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"
51 typedef enum MountMode {
54 MNT_IN_CONTAINER = 1 << 1,
57 typedef struct MountPoint {
63 bool (*condition_fn)(void);
67 /* The first three entries we might need before SELinux is up. The
68 * fourth (securityfs) is needed by IMA to load a custom policy. The
69 * other ones we can delay until SELinux and IMA are loaded. */
70 #define N_EARLY_MOUNT 5
72 static const MountPoint mount_table[] = {
73 { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
74 NULL, MNT_FATAL|MNT_IN_CONTAINER },
75 { "sysfs", "/sys", "sysfs", 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|MS_STRICTATIME,
83 use_smack, MNT_FATAL },
84 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
85 use_smack, 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 use_smack, MNT_FATAL },
95 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
96 NULL, MNT_FATAL|MNT_IN_CONTAINER },
97 { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
98 NULL, MNT_IN_CONTAINER },
100 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
101 NULL, MNT_IN_CONTAINER },
103 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV,
104 NULL, 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 */
120 /* Legacy cgroup mount points */
123 /* Legacy kernel file system */
125 /* Container bind mounts */
130 bool mount_point_is_api(const char *path) {
133 /* Checks if this mount point is considered "API", and hence
134 * should be ignored */
136 for (i = 0; i < ELEMENTSOF(mount_table); i ++)
137 if (path_equal(path, mount_table[i].where))
140 return path_startswith(path, "/sys/fs/cgroup/");
143 bool mount_point_ignore(const char *path) {
146 NULSTR_FOREACH(i, ignore_paths)
147 if (path_equal(path, i))
153 static int mount_one(const MountPoint *p, bool relabel) {
158 if (p->condition_fn && !p->condition_fn())
161 /* Relabel first, just in case */
163 label_fix(p->where, true, true);
165 r = path_is_mount_point(p->where, true);
172 /* Skip securityfs in a container */
173 if (!(p->mode & MNT_IN_CONTAINER) && detect_container(NULL) > 0)
176 /* The access mode here doesn't really matter too much, since
177 * the mounted file system will take precedence anyway. */
178 mkdir_p_label(p->where, 0755);
180 log_debug("Mounting %s to %s of type %s with options %s.",
191 log_full((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, "Failed to mount %s: %m", p->where);
192 return (p->mode & MNT_FATAL) ? -errno : 0;
195 /* Relabel again, since we now mounted something fresh here */
197 label_fix(p->where, false, false);
202 int mount_setup_early(void) {
206 assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));
208 /* Do a minimal mount of /proc and friends to enable the most
209 * basic stuff, such as SELinux */
210 for (i = 0; i < N_EARLY_MOUNT; i ++) {
213 j = mount_one(mount_table + i, false);
221 int mount_cgroup_controllers(char ***join_controllers) {
224 _cleanup_set_free_free_ Set *controllers = NULL;
225 _cleanup_fclose_ FILE *f;
227 /* Mount all available cgroup controllers that are built into the kernel. */
229 f = fopen("/proc/cgroups", "re");
231 log_error("Failed to enumerate cgroup controllers: %m");
235 controllers = set_new(string_hash_func, string_compare_func);
239 /* Ignore the header line */
240 (void) fgets(buf, sizeof(buf), f);
246 if (fscanf(f, "%ms %*i %*i %i", &controller, &enabled) != 2) {
251 log_error("Failed to parse /proc/cgroups.");
260 r = set_consume(controllers, controller);
262 log_error("Failed to add controller to set.");
271 .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV,
272 .mode = MNT_IN_CONTAINER,
275 _cleanup_free_ char *options = NULL, *controller;
277 controller = set_steal_first(controllers);
281 if (join_controllers)
282 for (k = join_controllers; *k; k++)
283 if (strv_find(*k, controller))
289 for (i = *k, j = *k; *i; i++) {
291 if (!streq(*i, controller)) {
292 char _cleanup_free_ *t;
294 t = set_remove(controllers, *i);
306 options = strv_join(*k, ",");
310 options = controller;
314 p.where = strappenda("/sys/fs/cgroup/", options);
317 r = mount_one(&p, true);
321 if (r > 0 && k && *k) {
324 for (i = *k; *i; i++) {
325 char *t = strappenda("/sys/fs/cgroup/", *i);
327 r = symlink(options, t);
328 if (r < 0 && errno != EEXIST) {
329 log_error("Failed to create symlink %s: %m", t);
341 const struct stat *sb,
343 struct FTW *ftwbuf) {
345 /* No need to label /dev twice in a row... */
346 if (_unlikely_(ftwbuf->level == 0))
349 label_fix(fpath, false, false);
351 /* /run/initramfs is static data and big, no need to
352 * dynamically relabel its contents at boot... */
353 if (_unlikely_(ftwbuf->level == 1 &&
355 streq(fpath, "/run/initramfs")))
356 return FTW_SKIP_SUBTREE;
361 int mount_setup(bool loaded_policy) {
365 for (i = 0; i < ELEMENTSOF(mount_table); i ++) {
366 r = mount_one(mount_table + i, true);
372 /* Nodes in devtmpfs and /run need to be manually updated for
373 * the appropriate labels, after mounting. The other virtual
374 * API file systems like /sys and /proc do not need that, they
375 * use the same label for all their files. */
377 usec_t before_relabel, after_relabel;
378 char timespan[FORMAT_TIMESPAN_MAX];
380 before_relabel = now(CLOCK_MONOTONIC);
382 nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
383 nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
385 after_relabel = now(CLOCK_MONOTONIC);
387 log_info("Relabelled /dev and /run in %s.",
388 format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel, 0));
391 /* Create a few default symlinks, which are normally created
392 * by udevd, but some scripts might need them before we start
396 /* Mark the root directory as shared in regards to mount
397 * propagation. The kernel defaults to "private", but we think
398 * it makes more sense to have a default of "shared" so that
399 * nspawn and the container tools work out of the box. If
400 * specific setups need other settings they can reset the
401 * propagation mode to private if needed. */
402 if (detect_container(NULL) <= 0)
403 if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
404 log_warning("Failed to set up the root directory for shared mount propagation: %m");
406 /* Create a few directories we always want around, Note that
407 * sd_booted() checks for /run/systemd/system, so this mkdir
408 * really needs to stay for good, otherwise software that
409 * copied sd-daemon.c into their sources will misdetect
411 mkdir_label("/run/systemd", 0755);
412 mkdir_label("/run/systemd/system", 0755);
413 mkdir_label("/run/systemd/inaccessible", 0000);