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 "mountpoint-util.h"
24 #include "namespace.h"
25 #include "path-util.h"
26 #include "selinux-util.h"
27 #include "socket-util.h"
28 #include "stat-util.h"
29 #include "string-table.h"
30 #include "string-util.h"
32 #include "umask-util.h"
33 #include "user-util.h"
36 #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC)
38 typedef enum MountMode
{
39 /* This is ordered by priority! */
54 typedef struct MountEntry
{
55 const char *path_const
; /* Memory allocated on stack or static */
57 bool ignore
:1; /* Ignore if path does not exist? */
58 bool has_prefix
:1; /* Already is prefixed by the root dir? */
59 bool read_only
:1; /* Shall this mount point be read-only? */
60 bool applied
:1; /* Already applied */
61 char *path_malloc
; /* Use this instead of 'path_const' if we had to allocate memory */
62 const char *source_const
; /* The source path, for bind mounts */
64 const char *options_const
;/* Mount options for tmpfs */
66 unsigned long flags
; /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */
70 /* If MountAPIVFS= is used, let's mount /sys and /proc into the it, but only as a fallback if the user hasn't mounted
71 * something there already. These mounts are hence overridden by any other explicitly configured mounts. */
72 static const MountEntry apivfs_table
[] = {
73 { "/proc", PROCFS
, false },
74 { "/dev", BIND_DEV
, false },
75 { "/sys", SYSFS
, false },
78 /* ProtectKernelTunables= option and the related filesystem APIs */
79 static const MountEntry protect_kernel_tunables_table
[] = {
80 { "/proc/acpi", READONLY
, true },
81 { "/proc/apm", READONLY
, true }, /* Obsolete API, there's no point in permitting access to this, ever */
82 { "/proc/asound", READONLY
, true },
83 { "/proc/bus", READONLY
, true },
84 { "/proc/fs", READONLY
, true },
85 { "/proc/irq", READONLY
, true },
86 { "/proc/kallsyms", INACCESSIBLE
, true },
87 { "/proc/kcore", INACCESSIBLE
, true },
88 { "/proc/latency_stats", READONLY
, true },
89 { "/proc/mtrr", READONLY
, true },
90 { "/proc/scsi", READONLY
, true },
91 { "/proc/sys", READONLY
, false },
92 { "/proc/sysrq-trigger", READONLY
, true },
93 { "/proc/timer_stats", READONLY
, true },
94 { "/sys", READONLY
, false },
95 { "/sys/fs/bpf", READONLY
, true },
96 { "/sys/fs/cgroup", READWRITE
, false }, /* READONLY is set by ProtectControlGroups= option */
97 { "/sys/fs/selinux", READWRITE
, true },
98 { "/sys/kernel/debug", READONLY
, true },
99 { "/sys/kernel/tracing", READONLY
, true },
102 /* ProtectKernelModules= option */
103 static const MountEntry protect_kernel_modules_table
[] = {
105 { "/lib/modules", INACCESSIBLE
, true },
107 { "/usr/lib/modules", INACCESSIBLE
, true },
111 * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of
112 * system should be protected by ProtectSystem=
114 static const MountEntry protect_home_read_only_table
[] = {
115 { "/home", READONLY
, true },
116 { "/run/user", READONLY
, true },
117 { "/root", READONLY
, true },
120 /* ProtectHome=tmpfs table */
121 static const MountEntry protect_home_tmpfs_table
[] = {
122 { "/home", TMPFS
, true, .read_only
= true, .options_const
= "mode=0755", .flags
= MS_NODEV
|MS_STRICTATIME
},
123 { "/run/user", TMPFS
, true, .read_only
= true, .options_const
= "mode=0755", .flags
= MS_NODEV
|MS_STRICTATIME
},
124 { "/root", TMPFS
, true, .read_only
= true, .options_const
= "mode=0700", .flags
= MS_NODEV
|MS_STRICTATIME
},
127 /* ProtectHome=yes table */
128 static const MountEntry protect_home_yes_table
[] = {
129 { "/home", INACCESSIBLE
, true },
130 { "/run/user", INACCESSIBLE
, true },
131 { "/root", INACCESSIBLE
, true },
134 /* ProtectSystem=yes table */
135 static const MountEntry protect_system_yes_table
[] = {
136 { "/usr", READONLY
, false },
137 { "/boot", READONLY
, true },
138 { "/efi", READONLY
, true },
140 { "/lib", READONLY
, true },
141 { "/lib64", READONLY
, true },
142 { "/bin", READONLY
, true },
144 { "/sbin", READONLY
, true },
149 /* ProtectSystem=full includes ProtectSystem=yes */
150 static const MountEntry protect_system_full_table
[] = {
151 { "/usr", READONLY
, false },
152 { "/boot", READONLY
, true },
153 { "/efi", READONLY
, true },
154 { "/etc", READONLY
, false },
156 { "/lib", READONLY
, true },
157 { "/lib64", READONLY
, true },
158 { "/bin", READONLY
, true },
160 { "/sbin", READONLY
, true },
166 * ProtectSystem=strict table. In this strict mode, we mount everything
167 * read-only, except for /proc, /dev, /sys which are the kernel API VFS,
168 * which are left writable, but PrivateDevices= + ProtectKernelTunables=
169 * protect those, and these options should be fully orthogonal.
170 * (And of course /home and friends are also left writable, as ProtectHome=
171 * shall manage those, orthogonally).
