1 /* SPDX-License-Identifier: LGPL-2.1+ */
3 This file is part of systemd.
5 Copyright 2010 Lennart Poettering
7 systemd is free software; you can redistribute it and/or modify it
8 under the terms of the GNU Lesser General Public License as published by
9 the Free Software Foundation; either version 2.1 of the License, or
10 (at your option) any later version.
12 systemd is distributed in the hope that it will be useful, but
13 WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 Lesser General Public License for more details.
17 You should have received a copy of the GNU Lesser General Public License
18 along with systemd; If not, see <http://www.gnu.org/licenses/>.
25 #include <sys/mount.h>
30 #include "alloc-util.h"
31 #include "base-filesystem.h"
32 #include "dev-setup.h"
36 #include "loop-util.h"
37 #include "loopback-setup.h"
40 #include "mount-util.h"
41 #include "namespace.h"
42 #include "path-util.h"
43 #include "selinux-util.h"
44 #include "socket-util.h"
45 #include "stat-util.h"
46 #include "string-table.h"
47 #include "string-util.h"
49 #include "umask-util.h"
50 #include "user-util.h"
53 #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC)
55 typedef enum MountMode
{
56 /* This is ordered by priority! */
71 typedef struct MountEntry
{
72 const char *path_const
; /* Memory allocated on stack or static */
74 bool ignore
:1; /* Ignore if path does not exist? */
75 bool has_prefix
:1; /* Already is prefixed by the root dir? */
76 bool read_only
:1; /* Shall this mount point be read-only? */
77 char *path_malloc
; /* Use this instead of 'path_const' if we had to allocate memory */
78 const char *source_const
; /* The source path, for bind mounts */
80 const char *options_const
;/* Mount options for tmpfs */
82 unsigned long flags
; /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */
85 /* If MountAPIVFS= is used, let's mount /sys and /proc into the it, but only as a fallback if the user hasn't mounted
86 * something there already. These mounts are hence overriden by any other explicitly configured mounts. */
87 static const MountEntry apivfs_table
[] = {
88 { "/proc", PROCFS
, false },
89 { "/dev", BIND_DEV
, false },
90 { "/sys", SYSFS
, false },
93 /* ProtectKernelTunables= option and the related filesystem APIs */
94 static const MountEntry protect_kernel_tunables_table
[] = {
95 { "/proc/sys", READONLY
, false },
96 { "/proc/sysrq-trigger", READONLY
, true },
97 { "/proc/latency_stats", READONLY
, true },
98 { "/proc/mtrr", READONLY
, true },
99 { "/proc/apm", READONLY
, true }, /* Obsolete API, there's no point in permitting access to this, ever */
100 { "/proc/acpi", READONLY
, true },
101 { "/proc/timer_stats", READONLY
, true },
102 { "/proc/asound", READONLY
, true },
103 { "/proc/bus", READONLY
, true },
104 { "/proc/fs", READONLY
, true },
105 { "/proc/irq", READONLY
, true },
106 { "/sys", READONLY
, false },
107 { "/sys/kernel/debug", READONLY
, true },
108 { "/sys/kernel/tracing", READONLY
, true },
109 { "/sys/fs/bpf", READONLY
, true },
110 { "/sys/fs/cgroup", READWRITE
, false }, /* READONLY is set by ProtectControlGroups= option */
111 { "/sys/fs/selinux", READWRITE
, true },
114 /* ProtectKernelModules= option */
115 static const MountEntry protect_kernel_modules_table
[] = {
117 { "/lib/modules", INACCESSIBLE
, true },
119 { "/usr/lib/modules", INACCESSIBLE
, true },
123 * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of
124 * system should be protected by ProtectSystem=
126 static const MountEntry protect_home_read_only_table
[] = {
127 { "/home", READONLY
, true },
128 { "/run/user", READONLY
, true },
129 { "/root", READONLY
, true },
132 /* ProtectHome=tmpfs table */
133 static const MountEntry protect_home_tmpfs_table
[] = {
134 { "/home", TMPFS
, true, .read_only
= true, .options_const
= "mode=0755", .flags
= MS_NODEV
|MS_STRICTATIME
},
135 { "/run/user", TMPFS
, true, .read_only
= true, .options_const
= "mode=0755", .flags
= MS_NODEV
|MS_STRICTATIME
},
136 { "/root", TMPFS
, true, .read_only
= true, .options_const
= "mode=0700", .flags
= MS_NODEV
|MS_STRICTATIME
},
139 /* ProtectHome=yes table */
140 static const MountEntry protect_home_yes_table
[] = {
141 { "/home", INACCESSIBLE
, true },
142 { "/run/user", INACCESSIBLE
, true },
143 { "/root", INACCESSIBLE
, true },
146 /* ProtectSystem=yes table */
147 static const MountEntry protect_system_yes_table
[] = {
148 { "/usr", READONLY
, false },
149 { "/boot", READONLY
, true },
150 { "/efi", READONLY
, true },
152 { "/lib", READONLY
, true },
153 { "/lib64", READONLY
, true },
154 { "/bin", READONLY
, true },
156 { "/sbin", READONLY
, true },
161 /* ProtectSystem=full includes ProtectSystem=yes */
162 static const MountEntry protect_system_full_table
[] = {
163 { "/usr", READONLY
, false },
164 { "/boot", READONLY
, true },
165 { "/efi", READONLY
, true },
166 { "/etc", READONLY
, false },
168 { "/lib", READONLY
, true },
169 { "/lib64", READONLY
, true },
170 { "/bin", READONLY
, true },
172 { "/sbin", READONLY
, true },
178 * ProtectSystem=strict table. In this strict mode, we mount everything
179 * read-only, except for /proc, /dev, /sys which are the kernel API VFS,
180 * which are left writable, but PrivateDevices= + ProtectKernelTunables=
181 * protect those, and these options should be fully orthogonal.
