1 /* SPDX-License-Identifier: LGPL-2.1+ */
12 #include "alloc-util.h"
13 #include "base-filesystem.h"
14 #include "dev-setup.h"
18 #include "loop-util.h"
19 #include "loopback-setup.h"
22 #include "mount-util.h"
23 #include "namespace.h"
24 #include "path-util.h"
25 #include "selinux-util.h"
26 #include "socket-util.h"
27 #include "stat-util.h"
28 #include "string-table.h"
29 #include "string-util.h"
31 #include "umask-util.h"
32 #include "user-util.h"
35 #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC)
37 typedef enum MountMode
{
38 /* This is ordered by priority! */
53 typedef struct MountEntry
{
54 const char *path_const
; /* Memory allocated on stack or static */
56 bool ignore
:1; /* Ignore if path does not exist? */
57 bool has_prefix
:1; /* Already is prefixed by the root dir? */
58 bool read_only
:1; /* Shall this mount point be read-only? */
59 bool applied
:1; /* Already applied */
60 char *path_malloc
; /* Use this instead of 'path_const' if we had to allocate memory */
61 const char *source_const
; /* The source path, for bind mounts */
63 const char *options_const
;/* Mount options for tmpfs */
65 unsigned long flags
; /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */
69 /* If MountAPIVFS= is used, let's mount /sys and /proc into the it, but only as a fallback if the user hasn't mounted
70 * something there already. These mounts are hence overridden by any other explicitly configured mounts. */
71 static const MountEntry apivfs_table
[] = {
72 { "/proc", PROCFS
, false },
73 { "/dev", BIND_DEV
, false },
74 { "/sys", SYSFS
, false },
77 /* ProtectKernelTunables= option and the related filesystem APIs */
78 static const MountEntry protect_kernel_tunables_table
[] = {
79 { "/proc/acpi", READONLY
, true },
80 { "/proc/apm", READONLY
, true }, /* Obsolete API, there's no point in permitting access to this, ever */
81 { "/proc/asound", READONLY
, true },
82 { "/proc/bus", READONLY
, true },
83 { "/proc/fs", READONLY
, true },
84 { "/proc/irq", READONLY
, true },
85 { "/proc/kallsyms", INACCESSIBLE
, true },
86 { "/proc/kcore", INACCESSIBLE
, true },
87 { "/proc/latency_stats", READONLY
, true },
88 { "/proc/mtrr", READONLY
, true },
89 { "/proc/scsi", READONLY
, true },
90 { "/proc/sys", READONLY
, false },
91 { "/proc/sysrq-trigger", READONLY
, true },
92 { "/proc/timer_stats", READONLY
, true },
93 { "/sys", READONLY
, false },
94 { "/sys/fs/bpf", READONLY
, true },
95 { "/sys/fs/cgroup", READWRITE
, false }, /* READONLY is set by ProtectControlGroups= option */
96 { "/sys/fs/selinux", READWRITE
, true },
97 { "/sys/kernel/debug", READONLY
, true },
98 { "/sys/kernel/tracing", READONLY
, true },
101 /* ProtectKernelModules= option */
102 static const MountEntry protect_kernel_modules_table
[] = {
104 { "/lib/modules", INACCESSIBLE
, true },
106 { "/usr/lib/modules", INACCESSIBLE
, true },
110 * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of
111 * system should be protected by ProtectSystem=
113 static const MountEntry protect_home_read_only_table
[] = {
114 { "/home", READONLY
, true },
115 { "/run/user", READONLY
, true },
116 { "/root", READONLY
, true },
119 /* ProtectHome=tmpfs table */
120 static const MountEntry protect_home_tmpfs_table
[] = {
121 { "/home", TMPFS
, true, .read_only
= true, .options_const
= "mode=0755", .flags
= MS_NODEV
|MS_STRICTATIME
},
122 { "/run/user", TMPFS
, true, .read_only
= true, .options_const
= "mode=0755", .flags
= MS_NODEV
|MS_STRICTATIME
},
123 { "/root", TMPFS
, true, .read_only
= true, .options_const
= "mode=0700", .flags
= MS_NODEV
|MS_STRICTATIME
},
126 /* ProtectHome=yes table */
127 static const MountEntry protect_home_yes_table
[] = {
128 { "/home", INACCESSIBLE
, true },
129 { "/run/user", INACCESSIBLE
, true },
130 { "/root", INACCESSIBLE
, true },
133 /* ProtectSystem=yes table */
134 static const MountEntry protect_system_yes_table
[] = {
135 { "/usr", READONLY
, false },
136 { "/boot", READONLY
, true },
137 { "/efi", READONLY
, true },
139 { "/lib", READONLY
, true },
140 { "/lib64", READONLY
, true },
141 { "/bin", READONLY
, true },
143 { "/sbin", READONLY
, true },
148 /* ProtectSystem=full includes ProtectSystem=yes */
149 static const MountEntry protect_system_full_table
[] = {
150 { "/usr", READONLY
, false },
151 { "/boot", READONLY
, true },
152 { "/efi", READONLY
, true },
153 { "/etc", READONLY
, false },
155 { "/lib", READONLY
, true },
156 { "/lib64", READONLY
, true },
157 { "/bin", READONLY
, true },
159 { "/sbin", READONLY
, true },
165 * ProtectSystem=strict table. In this strict mode, we mount everything
166 * read-only, except for /proc, /dev, /sys which are the kernel API VFS,
167 * which are left writable, but PrivateDevices= + ProtectKernelTunables=
168 * protect those, and these options should be fully orthogonal.
169 * (And of course /home and friends are also left writable, as ProtectHome=
170 * shall manage those, orthogonally).