173 static const MountEntry protect_system_strict_table
[] = {
174 { "/", READONLY
, false },
175 { "/proc", READWRITE
, false }, /* ProtectKernelTunables= */
176 { "/sys", READWRITE
, false }, /* ProtectKernelTunables= */
177 { "/dev", READWRITE
, false }, /* PrivateDevices= */
178 { "/home", READWRITE
, true }, /* ProtectHome= */
179 { "/run/user", READWRITE
, true }, /* ProtectHome= */
180 { "/root", READWRITE
, true }, /* ProtectHome= */
183 static const char *mount_entry_path(const MountEntry
*p
) {
186 /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that,
187 * otherwise the stack/static ->path field is returned. */
189 return p
->path_malloc
?: p
->path_const
;
192 static bool mount_entry_read_only(const MountEntry
*p
) {
195 return p
->read_only
|| IN_SET(p
->mode
, READONLY
, INACCESSIBLE
);
198 static const char *mount_entry_source(const MountEntry
*p
) {
201 return p
->source_malloc
?: p
->source_const
;
204 static const char *mount_entry_options(const MountEntry
*p
) {
207 return p
->options_malloc
?: p
->options_const
;
210 static void mount_entry_done(MountEntry
*p
) {
213 p
->path_malloc
= mfree(p
->path_malloc
);
214 p
->source_malloc
= mfree(p
->source_malloc
);
215 p
->options_malloc
= mfree(p
->options_malloc
);
218 static int append_access_mounts(MountEntry
**p
, char **strv
, MountMode mode
, bool forcibly_require_prefix
) {
223 /* Adds a list of user-supplied READWRITE/READONLY/INACCESSIBLE entries */
225 STRV_FOREACH(i
, strv
) {
226 bool ignore
= false, needs_prefix
= false;
229 /* Look for any prefixes */
230 if (startswith(e
, "-")) {
234 if (startswith(e
, "+")) {
239 if (!path_is_absolute(e
))
240 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL
),
241 "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
) {
269 .options_const
= "mode=755",
270 .flags
= MS_NOSUID
|MS_NOEXEC
|MS_NODEV
|MS_STRICTATIME
,
277 static int append_bind_mounts(MountEntry
**p
, const BindMount
*binds
, size_t n
) {
282 for (i
= 0; i
< n
; i
++) {
283 const BindMount
*b
= binds
+ i
;
285 *((*p
)++) = (MountEntry
) {
286 .path_const
= b
->destination
,
287 .mode
= b
->recursive
? BIND_MOUNT_RECURSIVE
: BIND_MOUNT
,
288 .read_only
= b
->read_only
,
289 .source_const
= b
->source
,
290 .ignore
= b
->ignore_enoent
,
297 static int append_tmpfs_mounts(MountEntry
**p
, const TemporaryFileSystem
*tmpfs
, size_t n
) {
303 for (i
= 0; i
< n
; i
++) {
304 const TemporaryFileSystem
*t
= tmpfs
+ i
;
305 _cleanup_free_
char *o
= NULL
, *str
= NULL
;
309 if (!path_is_absolute(t
->path
))
310 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL
),
311 "Path is not absolute: %s",
314 str
= strjoin("mode=0755,", t
->options
);
318 r
= mount_option_mangle(str
, MS_NODEV
|MS_STRICTATIME
, &flags
, &o
);
320 return log_debug_errno(r
, "Failed to parse mount option '%s': %m", str
);
322 ro
= flags
& MS_RDONLY
;
326 *((*p
)++) = (MountEntry
) {
327 .path_const
= t
->path
,
330 .options_malloc
= TAKE_PTR(o
),
338 static int append_static_mounts(MountEntry
**p
, const MountEntry
*mounts
, size_t n
, bool ignore_protect
) {
344 /* Adds a list of static pre-defined entries */
346 for (i
= 0; i
< n
; i
++)
347 *((*p
)++) = (MountEntry
) {
348 .path_const
= mount_entry_path(mounts
+i
),
349 .mode
= mounts
[i
].mode
,
350 .ignore
= mounts
[i
].ignore
|| ignore_protect
,
356 static int append_protect_home(MountEntry
**p
, ProtectHome protect_home
, bool ignore_protect
) {
359 switch (protect_home
) {
361 case PROTECT_HOME_NO
:
364 case PROTECT_HOME_READ_ONLY
:
365 return append_static_mounts(p
, protect_home_read_only_table
, ELEMENTSOF(protect_home_read_only_table
), ignore_protect
);
367 case PROTECT_HOME_TMPFS
:
368 return append_static_mounts(p
, protect_home_tmpfs_table
, ELEMENTSOF(protect_home_tmpfs_table
), ignore_protect
);
370 case PROTECT_HOME_YES
:
371 return append_static_mounts(p
, protect_home_yes_table
, ELEMENTSOF(protect_home_yes_table
), ignore_protect
);
374 assert_not_reached("Unexpected ProtectHome= value");
378 static int append_protect_system(MountEntry
**p
, ProtectSystem protect_system
, bool ignore_protect
) {
381 switch (protect_system
) {
383 case PROTECT_SYSTEM_NO
:
386 case PROTECT_SYSTEM_STRICT
:
387 return append_static_mounts(p
, protect_system_strict_table
, ELEMENTSOF(protect_system_strict_table
), ignore_protect
);
389 case PROTECT_SYSTEM_YES
:
390 return append_static_mounts(p
, protect_system_yes_table
, ELEMENTSOF(protect_system_yes_table
), ignore_protect
);
392 case PROTECT_SYSTEM_FULL
:
393 return append_static_mounts(p
, protect_system_full_table
, ELEMENTSOF(protect_system_full_table
), ignore_protect
);
396 assert_not_reached("Unexpected ProtectSystem= value");
400 static int mount_path_compare(const MountEntry
*a
, const MountEntry
*b
) {
403 /* If the paths are not equal, then order prefixes first */
404 d
= path_compare(mount_entry_path(a
), mount_entry_path(b
));
408 /* If the paths are equal, check the mode */
409 return CMP((int) a
->mode
, (int) b
->mode
);
412 static int prefix_where_needed(MountEntry
*m
, size_t n
, const char *root_directory
) {
415 /* Prefixes all paths in the bind mount table with the root directory if the entry needs that. */
417 for (i
= 0; i
< n
; i
++) {
423 s
= prefix_root(root_directory
, mount_entry_path(m
+i
));
427 free_and_replace(m
[i
].path_malloc
, s
);
428 m
[i
].has_prefix
= true;
434 static void drop_duplicates(MountEntry
*m
, size_t *n
) {
435 MountEntry
*f
, *t
, *previous
;
440 /* Drops duplicate entries. Expects that the array is properly ordered already. */
