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"
36 #include "mount-setup.h"
37 #include "mount-util.h"
38 #include "path-util.h"
40 #include "smack-util.h"
42 #include "user-util.h"
46 typedef enum MountMode
{
49 MNT_IN_CONTAINER
= 1 << 1,
52 typedef struct MountPoint
{
58 bool (*condition_fn
)(void);
62 /* The first three entries we might need before SELinux is up. The
63 * fourth (securityfs) is needed by IMA to load a custom policy. The
64 * other ones we can delay until SELinux and IMA are loaded. When
65 * SMACK is enabled we need smackfs, too, so it's a fifth one. */
67 #define N_EARLY_MOUNT 5
69 #define N_EARLY_MOUNT 4
72 static const MountPoint mount_table
[] = {
73 { "sysfs", "/sys", "sysfs", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
74 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
75 { "proc", "/proc", "proc", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
76 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
77 { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID
|MS_STRICTATIME
,
78 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
79 { "securityfs", "/sys/kernel/security", "securityfs", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
82 { "smackfs", "/sys/fs/smackfs", "smackfs", "smackfsdef=*", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
83 mac_smack_use
, MNT_FATAL
},
84 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777,smackfsroot=*", MS_NOSUID
|MS_NODEV
|MS_STRICTATIME
,
85 mac_smack_use
, MNT_FATAL
},
87 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID
|MS_NODEV
|MS_STRICTATIME
,
88 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
89 { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID
), MS_NOSUID
|MS_NOEXEC
,
90 NULL
, MNT_IN_CONTAINER
},
92 { "tmpfs", "/run", "tmpfs", "mode=755,smackfsroot=*", MS_NOSUID
|MS_NODEV
|MS_STRICTATIME
,
93 mac_smack_use
, MNT_FATAL
},
95 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID
|MS_NODEV
|MS_STRICTATIME
,
96 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
97 { "cgroup", "/sys/fs/cgroup", "cgroup2", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
98 cg_is_unified_wanted
, MNT_FATAL
|MNT_IN_CONTAINER
},
99 { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
|MS_STRICTATIME
,
100 cg_is_legacy_wanted
, MNT_FATAL
|MNT_IN_CONTAINER
},
101 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd,xattr", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
102 cg_is_legacy_wanted
, MNT_IN_CONTAINER
},
103 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
104 cg_is_legacy_wanted
, MNT_FATAL
|MNT_IN_CONTAINER
},
105 { "pstore", "/sys/fs/pstore", "pstore", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
108 { "efivarfs", "/sys/firmware/efi/efivars", "efivarfs", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
109 is_efi_boot
, MNT_NONE
},
113 /* These are API file systems that might be mounted by other software,
114 * we just list them here so that we know that we should ignore them */
116 static const char ignore_paths
[] =
117 /* SELinux file systems */
119 /* Container bind mounts */
124 bool mount_point_is_api(const char *path
) {
127 /* Checks if this mount point is considered "API", and hence
128 * should be ignored */
130 for (i
= 0; i
< ELEMENTSOF(mount_table
); i
++)
131 if (path_equal(path
, mount_table
[i
].where
))
134 return path_startswith(path
, "/sys/fs/cgroup/");
137 bool mount_point_ignore(const char *path
) {
140 NULSTR_FOREACH(i
, ignore_paths
)
141 if (path_equal(path
, i
))
147 static int mount_one(const MountPoint
*p
, bool relabel
) {
152 if (p
->condition_fn
&& !p
->condition_fn())
155 /* Relabel first, just in case */
157 (void) label_fix(p
->where
, true, true);
159 r
= path_is_mount_point(p
->where
, AT_SYMLINK_FOLLOW
);
160 if (r
< 0 && r
!= -ENOENT
) {
161 log_full_errno((p
->mode
& MNT_FATAL
) ? LOG_ERR
: LOG_DEBUG
, r
, "Failed to determine whether %s is a mount point: %m", p
->where
);
162 return (p
->mode
& MNT_FATAL
) ? r
: 0;
167 /* Skip securityfs in a container */
168 if (!(p
->mode
& MNT_IN_CONTAINER
) && detect_container() > 0)
171 /* The access mode here doesn't really matter too much, since
172 * the mounted file system will take precedence anyway. */
174 (void) mkdir_p_label(p
->where
, 0755);
176 (void) mkdir_p(p
->where
, 0755);
178 log_debug("Mounting %s to %s of type %s with options %s.",
189 log_full_errno((p
->mode
& MNT_FATAL
) ? LOG_ERR
: LOG_DEBUG
, errno
, "Failed to mount %s at %s: %m", p
->type
, p
->where
);
190 return (p
->mode
& MNT_FATAL
) ? -errno
: 0;
193 /* Relabel again, since we now mounted something fresh here */
195 (void) label_fix(p
->where
, false, false);
200 static int mount_points_setup(unsigned n
, bool loaded_policy
