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/>.
25 #include <sys/mount.h>
28 #include "alloc-util.h"
29 //#include "bus-util.h"
30 #include "cgroup-util.h"
31 //#include "dev-setup.h"
32 //#include "efivars.h"
36 //#include "missing.h"
38 #include "mount-setup.h"
39 #include "mount-util.h"
40 #include "path-util.h"
42 //#include "smack-util.h"
44 #include "string-util.h"
45 #include "user-util.h"
49 typedef enum MountMode {
52 MNT_IN_CONTAINER = 1 << 1,
55 typedef struct MountPoint {
61 bool (*condition_fn)(void);
65 /* The first three entries we might need before SELinux is up. The
66 * fourth (securityfs) is needed by IMA to load a custom policy. The
67 * other ones we can delay until SELinux and IMA are loaded. When
68 * SMACK is enabled we need smackfs, too, so it's a fifth one. */
70 #define N_EARLY_MOUNT 5
72 #define N_EARLY_MOUNT 4
75 static const MountPoint mount_table[] = {
76 /// 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", "cgroup", "__DEVEL__sane_behavior", MS_NOSUID|MS_NOEXEC|MS_NODEV,
103 cg_is_unified_wanted, MNT_FATAL|MNT_IN_CONTAINER },
105 { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
106 cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
107 /// UNNEEDED by elogind
109 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
110 cg_is_legacy_wanted, MNT_IN_CONTAINER },
111 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV,
112 cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
114 { "cgroup", "/sys/fs/cgroup/elogind", "cgroup", "none,name=elogind,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
115 cg_is_legacy_wanted, MNT_IN_CONTAINER },
116 { "cgroup", "/sys/fs/cgroup/elogind", "cgroup", "none,name=elogind", MS_NOSUID|MS_NOEXEC|MS_NODEV,
117 cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
119 /// UNNEEDED by elogind
121 { "pstore", "/sys/fs/pstore", "pstore", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
124 { "efivarfs", "/sys/firmware/efi/efivars", "efivarfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
125 is_efi_boot, MNT_NONE },
127 { "kdbusfs", "/sys/fs/kdbus", "kdbusfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
128 is_kdbus_wanted, MNT_IN_CONTAINER },
132 /// UNNEEDED by elogind
134 /* These are API file systems that might be mounted by other software,
135 * we just list them here so that we know that we should ignore them */
137 static const char ignore_paths[] =
138 /* SELinux file systems */
140 /* Container bind mounts */
145 bool mount_point_is_api(const char *path) {
148 /* Checks if this mount point is considered "API", and hence
149 * should be ignored */
151 for (i = 0; i < ELEMENTSOF(mount_table); i ++)
152 if (path_equal(path, mount_table[i].where))
155 return path_startswith(path, "/sys/fs/cgroup/");
158 bool mount_point_ignore(const char *path) {
161 NULSTR_FOREACH(i, ignore_paths)
162 if (path_equal(path, i))
169 static int mount_one(const MountPoint *p, bool relabel) {
174 if (p->condition_fn && !p->condition_fn())
177 /* Relabel first, just in case */
179 label_fix(p->where, true, true);
181 r = path_is_mount_point(p->where, AT_SYMLINK_FOLLOW);
182 if (r < 0 && r != -ENOENT)
187 /* Skip securityfs in a container */
188 if (!(p->mode & MNT_IN_CONTAINER) && detect_container() > 0)
191 /* The access mode here doesn't really matter too much, since
192 * the mounted file system will take precedence anyway. */
194 mkdir_p_label(p->where, 0755);
196 mkdir_p(p->where, 0755);
198 log_debug("Mounting %s to %s of type %s with options %s.",
209 log_full((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, "Failed to mount %s at %s: %m", p->type, p->where);
210 return (p->mode & MNT_FATAL) ? -errno : 0;
213 /* Relabel again, since we now mounted something fresh here */
215 label_fix(p->where, false, false);
220 /// UNNEEDED by elogind
222 int mount_setup_early(void) {
226 assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));
228 /* Do a minimal mount of /proc and friends to enable the most