442 for (f
= m
, t
= m
, previous
= NULL
; f
< m
+ *n
; f
++) {
444 /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare()
445 * above. Note that we only drop duplicates that haven't been mounted yet. */
447 path_equal(mount_entry_path(f
), mount_entry_path(previous
)) &&
448 !f
->applied
&& !previous
->applied
) {
449 log_debug("%s is duplicate.", mount_entry_path(f
));
450 previous
->read_only
= previous
->read_only
|| mount_entry_read_only(f
); /* Propagate the read-only flag to the remaining entry */
463 static void drop_inaccessible(MountEntry
*m
, size_t *n
) {
465 const char *clear
= NULL
;
470 /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly
471 * ordered already. */
473 for (f
= m
, t
= m
; f
< m
+ *n
; f
++) {
475 /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop
476 * it, as inaccessible paths really should drop the entire subtree. */
477 if (clear
&& path_startswith(mount_entry_path(f
), clear
)) {
478 log_debug("%s is masked by %s.", mount_entry_path(f
), clear
);
483 clear
= f
->mode
== INACCESSIBLE
? mount_entry_path(f
) : NULL
;
492 static void drop_nop(MountEntry
*m
, size_t *n
) {
498 /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the
499 * list is ordered by prefixes. */
501 for (f
= m
, t
= m
; f
< m
+ *n
; f
++) {
503 /* Only suppress such subtrees for READONLY and READWRITE entries */
504 if (IN_SET(f
->mode
, READONLY
, READWRITE
)) {
508 /* Now let's find the first parent of the entry we are looking at. */
509 for (p
= t
-1; p
>= m
; p
--) {
510 if (path_startswith(mount_entry_path(f
), mount_entry_path(p
))) {
516 /* We found it, let's see if it's the same mode, if so, we can drop this entry */
517 if (found
&& p
->mode
== f
->mode
) {
518 log_debug("%s is redundant by %s", mount_entry_path(f
), mount_entry_path(p
));
531 static void drop_outside_root(const char *root_directory
, MountEntry
*m
, size_t *n
) {
541 /* Drops all mounts that are outside of the root directory. */
543 for (f
= m
, t
= m
; f
< m
+ *n
; f
++) {
545 if (!path_startswith(mount_entry_path(f
), root_directory
)) {
546 log_debug("%s is outside of root directory.", mount_entry_path(f
));
558 static int clone_device_node(
560 const char *temporary_mount
,
561 bool *make_devnode
) {
563 _cleanup_free_
char *sl
= NULL
;
564 const char *dn
, *bn
, *t
;
568 if (stat(d
, &st
) < 0) {
569 if (errno
== ENOENT
) {
570 log_debug_errno(errno
, "Device node '%s' to clone does not exist, ignoring.", d
);
574 return log_debug_errno(errno
, "Failed to stat() device node '%s' to clone, ignoring: %m", d
);
577 if (!S_ISBLK(st
.st_mode
) &&
578 !S_ISCHR(st
.st_mode
))
579 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL
),
580 "Device node '%s' to clone is not a device node, ignoring.",
583 dn
= strjoina(temporary_mount
, d
);
585 /* First, try to create device node properly */
587 mac_selinux_create_file_prepare(d
, st
.st_mode
);
588 r
= mknod(dn
, st
.st_mode
, st
.st_rdev
);
589 mac_selinux_create_file_clear();
593 return log_debug_errno(errno
, "mknod failed for %s: %m", d
);
595 /* This didn't work, let's not try this again for the next iterations. */
596 *make_devnode
= false;
599 /* We're about to fallback to bind-mounting the device
600 * node. So create a dummy bind-mount target. */
601 mac_selinux_create_file_prepare(d
, 0);
602 r
= mknod(dn
, S_IFREG
, 0);
603 mac_selinux_create_file_clear();
604 if (r
< 0 && errno
!= EEXIST
)
605 return log_debug_errno(errno
, "mknod() fallback failed for '%s': %m", d
);
607 /* Fallback to bind-mounting:
608 * The assumption here is that all used device nodes carry standard
609 * properties. Specifically, the devices nodes we bind-mount should
610 * either be owned by root:root or root:tty (e.g. /dev/tty, /dev/ptmx)
611 * and should not carry ACLs. */
612 if (mount(d
, dn
, NULL
, MS_BIND
, NULL
) < 0)
613 return log_debug_errno(errno
, "Bind mounting failed for '%s': %m", d
);
616 bn
= path_startswith(d
, "/dev/");
620 /* Create symlinks like /dev/char/1:9 → ../urandom */
621 if (asprintf(&sl
, "%s/dev/%s/%u:%u", temporary_mount
, S_ISCHR(st
.st_mode
) ? "char" : "block", major(st
.st_rdev
), minor(st
.st_rdev
)) < 0)
624 (void) mkdir_parents(sl
, 0755);
626 t
= strjoina("../", bn
);
628 if (symlink(t
, sl
) < 0)
629 log_debug_errno(errno
, "Failed to symlink '%s' to '%s', ignoring: %m", t
, sl
);
634 static int mount_private_dev(MountEntry
*m
) {
635 static const char devnodes
[] =
643 char temporary_mount
[] = "/tmp/namespace-dev-XXXXXX";
644 const char *d
, *dev
= NULL
, *devpts
= NULL
, *devshm
= NULL
, *devhugepages
= NULL
, *devmqueue
= NULL
, *devlog
= NULL
, *devptmx
= NULL
;
645 bool can_mknod
= true;
646 _cleanup_umask_ mode_t u
;
653 if (!mkdtemp(temporary_mount
))
654 return log_debug_errno(errno
, "Failed to create temporary directory '%s': %m", temporary_mount
);
656 dev
= strjoina(temporary_mount
, "/dev");
657 (void) mkdir(dev
, 0755);
658 if (mount("tmpfs", dev
, "tmpfs", DEV_MOUNT_OPTIONS
, "mode=755") < 0) {
659 r
= log_debug_errno(errno
, "Failed to mount tmpfs on '%s': %m", dev
);
663 devpts
= strjoina(temporary_mount
, "/dev/pts");
664 (void) mkdir(devpts
, 0755);
665 if (mount("/dev/pts", devpts
, NULL
, MS_BIND
, NULL
) < 0) {
666 r
= log_debug_errno(errno
, "Failed to bind mount /dev/pts on '%s': %m", devpts
);
670 /* /dev/ptmx can either be a device node or a symlink to /dev/pts/ptmx.
671 * When /dev/ptmx a device node, /dev/pts/ptmx has 000 permissions making it inaccessible.
672 * Thus, in that case make a clone.