182 * (And of course /home and friends are also left writable, as ProtectHome=
183 * shall manage those, orthogonally).
185 static const MountEntry protect_system_strict_table
[] = {
186 { "/", READONLY
, false },
187 { "/proc", READWRITE
, false }, /* ProtectKernelTunables= */
188 { "/sys", READWRITE
, false }, /* ProtectKernelTunables= */
189 { "/dev", READWRITE
, false }, /* PrivateDevices= */
190 { "/home", READWRITE
, true }, /* ProtectHome= */
191 { "/run/user", READWRITE
, true }, /* ProtectHome= */
192 { "/root", READWRITE
, true }, /* ProtectHome= */
195 static const char *mount_entry_path(const MountEntry
*p
) {
198 /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that,
199 * otherwise the stack/static ->path field is returned. */
201 return p
->path_malloc
?: p
->path_const
;
204 static bool mount_entry_read_only(const MountEntry
*p
) {
207 return p
->read_only
|| IN_SET(p
->mode
, READONLY
, INACCESSIBLE
);
210 static const char *mount_entry_source(const MountEntry
*p
) {
213 return p
->source_malloc
?: p
->source_const
;
216 static const char *mount_entry_options(const MountEntry
*p
) {
219 return p
->options_malloc
?: p
->options_const
;
222 static void mount_entry_done(MountEntry
*p
) {
225 p
->path_malloc
= mfree(p
->path_malloc
);
226 p
->source_malloc
= mfree(p
->source_malloc
);
227 p
->options_malloc
= mfree(p
->options_malloc
);
230 static int append_access_mounts(MountEntry
**p
, char **strv
, MountMode mode
, bool forcibly_require_prefix
) {
235 /* Adds a list of user-supplied READWRITE/READONLY/INACCESSIBLE entries */
237 STRV_FOREACH(i
, strv
) {
238 bool ignore
= false, needs_prefix
= false;
241 /* Look for any prefixes */
242 if (startswith(e
, "-")) {
246 if (startswith(e
, "+")) {
251 if (!path_is_absolute(e
))
254 *((*p
)++) = (MountEntry
) {
258 .has_prefix
= !needs_prefix
&& !forcibly_require_prefix
,
265 static int append_empty_dir_mounts(MountEntry
**p
, char **strv
) {
270 /* Adds tmpfs mounts to provide readable but empty directories. This is primarily used to implement the
271 * "/private/" boundary directories for DynamicUser=1. */
273 STRV_FOREACH(i
, strv
) {
275 *((*p
)++) = (MountEntry
) {
281 .options_const
= "mode=755",
282 .flags
= MS_NOSUID
|MS_NOEXEC
|MS_NODEV
|MS_STRICTATIME
,
289 static int append_bind_mounts(MountEntry
**p
, const BindMount
*binds
, unsigned n
) {
294 for (i
= 0; i
< n
; i
++) {
295 const BindMount
*b
= binds
+ i
;
297 *((*p
)++) = (MountEntry
) {
298 .path_const
= b
->destination
,
299 .mode
= b
->recursive
? BIND_MOUNT_RECURSIVE
: BIND_MOUNT
,
300 .read_only
= b
->read_only
,
301 .source_const
= b
->source
,
302 .ignore
= b
->ignore_enoent
,
309 static int append_tmpfs_mounts(MountEntry
**p
, const TemporaryFileSystem
*tmpfs
, unsigned n
) {
315 for (i
= 0; i
< n
; i
++) {
316 const TemporaryFileSystem
*t
= tmpfs
+ i
;
317 _cleanup_free_
char *o
= NULL
, *str
= NULL
;
318 unsigned long flags
= MS_NODEV
|MS_STRICTATIME
;
321 if (!path_is_absolute(t
->path
))
324 if (!isempty(t
->options
)) {
325 str
= strjoin("mode=0755,", t
->options
);
329 r
= mount_option_mangle(str
, MS_NODEV
|MS_STRICTATIME
, &flags
, &o
);
333 ro
= !!(flags
& MS_RDONLY
);
338 *((*p
)++) = (MountEntry
) {
339 .path_const
= t
->path
,
352 static int append_static_mounts(MountEntry
**p
, const MountEntry
*mounts
, unsigned n
, bool ignore_protect
) {
358 /* Adds a list of static pre-defined entries */
360 for (i
= 0; i
< n
; i
++)
361 *((*p
)++) = (MountEntry
) {
362 .path_const
= mount_entry_path(mounts
+i
),
363 .mode
= mounts
[i
].mode
,
364 .ignore
= mounts
[i
].ignore
|| ignore_protect
,
370 static int append_protect_home(MountEntry
**p
, ProtectHome protect_home
, bool ignore_protect
) {
373 switch (protect_home
) {
375 case PROTECT_HOME_NO
:
378 case PROTECT_HOME_READ_ONLY
:
379 return append_static_mounts(p
, protect_home_read_only_table
, ELEMENTSOF(protect_home_read_only_table
), ignore_protect
);
381 case PROTECT_HOME_TMPFS
:
382 return append_static_mounts(p
, protect_home_tmpfs_table
, ELEMENTSOF(protect_home_tmpfs_table
), ignore_protect
);
384 case PROTECT_HOME_YES
:
385 return append_static_mounts(p
, protect_home_yes_table
, ELEMENTSOF(protect_home_yes_table
), ignore_protect
);
388 assert_not_reached("Unexpected ProtectHome= value");
392 static int append_protect_system(MountEntry
**p
, ProtectSystem protect_system
, bool ignore_protect
) {
395 switch (protect_system
) {
397 case PROTECT_SYSTEM_NO
:
400 case PROTECT_SYSTEM_STRICT
:
401 return append_static_mounts(p
, protect_system_strict_table
, ELEMENTSOF(protect_system_strict_table
), ignore_protect
);
403 case PROTECT_SYSTEM_YES
