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 #if 0 /// UNNEEDED by elogind
77 { "sysfs", "/sys", "sysfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
78 NULL, MNT_FATAL|MNT_IN_CONTAINER },
79 { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
80 NULL, MNT_FATAL|MNT_IN_CONTAINER },
81 { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME,
82 NULL, MNT_FATAL|MNT_IN_CONTAINER },
83 { "securityfs", "/sys/kernel/security", "securityfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
86 { "smackfs", "/sys/fs/smackfs", "smackfs", "smackfsdef=*", MS_NOSUID|MS_NOEXEC|MS_NODEV,
87 mac_smack_use, MNT_FATAL },
88 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
89 mac_smack_use, MNT_FATAL },
91 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
92 NULL, MNT_FATAL|MNT_IN_CONTAINER },
93 { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC,
94 NULL, MNT_IN_CONTAINER },
96 { "tmpfs", "/run", "tmpfs", "mode=755,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
97 mac_smack_use, MNT_FATAL },
99 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
100 NULL, MNT_FATAL|MNT_IN_CONTAINER },
101 { "cgroup", "/sys/fs/cgroup", "cgroup", "__DEVEL__sane_behavior", MS_NOSUID|MS_NOEXEC|MS_NODEV,
102 cg_is_unified_wanted, MNT_FATAL|MNT_IN_CONTAINER },
104 { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
105 cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
106 #if 0 /// UNNEEDED by elogind
107 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
108 cg_is_legacy_wanted, MNT_IN_CONTAINER },
109 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV,
110 cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
112 { "cgroup", "/sys/fs/cgroup/elogind", "cgroup", "none,name=elogind,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
113 cg_is_legacy_wanted, MNT_IN_CONTAINER },
114 { "cgroup", "/sys/fs/cgroup/elogind", "cgroup", "none,name=elogind", MS_NOSUID|MS_NOEXEC|MS_NODEV,
115 cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
117 #if 0 /// UNNEEDED by elogind
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 },
124 { "kdbusfs", "/sys/fs/kdbus", "kdbusfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
125 is_kdbus_wanted, MNT_IN_CONTAINER },
129 #if 0 /// UNNEEDED by elogind
130 /* These are API file systems that might be mounted by other software,
131 * we just list them here so that we know that we should ignore them */
133 static const char ignore_paths[] =
134 /* SELinux file systems */
136 /* Container bind mounts */
141 bool mount_point_is_api(const char *path) {
144 /* Checks if this mount point is considered "API", and hence
145 * should be ignored */
147 for (i = 0; i < ELEMENTSOF(mount_table); i ++)
148 if (path_equal(path, mount_table[i].where))
151 return path_startswith(path, "/sys/fs/cgroup/");
154 bool mount_point_ignore(const char *path) {
157 NULSTR_FOREACH(i, ignore_paths)
158 if (path_equal(path, i))
165 static int mount_one(const MountPoint *p, bool relabel) {
170 if (p->condition_fn && !p->condition_fn())
173 /* Relabel first, just in case */
175 (void) label_fix(p->where, true, true);
177 r = path_is_mount_point(p->where, AT_SYMLINK_FOLLOW);
178 if (r < 0 && r != -ENOENT) {
179 log_full_errno((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, r, "Failed to determine whether %s is a mount point: %m", p->where);
180 return (p->mode & MNT_FATAL) ? r : 0;
185 /* Skip securityfs in a container */
186 if (!(p->mode & MNT_IN_CONTAINER) && detect_container() > 0)
189 /* The access mode here doesn't really matter too much, since
190 * the mounted file system will take precedence anyway. */
192 (void) mkdir_p_label(p->where, 0755);
194 (void) mkdir_p(p->where, 0755);
196 log_debug("Mounting %s to %s of type %s with options %s.",
207 log_full_errno((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, errno, "Failed to mount %s at %s: %m", p->type, p->where);
208 return (p->mode & MNT_FATAL) ? -errno : 0;
211 /* Relabel again, since we now mounted something fresh here */
213 (void) label_fix(p->where, false, false);
218 #if 0 /// UNNEEDED by elogind
219 int mount_setup_early(void) {
223 assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));
225 /* Do a minimal mount of /proc and friends to enable the most
226 * basic stuff, such as SELinux */
227 for (i = 0; i < N_EARLY_MOUNT; i ++) {
230 j = mount_one(mount_table + i, false);
