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"
35 //#include "missing.h"
37 #include "mount-setup.h"
38 #include "mount-util.h"
39 #include "path-util.h"
41 //#include "smack-util.h"
43 #include "user-util.h"
47 /// Additional includes needed by elogind
48 #include "string-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. When
69 * SMACK is enabled we need smackfs, too, so it's a fifth one. */
71 #define N_EARLY_MOUNT 5
73 #define N_EARLY_MOUNT 4
76 static const MountPoint mount_table[] = {
77 #if 0 /// UNNEEDED by elogind
78 { "sysfs", "/sys", "sysfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
79 NULL, MNT_FATAL|MNT_IN_CONTAINER },
80 { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
81 NULL, MNT_FATAL|MNT_IN_CONTAINER },
82 { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME,
83 NULL, MNT_FATAL|MNT_IN_CONTAINER },
84 { "securityfs", "/sys/kernel/security", "securityfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
87 { "smackfs", "/sys/fs/smackfs", "smackfs", "smackfsdef=*", MS_NOSUID|MS_NOEXEC|MS_NODEV,
88 mac_smack_use, MNT_FATAL },
89 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
90 mac_smack_use, MNT_FATAL },
92 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
93 NULL, MNT_FATAL|MNT_IN_CONTAINER },
94 { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC,
95 NULL, MNT_IN_CONTAINER },
97 { "tmpfs", "/run", "tmpfs", "mode=755,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
98 mac_smack_use, MNT_FATAL },
100 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
101 NULL, MNT_FATAL|MNT_IN_CONTAINER },
102 { "cgroup", "/sys/fs/cgroup", "cgroup2", "nsdelegate", MS_NOSUID|MS_NOEXEC|MS_NODEV,
103 cg_is_unified_wanted, MNT_IN_CONTAINER },
104 { "cgroup", "/sys/fs/cgroup", "cgroup2", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
105 cg_is_unified_wanted, MNT_IN_CONTAINER },
107 { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
108 cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
109 { "cgroup", "/sys/fs/cgroup/unified", "cgroup2", "nsdelegate", MS_NOSUID|MS_NOEXEC|MS_NODEV,
110 cg_is_hybrid_wanted, MNT_IN_CONTAINER },
111 { "cgroup", "/sys/fs/cgroup/unified", "cgroup2", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
112 cg_is_hybrid_wanted, MNT_IN_CONTAINER },
113 #if 0 /// UNNEEDED by elogind
114 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
115 cg_is_legacy_wanted, MNT_IN_CONTAINER },
116 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV,
117 cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
118 { "pstore", "/sys/fs/pstore", "pstore", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
121 { "efivarfs", "/sys/firmware/efi/efivars", "efivarfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
122 is_efi_boot, MNT_NONE },
125 { "cgroup", "/sys/fs/cgroup/elogind", "cgroup", "none,name=elogind,release_agent="SYSTEMD_CGROUP_AGENT_PATH",xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
126 cg_is_legacy_wanted, MNT_IN_CONTAINER },
127 { "cgroup", "/sys/fs/cgroup/elogind", "cgroup", "none,name=elogind,release_agent="SYSTEMD_CGROUP_AGENT_PATH, MS_NOSUID|MS_NOEXEC|MS_NODEV,
128 cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
132 #if 0 /// UNNEEDED by elogind
133 /* These are API file systems that might be mounted by other software,
134 * we just list them here so that we know that we should ignore them */
136 static const char ignore_paths[] =
137 /* SELinux file systems */
139 /* Container bind mounts */
144 bool mount_point_is_api(const char *path) {
147 /* Checks if this mount point is considered "API", and hence
148 * should be ignored */
150 for (i = 0; i < ELEMENTSOF(mount_table); i ++)
151 if (path_equal(path, mount_table[i].where))
154 return path_startswith(path, "/sys/fs/cgroup/");
157 bool mount_point_ignore(const char *path) {
160 NULSTR_FOREACH(i, ignore_paths)
161 if (path_equal(path, i))
168 static int mount_one(const MountPoint *p, bool relabel) {