172 static const MountEntry protect_system_strict_table
[] = {
173 { "/", READONLY
, false },
174 { "/proc", READWRITE
, false }, /* ProtectKernelTunables= */
175 { "/sys", READWRITE
, false }, /* ProtectKernelTunables= */
176 { "/dev", READWRITE
, false }, /* PrivateDevices= */
177 { "/home", READWRITE
, true }, /* ProtectHome= */
178 { "/run/user", READWRITE
, true }, /* ProtectHome= */
179 { "/root", READWRITE
, true }, /* ProtectHome= */
182 static const char *mount_entry_path(const MountEntry
*p
) {
185 /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that,
186 * otherwise the stack/static ->path field is returned. */
188 return p
->path_malloc
?: p
->path_const
;
191 static bool mount_entry_read_only(const MountEntry
*p
) {
194 return p
->read_only
|| IN_SET(p
->mode
, READONLY
, INACCESSIBLE
);
197 static const char *mount_entry_source(const MountEntry
*p
) {
200 return p
->source_malloc
?: p
->source_const
;
203 static const char *mount_entry_options(const MountEntry
*p
) {
206 return p
->options_malloc
?: p
->options_const
;
209 static void mount_entry_done(MountEntry
*p
) {
212 p
->path_malloc
= mfree(p
->path_malloc
);
213 p
->source_malloc
= mfree(p
->source_malloc
);
214 p
->options_malloc
= mfree(p
->options_malloc
);
217 static int append_access_mounts(MountEntry
**p
, char **strv
, MountMode mode
, bool forcibly_require_prefix
) {
222 /* Adds a list of user-supplied READWRITE/READONLY/INACCESSIBLE entries */
224 STRV_FOREACH(i
, strv
) {
225 bool ignore
= false, needs_prefix
= false;
228 /* Look for any prefixes */
229 if (startswith(e
, "-")) {
233 if (startswith(e
, "+")) {
238 if (!path_is_absolute(e
)) {
239 log_debug("Path is not absolute: %s", e
);
243 *((*p
)++) = (MountEntry
) {
247 .has_prefix
= !needs_prefix
&& !forcibly_require_prefix
,
254 static int append_empty_dir_mounts(MountEntry
**p
, char **strv
) {
259 /* Adds tmpfs mounts to provide readable but empty directories. This is primarily used to implement the
260 * "/private/" boundary directories for DynamicUser=1. */
262 STRV_FOREACH(i
, strv
) {
264 *((*p
)++) = (MountEntry
) {
270 .options_const
= "mode=755",
271 .flags
= MS_NOSUID
|MS_NOEXEC
|MS_NODEV
|MS_STRICTATIME
,
278 static int append_bind_mounts(MountEntry
**p
, const BindMount
*binds
, size_t n
) {
283 for (i
= 0; i
< n
; i
++) {
284 const BindMount
*b
= binds
+ i
;
286 *((*p
)++) = (MountEntry
) {
287 .path_const
= b
->destination
,
288 .mode
= b
->recursive
? BIND_MOUNT_RECURSIVE
: BIND_MOUNT
,
289 .read_only
= b
->read_only
,
290 .source_const
= b
->source
,
291 .ignore
= b
->ignore_enoent
,
298 static int append_tmpfs_mounts(MountEntry
**p
, const TemporaryFileSystem
*tmpfs
, size_t n
) {
304 for (i
= 0; i
< n
; i
++) {
305 const TemporaryFileSystem
*t
= tmpfs
+ i
;
306 _cleanup_free_
char *o
= NULL
, *str
= NULL
;
307 unsigned long flags
= MS_NODEV
|MS_STRICTATIME
;
310 if (!path_is_absolute(t
->path
)) {
311 log_debug("Path is not absolute: %s", t
->path
);
315 if (!isempty(t
->options
)) {
316 str
= strjoin("mode=0755,", t
->options
);
320 r
= mount_option_mangle(str
, MS_NODEV
|MS_STRICTATIME
, &flags
, &o
);
322 return log_debug_errno(r
, "Failed to parse mount option '%s': %m", str
);
324 ro
= flags
& MS_RDONLY
;
329 *((*p
)++) = (MountEntry
) {
330 .path_const
= t
->path
,
343 static int append_static_mounts(MountEntry
**p
, const MountEntry
*mounts
, size_t n
, bool ignore_protect
) {
349 /* Adds a list of static pre-defined entries */
351 for (i
= 0; i
< n
; i
++)
352 *((*p
)++) = (MountEntry
) {
353 .path_const
= mount_entry_path(mounts
+i
),
354 .mode
= mounts
[i
].mode
,
355 .ignore
= mounts
[i
].ignore
|| ignore_protect
,
361 static int append_protect_home(MountEntry
**p
, ProtectHome protect_home
, bool ignore_protect
) {
364 switch (protect_home
) {
366 case PROTECT_HOME_NO
:
369 case PROTECT_HOME_READ_ONLY
:
370 return append_static_mounts(p
, protect_home_read_only_table
, ELEMENTSOF(protect_home_read_only_table
), ignore_protect
);
372 case PROTECT_HOME_TMPFS
:
373 return append_static_mounts(p
, protect_home_tmpfs_table
, ELEMENTSOF(protect_home_tmpfs_table
), ignore_protect
);
375 case PROTECT_HOME_YES
:
376 return append_static_mounts(p
, protect_home_yes_table
, ELEMENTSOF(protect_home_yes_table
), ignore_protect
);
379 assert_not_reached("Unexpected ProtectHome= value");
383 static int append_protect_system(MountEntry
**p
, ProtectSystem protect_system
, bool ignore_protect
) {
386 switch (protect_system
) {
388 case PROTECT_SYSTEM_NO
:
391 case PROTECT_SYSTEM_STRICT
:
392 return append_static_mounts(p
, protect_system_strict_table
, ELEMENTSOF(protect_system_strict_table
), ignore_protect
);
394 case PROTECT_SYSTEM_YES
:
395 return append_static_mounts(p
, protect_system_yes_table
, ELEMENTSOF(protect_system_yes_table
), ignore_protect
);
397 case PROTECT_SYSTEM_FULL
:
398 return append_static_mounts(p
, protect_system_full_table
, ELEMENTSOF(protect_system_full_table
), ignore_protect
);
401 assert_not_reached("Unexpected ProtectSystem= value");
405 static int mount_path_compare(const void *a
, const void *b
) {
406 const MountEntry
*p
= a
, *q
= b
;
409 /* If the paths are not equal, then order prefixes first */
410 d
= path_compare(mount_entry_path(p
), mount_entry_path(q
));
414 /* If the paths are equal, check the mode */
415 if (p
->mode
< q
->mode
)
417 if (p
->mode
> q
->mode
)
423 static int prefix_where_needed(MountEntry
*m
, size_t n
, const char *root_directory
) {
426 /* Prefixes all paths in the bind mount table with the root directory if it is specified and the entry needs
432 for (i
= 0; i
< n
; i
++) {
438 s
= prefix_root(root_directory
, mount_entry_path(m
+i
));
442 free_and_replace(m
[i
].path_malloc
, s
);
443 m
[i
].has_prefix
= true;
449 static void drop_duplicates(MountEntry