:
404 return append_static_mounts(p
, protect_system_yes_table
, ELEMENTSOF(protect_system_yes_table
), ignore_protect
);
406 case PROTECT_SYSTEM_FULL
:
407 return append_static_mounts(p
, protect_system_full_table
, ELEMENTSOF(protect_system_full_table
), ignore_protect
);
410 assert_not_reached("Unexpected ProtectSystem= value");
414 static int mount_path_compare(const void *a
, const void *b
) {
415 const MountEntry
*p
= a
, *q
= b
;
418 /* If the paths are not equal, then order prefixes first */
419 d
= path_compare(mount_entry_path(p
), mount_entry_path(q
));
423 /* If the paths are equal, check the mode */
424 if (p
->mode
< q
->mode
)
427 if (p
->mode
> q
->mode
)
433 static int prefix_where_needed(MountEntry
*m
, unsigned n
, const char *root_directory
) {
436 /* Prefixes all paths in the bind mount table with the root directory if it is specified and the entry needs
442 for (i
= 0; i
< n
; i
++) {
448 s
= prefix_root(root_directory
, mount_entry_path(m
+i
));
452 free_and_replace(m
[i
].path_malloc
, s
);
453 m
[i
].has_prefix
= true;
459 static void drop_duplicates(MountEntry
*m
, unsigned *n
) {
460 MountEntry
*f
, *t
, *previous
;
465 /* Drops duplicate entries. Expects that the array is properly ordered already. */
467 for (f
= m
, t
= m
, previous
= NULL
; f
< m
+ *n
; f
++) {
469 /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare()
471 if (previous
&& path_equal(mount_entry_path(f
), mount_entry_path(previous
))) {
472 log_debug("%s is duplicate.", mount_entry_path(f
));
473 previous
->read_only
= previous
->read_only
|| mount_entry_read_only(f
); /* Propagate the read-only flag to the remaining entry */
486 static void drop_inaccessible(MountEntry
*m
, unsigned *n
) {
488 const char *clear
= NULL
;
493 /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly
494 * ordered already. */
496 for (f
= m
, t
= m
; f
< m
+ *n
; f
++) {
498 /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop
499 * it, as inaccessible paths really should drop the entire subtree. */
500 if (clear
&& path_startswith(mount_entry_path(f
), clear
)) {
501 log_debug("%s is masked by %s.", mount_entry_path(f
), clear
);
506 clear
= f
->mode
== INACCESSIBLE
? mount_entry_path(f
) : NULL
;
515 static void drop_nop(MountEntry
*m
, unsigned *n
) {
521 /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the
522 * list is ordered by prefixes. */
524 for (f
= m
, t
= m
; f
< m
+ *n
; f
++) {
526 /* Only suppress such subtrees for READONLY and READWRITE entries */
527 if (IN_SET(f
->mode
, READONLY
, READWRITE
)) {
531 /* Now let's find the first parent of the entry we are looking at. */
532 for (p
= t
-1; p
>= m
; p
--) {
533 if (path_startswith(mount_entry_path(f
), mount_entry_path(p
))) {
539 /* We found it, let's see if it's the same mode, if so, we can drop this entry */
540 if (found
&& p
->mode
== f
->mode
) {
541 log_debug("%s is redundant by %s", mount_entry_path(f
), mount_entry_path(p
));
554 static void drop_outside_root(const char *root_directory
, MountEntry
*m
, unsigned *n
) {
564 /* Drops all mounts that are outside of the root directory. */
566 for (f
= m
, t
= m
; f
< m
+ *n
; f
++) {
568 if (!path_startswith(mount_entry_path(f
), root_directory
)) {
569 log_debug("%s is outside of root directory.", mount_entry_path(f
));
581 static int clone_device_node(const char *d
, const char *temporary_mount
) {
586 if (stat(d
, &st
) < 0) {
592 if (!S_ISBLK(st
.st_mode
) &&
593 !S_ISCHR(st
.st_mode
))
599 dn
= strjoina(temporary_mount
, d
);
601 mac_selinux_create_file_prepare(d
, st
.st_mode
);
602 r
= mknod(dn
, st
.st_mode
, st
.st_rdev
);
603 mac_selinux_create_file_clear();
605 return log_debug_errno(errno
, "mknod failed for %s: %m", d
);
610 static int mount_private_dev(MountEntry
*m
) {
611 static const char devnodes
[] =
619 char temporary_mount
[] = "/tmp/namespace-dev-XXXXXX";
620 const char *d
, *dev
= NULL
, *devpts
= NULL
, *devshm
= NULL
, *devhugepages
= NULL
, *devmqueue
= NULL
, *devlog
= NULL
, *devptmx
= NULL
;
621 _cleanup_umask_ mode_t u
;
628 if (!mkdtemp(temporary_mount
))
631 dev
= strjoina(temporary_mount
, "/dev");
632 (void) mkdir(dev
, 0755);
633 if (mount("tmpfs", dev
, "tmpfs", DEV_MOUNT_OPTIONS
, "mode=755") < 0) {
638 devpts
= strjoina(temporary_mount
, "/dev/pts");
639 (void) mkdir(devpts
, 0755);
640 if (mount("/dev/pts", devpts
, NULL
, MS_BIND
, NULL
) < 0) {
645 /* /dev/ptmx can either be a device node or a symlink to /dev/pts/ptmx
646 * when /dev/ptmx a device node, /dev/pts/ptmx has 000 permissions making it inaccessible
647 * thus, in that case make a clone