173 if (p->condition_fn && !p->condition_fn())
176 /* Relabel first, just in case */
178 (void) label_fix(p->where, true, true);
180 r = path_is_mount_point(p->where, NULL, AT_SYMLINK_FOLLOW);
181 if (r < 0 && r != -ENOENT) {
182 log_full_errno((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, r, "Failed to determine whether %s is a mount point: %m", p->where);
183 return (p->mode & MNT_FATAL) ? r : 0;
188 /* Skip securityfs in a container */
189 if (!(p->mode & MNT_IN_CONTAINER) && detect_container() > 0)
192 /* The access mode here doesn't really matter too much, since
193 * the mounted file system will take precedence anyway. */
195 (void) mkdir_p_label(p->where, 0755);
197 (void) mkdir_p(p->where, 0755);
199 log_debug("Mounting %s to %s of type %s with options %s.",
210 log_full_errno((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, errno, "Failed to mount %s at %s: %m", p->type, p->where);
211 return (p->mode & MNT_FATAL) ? -errno : 0;
214 /* Relabel again, since we now mounted something fresh here */
216 (void) label_fix(p->where, false, false);
221 static int mount_points_setup(unsigned n, bool loaded_policy) {
225 for (i = 0; i < n; i ++) {
228 j = mount_one(mount_table + i, loaded_policy);
229 if (j != 0 && r >= 0)
236 #if 0 /// UNNEEDED by elogind
237 int mount_setup_early(void) {
238 assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));
240 /* Do a minimal mount of /proc and friends to enable the most
241 * basic stuff, such as SELinux */
242 return mount_points_setup(N_EARLY_MOUNT, false);
245 int mount_cgroup_controllers(char ***join_controllers) {
246 _cleanup_set_free_free_ Set *controllers = NULL;
249 if (!cg_is_legacy_wanted())
252 /* Mount all available cgroup controllers that are built into the kernel. */
254 controllers = set_new(&string_hash_ops);
258 r = cg_kernel_controllers(controllers);
260 return log_error_errno(r, "Failed to enumerate cgroup controllers: %m");
263 _cleanup_free_ char *options = NULL, *controller = NULL, *where = NULL;
267 .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV,
268 .mode = MNT_IN_CONTAINER,
272 controller = set_steal_first(controllers);
276 if (join_controllers)
277 for (k = join_controllers; *k; k++)
278 if (strv_find(*k, controller))
284 for (i = *k, j = *k; *i; i++) {
286 if (!streq(*i, controller)) {
287 _cleanup_free_ char *t;
289 t = set_remove(controllers, *i);
301 options = strv_join(*k, ",");
305 options = controller;
309 where = strappend("/sys/fs/cgroup/", options);
316 r = mount_one(&p, true);
320 if (r > 0 && k && *k) {
323 for (i = *k; *i; i++) {
324 _cleanup_free_ char *t = NULL;
326 t = strappend("/sys/fs/cgroup/", *i);
330 r = symlink(options, t);
332 #ifdef SMACK_RUN_LABEL
333 _cleanup_free_ char *src;
334 src = strappend("/sys/fs/cgroup/", options);
337 r = mac_smack_copy(t, src);
338 if (r < 0 && r != -EOPNOTSUPP)
339 return log_error_errno(r, "Failed to copy smack label from %s to %s: %m", src, t);
341 } else if (errno != EEXIST)
342 return log_error_errno(errno, "Failed to create symlink %s: %m", t);
347 /* Now that we mounted everything, let's make the tmpfs the
348 * cgroup file systems are mounted into read-only. */
349 (void) mount("tmpfs", "/sys/fs/cgroup", "tmpfs", MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME|MS_RDONLY, "mode=755");
354 #if HAVE_SELINUX || ENABLE_SMACK
357 const struct stat *sb,
359 struct FTW *ftwbuf) {
361 /* No need to label /dev twice in a row... */
362 if (_unlikely_(ftwbuf->level == 0))
365 label_fix(fpath, false, false);
367 /* /run/initramfs is static data and big, no need to
368 * dynamically relabel its contents at boot... */
369 if (_unlikely_(ftwbuf->level == 1 &&
371 streq(fpath, "/run/initramfs")))
372 return FTW_SKIP_SUBTREE;
379 int mount_setup(bool loaded_policy) {
382 r = mount_points_setup(ELEMENTSOF(mount_table), loaded_policy);
386 #if 0 /// elogind does not control /, /dev, /run and /run/systemd/* are setup elsewhere.
