src/core/namespace.c

   1 /* SPDX-License-Identifier: LGPL-2.1+ */
   2
   3 #include <errno.h>
   4 #include <linux/loop.h>
   5 #include <sched.h>
   6 #include <stdio.h>
   7 #include <sys/mount.h>
   8 #include <unistd.h>
   9 #include <linux/fs.h>
  10
  11 #include "alloc-util.h"
  12 #include "base-filesystem.h"
  13 #include "dev-setup.h"
  14 #include "fd-util.h"
  15 #include "format-util.h"
  16 #include "fs-util.h"
  17 #include "label.h"
  18 #include "loop-util.h"
  19 #include "loopback-setup.h"
  20 #include "mkdir.h"
  21 #include "mount-util.h"
  22 #include "mountpoint-util.h"
  23 #include "namespace-util.h"
  24 #include "namespace.h"
  25 #include "nulstr-util.h"
  26 #include "path-util.h"
  27 #include "selinux-util.h"
  28 #include "socket-util.h"
  29 #include "sort-util.h"
  30 #include "stat-util.h"
  31 #include "string-table.h"
  32 #include "string-util.h"
  33 #include "strv.h"
  34 #include "tmpfile-util.h"
  35 #include "umask-util.h"
  36 #include "user-util.h"
  37
  38 #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC)
  39
  40 typedef enum MountMode {
  41         /* This is ordered by priority! */
  42         INACCESSIBLE,
  43         BIND_MOUNT,
  44         BIND_MOUNT_RECURSIVE,
  45         PRIVATE_TMP,
  46         PRIVATE_DEV,
  47         BIND_DEV,
  48         EMPTY_DIR,
  49         SYSFS,
  50         PROCFS,
  51         READONLY,
  52         READWRITE,
  53         TMPFS,
  54         READWRITE_IMPLICIT, /* Should have the lowest priority. */
  55         _MOUNT_MODE_MAX,
  56 } MountMode;
  57
  58 typedef struct MountEntry {
  59         const char *path_const;   /* Memory allocated on stack or static */
  60         MountMode mode:5;
  61         bool ignore:1;            /* Ignore if path does not exist? */
  62         bool has_prefix:1;        /* Already is prefixed by the root dir? */
  63         bool read_only:1;         /* Shall this mount point be read-only? */
  64         bool nosuid:1;            /* Shall set MS_NOSUID on the mount itself */
  65         bool applied:1;           /* Already applied */
  66         char *path_malloc;        /* Use this instead of 'path_const' if we had to allocate memory */
  67         const char *source_const; /* The source path, for bind mounts */
  68         char *source_malloc;
  69         const char *options_const;/* Mount options for tmpfs */
  70         char *options_malloc;
  71         unsigned long flags;      /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */
  72         unsigned n_followed;
  73 } MountEntry;
  74
  75 /* If MountAPIVFS= is used, let's mount /sys and /proc into the it, but only as a fallback if the user hasn't mounted
  76  * something there already. These mounts are hence overridden by any other explicitly configured mounts. */
  77 static const MountEntry apivfs_table[] = {
  78         { "/proc",               PROCFS,       false },
  79         { "/dev",                BIND_DEV,     false },
  80         { "/sys",                SYSFS,        false },
  81 };
  82
  83 /* ProtectKernelTunables= option and the related filesystem APIs */
  84 static const MountEntry protect_kernel_tunables_table[] = {
  85         { "/proc/acpi",          READONLY,           true  },
  86         { "/proc/apm",           READONLY,           true  }, /* Obsolete API, there's no point in permitting access to this, ever */
  87         { "/proc/asound",        READONLY,           true  },
  88         { "/proc/bus",           READONLY,           true  },
  89         { "/proc/fs",            READONLY,           true  },
  90         { "/proc/irq",           READONLY,           true  },
  91         { "/proc/kallsyms",      INACCESSIBLE,       true  },
  92         { "/proc/kcore",         INACCESSIBLE,       true  },
  93         { "/proc/latency_stats", READONLY,           true  },
  94         { "/proc/mtrr",          READONLY,           true  },
  95         { "/proc/scsi",          READONLY,           true  },
  96         { "/proc/sys",           READONLY,           false },
  97         { "/proc/sysrq-trigger", READONLY,           true  },
  98         { "/proc/timer_stats",   READONLY,           true  },
  99         { "/sys",                READONLY,           false },
 100         { "/sys/fs/bpf",         READONLY,           true  },
 101         { "/sys/fs/cgroup",      READWRITE_IMPLICIT, false }, /* READONLY is set by ProtectControlGroups= option */
 102         { "/sys/fs/selinux",     READWRITE_IMPLICIT, true  },
 103         { "/sys/kernel/debug",   READONLY,           true  },
 104         { "/sys/kernel/tracing", READONLY,           true  },
 105 };
 106
 107 /* ProtectKernelModules= option */
 108 static const MountEntry protect_kernel_modules_table[] = {
 109 #if HAVE_SPLIT_USR
 110         { "/lib/modules",        INACCESSIBLE, true  },
 111 #endif
 112         { "/usr/lib/modules",    INACCESSIBLE, true  },
 113 };
 114
 115 /* ProtectKernelLogs= option */
 116 static const MountEntry protect_kernel_logs_table[] = {
 117         { "/proc/kmsg",          INACCESSIBLE, true },
 118         { "/dev/kmsg",           INACCESSIBLE, true },
 119 };
 120
 121 /*
 122  * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of
 123  * system should be protected by ProtectSystem=
 124  */
 125 static const MountEntry protect_home_read_only_table[] = {
 126         { "/home",               READONLY,     true  },
 127         { "/run/user",           READONLY,     true  },
 128         { "/root",               READONLY,     true  },
 129 };
 130
 131 /* ProtectHome=tmpfs table */
 132 static const MountEntry protect_home_tmpfs_table[] = {
 133         { "/home",               TMPFS,        true, .read_only = true, .options_const = "mode=0755", .flags = MS_NODEV|MS_STRICTATIME },
 134         { "/run/user",           TMPFS,        true, .read_only = true, .options_const = "mode=0755", .flags = MS_NODEV|MS_STRICTATIME },
 135         { "/root",               TMPFS,        true, .read_only = true, .options_const = "mode=0700", .flags = MS_NODEV|MS_STRICTATIME },
 136 };
 137
 138 /* ProtectHome=yes table */
 139 static const MountEntry protect_home_yes_table[] = {
 140         { "/home",               INACCESSIBLE, true  },
 141         { "/run/user",           INACCESSIBLE, true  },
 142         { "/root",               INACCESSIBLE, true  },
 143 };
 144
 145 /* ProtectSystem=yes table */
 146 static const MountEntry protect_system_yes_table[] = {
 147         { "/usr",                READONLY,     false },
 148         { "/boot",               READONLY,     true  },
 149         { "/efi",                READONLY,     true  },
 150 #if HAVE_SPLIT_USR
 151         { "/lib",                READONLY,     true  },
 152         { "/lib64",              READONLY,     true  },
 153         { "/bin",                READONLY,     true  },
 154 #  if HAVE_SPLIT_BIN
 155         { "/sbin",               READONLY,     true  },
 156 #  endif
 157 #endif
 158 };
 159
 160 /* ProtectSystem=full includes ProtectSystem=yes */
 161 static const MountEntry protect_system_full_table[] = {
 162         { "/usr",                READONLY,     false },
 163         { "/boot",               READONLY,     true  },
 164         { "/efi",                READONLY,     true  },
 165         { "/etc",                READONLY,     false },
 166 #if HAVE_SPLIT_USR
 167         { "/lib",                READONLY,     true  },
 168         { "/lib64",              READONLY,     true  },
 169         { "/bin",                READONLY,     true  },
 170 #  if HAVE_SPLIT_BIN
 171         { "/sbin",               READONLY,     true  },
 172 #  endif
 173 #endif
 174 };
 175
 176 /*
 177  * ProtectSystem=strict table. In this strict mode, we mount everything
 178  * read-only, except for /proc, /dev, /sys which are the kernel API VFS,
 179  * which are left writable, but PrivateDevices= + ProtectKernelTunables=
 180  * protect those, and these options should be fully orthogonal.
 181  * (And of course /home and friends are also left writable, as ProtectHome=
 182  * shall manage those, orthogonally).
