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 filesystem
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 filesystem in the mount namespace of the calling process.
22 More precisely, it moves the root filesystem to the
23 directory \fIput_old\fP and makes \fInew_root\fP the new root filesystem.
24 The calling process must have the
26 capability in the user namespace that owns the caller's mount namespace.
29 may or may not change the current root and the current
30 working directory of any processes or threads that
31 use the old root directory and which are in
32 the same mount namespace as the caller of
36 should ensure that processes with root or current working directory
37 at the old root operate correctly in either case.
38 An easy way to ensure this is to change their
39 root and current working directory to \fInew_root\fP before invoking
43 may or may not affect the calling process's current working directory.
44 It is therefore recommended to call
45 \fBchdir("/")\fP immediately after
48 The paragraph above is intentionally vague because at the time when
50 was first implemented, it was unclear whether its affect
51 on other process's root and current working directories\(emand
52 the caller's current working directory\(emmight change in the future.
53 However, the behavior has remained consistent since this system call
54 was first implemented:
56 changes the root directory and the current working directory
57 of each process or thread in the same mount namespace to
59 if they point to the old root directory.
63 does not change the caller's current working directory
64 (unless it is on the old root directory),
65 and thus it should be followed by a
66 \fBchdir("/")\fP call.
68 The following restrictions apply:
75 \fInew_root\fP and \fIput_old\fP must not be on the same filesystem as
81 (but can be a bind mounted directory on the current root filesystem).
83 \fIput_old\fP must be at or underneath \fInew_root\fP;
84 that is, adding a nonnegative
85 number of \fI/..\fP to the string pointed to by \fIput_old\fP must yield
86 the same directory as \fInew_root\fP.
89 must be a mount point.
90 (If it is not otherwise a mount point, it suffices to bind mount
94 The propagation type of the parent mount of
96 and the parent mount of the current root directory must not be
100 is an existing mount point, its propagation type must not be
102 These restrictions ensure that
104 never propagates any changes to another mount namespace.
106 The current root directory must be a mount point.
108 On success, zero is returned.
109 On error, \-1 is returned, and
110 \fIerrno\fP is set appropriately.
113 may fail with any of the same errors as
115 Additionally, it may fail with the following errors:
118 .\" Reconfirmed that the following error occurs on Linux 5.0 by
119 .\" specifying 'new_root' as "/rootfs" and 'put_old' as
120 .\" "/rootfs/oldrootfs", and *not* bind mounting "/rootfs" on top of
121 .\" itself. Of course, this is an odd situation, since a later check
122 .\" in the kernel code will in any case yield EINVAL if 'new_root' is
123 .\" not a mount point. However, when the system call was first added,
124 .\" 'new_root' was not required to be a mount point. So, this
125 .\" error is nowadays probably just the result of crufty accumulation.
126 .\" This error can also occur if we bind mount "/" on top of itself
127 .\" and try to specify "/" as the 'new' (again, an odd situation). So,
128 .\" the EBUSY check in the kernel does still seem necessary to prevent
129 .\" that case. Furthermore, the "or put_old" piece is probably
130 .\" redundant text (although the check is in the kernel), since,
131 .\" in another check, 'put_old' is required to be under 'new_root'.
135 is on the current root filesystem.
136 (This error covers the pathological case where
143 is not a mount point.
146 \fIput_old\fP is not underneath \fInew_root\fP.
149 The current root directory is not a mount point
150 (because of an earlier
154 The current root is on the rootfs (initial ramfs) filesystem; see NOTES.
157 Either the mount point at
159 or the parent mount of that mount point,
165 is a mount point and has the propagation type
169 \fInew_root\fP or \fIput_old\fP is not a directory.
172 The calling process does not have the
177 was introduced in Linux 2.3.41.
180 is Linux-specific and hence is not portable.
182 Glibc does not provide a wrapper for this system call; call it using
185 A command-line interface for this system call is provided by
189 allows the caller to switch to a new root filesystem while at the same time
190 placing the old root mount at a location under
192 from where it can subsequently be unmounted.
193 (The fact that it moves all processes that have a root directory
194 or current working directory on the old root filesystem to the
195 new root filesystem frees the old root filesystem of users,
196 allowing it to be unmounted more easily.)
200 is during system startup, when the
201 system mounts a temporary root filesystem (e.g., an \fBinitrd\fP), then
202 mounts the real root filesystem, and eventually turns the latter into
203 the current root of all relevant processes or threads.
