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 },
90 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV,
91 NULL, MNT_IN_CONTAINER },
92 { "pstore", "/sys/fs/pstore", "pstore", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
95 { "efivarfs", "/sys/firmware/efi/efivars", "efivarfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
96 is_efi_boot, MNT_NONE },
100 /* These are API file systems that might be mounted by other software,
101 * we just list them here so that we know that we should ignore them */
103 static const char ignore_paths[] =
104 /* SELinux file systems */
107 /* Legacy cgroup mount points */
110 /* Legacy kernel file system */
112 /* Container bind mounts */
117 bool mount_point_is_api(const char *path) {
120 /* Checks if this mount point is considered "API", and hence
121 * should be ignored */
123 for (i = 0; i < ELEMENTSOF(mount_table); i ++)
124 if (path_equal(path, mount_table[i].where))
127 return path_startswith(path, "/sys/fs/cgroup/");
130 bool mount_point_ignore(const char *path) {
133 NULSTR_FOREACH(i, ignore_paths)
134 if (path_equal(path, i))
140 static int mount_one(const MountPoint *p, bool relabel) {
145 if (p->condition_fn && !p->condition_fn())
148 /* Relabel first, just in case */
150 label_fix(p->where, true, true);
152 r = path_is_mount_point(p->where, true);
159 /* Skip securityfs in a container */
160 if (!(p->mode & MNT_IN_CONTAINER) && detect_container(NULL) > 0)
163 /* The access mode here doesn't really matter too much, since
164 * the mounted file system will take precedence anyway. */
165 mkdir_p_label(p->where, 0755);
167 log_debug("Mounting %s to %s of type %s with options %s.",
178 log_full((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, "Failed to mount %s: %s", p->where, strerror(errno));
179 return (p->mode & MNT_FATAL) ? -errno : 0;
182 /* Relabel again, since we now mounted something fresh here */
184 label_fix(p->where, false, false);
189 int mount_setup_early(void) {
193 assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));
195 /* Do a minimal mount of /proc and friends to enable the most
196 * basic stuff, such as SELinux */
197 for (i = 0; i < N_EARLY_MOUNT; i ++) {
200 j = mount_one(mount_table + i, false);
208 int mount_cgroup_controllers(char ***join_controllers) {
214 /* Mount all available cgroup controllers that are built into the kernel. */
216 f = fopen("/proc/cgroups", "re");
218 log_error("Failed to enumerate cgroup controllers: %m");
222 controllers = set_new(string_hash_func, string_compare_func);
228 /* Ignore the header line */
229 (void) fgets(buf, sizeof(buf), f);
235 if (fscanf(f, "%ms %*i %*i %i", &controller, &enabled) != 2) {
240 log_error("Failed to parse /proc/cgroups.");
250 r = set_put(controllers, controller);
252 log_error("Failed to add controller to set.");
260 char *controller, *where, *options;
263 controller = set_steal_first(controllers);
267 if (join_controllers)
268 for (k = join_controllers; *k; k++)
269 if (strv_find(*k, controller))
275 for (i = *k, j = *k; *i; i++) {
277 if (!streq(*i, controller)) {
280 t = set_remove(controllers, *i);
293 options = strv_join(*k, ",");
301 options = controller;
305 where = strappend("/sys/fs/cgroup/", options);
317 p.flags = MS_NOSUID|MS_NOEXEC|MS_NODEV;
319 r = mount_one(&p, true);
328 if (r > 0 && k && *k) {
331 for (i = *k; *i; i++) {
334 t = strappend("/sys/fs/cgroup/", *i);
341 r = symlink(options, t);
344 if (r < 0 && errno != EEXIST) {
345 log_error("Failed to create symlink: %m");
359 set_free_free(controllers);
368 const struct stat *sb,
370 struct FTW *ftwbuf) {
372 /* No need to label /dev twice in a row... */
373 if (_unlikely_(ftwbuf->level == 0))
376 label_fix(fpath, false, false);
378 /* /run/initramfs is static data and big, no need to
379 * dynamically relabel its contents at boot... */
380 if (_unlikely_(ftwbuf->level == 1 &&
382 streq(fpath, "/run/initramfs")))
383 return FTW_SKIP_SUBTREE;
388 int mount_setup(bool loaded_policy) {
390 static const char relabel[] =
391 "/run/initramfs/root-fsck\0"
392 "/run/initramfs/shutdown\0";
398 for (i = 0; i < ELEMENTSOF(mount_table); i ++) {
399 r = mount_one(mount_table + i, true);
405 /* Nodes in devtmpfs and /run need to be manually updated for
406 * the appropriate labels, after mounting. The other virtual
407 * API file systems like /sys and /proc do not need that, they
408 * use the same label for all their files. */
410 usec_t before_relabel, after_relabel;
411 char timespan[FORMAT_TIMESPAN_MAX];
413 before_relabel = now(CLOCK_MONOTONIC);
415 nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
416 nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
418 /* Explicitly relabel these */
419 NULSTR_FOREACH(j, relabel)
420 label_fix(j, true, false);
422 after_relabel = now(CLOCK_MONOTONIC);
424 log_info("Relabelled /dev and /run in %s.",
425 format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel));
428 /* Create a few default symlinks, which are normally created
429 * by udevd, but some scripts might need them before we start
433 /* Mark the root directory as shared in regards to mount
434 * propagation. The kernel defaults to "private", but we think
435 * it makes more sense to have a default of "shared" so that
436 * nspawn and the container tools work out of the box. If
437 * specific setups need other settings they can reset the
438 * propagation mode to private if needed. */
439 if (detect_container(NULL) <= 0)
440 if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
441 log_warning("Failed to set up the root directory for shared mount propagation: %m");
443 /* Create a few directories we always want around, Note that
444 * sd_booted() checks for /run/systemd/system, so this mkdir
445 * really needs to stay for good, otherwise software that
446 * copied sd-daemon.c into their sources will misdetect
448 mkdir_label("/run/systemd", 0755);
449 mkdir_label("/run/systemd/system", 0755);
450 mkdir_label("/run/systemd/inaccessible", 0000);