summaryrefslogtreecommitdiffstats
path: root/src/core/mount-setup.c
blob: 3ce6164b06fd0f06bd467ac62a49016ab006a19a (plain)
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
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
/* SPDX-License-Identifier: LGPL-2.1+ */

#include <errno.h>
#include <ftw.h>
#include <stdlib.h>
#include <sys/mount.h>
#include <sys/statvfs.h>
#include <unistd.h>

#include "alloc-util.h"
#include "bus-util.h"
#include "cgroup-util.h"
#include "dev-setup.h"
#include "dirent-util.h"
#include "efivars.h"
#include "fd-util.h"
#include "fileio.h"
#include "fs-util.h"
#include "label.h"
#include "log.h"
#include "macro.h"
#include "missing.h"
#include "mkdir.h"
#include "mount-setup.h"
#include "mountpoint-util.h"
#include "path-util.h"
#include "set.h"
#include "smack-util.h"
#include "strv.h"
#include "user-util.h"
#include "util.h"
#include "virt.h"

typedef enum MountMode {
        MNT_NONE           = 0,
        MNT_FATAL          = 1 << 0,
        MNT_IN_CONTAINER   = 1 << 1,
        MNT_CHECK_WRITABLE = 1 << 2,
} MountMode;

typedef struct MountPoint {
        const char *what;
        const char *where;
        const char *type;
        const char *options;
        unsigned long flags;
        bool (*condition_fn)(void);
        MountMode mode;
} MountPoint;

/* The first three entries we might need before SELinux is up. The
 * fourth (securityfs) is needed by IMA to load a custom policy. The
 * other ones we can delay until SELinux and IMA are loaded. When
 * SMACK is enabled we need smackfs, too, so it's a fifth one. */
#if ENABLE_SMACK
#define N_EARLY_MOUNT 5
#else
#define N_EARLY_MOUNT 4
#endif

static const MountPoint mount_table[] = {
        { "sysfs",       "/sys",                      "sysfs",      NULL,                      MS_NOSUID|MS_NOEXEC|MS_NODEV,
          NULL,          MNT_FATAL|MNT_IN_CONTAINER },
        { "proc",        "/proc",                     "proc",       NULL,                      MS_NOSUID|MS_NOEXEC|MS_NODEV,
          NULL,          MNT_FATAL|MNT_IN_CONTAINER },
        { "devtmpfs",    "/dev",                      "devtmpfs",   "mode=755",                MS_NOSUID|MS_STRICTATIME,
          NULL,          MNT_FATAL|MNT_IN_CONTAINER },
        { "securityfs",  "/sys/kernel/security",      "securityfs", NULL,                      MS_NOSUID|MS_NOEXEC|MS_NODEV,
          NULL,          MNT_NONE                   },
#if ENABLE_SMACK
        { "smackfs",     "/sys/fs/smackfs",           "smackfs",    "smackfsdef=*",            MS_NOSUID|MS_NOEXEC|MS_NODEV,
          mac_smack_use, MNT_FATAL                  },
        { "tmpfs",       "/dev/shm",                  "tmpfs",      "mode=1777,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME,
          mac_smack_use, MNT_FATAL                  },
#endif
        { "tmpfs",       "/dev/shm",                  "tmpfs",      "mode=1777",               MS_NOSUID|MS_NODEV|MS_STRICTATIME,
          NULL,          MNT_FATAL|MNT_IN_CONTAINER },
        { "devpts",      "/dev/pts",                  "devpts",     "mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC,
          NULL,          MNT_IN_CONTAINER           },
#if ENABLE_SMACK
        { "tmpfs",       "/run",                      "tmpfs",      "mode=755,smackfsroot=*",  MS_NOSUID|MS_NODEV|MS_STRICTATIME,
          mac_smack_use, MNT_FATAL                  },
#endif
        { "tmpfs",       "/run",                      "tmpfs",      "mode=755",                MS_NOSUID|MS_NODEV|MS_STRICTATIME,
          NULL,          MNT_FATAL|MNT_IN_CONTAINER },
        { "cgroup2",     "/sys/fs/cgroup",            "cgroup2",    "nsdelegate",              MS_NOSUID|MS_NOEXEC|MS_NODEV,
          cg_is_unified_wanted, MNT_IN_CONTAINER|MNT_CHECK_WRITABLE },
        { "cgroup2",     "/sys/fs/cgroup",            "cgroup2",    NULL,                      MS_NOSUID|MS_NOEXEC|MS_NODEV,
          cg_is_unified_wanted, MNT_IN_CONTAINER|MNT_CHECK_WRITABLE },
        { "tmpfs",       "/sys/fs/cgroup",            "tmpfs",      "mode=755",                MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
          cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
        { "cgroup2",     "/sys/fs/cgroup/unified",    "cgroup2",    "nsdelegate",              MS_NOSUID|MS_NOEXEC|MS_NODEV,
          cg_is_hybrid_wanted, MNT_IN_CONTAINER|MNT_CHECK_WRITABLE },
        { "cgroup2",     "/sys/fs/cgroup/unified",    "cgroup2",    NULL,                      MS_NOSUID|MS_NOEXEC|MS_NODEV,
          cg_is_hybrid_wanted, MNT_IN_CONTAINER|MNT_CHECK_WRITABLE },
        { "cgroup",      "/sys/fs/cgroup/systemd",    "cgroup",     "none,name=systemd,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV,
          cg_is_legacy_wanted, MNT_IN_CONTAINER     },
        { "cgroup",      "/sys/fs/cgroup/systemd",    "cgroup",     "none,name=systemd",       MS_NOSUID|MS_NOEXEC|MS_NODEV,
          cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
        { "pstore",      "/sys/fs/pstore",            "pstore",     NULL,                      MS_NOSUID|MS_NOEXEC|MS_NODEV,
          NULL,          MNT_NONE                   },
#if ENABLE_EFI
        { "efivarfs",    "/sys/firmware/efi/efivars", "efivarfs",   NULL,                      MS_NOSUID|MS_NOEXEC|MS_NODEV,
          is_efi_boot,   MNT_NONE                   },
#endif
        { "bpf",         "/sys/fs/bpf",               "bpf",        "mode=700",                MS_NOSUID|MS_NOEXEC|MS_NODEV,
          NULL,          MNT_NONE,                  },
};