 183  */
 184 static const MountEntry protect_system_strict_table[] = {
 185         { "/",                   READONLY,           false },
 186         { "/proc",               READWRITE_IMPLICIT, false },      /* ProtectKernelTunables= */
 187         { "/sys",                READWRITE_IMPLICIT, false },      /* ProtectKernelTunables= */
 188         { "/dev",                READWRITE_IMPLICIT, false },      /* PrivateDevices= */
 189         { "/home",               READWRITE_IMPLICIT, true  },      /* ProtectHome= */
 190         { "/run/user",           READWRITE_IMPLICIT, true  },      /* ProtectHome= */
 191         { "/root",               READWRITE_IMPLICIT, true  },      /* ProtectHome= */
 192 };
 193
 194 static const char * const mount_mode_table[_MOUNT_MODE_MAX] = {
 195         [INACCESSIBLE]         = "inaccessible",
 196         [BIND_MOUNT]           = "bind",
 197         [BIND_MOUNT_RECURSIVE] = "rbind",
 198         [PRIVATE_TMP]          = "private-tmp",
 199         [PRIVATE_DEV]          = "private-dev",
 200         [BIND_DEV]             = "bind-dev",
 201         [EMPTY_DIR]            = "empty",
 202         [SYSFS]                = "sysfs",
 203         [PROCFS]               = "procfs",
 204         [READONLY]             = "read-only",
 205         [READWRITE]            = "read-write",
 206         [TMPFS]                = "tmpfs",
 207         [READWRITE_IMPLICIT]   = "rw-implicit",
 208 };
 209
 210 DEFINE_PRIVATE_STRING_TABLE_LOOKUP_TO_STRING(mount_mode, MountMode);
 211
 212 static const char *mount_entry_path(const MountEntry *p) {
 213         assert(p);
 214
 215         /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that,
 216          * otherwise the stack/static ->path field is returned. */
 217
 218         return p->path_malloc ?: p->path_const;
 219 }
 220
 221 static bool mount_entry_read_only(const MountEntry *p) {
 222         assert(p);
 223
 224         return p->read_only || IN_SET(p->mode, READONLY, INACCESSIBLE);
 225 }
 226
 227 static const char *mount_entry_source(const MountEntry *p) {
 228         assert(p);
 229
 230         return p->source_malloc ?: p->source_const;
 231 }
 232
 233 static const char *mount_entry_options(const MountEntry *p) {
 234         assert(p);
 235
 236         return p->options_malloc ?: p->options_const;
 237 }
 238
 239 static void mount_entry_done(MountEntry *p) {
 240         assert(p);
 241
 242         p->path_malloc = mfree(p->path_malloc);
 243         p->source_malloc = mfree(p->source_malloc);
 244         p->options_malloc = mfree(p->options_malloc);
 245 }
 246
 247 static int append_access_mounts(MountEntry **p, char **strv, MountMode mode, bool forcibly_require_prefix) {
 248         char **i;
 249
 250         assert(p);
 251
 252         /* Adds a list of user-supplied READWRITE/READWRITE_IMPLICIT/READONLY/INACCESSIBLE entries */
 253
 254         STRV_FOREACH(i, strv) {
 255                 bool ignore = false, needs_prefix = false;
 256                 const char *e = *i;
 257
 258                 /* Look for any prefixes */
 259                 if (startswith(e, "-")) {
 260                         e++;
 261                         ignore = true;
 262                 }
 263                 if (startswith(e, "+")) {
 264                         e++;
 265                         needs_prefix = true;
 266                 }
 267
 268                 if (!path_is_absolute(e))
 269                         return log_debug_errno(SYNTHETIC_ERRNO(EINVAL),
 270                                                "Path is not absolute: %s", e);
 271
 272                 *((*p)++) = (MountEntry) {
 273                         .path_const = e,
 274                         .mode = mode,
 275                         .ignore = ignore,
 276                         .has_prefix = !needs_prefix && !forcibly_require_prefix,
 277                 };
 278         }
 279
 280         return 0;
 281 }
 282
 283 static int append_empty_dir_mounts(MountEntry **p, char **strv) {
 284         char **i;
 285
 286         assert(p);
 287
 288         /* Adds tmpfs mounts to provide readable but empty directories. This is primarily used to implement the
 289          * "/private/" boundary directories for DynamicUser=1. */
 290
 291         STRV_FOREACH(i, strv) {
 292
 293                 *((*p)++) = (MountEntry) {
 294                         .path_const = *i,
 295                         .mode = EMPTY_DIR,
 296                         .ignore = false,
 297                         .read_only = true,
 298                         .options_const = "mode=755",
 299                         .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
 300                 };
 301         }
 302
 303         return 0;
 304 }
 305
 306 static int append_bind_mounts(MountEntry **p, const BindMount *binds, size_t n) {
 307         size_t i;
 308
 309         assert(p);
 310
 311         for (i = 0; i < n; i++) {
 312                 const BindMount *b = binds + i;
 313
 314                 *((*p)++) = (MountEntry) {
 315                         .path_const = b->destination,
 316                         .mode = b->recursive ? BIND_MOUNT_RECURSIVE : BIND_MOUNT,
 317                         .read_only = b->read_only,
 318                         .nosuid = b->nosuid,
 319                         .source_const = b->source,
 320                         .ignore = b->ignore_enoent,
 321                 };
 322         }
 323
 324         return 0;
 325 }
 326
 327 static int append_tmpfs_mounts(MountEntry **p, const TemporaryFileSystem *tmpfs, size_t n) {
 328         size_t i;
 329         int r;
 330
 331         assert(p);
 332
 333         for (i = 0; i < n; i++) {
 334                 const TemporaryFileSystem *t = tmpfs + i;
 335                 _cleanup_free_ char *o = NULL, *str = NULL;
 336                 unsigned long flags;
 337                 bool ro = false;
 338
 339                 if (!path_is_absolute(t->path))
 340                         return log_debug_errno(SYNTHETIC_ERRNO(EINVAL),
 341                                                "Path is not absolute: %s",
 342                                                t->path);
 343
 344                 str = strjoin("mode=0755,", t->options);
 345                 if (!str)
 346                         return -ENOMEM;
 347
 348                 r = mount_option_mangle(str, MS_NODEV|MS_STRICTATIME, &flags, &o);
 349                 if (r < 0)
 350                         return log_debug_errno(r, "Failed to parse mount option '%s': %m", str);
 351
 352                 ro = flags & MS_RDONLY;
 353                 if (ro)
 354                         flags ^= MS_RDONLY;
 355
 356                 *((*p)++) = (MountEntry) {
 357                         .path_const = t->path,
 358                         .mode = TMPFS,
 359                         .read_only = ro,
 360                         .options_malloc = TAKE_PTR(o),
 361                         .flags = flags,
 362                 };
 363         }
 364
 365         return 0;
 366 }
 367
 368 static int append_static_mounts(MountEntry **p, const MountEntry *mounts, size_t n, bool ignore_protect) {
 369         size_t i;
 370
 371         assert(p);
 372         assert(mounts);
 373
 374         /* Adds a list of static pre-defined entries */
 375
 376         for (i = 0; i < n; i++)
 377                 *((*p)++) = (MountEntry) {
 378                         .path_const = mount_entry_path(mounts+i),
 379                         .mode = mounts[i].mode,
 380                         .ignore = mounts[i].ignore || ignore_protect,
 381                 };
 382
 383         return 0;
 384 }
 385
 386 static int append_protect_home(MountEntry **p, ProtectHome protect_home, bool ignore_protect) {
 387         assert(p);
 388
 389         switch (protect_home) {
 390
 391         case PROTECT_HOME_NO:
 392                 return 0;
 393
 394         case PROTECT_HOME_READ_ONLY:
 395                 return append_static_mounts(p, protect_home_read_only_table, ELEMENTSOF(protect_home_read_only_table), ignore_protect);
 396
 397         case PROTECT_HOME_TMPFS:
 398                 return append_static_mounts(p, protect_home_tmpfs_table, ELEMENTSOF(protect_home_tmpfs_table), ignore_protect);
 399
 400         case PROTECT_HOME_YES:
 401                 return append_static_mounts(p, protect_home_yes_table, ELEMENTSOF(protect_home_yes_table), ignore_protect);
 402
 403         default:
 404                 assert_not_reached("Unexpected ProtectHome= value");
 405         }
 406 }
 407
 408 static int append_protect_system(MountEntry **p, ProtectSystem protect_system, bool ignore_protect) {
 409         assert(p);
 410
 411         switch (protect_system) {
 412
 413         case PROTECT_SYSTEM_NO:
 414                 return 0;
 415
 416         case PROTECT_SYSTEM_STRICT:
 417                 return append_static_mounts(p, protect_system_strict_table, ELEMENTSOF(protect_system_strict_table), ignore_protect);
 418
 419         case PROTECT_SYSTEM_YES:
 420                 return append_static_mounts(p, protect_system_yes_table, ELEMENTSOF(protect_system_yes_table), ignore_protect);
 421
 422         case PROTECT_SYSTEM_FULL:
 423                 return append_static_mounts(p, protect_system_full_table, ELEMENTSOF(protect_system_full_table), ignore_protect);
 424
 425         default:
 426                 assert_not_reached("Unexpected ProtectSystem= value");
 427         }
 428 }
 429
 430 static int mount_path_compare(const MountEntry *a, const MountEntry *b) {
 431         int d;
 432
 433         /* If the paths are not equal, then order prefixes first */
 434         d = path_compare(mount_entry_path(a), mount_entry_path(b));
 435         if (d != 0)
 436                 return d;
 437
 438         /* If the paths are equal, check the mode */
 439         return CMP((int) a->mode, (int) b->mode);
 440 }
 441
 442 static int prefix_where_needed(MountEntry *m, size_t n, const char *root_directory) {
 443         size_t i;
 444
 445         /* Prefixes all paths in the bind mount table with the root directory if the entry needs that. */
