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/>.
23 #include <sys/mount.h>
28 #include <sys/types.h>
30 #include <sys/syscall.h>
37 #include "path-util.h"
38 #include "namespace.h"
41 #include "loopback-setup.h"
43 #include "dev-setup.h"
46 typedef enum MountMode {
47 /* This is ordered by priority! */
56 typedef struct BindMount {
63 static int append_mounts(BindMount **p, char **strv, MountMode mode) {
68 STRV_FOREACH(i, strv) {
72 if ((mode == INACCESSIBLE || mode == READONLY || mode == READWRITE) && (*i)[0] == '-') {
77 if (!path_is_absolute(*i))
88 static int mount_path_compare(const void *a, const void *b) {
89 const BindMount *p = a, *q = b;
91 if (path_equal(p->path, q->path)) {
93 /* If the paths are equal, check the mode */
94 if (p->mode < q->mode)
97 if (p->mode > q->mode)
103 /* If the paths are not equal, then order prefixes first */
104 if (path_startswith(p->path, q->path))
107 if (path_startswith(q->path, p->path))
113 static void drop_duplicates(BindMount *m, unsigned *n) {
114 BindMount *f, *t, *previous;
119 for (f = m, t = m, previous = NULL; f < m+*n; f++) {
121 /* The first one wins */
122 if (previous && path_equal(f->path, previous->path))
136 static int mount_dev(BindMount *m) {
137 static const char devnodes[] =
145 char temporary_mount[] = "/tmp/namespace-dev-XXXXXX";
146 const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devkdbus = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL;
147 _cleanup_umask_ mode_t u;
154 if (!mkdtemp(temporary_mount))
157 dev = strappenda(temporary_mount, "/dev");
159 if (mount("tmpfs", dev, "tmpfs", MS_NOSUID|MS_STRICTATIME, "mode=755") < 0) {
164 devpts = strappenda(temporary_mount, "/dev/pts");
166 if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) {
171 devptmx = strappenda(temporary_mount, "/dev/ptmx");
172 symlink("pts/ptmx", devptmx);
174 devshm = strappenda(temporary_mount, "/dev/shm");
175 mkdir(devshm, 01777);
176 r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL);
182 devmqueue = strappenda(temporary_mount, "/dev/mqueue");
183 mkdir(devmqueue, 0755);
184 mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL);
186 devkdbus = strappenda(temporary_mount, "/dev/kdbus");
187 mkdir(devkdbus, 0755);
188 mount("/dev/kdbus", devkdbus, NULL, MS_BIND, NULL);
190 devhugepages = strappenda(temporary_mount, "/dev/hugepages");
191 mkdir(devhugepages, 0755);
192 mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL);
194 devlog = strappenda(temporary_mount, "/dev/log");
195 symlink("/run/systemd/journal/dev-log", devlog);
197 NULSTR_FOREACH(d, devnodes) {
198 _cleanup_free_ char *dn = NULL;
211 if (!S_ISBLK(st.st_mode) &&
212 !S_ISCHR(st.st_mode)) {
220 dn = strappend(temporary_mount, d);
226 r = mknod(dn, st.st_mode, st.st_rdev);
233 dev_setup(temporary_mount);
235 if (mount(dev, "/dev/", NULL, MS_MOVE, NULL) < 0) {
241 rmdir(temporary_mount);
256 umount(devhugepages);
266 rmdir(temporary_mount);
271 static int apply_mount(
274 const char *var_tmp_dir) {
287 what = "/run/systemd/inaccessible";
299 case PRIVATE_VAR_TMP:
304 assert_not_reached("Unknown mode");
309 r = mount(what, m->path, NULL, MS_BIND|MS_REC, NULL);
311 log_debug("Successfully mounted %s to %s", what, m->path);
312 else if (m->ignore && errno == ENOENT)
318 static int make_read_only(BindMount *m) {
323 if (m->mode != INACCESSIBLE && m->mode != READONLY)
326 r = mount(NULL, m->path, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_REC, NULL);
327 if (r < 0 && !(m->ignore && errno == ENOENT))
334 char** read_write_dirs,
335 char** read_only_dirs,
336 char** inaccessible_dirs,
340 ProtectHome protect_home,
341 ProtectSystem protect_system,
342 unsigned mount_flags) {
344 BindMount *m, *mounts = NULL;
348 if (mount_flags == 0)
349 mount_flags = MS_SHARED;
351 if (unshare(CLONE_NEWNS) < 0)
354 n = !!tmp_dir + !!var_tmp_dir +
355 strv_length(read_write_dirs) +
356 strv_length(read_only_dirs) +
357 strv_length(inaccessible_dirs) +
359 (protect_home != PROTECT_HOME_NO ? 2 : 0) +
360 (protect_system != PROTECT_SYSTEM_NO ? 2 : 0) +
361 (protect_system == PROTECT_SYSTEM_FULL ? 1 : 0);
364 m = mounts = (BindMount *) alloca(n * sizeof(BindMount));
