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"
46 typedef struct MountPoint {
55 /* The first three entries we might need before SELinux is up. The
56 * fourth (securityfs) is needed by IMA to load a custom policy. The
57 * other ones we can delay until SELinux and IMA are loaded. */
58 #define N_EARLY_MOUNT 4
60 static const MountPoint mount_table[] = {
61 { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
62 { "sysfs", "/sys", "sysfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
63 { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true },
64 { "securityfs", "/sys/kernel/security", "securityfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, false },
65 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
66 { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, false },
67 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
68 { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, false },
69 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV, false },
72 /* These are API file systems that might be mounted by other software,
73 * we just list them here so that we know that we should ignore them */
75 static const char ignore_paths[] =
76 /* SELinux file systems */
79 /* Legacy cgroup mount points */
82 /* Legacy kernel file system */
84 /* Container bind mounts */
92 bool mount_point_is_api(const char *path) {
95 /* Checks if this mount point is considered "API", and hence
96 * should be ignored */
98 for (i = 0; i < ELEMENTSOF(mount_table); i ++)
99 if (path_equal(path, mount_table[i].where))
102 return path_startswith(path, "/sys/fs/cgroup/");
105 bool mount_point_ignore(const char *path) {
108 NULSTR_FOREACH(i, ignore_paths)
109 if (path_equal(path, i))
115 static int mount_one(const MountPoint *p, bool relabel) {
120 /* Relabel first, just in case */
122 label_fix(p->where, true);
124 if ((r = path_is_mount_point(p->where, true)) < 0)
130 /* The access mode here doesn't really matter too much, since
131 * the mounted file system will take precedence anyway. */
132 mkdir_p(p->where, 0755);
134 log_debug("Mounting %s to %s of type %s with options %s.",
145 log_full(p->fatal ? LOG_ERR : LOG_DEBUG, "Failed to mount %s: %s", p->where, strerror(errno));
146 return p->fatal ? -errno : 0;
149 /* Relabel again, since we now mounted something fresh here */
151 label_fix(p->where, false);
156 int mount_setup_early(void) {
160 assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));
162 /* Do a minimal mount of /proc and friends to enable the most
163 * basic stuff, such as SELinux */
164 for (i = 0; i < N_EARLY_MOUNT; i ++) {
167 j = mount_one(mount_table + i, false);
175 int mount_cgroup_controllers(char ***join_controllers) {
181 /* Mount all available cgroup controllers that are built into the kernel. */
183 f = fopen("/proc/cgroups", "re");
185 log_error("Failed to enumerate cgroup controllers: %m");
189 controllers = set_new(string_hash_func, string_compare_func);
192 log_error("Failed to allocate controller set.");
196 /* Ignore the header line */
197 (void) fgets(buf, sizeof(buf), f);
203 if (fscanf(f, "%ms %*i %*i %i", &controller, &enabled) != 2) {
208 log_error("Failed to parse /proc/cgroups.");
218 r = set_put(controllers, controller);
220 log_error("Failed to add controller to set.");
228 char *controller, *where, *options;
231 controller = set_steal_first(controllers);
235 if (join_controllers)
236 for (k = join_controllers; *k; k++)
237 if (strv_find(*k, controller))
243 for (i = *k, j = *k; *i; i++) {
245 if (!streq(*i, controller)) {
248 t = set_remove(controllers, *i);
261 options = strv_join(*k, ",");
263 log_error("Failed to join options");
270 options = controller;
274 where = strappend("/sys/fs/cgroup/", options);
276 log_error("Failed to build path");
287 p.flags = MS_NOSUID|MS_NOEXEC|MS_NODEV;
290 r = mount_one(&p, true);
299 if (r > 0 && k && *k) {
302 for (i = *k; *i; i++) {
305 t = strappend("/sys/fs/cgroup/", *i);
307 log_error("Failed to build path");
313 r = symlink(options, t);
316 if (r < 0 && errno != EEXIST) {
317 log_error("Failed to create symlink: %m");
331 set_free_free(controllers);
340 const struct stat *sb,
342 struct FTW *ftwbuf) {
344 /* No need to label /dev twice in a row... */
345 if (_unlikely_(ftwbuf->level == 0))
348 label_fix(fpath, true);
350 /* /run/initramfs is static data and big, no need to
351 * dynamically relabel its contents at boot... */
352 if (_unlikely_(ftwbuf->level == 1 &&
354 streq(fpath, "/run/initramfs")))
355 return FTW_SKIP_SUBTREE;
360 int mount_setup(bool loaded_policy) {
362 static const char relabel[] =
363 "/run/initramfs/root-fsck\0"
364 "/run/initramfs/shutdown\0";
370 for (i = 0; i < ELEMENTSOF(mount_table); i ++) {
371 r = mount_one(mount_table + i, true);
377 /* Nodes in devtmpfs and /run need to be manually updated for
378 * the appropriate labels, after mounting. The other virtual
379 * API file systems like /sys and /proc do not need that, they
380 * use the same label for all their files. */
382 usec_t before_relabel, after_relabel;
383 char timespan[FORMAT_TIMESPAN_MAX];
385 before_relabel = now(CLOCK_MONOTONIC);
387 nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
388 nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
390 /* Explicitly relabel these */
391 NULSTR_FOREACH(j, relabel)
394 after_relabel = now(CLOCK_MONOTONIC);
396 log_info("Relabelled /dev and /run in %s.",
397 format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel));
400 /* Create a few default symlinks, which are normally created
401 * by udevd, but some scripts might need them before we start
405 /* Create a few directories we always want around */
406 label_mkdir("/run/systemd", 0755);
407 label_mkdir("/run/systemd/system", 0755);