649 * in nspawn and other containers it will be a symlink, in that case make it a symlink
651 r
= is_symlink("/dev/ptmx");
655 devptmx
= strjoina(temporary_mount
, "/dev/ptmx");
656 if (symlink("pts/ptmx", devptmx
) < 0) {
661 r
= clone_device_node("/dev/ptmx", temporary_mount
);
670 devshm
= strjoina(temporary_mount
, "/dev/shm");
671 (void) mkdir(devshm
, 0755);
672 r
= mount("/dev/shm", devshm
, NULL
, MS_BIND
, NULL
);
678 devmqueue
= strjoina(temporary_mount
, "/dev/mqueue");
679 (void) mkdir(devmqueue
, 0755);
680 (void) mount("/dev/mqueue", devmqueue
, NULL
, MS_BIND
, NULL
);
682 devhugepages
= strjoina(temporary_mount
, "/dev/hugepages");
683 (void) mkdir(devhugepages
, 0755);
684 (void) mount("/dev/hugepages", devhugepages
, NULL
, MS_BIND
, NULL
);
686 devlog
= strjoina(temporary_mount
, "/dev/log");
687 (void) symlink("/run/systemd/journal/dev-log", devlog
);
689 NULSTR_FOREACH(d
, devnodes
) {
690 r
= clone_device_node(d
, temporary_mount
);
695 dev_setup(temporary_mount
, UID_INVALID
, GID_INVALID
);
697 /* Create the /dev directory if missing. It is more likely to be
698 * missing when the service is started with RootDirectory. This is
699 * consistent with mount units creating the mount points when missing.
701 (void) mkdir_p_label(mount_entry_path(m
), 0755);
703 /* Unmount everything in old /dev */
704 umount_recursive(mount_entry_path(m
), 0);
705 if (mount(dev
, mount_entry_path(m
), NULL
, MS_MOVE
, NULL
) < 0) {
711 rmdir(temporary_mount
);
723 umount(devhugepages
);
730 rmdir(temporary_mount
);
735 static int mount_bind_dev(const MountEntry
*m
) {
740 /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's
741 * /dev. This is only used when RootDirectory= is set. */
743 (void) mkdir_p_label(mount_entry_path(m
), 0755);
745 r
= path_is_mount_point(mount_entry_path(m
), NULL
, 0);
747 return log_debug_errno(r
, "Unable to determine whether /dev is already mounted: %m");
748 if (r
> 0) /* make this a NOP if /dev is already a mount point */
751 if (mount("/dev", mount_entry_path(m
), NULL
, MS_BIND
|MS_REC
, NULL
) < 0)
752 return log_debug_errno(errno
, "Failed to bind mount %s: %m", mount_entry_path(m
));
757 static int mount_sysfs(const MountEntry
*m
) {
762 (void) mkdir_p_label(mount_entry_path(m
), 0755);
764 r
= path_is_mount_point(mount_entry_path(m
), NULL
, 0);
766 return log_debug_errno(r
, "Unable to determine whether /sys is already mounted: %m");
767 if (r
> 0) /* make this a NOP if /sys is already a mount point */
770 /* Bind mount the host's version so that we get all child mounts of it, too. */
771 if (mount("/sys", mount_entry_path(m
), NULL
, MS_BIND
|MS_REC
, NULL
) < 0)
772 return log_debug_errno(errno
, "Failed to mount %s: %m", mount_entry_path(m
));
777 static int mount_procfs(const MountEntry
*m
) {
782 (void) mkdir_p_label(mount_entry_path(m
), 0755);
784 r
= path_is_mount_point(mount_entry_path(m
), NULL
, 0);
786 return log_debug_errno(r
, "Unable to determine whether /proc is already mounted: %m");
787 if (r
> 0) /* make this a NOP if /proc is already a mount point */
790 /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */
791 if (mount("proc", mount_entry_path(m
), "proc", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
, NULL
) < 0)
792 return log_debug_errno(errno
, "Failed to mount %s: %m", mount_entry_path(m
));
797 static int mount_tmpfs(const MountEntry
*m
) {
800 /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */
802 (void) mkdir_p_label(mount_entry_path(m
), 0755);
803 (void) umount_recursive(mount_entry_path(m
), 0);
805 if (mount("tmpfs", mount_entry_path(m
), "tmpfs", m
->flags
, mount_entry_options(m
)) < 0)
806 return log_debug_errno(errno
, "Failed to mount %s: %m", mount_entry_path(m
));
811 static int mount_entry_chase(
812 const char *root_directory
,
815 bool chase_nonexistent
,
823 /* Since mount() will always follow symlinks and we need to take the different root directory into account we
824 * chase the symlinks on our own first. This is called for the destination path, as well as the source path (if
825 * that applies). The result is stored in "location". */
827 r
= chase_symlinks(path
, root_directory
, CHASE_TRAIL_SLASH
| (chase_nonexistent
? CHASE_NONEXISTENT
: 0), &chased
);
828 if (r
== -ENOENT
&& m
->ignore
) {
829 log_debug_errno(r
, "Path %s does not exist, ignoring.", path
);
833 return log_debug_errno(r
, "Failed to follow symlinks on %s: %m", path
);
835 log_debug("Followed symlinks %s → %s.", path
, chased
);
843 static int apply_mount(
844 const char *root_directory
,
847 bool rbind
= true, make
= false;
853 r
= mount_entry_chase(root_directory
, m