*m
, size_t *n
) {
450 MountEntry
*f
, *t
, *previous
;
455 /* Drops duplicate entries. Expects that the array is properly ordered already. */
457 for (f
= m
, t
= m
, previous
= NULL
; f
< m
+ *n
; f
++) {
459 /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare()
460 * above. Note that we only drop duplicates that haven't been mounted yet. */
462 path_equal(mount_entry_path(f
), mount_entry_path(previous
)) &&
463 !f
->applied
&& !previous
->applied
) {
464 log_debug("%s is duplicate.", mount_entry_path(f
));
465 previous
->read_only
= previous
->read_only
|| mount_entry_read_only(f
); /* Propagate the read-only flag to the remaining entry */
478 static void drop_inaccessible(MountEntry
*m
, size_t *n
) {
480 const char *clear
= NULL
;
485 /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly
486 * ordered already. */
488 for (f
= m
, t
= m
; f
< m
+ *n
; f
++) {
490 /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop
491 * it, as inaccessible paths really should drop the entire subtree. */
492 if (clear
&& path_startswith(mount_entry_path(f
), clear
)) {
493 log_debug("%s is masked by %s.", mount_entry_path(f
), clear
);
498 clear
= f
->mode
== INACCESSIBLE
? mount_entry_path(f
) : NULL
;
507 static void drop_nop(MountEntry
*m
, size_t *n
) {
513 /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the
514 * list is ordered by prefixes. */
516 for (f
= m
, t
= m
; f
< m
+ *n
; f
++) {
518 /* Only suppress such subtrees for READONLY and READWRITE entries */
519 if (IN_SET(f
->mode
, READONLY
, READWRITE
)) {
523 /* Now let's find the first parent of the entry we are looking at. */
524 for (p
= t
-1; p
>= m
; p
--) {
525 if (path_startswith(mount_entry_path(f
), mount_entry_path(p
))) {
531 /* We found it, let's see if it's the same mode, if so, we can drop this entry */
532 if (found
&& p
->mode
== f
->mode
) {
533 log_debug("%s is redundant by %s", mount_entry_path(f
), mount_entry_path(p
));
546 static void drop_outside_root(const char *root_directory
, MountEntry
*m
, size_t *n
) {
556 /* Drops all mounts that are outside of the root directory. */
558 for (f
= m
, t
= m
; f
< m
+ *n
; f
++) {
560 if (!path_startswith(mount_entry_path(f
), root_directory
)) {
561 log_debug("%s is outside of root directory.", mount_entry_path(f
));
573 static int clone_device_node(const char *d
, const char *temporary_mount
, bool *make_devnode
) {
578 if (stat(d
, &st
) < 0) {
584 if (!S_ISBLK(st
.st_mode
) &&
585 !S_ISCHR(st
.st_mode
))
591 dn
= strjoina(temporary_mount
, d
);
594 mac_selinux_create_file_prepare(d
, st
.st_mode
);
595 r
= mknod(dn
, st
.st_mode
, st
.st_rdev
);
596 mac_selinux_create_file_clear();
601 return log_debug_errno(errno
, "mknod failed for %s: %m", d
);
603 *make_devnode
= false;
606 /* We're about to fallback to bind-mounting the device
607 * node. So create a dummy bind-mount target. */
608 mac_selinux_create_file_prepare(d
, 0);
609 r
= mknod(dn
, S_IFREG
, 0);
610 mac_selinux_create_file_clear();
612 if (r
< 0 && errno
!= EEXIST
)
613 return log_debug_errno(errno
, "mknod fallback failed for %s: %m", d
);
615 /* Fallback to bind-mounting:
616 * The assumption here is that all used device nodes carry standard
617 * properties. Specifically, the devices nodes we bind-mount should
618 * either be owned by root:root or root:tty (e.g. /dev/tty, /dev/ptmx)
619 * and should not carry ACLs. */
620 if (mount(d
, dn
, NULL
, MS_BIND
, NULL
) < 0)
621 return log_debug_errno(errno
, "mount failed for %s: %m", d
);
626 static int mount_private_dev(MountEntry
*m
) {
627 static const char devnodes
[] =
635 char temporary_mount
[] = "/tmp/namespace-dev-XXXXXX";
636 const char *d
, *dev
= NULL
, *devpts
= NULL
, *devshm
= NULL
, *devhugepages
= NULL
, *devmqueue
= NULL
, *devlog
= NULL
, *devptmx
= NULL
;
637 bool can_mknod
= true;
638 _cleanup_umask_ mode_t u
;
645 if (!mkdtemp(temporary_mount
))
648 dev
= strjoina(temporary_mount
, "/dev");
649 (void) mkdir(dev
, 0755);
650 if (mount("tmpfs", dev
, "tmpfs", DEV_MOUNT_OPTIONS
, "mode=755") < 0) {
655 devpts
= strjoina(temporary_mount
, "/dev/pts");
656 (void) mkdir(devpts
, 0755);
657 if (mount("/dev/pts", devpts
, NULL
, MS_BIND
, NULL
) < 0) {
662 /* /dev/ptmx can either be a device node or a symlink to /dev/pts/ptmx
663 * when /dev/ptmx a device node, /dev/pts/ptmx has 000 permissions making it inaccessible
664 * thus, in that case make a clone
666 * in nspawn and other containers it will be a symlink, in that case make it a symlink
668 r
= is_symlink("/dev/ptmx");
672 devptmx
= strjoina(temporary_mount
, "/dev/ptmx");
673 if (symlink("pts/ptmx", devptmx
) < 0) {
678 r
= clone_device_node("/dev/ptmx", temporary_mount
, &can_mknod
);
683 devshm
= strjoina(temporary_mount
, "/dev/shm");
684 (void) mkdir(devshm
, 0755);
685 r
= mount("/dev/shm", devshm
, NULL
, MS_BIND
, NULL
);
691 devmqueue
= strjoina(temporary_mount
, "/dev/mqueue");
692 (void) mkdir(devmqueue
, 0755);
693 (void) mount("/dev/mqueue", devmqueue
, NULL
, MS_BIND
, NULL
);
695 devhugepages
= strjoina(temporary_mount
, "/dev/hugepages");
696 (void) mkdir(devhugepages
, 0755);
697 (void) mount("/dev/hugepages", devhugepages
, NULL
, MS_BIND
, NULL
);
699 devlog
= strjoina(temporary_mount
, "/dev/log");
700 (void) symlink("/run/systemd/journal/dev-log", devlog
);
702 NULSTR_FOREACH(d
, devnodes
) {
703 r
= clone_device_node(d
, temporary_mount
, &can_mknod
);
704 /* ENXIO means the the *source* is not a device file, skip creation in that case */
705 if (r
< 0 && r
!= -ENXIO
)
709 dev_setup(temporary_mount
, UID_INVALID
, GID_INVALID
);
711 /* Create the /dev directory if missing. It is more likely to be
712 * missing when the service is started with RootDirectory. This is
713 * consistent with mount units creating the mount points when missing.
