busybox/util-linux/switch_root.c

/* vi: set sw=4 ts=4: */
/* Copyright 2005 Rob Landley <rob@landley.net>
 *
 * Switch from rootfs to another filesystem as the root of the mount tree.
 *
 * Licensed under GPL version 2, see file LICENSE in this tarball for details.
 */
#include <sys/vfs.h>
#include <sys/mount.h>
#include "libbb.h"
// Make up for header deficiencies
#ifndef RAMFS_MAGIC
# define RAMFS_MAGIC ((unsigned)0x858458f6)
#endif
#ifndef TMPFS_MAGIC
# define TMPFS_MAGIC ((unsigned)0x01021994)
#endif
#ifndef MS_MOVE
# define MS_MOVE     8192
#endif

// Recursively delete contents of rootfs
static void delete_contents(const char *directory, dev_t rootdev)
{
	DIR *dir;
	struct dirent *d;
	struct stat st;

	// Don't descend into other filesystems
	if (lstat(directory, &st) || st.st_dev != rootdev)
		return;

	// Recursively delete the contents of directories
	if (S_ISDIR(st.st_mode)) {
		dir = opendir(directory);
		if (dir) {
			while ((d = readdir(dir))) {
				char *newdir = d->d_name;

				// Skip . and ..
				if (DOT_OR_DOTDOT(newdir))
					continue;

				// Recurse to delete contents
				newdir = concat_path_file(directory, newdir);
				delete_contents(newdir, rootdev);
				free(newdir);
			}
			closedir(dir);

			// Directory should now be empty, zap it
			rmdir(directory);
		}
	} else {
		// It wasn't a directory, zap it
		unlink(directory);
	}
}

int switch_root_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
int switch_root_main(int argc UNUSED_PARAM, char **argv)
{
	char *newroot, *console = NULL;
	struct stat st;
	struct statfs stfs;
	dev_t rootdev;

	// Parse args (-c console)
	opt_complementary = "-2"; // minimum 2 params
	getopt32(argv, "+c:", &console); // '+': stop at first non-option
	argv += optind;
	newroot = *argv++;

	// Change to new root directory and verify it's a different fs
	xchdir(newroot);
	xstat("/", &st);
	rootdev = st.st_dev;
	xstat(".", &st);
	if (st.st_dev == rootdev || getpid() != 1) {
		// Show usage, it says new root must be a mountpoint
		// and we must be PID 1
		bb_show_usage();
	}

	// Additional sanity checks: we're about to rm -rf /, so be REALLY SURE
	// we mean it. I could make this a CONFIG option, but I would get email
	// from all the people who WILL destroy their filesystems.
	if (stat("/init", &st) != 0 || !S_ISREG(st.st_mode)) {
		bb_error_msg_and_die("/init is not a regular file");
	}
	statfs("/", &stfs); // this never fails
	if ((unsigned)stfs.f_type != RAMFS_MAGIC
	 && (unsigned)stfs.f_type != TMPFS_MAGIC
	) {
		bb_error_msg_and_die("root filesystem is not ramfs/tmpfs");
	}

	// Zap everything out of rootdev
	delete_contents("/", rootdev);

	// Overmount / with newdir and chroot into it
	if (mount(".", "/", NULL, MS_MOVE, NULL)) {
		// For example, fails when newroot is not a mountpoint
		bb_perror_msg_and_die("error moving root");
	}
	xchroot(".");
	// The chdir is needed to recalculate "." and ".." links
	xchdir("/");

	// If a new console specified, redirect stdin/stdout/stderr to it
	if (console) {
		close(0);
		xopen(console, O_RDWR);
		xdup2(0, 1);
		xdup2(0, 2);
	}

	// Exec real init
	execv(argv[0], argv);
	bb_perror_msg_and_die("can't execute '%s'", argv[0]);
}

/*
From: Rob Landley <rob@landley.net>
Date: Tue, Jun 16, 2009 at 7:47 PM
Subject: Re: switch_root...

...
...
...