673 * In nspawn and other containers it will be a symlink, in that case make it a symlink. */
674 r
= is_symlink("/dev/ptmx");
676 log_debug_errno(r
, "Failed to detect whether /dev/ptmx is a symlink or not: %m");
679 devptmx
= strjoina(temporary_mount
, "/dev/ptmx");
680 if (symlink("pts/ptmx", devptmx
) < 0) {
681 r
= log_debug_errno(errno
, "Failed to create a symlink '%s' to pts/ptmx: %m", devptmx
);
685 r
= clone_device_node("/dev/ptmx", temporary_mount
, &can_mknod
);
690 devshm
= strjoina(temporary_mount
, "/dev/shm");
691 (void) mkdir(devshm
, 0755);
692 r
= mount("/dev/shm", devshm
, NULL
, MS_BIND
, NULL
);
694 r
= log_debug_errno(errno
, "Failed to bind mount /dev/shm on '%s': %m", devshm
);
698 devmqueue
= strjoina(temporary_mount
, "/dev/mqueue");
699 (void) mkdir(devmqueue
, 0755);
700 if (mount("/dev/mqueue", devmqueue
, NULL
, MS_BIND
, NULL
) < 0)
701 log_debug_errno(errno
, "Failed to bind mount /dev/mqueue on '%s', ignoring: %m", devmqueue
);
703 devhugepages
= strjoina(temporary_mount
, "/dev/hugepages");
704 (void) mkdir(devhugepages
, 0755);
705 if (mount("/dev/hugepages", devhugepages
, NULL
, MS_BIND
, NULL
) < 0)
706 log_debug_errno(errno
, "Failed to bind mount /dev/hugepages on '%s', ignoring: %m", devhugepages
);
708 devlog
= strjoina(temporary_mount
, "/dev/log");
709 if (symlink("/run/systemd/journal/dev-log", devlog
) < 0)
710 log_debug_errno(errno
, "Failed to create a symlink '%s' to /run/systemd/journal/dev-log, ignoring: %m", devlog
);
712 NULSTR_FOREACH(d
, devnodes
) {
713 r
= clone_device_node(d
, temporary_mount
, &can_mknod
);
714 /* ENXIO means the the *source* is not a device file, skip creation in that case */
715 if (r
< 0 && r
!= -ENXIO
)
719 r
= dev_setup(temporary_mount
, UID_INVALID
, GID_INVALID
);
721 log_debug_errno(r
, "Failed to setup basic device tree at '%s', ignoring: %m", temporary_mount
);
723 /* Create the /dev directory if missing. It is more likely to be
724 * missing when the service is started with RootDirectory. This is
725 * consistent with mount units creating the mount points when missing.
727 (void) mkdir_p_label(mount_entry_path(m
), 0755);
729 /* Unmount everything in old /dev */
730 r
= umount_recursive(mount_entry_path(m
), 0);
732 log_debug_errno(r
, "Failed to unmount directories below '%s', ignoring: %m", mount_entry_path(m
));
734 if (mount(dev
, mount_entry_path(m
), NULL
, MS_MOVE
, NULL
) < 0) {
735 r
= log_debug_errno(errno
, "Failed to move mount point '%s' to '%s': %m", dev
, mount_entry_path(m
));
740 rmdir(temporary_mount
);
752 umount(devhugepages
);
759 rmdir(temporary_mount
);
764 static int mount_bind_dev(const MountEntry
*m
) {
769 /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's
770 * /dev. This is only used when RootDirectory= is set. */
772 (void) mkdir_p_label(mount_entry_path(m
), 0755);
774 r
= path_is_mount_point(mount_entry_path(m
), NULL
, 0);
776 return log_debug_errno(r
, "Unable to determine whether /dev is already mounted: %m");
777 if (r
> 0) /* make this a NOP if /dev is already a mount point */
780 if (mount("/dev", mount_entry_path(m
), NULL
, MS_BIND
|MS_REC
, NULL
) < 0)
781 return log_debug_errno(errno
, "Failed to bind mount %s: %m", mount_entry_path(m
));
786 static int mount_sysfs(const MountEntry
*m
) {
791 (void) mkdir_p_label(mount_entry_path(m
), 0755);
793 r
= path_is_mount_point(mount_entry_path(m
), NULL
, 0);
795 return log_debug_errno(r
, "Unable to determine whether /sys is already mounted: %m");
796 if (r
> 0) /* make this a NOP if /sys is already a mount point */
799 /* Bind mount the host's version so that we get all child mounts of it, too. */
800 if (mount("/sys", mount_entry_path(m
), NULL
, MS_BIND
|MS_REC
, NULL
) < 0)
801 return log_debug_errno(errno
, "Failed to mount %s: %m", mount_entry_path(m
));
806 static int mount_procfs(const MountEntry
*m
) {
811 (void) mkdir_p_label(mount_entry_path(m
), 0755);
813 r
= path_is_mount_point(mount_entry_path(m
), NULL
, 0);
815 return log_debug_errno(r
, "Unable to determine whether /proc is already mounted: %m");
816 if (r
> 0) /* make this a NOP if /proc is already a mount point */
819 /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */
820 if (mount("proc", mount_entry_path(m
), "proc", MS_NOSUID
|MS_NOEXEC
|MS_NODEV
, NULL
) < 0)
821 return log_debug_errno(errno
, "Failed to mount %s: %m", mount_entry_path(m
));
826 static int mount_tmpfs(const MountEntry
*m
) {
829 /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */
831 (void) mkdir_p_label(mount_entry_path(m
), 0755);
832 (void) umount_recursive(mount_entry_path(m
), 0);
834 if (mount("tmpfs", mount_entry_path(m
), "tmpfs", m
->flags
, mount_entry_options(m
)) < 0)
835 return log_debug_errno(errno
, "Failed to mount %s: %m", mount_entry_path(m
));
840 static int follow_symlink(
841 const char *root_directory
,
844 _cleanup_free_
char *target
= NULL
;
847 /* Let's chase symlinks, but only one step at a time. That's because depending where the symlink points we
848 * might need to change the order in which we mount stuff. Hence: let's normalize piecemeal, and do one step at
849 * a time by specifying CHASE_STEP. This function returns 0 if we resolved one step, and > 0 if we reached the
850 * end and already have a fully normalized name. */
852 r
= chase_symlinks(mount_entry_path(m
), root_directory
, CHASE_STEP
|CHASE_NONEXISTENT
, &target
);
854 return log_debug_errno(r
, "Failed to chase symlinks '%s': %m", mount_entry_path(m
));
855 if (r
> 0) /* Reached the end, nothing more to resolve */
858 if (m
->n_followed
>= CHASE_SYMLINKS_MAX
) /* put a boundary on things */
859 return log_debug_errno(SYNTHETIC_ERRNO(ELOOP
