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/>.
22 #include <sys/mount.h>
32 #include "mount-setup.h"
33 #include "dev-setup.h"
41 #include "path-util.h"
49 typedef struct MountPoint {
59 /* The first three entries we might need before SELinux is up. The
60 * fourth (securityfs) is needed by IMA to load a custom policy. The
61 * other ones we can delay until SELinux and IMA are loaded. */
62 #define N_EARLY_MOUNT 4
64 static const MountPoint mount_table[] = {
65 { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true, true },
66 { "sysfs", "/sys", "sysfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true, true },
67 { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true, true },
68 { "securityfs", "/sys/kernel/security", "securityfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, false, false },
69 { "efivarfs", "/sys/firmware/efi/efivars", "efivarfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, false, false },
70 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true, true },
71 { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, false, true },
72 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true, true },
73 { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, false, true },
74 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV, false, true },
77 /* These are API file systems that might be mounted by other software,
78 * we just list them here so that we know that we should ignore them */
80 static const char ignore_paths[] =
81 /* SELinux file systems */
84 /* Legacy cgroup mount points */
87 /* Legacy kernel file system */
89 /* Container bind mounts */
94 bool mount_point_is_api(const char *path) {
97 /* Checks if this mount point is considered "API", and hence
98 * should be ignored */
100 for (i = 0; i < ELEMENTSOF(mount_table); i ++)
101 if (path_equal(path, mount_table[i].where))
104 return path_startswith(path, "/sys/fs/cgroup/");
107 bool mount_point_ignore(const char *path) {
110 NULSTR_FOREACH(i, ignore_paths)
111 if (path_equal(path, i))
117 static int mount_one(const MountPoint *p, bool relabel) {
122 /* Relabel first, just in case */
124 label_fix(p->where, true, true);
126 r = path_is_mount_point(p->where, true);
133 /* Skip securityfs in a container */
134 if (!p->in_container && detect_container(NULL) > 0)
137 /* The access mode here doesn't really matter too much, since
138 * the mounted file system will take precedence anyway. */
139 mkdir_p_label(p->where, 0755);
141 log_debug("Mounting %s to %s of type %s with options %s.",
152 log_full(p->fatal ? LOG_ERR : LOG_DEBUG, "Failed to mount %s: %s", p->where, strerror(errno));
153 return p->fatal ? -errno : 0;
156 /* Relabel again, since we now mounted something fresh here */
158 label_fix(p->where, false, false);
163 int mount_setup_early(void) {
167 assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));
169 /* Do a minimal mount of /proc and friends to enable the most
170 * basic stuff, such as SELinux */
171 for (i = 0; i < N_EARLY_MOUNT; i ++) {
174 j = mount_one(mount_table + i, false);
182 int mount_cgroup_controllers(char ***join_controllers) {
188 /* Mount all available cgroup controllers that are built into the kernel. */
190 f = fopen("/proc/cgroups", "re");
192 log_error("Failed to enumerate cgroup controllers: %m");
196 controllers = set_new(string_hash_func, string_compare_func);
202 /* Ignore the header line */
203 (void) fgets(buf, sizeof(buf), f);
209 if (fscanf(f, "%ms %*i %*i %i", &controller, &enabled) != 2) {
214 log_error("Failed to parse /proc/cgroups.");
224 r = set_put(controllers, controller);
226 log_error("Failed to add controller to set.");
234 char *controller, *where, *options;
237 controller = set_steal_first(controllers);
241 if (join_controllers)
242 for (k = join_controllers; *k; k++)
243 if (strv_find(*k, controller))
249 for (i = *k, j = *k; *i; i++) {
251 if (!streq(*i, controller)) {
254 t = set_remove(controllers, *i);
267 options = strv_join(*k, ",");
275 options = controller;
279 where = strappend("/sys/fs/cgroup/", options);
291 p.flags = MS_NOSUID|MS_NOEXEC|MS_NODEV;
294 r = mount_one(&p, true);
303 if (r > 0 && k && *k) {
306 for (i = *k; *i; i++) {
309 t = strappend("/sys/fs/cgroup/", *i);
316 r = symlink(options, t);
319 if (r < 0 && errno != EEXIST) {
320 log_error("Failed to create symlink: %m");
334 set_free_free(controllers);
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;
363 int mount_setup(bool loaded_policy) {
365 static const char relabel[] =
366 "/run/initramfs/root-fsck\0"
367 "/run/initramfs/shutdown\0";
373 for (i = 0; i < ELEMENTSOF(mount_table); i ++) {
374 r = mount_one(mount_table + i, true);
380 /* Nodes in devtmpfs and /run need to be manually updated for
381 * the appropriate labels, after mounting. The other virtual
382 * API file systems like /sys and /proc do not need that, they
383 * use the same label for all their files. */
385 usec_t before_relabel, after_relabel;
386 char timespan[FORMAT_TIMESPAN_MAX];
388 before_relabel = now(CLOCK_MONOTONIC);
390 nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
391 nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
393 /* Explicitly relabel these */
394 NULSTR_FOREACH(j, relabel)
395 label_fix(j, true, false);
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));
403 /* Create a few default symlinks, which are normally created
404 * by udevd, but some scripts might need them before we start
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(NULL) <= 0)
415 if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
416 log_warning("Failed to set up the root directory for shared mount propagation: %m");
418 /* Create a few directories we always want around */
419 mkdir_label("/run/systemd", 0755);
420 mkdir_label("/run/systemd/system", 0755);