that old root mount point is unmounted.
234 For many years, this manual page carried the following text:
238 may or may not change the current root and the current
239 working directory of any processes or threads which use the old
243 must ensure that processes with root or current working directory
244 at the old root operate correctly in either case.
245 An easy way to ensure this is to change their
246 root and current working directory to \fInew_root\fP before invoking
250 This text, written before the system call implementation was
251 even finalized in the kernel, was probably intended to warn users
252 at that time that the implementation might change before final release.
253 However, the behavior stated in DESCRIPTION
254 has remained consistent since this system call
255 was first implemented and will not change now.
258 .\" Would it be better, because simpler, to use unshare(2)
259 .\" rather than clone(2) in the example below?
261 The program below demonstrates the use of
263 inside a mount namespace that is created using
265 After pivoting to the root directory named in the program's
266 first command-line argument, the child created by
268 then executes the program named in the remaining command-line arguments.
270 We demonstrate the program by creating a directory that will serve as
271 the new root filesystem and placing a copy of the (statically linked)
273 executable in that directory.
277 $ \fBmkdir /tmp/rootfs\fP
278 $ \fBls \-id /tmp/rootfs\fP # Show inode number of new root directory
280 $ \fBcp $(which busybox) /tmp/rootfs\fP
281 $ \fBPS1='bbsh$ ' sudo ./pivot_root_demo /tmp/rootfs /busybox sh\fP
283 bbsh$ \fBbusybox ln busybox ln\fP
284 bbsh$ \fBln busybox echo\fP
285 bbsh$ \fBln busybox ls\fP
288 bbsh$ \fBls \-id /\fP # Compare with inode number above
290 bbsh$ \fBecho \(aqhello world\(aq\fP
298 /* pivot_root_demo.c */
305 #include <sys/wait.h>
306 #include <sys/syscall.h>
307 #include <sys/mount.h>
308 #include <sys/stat.h>
311 #define errExit(msg) do { perror(msg); exit(EXIT_FAILURE); \e
315 pivot_root(const char *new_root, const char *put_old)
317 return syscall(SYS_pivot_root, new_root, put_old);
320 #define STACK_SIZE (1024 * 1024)
322 static int /* Startup function for cloned child */
326 char *new_root = args[0];
327 const char *put_old = "/oldrootfs";
330 /* Ensure that \(aqnew_root\(aq and its parent mount don\(aqt have
331 shared propagation (which would cause pivot_root() to
332 return an error), and prevent propagation of mount
333 events to the initial mount namespace */
335 if (mount(NULL, "/", NULL, MS_REC | MS_PRIVATE, NULL) == 1)
336 errExit("mount\-MS_PRIVATE");
338 /* Ensure that \(aqnew_root\(aq is a mount point */
340 if (mount(new_root, new_root, NULL, MS_BIND, NULL) == \-1)
341 errExit("mount\-MS_BIND");
343 /* Create directory to which old root will be pivoted */
345 snprintf(path, sizeof(path), "%s/%s", new_root, put_old);
346 if (mkdir(path, 0777) == \-1)
349 /* And pivot the root filesystem */
351 if (pivot_root(new_root, path) == \-1)
352 errExit("pivot_root");
354 /* Switch the current working directory to "/" */
356 if (chdir("/") == \-1)
359 /* Unmount old root and remove mount point */
361 if (umount2(put_old, MNT_DETACH) == \-1)
363 if (rmdir(put_old) == \-1)
366 /* Execute the command specified in argv[1]... */
368 execv(args[1], &args[1]);
373 main(int argc, char *argv[])
375 /* Create a child process in a new mount namespace */
377 char *stack = malloc(STACK_SIZE);
381 if (clone(child, stack + STACK_SIZE,
382 CLONE_NEWNS | SIGCHLD, &argv[1]) == \-1)
385 /* Parent falls through to here; wait for child */
387 if (wait(NULL) == \-1)
399 .BR mount_namespaces (7),