, mount_entry_path(m
), !IN_SET(m
->mode
, INACCESSIBLE
, READONLY
, READWRITE
), &m
->path_malloc
);
857 log_debug("Applying namespace mount on %s", mount_entry_path(m
));
864 /* First, get rid of everything that is below if there
865 * is anything... Then, overmount it with an
866 * inaccessible path. */
867 (void) umount_recursive(mount_entry_path(m
), 0);
869 if (lstat(mount_entry_path(m
), &target
) < 0)
870 return log_debug_errno(errno
, "Failed to lstat() %s to determine what to mount over it: %m", mount_entry_path(m
));
872 what
= mode_to_inaccessible_node(target
.st_mode
);
874 log_debug("File type not supported for inaccessible mounts. Note that symlinks are not allowed");
882 r
= path_is_mount_point(mount_entry_path(m
), root_directory
, 0);
884 return log_debug_errno(r
, "Failed to determine whether %s is already a mount point: %m", mount_entry_path(m
));
885 if (r
> 0) /* Nothing to do here, it is already a mount. We just later toggle the MS_RDONLY bit for the mount point if needed. */
887 /* This isn't a mount point yet, let's make it one. */
888 what
= mount_entry_path(m
);
895 case BIND_MOUNT_RECURSIVE
:
896 /* Also chase the source mount */
898 r
= mount_entry_chase(root_directory
, m
, mount_entry_source(m
), false, &m
->source_malloc
);
902 what
= mount_entry_source(m
);
908 return mount_tmpfs(m
);
911 what
= mount_entry_source(m
);
916 return mount_private_dev(m
);
919 return mount_bind_dev(m
);
922 return mount_sysfs(m
);
925 return mount_procfs(m
);
928 assert_not_reached("Unknown mode");
933 if (mount(what
, mount_entry_path(m
), NULL
, MS_BIND
|(rbind
? MS_REC
: 0), NULL
) < 0) {
934 bool try_again
= false;
937 if (r
== -ENOENT
&& make
) {
940 /* Hmm, either the source or the destination are missing. Let's see if we can create the destination, then try again */
942 if (stat(what
, &st
) >= 0) {
944 (void) mkdir_parents(mount_entry_path(m
), 0755);
946 if (S_ISDIR(st
.st_mode
))
947 try_again
= mkdir(mount_entry_path(m
), 0755) >= 0;
949 try_again
= touch(mount_entry_path(m
)) >= 0;
954 if (mount(what
, mount_entry_path(m
), NULL
, MS_BIND
|(rbind
? MS_REC
: 0), NULL
) < 0)
961 return log_debug_errno(r
, "Failed to mount %s to %s: %m", what
, mount_entry_path(m
));
964 log_debug("Successfully mounted %s to %s", what
, mount_entry_path(m
));
968 static int make_read_only(const MountEntry
*m
, char **blacklist
, FILE *proc_self_mountinfo
) {
972 assert(proc_self_mountinfo
);
974 if (mount_entry_read_only(m
)) {
975 if (IN_SET(m
->mode
, EMPTY_DIR
, TMPFS
)) {
976 /* Make superblock readonly */
977 if (mount(NULL
, mount_entry_path(m
), NULL
, MS_REMOUNT
| MS_RDONLY
| m
->flags
, mount_entry_options(m
)) < 0)
980 r
= bind_remount_recursive_with_mountinfo(mount_entry_path(m
), true, blacklist
, proc_self_mountinfo
);
981 } else if (m
->mode
== PRIVATE_DEV
) {
982 /* Superblock can be readonly but the submounts can't */
983 if (mount(NULL
, mount_entry_path(m
), NULL
, MS_REMOUNT
|DEV_MOUNT_OPTIONS
|MS_RDONLY
, NULL
) < 0)
988 /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked read-only
989 * already stays this way. This improves compatibility with container managers, where we won't attempt to undo
990 * read-only mounts already applied. */
992 if (r
== -ENOENT
&& m
->ignore
)
998 static bool namespace_info_mount_apivfs(const char *root_directory
, const NamespaceInfo
*ns_info
) {
1002 * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=,
1003 * since to protect the API VFS mounts, they need to be around in the
1004 * first place... and RootDirectory= or RootImage= need to be set.
1007 /* root_directory should point to a mount point */
1008 return root_directory
&&
1009 (ns_info
->mount_apivfs
||
1010 ns_info
->protect_control_groups
||
1011 ns_info
->protect_kernel_tunables
);
1014 static unsigned namespace_calculate_mounts(
1015 const char* root_directory
,
1016 const NamespaceInfo
*ns_info
,
1017 char** read_write_paths
,
1018 char** read_only_paths
,
1019 char** inaccessible_paths
,
1020 char** empty_directories
,
1021 unsigned n_bind_mounts
,
1022 unsigned n_temporary_filesystems
,
1023 const char* tmp_dir
,
1024 const char* var_tmp_dir
,
1025 ProtectHome protect_home
,
1026 ProtectSystem protect_system
) {
1028 unsigned protect_home_cnt
;
1029 unsigned protect_system_cnt
=
1030 (protect_system
== PROTECT_SYSTEM_STRICT
?
1031 ELEMENTSOF(protect_system_strict_table
) :
1032 ((protect_system
== PROTECT_SYSTEM_FULL
) ?
1033 ELEMENTSOF(protect_system_full_table
) :
1034 ((protect_system
== PROTECT_SYSTEM_YES
) ?
1035 ELEMENTSOF(protect_system_yes_table
) : 0)));
1038 (protect_home
== PROTECT_HOME_YES
?
