2 This file is part of systemd.
4 Copyright 2010 Lennart Poettering
6 systemd is free software; you can redistribute it and/or modify it
7 under the terms of the GNU Lesser General Public License as published by
8 the Free Software Foundation; either version 2.1 of the License, or
9 (at your option) any later version.
11 systemd is distributed in the hope that it will be useful, but
12 WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
16 You should have received a copy of the GNU Lesser General Public License
17 along with systemd; If not, see <http://www.gnu.org/licenses/>.
23 #include <sys/mount.h>
26 #include "alloc-util.h"
28 #include "cgroup-util.h"
29 #include "dev-setup.h"
37 #include "mount-setup.h"
38 #include "mount-util.h"
39 #include "path-util.h"
41 #include "smack-util.h"
43 #include "user-util.h"
47 typedef enum MountMode
{
50 MNT_IN_CONTAINER
= 1 << 1,
53 typedef struct MountPoint
{
59 bool (*condition_fn
)(void);
63 /* The first three entries we might need before SELinux is up. The
64 * fourth (securityfs) is needed by IMA to load a custom policy. The
65 * other ones we can delay until SELinux and IMA are loaded. When
66 * SMACK is enabled we need smackfs, too, so it's a fifth one. */
68 #define N_EARLY_MOUNT 5
70 #define N_EARLY_MOUNT 4
73 static const MountPoint mount_table
[] = {
74 { "sysfs", "/sys", "sysfs", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
75 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
76 { "proc", "/proc", "proc", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
77 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
78 { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID
|MS_STRICTATIME
,
79 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
80 { "securityfs", "/sys/kernel/security", "securityfs", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
83 { "smackfs", "/sys/fs/smackfs", "smackfs", "smackfsdef=*", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
84 mac_smack_use
, MNT_FATAL
},
85 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777,smackfsroot=*", MS_NOSUID
|MS_NODEV
|MS_STRICTATIME
,
86 mac_smack_use
, MNT_FATAL
},
88 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID
|MS_NODEV
|MS_STRICTATIME
,
89 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
90 { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID
), MS_NOSUID
|MS_NOEXEC
,
91 NULL
, MNT_IN_CONTAINER
},
93 { "tmpfs", "/run", "tmpfs", "mode=755,smackfsroot=*", MS_NOSUID
|MS_NODEV
|MS_STRICTATIME
,
94 mac_smack_use
, MNT_FATAL
},
96 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID
|MS_NODEV
|MS_STRICTATIME
,
97 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
98 { "cgroup", "/sys/fs/cgroup", "cgroup2", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
99 cg_is_unified_wanted
, MNT_IN_CONTAINER
},
100 { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
|MS_STRICTATIME
,
101 cg_is_legacy_wanted
, MNT_FATAL
|MNT_IN_CONTAINER
},
102 { "cgroup", "/sys/fs/cgroup/unified", "cgroup2", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
103 cg_is_hybrid_wanted
, MNT_IN_CONTAINER
},
104 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd,xattr", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
105 cg_is_legacy_wanted
, MNT_IN_CONTAINER
},
106 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
107 cg_is_legacy_wanted
, MNT_FATAL
|MNT_IN_CONTAINER
},
108 { "pstore", "/sys/fs/pstore", "pstore", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
111 { "efivarfs", "/sys/firmware/efi/efivars", "efivarfs", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
112 is_efi_boot
, MNT_NONE
},
116 /* These are API file systems that might be mounted by other software,