),
860 "Symlink loop on '%s'.",
861 mount_entry_path(m
));
863 log_debug("Followed mount entry path symlink %s → %s.", mount_entry_path(m
), target
);
865 free_and_replace(m
->path_malloc
, target
);
866 m
->has_prefix
= true;
873 static int apply_mount(
874 const char *root_directory
,
877 bool rbind
= true, make
= false;
883 log_debug("Applying namespace mount on %s", mount_entry_path(m
));
890 /* First, get rid of everything that is below if there
891 * is anything... Then, overmount it with an
892 * inaccessible path. */
893 (void) umount_recursive(mount_entry_path(m
), 0);
895 if (lstat(mount_entry_path(m
), &target
) < 0) {
896 if (errno
== ENOENT
&& m
->ignore
)
899 return log_debug_errno(errno
, "Failed to lstat() %s to determine what to mount over it: %m", mount_entry_path(m
));
902 what
= mode_to_inaccessible_node(target
.st_mode
);
904 return log_debug_errno(SYNTHETIC_ERRNO(ELOOP
),
905 "File type not supported for inaccessible mounts. Note that symlinks are not allowed");
911 r
= path_is_mount_point(mount_entry_path(m
), root_directory
, 0);
912 if (r
== -ENOENT
&& m
->ignore
)
915 return log_debug_errno(r
, "Failed to determine whether %s is already a mount point: %m", mount_entry_path(m
));
916 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. */
918 /* This isn't a mount point yet, let's make it one. */
919 what
= mount_entry_path(m
);
926 case BIND_MOUNT_RECURSIVE
: {
927 _cleanup_free_
char *chased
= NULL
;
929 /* Since mount() will always follow symlinks we chase the symlinks on our own first. Note that bind
930 * mount source paths are always relative to the host root, hence we pass NULL as root directory to
931 * chase_symlinks() here. */
933 r
= chase_symlinks(mount_entry_source(m
), NULL
, CHASE_TRAIL_SLASH
, &chased
);
934 if (r
== -ENOENT
&& m
->ignore
) {
935 log_debug_errno(r
, "Path %s does not exist, ignoring.", mount_entry_source(m
));
939 return log_debug_errno(r
, "Failed to follow symlinks on %s: %m", mount_entry_source(m
));
941 log_debug("Followed source symlinks %s → %s.", mount_entry_source(m
), chased
);
943 free_and_replace(m
->source_malloc
, chased
);
945 what
= mount_entry_source(m
);
952 return mount_tmpfs(m
);
955 what
= mount_entry_source(m
);
960 return mount_private_dev(m
);
963 return mount_bind_dev(m
);
966 return mount_sysfs(m
);
969 return mount_procfs(m
);
972 assert_not_reached("Unknown mode");
977 if (mount(what
, mount_entry_path(m
), NULL
, MS_BIND
|(rbind
? MS_REC
: 0), NULL
) < 0) {
978 bool try_again
= false;
981 if (r
== -ENOENT
&& make
) {
984 /* Hmm, either the source or the destination are missing. Let's see if we can create the destination, then try again */
986 if (stat(what
, &st
) < 0)
987 log_debug_errno(errno
, "Mount point source '%s' is not accessible: %m", what
);
991 (void) mkdir_parents(mount_entry_path(m
), 0755);
993 if (S_ISDIR(st
.st_mode
))
994 q
= mkdir(mount_entry_path(m
), 0755) < 0 ? -errno
: 0;
996 q
= touch(mount_entry_path(m
));
999 log_debug_errno(q
, "Failed to create destination mount point node '%s': %m", mount_entry_path(m
));
1006 if (mount(what
, mount_entry_path(m
), NULL
, MS_BIND
|(rbind
? MS_REC
: 0), NULL
) < 0)
1013 return log_debug_errno(r
, "Failed to mount %s to %s: %m", what
, mount_entry_path(m
));
1016 log_debug("Successfully mounted %s to %s", what
, mount_entry_path(m
));
1020 /* Change the per-mount readonly flag on an existing mount */
1021 static int remount_bind_readonly(const char *path
, unsigned long orig_flags
) {
1024 r
= mount(NULL
, path
, NULL
, MS_REMOUNT
| MS_BIND
| MS_RDONLY
| orig_flags
, NULL
);
1026 return r
< 0 ? -errno
: 0;
1029 static int make_read_only(const MountEntry
*m
, char **blacklist
, FILE *proc_self_mountinfo
) {
1030 bool submounts
= false;
1034 assert(proc_self_mountinfo
);
1036 if (mount_entry_read_only(m
)) {
1037 if (IN_SET(m
->mode
, EMPTY_DIR
, TMPFS
)) {
1038 r
= remount_bind_readonly(mount_entry_path(m
), m
->flags
);
1041 r
= bind_remount_recursive_with_mountinfo(mount_entry_path(m
), true, blacklist
, proc_self_mountinfo
);
1043 } else if (m
->mode
== PRIVATE_DEV
) {
1044 /* Set /dev readonly, but not submounts like /dev/shm. Also, we only set the per-mount read-only flag.
1045 * We can't set it on the superblock, if we are inside a user namespace and running Linux <= 4.17. */
1046 r
= remount_bind_readonly(mount_entry_path(m
), DEV_MOUNT_OPTIONS
);
1050 /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked read-only
1051 * already stays this way. This improves compatibility with container managers, where we won't attempt to undo
1052 * read-only mounts already applied. */
1054 if (r
== -ENOENT
&& m
->ignore
)
1058 return log_debug_errno(r
, "Failed to re-mount '%s'%s read-only: %m", mount_entry_path(m
),
1059 submounts
? " and its submounts" : "");
1064 static bool namespace_info_mount_apivfs(const NamespaceInfo
*ns_info
) {
1068 * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=,
1069 * since to protect the API VFS mounts, they need to be around in the
1073 return ns_info
->mount_apivfs
||
1074 ns_info
->protect_control_groups
||
1075 ns_info
->protect_kernel_tunables
;
1078 static size_t namespace_calculate_mounts(
1079 const NamespaceInfo
*ns_info
,
1080 char** read_write_paths
,
1081 char** read_only_paths
,
1082 char** inaccessible_paths
,
1083 char** empty_directories
,
1084 size_t n_bind_mounts
,
1085 size_t n_temporary_filesystems
,
1086 const char* tmp_dir
,
1087 const char* var_tmp_dir
,
1088 ProtectHome protect_home
,
1089 ProtectSystem protect_system
) {
1091 size_t protect_home_cnt
;
1092 size_t protect_system_cnt
=
1093 (protect_system
== PROTECT_SYSTEM_STRICT
?