/* These are API file systems that might be mounted by other software,
 * we just list them here so that we know that we should ignore them */

static const char ignore_paths[] =
        /* SELinux file systems */
        "/sys/fs/selinux\0"
        /* Container bind mounts */
        "/proc/sys\0"
        "/dev/console\0"
        "/proc/kmsg\0";

bool mount_point_is_api(const char *path) {
        unsigned i;

        /* Checks if this mount point is considered "API", and hence
         * should be ignored */

        for (i = 0; i < ELEMENTSOF(mount_table); i ++)
                if (path_equal(path, mount_table[i].where))
                        return true;

        return path_startswith(path, "/sys/fs/cgroup/");
}

bool mount_point_ignore(const char *path) {
        const char *i;

        NULSTR_FOREACH(i, ignore_paths)
                if (path_equal(path, i))
                        return true;

        return false;
}

static int mount_one(const MountPoint *p, bool relabel) {
        int r, priority;

        assert(p);

        priority = (p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG;

        if (p->condition_fn && !p->condition_fn())
                return 0;

        /* Relabel first, just in case */
        if (relabel)
                (void) label_fix(p->where, LABEL_IGNORE_ENOENT|LABEL_IGNORE_EROFS);

        r = path_is_mount_point(p->where, NULL, AT_SYMLINK_FOLLOW);
        if (r < 0 && r != -ENOENT) {
                log_full_errno(priority, r, "Failed to determine whether %s is a mount point: %m", p->where);
                return (p->mode & MNT_FATAL) ? r : 0;
        }
        if (r > 0)
                return 0;

        /* Skip securityfs in a container */
        if (!(p->mode & MNT_IN_CONTAINER) && detect_container() > 0)
                return 0;