1039 ELEMENTSOF(protect_home_yes_table
) :
1040 ((protect_home
== PROTECT_HOME_READ_ONLY
) ?
1041 ELEMENTSOF(protect_home_read_only_table
) :
1042 ((protect_home
== PROTECT_HOME_TMPFS
) ?
1043 ELEMENTSOF(protect_home_tmpfs_table
) : 0)));
1045 return !!tmp_dir
+ !!var_tmp_dir
+
1046 strv_length(read_write_paths
) +
1047 strv_length(read_only_paths
) +
1048 strv_length(inaccessible_paths
) +
1049 strv_length(empty_directories
) +
1051 n_temporary_filesystems
+
1052 ns_info
->private_dev
+
1053 (ns_info
->protect_kernel_tunables
? ELEMENTSOF(protect_kernel_tunables_table
) : 0) +
1054 (ns_info
->protect_control_groups
? 1 : 0) +
1055 (ns_info
->protect_kernel_modules
? ELEMENTSOF(protect_kernel_modules_table
) : 0) +
1056 protect_home_cnt
+ protect_system_cnt
+
1057 (namespace_info_mount_apivfs(root_directory
, ns_info
) ? ELEMENTSOF(apivfs_table
) : 0);
1060 int setup_namespace(
1061 const char* root_directory
,
1062 const char* root_image
,
1063 const NamespaceInfo
*ns_info
,
1064 char** read_write_paths
,
1065 char** read_only_paths
,
1066 char** inaccessible_paths
,
1067 char** empty_directories
,
1068 const BindMount
*bind_mounts
,
1069 unsigned n_bind_mounts
,
1070 const TemporaryFileSystem
*temporary_filesystems
,
1071 unsigned n_temporary_filesystems
,
1072 const char* tmp_dir
,
1073 const char* var_tmp_dir
,
1074 ProtectHome protect_home
,
1075 ProtectSystem protect_system
,
1076 unsigned long mount_flags
,
1077 DissectImageFlags dissect_image_flags
) {
1079 _cleanup_(loop_device_unrefp
) LoopDevice
*loop_device
= NULL
;
1080 _cleanup_(decrypted_image_unrefp
) DecryptedImage
*decrypted_image
= NULL
;
1081 _cleanup_(dissected_image_unrefp
) DissectedImage
*dissected_image
= NULL
;
1082 _cleanup_free_
void *root_hash
= NULL
;
1083 MountEntry
*m
, *mounts
= NULL
;
1084 size_t root_hash_size
= 0;
1085 bool make_slave
= false;
1088 bool require_prefix
= false;
1093 if (mount_flags
== 0)
1094 mount_flags
= MS_SHARED
;
1097 dissect_image_flags
|= DISSECT_IMAGE_REQUIRE_ROOT
;
1099 if (protect_system
== PROTECT_SYSTEM_STRICT
&& strv_isempty(read_write_paths
))
1100 dissect_image_flags
|= DISSECT_IMAGE_READ_ONLY
;
1102 r
= loop_device_make_by_path(root_image
,
1103 dissect_image_flags
& DISSECT_IMAGE_READ_ONLY
? O_RDONLY
: O_RDWR
,
1108 r
= root_hash_load(root_image
, &root_hash
, &root_hash_size
);
1112 r
= dissect_image(loop_device
->fd
, root_hash
, root_hash_size
, dissect_image_flags
, &dissected_image
);
1116 r
= dissected_image_decrypt(dissected_image
, NULL
, root_hash
, root_hash_size
, dissect_image_flags
, &decrypted_image
);
1122 root
= root_directory
;
1123 else if (root_image
|| n_bind_mounts
> 0 || n_temporary_filesystems
> 0) {
1125 /* If we are booting from an image, create a mount point for the image, if it's still missing. We use
1126 * the same mount point for all images, which is safe, since they all live in their own namespaces
1127 * after all, and hence won't see each other. We also use such a root directory whenever there are bind
1128 * mounts configured, so that their source mounts are never obstructed by mounts we already applied
1129 * while we are applying them. */
1131 root
= "/run/systemd/unit-root";
1132 (void) mkdir_label(root
, 0700);
1133 require_prefix
= true;
1137 n_mounts
= namespace_calculate_mounts(
1145 n_temporary_filesystems
,
1146 tmp_dir
, var_tmp_dir
,
1147 protect_home
, protect_system
);
1149 /* Set mount slave mode */
1150 if (root
|| n_mounts
> 0)
1154 m
= mounts
= (MountEntry
*) alloca0(n_mounts
* sizeof(MountEntry
));
1155 r
= append_access_mounts(&m
, read_write_paths
, READWRITE
, require_prefix
);
1159 r
= append_access_mounts(&m
, read_only_paths
, READONLY
, require_prefix
);
1163 r
= append_access_mounts(&m
, inaccessible_paths
, INACCESSIBLE
, require_prefix
);
1167 r
= append_empty_dir_mounts(&m
, empty_directories
);
1171 r
= append_bind_mounts(&m
, bind_mounts
, n_bind_mounts
);
1175 r
= append_tmpfs_mounts(&m
, temporary_filesystems
, n_temporary_filesystems
);
1180 *(m
++) = (MountEntry
) {
1181 .path_const
= "/tmp",
1182 .mode
= PRIVATE_TMP
,
1183 .source_const
= tmp_dir
,
1188 *(m
++) = (MountEntry
) {
1189 .path_const
= "/var/tmp",
1190 .mode
= PRIVATE_TMP
,
1191 .source_const