1094 ELEMENTSOF(protect_system_strict_table
) :
1095 ((protect_system
== PROTECT_SYSTEM_FULL
) ?
1096 ELEMENTSOF(protect_system_full_table
) :
1097 ((protect_system
== PROTECT_SYSTEM_YES
) ?
1098 ELEMENTSOF(protect_system_yes_table
) : 0)));
1101 (protect_home
== PROTECT_HOME_YES
?
1102 ELEMENTSOF(protect_home_yes_table
) :
1103 ((protect_home
== PROTECT_HOME_READ_ONLY
) ?
1104 ELEMENTSOF(protect_home_read_only_table
) :
1105 ((protect_home
== PROTECT_HOME_TMPFS
) ?
1106 ELEMENTSOF(protect_home_tmpfs_table
) : 0)));
1108 return !!tmp_dir
+ !!var_tmp_dir
+
1109 strv_length(read_write_paths
) +
1110 strv_length(read_only_paths
) +
1111 strv_length(inaccessible_paths
) +
1112 strv_length(empty_directories
) +
1114 n_temporary_filesystems
+
1115 ns_info
->private_dev
+
1116 (ns_info
->protect_kernel_tunables
? ELEMENTSOF(protect_kernel_tunables_table
) : 0) +
1117 (ns_info
->protect_control_groups
? 1 : 0) +
1118 (ns_info
->protect_kernel_modules
? ELEMENTSOF(protect_kernel_modules_table
) : 0) +
1119 protect_home_cnt
+ protect_system_cnt
+
1120 (ns_info
->protect_hostname
? 2 : 0) +
1121 (namespace_info_mount_apivfs(ns_info
) ? ELEMENTSOF(apivfs_table
) : 0);
1124 static void normalize_mounts(const char *root_directory
, MountEntry
*mounts
, size_t *n_mounts
) {
1125 assert(root_directory
);
1127 assert(mounts
|| *n_mounts
== 0);
1129 typesafe_qsort(mounts
, *n_mounts
, mount_path_compare
);
1131 drop_duplicates(mounts
, n_mounts
);
1132 drop_outside_root(root_directory
, mounts
, n_mounts
);
1133 drop_inaccessible(mounts
, n_mounts
);
1134 drop_nop(mounts
, n_mounts
);
1137 int setup_namespace(
1138 const char* root_directory
,
1139 const char* root_image
,
1140 const NamespaceInfo
*ns_info
,
1141 char** read_write_paths
,
1142 char** read_only_paths
,
1143 char** inaccessible_paths
,
1144 char** empty_directories
,
1145 const BindMount
*bind_mounts
,
1146 size_t n_bind_mounts
,
1147 const TemporaryFileSystem
*temporary_filesystems
,
1148 size_t n_temporary_filesystems
,
1149 const char* tmp_dir
,
1150 const char* var_tmp_dir
,
1151 ProtectHome protect_home
,
1152 ProtectSystem protect_system
,
1153 unsigned long mount_flags
,
1154 DissectImageFlags dissect_image_flags
) {
1156 _cleanup_(loop_device_unrefp
) LoopDevice
*loop_device
= NULL
;
1157 _cleanup_(decrypted_image_unrefp
) DecryptedImage
*decrypted_image
= NULL
;
1158 _cleanup_(dissected_image_unrefp
) DissectedImage
*dissected_image
= NULL
;
1159 _cleanup_free_
void *root_hash
= NULL
;
1160 MountEntry
*m
, *mounts
= NULL
;
1161 size_t n_mounts
, root_hash_size
= 0;
1162 bool require_prefix
= false;
1168 if (mount_flags
== 0)
1169 mount_flags
= MS_SHARED
;
1172 dissect_image_flags
|= DISSECT_IMAGE_REQUIRE_ROOT
;
1174 if (protect_system
== PROTECT_SYSTEM_STRICT
&&
1175 protect_home
!= PROTECT_HOME_NO
&&
1176 strv_isempty(read_write_paths
))
1177 dissect_image_flags
|= DISSECT_IMAGE_READ_ONLY
;
1179 r
= loop_device_make_by_path(root_image
,
1180 dissect_image_flags
& DISSECT_IMAGE_READ_ONLY
? O_RDONLY
: O_RDWR
,
1183 return log_debug_errno(r
, "Failed to create loop device for root image: %m");
1185 r
= root_hash_load(root_image
, &root_hash
, &root_hash_size
);
1187 return log_debug_errno(r
, "Failed to load root hash: %m");
1189 r
= dissect_image(loop_device
->fd
, root_hash
, root_hash_size
, dissect_image_flags
, &dissected_image
);
1191 return log_debug_errno(r
, "Failed to dissect image: %m");
1193 r
= dissected_image_decrypt(dissected_image
, NULL
, root_hash
, root_hash_size
, dissect_image_flags
, &decrypted_image
);
1195 return log_debug_errno(r
, "Failed to decrypt dissected image: %m");
1199 root
= root_directory
;
1201 /* Always create the mount namespace in a temporary directory, instead of operating
1202 * directly in the root. The temporary directory prevents any mounts from being
1203 * potentially obscured my other mounts we already applied.