If you're _not_ running out of init_ramfs (if for example you're using initrd
instead), you probably shouldn't use switch_root because it's the wrong tool.

Basically what the sucker does is something like the following shell script:

 find / -xdev | xargs rm -rf
 cd "$1"
 shift
 mount --move . /
 exec chroot . "$@"

There are a couple reasons that won't work as a shell script:

1) If you delete the commands out of your $PATH, your shell scripts can't run
more commands, but you can't start using dynamically linked _new_ commands
until after you do the chroot because the path to the dynamic linker is wrong.
So there's a step that needs to be sort of atomic but can't be as a shell
script.  (You can work around this with static linking or very carefully laid
out paths and sequencing, but it's brittle, ugly, and non-obvious.)

2) The "find | rm" bit will acually delete everything because the mount points
still show up (even if their contents don't), and rm -rf will then happily zap
that.  So the first line is an oversimplification of what you need to do _not_
to descend into other filesystems and delete their contents.

The reason we do this is to free up memory, by the way.  Since initramfs is a
ramfs, deleting its contents frees up the memory it uses.  (We leave it with
one remaining dentry for the new mount point, but that's ok.)

Note that you cannot ever umount rootfs, for approximately the same reason you
can't kill PID 1.  The kernel tracks mount points as a doubly linked list, and
the pointer to the start/end of that list always points to an entry that's
known to be there (rootfs), so it never has to worry about moving that pointer
and it never has to worry about the list being empty.  (Back around 2.6.13
there _was_ a bug that let you umount rootfs, and the system locked hard the
instant you did so endlessly looping to find the end of the mount list and
never stopping.  They fixed it.)

Oh, and the reason we mount --move _and_ do the chroot is due to the way "/"
works.  Each process has two special symlinks, ".", and "/".  Each of them
points to the dentry of a directory, and give you a location paths can start
from.  (Historically ".." was also special, because you could enter a
directory via a symlink so backing out to the directory you came from doesn't
necessarily mean the one physically above where "." points to.  These days I
think it's just handed off to the filesystem.)

Anyway, path resolution starts with "." or "/" (although the "./" at the start
of the path may be implicit), meaning it's relative to one of those two
directories.  Your current directory, and your current root directory.  The
chdir() syscall changes where "." points to, and the chroot() syscall changes
where "/" points to.  (Again, both are per-process which is why chroot only
affects your current process and its child processes.)

Note that chroot() does _not_ change where "." points to, and back before they
put crazy security checks into the kernel your current directory could be
somewhere you could no longer access after the chroot.  (The command line
chroot does a cd as well, the chroot _syscall_ is what I'm talking about.)

The reason mounting something new over / has no obvious effect is the same
reason mounting something over your current directory has no obvious effect:
the . and / links aren't recalculated after a mount, so they still point to
the same dentry they did before, even if that dentry is no longer accessible
by other means.  Note that "cd ." is a NOP, and "chroot /" is a nop; both look
up the cached dentry and set it right back.  They don't re-parse any paths,
because they're what all paths your process uses would be relative to.

That's why the careful sequencing above: we cd into the new mount point before
we do the mount --move.  Moving the mount point would otherwise make it
totally inaccessible to is because cd-ing to the old path wouldn't give it to
us anymore, and cd "/" just gives us the cached dentry from when the process
was created (in this case the old initramfs one).  But the "." symlink gives
us the dentry of the filesystem we just moved, so we can then "chroot ." to
copy that dentry to "/" and get the new filesystem.  If we _didn't_ save that
dentry in "." we couldn't get it back after the mount --move.