715 (void) mkdir_p_label(mount_entry_path(m
), 0755);
717 /* Unmount everything in old /dev */
718 umount_recursive(mount_entry_path(m
), 0);
719 if (mount(dev
, mount_entry_path(m
), NULL
, MS_MOVE
, NULL
) < 0) {
725 rmdir(temporary_mount
);
737 umount(devhugepages
);
744 rmdir(temporary_mount
);
749 static int mount_bind_dev(const MountEntry
*m
) {
754 /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's
755 * /dev. This is only used when RootDirectory= is set. */
757 (void) mkdir_p_label(mount_entry_path(m
), 0755);
759 r
= path_is_mount_point(mount_entry_path(m
), NULL
, 0);
761 return log_debug_errno(r
, "Unable to determine whether /dev is already mounted: %m");
762 if (r
> 0) /* make this a NOP if /dev is already a mount point */
765 if (mount("/dev", mount_entry_path(m
), NULL
, MS_BIND
|MS_REC
, NULL
) < 0)
766 return log_debug_errno(errno
, "Failed to bind mount %s: %m", mount_entry_path(m
));
771 static int mount_sysfs(const MountEntry
*m
) {
776 (void) mkdir_p_label(mount_entry_path(m
), 0755);
778 r
= path_is_mount_point(mount_entry_path(m
), NULL
, 0);
780 return log_debug_errno(r
, "Unable to determine whether /sys is already mounted: %m");
781 if (r
> 0) /* make this a NOP if /sys is already a mount point */
784 /* Bind mount the host's version so that we get all child mounts of it, too. */
785 if (mount("/sys", mount_entry_path(m
), NULL
, MS_BIND
|MS_REC
, NULL
) < 0)
786 return log_debug_errno(errno
, "Failed to mount %s: %m", mount_entry_path(m
));
791 static int mount_procfs(const MountEntry
*m
) {
796 (void) mkdir_p_label(mount_entry_path(m
), 0755);
798 r
= path_is_mount_point(mount_entry_path(m
), NULL
, 0);
800 return log_debug_errno(r
, "Unable to determine whether /proc is already mounted: %m");
801 if (r
> 0) /* make this a NOP if /proc is already a mount point */
804 /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */
805 if (mount("proc", mount_entry_path(m
), "proc", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
, NULL
) < 0)
806 return log_debug_errno(errno
, "Failed to mount %s: %m", mount_entry_path(m
));
811 static int mount_tmpfs(const MountEntry
*m
) {
814 /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */
816 (void) mkdir_p_label(mount_entry_path(m
), 0755);
817 (void) umount_recursive(mount_entry_path(m
), 0);
819 if (mount("tmpfs", mount_entry_path(m
), "tmpfs", m
->flags
, mount_entry_options(m
)) < 0)
820 return log_debug_errno(errno
, "Failed to mount %s: %m", mount_entry_path(m
));
825 static int follow_symlink(
826 const char *root_directory
,
829 _cleanup_free_
char *target
= NULL
;
832 /* Let's chase symlinks, but only one step at a time. That's because depending where the symlink points we
833 * might need to change the order in which we mount stuff. Hence: let's normalize piecemeal, and do one step at
834 * a time by specifying CHASE_STEP. This function returns 0 if we resolved one step, and > 0 if we reached the
835 * end and already have a fully normalized name. */
837 r
= chase_symlinks(mount_entry_path(m
), root_directory
, CHASE_STEP
|CHASE_NONEXISTENT
, &target
);
839 return log_debug_errno(r
, "Failed to chase symlinks '%s': %m", mount_entry_path(m
));
840 if (r
> 0) /* Reached the end, nothing more to resolve */
843 if (m
->n_followed
>= CHASE_SYMLINKS_MAX
) { /* put a boundary on things */
844 log_debug("Symlink loop on '%s'.", mount_entry_path(m
));
848 log_debug("Followed mount entry path symlink %s → %s.", mount_entry_path(m
), target
);
850 free_and_replace(m
->path_malloc
, target
);
851 m
->has_prefix
= true;
858 static int apply_mount(
859 const char *root_directory
,
862 bool rbind
= true, make
= false;
868 log_debug("Applying namespace mount on %s", mount_entry_path(m
));
875 /* First, get rid of everything that is below if there
876 * is anything... Then, overmount it with an
877 * inaccessible path. */
878 (void) umount_recursive(mount_entry_path(m
), 0);
880 if (lstat(mount_entry_path(m
), &target
) < 0) {
881 if (errno
== ENOENT
&& m
->ignore
)
884 return log_debug_errno(errno
, "Failed to lstat() %s to determine what to mount over it: %m", mount_entry_path(m
));
887 what
= mode_to_inaccessible_node(target
.st_mode
);
889 log_debug("File type not supported for inaccessible mounts. Note that symlinks are not allowed");
897 r
= path_is_mount_point(mount_entry_path(m
), root_directory
, 0);
898 if (r
== -ENOENT
&& m
->ignore
)