 446
 447         for (i = 0; i < n; i++) {
 448                 char *s;
 449
 450                 if (m[i].has_prefix)
 451                         continue;
 452
 453                 s = path_join(root_directory, mount_entry_path(m+i));
 454                 if (!s)
 455                         return -ENOMEM;
 456
 457                 free_and_replace(m[i].path_malloc, s);
 458                 m[i].has_prefix = true;
 459         }
 460
 461         return 0;
 462 }
 463
 464 static void drop_duplicates(MountEntry *m, size_t *n) {
 465         MountEntry *f, *t, *previous;
 466
 467         assert(m);
 468         assert(n);
 469
 470         /* Drops duplicate entries. Expects that the array is properly ordered already. */
 471
 472         for (f = m, t = m, previous = NULL; f < m + *n; f++) {
 473
 474                 /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare()
 475                  * above. Note that we only drop duplicates that haven't been mounted yet. */
 476                 if (previous &&
 477                     path_equal(mount_entry_path(f), mount_entry_path(previous)) &&
 478                     !f->applied && !previous->applied) {
 479                         log_debug("%s (%s) is duplicate.", mount_entry_path(f), mount_mode_to_string(f->mode));
 480                         previous->read_only = previous->read_only || mount_entry_read_only(f); /* Propagate the read-only flag to the remaining entry */
 481                         mount_entry_done(f);
 482                         continue;
 483                 }
 484
 485                 *t = *f;
 486                 previous = t;
 487                 t++;
 488         }
 489
 490         *n = t - m;
 491 }
 492
 493 static void drop_inaccessible(MountEntry *m, size_t *n) {
 494         MountEntry *f, *t;
 495         const char *clear = NULL;
 496
 497         assert(m);
 498         assert(n);
 499
 500         /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly
 501          * ordered already. */
 502
 503         for (f = m, t = m; f < m + *n; f++) {
 504
 505                 /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop
 506                  * it, as inaccessible paths really should drop the entire subtree. */
 507                 if (clear && path_startswith(mount_entry_path(f), clear)) {
 508                         log_debug("%s is masked by %s.", mount_entry_path(f), clear);
 509                         mount_entry_done(f);
 510                         continue;
 511                 }
 512
 513                 clear = f->mode == INACCESSIBLE ? mount_entry_path(f) : NULL;
 514
 515                 *t = *f;
 516                 t++;
 517         }
 518
 519         *n = t - m;
 520 }
 521
 522 static void drop_nop(MountEntry *m, size_t *n) {
 523         MountEntry *f, *t;
 524
 525         assert(m);
 526         assert(n);
 527
 528         /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the
 529          * list is ordered by prefixes. */
 530
 531         for (f = m, t = m; f < m + *n; f++) {
 532
 533                 /* Only suppress such subtrees for READONLY, READWRITE and READWRITE_IMPLICIT entries */
 534                 if (IN_SET(f->mode, READONLY, READWRITE, READWRITE_IMPLICIT)) {
 535                         MountEntry *p;
 536                         bool found = false;
 537
 538                         /* Now let's find the first parent of the entry we are looking at. */
 539                         for (p = t-1; p >= m; p--) {
 540                                 if (path_startswith(mount_entry_path(f), mount_entry_path(p))) {
 541                                         found = true;
 542                                         break;
 543                                 }
 544                         }
 545
 546                         /* We found it, let's see if it's the same mode, if so, we can drop this entry */
 547                         if (found && p->mode == f->mode) {
 548                                 log_debug("%s (%s) is made redundant by %s (%s)",
 549                                           mount_entry_path(f), mount_mode_to_string(f->mode),
 550                                           mount_entry_path(p), mount_mode_to_string(p->mode));
 551                                 mount_entry_done(f);
 552                                 continue;
 553                         }
 554                 }
 555
 556                 *t = *f;
 557                 t++;
 558         }
 559
 560         *n = t - m;
 561 }
 562
 563 static void drop_outside_root(const char *root_directory, MountEntry *m, size_t *n) {
 564         MountEntry *f, *t;
 565
 566         assert(m);
 567         assert(n);
 568
 569         /* Nothing to do */
 570         if (!root_directory)
 571                 return;
 572
 573         /* Drops all mounts that are outside of the root directory. */
 574
 575         for (f = m, t = m; f < m + *n; f++) {
 576
 577                 if (!path_startswith(mount_entry_path(f), root_directory)) {
 578                         log_debug("%s is outside of root directory.", mount_entry_path(f));
 579                         mount_entry_done(f);
 580                         continue;
 581                 }
 582
 583                 *t = *f;
 584                 t++;
 585         }
 586
 587         *n = t - m;
 588 }
 589
 590 static int clone_device_node(
 591                 const char *d,
 592                 const char *temporary_mount,
 593                 bool *make_devnode) {
 594
 595         _cleanup_free_ char *sl = NULL;
 596         const char *dn, *bn, *t;
 597         struct stat st;
 598         int r;
 599
 600         if (stat(d, &st) < 0) {
 601                 if (errno == ENOENT) {
 602                         log_debug_errno(errno, "Device node '%s' to clone does not exist, ignoring.", d);
 603                         return -ENXIO;
 604                 }
 605
 606                 return log_debug_errno(errno, "Failed to stat() device node '%s' to clone, ignoring: %m", d);
 607         }
 608
 609         if (!S_ISBLK(st.st_mode) &&
 610             !S_ISCHR(st.st_mode))
 611                 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL),
 612                                        "Device node '%s' to clone is not a device node, ignoring.",
 613                                        d);
 614
 615         dn = strjoina(temporary_mount, d);
 616
 617         /* First, try to create device node properly */
 618         if (*make_devnode) {
 619                 mac_selinux_create_file_prepare(d, st.st_mode);
 620                 r = mknod(dn, st.st_mode, st.st_rdev);
 621                 mac_selinux_create_file_clear();
 622                 if (r >= 0)
 623                         goto add_symlink;
 624                 if (errno != EPERM)
 625                         return log_debug_errno(errno, "mknod failed for %s: %m", d);
 626
 627                 /* This didn't work, let's not try this again for the next iterations. */
 628                 *make_devnode = false;
 629         }
 630
 631         /* We're about to fallback to bind-mounting the device
 632          * node. So create a dummy bind-mount target. */
 633         mac_selinux_create_file_prepare(d, 0);
 634         r = mknod(dn, S_IFREG, 0);
 635         mac_selinux_create_file_clear();
 636         if (r < 0 && errno != EEXIST)
 637                 return log_debug_errno(errno, "mknod() fallback failed for '%s': %m", d);
 638
 639         /* Fallback to bind-mounting:
 640          * The assumption here is that all used device nodes carry standard
 641          * properties. Specifically, the devices nodes we bind-mount should
 642          * either be owned by root:root or root:tty (e.g. /dev/tty, /dev/ptmx)
 643          * and should not carry ACLs. */
 644         if (mount(d, dn, NULL, MS_BIND, NULL) < 0)
 645                 return log_debug_errno(errno, "Bind mounting failed for '%s': %m", d);
 646
 647 add_symlink:
 648         bn = path_startswith(d, "/dev/");
 649         if (!bn)
 650                 return 0;
 651
 652         /* Create symlinks like /dev/char/1:9 → ../urandom */
 653         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)
 654                 return log_oom();
 655
 656         (void) mkdir_parents(sl, 0755);
 657
 658         t = strjoina("../", bn);
 659
 660         if (symlink(t, sl) < 0)
 661                 log_debug_errno(errno, "Failed to symlink '%s' to '%s', ignoring: %m", t, sl);
 662
 663         return 0;
 664 }
 665
 666 static int mount_private_dev(MountEntry *m) {
 667         static const char devnodes[] =
 668                 "/dev/null\0"
 669                 "/dev/zero\0"
 670                 "/dev/full\0"
 671                 "/dev/random\0"
 672                 "/dev/urandom\0"
 673                 "/dev/tty\0";
 674
 675         char temporary_mount[] = "/tmp/namespace-dev-XXXXXX";
 676         const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL;
 677         bool can_mknod = true;
 678         _cleanup_umask_ mode_t u;
 679         int r;
 680
 681         assert(m);
 682
 683         u = umask(0000);
 684
 685         if (!mkdtemp(temporary_mount))
 686                 return log_debug_errno(errno, "Failed to create temporary directory '%s': %m", temporary_mount);
 687
 688         dev = strjoina(temporary_mount, "/dev");
 689         (void) mkdir(dev, 0755);
 690         if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755") < 0) {
 691                 r = log_debug_errno(errno, "Failed to mount tmpfs on '%s': %m", dev);
 692                 goto fail;
 693         }
 694
 695         devpts = strjoina(temporary_mount, "/dev/pts");
 696         (void) mkdir(devpts, 0755);
 697         if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) {
 698                 r = log_debug_errno(errno, "Failed to bind mount /dev/pts on '%s': %m", devpts);
 699                 goto fail;
 700         }
 701
 702         /* /dev/ptmx can either be a device node or a symlink to /dev/pts/ptmx.
