1 /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
4 This file is part of systemd.
6 Copyright 2010 Lennart Poettering
8 systemd is free software; you can redistribute it and/or modify it
9 under the terms of the GNU Lesser General Public License as published by
10 the Free Software Foundation; either version 2.1 of the License, or
11 (at your option) any later version.
13 systemd is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 Lesser General Public License for more details.
18 You should have received a copy of the GNU Lesser General Public License
19 along with systemd; If not, see <http://www.gnu.org/licenses/>.
22 #include <sys/mount.h>
32 #include "mount-setup.h"
33 #include "dev-setup.h"
41 #include "path-util.h"
50 typedef enum MountMode
{
53 MNT_IN_CONTAINER
= 1 << 1,
56 typedef struct MountPoint
{
62 bool (*condition_fn
)(void);
66 /* The first three entries we might need before SELinux is up. The
67 * fourth (securityfs) is needed by IMA to load a custom policy. The
68 * other ones we can delay until SELinux and IMA are loaded. */
69 #define N_EARLY_MOUNT 5
71 static const MountPoint mount_table
[] = {
72 { "proc", "/proc", "proc", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
73 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
74 { "sysfs", "/sys", "sysfs", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
75 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
76 { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID
|MS_STRICTATIME
,
77 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
78 { "securityfs", "/sys/kernel/security", "securityfs", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
80 { "smackfs", "/sys/fs/smackfs", "smackfs", "smackfsdef=*", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
|MS_STRICTATIME
,
82 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID
|MS_NODEV
|MS_STRICTATIME
,
83 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
84 { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID
), MS_NOSUID
|MS_NOEXEC
,
85 NULL
, MNT_IN_CONTAINER
},
86 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID
|MS_NODEV
|MS_STRICTATIME
,
87 NULL
, MNT_FATAL
|MNT_IN_CONTAINER
},
88 { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
|MS_STRICTATIME
,
89 NULL
, MNT_IN_CONTAINER
},
91 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd,xattr", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
92 NULL
, MNT_IN_CONTAINER
},
94 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
95 NULL
, MNT_IN_CONTAINER
},
96 { "pstore", "/sys/fs/pstore", "pstore", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
99 { "efivarfs", "/sys/firmware/efi/efivars", "efivarfs", NULL
, MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
100 is_efi_boot
, MNT_NONE
},
104 /* These are API file systems that might be mounted by other software,
105 * we just list them here so that we know that we should ignore them */
107 static const char ignore_paths
[] =
108 /* SELinux file systems */
111 /* Legacy cgroup mount points */
114 /* Legacy kernel file system */
116 /* Container bind mounts */
121 bool mount_point_is_api(const char *path
) {
124 /* Checks if this mount point is considered "API", and hence
125 * should be ignored */
127 for (i
= 0; i
< ELEMENTSOF(mount_table
); i
++)
128 if (path_equal(path
, mount_table
[i
].where
))
131 return path_startswith(path
, "/sys/fs/cgroup/");
134 bool mount_point_ignore(const char *path
) {
137 NULSTR_FOREACH(i
, ignore_paths
)
138 if (path_equal(path
, i
))
144 static int mount_one(const MountPoint
*p
, bool relabel
) {
149 if (p
->condition_fn
&& !p
->condition_fn())
152 /* Relabel first, just in case */
154 label_fix(p
->where
, true, true);
156 r
= path_is_mount_point(p
->where
, true);
163 /* Skip securityfs in a container */
164 if (!(p
->mode
& MNT_IN_CONTAINER
) && detect_container(NULL
) > 0)
167 /* The access mode here doesn't really matter too much, since
168 * the mounted file system will take precedence anyway. */
169 mkdir_p_label(p
->where
, 0755);
171 log_debug("Mounting %s to %s of type %s with options %s.",
182 log_full((p
->mode
& MNT_FATAL
) ? LOG_ERR
: LOG_DEBUG
, "Failed to mount %s: %s", p
->where
, strerror(errno
));
183 return (p