387 #if HAVE_SELINUX || ENABLE_SMACK
388 /* Nodes in devtmpfs and /run need to be manually updated for
389 * the appropriate labels, after mounting. The other virtual
390 * API file systems like /sys and /proc do not need that, they
391 * use the same label for all their files. */
393 usec_t before_relabel, after_relabel;
394 char timespan[FORMAT_TIMESPAN_MAX];
396 before_relabel = now(CLOCK_MONOTONIC);
398 nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
399 nftw("/dev/shm", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
400 nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
402 /* Temporarily remount the root cgroup filesystem to give it a proper label. */
403 r = cg_all_unified();
405 (void) mount(NULL, "/sys/fs/cgroup", NULL, MS_REMOUNT, NULL);
406 label_fix("/sys/fs/cgroup", false, false);
407 nftw("/sys/fs/cgroup", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
408 (void) mount(NULL, "/sys/fs/cgroup", NULL, MS_REMOUNT|MS_RDONLY, NULL);
410 return log_error_errno(r, "Failed to determine whether we are in all unified mode: %m");
412 after_relabel = now(CLOCK_MONOTONIC);
414 log_info("Relabelled /dev, /run and /sys/fs/cgroup in %s.",
415 format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel, 0));
419 /* Create a few default symlinks, which are normally created
420 * by udevd, but some scripts might need them before we start
422 dev_setup(NULL, UID_INVALID, GID_INVALID);
424 /* Mark the root directory as shared in regards to mount propagation. The kernel defaults to "private", but we
425 * think it makes more sense to have a default of "shared" so that nspawn and the container tools work out of
426 * the box. If specific setups need other settings they can reset the propagation mode to private if
427 * needed. Note that we set this only when we are invoked directly by the kernel. If we are invoked by a
428 * container manager we assume the container manager knows what it is doing (for example, because it set up
429 * some directories with different propagation modes). */
430 if (detect_container() <= 0)
431 if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
432 log_warning_errno(errno, "Failed to set up the root directory for shared mount propagation: %m");
435 /* Create a few directories we always want around, Note that sd_booted() checks for /run/systemd/system, so
436 * this mkdir really needs to stay for good, otherwise software that copied sd-daemon.c into their sources will
437 * misdetect systemd. */
438 (void) mkdir_label("/run/systemd", 0755);
439 #if 0 /// Yeah, but elogind is not used with systemd, so this directory would be toxic.
440 (void) mkdir_label("/run/systemd/system", 0755);
443 /* Set up inaccessible items */
444 (void) mkdir_label("/run/systemd/inaccessible", 0000);
445 (void) mknod("/run/systemd/inaccessible/reg", S_IFREG | 0000, 0);
446 (void) mkdir_label("/run/systemd/inaccessible/dir", 0000);
447 (void) mknod("/run/systemd/inaccessible/chr", S_IFCHR | 0000, makedev(0, 0));
448 (void) mknod("/run/systemd/inaccessible/blk", S_IFBLK | 0000, makedev(0, 0));
449 (void) mkfifo("/run/systemd/inaccessible/fifo", 0000);
450 (void) mknod("/run/systemd/inaccessible/sock", S_IFSOCK | 0000, 0);