 703          * When /dev/ptmx a device node, /dev/pts/ptmx has 000 permissions making it inaccessible.
 704          * Thus, in that case make a clone.
 705          * In nspawn and other containers it will be a symlink, in that case make it a symlink. */
 706         r = is_symlink("/dev/ptmx");
 707         if (r < 0) {
 708                 log_debug_errno(r, "Failed to detect whether /dev/ptmx is a symlink or not: %m");
 709                 goto fail;
 710         } else if (r > 0) {
 711                 devptmx = strjoina(temporary_mount, "/dev/ptmx");
 712                 if (symlink("pts/ptmx", devptmx) < 0) {
 713                         r = log_debug_errno(errno, "Failed to create a symlink '%s' to pts/ptmx: %m", devptmx);
 714                         goto fail;
 715                 }
 716         } else {
 717                 r = clone_device_node("/dev/ptmx", temporary_mount, &can_mknod);
 718                 if (r < 0)
 719                         goto fail;
 720         }
 721
 722         devshm = strjoina(temporary_mount, "/dev/shm");
 723         (void) mkdir(devshm, 0755);
 724         r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL);
 725         if (r < 0) {
 726                 r = log_debug_errno(errno, "Failed to bind mount /dev/shm on '%s': %m", devshm);
 727                 goto fail;
 728         }
 729
 730         devmqueue = strjoina(temporary_mount, "/dev/mqueue");
 731         (void) mkdir(devmqueue, 0755);
 732         if (mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL) < 0)
 733                 log_debug_errno(errno, "Failed to bind mount /dev/mqueue on '%s', ignoring: %m", devmqueue);
 734
 735         devhugepages = strjoina(temporary_mount, "/dev/hugepages");
 736         (void) mkdir(devhugepages, 0755);
 737         if (mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL) < 0)
 738                 log_debug_errno(errno, "Failed to bind mount /dev/hugepages on '%s', ignoring: %m", devhugepages);
 739
 740         devlog = strjoina(temporary_mount, "/dev/log");
 741         if (symlink("/run/systemd/journal/dev-log", devlog) < 0)
 742                 log_debug_errno(errno, "Failed to create a symlink '%s' to /run/systemd/journal/dev-log, ignoring: %m", devlog);
 743
 744         NULSTR_FOREACH(d, devnodes) {
 745                 r = clone_device_node(d, temporary_mount, &can_mknod);
 746                 /* ENXIO means the the *source* is not a device file, skip creation in that case */
 747                 if (r < 0 && r != -ENXIO)
 748                         goto fail;
 749         }
 750
 751         r = dev_setup(temporary_mount, UID_INVALID, GID_INVALID);
 752         if (r < 0)
 753                 log_debug_errno(r, "Failed to setup basic device tree at '%s', ignoring: %m", temporary_mount);
 754
 755         /* Create the /dev directory if missing. It is more likely to be
 756          * missing when the service is started with RootDirectory. This is
 757          * consistent with mount units creating the mount points when missing.
 758          */
 759         (void) mkdir_p_label(mount_entry_path(m), 0755);
 760
 761         /* Unmount everything in old /dev */
 762         r = umount_recursive(mount_entry_path(m), 0);
 763         if (r < 0)
 764                 log_debug_errno(r, "Failed to unmount directories below '%s', ignoring: %m", mount_entry_path(m));
 765
 766         if (mount(dev, mount_entry_path(m), NULL, MS_MOVE, NULL) < 0) {
 767                 r = log_debug_errno(errno, "Failed to move mount point '%s' to '%s': %m", dev, mount_entry_path(m));
 768                 goto fail;
 769         }
 770
 771         (void) rmdir(dev);
 772         (void) rmdir(temporary_mount);
 773
 774         return 0;
 775
 776 fail:
 777         if (devpts)
 778                 (void) umount(devpts);
 779
 780         if (devshm)
 781                 (void) umount(devshm);
 782
 783         if (devhugepages)
 784                 (void) umount(devhugepages);
 785
 786         if (devmqueue)
 787                 (void) umount(devmqueue);
 788
 789         (void) umount(dev);
 790         (void) rmdir(dev);
 791         (void) rmdir(temporary_mount);
 792
 793         return r;
 794 }
 795
 796 static int mount_bind_dev(const MountEntry *m) {
 797         int r;
 798
 799         assert(m);
 800
 801         /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's
 802          * /dev. This is only used when RootDirectory= is set. */
 803
 804         (void) mkdir_p_label(mount_entry_path(m), 0755);
 805
 806         r = path_is_mount_point(mount_entry_path(m), NULL, 0);
 807         if (r < 0)
 808                 return log_debug_errno(r, "Unable to determine whether /dev is already mounted: %m");
 809         if (r > 0) /* make this a NOP if /dev is already a mount point */
 810                 return 0;
 811
 812         if (mount("/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0)
 813                 return log_debug_errno(errno, "Failed to bind mount %s: %m", mount_entry_path(m));
 814
 815         return 1;
 816 }
 817
 818 static int mount_sysfs(const MountEntry *m) {
 819         int r;
 820
 821         assert(m);
 822
 823         (void) mkdir_p_label(mount_entry_path(m), 0755);
 824
 825         r = path_is_mount_point(mount_entry_path(m), NULL, 0);
 826         if (r < 0)
 827                 return log_debug_errno(r, "Unable to determine whether /sys is already mounted: %m");
 828         if (r > 0) /* make this a NOP if /sys is already a mount point */
 829                 return 0;
 830
 831         /* Bind mount the host's version so that we get all child mounts of it, too. */
 832         if (mount("/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0)
 833                 return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m));
 834
 835         return 1;
 836 }
 837
 838 static int mount_procfs(const MountEntry *m) {
 839         int r;
 840
 841         assert(m);
 842
 843         (void) mkdir_p_label(mount_entry_path(m), 0755);
 844
 845         r = path_is_mount_point(mount_entry_path(m), NULL, 0);
 846         if (r < 0)
 847                 return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m");
 848         if (r > 0) /* make this a NOP if /proc is already a mount point */
 849                 return 0;
 850
 851         /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */
 852         if (mount("proc", mount_entry_path(m), "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL) < 0)
 853                 return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m));
 854
 855         return 1;
 856 }
 857
 858 static int mount_tmpfs(const MountEntry *m) {
 859         assert(m);
 860
 861         /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */
 862
 863         (void) mkdir_p_label(mount_entry_path(m), 0755);
 864         (void) umount_recursive(mount_entry_path(m), 0);
 865
 866         if (mount("tmpfs", mount_entry_path(m), "tmpfs", m->flags, mount_entry_options(m)) < 0)
 867                 return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m));
 868
 869         return 1;
 870 }
 871
 872 static int follow_symlink(
 873                 const char *root_directory,
 874                 MountEntry *m) {
 875
 876         _cleanup_free_ char *target = NULL;
 877         int r;
 878
 879         /* Let's chase symlinks, but only one step at a time. That's because depending where the symlink points we
 880          * might need to change the order in which we mount stuff. Hence: let's normalize piecemeal, and do one step at
 881          * a time by specifying CHASE_STEP. This function returns 0 if we resolved one step, and > 0 if we reached the
 882          * end and already have a fully normalized name. */
 883
 884         r = chase_symlinks(mount_entry_path(m), root_directory, CHASE_STEP|CHASE_NONEXISTENT, &target, NULL);
 885         if (r < 0)
 886                 return log_debug_errno(r, "Failed to chase symlinks '%s': %m", mount_entry_path(m));
 887         if (r > 0) /* Reached the end, nothing more to resolve */
 888                 return 1;
 889
 890         if (m->n_followed >= CHASE_SYMLINKS_MAX) /* put a boundary on things */
 891                 return log_debug_errno(SYNTHETIC_ERRNO(ELOOP),
 892                                        "Symlink loop on '%s'.",
 893                                        mount_entry_path(m));
 894
 895         log_debug("Followed mount entry path symlink %s → %s.", mount_entry_path(m), target);
 896
 897         free_and_replace(m->path_malloc, target);
 898         m->has_prefix = true;
 899
 900         m->n_followed ++;
 901
 902         return 0;
 903 }
 904
 905 static int apply_mount(
 906                 const char *root_directory,
 907                 MountEntry *m) {
 908
 909         _cleanup_free_ char *inaccessible = NULL;