231 if (j != 0 && r >= 0)
238 int mount_cgroup_controllers(char ***join_controllers) {
239 _cleanup_set_free_free_ Set *controllers = NULL;
242 if (!cg_is_legacy_wanted())
245 /* Mount all available cgroup controllers that are built into the kernel. */
247 controllers = set_new(&string_hash_ops);
251 r = cg_kernel_controllers(controllers);
253 return log_error_errno(r, "Failed to enumerate cgroup controllers: %m");
256 _cleanup_free_ char *options = NULL, *controller = NULL, *where = NULL;
260 .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV,
261 .mode = MNT_IN_CONTAINER,
265 controller = set_steal_first(controllers);
269 if (join_controllers)
270 for (k = join_controllers; *k; k++)
271 if (strv_find(*k, controller))
277 for (i = *k, j = *k; *i; i++) {
279 if (!streq(*i, controller)) {
280 _cleanup_free_ char *t;
282 t = set_remove(controllers, *i);
294 options = strv_join(*k, ",");
298 options = controller;
302 where = strappend("/sys/fs/cgroup/", options);
309 r = mount_one(&p, true);
313 if (r > 0 && k && *k) {
316 for (i = *k; *i; i++) {
317 _cleanup_free_ char *t = NULL;
319 t = strappend("/sys/fs/cgroup/", *i);
323 r = symlink(options, t);
325 #ifdef SMACK_RUN_LABEL
326 _cleanup_free_ char *src;
327 src = strappend("/sys/fs/cgroup/", options);
330 r = mac_smack_copy(t, src);
331 if (r < 0 && r != -EOPNOTSUPP)
332 return log_error_errno(r, "Failed to copy smack label from %s to %s: %m", src, t);
334 } else if (errno != EEXIST)
335 return log_error_errno(errno, "Failed to create symlink %s: %m", t);
340 /* Now that we mounted everything, let's make the tmpfs the
341 * cgroup file systems are mounted into read-only. */
342 (void) mount("tmpfs", "/sys/fs/cgroup", "tmpfs", MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME|MS_RDONLY, "mode=755");
347 #if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
350 const struct stat *sb,
352 struct FTW *ftwbuf) {
354 /* No need to label /dev twice in a row... */
355 if (_unlikely_(ftwbuf->level == 0))
358 label_fix(fpath, false, false);
360 /* /run/initramfs is static data and big, no need to
361 * dynamically relabel its contents at boot... */
362 if (_unlikely_(ftwbuf->level == 1 &&
364 streq(fpath, "/run/initramfs")))
365 return FTW_SKIP_SUBTREE;
372 int mount_setup(bool loaded_policy) {
376 for (i = 0; i < ELEMENTSOF(mount_table); i ++) {
379 j = mount_one(mount_table + i, loaded_policy);
380 if (j != 0 && r >= 0)
388 #if 0 /// elogind does not control /, /dev, /run and /run/systemd/* are setup elsewhere.
389 #if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
390 /* Nodes in devtmpfs and /run need to be manually updated for
391 * the appropriate labels, after mounting. The other virtual
392 * API file systems like /sys and /proc do not need that, they
393 * use the same label for all their files. */
395 usec_t before_relabel, after_relabel;
396 char timespan[FORMAT_TIMESPAN_MAX];
398 before_relabel = now(CLOCK_MONOTONIC);
400 nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
401 nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
403 after_relabel = now(CLOCK_MONOTONIC);
405 log_info("Relabelled /dev and /run in %s.",
406 format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel, 0));
410 /* Create a few default symlinks, which are normally created
411 * by udevd, but some scripts might need them before we start
413 dev_setup(NULL, UID_INVALID, GID_INVALID);
415 /* Mark the root directory as shared in regards to mount
416 * propagation. The kernel defaults to "private", but we think
417 * it makes more sense to have a default of "shared" so that
418 * nspawn and the container tools work out of the box. If
419 * specific setups need other settings they can reset the
420 * propagation mode to private if needed. */
421 if (detect_container() <= 0)
422 if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
423 log_warning_errno(errno, "Failed to set up the root directory for shared mount propagation: %m");
425 /* Create a few directories we always want around, Note that
426 * sd_booted() checks for /run/systemd/system, so this mkdir
427 * really needs to stay for good, otherwise software that
428 * copied sd-daemon.c into their sources will misdetect
430 mkdir_label("/run/systemd", 0755);
431 mkdir_label("/run/systemd/system", 0755);
432 mkdir_label("/run/systemd/inaccessible", 0000);