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 label_fix(p->where, true, true);
177 r = path_is_mount_point(p->where, AT_SYMLINK_FOLLOW);
178 if (r < 0 && r != -ENOENT)
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 mkdir_p_label(p->where, 0755);
192 mkdir_p(p->where, 0755);
194 log_debug("Mounting %s to %s of type %s with options %s.",
205 log_full((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, "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 label_fix(p->where, false, false);
216 #if 0 /// UNNEEDED by elogind
217 int mount_setup_early(void) {
221 assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));
223 /* Do a minimal mount of /proc and friends to enable the most
224 * basic stuff, such as SELinux */
225 for (i = 0; i < N_EARLY_MOUNT; i ++) {
228 j = mount_one(mount_table + i, false);
229 if (j != 0 && r >= 0)
236 int mount_cgroup_controllers(char ***join_controllers) {
237 _cleanup_set_free_free_ Set *controllers = NULL;
240 if (!cg_is_legacy_wanted())
243 /* Mount all available cgroup controllers that are built into the kernel. */
245 controllers = set_new(&string_hash_ops);
249 r = cg_kernel_controllers(controllers);
251 return log_error_errno(r, "Failed to enumerate cgroup controllers: %m");
254 _cleanup_free_ char *options = NULL, *controller = NULL, *where = NULL;
258 .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV,
259 .mode = MNT_IN_CONTAINER,
263 controller = set_steal_first(controllers);
267 if (join_controllers)
268 for (k = join_controllers; *k; k++)
269 if (strv_find(*k, controller))
275 for (i = *k, j = *k; *i; i++) {
277 if (!streq(*i, controller)) {
278 _cleanup_free_ char *t;
280 t = set_remove(controllers, *i);
292 options = strv_join(*k, ",");
296 options = controller;
300 where = strappend("/sys/fs/cgroup/", options);
307 r = mount_one(&p, true);
311 if (r > 0 && k && *k) {
314 for (i = *k; *i; i++) {
315 _cleanup_free_ char *t = NULL;
317 t = strappend("/sys/fs/cgroup/", *i);
321 r = symlink(options, t);
322 if (r < 0 && errno != EEXIST)
323 return log_error_errno(errno, "Failed to create symlink %s: %m", t);
324 #ifdef SMACK_RUN_LABEL
325 r = mac_smack_copy(t, options);
326 if (r < 0 && r != -EOPNOTSUPP)
327 return log_error_errno(r, "Failed to copy smack label from %s to %s: %m", options, t);
333 /* Now that we mounted everything, let's make the tmpfs the
334 * cgroup file systems are mounted into read-only. */
335 (void) mount("tmpfs", "/sys/fs/cgroup", "tmpfs", MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME|MS_RDONLY, "mode=755");
340 #if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
343 const struct stat *sb,
345 struct FTW *ftwbuf) {
347 /* No need to label /dev twice in a row... */
348 if (_unlikely_(ftwbuf->level == 0))
351 label_fix(fpath, false, false);
353 /* /run/initramfs is static data and big, no need to
354 * dynamically relabel its contents at boot... */
355 if (_unlikely_(ftwbuf->level == 1 &&
357 streq(fpath, "/run/initramfs")))
358 return FTW_SKIP_SUBTREE;
365 int mount_setup(bool loaded_policy) {
369 for (i = 0; i < ELEMENTSOF(mount_table); i ++) {
372 j = mount_one(mount_table + i, loaded_policy);
373 if (j != 0 && r >= 0)
381 #if 0 /// elogind does not control /, /dev, /run and /run/systemd/* are setup elsewhere.
382 #if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
383 /* Nodes in devtmpfs and /run need to be manually updated for
384 * the appropriate labels, after mounting. The other virtual
385 * API file systems like /sys and /proc do not need that, they
386 * use the same label for all their files. */
388 usec_t before_relabel, after_relabel;
389 char timespan[FORMAT_TIMESPAN_MAX];
391 before_relabel = now(CLOCK_MONOTONIC);
393 nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
394 nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
396 after_relabel = now(CLOCK_MONOTONIC);
398 log_info("Relabelled /dev and /run in %s.",
399 format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel, 0));
403 /* Create a few default symlinks, which are normally created
404 * by udevd, but some scripts might need them before we start
406 dev_setup(NULL, UID_INVALID, GID_INVALID);
408 /* Mark the root directory as shared in regards to mount
409 * propagation. The kernel defaults to "private", but we think
410 * it makes more sense to have a default of "shared" so that
411 * nspawn and the container tools work out of the box. If
412 * specific setups need other settings they can reset the
413 * propagation mode to private if needed. */
414 if (detect_container() <= 0)
415 if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
416 log_warning_errno(errno, "Failed to set up the root directory for shared mount propagation: %m");
418 /* Create a few directories we always want around, Note that
419 * sd_booted() checks for /run/systemd/system, so this mkdir
420 * really needs to stay for good, otherwise software that
421 * copied sd-daemon.c into their sources will misdetect
423 mkdir_label("/run/systemd", 0755);
424 mkdir_label("/run/systemd/system", 0755);
425 mkdir_label("/run/systemd/inaccessible", 0000);