 910         bool rbind = true, make = false;
 911         const char *what;
 912         int r;
 913
 914         assert(m);
 915
 916         log_debug("Applying namespace mount on %s", mount_entry_path(m));
 917
 918         switch (m->mode) {
 919
 920         case INACCESSIBLE: {
 921                 _cleanup_free_ char *tmp = NULL;
 922                 const char *runtime_dir;
 923                 struct stat target;
 924
 925                 /* First, get rid of everything that is below if there
 926                  * is anything... Then, overmount it with an
 927                  * inaccessible path. */
 928                 (void) umount_recursive(mount_entry_path(m), 0);
 929
 930                 if (lstat(mount_entry_path(m), &target) < 0) {
 931                         if (errno == ENOENT && m->ignore)
 932                                 return 0;
 933
 934                         return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", mount_entry_path(m));
 935                 }
 936
 937                 if (geteuid() == 0)
 938                         runtime_dir = "/run/systemd";
 939                 else {
 940                         if (asprintf(&tmp, "/run/user/"UID_FMT, geteuid()) < 0)
 941                                 log_oom();
 942
 943                         runtime_dir = tmp;
 944                 }
 945
 946                 r = mode_to_inaccessible_node(runtime_dir, target.st_mode, &inaccessible);
 947                 if (r < 0)
 948                         return log_debug_errno(SYNTHETIC_ERRNO(ELOOP),
 949                                                "File type not supported for inaccessible mounts. Note that symlinks are not allowed");
 950                 what = inaccessible;
 951                 break;
 952         }
 953
 954         case READONLY:
 955         case READWRITE:
 956         case READWRITE_IMPLICIT:
 957                 r = path_is_mount_point(mount_entry_path(m), root_directory, 0);
 958                 if (r == -ENOENT && m->ignore)
 959                         return 0;
 960                 if (r < 0)
 961                         return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", mount_entry_path(m));
 962                 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. */
 963                         return 0;
 964                 /* This isn't a mount point yet, let's make it one. */
 965                 what = mount_entry_path(m);
 966                 break;
 967
 968         case BIND_MOUNT:
 969                 rbind = false;
 970
 971                 _fallthrough_;
 972         case BIND_MOUNT_RECURSIVE: {
 973                 _cleanup_free_ char *chased = NULL;
 974
 975                 /* Since mount() will always follow symlinks we chase the symlinks on our own first. Note that bind
 976                  * mount source paths are always relative to the host root, hence we pass NULL as root directory to
 977                  * chase_symlinks() here. */
 978
 979                 r = chase_symlinks(mount_entry_source(m), NULL, CHASE_TRAIL_SLASH, &chased, NULL);
 980                 if (r == -ENOENT && m->ignore) {
 981                         log_debug_errno(r, "Path %s does not exist, ignoring.", mount_entry_source(m));
 982                         return 0;
 983                 }
 984                 if (r < 0)
 985                         return log_debug_errno(r, "Failed to follow symlinks on %s: %m", mount_entry_source(m));
 986
 987                 log_debug("Followed source symlinks %s → %s.", mount_entry_source(m), chased);
 988
 989                 free_and_replace(m->source_malloc, chased);
 990
 991                 what = mount_entry_source(m);
 992                 make = true;
 993                 break;
 994         }
 995
 996         case EMPTY_DIR:
 997         case TMPFS:
 998                 return mount_tmpfs(m);
 999
1000         case PRIVATE_TMP:
1001                 what = mount_entry_source(m);
1002                 make = true;
1003                 break;
1004
1005         case PRIVATE_DEV:
1006                 return mount_private_dev(m);
1007
1008         case BIND_DEV:
1009                 return mount_bind_dev(m);
1010
1011         case SYSFS:
1012                 return mount_sysfs(m);
1013
1014         case PROCFS:
1015                 return mount_procfs(m);
1016
1017         default:
1018                 assert_not_reached("Unknown mode");
1019         }
1020
1021         assert(what);
1022
1023         if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) {
1024                 bool try_again = false;
1025                 r = -errno;
1026
1027                 if (r == -ENOENT && make) {
1028                         struct stat st;
1029
1030                         /* Hmm, either the source or the destination are missing. Let's see if we can create the destination, then try again */
1031
1032                         if (stat(what, &st) < 0)
1033                                 log_debug_errno(errno, "Mount point source '%s' is not accessible: %m", what);
1034                         else {
1035                                 int q;
1036
1037                                 (void) mkdir_parents(mount_entry_path(m), 0755);
1038
1039                                 if (S_ISDIR(st.st_mode))
1040                                         q = mkdir(mount_entry_path(m), 0755) < 0 ? -errno : 0;
1041                                 else
1042                                         q = touch(mount_entry_path(m));
1043
1044                                 if (q < 0)
1045                                         log_debug_errno(q, "Failed to create destination mount point node '%s': %m", mount_entry_path(m));
1046                                 else
1047                                         try_again = true;
1048                         }
1049                 }
1050
1051                 if (try_again) {
1052                         if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0)
1053                                 r = -errno;
1054                         else
1055                                 r = 0;
1056                 }
1057
1058                 if (r < 0)
1059                         return log_debug_errno(r, "Failed to mount %s to %s: %m", what, mount_entry_path(m));
1060         }
1061
1062         log_debug("Successfully mounted %s to %s", what, mount_entry_path(m));
1063         return 0;
1064 }
1065
1066 static int make_read_only(const MountEntry *m, char **blacklist, FILE *proc_self_mountinfo) {
1067         unsigned long new_flags = 0, flags_mask = 0;
1068         bool submounts = false;
1069         int r = 0;
1070
1071         assert(m);
1072         assert(proc_self_mountinfo);
1073
1074         if (mount_entry_read_only(m) || m->mode == PRIVATE_DEV) {
1075                 new_flags |= MS_RDONLY;
1076                 flags_mask |= MS_RDONLY;
1077         }
1078
1079         if (m->nosuid) {
1080                 new_flags |= MS_NOSUID;
1081                 flags_mask |= MS_NOSUID;
1082         }
1083
1084         if (flags_mask == 0) /* No Change? */
1085                 return 0;
1086
1087         /* We generally apply these changes recursively, except for /dev, and the cases we know there's
1088          * nothing further down.  Set /dev readonly, but not submounts like /dev/shm. Also, we only set the
1089          * per-mount read-only flag.  We can't set it on the superblock, if we are inside a user namespace
1090          * and running Linux <= 4.17. */
1091         submounts =
1092                 mount_entry_read_only(m) &&
1093                 !IN_SET(m->mode, EMPTY_DIR, TMPFS);
1094         if (submounts)
1095                 r = bind_remount_recursive_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, blacklist, proc_self_mountinfo);
1096         else
1097                 r = bind_remount_one_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, proc_self_mountinfo);
1098
1099         /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked
1100          * read-only already stays this way. This improves compatibility with container managers, where we
1101          * won't attempt to undo read-only mounts already applied. */
1102
1103         if (r == -ENOENT && m->ignore)
1104                 return 0;
1105         if (r < 0)
1106                 return log_debug_errno(r, "Failed to re-mount '%s'%s: %m", mount_entry_path(m),
1107                                        submounts ? " and its submounts" : "");
1108         return 0;
1109 }
1110
1111 static bool namespace_info_mount_apivfs(const NamespaceInfo *ns_info) {
1112         assert(ns_info);
1113
1114         /*
1115          * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=,
1116          * since to protect the API VFS mounts, they need to be around in the
1117          * first place...
