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"
27 //#include "bus-util.h"
28 #include "cgroup-util.h"
29 //#include "dev-setup.h"
30 //#include "efivars.h"
34 //#include "missing.h"
36 #include "mount-setup.h"
37 #include "mount-util.h"
38 #include "path-util.h"
40 //#include "smack-util.h"
41 #include "string-util.h"
43 #include "user-util.h"
47 typedef enum MountMode {
50 MNT_IN_CONTAINER = 1 << 1,
53 typedef struct MountPoint {
59 bool (*condition_fn)(void);
63 /* The first three entries we might need before SELinux is up. The
64 * fourth (securityfs) is needed by IMA to load a custom policy. The
65 * other ones we can delay until SELinux and IMA are loaded. When
66 * SMACK is enabled we need smackfs, too, so it's a fifth one. */
68 #define N_EARLY_MOUNT 5
70 #define N_EARLY_MOUNT 4
73 static const MountPoint mount_table[] = {
74 #if 0 /// UNNEEDED by elogind
75 { "sysfs", "/sys", "sysfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
76 NULL, MNT_FATAL|MNT_IN_CONTAINER },
77 { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
78 NULL, MNT_FATAL|MNT_IN_CONTAINER },
79 { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME,
80 NULL, MNT_FATAL|MNT_IN_CONTAINER },
81 { "securityfs", "/sys/kernel/security", "securityfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
84 { "smackfs", "/sys/fs/smackfs", "smackfs", "smackfsdef=*", MS_NOSUID|MS_NOEXEC|MS_NODEV,
85 mac_smack_use, MNT_FATAL },
86 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
87 mac_smack_use, MNT_FATAL },
89 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
90 NULL, MNT_FATAL|MNT_IN_CONTAINER },
91 { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC,
92 NULL, MNT_IN_CONTAINER },
94 { "tmpfs", "/run", "tmpfs", "mode=755,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
95 mac_smack_use, MNT_FATAL },
97 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
98 NULL, MNT_FATAL|MNT_IN_CONTAINER },
99 { "cgroup", "/sys/fs/cgroup", "cgroup", "__DEVEL__sane_behavior", MS_NOSUID|MS_NOEXEC|MS_NODEV,
100 cg_is_unified_wanted, MNT_FATAL|MNT_IN_CONTAINER },
102 { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
103 cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
104 #if 0 /// UNNEEDED by elogind
105 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
106 cg_is_legacy_wanted, MNT_IN_CONTAINER },
107 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV,
108 cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
110 { "cgroup", "/sys/fs/cgroup/elogind", "cgroup", "none,name=elogind,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
111 cg_is_legacy_wanted, MNT_IN_CONTAINER },
112 { "cgroup", "/sys/fs/cgroup/elogind", "cgroup", "none,name=elogind", MS_NOSUID|MS_NOEXEC|MS_NODEV,
113 cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
115 #if 0 /// UNNEEDED by elogind
116 { "pstore", "/sys/fs/pstore", "pstore", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
119 { "efivarfs", "/sys/firmware/efi/efivars", "efivarfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
120 is_efi_boot, MNT_NONE },
122 { "kdbusfs", "/sys/fs/kdbus", "kdbusfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
123 is_kdbus_wanted, MNT_IN_CONTAINER },
127 #if 0 /// UNNEEDED by elogind
128 /* These are API file systems that might be mounted by other software,
129 * we just list them here so that we know that we should ignore them */
131 static const char ignore_paths[] =
132 /* SELinux file systems */
134 /* Container bind mounts */
139 bool mount_point_is_api(const char *path) {
142 /* Checks if this mount point is considered "API", and hence
143 * should be ignored */
145 for (i = 0; i < ELEMENTSOF(mount_table); i ++)
146 if (path_equal(path, mount_table[i].where))
149 return path_startswith(path, "/sys/fs/cgroup/");
152 bool mount_point_ignore(const char *path) {
155 NULSTR_FOREACH(i, ignore_paths)
156 if (path_equal(path, i))
163 static int mount_one(const MountPoint *p, bool relabel) {
168 if (p->condition_fn && !p->condition_fn())
171 /* Relabel first, just in case */
173 (void) label_fix(p->where, true, true);
175 r = path_is_mount_point(p->where, AT_SYMLINK_FOLLOW);
176 if (r < 0 && r != -ENOENT) {
177 log_full_errno((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, r, "Failed to determine whether %s is a mount point: %m", p->where);
178 return (p->mode & MNT_FATAL) ? r : 0;
183 /* Skip securityfs in a container */
184 if (!(p->mode & MNT_IN_CONTAINER) && detect_container() > 0)
187 /* The access mode here doesn't really matter too much, since
188 * the mounted file system will take precedence anyway. */
190 (void) mkdir_p_label(p->where, 0755);
192 (void) mkdir_p(p->where, 0755);
194 log_debug("Mounting %s to %s of type %s with options %s.",
205 log_full_errno((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, errno, "Failed to mount %s at %s: %m", p->type, p->where);