204 A modern use is to set up a root filesystem during
205 the creation of a container.
209 modifies process root and current working directories in the
210 manner noted in DESCRIPTION
211 is necessary in order to prevent kernel threads from keeping the old
212 root directory busy with their root and current working directory,
213 even if they never access
214 the filesystem in any way.
219 may be the same directory.
220 In particular, the following sequence allows a pivot-root operation
221 without needing to create and remove a temporary directory:
226 pivot_root(".", ".");
227 umount2(".", MNT_DETACH);
231 This sequence succeeds because the
233 call stacks the old root mount point
235 on top of the new root mount point at
237 At that point, the calling process's root directory and current
238 working directory refer to the new root mount point
240 During the subsequent
246 and then moves up the list of mounts stacked at
252 The rootfs (initial ramfs) cannot be
254 The recommended method of changing the root filesystem in this case is
255 to delete everything in rootfs, overmount rootfs with the new root, attach
256 .IR stdin / stdout / stderr
261 Helper programs for this process exist; see
265 The program below demonstrates the use of
267 inside a mount namespace that is created using
269 After pivoting to the root directory named in the program's
270 first command-line argument, the child created by
272 then executes the program named in the remaining command-line arguments.
274 We demonstrate the program by creating a directory that will serve as
275 the new root filesystem and placing a copy of the (statically linked)
277 executable in that directory.
281 $ \fBmkdir /tmp/rootfs\fP
282 $ \fBls \-id /tmp/rootfs\fP # Show inode number of new root directory
284 $ \fBcp $(which busybox) /tmp/rootfs\fP
285 $ \fBPS1='bbsh$ ' sudo ./pivot_root_demo /tmp/rootfs /busybox sh\fP
287 bbsh$ \fBbusybox ln busybox ln\fP
288 bbsh$ \fBln busybox echo\fP
289 bbsh$ \fBln busybox ls\fP
292 bbsh$ \fBls \-id /\fP # Compare with inode number above
294 bbsh$ \fBecho \(aqhello world\(aq\fP
302 /* pivot_root_demo.c */
309 #include <sys/wait.h>
310 #include <sys/syscall.h>
311 #include <sys/mount.h>
312 #include <sys/stat.h>
315 #define errExit(msg) do { perror(msg); exit(EXIT_FAILURE); \e
319 pivot_root(const char *new_root, const char *put_old)
321 return syscall(SYS_pivot_root, new_root, put_old);
324 #define STACK_SIZE (1024 * 1024)
326 static int /* Startup function for cloned child */
330 char *new_root = args[0];
331 const char *put_old = "/oldrootfs";
334 /* Ensure that \(aqnew_root\(aq and its parent mount don\(aqt have
335 shared propagation (which would cause pivot_root() to
336 return an error), and prevent propagation of mount
337 events to the initial mount namespace */
339 if (mount(NULL, "/", NULL, MS_REC | MS_PRIVATE, NULL) == 1)
340 errExit("mount\-MS_PRIVATE");
342 /* Ensure that \(aqnew_root\(aq is a mount point */
344 if (mount(new_root, new_root, NULL, MS_BIND, NULL) == \-1)
345 errExit("mount\-MS_BIND");
347 /* Create directory to which old root will be pivoted */
349 snprintf(path, sizeof(path), "%s/%s", new_root, put_old);
350 if (mkdir(path, 0777) == \-1)
353 /* And pivot the root filesystem */
355 if (pivot_root(new_root, path) == \-1)
356 errExit("pivot_root");
358 /* Switch the current working directory to "/" */
360 if (chdir("/") == \-1)
363 /* Unmount old root and remove mount point */
365 if (umount2(put_old, MNT_DETACH) == \-1)
367 if (rmdir(put_old) == \-1)
370 /* Execute the command specified in argv[1]... */
372 execv(args[1], &args[1]);
377 main(int argc, char *argv[])
379 /* Create a child process in a new mount namespace */
381 char *stack = malloc(STACK_SIZE);
385 if (clone(child, stack + STACK_SIZE,
386 CLONE_NEWNS | SIGCHLD, &argv[1]) == \-1)
389 /* Parent falls through to here; wait for child */
391 if (wait(NULL) == \-1)
403 .BR mount_namespaces (7),