1118          */
1119
1120         return ns_info->mount_apivfs ||
1121                 ns_info->protect_control_groups ||
1122                 ns_info->protect_kernel_tunables;
1123 }
1124
1125 static size_t namespace_calculate_mounts(
1126                 const NamespaceInfo *ns_info,
1127                 char** read_write_paths,
1128                 char** read_only_paths,
1129                 char** inaccessible_paths,
1130                 char** empty_directories,
1131                 size_t n_bind_mounts,
1132                 size_t n_temporary_filesystems,
1133                 const char* tmp_dir,
1134                 const char* var_tmp_dir,
1135                 ProtectHome protect_home,
1136                 ProtectSystem protect_system) {
1137
1138         size_t protect_home_cnt;
1139         size_t protect_system_cnt =
1140                 (protect_system == PROTECT_SYSTEM_STRICT ?
1141                  ELEMENTSOF(protect_system_strict_table) :
1142                  ((protect_system == PROTECT_SYSTEM_FULL) ?
1143                   ELEMENTSOF(protect_system_full_table) :
1144                   ((protect_system == PROTECT_SYSTEM_YES) ?
1145                    ELEMENTSOF(protect_system_yes_table) : 0)));
1146
1147         protect_home_cnt =
1148                 (protect_home == PROTECT_HOME_YES ?
1149                  ELEMENTSOF(protect_home_yes_table) :
1150                  ((protect_home == PROTECT_HOME_READ_ONLY) ?
1151                   ELEMENTSOF(protect_home_read_only_table) :
1152                   ((protect_home == PROTECT_HOME_TMPFS) ?
1153                    ELEMENTSOF(protect_home_tmpfs_table) : 0)));
1154
1155         return !!tmp_dir + !!var_tmp_dir +
1156                 strv_length(read_write_paths) +
1157                 strv_length(read_only_paths) +
1158                 strv_length(inaccessible_paths) +
1159                 strv_length(empty_directories) +
1160                 n_bind_mounts +
1161                 n_temporary_filesystems +
1162                 ns_info->private_dev +
1163                 (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) +
1164                 (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) +
1165                 (ns_info->protect_kernel_logs ? ELEMENTSOF(protect_kernel_logs_table) : 0) +
1166                 (ns_info->protect_control_groups ? 1 : 0) +
1167                 protect_home_cnt + protect_system_cnt +
1168                 (ns_info->protect_hostname ? 2 : 0) +
1169                 (namespace_info_mount_apivfs(ns_info) ? ELEMENTSOF(apivfs_table) : 0);
1170 }
1171
1172 static void normalize_mounts(const char *root_directory, MountEntry *mounts, size_t *n_mounts) {
1173         assert(root_directory);
1174         assert(n_mounts);
1175         assert(mounts || *n_mounts == 0);
1176
1177         typesafe_qsort(mounts, *n_mounts, mount_path_compare);
1178
1179         drop_duplicates(mounts, n_mounts);
1180         drop_outside_root(root_directory, mounts, n_mounts);
1181         drop_inaccessible(mounts, n_mounts);
1182         drop_nop(mounts, n_mounts);
1183 }
1184
1185 static bool root_read_only(
1186                 char **read_only_paths,
1187                 ProtectSystem protect_system) {
1188
1189         /* Determine whether the root directory is going to be read-only given the configured settings. */
1190
1191         if (protect_system == PROTECT_SYSTEM_STRICT)
1192                 return true;
1193
1194         if (path_strv_contains(read_only_paths, "/"))
1195                 return true;
1196
1197         return false;
1198 }
1199
1200 static bool home_read_only(
1201                 char** read_only_paths,
1202                 char** inaccessible_paths,
1203                 char** empty_directories,
1204                 const BindMount *bind_mounts,
1205                 size_t n_bind_mounts,
1206                 const TemporaryFileSystem *temporary_filesystems,
1207                 size_t n_temporary_filesystems,
1208                 ProtectHome protect_home) {
1209
1210         size_t i;
1211
1212         /* Determine whether the /home directory is going to be read-only given the configured settings. Yes,
1213          * this is a bit sloppy, since we don't bother checking for cases where / is affected by multiple
1214          * settings. */
1215
1216         if (protect_home != PROTECT_HOME_NO)
1217                 return true;
1218
1219         if (path_strv_contains(read_only_paths, "/home") ||
1220             path_strv_contains(inaccessible_paths, "/home") ||
1221             path_strv_contains(empty_directories, "/home"))
1222                 return true;
1223
1224         for (i = 0; i < n_temporary_filesystems; i++)
1225                 if (path_equal(temporary_filesystems[i].path, "/home"))
1226                         return true;
1227
1228         /* If /home is overmounted with some dir from the host it's not writable. */
1229         for (i = 0; i < n_bind_mounts; i++)
1230                 if (path_equal(bind_mounts[i].destination, "/home"))
1231                         return true;
1232
1233         return false;
1234 }
1235
1236 int setup_namespace(
1237                 const char* root_directory,
1238                 const char* root_image,
1239                 const NamespaceInfo *ns_info,
1240                 char** read_write_paths,
1241                 char** read_only_paths,
1242                 char** inaccessible_paths,
1243                 char** empty_directories,
1244                 const BindMount *bind_mounts,
1245                 size_t n_bind_mounts,
1246                 const TemporaryFileSystem *temporary_filesystems,
1247                 size_t n_temporary_filesystems,
1248                 const char* tmp_dir,
1249                 const char* var_tmp_dir,
1250                 ProtectHome protect_home,
1251                 ProtectSystem protect_system,
1252                 unsigned long mount_flags,
1253                 DissectImageFlags dissect_image_flags,
1254                 char **error_path) {
1255
1256         _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL;
1257         _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL;
1258         _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL;
1259         _cleanup_free_ void *root_hash = NULL;
1260         MountEntry *m = NULL, *mounts = NULL;
1261         size_t n_mounts, root_hash_size = 0;
1262         bool require_prefix = false;
1263         const char *root;
1264         int r = 0;
1265
1266         assert(ns_info);
1267
1268         if (mount_flags == 0)
1269                 mount_flags = MS_SHARED;
1270
1271         if (root_image) {
1272                 dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT;
1273
1274                 /* Make the whole image read-only if we can determine that we only access it in a read-only fashion. */
1275                 if (root_read_only(read_only_paths,
1276                                    protect_system) &&
1277                     home_read_only(read_only_paths, inaccessible_paths, empty_directories,
1278                                    bind_mounts, n_bind_mounts, temporary_filesystems, n_temporary_filesystems,
1279                                    protect_home) &&
1280                     strv_isempty(read_write_paths))
1281                         dissect_image_flags |= DISSECT_IMAGE_READ_ONLY;
1282
1283                 r = loop_device_make_by_path(root_image,
1284                                              FLAGS_SET(dissect_image_flags, DISSECT_IMAGE_READ_ONLY) ? O_RDONLY : -1 /* < 0 means writable if possible, read-only as fallback */,
1285                                              LO_FLAGS_PARTSCAN,
1286                                              &loop_device);
1287                 if (r < 0)
1288                         return log_debug_errno(r, "Failed to create loop device for root image: %m");
1289
1290                 r = root_hash_load(root_image, &root_hash, &root_hash_size);
1291                 if (r < 0)
1292                         return log_debug_errno(r, "Failed to load root hash: %m");
1293
1294                 r = dissect_image(loop_device->fd, root_hash, root_hash_size, dissect_image_flags, &dissected_image);
1295                 if (r < 0)
1296                         return log_debug_errno(r, "Failed to dissect image: %m");
1297
1298                 r = dissected_image_decrypt(dissected_image, NULL, root_hash, root_hash_size, dissect_image_flags, &decrypted_image);
1299                 if (r < 0)
1300                         return log_debug_errno(r, "Failed to decrypt dissected image: %m");
1301         }
1302
1303         if (root_directory)
1304                 root = root_directory;
1305         else {
1306                 /* Always create the mount namespace in a temporary directory, instead of operating
1307                  * directly in the root. The temporary directory prevents any mounts from being
1308                  * potentially obscured my other mounts we already applied.