901 return log_debug_errno(r
, "Failed to determine whether %s is already a mount point: %m", mount_entry_path(m
));
902 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. */
904 /* This isn't a mount point yet, let's make it one. */
905 what
= mount_entry_path(m
);
912 case BIND_MOUNT_RECURSIVE
: {
913 _cleanup_free_
char *chased
= NULL
;
915 /* Since mount() will always follow symlinks we chase the symlinks on our own first. Note that bind
916 * mount source paths are always relative to the host root, hence we pass NULL as root directory to
917 * chase_symlinks() here. */
919 r
= chase_symlinks(mount_entry_source(m
), NULL
, CHASE_TRAIL_SLASH
, &chased
);
920 if (r
== -ENOENT
&& m
->ignore
) {
921 log_debug_errno(r
, "Path %s does not exist, ignoring.", mount_entry_source(m
));
925 return log_debug_errno(r
, "Failed to follow symlinks on %s: %m", mount_entry_source(m
));
927 log_debug("Followed source symlinks %s → %s.", mount_entry_source(m
), chased
);
929 free_and_replace(m
->source_malloc
, chased
);
931 what
= mount_entry_source(m
);
938 return mount_tmpfs(m
);
941 what
= mount_entry_source(m
);
946 return mount_private_dev(m
);
949 return mount_bind_dev(m
);
952 return mount_sysfs(m
);
955 return mount_procfs(m
);
958 assert_not_reached("Unknown mode");
963 if (mount(what
, mount_entry_path(m
), NULL
, MS_BIND
|(rbind
? MS_REC
: 0), NULL
) < 0) {
964 bool try_again
= false;
967 if (r
== -ENOENT
&& make
) {
970 /* Hmm, either the source or the destination are missing. Let's see if we can create the destination, then try again */
972 if (stat(what
, &st
) < 0)
973 log_debug_errno(errno
, "Mount point source '%s' is not accessible: %m", what
);
977 (void) mkdir_parents(mount_entry_path(m
), 0755);
979 if (S_ISDIR(st
.st_mode
))
980 q
= mkdir(mount_entry_path(m
), 0755) < 0 ? -errno
: 0;
982 q
= touch(mount_entry_path(m
));
985 log_debug_errno(q
, "Failed to create destination mount point node '%s': %m", mount_entry_path(m
));
992 if (mount(what
, mount_entry_path(m
), NULL
, MS_BIND
|(rbind
? MS_REC
: 0), NULL
) < 0)
999 return log_debug_errno(r
, "Failed to mount %s to %s: %m", what
, mount_entry_path(m
));
1002 log_debug("Successfully mounted %s to %s", what
, mount_entry_path(m
));
1006 static int make_read_only(const MountEntry
*m
, char **blacklist
, FILE *proc_self_mountinfo
) {
1007 bool submounts
= false;
1011 assert(proc_self_mountinfo
);
1013 if (mount_entry_read_only(m
)) {
1014 if (IN_SET(m
->mode
, EMPTY_DIR
, TMPFS
)) {
1015 /* Make superblock readonly */
1016 if (mount(NULL
, mount_entry_path(m
), NULL
, MS_REMOUNT
| MS_RDONLY
| m
->flags
, mount_entry_options(m
)) < 0)
1020 r
= bind_remount_recursive_with_mountinfo(mount_entry_path(m
), true, blacklist
, proc_self_mountinfo
);
1022 } else if (m
->mode
== PRIVATE_DEV
) {
1023 /* Superblock can be readonly but the submounts can't */
1024 if (mount(NULL
, mount_entry_path(m
), NULL
, MS_REMOUNT
|DEV_MOUNT_OPTIONS
|MS_RDONLY
, NULL
) < 0)
1029 /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked read-only
1030 * already stays this way. This improves compatibility with container managers, where we won't attempt to undo
1031 * read-only mounts already applied. */
1033 if (r
== -ENOENT
&& m
->ignore
)
1037 return log_debug_errno(r
, "Failed to re-mount '%s'%s read-only: %m", mount_entry_path(m
),
1038 submounts
? " and its submounts" : "");
1043 static bool namespace_info_mount_apivfs(const NamespaceInfo
*ns_info
) {
1047 * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=,
1048 * since to protect the API VFS mounts, they need to be around in the
1052 return ns_info
->mount_apivfs
||
1053 ns_info
->protect_control_groups
||
1054 ns_info
->protect_kernel_tunables
;
1057 static size_t namespace_calculate_mounts(
1058 const NamespaceInfo
*ns_info
,
1059 char** read_write_paths
,
1060 char** read_only_paths
,
1061 char** inaccessible_paths
,
1062 char** empty_directories
,
1063 size_t n_bind_mounts
,
1064 size_t n_temporary_filesystems
,
1065 const char* tmp_dir
,
1066 const char* var_tmp_dir
,
1067 ProtectHome protect_home
,
1068 ProtectSystem protect_system
) {
1070 size_t protect_home_cnt
;
1071 size_t protect_system_cnt
=
1072 (protect_system
== PROTECT_SYSTEM_STRICT
?
1073 ELEMENTSOF(protect_system_strict_table
) :
1074 ((protect_system
== PROTECT_SYSTEM_FULL
) ?
1075 ELEMENTSOF(protect_system_full_table
) :
1076 ((protect_system
== PROTECT_SYSTEM_YES
) ?