        /* The access mode here doesn't really matter too much, since
         * the mounted file system will take precedence anyway. */
        if (relabel)
                (void) mkdir_p_label(p->where, 0755);
        else
                (void) mkdir_p(p->where, 0755);

        log_debug("Mounting %s to %s of type %s with options %s.",
                  p->what,
                  p->where,
                  p->type,
                  strna(p->options));

        if (mount(p->what,
                  p->where,
                  p->type,
                  p->flags,
                  p->options) < 0) {
                log_full_errno(priority, errno, "Failed to mount %s at %s: %m", p->type, p->where);
                return (p->mode & MNT_FATAL) ? -errno : 0;
        }

        /* Relabel again, since we now mounted something fresh here */
        if (relabel)
                (void) label_fix(p->where, 0);

        if (p->mode & MNT_CHECK_WRITABLE) {
                if (access(p->where, W_OK) < 0) {
                        r = -errno;

                        (void) umount(p->where);
                        (void) rmdir(p->where);

                        log_full_errno(priority, r, "Mount point %s not writable after mounting: %m", p->where);
                        return (p->mode & MNT_FATAL) ? r : 0;
                }
        }

        return 1;
}

static int mount_points_setup(unsigned n, bool loaded_policy) {
        unsigned i;
        int r = 0;

        for (i = 0; i < n; i ++) {
                int j;

                j = mount_one(mount_table + i, loaded_policy);
                if (j != 0 && r >= 0)
                        r = j;
        }

        return r;
}

int mount_setup_early(void) {
        assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));

        /* Do a minimal mount of /proc and friends to enable the most
         * basic stuff, such as SELinux */
        return mount_points_setup(N_EARLY_MOUNT, false);
}

static const char *join_with(const char *controller) {

        static const char* const pairs[] = {
                "cpu", "cpuacct",
                "net_cls", "net_prio",
                NULL
        };

        const char *const *x, *const *y;

        assert(controller);

        /* This will lookup which controller to mount another controller with. Input is a controller name, and output
         * is the other controller name. The function works both ways: you can input one and get the other, and input
         * the other to get the one. */

        STRV_FOREACH_PAIR(x, y, pairs) {
                if (streq(controller, *x))
                        return *y;
                if (streq(controller, *y))
                        return *x;
        }

        return NULL;
}

static int symlink_controller(const char *target, const char *alias) {
        const char *a;
        int r;

        assert(target);
        assert(alias);

        a = strjoina("/sys/fs/cgroup/", alias);

        r = symlink_idempotent(target, a, false);
        if (r < 0)
                return log_error_errno(r, "Failed to create symlink %s: %m", a);

#ifdef SMACK_RUN_LABEL
        const char *p;

        p = strjoina("/sys/fs/cgroup/", target);

        r = mac_smack_copy(a, p);
        if (r < 0 && r != -EOPNOTSUPP)
                return log_error_errno(r, "Failed to copy smack label from %s to %s: %m", p, a);
#endif

        return 0;
}

int mount_cgroup_controllers(void) {
        _cleanup_set_free_free_ Set *controllers = NULL;
        int r;

        if (!cg_is_legacy_wanted())
                return 0;

        /* Mount all available cgroup controllers that are built into the kernel. */
        r = cg_kernel_controllers(&controllers);
        if (r < 0)
                return log_error_errno(r, "Failed to enumerate cgroup controllers: %m");

        for (;;) {
                _cleanup_free_ char *options = NULL, *controller = NULL, *where = NULL;
                const char *other_controller;
                MountPoint p = {
                        .what = "cgroup",
                        .type = "cgroup",
                        .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV,
                        .mode = MNT_IN_CONTAINER,
                };

                controller = set_steal_first(controllers);
                if (!controller)
                        break;

                /* Check if we shall mount this together with another controller */
                other_controller = join_with(controller);
                if (other_controller) {
                        _cleanup_free_ char *c = NULL;

                        /* Check if the other controller is actually available in the kernel too */
                        c = set_remove(controllers, other_controller);
                        if (c) {

                                /* Join the two controllers into one string, and maintain a stable ordering */
                                if (strcmp(controller, other_controller) < 0)
                                        options = strjoin(controller, ",", other_controller);
                                else
                                        options = strjoin(other_controller, ",", controller);
                                if (!options)
                                        return log_oom();
                        }
                }