(Yes, this is all screwy and I had to email questions to Linus Torvalds to get
it straight myself.  I keep meaning to write up a "how mount actually works"
document someday...)
*/
1	/* vi: set sw=4 ts=4: */
2	/* Copyright 2005 Rob Landley <rob@landley.net>
3	*
4	* Switch from rootfs to another filesystem as the root of the mount tree.
5	*
6	* Licensed under GPL version 2, see file LICENSE in this tarball for details.
7	*/
8	#include <sys/vfs.h>
9	#include <sys/mount.h>
10	#include "libbb.h"
11	// Make up for header deficiencies
12	#ifndef RAMFS_MAGIC
13	# define RAMFS_MAGIC ((unsigned)0x858458f6)
14	#endif
15	#ifndef TMPFS_MAGIC
16	# define TMPFS_MAGIC ((unsigned)0x01021994)
17	#endif
18	#ifndef MS_MOVE
19	# define MS_MOVE 8192
20	#endif
21
22	// Recursively delete contents of rootfs
23	static void delete_contents(const char *directory, dev_t rootdev)
24	{
25	DIR *dir;
26	struct dirent *d;
27	struct stat st;
28
29	// Don't descend into other filesystems
30	if (lstat(directory, &st) \|\| st.st_dev != rootdev)
31	return;
32
33	// Recursively delete the contents of directories
34	if (S_ISDIR(st.st_mode)) {
35	dir = opendir(directory);
36	if (dir) {
37	while ((d = readdir(dir))) {
38	char *newdir = d->d_name;
39
40	// Skip . and ..
41	if (DOT_OR_DOTDOT(newdir))
42	continue;
43
44	// Recurse to delete contents
45	newdir = concat_path_file(directory, newdir);
46	delete_contents(newdir, rootdev);
47	free(newdir);
48	}
49	closedir(dir);
50
51	// Directory should now be empty, zap it
52	rmdir(directory);
53	}
54	} else {
55	// It wasn't a directory, zap it
56	unlink(directory);
57	}
58	}
59
60	int switch_root_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
61	int switch_root_main(int argc UNUSED_PARAM, char **argv)
62	{
63	char newroot, console = NULL;
64	struct stat st;
65	struct statfs stfs;
66	dev_t rootdev;
67
68	// Parse args (-c console)
69	opt_complementary = "-2"; // minimum 2 params
70	getopt32(argv, "+c:", &console); // '+': stop at first non-option
71	argv += optind;
72	newroot = *argv++;
73
74	// Change to new root directory and verify it's a different fs
75	xchdir(newroot);
76	xstat("/", &st);
77	rootdev = st.st_dev;
78	xstat(".", &st);
79	if (st.st_dev == rootdev \|\| getpid() != 1) {
80	// Show usage, it says new root must be a mountpoint
81	// and we must be PID 1
82	bb_show_usage();
83	}
84
85	// Additional sanity checks: we're about to rm -rf /, so be REALLY SURE
86	// we mean it. I could make this a CONFIG option, but I would get email
87	// from all the people who WILL destroy their filesystems.
88	if (stat("/init", &st) != 0 \|\| !S_ISREG(st.st_mode)) {
89	bb_error_msg_and_die("/init is not a regular file");
90	}
91	statfs("/", &stfs); // this never fails
92	if ((unsigned)stfs.f_type != RAMFS_MAGIC
93	&& (unsigned)stfs.f_type != TMPFS_MAGIC
94	) {
95	bb_error_msg_and_die("root filesystem is not ramfs/tmpfs");
96	}
97
98	// Zap everything out of rootdev
99	delete_contents("/", rootdev);
100
101	// Overmount / with newdir and chroot into it
102	if (mount(".", "/", NULL, MS_MOVE, NULL)) {
103	// For example, fails when newroot is not a mountpoint
104	bb_perror_msg_and_die("error moving root");
105	}
106	xchroot(".");
107	// The chdir is needed to recalculate "." and ".." links
108	xchdir("/");
109
110	// If a new console specified, redirect stdin/stdout/stderr to it
111	if (console) {
112	close(0);
113	xopen(console, O_RDWR);
114	xdup2(0, 1);
115	xdup2(0, 2);
116	}
117
118	// Exec real init
119	execv(argv[0], argv);
120	bb_perror_msg_and_die("can't execute '%s'", argv[0]);
121	}
122
123	/*
124	From: Rob Landley <rob@landley.net>
125	Date: Tue, Jun 16, 2009 at 7:47 PM
126	Subject: Re: switch_root...
127
128	...
129	...
130	...
131
132	If you're _not_ running out of init_ramfs (if for example you're using initrd
133	instead), you probably shouldn't use switch_root because it's the wrong tool.
134
135	Basically what the sucker does is something like the following shell script:
136
137	find / -xdev \| xargs rm -rf
138	cd "$1"
139	shift
140	mount --move . /
141	exec chroot . "$@"
142
143	There are a couple reasons that won't work as a shell script:
144
145	1) If you delete the commands out of your $PATH, your shell scripts can't run