1077 ELEMENTSOF(protect_system_yes_table
) : 0)));
1080 (protect_home
== PROTECT_HOME_YES
?
1081 ELEMENTSOF(protect_home_yes_table
) :
1082 ((protect_home
== PROTECT_HOME_READ_ONLY
) ?
1083 ELEMENTSOF(protect_home_read_only_table
) :
1084 ((protect_home
== PROTECT_HOME_TMPFS
) ?
1085 ELEMENTSOF(protect_home_tmpfs_table
) : 0)));
1087 return !!tmp_dir
+ !!var_tmp_dir
+
1088 strv_length(read_write_paths
) +
1089 strv_length(read_only_paths
) +
1090 strv_length(inaccessible_paths
) +
1091 strv_length(empty_directories
) +
1093 n_temporary_filesystems
+
1094 ns_info
->private_dev
+
1095 (ns_info
->protect_kernel_tunables
? ELEMENTSOF(protect_kernel_tunables_table
) : 0) +
1096 (ns_info
->protect_control_groups
? 1 : 0) +
1097 (ns_info
->protect_kernel_modules
? ELEMENTSOF(protect_kernel_modules_table
) : 0) +
1098 protect_home_cnt
+ protect_system_cnt
+
1099 (namespace_info_mount_apivfs(ns_info
) ? ELEMENTSOF(apivfs_table
) : 0);
1102 static void normalize_mounts(const char *root_directory
, MountEntry
*mounts
, size_t *n_mounts
) {
1103 assert(root_directory
);
1105 assert(mounts
|| *n_mounts
== 0);
1107 qsort_safe(mounts
, *n_mounts
, sizeof(MountEntry
), mount_path_compare
);
1109 drop_duplicates(mounts
, n_mounts
);
1110 drop_outside_root(root_directory
, mounts
, n_mounts
);
1111 drop_inaccessible(mounts
, n_mounts
);
1112 drop_nop(mounts
, n_mounts
);
1115 int setup_namespace(
1116 const char* root_directory
,
1117 const char* root_image
,
1118 const NamespaceInfo
*ns_info
,
1119 char** read_write_paths
,
1120 char** read_only_paths
,
1121 char** inaccessible_paths
,
1122 char** empty_directories
,
1123 const BindMount
*bind_mounts
,
1124 size_t n_bind_mounts
,
1125 const TemporaryFileSystem
*temporary_filesystems
,
1126 size_t n_temporary_filesystems
,
1127 const char* tmp_dir
,
1128 const char* var_tmp_dir
,
1129 ProtectHome protect_home
,
1130 ProtectSystem protect_system
,
1131 unsigned long mount_flags
,
1132 DissectImageFlags dissect_image_flags
) {
1134 _cleanup_(loop_device_unrefp
) LoopDevice
*loop_device
= NULL
;
1135 _cleanup_(decrypted_image_unrefp
) DecryptedImage
*decrypted_image
= NULL
;
1136 _cleanup_(dissected_image_unrefp
) DissectedImage
*dissected_image
= NULL
;
1137 _cleanup_free_
void *root_hash
= NULL
;
1138 MountEntry
*m
, *mounts
= NULL
;
1139 size_t n_mounts
, root_hash_size
= 0;
1140 bool require_prefix
= false;
1146 if (mount_flags
== 0)
1147 mount_flags
= MS_SHARED
;
1150 dissect_image_flags
|= DISSECT_IMAGE_REQUIRE_ROOT
;
1152 if (protect_system
== PROTECT_SYSTEM_STRICT
&&
1153 protect_home
!= PROTECT_HOME_NO
&&
1154 strv_isempty(read_write_paths
))
1155 dissect_image_flags
|= DISSECT_IMAGE_READ_ONLY
;
1157 r
= loop_device_make_by_path(root_image
,
1158 dissect_image_flags
& DISSECT_IMAGE_READ_ONLY
? O_RDONLY
: O_RDWR
,
1161 return log_debug_errno(r
, "Failed to create loop device for root image: %m");
1163 r
= root_hash_load(root_image
, &root_hash
, &root_hash_size
);
1165 return log_debug_errno(r
, "Failed to load root hash: %m");
1167 r
= dissect_image(loop_device
->fd
, root_hash
, root_hash_size
, dissect_image_flags
, &dissected_image
);
1169 return log_debug_errno(r
, "Failed to dissect image: %m");
1171 r
= dissected_image_decrypt(dissected_image
, NULL
, root_hash
, root_hash_size
, dissect_image_flags
, &decrypted_image
);
1173 return log_debug_errno(r
, "Failed to decrypt dissected image: %m");
1177 root
= root_directory
;
1179 /* Always create the mount namespace in a temporary directory, instead of operating
1180 * directly in the root. The temporary directory prevents any mounts from being
1181 * potentially obscured my other mounts we already applied.