= var_tmp_dir
,
1195 if (ns_info
->private_dev
) {
1196 *(m
++) = (MountEntry
) {
1197 .path_const
= "/dev",
1198 .mode
= PRIVATE_DEV
,
1202 if (ns_info
->protect_kernel_tunables
) {
1203 r
= append_static_mounts(&m
, protect_kernel_tunables_table
, ELEMENTSOF(protect_kernel_tunables_table
), ns_info
->ignore_protect_paths
);
1208 if (ns_info
->protect_kernel_modules
) {
1209 r
= append_static_mounts(&m
, protect_kernel_modules_table
, ELEMENTSOF(protect_kernel_modules_table
), ns_info
->ignore_protect_paths
);
1214 if (ns_info
->protect_control_groups
) {
1215 *(m
++) = (MountEntry
) {
1216 .path_const
= "/sys/fs/cgroup",
1221 r
= append_protect_home(&m
, protect_home
, ns_info
->ignore_protect_paths
);
1225 r
= append_protect_system(&m
, protect_system
, false);
1229 if (namespace_info_mount_apivfs(root
, ns_info
)) {
1230 r
= append_static_mounts(&m
, apivfs_table
, ELEMENTSOF(apivfs_table
), ns_info
->ignore_protect_paths
);
1235 assert(mounts
+ n_mounts
== m
);
1237 /* Prepend the root directory where that's necessary */
1238 r
= prefix_where_needed(mounts
, n_mounts
, root
);
1242 qsort(mounts
, n_mounts
, sizeof(MountEntry
), mount_path_compare
);
1244 drop_duplicates(mounts
, &n_mounts
);
1245 drop_outside_root(root
, mounts
, &n_mounts
);
1246 drop_inaccessible(mounts
, &n_mounts
);
1247 drop_nop(mounts
, &n_mounts
);
1250 if (unshare(CLONE_NEWNS
) < 0) {
1256 /* Remount / as SLAVE so that nothing now mounted in the namespace
1257 shows up in the parent */
1258 if (mount(NULL
, "/", NULL
, MS_SLAVE
|MS_REC
, NULL
) < 0) {
1265 /* A root image is specified, mount it to the right place */
1266 r
= dissected_image_mount(dissected_image
, root
, UID_INVALID
, dissect_image_flags
);
1270 if (decrypted_image
) {
1271 r
= decrypted_image_relinquish(decrypted_image
);
1276 loop_device_relinquish(loop_device
);
1278 } else if (root_directory
) {
1280 /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */
1281 r
= path_is_mount_point(root
, NULL
, AT_SYMLINK_FOLLOW
);
1285 if (mount(root
, root
, NULL
, MS_BIND
|MS_REC
, NULL
) < 0) {
1293 /* Let's mount the main root directory to the root directory to use */
1294 if (mount("/", root
, NULL
, MS_BIND
|MS_REC
, NULL
) < 0) {
1300 /* Try to set up the new root directory before mounting anything else there. */
1301 if (root_image
|| root_directory
)
1302 (void) base_filesystem_create(root
, UID_INVALID
, GID_INVALID
);
1305 _cleanup_fclose_
FILE *proc_self_mountinfo
= NULL
;
1309 /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of /proc.
1310 * For example, this is the case with the option: 'InaccessiblePaths=/proc' */
1311 proc_self_mountinfo
= fopen("/proc/self/mountinfo", "re");
1312 if (!proc_self_mountinfo
) {
1317 /* First round, add in all special mounts we need */
1318 for (m
= mounts
; m
< mounts
+ n_mounts
; ++m
) {
1319 r
= apply_mount(root
, m
);
1324 /* Create a blacklist we can pass to bind_mount_recursive() */
1325 blacklist
= newa(char*, n_mounts
+1);
1326 for (j
= 0; j
< n_mounts
; j
++)
1327 blacklist
[j
] = (char*) mount_entry_path(mounts
+j
);
1328 blacklist
[j
] = NULL
;
1330 /* Second round, flip the ro bits if necessary. */
1331 for (m
= mounts
; m
< mounts
+ n_mounts
; ++m
) {
1332 r
= make_read_only(m
, blacklist
, proc_self_mountinfo
);
1339 /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */
1340 r
= mount_move_root(root
);
1345 /* Remount / as the desired mode. Note that this will not
1346 * reestablish propagation from our side to the host, since
1347 * what's disconnected is disconnected. */
1348 if (mount(NULL
, "/", NULL
, mount_flags
| MS_REC
, NULL
) < 0) {
1356 for (m
= mounts
; m
< mounts
+ n_mounts
; m
++)
1357 mount_entry_done(m
);
1362 void bind_mount_free_many(BindMount
*b
, unsigned n
) {
1365 assert(b
|| n
== 0);
1367 for (i
= 0; i
< n
; i
++) {
1369 free(b
[i
].destination
);
1375 int bind_mount_add(BindMount
**b
, unsigned *n
, const BindMount
*item
) {
1376 _cleanup_free_
char *s
= NULL
, *d
= NULL
;
1383 s
= strdup(item
->source
);
1387 d
= strdup(item
->destination
);
1391 c
= reallocarray(*b
, *n
+ 1, sizeof(BindMount
));
1397 c
[(*n
) ++] = (BindMount
) {
1398 .source
= TAKE_PTR(s
),
1399 .destination
= TAKE_PTR(d
),