365 r = append_mounts(&m, read_write_dirs, READWRITE);
369 r = append_mounts(&m, read_only_dirs, READONLY);
373 r = append_mounts(&m, inaccessible_dirs, INACCESSIBLE);
379 m->mode = PRIVATE_TMP;
384 m->path = "/var/tmp";
385 m->mode = PRIVATE_VAR_TMP;
391 m->mode = PRIVATE_DEV;
395 if (protect_home != PROTECT_HOME_NO) {
396 r = append_mounts(&m, STRV_MAKE("-/home", "-/run/user"), protect_home == PROTECT_HOME_READ_ONLY ? READONLY : INACCESSIBLE);
401 if (protect_system != PROTECT_SYSTEM_NO) {
402 r = append_mounts(&m, protect_system == PROTECT_SYSTEM_FULL ? STRV_MAKE("/usr", "/etc", "-/boot") : STRV_MAKE("/usr", "-/boot"), READONLY);
407 assert(mounts + n == m);
409 qsort(mounts, n, sizeof(BindMount), mount_path_compare);
410 drop_duplicates(mounts, &n);
414 /* Remount / as SLAVE so that nothing now mounted in the namespace
415 shows up in the parent */
416 if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0)
419 for (m = mounts; m < mounts + n; ++m) {
420 r = apply_mount(m, tmp_dir, var_tmp_dir);
425 for (m = mounts; m < mounts + n; ++m) {
426 r = make_read_only(m);
432 /* Remount / as the desired mode. Not that this will not
433 * reestablish propagation from our side to the host, since
434 * what's disconnected is disconnected. */
435 if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) {
444 for (m = mounts; m < mounts + n; ++m)
446 umount2(m->path, MNT_DETACH);
452 static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) {
453 _cleanup_free_ char *x = NULL;
454 char bid[SD_ID128_STRING_MAX];
462 /* We include the boot id in the directory so that after a
463 * reboot we can easily identify obsolete directories. */
465 r = sd_id128_get_boot(&boot_id);
469 x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX", NULL);
477 RUN_WITH_UMASK(0000) {
480 y = strappenda(x, "/tmp");
482 if (mkdir(y, 0777 | S_ISVTX) < 0)
492 int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) {
500 r = setup_one_tmp_dir(id, "/tmp", &a);
504 r = setup_one_tmp_dir(id, "/var/tmp", &b);
508 t = strappenda(a, "/tmp");
522 int setup_netns(int netns_storage_socket[2]) {
523 _cleanup_close_ int netns = -1;
525 struct cmsghdr cmsghdr;
526 uint8_t buf[CMSG_SPACE(sizeof(int))];
529 .msg_control = &control,
530 .msg_controllen = sizeof(control),
532 struct cmsghdr *cmsg;
535 assert(netns_storage_socket);
536 assert(netns_storage_socket[0] >= 0);
537 assert(netns_storage_socket[1] >= 0);
539 /* We use the passed socketpair as a storage buffer for our
540 * namespace reference fd. Whatever process runs this first
541 * shall create a new namespace, all others should just join
542 * it. To serialize that we use a file lock on the socket
545 * It's a bit crazy, but hey, works great! */
547 if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0)
550 if (recvmsg(netns_storage_socket[0], &mh, MSG_DONTWAIT|MSG_CMSG_CLOEXEC) < 0) {
551 if (errno != EAGAIN) {
556 /* Nothing stored yet, so let's create a new namespace */
558 if (unshare(CLONE_NEWNET) < 0) {
565 netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY);
573 /* Yay, found something, so let's join the namespace */
575 for (cmsg = CMSG_FIRSTHDR(&mh); cmsg; cmsg = CMSG_NXTHDR(&mh, cmsg)) {
576 if (cmsg->cmsg_level == SOL_SOCKET && cmsg->cmsg_type == SCM_RIGHTS) {
577 assert(cmsg->cmsg_len == CMSG_LEN(sizeof(int)));
578 netns = *(int*) CMSG_DATA(cmsg);
582 if (setns(netns, CLONE_NEWNET) < 0) {
590 cmsg = CMSG_FIRSTHDR(&mh);
591 cmsg->cmsg_level = SOL_SOCKET;
592 cmsg->cmsg_type = SCM_RIGHTS;
593 cmsg->cmsg_len = CMSG_LEN(sizeof(int));
594 memcpy(CMSG_DATA(cmsg), &netns, sizeof(int));
595 mh.msg_controllen = cmsg->cmsg_len;
597 if (sendmsg(netns_storage_socket[1], &mh, MSG_DONTWAIT|MSG_NOSIGNAL) < 0) {
603 lockf(netns_storage_socket[0], F_ULOCK, 0);
608 static const char *const protect_home_table[_PROTECT_HOME_MAX] = {
609 [PROTECT_HOME_NO] = "no",
610 [PROTECT_HOME_YES] = "yes",
611 [PROTECT_HOME_READ_ONLY] = "read-only",
614 DEFINE_STRING_TABLE_LOOKUP(protect_home, ProtectHome);
616 static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = {
617 [PROTECT_SYSTEM_NO] = "no",
618 [PROTECT_SYSTEM_YES] = "yes",
619 [PROTECT_SYSTEM_FULL] = "full",
622 DEFINE_STRING_TABLE_LOOKUP(protect_system, ProtectSystem);