1309                  * We use the same mount point for all images, which is safe, since they all live
1310                  * in their own namespaces after all, and hence won't see each other. */
1311
1312                 root = "/run/systemd/unit-root";
1313                 (void) mkdir_label(root, 0700);
1314                 require_prefix = true;
1315         }
1316
1317         n_mounts = namespace_calculate_mounts(
1318                         ns_info,
1319                         read_write_paths,
1320                         read_only_paths,
1321                         inaccessible_paths,
1322                         empty_directories,
1323                         n_bind_mounts,
1324                         n_temporary_filesystems,
1325                         tmp_dir, var_tmp_dir,
1326                         protect_home, protect_system);
1327
1328         if (n_mounts > 0) {
1329                 m = mounts = new0(MountEntry, n_mounts);
1330                 if (!mounts)
1331                         return -ENOMEM;
1332
1333                 r = append_access_mounts(&m, read_write_paths, READWRITE, require_prefix);
1334                 if (r < 0)
1335                         goto finish;
1336
1337                 r = append_access_mounts(&m, read_only_paths, READONLY, require_prefix);
1338                 if (r < 0)
1339                         goto finish;
1340
1341                 r = append_access_mounts(&m, inaccessible_paths, INACCESSIBLE, require_prefix);
1342                 if (r < 0)
1343                         goto finish;
1344
1345                 r = append_empty_dir_mounts(&m, empty_directories);
1346                 if (r < 0)
1347                         goto finish;
1348
1349                 r = append_bind_mounts(&m, bind_mounts, n_bind_mounts);
1350                 if (r < 0)
1351                         goto finish;
1352
1353                 r = append_tmpfs_mounts(&m, temporary_filesystems, n_temporary_filesystems);
1354                 if (r < 0)
1355                         goto finish;
1356
1357                 if (tmp_dir) {
1358                         *(m++) = (MountEntry) {
1359                                 .path_const = "/tmp",
1360                                 .mode = PRIVATE_TMP,
1361                                 .source_const = tmp_dir,
1362                         };
1363                 }
1364
1365                 if (var_tmp_dir) {
1366                         *(m++) = (MountEntry) {
1367                                 .path_const = "/var/tmp",
1368                                 .mode = PRIVATE_TMP,
1369                                 .source_const = var_tmp_dir,
1370                         };
1371                 }
1372
1373                 if (ns_info->private_dev) {
1374                         *(m++) = (MountEntry) {
1375                                 .path_const = "/dev",
1376                                 .mode = PRIVATE_DEV,
1377                                 .flags = DEV_MOUNT_OPTIONS,
1378                         };
1379                 }
1380
1381                 if (ns_info->protect_kernel_tunables) {
1382                         r = append_static_mounts(&m, protect_kernel_tunables_table, ELEMENTSOF(protect_kernel_tunables_table), ns_info->ignore_protect_paths);
1383                         if (r < 0)
1384                                 goto finish;
1385                 }
1386
1387                 if (ns_info->protect_kernel_modules) {
1388                         r = append_static_mounts(&m, protect_kernel_modules_table, ELEMENTSOF(protect_kernel_modules_table), ns_info->ignore_protect_paths);
1389                         if (r < 0)
1390                                 goto finish;
1391                 }
1392
1393                 if (ns_info->protect_kernel_logs) {
1394                         r = append_static_mounts(&m, protect_kernel_logs_table, ELEMENTSOF(protect_kernel_logs_table), ns_info->ignore_protect_paths);
1395                         if (r < 0)
1396                                 goto finish;
1397                 }
1398
1399                 if (ns_info->protect_control_groups) {
1400                         *(m++) = (MountEntry) {
1401                                 .path_const = "/sys/fs/cgroup",
1402                                 .mode = READONLY,
1403                         };
1404                 }
1405
1406                 r = append_protect_home(&m, protect_home, ns_info->ignore_protect_paths);
1407                 if (r < 0)
1408                         goto finish;
1409
1410                 r = append_protect_system(&m, protect_system, false);
1411                 if (r < 0)
1412                         goto finish;
1413
1414                 if (namespace_info_mount_apivfs(ns_info)) {
1415                         r = append_static_mounts(&m, apivfs_table, ELEMENTSOF(apivfs_table), ns_info->ignore_protect_paths);
1416                         if (r < 0)
1417                                 goto finish;
1418                 }
1419
1420                 if (ns_info->protect_hostname) {
1421                         *(m++) = (MountEntry) {
1422                                 .path_const = "/proc/sys/kernel/hostname",
1423                                 .mode = READONLY,
1424                         };
1425                         *(m++) = (MountEntry) {
1426                                 .path_const = "/proc/sys/kernel/domainname",
1427                                 .mode = READONLY,
1428                         };
1429                 }
1430
1431                 assert(mounts + n_mounts == m);
1432
1433                 /* Prepend the root directory where that's necessary */
1434                 r = prefix_where_needed(mounts, n_mounts, root);
1435                 if (r < 0)
1436                         goto finish;
1437
1438                 normalize_mounts(root, mounts, &n_mounts);
1439         }
1440
1441         /* All above is just preparation, figuring out what to do. Let's now actually start doing something. */
1442
1443         if (unshare(CLONE_NEWNS) < 0) {
1444                 r = log_debug_errno(errno, "Failed to unshare the mount namespace: %m");
1445                 if (IN_SET(r, -EACCES, -EPERM, -EOPNOTSUPP, -ENOSYS))
1446                         /* If the kernel doesn't support namespaces, or when there's a MAC or seccomp filter in place
1447                          * that doesn't allow us to create namespaces (or a missing cap), then propagate a recognizable
1448                          * error back, which the caller can use to detect this case (and only this) and optionally
1449                          * continue without namespacing applied. */
1450                         r = -ENOANO;
1451
1452                 goto finish;
1453         }
1454
1455         /* Remount / as SLAVE so that nothing now mounted in the namespace
1456          * shows up in the parent */
1457         if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) {
1458                 r = log_debug_errno(errno, "Failed to remount '/' as SLAVE: %m");
1459                 goto finish;
1460         }
1461
1462         if (root_image) {
1463                 /* A root image is specified, mount it to the right place */
1464                 r = dissected_image_mount(dissected_image, root, UID_INVALID, dissect_image_flags);
1465                 if (r < 0) {
1466                         log_debug_errno(r, "Failed to mount root image: %m");
1467                         goto finish;
1468                 }
1469
1470                 if (decrypted_image) {
1471                         r = decrypted_image_relinquish(decrypted_image);
1472                         if (r < 0) {
1473                                 log_debug_errno(r, "Failed to relinquish decrypted image: %m");
1474                                 goto finish;
1475                         }
1476                 }
1477
1478                 loop_device_relinquish(loop_device);
1479
1480         } else if (root_directory) {
1481
1482                 /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */
1483                 r = path_is_mount_point(root, NULL, AT_SYMLINK_FOLLOW);
1484                 if (r < 0) {
1485                         log_debug_errno(r, "Failed to detect that %s is a mount point or not: %m", root);
1486                         goto finish;
1487                 }
1488                 if (r == 0) {
1489                         if (mount(root, root, NULL, MS_BIND|MS_REC, NULL) < 0) {
1490                                 r = log_debug_errno(errno, "Failed to bind mount '%s': %m", root);
1491                                 goto finish;
1492                         }
1493                 }
1494
1495         } else {
1496
1497                 /* Let's mount the main root directory to the root directory to use */
1498                 if (mount("/", root, NULL, MS_BIND|MS_REC, NULL) < 0) {
1499                         r = log_debug_errno(errno, "Failed to bind mount '/' on '%s': %m", root);
1500                         goto finish;
1501                 }
1502         }
1503
1504         /* Try to set up the new root directory before mounting anything else there. */
1505         if (root_image || root_directory)
1506                 (void) base_filesystem_create(root, UID_INVALID, GID_INVALID);
1507
1508         if (n_mounts > 0) {
1509                 _cleanup_fclose_ FILE *proc_self_mountinfo = NULL;
1510                 _cleanup_free_ char **blacklist = NULL;
1511                 size_t j;
1512
1513                 /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of /proc.
