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"
45 #include "smack-util.h"
48 typedef enum MountMode {
51 MNT_IN_CONTAINER = 1 << 1,
54 typedef struct MountPoint {
60 bool (*condition_fn)(void);
64 /* The first three entries we might need before SELinux is up. The
65 * fourth (securityfs) is needed by IMA to load a custom policy. The
66 * other ones we can delay until SELinux and IMA are loaded. */
67 #define N_EARLY_MOUNT 5
69 static const MountPoint mount_table[] = {
70 { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
71 NULL, MNT_FATAL|MNT_IN_CONTAINER },
72 { "sysfs", "/sys", "sysfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
73 NULL, MNT_FATAL|MNT_IN_CONTAINER },
74 { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME,
75 NULL, MNT_FATAL|MNT_IN_CONTAINER },
76 { "securityfs", "/sys/kernel/security", "securityfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
79 { "smackfs", "/sys/fs/smackfs", "smackfs", "smackfsdef=*", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
80 use_smack, MNT_FATAL },
81 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
82 use_smack, MNT_FATAL },
84 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
85 NULL, MNT_FATAL|MNT_IN_CONTAINER },
86 { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC,
87 NULL, MNT_IN_CONTAINER },
89 { "tmpfs", "/run", "tmpfs", "mode=755,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
90 use_smack, MNT_FATAL },
92 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
93 NULL, MNT_FATAL|MNT_IN_CONTAINER },
94 { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
95 NULL, MNT_IN_CONTAINER },
97 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
98 NULL, MNT_IN_CONTAINER },
100 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV,
101 NULL, MNT_IN_CONTAINER },
102 { "pstore", "/sys/fs/pstore", "pstore", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
105 { "efivarfs", "/sys/firmware/efi/efivars", "efivarfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
106 is_efi_boot, MNT_NONE },
110 /* These are API file systems that might be mounted by other software,
111 * we just list them here so that we know that we should ignore them */
113 static const char ignore_paths[] =
114 /* SELinux file systems */
117 /* Legacy cgroup mount points */
120 /* Legacy kernel file system */
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 label_fix(p->where, true, true);
162 r = path_is_mount_point(p->where, true);
169 /* Skip securityfs in a container */
170 if (!(p->mode & MNT_IN_CONTAINER) && detect_container(NULL) > 0)
173 /* The access mode here doesn't really matter too much, since
174 * the mounted file system will take precedence anyway. */
175 mkdir_p_label(p->where, 0755);
177 log_debug("Mounting %s to %s of type %s with options %s.",
188 log_full((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, "Failed to mount %s: %m", p->where);
189 return (p->mode & MNT_FATAL) ? -errno : 0;
192 /* Relabel again, since we now mounted something fresh here */
194 label_fix(p->where, false, false);
199 int mount_setup_early(void) {
203 assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));
205 /* Do a minimal mount of /proc and friends to enable the most
206 * basic stuff, such as SELinux */
207 for (i = 0; i < N_EARLY_MOUNT; i ++) {
210 j = mount_one(mount_table + i, false);
218 int mount_cgroup_controllers(char ***join_controllers) {
219 _cleanup_set_free_free_ Set *controllers = NULL;
220 _cleanup_fclose_ FILE *f;
224 /* Mount all available cgroup controllers that are built into the kernel. */
226 f = fopen("/proc/cgroups", "re");
228 log_error("Failed to enumerate cgroup controllers: %m");
232 controllers = set_new(string_hash_func, string_compare_func);
236 /* Ignore the header line */
237 (void) fgets(buf, sizeof(buf), f);
243 if (fscanf(f, "%ms %*i %*i %i", &controller, &enabled) != 2) {
248 log_error("Failed to parse /proc/cgroups.");
257 r = set_consume(controllers, controller);
259 log_error("Failed to add controller to set.");
265 _cleanup_free_ char *options = NULL, *controller = NULL, *where = NULL;
269 .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV,
270 .mode = MNT_IN_CONTAINER,
274 controller = set_steal_first(controllers);
278 if (join_controllers)
279 for (k = join_controllers; *k; k++)
280 if (strv_find(*k, controller))
286 for (i = *k, j = *k; *i; i++) {
288 if (!streq(*i, controller)) {
289 _cleanup_free_ char *t;
291 t = set_remove(controllers, *i);
303 options = strv_join(*k, ",");
307 options = controller;
311 where = strappend("/sys/fs/cgroup/", options);
318 r = mount_one(&p, true);
322 if (r > 0 && k && *k) {
325 for (i = *k; *i; i++) {
326 _cleanup_free_ char *t = NULL;
328 t = strappend("/sys/fs/cgroup/", *i);
332 r = symlink(options, t);
333 if (r < 0 && errno != EEXIST) {
334 log_error("Failed to create symlink %s: %m", t);
341 /* Now that we mounted everything, let's make the tmpfs the
342 * cgroup file systems are mounted into read-only. */
343 mount("tmpfs", "/sys/fs/cgroup", "tmpfs", MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME|MS_RDONLY, "mode=755");
350 const struct stat *sb,
352 struct FTW *ftwbuf) {
354 /* No need to label /dev twice in a row... */
355 if (_unlikely_(ftwbuf->level == 0))
358 label_fix(fpath, false, false);
360 /* /run/initramfs is static data and big, no need to
361 * dynamically relabel its contents at boot... */
362 if (_unlikely_(ftwbuf->level == 1 &&
364 streq(fpath, "/run/initramfs")))
365 return FTW_SKIP_SUBTREE;
370 int mount_setup(bool loaded_policy) {
374 for (i = 0; i < ELEMENTSOF(mount_table); i ++) {
375 r = mount_one(mount_table + i, true);
381 /* Nodes in devtmpfs and /run need to be manually updated for
382 * the appropriate labels, after mounting. The other virtual
383 * API file systems like /sys and /proc do not need that, they
384 * use the same label for all their files. */
386 usec_t before_relabel, after_relabel;
387 char timespan[FORMAT_TIMESPAN_MAX];
389 before_relabel = now(CLOCK_MONOTONIC);
391 nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
392 nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
394 after_relabel = now(CLOCK_MONOTONIC);
396 log_info("Relabelled /dev and /run in %s.",
397 format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel, 0));
400 /* Create a few default symlinks, which are normally created
401 * by udevd, but some scripts might need them before we start
405 /* Mark the root directory as shared in regards to mount
406 * propagation. The kernel defaults to "private", but we think
407 * it makes more sense to have a default of "shared" so that
408 * nspawn and the container tools work out of the box. If
409 * specific setups need other settings they can reset the
410 * propagation mode to private if needed. */
411 if (detect_container(NULL) <= 0)
412 if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
413 log_warning("Failed to set up the root directory for shared mount propagation: %m");
415 /* Create a few directories we always want around, Note that
416 * sd_booted() checks for /run/systemd/system, so this mkdir
417 * really needs to stay for good, otherwise software that
418 * copied sd-daemon.c into their sources will misdetect
420 mkdir_label("/run/systemd", 0755);
421 mkdir_label("/run/systemd/system", 0755);
422 mkdir_label("/run/systemd/inaccessible", 0000);