1400 .read_only
= item
->read_only
,
1401 .recursive
= item
->recursive
,
1402 .ignore_enoent
= item
->ignore_enoent
,
1408 void temporary_filesystem_free_many(TemporaryFileSystem
*t
, unsigned n
) {
1411 assert(t
|| n
== 0);
1413 for (i
= 0; i
< n
; i
++) {
1421 int temporary_filesystem_add(
1422 TemporaryFileSystem
**t
,
1425 const char *options
) {
1427 _cleanup_free_
char *p
= NULL
, *o
= NULL
;
1428 TemporaryFileSystem
*c
;
1438 if (!isempty(options
)) {
1439 o
= strdup(options
);
1444 c
= reallocarray(*t
, *n
+ 1, sizeof(TemporaryFileSystem
));
1450 c
[(*n
) ++] = (TemporaryFileSystem
) {
1451 .path
= TAKE_PTR(p
),
1452 .options
= TAKE_PTR(o
),
1458 static int setup_one_tmp_dir(const char *id
, const char *prefix
, char **path
) {
1459 _cleanup_free_
char *x
= NULL
;
1460 char bid
[SD_ID128_STRING_MAX
];
1468 /* We include the boot id in the directory so that after a
1469 * reboot we can easily identify obsolete directories. */
1471 r
= sd_id128_get_boot(&boot_id
);
1475 x
= strjoin(prefix
, "/systemd-private-", sd_id128_to_string(boot_id
, bid
), "-", id
, "-XXXXXX");
1479 RUN_WITH_UMASK(0077)
1483 RUN_WITH_UMASK(0000) {
1486 y
= strjoina(x
, "/tmp");
1488 if (mkdir(y
, 0777 | S_ISVTX
) < 0)
1492 *path
= TAKE_PTR(x
);
1497 int setup_tmp_dirs(const char *id
, char **tmp_dir
, char **var_tmp_dir
) {
1503 assert(var_tmp_dir
);
1505 r
= setup_one_tmp_dir(id
, "/tmp", &a
);
1509 r
= setup_one_tmp_dir(id
, "/var/tmp", &b
);
1513 t
= strjoina(a
, "/tmp");
1527 int setup_netns(int netns_storage_socket
[2]) {
1528 _cleanup_close_
int netns
= -1;
1531 assert(netns_storage_socket
);
1532 assert(netns_storage_socket
[0] >= 0);
1533 assert(netns_storage_socket
[1] >= 0);
1535 /* We use the passed socketpair as a storage buffer for our
1536 * namespace reference fd. Whatever process runs this first
1537 * shall create a new namespace, all others should just join
1538 * it. To serialize that we use a file lock on the socket
1541 * It's a bit crazy, but hey, works great! */
1543 if (lockf(netns_storage_socket
[0], F_LOCK
, 0) < 0)
1546 netns
= receive_one_fd(netns_storage_socket
[0], MSG_DONTWAIT
);
1547 if (netns
== -EAGAIN
) {
1548 /* Nothing stored yet, so let's create a new namespace */
1550 if (unshare(CLONE_NEWNET
) < 0) {
1557 netns
= open("/proc/self/ns/net", O_RDONLY
|O_CLOEXEC
|O_NOCTTY
);
1565 } else if (netns
< 0) {
1570 /* Yay, found something, so let's join the namespace */
1571 if (setns(netns
, CLONE_NEWNET
) < 0) {
1579 q
= send_one_fd(netns_storage_socket
[1], netns
, MSG_DONTWAIT
);
1586 (void) lockf(netns_storage_socket
[0], F_ULOCK
, 0);
1590 bool ns_type_supported(NamespaceType type
) {
1591 const char *t
, *ns_proc
;
1593 t
= namespace_type_to_string(type
);
1594 if (!t
) /* Don't know how to translate this? Then it's not supported */
1597 ns_proc
= strjoina("/proc/self/ns/", t
);
1598 return access(ns_proc
, F_OK
) == 0;
1601 static const char *const protect_home_table
[_PROTECT_HOME_MAX
] = {
1602 [PROTECT_HOME_NO
] = "no",
1603 [PROTECT_HOME_YES
] = "yes",
1604 [PROTECT_HOME_READ_ONLY
] = "read-only",
1605 [PROTECT_HOME_TMPFS
] = "tmpfs",
1608 DEFINE_STRING_TABLE_LOOKUP(protect_home
, ProtectHome
);
1610 ProtectHome
parse_protect_home_or_bool(const char *s
) {
1613 r
= parse_boolean(s
);
1615 return PROTECT_HOME_YES
;
1617 return PROTECT_HOME_NO
;
1619 return protect_home_from_string(s
);
1622 static const char *const protect_system_table
[_PROTECT_SYSTEM_MAX
] = {
1623 [PROTECT_SYSTEM_NO
] = "no",
1624 [PROTECT_SYSTEM_YES
] = "yes",
1625 [PROTECT_SYSTEM_FULL
] = "full",
1626 [PROTECT_SYSTEM_STRICT
] = "strict",
1629 DEFINE_STRING_TABLE_LOOKUP(protect_system
, ProtectSystem
);
1631 ProtectSystem
parse_protect_system_or_bool(const char *s
) {
1634 r
= parse_boolean(s
);
1636 return PROTECT_SYSTEM_YES
;
1638 return PROTECT_SYSTEM_NO
;
1640 return protect_system_from_string(s
);
1643 static const char* const namespace_type_table
[] = {
1644 [NAMESPACE_MOUNT
] = "mnt",
1645 [NAMESPACE_CGROUP
] = "cgroup",
1646 [NAMESPACE_UTS
] = "uts",
1647 [NAMESPACE_IPC
] = "ipc",
1648 [NAMESPACE_USER
] = "user",
1649 [NAMESPACE_PID
] = "pid",
1650 [NAMESPACE_NET
] = "net",
1653 DEFINE_STRING_TABLE_LOOKUP(namespace_type
, NamespaceType
);