                /* The simple case, where there's only one controller to mount together */
                if (!options)
                        options = TAKE_PTR(controller);

                where = strappend("/sys/fs/cgroup/", options);
                if (!where)
                        return log_oom();

                p.where = where;
                p.options = options;

                r = mount_one(&p, true);
                if (r < 0)
                        return r;

                /* Create symlinks from the individual controller names, in case we have a joined mount */
                if (controller)
                        (void) symlink_controller(options, controller);
                if (other_controller)
                        (void) symlink_controller(options, other_controller);
        }

        /* Now that we mounted everything, let's make the tmpfs the cgroup file systems are mounted into read-only. */
        (void) mount("tmpfs", "/sys/fs/cgroup", "tmpfs", MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME|MS_RDONLY, "mode=755");

        return 0;
}

#if HAVE_SELINUX || ENABLE_SMACK
static int nftw_cb(
                const char *fpath,
                const struct stat *sb,
                int tflag,
                struct FTW *ftwbuf) {

        /* No need to label /dev twice in a row... */
        if (_unlikely_(ftwbuf->level == 0))
                return FTW_CONTINUE;

        (void) label_fix(fpath, 0);

        /* /run/initramfs is static data and big, no need to
         * dynamically relabel its contents at boot... */
        if (_unlikely_(ftwbuf->level == 1 &&
                      tflag == FTW_D &&
                      streq(fpath, "/run/initramfs")))
                return FTW_SKIP_SUBTREE;

        return FTW_CONTINUE;
};

static int relabel_cgroup_filesystems(void) {
        int r;
        struct statfs st;

        r = cg_all_unified();
        if (r == 0) {
                /* Temporarily remount the root cgroup filesystem to give it a proper label. Do this
                   only when the filesystem has been already populated by a previous instance of systemd
                   running from initrd. Otherwise don't remount anything and leave the filesystem read-write
                   for the cgroup filesystems to be mounted inside. */
                if (statfs("/sys/fs/cgroup", &st) < 0)
                        return log_error_errno(errno, "Failed to determine mount flags for /sys/fs/cgroup: %m");

                if (st.f_flags & ST_RDONLY)
                        (void) mount(NULL, "/sys/fs/cgroup", NULL, MS_REMOUNT, NULL);

                (void) label_fix("/sys/fs/cgroup", 0);
                (void) nftw("/sys/fs/cgroup", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);

                if (st.f_flags & ST_RDONLY)
                        (void) mount(NULL, "/sys/fs/cgroup", NULL, MS_REMOUNT|MS_RDONLY, NULL);

        } else if (r < 0)
                return log_error_errno(r, "Failed to determine whether we are in all unified mode: %m");

        return 0;
}

static int relabel_extra(void) {
        _cleanup_closedir_ DIR *d = NULL;
        int r, c = 0;

        /* Support for relabelling additional files or directories after loading the policy. For this, code in the
         * initrd simply has to drop in *.relabel files into /run/systemd/relabel-extra.d/. We'll read all such files
         * expecting one absolute path by line and will relabel each (and everyone below that in case the path refers
         * to a directory). These drop-in files are supposed to be absolutely minimal, and do not understand comments
         * and such. After the operation succeeded the files are removed, and the drop-in directory as well, if
         * possible.
         */

        d = opendir("/run/systemd/relabel-extra.d/");
        if (!d) {
                if (errno == ENOENT)
                        return 0;

                return log_warning_errno(errno, "Failed to open /run/systemd/relabel-extra.d/, ignoring: %m");
        }

        for (;;) {
                _cleanup_fclose_ FILE *f = NULL;
                _cleanup_close_ int fd = -1;
                struct dirent *de;

                errno = 0;
                de = readdir_no_dot(d);
                if (!de) {
                        if (errno != 0)
                                return log_error_errno(errno, "Failed read directory /run/systemd/relabel-extra.d/, ignoring: %m");
                        break;
                }

                if (hidden_or_backup_file(de->d_name))
                        continue;

                if (!endswith(de->d_name, ".relabel"))
                        continue;

                if (!IN_SET(de->d_type, DT_REG, DT_UNKNOWN))
                        continue;

                fd = openat(dirfd(d), de->d_name, O_RDONLY|O_CLOEXEC|O_NONBLOCK);
                if (fd < 0) {
                        log_warning_errno(errno, "Failed to open /run/systemd/relabel-extra.d/%s, ignoring: %m", de->d_name);
                        continue;
                }

                f = fdopen(fd, "r");
                if (!f) {
                        log_warning_errno(errno, "Failed to convert file descriptor into file object, ignoring: %m");
                        continue;
                }
                TAKE_FD(fd);

                for (;;) {
                        _cleanup_free_ char *line = NULL;