->mode
& MNT_FATAL
) ? -errno
: 0;
186 /* Relabel again, since we now mounted something fresh here */
188 label_fix(p
->where
, false, false);
193 int mount_setup_early(void) {
197 assert_cc(N_EARLY_MOUNT
<= ELEMENTSOF(mount_table
));
199 /* Do a minimal mount of /proc and friends to enable the most
200 * basic stuff, such as SELinux */
201 for (i
= 0; i
< N_EARLY_MOUNT
; i
++) {
204 j
= mount_one(mount_table
+ i
, false);
212 int mount_cgroup_controllers(char ***join_controllers
) {
215 _cleanup_set_free_free_ Set
*controllers
= NULL
;
216 _cleanup_fclose_
FILE *f
;
218 /* Mount all available cgroup controllers that are built into the kernel. */
220 f
= fopen("/proc/cgroups", "re");
222 log_error("Failed to enumerate cgroup controllers: %m");
226 controllers
= set_new(string_hash_func
, string_compare_func
);
230 /* Ignore the header line */
231 (void) fgets(buf
, sizeof(buf
), f
);
237 if (fscanf(f
, "%ms %*i %*i %i", &controller
, &enabled
) != 2) {
242 log_error("Failed to parse /proc/cgroups.");
251 r
= set_consume(controllers
, controller
);
253 log_error("Failed to add controller to set.");
262 .flags
= MS_NOSUID
|MS_NOEXEC
|MS_NODEV
,
263 .mode
= MNT_IN_CONTAINER
,
266 _cleanup_free_
char *options
= NULL
, *controller
;
268 controller
= set_steal_first(controllers
);
272 if (join_controllers
)
273 for (k
= join_controllers
; *k
; k
++)
274 if (strv_find(*k
, controller
))
280 for (i
= *k
, j
= *k
; *i
; i
++) {
282 if (!streq(*i
, controller
)) {
283 char _cleanup_free_
*t
;
285 t
= set_remove(controllers
, *i
);
297 options
= strv_join(*k
, ",");
301 options
= controller
;
305 p
.where
= strappenda("/sys/fs/cgroup/", options
);
308 r
= mount_one(&p
, true);
312 if (r
> 0 && k
&& *k
) {
315 for (i
= *k
; *i
; i
++) {
316 char *t
= strappenda("/sys/fs/cgroup/", *i
);
318 r
= symlink(options
, t
);
319 if (r
< 0 && errno
!= EEXIST
) {
320 log_error("Failed to create symlink %s: %m", t
);
332 const struct stat
*sb
,
334 struct FTW
*ftwbuf
) {
336 /* No need to label /dev twice in a row... */
337 if (_unlikely_(ftwbuf
->level
== 0))
340 label_fix(fpath
, false, false);
342 /* /run/initramfs is static data and big, no need to
343 * dynamically relabel its contents at boot... */
344 if (_unlikely_(ftwbuf
->level
== 1 &&
346 streq(fpath
, "/run/initramfs")))
347 return FTW_SKIP_SUBTREE
;
352 int mount_setup(bool loaded_policy
) {
356 for (i
= 0; i
< ELEMENTSOF(mount_table
); i
++) {
357 r
= mount_one(mount_table
+ i
, true);
363 /* Nodes in devtmpfs and /run need to be manually updated for
364 * the appropriate labels, after mounting. The other virtual
365 * API file systems like /sys and /proc do not need that, they
366 * use the same label for all their files. */
368 usec_t before_relabel
, after_relabel
;
369 char timespan
[FORMAT_TIMESPAN_MAX
];
371 before_relabel
= now(CLOCK_MONOTONIC
);
373 nftw("/dev", nftw_cb
, 64, FTW_MOUNT
|FTW_PHYS
|FTW_ACTIONRETVAL
);
374 nftw("/run", nftw_cb
, 64, FTW_MOUNT
|FTW_PHYS
|FTW_ACTIONRETVAL
);
376 after_relabel
= now(CLOCK_MONOTONIC
);
378 log_info("Relabelled /dev and /run in %s.",
379 format_timespan(timespan
, sizeof(timespan
), after_relabel
- before_relabel
, 0));
382 /* Create a few default symlinks, which are normally created
383 * by udevd, but some scripts might need them before we start
387 /* Mark the root directory as shared in regards to mount
388 * propagation. The kernel defaults to "private", but we think
389 * it makes more sense to have a default of "shared" so that
390 * nspawn and the container tools work out of the box. If
391 * specific setups need other settings they can reset the
392 * propagation mode to private if needed. */
393 if (detect_container(NULL
) <= 0)
394 if (mount(NULL
, "/", NULL
, MS_REC
|MS_SHARED
, NULL
) < 0)
395 log_warning("Failed to set up the root directory for shared mount propagation: %m");
397 /* Create a few directories we always want around, Note that
398 * sd_booted() checks for /run/systemd/system, so this mkdir
399 * really needs to stay for good, otherwise software that
400 * copied sd-daemon.c into their sources will misdetect
402 mkdir_label("/run/systemd", 0755);
403 mkdir_label("/run/systemd/system", 0755);
404 mkdir_label("/run/systemd/inaccessible", 0000);