117 * we just list them here so that we know that we should ignore them */
119 static const char ignore_paths
[] =
120 /* SELinux file systems */
122 /* Container bind mounts */
127 bool mount_point_is_api(const char *path
) {
130 /* Checks if this mount point is considered "API", and hence
131 * should be ignored */
133 for (i
= 0; i
< ELEMENTSOF(mount_table
); i
++)
134 if (path_equal(path
, mount_table
[i
].where
))
137 return path_startswith(path
, "/sys/fs/cgroup/");
140 bool mount_point_ignore(const char *path
) {
143 NULSTR_FOREACH(i
, ignore_paths
)
144 if (path_equal(path
, i
))
150 static int mount_one(const MountPoint
*p
, bool relabel
) {
155 if (p
->condition_fn
&& !p
->condition_fn())
158 /* Relabel first, just in case */
160 (void) label_fix(p
->where
, true, true);
162 r
= path_is_mount_point(p
->where
, NULL
, AT_SYMLINK_FOLLOW
);
163 if (r
< 0 && r
!= -ENOENT
) {
164 log_full_errno((p
->mode
& MNT_FATAL
) ? LOG_ERR
: LOG_DEBUG
, r
, "Failed to determine whether %s is a mount point: %m", p
->where
);
165 return (p
->mode
& MNT_FATAL
) ? r
: 0;
170 /* Skip securityfs in a container */
171 if (!(p
->mode
& MNT_IN_CONTAINER
) && detect_container() > 0)
174 /* The access mode here doesn't really matter too much, since
175 * the mounted file system will take precedence anyway. */
177 (void) mkdir_p_label(p
->where
, 0755);
179 (void) mkdir_p(p
->where
, 0755);
181 log_debug("Mounting %s to %s of type %s with options %s.",
192 log_full_errno((p
->mode
& MNT_FATAL
) ? LOG_ERR
: LOG_DEBUG
, errno
, "Failed to mount %s at %s: %m", p
->type
, p
->where
);
193 return (p
->mode
& MNT_FATAL
) ? -errno
: 0;
196 /* Relabel again, since we now mounted something fresh here */
198 (void) label_fix(p
->where
, false, false);
203 static int mount_points_setup(unsigned n
, bool loaded_policy
) {
207 for (i
= 0; i
< n
; i
++) {
210 j
= mount_one(mount_table
+ i
, loaded_policy
);
211 if (j
!= 0 && r
>= 0)
218 int mount_setup_early(void) {
219 assert_cc(N_EARLY_MOUNT
<= ELEMENTSOF(mount_table
));
221 /* Do a minimal mount of /proc and friends to enable the most
222 * basic stuff, such as SELinux */
223 return mount_points_setup(N_EARLY_MOUNT
, false);
226 int mount_cgroup_controllers(char ***join_controllers
) {
227 _cleanup_set_free_free_ Set
*controllers
= NULL
;
230 if (!cg_is_legacy_wanted())
233 /* Mount all available cgroup controllers that are built into the kernel. */
235 controllers
= set_new(&string_hash_ops
);
239 r
= cg_kernel_controllers(controllers
);
241 return log_error_errno(r
, "Failed to enumerate cgroup controllers: %m");
244 _cleanup_free_
char *options
= NULL
, *controller
= NULL
, *where
= NULL
;
248 .flags
= MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
249 .mode
= MNT_IN_CONTAINER
,
253 controller
= set_steal_first(controllers
);
257 if (join_controllers
)
258 for (k
= join_controllers
; *k
; k
++)
259 if (strv_find(*k
, controller
))
265 for (i
= *k
, j
= *k
; *i
; i
++) {
267 if (!streq(*i
, controller
)) {
268 _cleanup_free_
char *t
;
270 t
= set_remove(controllers
, *i
);
282 options
= strv_join(*k
, ",");
286 options
= controller
;
290 where
= strappend("/sys/fs/cgroup/", options
);
297 r
= mount_one(&p
, true);
301 if (r
> 0 && k
&& *k
) {
304 for (i
= *k
; *i
; i
++) {
305 _cleanup_free_
char *t
= NULL
;
307 t
= strappend("/sys/fs/cgroup/", *i
);
311 r
= symlink(options
, t
);
313 #ifdef SMACK_RUN_LABEL
314 _cleanup_free_
char *src
;
315 src
= strappend("/sys/fs/cgroup/", options
);
318 r
= mac_smack_copy(t
, src
);
319 if (r
< 0 && r
!= -EOPNOTSUPP
)
320 return log_error_errno(r
, "Failed to copy smack label from %s to %s: %m", src
