1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
|
/* SPDX-License-Identifier: LGPL-2.1-or-later */
#include <sched.h>
#include <sys/mount.h>
#include <linux/fs.h>
#include "alloc-util.h"
#include "fd-util.h"
#include "format-util.h"
#include "home-util.h"
#include "homework-mount.h"
#include "homework.h"
#include "missing_mount.h"
#include "missing_syscall.h"
#include "mkdir.h"
#include "mount-util.h"
#include "namespace-util.h"
#include "path-util.h"
#include "string-util.h"
#include "user-util.h"
static const char *mount_options_for_fstype(const char *fstype) {
const char *e;
char *n;
assert(fstype);
/* Allow overriding our built-in defaults with an environment variable */
n = strjoina("SYSTEMD_HOME_MOUNT_OPTIONS_", fstype);
e = getenv(ascii_strupper(n));
if (e)
return e;
if (streq(fstype, "ext4"))
return "noquota,user_xattr";
if (streq(fstype, "xfs"))
return "noquota";
if (streq(fstype, "btrfs"))
return "noacl,compress=zstd:1";
return NULL;
}
int home_mount_node(
const char *node,
const char *fstype,
bool discard,
unsigned long flags,
const char *extra_mount_options) {
_cleanup_free_ char *joined = NULL;
const char *default_options;
int r;
assert(node);
assert(fstype);
default_options = mount_options_for_fstype(fstype);
if (default_options) {
if (!strextend_with_separator(&joined, ",", default_options))
return log_oom();
}
if (!strextend_with_separator(&joined, ",", discard ? "discard" : "nodiscard"))
return log_oom();
if (extra_mount_options) {
if (!strextend_with_separator(&joined, ",", extra_mount_options))
return log_oom();
}
r = mount_nofollow_verbose(LOG_ERR, node, HOME_RUNTIME_WORK_DIR, fstype, flags|MS_RELATIME, joined);
if (r < 0)
return r;
log_info("Mounting file system completed.");
return 0;
}
int home_unshare_and_mkdir(void) {
int r;
if (unshare(CLONE_NEWNS) < 0)
return log_error_errno(errno, "Couldn't unshare file system namespace: %m");
assert(path_startswith(HOME_RUNTIME_WORK_DIR, "/run"));
r = mount_nofollow_verbose(LOG_ERR, "/run", "/run", NULL, MS_SLAVE|MS_REC, NULL); /* Mark /run as MS_SLAVE in our new namespace */
if (r < 0)
return r;
(void) mkdir_p(HOME_RUNTIME_WORK_DIR, 0700);
return 0;
}
int home_unshare_and_mount(
const char *node,
const char *fstype,
bool discard,
unsigned long flags,
const char *extra_mount_options) {
int r;
assert(node);
assert(fstype);
r = home_unshare_and_mkdir();
if (r < 0)
return r;
r = home_mount_node(node, fstype, discard, flags, extra_mount_options);
if (r < 0)
return r;
r = mount_nofollow_verbose(LOG_ERR, NULL, HOME_RUNTIME_WORK_DIR, NULL, MS_PRIVATE, NULL);
if (r < 0) {
(void) umount_verbose(LOG_ERR, HOME_RUNTIME_WORK_DIR, UMOUNT_NOFOLLOW);
return r;
}
return 0;
}
int home_move_mount(const char *mount_suffix, const char *target) {
_cleanup_free_ char *subdir = NULL;
const char *d;
int r;
assert(target);
/* If 'mount_suffix' is set, then we'll mount a subdir of the source mount into the host. If it's
* NULL we'll move the mount itself */
if (mount_suffix) {
subdir = path_join(HOME_RUNTIME_WORK_DIR, mount_suffix);
if (!subdir)
return log_oom();
d = subdir;
} else
d = HOME_RUNTIME_WORK_DIR;
(void) mkdir_p(target, 0700);
r = mount_nofollow_verbose(LOG_ERR, d, target, NULL, MS_BIND, NULL);
if (r < 0)
return r;
r = umount_recursive(HOME_RUNTIME_WORK_DIR, 0);
if (r < 0)
return log_error_errno(r, "Failed to unmount %s: %m", HOME_RUNTIME_WORK_DIR);
log_info("Moving to final mount point %s completed.", target);
return 0;
}
static int append_identity_range(char **text, uid_t start, uid_t next_start, uid_t exclude) {
/* Creates an identity range ranging from 'start' to 'next_start-1'. Excludes the UID specified by 'exclude' if
* it is in that range. */
assert(text);
if (next_start <= start) /* Empty range? */
return 0;
if (exclude < start || exclude >= next_start) /* UID to exclude it outside of the range? */
return strextendf(text, UID_FMT " " UID_FMT " " UID_FMT "\n", start, start, next_start - start);
if (start == exclude && next_start == exclude + 1) /* The only UID in the range is the one to exclude? */
return 0;
if (exclude == start) /* UID to exclude at beginning of range? */
return strextendf(text, UID_FMT " " UID_FMT " " UID_FMT "\n", start+1, start+1, next_start - start - 1);