1182 * We use the same mount point for all images, which is safe, since they all live
1183 * in their own namespaces after all, and hence won't see each other. */
1185 root
= "/run/systemd/unit-root";
1186 (void) mkdir_label(root
, 0700);
1187 require_prefix
= true;
1190 n_mounts
= namespace_calculate_mounts(
1197 n_temporary_filesystems
,
1198 tmp_dir
, var_tmp_dir
,
1199 protect_home
, protect_system
);
1202 m
= mounts
= (MountEntry
*) alloca0(n_mounts
* sizeof(MountEntry
));
1203 r
= append_access_mounts(&m
, read_write_paths
, READWRITE
, require_prefix
);
1207 r
= append_access_mounts(&m
, read_only_paths
, READONLY
, require_prefix
);
1211 r
= append_access_mounts(&m
, inaccessible_paths
, INACCESSIBLE
, require_prefix
);
1215 r
= append_empty_dir_mounts(&m
, empty_directories
);
1219 r
= append_bind_mounts(&m
, bind_mounts
, n_bind_mounts
);
1223 r
= append_tmpfs_mounts(&m
, temporary_filesystems
, n_temporary_filesystems
);
1228 *(m
++) = (MountEntry
) {
1229 .path_const
= "/tmp",
1230 .mode
= PRIVATE_TMP
,
1231 .source_const
= tmp_dir
,
1236 *(m
++) = (MountEntry
) {
1237 .path_const
= "/var/tmp",
1238 .mode
= PRIVATE_TMP
,
1239 .source_const
= var_tmp_dir
,
1243 if (ns_info
->private_dev
) {
1244 *(m
++) = (MountEntry
) {
1245 .path_const
= "/dev",
1246 .mode
= PRIVATE_DEV
,
1250 if (ns_info
->protect_kernel_tunables
) {
1251 r
= append_static_mounts(&m
, protect_kernel_tunables_table
, ELEMENTSOF(protect_kernel_tunables_table
), ns_info
->ignore_protect_paths
);
1256 if (ns_info
->protect_kernel_modules
) {
1257 r
= append_static_mounts(&m
, protect_kernel_modules_table
, ELEMENTSOF(protect_kernel_modules_table
), ns_info
->ignore_protect_paths
);
1262 if (ns_info
->protect_control_groups
) {
1263 *(m
++) = (MountEntry
) {
1264 .path_const
= "/sys/fs/cgroup",
1269 r
= append_protect_home(&m
, protect_home
, ns_info
->ignore_protect_paths
);
1273 r
= append_protect_system(&m
, protect_system
, false);
1277 if (namespace_info_mount_apivfs(ns_info
)) {
1278 r
= append_static_mounts(&m
, apivfs_table
, ELEMENTSOF(apivfs_table
), ns_info
->ignore_protect_paths
);
1283 assert(mounts
+ n_mounts
== m
);
1285 /* Prepend the root directory where that's necessary */
1286 r
= prefix_where_needed(mounts
, n_mounts
, root
);
1290 normalize_mounts(root
, mounts
, &n_mounts
);
1293 if (unshare(CLONE_NEWNS
) < 0) {
1294 r
= log_debug_errno(errno
, "Failed to unshare the mount namespace: %m");
1298 /* Remount / as SLAVE so that nothing now mounted in the namespace
1299 * shows up in the parent */
1300 if (mount(NULL
, "/", NULL
, MS_SLAVE
|MS_REC
, NULL
) < 0) {
1301 r
= log_debug_errno(errno
, "Failed to remount '/' as SLAVE: %m");
1306 /* A root image is specified, mount it to the right place */
1307 r
= dissected_image_mount(dissected_image
, root
, UID_INVALID
, dissect_image_flags
);
1309 log_debug_errno(r
, "Failed to mount root image: %m");
1313 if (decrypted_image
) {
1314 r
= decrypted_image_relinquish(decrypted_image
);
1316 log_debug_errno(r
, "Failed to relinquish decrypted image: %m");
1321 loop_device_relinquish(loop_device
);
1323 } else if (root_directory
) {
1325 /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */
1326 r
= path_is_mount_point(root
, NULL
, AT_SYMLINK_FOLLOW
);
1328 log_debug_errno(r
, "Failed to detect that %s is a mount point or not: %m", root
);
1332 if (mount(root
, root
, NULL
, MS_BIND
|MS_REC
, NULL
) < 0) {
1333 r
= log_debug_errno(errno
, "Failed to bind mount '%s': %m", root
);
1340 /* Let's mount the main root directory to the root directory to use */
1341 if (mount("/", root
, NULL
, MS_BIND
|MS_REC
, NULL
) < 0) {
1342 r
= log_debug_errno(errno
, "Failed to bind mount '/' on '%s': %m", root
);
1347 /* Try to set up the new root directory before mounting anything else there. */
1348 if (root_image
|| root_directory
)
1349 (void) base_filesystem_create(root
, UID_INVALID
, GID_INVALID
);
1352 _cleanup_fclose_
FILE *proc_self_mountinfo
= NULL
;
1356 /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of /proc.
1357 * For example, this is the case with the option: 'InaccessiblePaths=/proc' */
1358 proc_self_mountinfo
= fopen("/proc/self/mountinfo", "re");
1359 if (!proc_self_mountinfo
) {
1360 r
= log_debug_errno(errno
, "Failed to open /proc/self/mountinfo: %m");
1364 /* First round, establish all mounts we need */
1368 for (m
= mounts
; m
< mounts
+ n_mounts
; ++m
) {
1373 r
= follow_symlink(root
, m
);
1377 /* We hit a symlinked mount point. The entry got rewritten and might point to a
1378 * very different place now. Let's normalize the changed list, and start from
1379 * the beginning. After all to mount the entry at the new location we might
1380 * need some other mounts first */
1385 r
= apply_mount(root
, m
);
1395 normalize_mounts(root
, mounts
, &n_mounts
);
1398 /* Create a blacklist we can pass to bind_mount_recursive() */
1399 blacklist
= newa(char*, n_mounts
+1);
1400 for (j
= 0; j
< n_mounts
; j
++)
1401 blacklist
[j
] = (char*) mount_entry_path(mounts
+j
);
1402 blacklist
[j
] = NULL
;
1404 /* Second round, flip the ro bits if necessary. */