146	more commands, but you can't start using dynamically linked _new_ commands
147	until after you do the chroot because the path to the dynamic linker is wrong.
148	So there's a step that needs to be sort of atomic but can't be as a shell
149	script. (You can work around this with static linking or very carefully laid
150	out paths and sequencing, but it's brittle, ugly, and non-obvious.)
151
152	2) The "find \| rm" bit will acually delete everything because the mount points
153	still show up (even if their contents don't), and rm -rf will then happily zap
154	that. So the first line is an oversimplification of what you need to do _not_
155	to descend into other filesystems and delete their contents.
156
157	The reason we do this is to free up memory, by the way. Since initramfs is a
158	ramfs, deleting its contents frees up the memory it uses. (We leave it with
159	one remaining dentry for the new mount point, but that's ok.)
160
161	Note that you cannot ever umount rootfs, for approximately the same reason you
162	can't kill PID 1. The kernel tracks mount points as a doubly linked list, and
163	the pointer to the start/end of that list always points to an entry that's
164	known to be there (rootfs), so it never has to worry about moving that pointer
165	and it never has to worry about the list being empty. (Back around 2.6.13
166	there _was_ a bug that let you umount rootfs, and the system locked hard the
167	instant you did so endlessly looping to find the end of the mount list and
168	never stopping. They fixed it.)
169
170	Oh, and the reason we mount --move _and_ do the chroot is due to the way "/"
171	works. Each process has two special symlinks, ".", and "/". Each of them
172	points to the dentry of a directory, and give you a location paths can start
173	from. (Historically ".." was also special, because you could enter a
174	directory via a symlink so backing out to the directory you came from doesn't
175	necessarily mean the one physically above where "." points to. These days I
176	think it's just handed off to the filesystem.)
177
178	Anyway, path resolution starts with "." or "/" (although the "./" at the start
179	of the path may be implicit), meaning it's relative to one of those two
180	directories. Your current directory, and your current root directory. The
181	chdir() syscall changes where "." points to, and the chroot() syscall changes
182	where "/" points to. (Again, both are per-process which is why chroot only
183	affects your current process and its child processes.)
184
185	Note that chroot() does _not_ change where "." points to, and back before they
186	put crazy security checks into the kernel your current directory could be
187	somewhere you could no longer access after the chroot. (The command line
188	chroot does a cd as well, the chroot _syscall_ is what I'm talking about.)
189
190	The reason mounting something new over / has no obvious effect is the same
191	reason mounting something over your current directory has no obvious effect:
192	the . and / links aren't recalculated after a mount, so they still point to
193	the same dentry they did before, even if that dentry is no longer accessible
194	by other means. Note that "cd ." is a NOP, and "chroot /" is a nop; both look
195	up the cached dentry and set it right back. They don't re-parse any paths,
196	because they're what all paths your process uses would be relative to.
197
198	That's why the careful sequencing above: we cd into the new mount point before
199	we do the mount --move. Moving the mount point would otherwise make it
200	totally inaccessible to is because cd-ing to the old path wouldn't give it to
201	us anymore, and cd "/" just gives us the cached dentry from when the process
202	was created (in this case the old initramfs one). But the "." symlink gives
203	us the dentry of the filesystem we just moved, so we can then "chroot ." to
204	copy that dentry to "/" and get the new filesystem. If we _didn't_ save that
205	dentry in "." we couldn't get it back after the mount --move.
206
207	(Yes, this is all screwy and I had to email questions to Linus Torvalds to get
208	it straight myself. I keep meaning to write up a "how mount actually works"
209	document someday...)
210	*/