1204 * We use the same mount point for all images, which is safe, since they all live
1205 * in their own namespaces after all, and hence won't see each other. */
1207 root
= "/run/systemd/unit-root";
1208 (void) mkdir_label(root
, 0700);
1209 require_prefix
= true;
1212 n_mounts
= namespace_calculate_mounts(
1219 n_temporary_filesystems
,
1220 tmp_dir
, var_tmp_dir
,
1221 protect_home
, protect_system
);
1224 m
= mounts
= (MountEntry
*) alloca0(n_mounts
* sizeof(MountEntry
));
1225 r
= append_access_mounts(&m
, read_write_paths
, READWRITE
, require_prefix
);
1229 r
= append_access_mounts(&m
, read_only_paths
, READONLY
, require_prefix
);
1233 r
= append_access_mounts(&m
, inaccessible_paths
, INACCESSIBLE
, require_prefix
);
1237 r
= append_empty_dir_mounts(&m
, empty_directories
);
1241 r
= append_bind_mounts(&m
, bind_mounts
, n_bind_mounts
);
1245 r
= append_tmpfs_mounts(&m
, temporary_filesystems
, n_temporary_filesystems
);
1250 *(m
++) = (MountEntry
) {
1251 .path_const
= "/tmp",
1252 .mode
= PRIVATE_TMP
,
1253 .source_const
= tmp_dir
,
1258 *(m
++) = (MountEntry
) {
1259 .path_const
= "/var/tmp",
1260 .mode
= PRIVATE_TMP
,
1261 .source_const
= var_tmp_dir
,
1265 if (ns_info
->private_dev
) {
1266 *(m
++) = (MountEntry
) {
1267 .path_const
= "/dev",
1268 .mode
= PRIVATE_DEV
,
1272 if (ns_info
->protect_kernel_tunables
) {
1273 r
= append_static_mounts(&m
, protect_kernel_tunables_table
, ELEMENTSOF(protect_kernel_tunables_table
), ns_info
->ignore_protect_paths
);
1278 if (ns_info
->protect_kernel_modules
) {
1279 r
= append_static_mounts(&m
, protect_kernel_modules_table
, ELEMENTSOF(protect_kernel_modules_table
), ns_info
->ignore_protect_paths
);
1284 if (ns_info
->protect_control_groups
) {
1285 *(m
++) = (MountEntry
) {
1286 .path_const
= "/sys/fs/cgroup",
1291 r
= append_protect_home(&m
, protect_home
, ns_info
->ignore_protect_paths
);
1295 r
= append_protect_system(&m
, protect_system
, false);
1299 if (namespace_info_mount_apivfs(ns_info
)) {
1300 r
= append_static_mounts(&m
, apivfs_table
, ELEMENTSOF(apivfs_table
), ns_info
->ignore_protect_paths
);
1305 if (ns_info
->protect_hostname
) {
1306 *(m
++) = (MountEntry
) {
1307 .path_const
= "/proc/sys/kernel/hostname",
1310 *(m
++) = (MountEntry
) {
1311 .path_const
= "/proc/sys/kernel/domainname",
1316 assert(mounts
+ n_mounts
== m
);
1318 /* Prepend the root directory where that's necessary */
1319 r
= prefix_where_needed(mounts
, n_mounts
, root
);
1323 normalize_mounts(root
, mounts
, &n_mounts
);
1326 /* All above is just preparation, figuring out what to do. Let's now actually start doing something. */
1328 if (unshare(CLONE_NEWNS
) < 0) {
1329 r
= log_debug_errno(errno
, "Failed to unshare the mount namespace: %m");
1330 if (IN_SET(r
, -EACCES
, -EPERM
, -EOPNOTSUPP
, -ENOSYS
))
1331 /* If the kernel doesn't support namespaces, or when there's a MAC or seccomp filter in place
1332 * that doesn't allow us to create namespaces (or a missing cap), then propagate a recognizable
1333 * error back, which the caller can use to detect this case (and only this) and optionally
1334 * continue without namespacing applied. */
1340 /* Remount / as SLAVE so that nothing now mounted in the namespace
1341 * shows up in the parent */
1342 if (mount(NULL
, "/", NULL
, MS_SLAVE
|MS_REC
, NULL
) < 0) {
1343 r
= log_debug_errno(errno
, "Failed to remount '/' as SLAVE: %m");
1348 /* A root image is specified, mount it to the right place */
1349 r
= dissected_image_mount(dissected_image
, root
, UID_INVALID
, dissect_image_flags
);
1351 log_debug_errno(r
, "Failed to mount root image: %m");
1355 if (decrypted_image
) {
1356 r
= decrypted_image_relinquish(decrypted_image
);
1358 log_debug_errno(r
, "Failed to relinquish decrypted image: %m");
1363 loop_device_relinquish(loop_device
);
1365 } else if (root_directory
) {
1367 /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */
1368 r
= path_is_mount_point(root
, NULL
, AT_SYMLINK_FOLLOW
);
1370 log_debug_errno(r
, "Failed to detect that %s is a mount point or not: %m", root
);
1374 if (mount(root
, root
, NULL
, MS_BIND
|MS_REC
, NULL
) < 0) {
1375 r
= log_debug_errno(errno
, "Failed to bind mount '%s': %m", root
);
1382 /* Let's mount the main root directory to the root directory to use */
1383 if (mount("/", root
, NULL
, MS_BIND
|MS_REC
, NULL
) < 0) {
1384 r
= log_debug_errno(errno
, "Failed to bind mount '/' on '%s': %m", root
);
1389 /* Try to set up the new root directory before mounting anything else there. */
1390 if (root_image
|| root_directory
)
1391 (void) base_filesystem_create(root
, UID_INVALID
, GID_INVALID
);
1394 _cleanup_fclose_
FILE *proc_self_mountinfo
= NULL
;
1398 /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of /proc.