1405 for (m
= mounts
; m
< mounts
+ n_mounts
; ++m
) {
1406 r
= make_read_only(m
, blacklist
, proc_self_mountinfo
);
1412 /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */
1413 r
= mount_move_root(root
);
1415 log_debug_errno(r
, "Failed to mount root with MS_MOVE: %m");
1419 /* Remount / as the desired mode. Note that this will not
1420 * reestablish propagation from our side to the host, since
1421 * what's disconnected is disconnected. */
1422 if (mount(NULL
, "/", NULL
, mount_flags
| MS_REC
, NULL
) < 0) {
1423 r
= log_debug_errno(errno
, "Failed to remount '/' with desired mount flags: %m");
1430 for (m
= mounts
; m
< mounts
+ n_mounts
; m
++)
1431 mount_entry_done(m
);
1436 void bind_mount_free_many(BindMount
*b
, size_t n
) {
1439 assert(b
|| n
== 0);
1441 for (i
= 0; i
< n
; i
++) {
1443 free(b
[i
].destination
);
1449 int bind_mount_add(BindMount
**b
, size_t *n
, const BindMount
*item
) {
1450 _cleanup_free_
char *s
= NULL
, *d
= NULL
;
1457 s
= strdup(item
->source
);
1461 d
= strdup(item
->destination
);
1465 c
= reallocarray(*b
, *n
+ 1, sizeof(BindMount
));
1471 c
[(*n
) ++] = (BindMount
) {
1472 .source
= TAKE_PTR(s
),
1473 .destination
= TAKE_PTR(d
),
1474 .read_only
= item
->read_only
,
1475 .recursive
= item
->recursive
,
1476 .ignore_enoent
= item
->ignore_enoent
,
1482 void temporary_filesystem_free_many(TemporaryFileSystem
*t
, size_t n
) {
1485 assert(t
|| n
== 0);
1487 for (i
= 0; i
< n
; i
++) {
1495 int temporary_filesystem_add(
1496 TemporaryFileSystem
**t
,
1499 const char *options
) {
1501 _cleanup_free_
char *p
= NULL
, *o
= NULL
;
1502 TemporaryFileSystem
*c
;
1512 if (!isempty(options
)) {
1513 o
= strdup(options
);
1518 c
= reallocarray(*t
, *n
+ 1, sizeof(TemporaryFileSystem
));
1524 c
[(*n
) ++] = (TemporaryFileSystem
) {
1525 .path
= TAKE_PTR(p
),
1526 .options
= TAKE_PTR(o
),
1532 static int setup_one_tmp_dir(const char *id
, const char *prefix
, char **path
) {
1533 _cleanup_free_
char *x
= NULL
;
1534 char bid
[SD_ID128_STRING_MAX
];
1542 /* We include the boot id in the directory so that after a
1543 * reboot we can easily identify obsolete directories. */
1545 r
= sd_id128_get_boot(&boot_id
);
1549 x
= strjoin(prefix
, "/systemd-private-", sd_id128_to_string(boot_id
, bid
), "-", id
, "-XXXXXX");
1553 RUN_WITH_UMASK(0077)
1557 RUN_WITH_UMASK(0000) {
1560 y
= strjoina(x
, "/tmp");
1562 if (mkdir(y
, 0777 | S_ISVTX
) < 0)
1566 *path
= TAKE_PTR(x
);
1571 int setup_tmp_dirs(const char *id
, char **tmp_dir
, char **var_tmp_dir
) {
1577 assert(var_tmp_dir
);
1579 r
= setup_one_tmp_dir(id
, "/tmp", &a
);
1583 r
= setup_one_tmp_dir(id
, "/var/tmp", &b
);
1587 t
= strjoina(a
, "/tmp");
1601 int setup_netns(int netns_storage_socket
[2]) {
1602 _cleanup_close_
int netns
= -1;
1605 assert(netns_storage_socket
);
1606 assert(netns_storage_socket
[0] >= 0);
1607 assert(netns_storage_socket
[1] >= 0);
1609 /* We use the passed socketpair as a storage buffer for our
1610 * namespace reference fd. Whatever process runs this first
1611 * shall create a new namespace, all others should just join
1612 * it. To serialize that we use a file lock on the socket
1615 * It's a bit crazy, but hey, works great! */
1617 if (lockf(netns_storage_socket
[0], F_LOCK
, 0) < 0)
1620 netns
= receive_one_fd(netns_storage_socket
[0], MSG_DONTWAIT
);
1621 if (netns
== -EAGAIN
) {
1622 /* Nothing stored yet, so let's create a new namespace */
1624 if (unshare(CLONE_NEWNET
) < 0) {
1631 netns
= open("/proc/self/ns/net", O_RDONLY
|O_CLOEXEC
|O_NOCTTY
);
1639 } else if (netns
< 0) {
1644 /* Yay, found something, so let's join the namespace */
1645 if (setns(netns
, CLONE_NEWNET
) < 0) {
1653 q
= send_one_fd(netns_storage_socket
[1], netns
, MSG_DONTWAIT
);
1660 (void) lockf(netns_storage_socket
[0], F_ULOCK
, 0);
1664 bool ns_type_supported(NamespaceType type
) {
1665 const char *t
, *ns_proc
;
1667 t
= namespace_type_to_string(type
);
1668 if (!t
) /* Don't know how to translate this? Then it's not supported */
1671 ns_proc
= strjoina("/proc/self/ns/", t
);
1672 return access(ns_proc
, F_OK
) == 0;
1675 static const char *const protect_home_table
[_PROTECT_HOME_MAX
] = {
1676 [PROTECT_HOME_NO
] = "no",
1677 [PROTECT_HOME_YES
] = "yes",
1678 [PROTECT_HOME_READ_ONLY
] = "read-only",
1679 [PROTECT_HOME_TMPFS
] = "tmpfs",
1682 DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_home
, ProtectHome
, PROTECT_HOME_YES
);
1684 static const char *const protect_system_table
[_PROTECT_SYSTEM_MAX
] = {
1685 [PROTECT_SYSTEM_NO
] = "no",
1686 [PROTECT_SYSTEM_YES
] = "yes",
1687 [PROTECT_SYSTEM_FULL
] = "full",
1688 [PROTECT_SYSTEM_STRICT
] = "strict",
1691 DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_system
, ProtectSystem
, PROTECT_SYSTEM_YES
);
1693 static const char* const namespace_type_table
[] = {
1694 [NAMESPACE_MOUNT
] = "mnt",
1695 [NAMESPACE_CGROUP
] = "cgroup",
1696 [NAMESPACE_UTS
] = "uts",
1697 [NAMESPACE_IPC
] = "ipc",
1698 [NAMESPACE_USER
] = "user",
1699 [NAMESPACE_PID
] = "pid",
1700 [NAMESPACE_NET
] = "net",
1703 DEFINE_STRING_TABLE_LOOKUP(namespace_type
, NamespaceType
);