) {
204 for (i
= 0; i
< n
; i
++) {
207 j
= mount_one(mount_table
+ i
, loaded_policy
);
208 if (j
!= 0 && r
>= 0)
215 int mount_setup_early(void) {
216 assert_cc(N_EARLY_MOUNT
<= ELEMENTSOF(mount_table
));
218 /* Do a minimal mount of /proc and friends to enable the most
219 * basic stuff, such as SELinux */
220 return mount_points_setup(N_EARLY_MOUNT
, false);
223 int mount_cgroup_controllers(char ***join_controllers
) {
224 _cleanup_set_free_free_ Set
*controllers
= NULL
;
227 if (!cg_is_legacy_wanted())
230 /* Mount all available cgroup controllers that are built into the kernel. */
232 controllers
= set_new(&string_hash_ops
);
236 r
= cg_kernel_controllers(controllers
);
238 return log_error_errno(r
, "Failed to enumerate cgroup controllers: %m");
241 _cleanup_free_
char *options
= NULL
, *controller
= NULL
, *where
= NULL
;
245 .flags
= MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
246 .mode
= MNT_IN_CONTAINER
,
250 controller
= set_steal_first(controllers
);
254 if (join_controllers
)
255 for (k
= join_controllers
; *k
; k
++)
256 if (strv_find(*k
, controller
))
262 for (i
= *k
, j
= *k
; *i
; i
++) {
264 if (!streq(*i
, controller
)) {
265 _cleanup_free_
char *t
;
267 t
= set_remove(controllers
, *i
);
279 options
= strv_join(*k
, ",");
283 options
= controller
;
287 where
= strappend("/sys/fs/cgroup/", options
);
294 r
= mount_one(&p
, true);
298 if (r
> 0 && k
&& *k
) {
301 for (i
= *k
; *i
; i
++) {
302 _cleanup_free_
char *t
= NULL
;
304 t
= strappend("/sys/fs/cgroup/", *i
);
308 r
= symlink(options
, t
);
310 #ifdef SMACK_RUN_LABEL
311 _cleanup_free_
char *src
;
312 src
= strappend("/sys/fs/cgroup/", options
);
315 r
= mac_smack_copy(t
, src
);
316 if (r
< 0 && r
!= -EOPNOTSUPP
)
317 return log_error_errno(r
, "Failed to copy smack label from %s to %s: %m", src
, t
);
319 } else if (errno
!= EEXIST
)
320 return log_error_errno(errno
, "Failed to create symlink %s: %m", t
);
325 /* Now that we mounted everything, let's make the tmpfs the
326 * cgroup file systems are mounted into read-only. */
327 (void) mount("tmpfs", "/sys/fs/cgroup", "tmpfs", MS_REMOUNT
|MS_NOSUID
|MS_NOEXEC
|MS_NODEV
|MS_STRICTATIME
|MS_RDONLY
, "mode=755");
332 #if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
335 const struct stat
*sb
,
337 struct FTW
*ftwbuf
) {
339 /* No need to label /dev twice in a row... */
340 if (_unlikely_(ftwbuf
->level
== 0))
343 label_fix(fpath
, false, false);
345 /* /run/initramfs is static data and big, no need to
346 * dynamically relabel its contents at boot... */
347 if (_unlikely_(ftwbuf
->level
== 1 &&
349 streq(fpath
, "/run/initramfs")))
350 return FTW_SKIP_SUBTREE
;
356 int mount_setup(bool loaded_policy
) {
359 r
= mount_points_setup(ELEMENTSOF(mount_table
), loaded_policy
);
364 #if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
365 /* Nodes in devtmpfs and /run need to be manually updated for
366 * the appropriate labels, after mounting. The other virtual
367 * API file systems like /sys and /proc do not need that, they
368 * use the same label for all their files. */
370 usec_t before_relabel
, after_relabel
;
371 char timespan
[FORMAT_TIMESPAN_MAX
];
373 before_relabel
= now(CLOCK_MONOTONIC
);
375 nftw("/dev", nftw_cb
, 64, FTW_MOUNT
|FTW_PHYS
|FTW_ACTIONRETVAL
);
376 nftw("/dev/shm", nftw_cb
, 64, FTW_MOUNT
|FTW_PHYS
|FTW_ACTIONRETVAL
);
377 nftw("/run", nftw_cb
, 64, FTW_MOUNT
|FTW_PHYS
|FTW_ACTIONRETVAL
);
379 after_relabel
= now(CLOCK_MONOTONIC
);
381 log_info("Relabelled /dev and /run in %s.",
382 format_timespan(timespan
, sizeof(timespan
), after_relabel
- before_relabel
, 0));
386 /* Create a few default symlinks, which are normally created
387 * by udevd, but some scripts might need them before we start
389 dev_setup(NULL
, UID_INVALID
, GID_INVALID
);
391 /* Mark the root directory as shared in regards to mount
392 * propagation. The kernel defaults to "private", but we think
393 * it makes more sense to have a default of "shared" so that
394 * nspawn and the container tools work out of the box. If
395 * specific setups need other settings they can reset the
396 * propagation mode to private if needed. */
397 if (detect_container() <= 0)
398 if (mount(NULL
, "/", NULL
, MS_REC
|MS_SHARED
, NULL
) < 0)
399 log_warning_errno(errno
, "Failed to set up the root directory for shared mount propagation: %m");
401 /* Create a few directories we always want around, Note that
402 * sd_booted() checks for /run/systemd/system, so this mkdir
403 * really needs to stay for good, otherwise software that
404 * copied sd-daemon.c into their sources will misdetect
406 mkdir_label("/run/systemd", 0755);
407 mkdir_label("/run/systemd/system", 0755);
408 mkdir_label("/run/systemd/inaccessible", 0000);