206 return (p->mode & MNT_FATAL) ? -errno : 0;
209 /* Relabel again, since we now mounted something fresh here */
211 (void) label_fix(p->where, false, false);
216 static int mount_points_setup(unsigned n, bool loaded_policy) {
220 for (i = 0; i < n; i ++) {
223 j = mount_one(mount_table + i, loaded_policy);
224 if (j != 0 && r >= 0)
231 #if 0 /// UNNEEDED by elogind
232 int mount_setup_early(void) {
233 assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));
235 /* Do a minimal mount of /proc and friends to enable the most
236 * basic stuff, such as SELinux */
237 return mount_points_setup(N_EARLY_MOUNT, false);
240 int mount_cgroup_controllers(char ***join_controllers) {
241 _cleanup_set_free_free_ Set *controllers = NULL;
244 if (!cg_is_legacy_wanted())
247 /* Mount all available cgroup controllers that are built into the kernel. */
249 controllers = set_new(&string_hash_ops);
253 r = cg_kernel_controllers(controllers);
255 return log_error_errno(r, "Failed to enumerate cgroup controllers: %m");
258 _cleanup_free_ char *options = NULL, *controller = NULL, *where = NULL;
262 .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV,
263 .mode = MNT_IN_CONTAINER,
267 controller = set_steal_first(controllers);
271 if (join_controllers)
272 for (k = join_controllers; *k; k++)
273 if (strv_find(*k, controller))
279 for (i = *k, j = *k; *i; i++) {
281 if (!streq(*i, controller)) {
282 _cleanup_free_ char *t;
284 t = set_remove(controllers, *i);
296 options = strv_join(*k, ",");
300 options = controller;
304 where = strappend("/sys/fs/cgroup/", options);
311 r = mount_one(&p, true);
315 if (r > 0 && k && *k) {
318 for (i = *k; *i; i++) {
319 _cleanup_free_ char *t = NULL;
321 t = strappend("/sys/fs/cgroup/", *i);
325 r = symlink(options, t);
327 #ifdef SMACK_RUN_LABEL
328 _cleanup_free_ char *src;
329 src = strappend("/sys/fs/cgroup/", options);
332 r = mac_smack_copy(t, src);
333 if (r < 0 && r != -EOPNOTSUPP)
334 return log_error_errno(r, "Failed to copy smack label from %s to %s: %m", src, t);
336 } else if (errno != EEXIST)
337 return log_error_errno(errno, "Failed to create symlink %s: %m", t);
342 /* Now that we mounted everything, let's make the tmpfs the
343 * cgroup file systems are mounted into read-only. */
344 (void) mount("tmpfs", "/sys/fs/cgroup", "tmpfs", MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME|MS_RDONLY, "mode=755");
349 #if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
352 const struct stat *sb,
354 struct FTW *ftwbuf) {
356 /* No need to label /dev twice in a row... */
357 if (_unlikely_(ftwbuf->level == 0))
360 label_fix(fpath, false, false);
362 /* /run/initramfs is static data and big, no need to
363 * dynamically relabel its contents at boot... */
364 if (_unlikely_(ftwbuf->level == 1 &&
366 streq(fpath, "/run/initramfs")))
367 return FTW_SKIP_SUBTREE;
374 int mount_setup(bool loaded_policy) {
377 r = mount_points_setup(ELEMENTSOF(mount_table), loaded_policy);
382 #if 0 /// elogind does not control /, /dev, /run and /run/systemd/* are setup elsewhere.
383 #if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
384 /* Nodes in devtmpfs and /run need to be manually updated for
385 * the appropriate labels, after mounting. The other virtual
386 * API file systems like /sys and /proc do not need that, they
387 * use the same label for all their files. */
389 usec_t before_relabel, after_relabel;
390 char timespan[FORMAT_TIMESPAN_MAX];
392 before_relabel = now(CLOCK_MONOTONIC);
394 nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
395 nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
397 after_relabel = now(CLOCK_MONOTONIC);
399 log_info("Relabelled /dev and /run in %s.",
400 format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel, 0));
404 /* Create a few default symlinks, which are normally created
405 * by udevd, but some scripts might need them before we start
407 dev_setup(NULL, UID_INVALID, GID_INVALID);
409 /* Mark the root directory as shared in regards to mount
410 * propagation. The kernel defaults to "private", but we think
411 * it makes more sense to have a default of "shared" so that
412 * nspawn and the container tools work out of the box. If
413 * specific setups need other settings they can reset the
414 * propagation mode to private if needed. */
415 if (detect_container() <= 0)
416 if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
417 log_warning_errno(errno, "Failed to set up the root directory for shared mount propagation: %m");
419 /* Create a few directories we always want around, Note that
420 * sd_booted() checks for /run/systemd/system, so this mkdir
421 * really needs to stay for good, otherwise software that
422 * copied sd-daemon.c into their sources will misdetect
424 mkdir_label("/run/systemd", 0755);
425 mkdir_label("/run/systemd/system", 0755);
426 mkdir_label("/run/systemd/inaccessible", 0000);