1399 * For example, this is the case with the option: 'InaccessiblePaths=/proc' */
1400 proc_self_mountinfo
= fopen("/proc/self/mountinfo", "re");
1401 if (!proc_self_mountinfo
) {
1402 r
= log_debug_errno(errno
, "Failed to open /proc/self/mountinfo: %m");
1406 /* First round, establish all mounts we need */
1410 for (m
= mounts
; m
< mounts
+ n_mounts
; ++m
) {
1415 r
= follow_symlink(root
, m
);
1419 /* We hit a symlinked mount point. The entry got rewritten and might point to a
1420 * very different place now. Let's normalize the changed list, and start from
1421 * the beginning. After all to mount the entry at the new location we might
1422 * need some other mounts first */
1427 r
= apply_mount(root
, m
);
1437 normalize_mounts(root
, mounts
, &n_mounts
);
1440 /* Create a blacklist we can pass to bind_mount_recursive() */
1441 blacklist
= newa(char*, n_mounts
+1);
1442 for (j
= 0; j
< n_mounts
; j
++)
1443 blacklist
[j
] = (char*) mount_entry_path(mounts
+j
);
1444 blacklist
[j
] = NULL
;
1446 /* Second round, flip the ro bits if necessary. */
1447 for (m
= mounts
; m
< mounts
+ n_mounts
; ++m
) {
1448 r
= make_read_only(m
, blacklist
, proc_self_mountinfo
);
1454 /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */
1455 r
= mount_move_root(root
);
1457 log_debug_errno(r
, "Failed to mount root with MS_MOVE: %m");
1461 /* Remount / as the desired mode. Note that this will not
1462 * reestablish propagation from our side to the host, since
1463 * what's disconnected is disconnected. */
1464 if (mount(NULL
, "/", NULL
, mount_flags
| MS_REC
, NULL
) < 0) {
1465 r
= log_debug_errno(errno
, "Failed to remount '/' with desired mount flags: %m");
1472 for (m
= mounts
; m
< mounts
+ n_mounts
; m
++)
1473 mount_entry_done(m
);
1478 void bind_mount_free_many(BindMount
*b
, size_t n
) {
1481 assert(b
|| n
== 0);
1483 for (i
= 0; i
< n
; i
++) {
1485 free(b
[i
].destination
);
1491 int bind_mount_add(BindMount
**b
, size_t *n
, const BindMount
*item
) {
1492 _cleanup_free_
char *s
= NULL
, *d
= NULL
;
1499 s
= strdup(item
->source
);
1503 d
= strdup(item
->destination
);
1507 c
= reallocarray(*b
, *n
+ 1, sizeof(BindMount
));
1513 c
[(*n
) ++] = (BindMount
) {
1514 .source
= TAKE_PTR(s
),
1515 .destination
= TAKE_PTR(d
),
1516 .read_only
= item
->read_only
,
1517 .recursive
= item
->recursive
,
1518 .ignore_enoent
= item
->ignore_enoent
,
1524 void temporary_filesystem_free_many(TemporaryFileSystem
*t
, size_t n
) {
1527 assert(t
|| n
== 0);
1529 for (i
= 0; i
< n
; i
++) {
1537 int temporary_filesystem_add(
1538 TemporaryFileSystem
**t
,
1541 const char *options
) {
1543 _cleanup_free_
char *p
= NULL
, *o
= NULL
;
1544 TemporaryFileSystem
*c
;
1554 if (!isempty(options
)) {
1555 o
= strdup(options
);
1560 c
= reallocarray(*t
, *n
+ 1, sizeof(TemporaryFileSystem
));
1566 c
[(*n
) ++] = (TemporaryFileSystem
) {
1567 .path
= TAKE_PTR(p
),
1568 .options
= TAKE_PTR(o
),
1574 static int setup_one_tmp_dir(const char *id
, const char *prefix
, char **path
) {
1575 _cleanup_free_
char *x
= NULL
;
1576 char bid
[SD_ID128_STRING_MAX
];
1584 /* We include the boot id in the directory so that after a
1585 * reboot we can easily identify obsolete directories. */
1587 r
= sd_id128_get_boot(&boot_id
);
1591 x
= strjoin(prefix
, "/systemd-private-", sd_id128_to_string(boot_id
, bid
), "-", id
, "-XXXXXX");
1595 RUN_WITH_UMASK(0077)
1599 RUN_WITH_UMASK(0000) {
1602 y
= strjoina(x
, "/tmp");
1604 if (mkdir(y
, 0777 | S_ISVTX
) < 0)
1608 *path
= TAKE_PTR(x
);
1613 int setup_tmp_dirs(const char *id
, char **tmp_dir
, char **var_tmp_dir
) {
1619 assert(var_tmp_dir
);
1621 r
= setup_one_tmp_dir(id
, "/tmp", &a
);
1625 r
= setup_one_tmp_dir(id
, "/var/tmp", &b
);
1629 t
= strjoina(a
, "/tmp");
1643 int setup_netns(int netns_storage_socket
[static 2]) {
1644 _cleanup_close_
int netns
= -1;
1647 assert(netns_storage_socket
);
1648 assert(netns_storage_socket
[0] >= 0);
1649 assert(netns_storage_socket
[1] >= 0);
1651 /* We use the passed socketpair as a storage buffer for our
1652 * namespace reference fd. Whatever process runs this first
1653 * shall create a new namespace, all others should just join
1654 * it. To serialize that we use a file lock on the socket
1657 * It's a bit crazy, but hey, works great! */
1659 if (lockf(netns_storage_socket
[0], F_LOCK
, 0) < 0)
1662 netns
= receive_one_fd(netns_storage_socket
[0], MSG_DONTWAIT
);
1663 if (netns
== -EAGAIN
) {
1664 /* Nothing stored yet, so let's create a new namespace */
1666 if (unshare(CLONE_NEWNET
) < 0) {
1673 netns
= open("/proc/self/ns/net", O_RDONLY
|O_CLOEXEC
|O_NOCTTY
);
1681 } else if (netns
< 0) {
1686 /* Yay, found something, so let's join the namespace */
1687 if (setns(netns
, CLONE_NEWNET
) < 0) {
1695 q
= send_one_fd(netns_storage_socket
[1], netns
, MSG_DONTWAIT
);
1702 (void) lockf(netns_storage_socket
[0], F_ULOCK
, 0);
1706 bool ns_type_supported(NamespaceType type
) {
1707 const char *t
, *ns_proc
;
1709 t
= namespace_type_to_string(type
);
1710 if (!t
) /* Don't know how to translate this? Then it's not supported */
1713 ns_proc
= strjoina("/proc/self/ns/", t
);
1714 return access(ns_proc
, F_OK
) == 0;
1717 static const char *const protect_home_table
[_PROTECT_HOME_MAX
] = {
1718 [PROTECT_HOME_NO
] = "no",
1719 [PROTECT_HOME_YES
] = "yes",
1720 [PROTECT_HOME_READ_ONLY
] = "read-only",
1721 [PROTECT_HOME_TMPFS
] = "tmpfs",
1724 DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_home
, ProtectHome
, PROTECT_HOME_YES
);
1726 static const char *const protect_system_table
[_PROTECT_SYSTEM_MAX
] = {
1727 [PROTECT_SYSTEM_NO
] = "no",
1728 [PROTECT_SYSTEM_YES
] = "yes",
1729 [PROTECT_SYSTEM_FULL
] = "full",
1730 [PROTECT_SYSTEM_STRICT
] = "strict",
1733 DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_system
, ProtectSystem
, PROTECT_SYSTEM_YES
);
1735 static const char* const namespace_type_table
[] = {
1736 [NAMESPACE_MOUNT
] = "mnt",
1737 [NAMESPACE_CGROUP
] = "cgroup",
1738 [NAMESPACE_UTS
] = "uts",
1739 [NAMESPACE_IPC
] = "ipc",
1740 [NAMESPACE_USER
] = "user",
1741 [NAMESPACE_PID
] = "pid",
1742 [NAMESPACE_NET
] = "net",
1745 DEFINE_STRING_TABLE_LOOKUP(namespace_type
, NamespaceType
);