, t
);
322 } else if (errno
!= EEXIST
)
323 return log_error_errno(errno
, "Failed to create symlink %s: %m", t
);
328 /* Now that we mounted everything, let's make the tmpfs the
329 * cgroup file systems are mounted into read-only. */
330 (void) mount("tmpfs", "/sys/fs/cgroup", "tmpfs", MS_REMOUNT
|MS_NOSUID
|MS_NOEXEC
|MS_NODEV
|MS_STRICTATIME
|MS_RDONLY
, "mode=755");
335 #if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
338 const struct stat
*sb
,
340 struct FTW
*ftwbuf
) {
342 /* No need to label /dev twice in a row... */
343 if (_unlikely_(ftwbuf
->level
== 0))
346 label_fix(fpath
, false, false);
348 /* /run/initramfs is static data and big, no need to
349 * dynamically relabel its contents at boot... */
350 if (_unlikely_(ftwbuf
->level
== 1 &&
352 streq(fpath
, "/run/initramfs")))
353 return FTW_SKIP_SUBTREE
;
359 int mount_setup(bool loaded_policy
) {
362 r
= mount_points_setup(ELEMENTSOF(mount_table
), loaded_policy
);
366 #if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
367 /* Nodes in devtmpfs and /run need to be manually updated for
368 * the appropriate labels, after mounting. The other virtual
369 * API file systems like /sys and /proc do not need that, they
370 * use the same label for all their files. */
372 usec_t before_relabel
, after_relabel
;
373 char timespan
[FORMAT_TIMESPAN_MAX
];
375 before_relabel
= now(CLOCK_MONOTONIC
);
377 nftw("/dev", nftw_cb
, 64, FTW_MOUNT
|FTW_PHYS
|FTW_ACTIONRETVAL
);
378 nftw("/dev/shm", nftw_cb
, 64, FTW_MOUNT
|FTW_PHYS
|FTW_ACTIONRETVAL
);
379 nftw("/run", nftw_cb
, 64, FTW_MOUNT
|FTW_PHYS
|FTW_ACTIONRETVAL
);
381 after_relabel
= now(CLOCK_MONOTONIC
);
383 log_info("Relabelled /dev and /run in %s.",
384 format_timespan(timespan
, sizeof(timespan
), after_relabel
- before_relabel
, 0));
388 /* Create a few default symlinks, which are normally created
389 * by udevd, but some scripts might need them before we start
391 dev_setup(NULL
, UID_INVALID
, GID_INVALID
);
393 /* Mark the root directory as shared in regards to mount propagation. The kernel defaults to "private", but we
394 * think it makes more sense to have a default of "shared" so that nspawn and the container tools work out of
395 * the box. If specific setups need other settings they can reset the propagation mode to private if
396 * needed. Note that we set this only when we are invoked directly by the kernel. If we are invoked by a
397 * container manager we assume the container manager knows what it is doing (for example, because it set up
398 * some directories with different propagation modes). */
399 if (detect_container() <= 0)
400 if (mount(NULL
, "/", NULL
, MS_REC
|MS_SHARED
, NULL
) < 0)
401 log_warning_errno(errno
, "Failed to set up the root directory for shared mount propagation: %m");
403 /* Create a few directories we always want around, Note that sd_booted() checks for /run/systemd/system, so
404 * this mkdir really needs to stay for good, otherwise software that copied sd-daemon.c into their sources will
405 * misdetect systemd. */
406 (void) mkdir_label("/run/systemd", 0755);
407 (void) mkdir_label("/run/systemd/system", 0755);
409 /* Set up inaccessible items */
410 (void) mkdir_label("/run/systemd/inaccessible", 0000);
411 (void) mknod("/run/systemd/inaccessible/reg", S_IFREG
| 0000, 0);
412 (void) mkdir_label("/run/systemd/inaccessible/dir", 0000);
413 (void) mknod("/run/systemd/inaccessible/chr", S_IFCHR
| 0000, makedev(0, 0));
414 (void) mknod("/run/systemd/inaccessible/blk", S_IFBLK
| 0000, makedev(0, 0));
415 (void) mkfifo("/run/systemd/inaccessible/fifo", 0000);
416 (void) mknod("/run/systemd/inaccessible/sock", S_IFSOCK
| 0000, 0);