                        r = read_line(f, LONG_LINE_MAX, &line);
                        if (r < 0) {
                                log_warning_errno(r, "Failed to read from /run/systemd/relabel-extra.d/%s, ignoring: %m", de->d_name);
                                break;
                        }
                        if (r == 0) /* EOF */
                                break;

                        path_simplify(line, true);

                        if (!path_is_normalized(line)) {
                                log_warning("Path to relabel is not normalized, ignoring: %s", line);
                                continue;
                        }

                        if (!path_is_absolute(line)) {
                                log_warning("Path to relabel is not absolute, ignoring: %s", line);
                                continue;
                        }

                        log_debug("Relabelling additional file/directory '%s'.", line);
                        (void) nftw(line, nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
                        c++;
                }

                if (unlinkat(dirfd(d), de->d_name, 0) < 0)
                        log_warning_errno(errno, "Failed to remove /run/systemd/relabel-extra.d/%s, ignoring: %m", de->d_name);
        }

        /* Remove when we completing things. */
        if (rmdir("/run/systemd/relabel-extra.d") < 0)
                log_warning_errno(errno, "Failed to remove /run/systemd/relabel-extra.d/ directory: %m");

        return c;
}
#endif

int mount_setup(bool loaded_policy) {
        int r = 0;

        r = mount_points_setup(ELEMENTSOF(mount_table), loaded_policy);
        if (r < 0)
                return r;

#if HAVE_SELINUX || ENABLE_SMACK
        /* Nodes in devtmpfs and /run need to be manually updated for
         * the appropriate labels, after mounting. The other virtual
         * API file systems like /sys and /proc do not need that, they
         * use the same label for all their files. */
        if (loaded_policy) {
                usec_t before_relabel, after_relabel;
                char timespan[FORMAT_TIMESPAN_MAX];
                const char *i;
                int n_extra;

                before_relabel = now(CLOCK_MONOTONIC);

                FOREACH_STRING(i, "/dev", "/dev/shm", "/run")
                        (void) nftw(i, nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);

                (void) relabel_cgroup_filesystems();

                n_extra = relabel_extra();

                after_relabel = now(CLOCK_MONOTONIC);

                log_info("Relabelled /dev, /dev/shm, /run, /sys/fs/cgroup%s in %s.",
                         n_extra > 0 ? ", additional files" : "",
                         format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel, 0));
        }
#endif

        /* Create a few default symlinks, which are normally created
         * by udevd, but some scripts might need them before we start
         * udevd. */
        dev_setup(NULL, UID_INVALID, GID_INVALID);

        /* Mark the root directory as shared in regards to mount propagation. The kernel defaults to "private", but we
         * think it makes more sense to have a default of "shared" so that nspawn and the container tools work out of
         * the box. If specific setups need other settings they can reset the propagation mode to private if
         * needed. Note that we set this only when we are invoked directly by the kernel. If we are invoked by a
         * container manager we assume the container manager knows what it is doing (for example, because it set up
         * some directories with different propagation modes). */
        if (detect_container() <= 0)
                if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
                        log_warning_errno(errno, "Failed to set up the root directory for shared mount propagation: %m");

        /* Create a few directories we always want around, Note that sd_booted() checks for /run/systemd/system, so
         * this mkdir really needs to stay for good, otherwise software that copied sd-daemon.c into their sources will
         * misdetect systemd. */
        (void) mkdir_label("/run/systemd", 0755);
        (void) mkdir_label("/run/systemd/system", 0755);

        /* Also create /run/systemd/inaccessible nodes, so that we always have something to mount inaccessible nodes
         * from. */
        (void) make_inaccessible_nodes(NULL, UID_INVALID, GID_INVALID);

        return 0;
}