if (exclude == next_start - 1) /* UID to exclude at end of range? */
return strextendf(text, UID_FMT " " UID_FMT " " UID_FMT "\n", start, start, next_start - start - 1);
return strextendf(text,
UID_FMT " " UID_FMT " " UID_FMT "\n"
UID_FMT " " UID_FMT " " UID_FMT "\n",
start, start, exclude - start,
exclude + 1, exclude + 1, next_start - exclude - 1);
}
static int make_userns(uid_t stored_uid, uid_t exposed_uid) {
_cleanup_free_ char *text = NULL;
_cleanup_close_ int userns_fd = -1;
int r;
assert(uid_is_valid(stored_uid));
assert(uid_is_valid(exposed_uid));
assert_cc(HOME_UID_MIN <= HOME_UID_MAX);
assert_cc(HOME_UID_MAX < UID_NOBODY);
/* Map everything below the homed UID range to itself (except for the UID we actually care about if
* it is inside this range) */
r = append_identity_range(&text, 0, HOME_UID_MIN, stored_uid);
if (r < 0)
return log_oom();
/* Now map the UID we are doing this for to the target UID. */
r = strextendf(&text, UID_FMT " " UID_FMT " " UID_FMT "\n", stored_uid, exposed_uid, 1);
if (r < 0)
return log_oom();
/* Map everything above the homed UID range to itself (again, excluding the UID we actually care
* about if it is in that range). Also we leave "nobody" itself excluded) */
r = append_identity_range(&text, HOME_UID_MAX, UID_NOBODY, stored_uid);
if (r < 0)
return log_oom();
/* Leave everything else unmapped, starting from UID_NOBODY itself. Specifically, this means the
* whole space outside of 16bit remains unmapped */
log_debug("Creating userns with mapping:\n%s", text);
userns_fd = userns_acquire(text, text); /* same uid + gid mapping */
if (userns_fd < 0)
return log_error_errno(userns_fd, "Failed to allocate user namespace: %m");
return TAKE_FD(userns_fd);
}
int home_shift_uid(int dir_fd, const char *target, uid_t stored_uid, uid_t exposed_uid, int *ret_mount_fd) {
_cleanup_close_ int mount_fd = -1, userns_fd = -1;
int r;
assert(dir_fd >= 0);
assert(uid_is_valid(stored_uid));
assert(uid_is_valid(exposed_uid));
/* Let's try to set up a UID mapping for this directory. This is called when first creating a home
* directory or when activating it again. We do this as optimization only, to avoid having to
* recursively chown() things on each activation. If the kernel or file system doesn't support this
* scheme we'll handle this gracefully, and not do anything, so that the later recursive chown()ing
* then fixes up things for us. Note that the chown()ing is smart enough to skip things if they look
* alright already.
*
* Note that this always creates a new mount (i.e. we use OPEN_TREE_CLONE), since applying idmaps is
* not allowed once the mount is put in place. */
mount_fd = open_tree(dir_fd, "", AT_EMPTY_PATH | OPEN_TREE_CLONE | OPEN_TREE_CLOEXEC);
if (mount_fd < 0) {
if (ERRNO_IS_NOT_SUPPORTED(errno)) {
log_debug_errno(errno, "The open_tree() syscall is not supported, not setting up UID shift mount: %m");
if (ret_mount_fd)
*ret_mount_fd = -1;
return 0;
}
return log_error_errno(errno, "Failed to open tree of home directory: %m");
}
userns_fd = make_userns(stored_uid, exposed_uid);
if (userns_fd < 0)
return userns_fd;
/* Set the user namespace mapping attribute on the cloned mount point */
if (mount_setattr(mount_fd, "", AT_EMPTY_PATH,
&(struct mount_attr) {
.attr_set = MOUNT_ATTR_IDMAP,
.userns_fd = userns_fd,
}, MOUNT_ATTR_SIZE_VER0) < 0) {
if (ERRNO_IS_NOT_SUPPORTED(errno) || errno == EINVAL) { /* EINVAL is documented in mount_attr() as fs doesn't support idmapping */
log_debug_errno(errno, "UID/GID mapping for shifted mount not available, not setting it up: %m");
if (ret_mount_fd)
*ret_mount_fd = -1;
return 0;
}
return log_error_errno(errno, "Failed to apply UID/GID mapping: %m");
}
if (target)
r = move_mount(mount_fd, "", AT_FDCWD, target, MOVE_MOUNT_F_EMPTY_PATH);
else
r = move_mount(mount_fd, "", dir_fd, "", MOVE_MOUNT_F_EMPTY_PATH|MOVE_MOUNT_T_EMPTY_PATH);
if (r < 0)
return log_error_errno(errno, "Failed to apply UID/GID map: %m");
log_debug("Applied uidmap mount to %s. Mapping is " UID_FMT " → " UID_FMT ".", strna(target), stored_uid, exposed_uid);
if (ret_mount_fd)
*ret_mount_fd = TAKE_FD(mount_fd);
return 1;
}
|