229 * basic stuff, such as SELinux */
230 for (i = 0; i < N_EARLY_MOUNT; i ++) {
233 j = mount_one(mount_table + i, false);
234 if (j != 0 && r >= 0)
241 int mount_cgroup_controllers(char ***join_controllers) {
242 _cleanup_set_free_free_ Set *controllers = NULL;
245 if (!cg_is_legacy_wanted())
248 /* Mount all available cgroup controllers that are built into the kernel. */
250 controllers = set_new(&string_hash_ops);
254 r = cg_kernel_controllers(controllers);
256 return log_error_errno(r, "Failed to enumerate cgroup controllers: %m");
259 _cleanup_free_ char *options = NULL, *controller = NULL, *where = NULL;
263 .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV,
264 .mode = MNT_IN_CONTAINER,
268 controller = set_steal_first(controllers);
272 if (join_controllers)
273 for (k = join_controllers; *k; k++)
274 if (strv_find(*k, controller))
280 for (i = *k, j = *k; *i; i++) {
282 if (!streq(*i, controller)) {
283 _cleanup_free_ char *t;
285 t = set_remove(controllers, *i);
297 options = strv_join(*k, ",");
301 options = controller;
305 where = strappend("/sys/fs/cgroup/", options);
312 r = mount_one(&p, true);
316 if (r > 0 && k && *k) {
319 for (i = *k; *i; i++) {
320 _cleanup_free_ char *t = NULL;
322 t = strappend("/sys/fs/cgroup/", *i);
326 r = symlink(options, t);
327 if (r < 0 && errno != EEXIST)
328 return log_error_errno(errno, "Failed to create symlink %s: %m", t);
329 #ifdef SMACK_RUN_LABEL
330 r = mac_smack_copy(t, options);
331 if (r < 0 && r != -EOPNOTSUPP)
332 return log_error_errno(r, "Failed to copy smack label from %s to %s: %m", options, t);
338 /* Now that we mounted everything, let's make the tmpfs the
339 * cgroup file systems are mounted into read-only. */
340 (void) mount("tmpfs", "/sys/fs/cgroup", "tmpfs", MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME|MS_RDONLY, "mode=755");
345 #if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
348 const struct stat *sb,
350 struct FTW *ftwbuf) {
352 /* No need to label /dev twice in a row... */
353 if (_unlikely_(ftwbuf->level == 0))
356 label_fix(fpath, false, false);
358 /* /run/initramfs is static data and big, no need to
359 * dynamically relabel its contents at boot... */
360 if (_unlikely_(ftwbuf->level == 1 &&
362 streq(fpath, "/run/initramfs")))
363 return FTW_SKIP_SUBTREE;
370 int mount_setup(bool loaded_policy) {
374 for (i = 0; i < ELEMENTSOF(mount_table); i ++) {
377 j = mount_one(mount_table + i, loaded_policy);
378 if (j != 0 && r >= 0)
385 /// elogind does not control /, /dev, /run and /run/systemd/* are setup elsewhere.
387 #if defined(HAVE_SELINUX) || defined(HAVE_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("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
401 after_relabel = now(CLOCK_MONOTONIC);
403 log_info("Relabelled /dev and /run in %s.",
404 format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel, 0));
408 /* Create a few default symlinks, which are normally created
409 * by udevd, but some scripts might need them before we start
411 dev_setup(NULL, UID_INVALID, GID_INVALID);
413 /* Mark the root directory as shared in regards to mount
414 * propagation. The kernel defaults to "private", but we think
415 * it makes more sense to have a default of "shared" so that
416 * nspawn and the container tools work out of the box. If
417 * specific setups need other settings they can reset the
418 * propagation mode to private if needed. */
419 if (detect_container() <= 0)
420 if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
421 log_warning_errno(errno, "Failed to set up the root directory for shared mount propagation: %m");
423 /* Create a few directories we always want around, Note that
424 * sd_booted() checks for /run/systemd/system, so this mkdir
425 * really needs to stay for good, otherwise software that
426 * copied sd-daemon.c into their sources will misdetect
428 mkdir_label("/run/systemd", 0755);
429 mkdir_label("/run/systemd/system", 0755);
430 mkdir_label("/run/systemd/inaccessible", 0000);