1514                  * For example, this is the case with the option: 'InaccessiblePaths=/proc' */
1515                 proc_self_mountinfo = fopen("/proc/self/mountinfo", "re");
1516                 if (!proc_self_mountinfo) {
1517                         r = log_debug_errno(errno, "Failed to open /proc/self/mountinfo: %m");
1518                         if (error_path)
1519                                 *error_path = strdup("/proc/self/mountinfo");
1520                         goto finish;
1521                 }
1522
1523                 /* First round, establish all mounts we need */
1524                 for (;;) {
1525                         bool again = false;
1526
1527                         for (m = mounts; m < mounts + n_mounts; ++m) {
1528
1529                                 if (m->applied)
1530                                         continue;
1531
1532                                 r = follow_symlink(root, m);
1533                                 if (r < 0) {
1534                                         if (error_path && mount_entry_path(m))
1535                                                 *error_path = strdup(mount_entry_path(m));
1536                                         goto finish;
1537                                 }
1538                                 if (r == 0) {
1539                                         /* We hit a symlinked mount point. The entry got rewritten and might point to a
1540                                          * very different place now. Let's normalize the changed list, and start from
1541                                          * the beginning. After all to mount the entry at the new location we might
1542                                          * need some other mounts first */
1543                                         again = true;
1544                                         break;
1545                                 }
1546
1547                                 r = apply_mount(root, m);
1548                                 if (r < 0) {
1549                                         if (error_path && mount_entry_path(m))
1550                                                 *error_path = strdup(mount_entry_path(m));
1551                                         goto finish;
1552                                 }
1553
1554                                 m->applied = true;
1555                         }
1556
1557                         if (!again)
1558                                 break;
1559
1560                         normalize_mounts(root, mounts, &n_mounts);
1561                 }
1562
1563                 /* Create a blacklist we can pass to bind_mount_recursive() */
1564                 blacklist = new(char*, n_mounts+1);
1565                 if (!blacklist) {
1566                         r = -ENOMEM;
1567                         goto finish;
1568                 }
1569                 for (j = 0; j < n_mounts; j++)
1570                         blacklist[j] = (char*) mount_entry_path(mounts+j);
1571                 blacklist[j] = NULL;
1572
1573                 /* Second round, flip the ro bits if necessary. */
1574                 for (m = mounts; m < mounts + n_mounts; ++m) {
1575                         r = make_read_only(m, blacklist, proc_self_mountinfo);
1576                         if (r < 0) {
1577                                 if (error_path && mount_entry_path(m))
1578                                         *error_path = strdup(mount_entry_path(m));
1579                                 goto finish;
1580                         }
1581                 }
1582         }
1583
1584         /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */
1585         r = mount_move_root(root);
1586         if (r < 0) {
1587                 log_debug_errno(r, "Failed to mount root with MS_MOVE: %m");
1588                 goto finish;
1589         }
1590
1591         /* Remount / as the desired mode. Note that this will not
1592          * reestablish propagation from our side to the host, since
1593          * what's disconnected is disconnected. */
1594         if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) {
1595                 r = log_debug_errno(errno, "Failed to remount '/' with desired mount flags: %m");
1596                 goto finish;
1597         }
1598
1599         r = 0;
1600
1601 finish:
1602         for (m = mounts; m < mounts + n_mounts; m++)
1603                 mount_entry_done(m);
1604
1605         free(mounts);
1606
1607         return r;
1608 }
1609
1610 void bind_mount_free_many(BindMount *b, size_t n) {
1611         size_t i;
1612
1613         assert(b || n == 0);
1614
1615         for (i = 0; i < n; i++) {
1616                 free(b[i].source);
1617                 free(b[i].destination);
1618         }
1619
1620         free(b);
1621 }
1622
1623 int bind_mount_add(BindMount **b, size_t *n, const BindMount *item) {
1624         _cleanup_free_ char *s = NULL, *d = NULL;
1625         BindMount *c;
1626
1627         assert(b);
1628         assert(n);
1629         assert(item);
1630
1631         s = strdup(item->source);
1632         if (!s)
1633                 return -ENOMEM;
1634
1635         d = strdup(item->destination);
1636         if (!d)
1637                 return -ENOMEM;
1638
1639         c = reallocarray(*b, *n + 1, sizeof(BindMount));
1640         if (!c)
1641                 return -ENOMEM;
1642
1643         *b = c;
1644
1645         c[(*n) ++] = (BindMount) {
1646                 .source = TAKE_PTR(s),
1647                 .destination = TAKE_PTR(d),
1648                 .read_only = item->read_only,
1649                 .nosuid = item->nosuid,
1650                 .recursive = item->recursive,
1651                 .ignore_enoent = item->ignore_enoent,
1652         };
1653
1654         return 0;
1655 }
1656
1657 void temporary_filesystem_free_many(TemporaryFileSystem *t, size_t n) {
1658         size_t i;
1659
1660         assert(t || n == 0);
1661
1662         for (i = 0; i < n; i++) {
1663                 free(t[i].path);
1664                 free(t[i].options);
1665         }
1666
1667         free(t);
1668 }
1669
1670 int temporary_filesystem_add(
1671                 TemporaryFileSystem **t,
1672                 size_t *n,
1673                 const char *path,
1674                 const char *options) {
1675
1676         _cleanup_free_ char *p = NULL, *o = NULL;
1677         TemporaryFileSystem *c;
1678
1679         assert(t);
1680         assert(n);
1681         assert(path);
1682
1683         p = strdup(path);
1684         if (!p)
1685                 return -ENOMEM;
1686
1687         if (!isempty(options)) {
1688                 o = strdup(options);
1689                 if (!o)
1690                         return -ENOMEM;
1691         }
1692
1693         c = reallocarray(*t, *n + 1, sizeof(TemporaryFileSystem));
1694         if (!c)
1695                 return -ENOMEM;
1696
1697         *t = c;
1698
1699         c[(*n) ++] = (TemporaryFileSystem) {
1700                 .path = TAKE_PTR(p),
1701                 .options = TAKE_PTR(o),
1702         };
1703
1704         return 0;
1705 }
1706
1707 static int make_tmp_prefix(const char *prefix) {
1708         _cleanup_free_ char *t = NULL;
1709         int r;
1710
1711         /* Don't do anything unless we know the dir is actually missing */
1712         r = access(prefix, F_OK);
1713         if (r >= 0)
1714                 return 0;
1715         if (errno != ENOENT)
1716                 return -errno;
1717
1718         r = mkdir_parents(prefix, 0755);
1719         if (r < 0)
1720                 return r;
1721
1722         r = tempfn_random(prefix, NULL, &t);
1723         if (r < 0)
1724                 return r;
1725
1726         if (mkdir(t, 0777) < 0)
1727                 return -errno;
1728
1729         if (chmod(t, 01777) < 0) {
1730                 r = -errno;
1731                 (void) rmdir(t);
1732                 return r;
1733         }
1734
1735         if (rename(t, prefix) < 0) {
1736                 r = -errno;
1737                 (void) rmdir(t);
1738                 return r == -EEXIST ? 0 : r; /* it's fine if someone else created the dir by now */
1739         }
1740
1741         return 0;
1742
1743 }
1744
1745 static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) {
1746         _cleanup_free_ char *x = NULL;
1747         char bid[SD_ID128_STRING_MAX];
1748         sd_id128_t boot_id;
1749         int r;
1750
1751         assert(id);
1752         assert(prefix);
1753         assert(path);
1754
1755         /* We include the boot id in the directory so that after a
1756          * reboot we can easily identify obsolete directories. */
1757
1758         r = sd_id128_get_boot(&boot_id);
1759         if (r < 0)
1760                 return r;
1761
1762         x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX");
1763         if (!x)
1764                 return -ENOMEM;
1765
1766         r = make_tmp_prefix(prefix);
1767         if (r < 0)
1768                 return r;
1769
1770         RUN_WITH_UMASK(0077)
1771                 if (!mkdtemp(x))
1772                         return -errno;
1773
1774         RUN_WITH_UMASK(0000) {
1775                 char *y;
1776
1777                 y = strjoina(x, "/tmp");
1778
1779                 if (mkdir(y, 0777 | S_ISVTX) < 0)
1780                         return -errno;
1781         }
1782
1783         *path = TAKE_PTR(x);
1784
1785         return 0;
1786 }
1787
1788 int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) {
1789         char *a, *b;
1790         int r;
1791
1792         assert(id);
1793         assert(tmp_dir);
1794         assert(var_tmp_dir);
1795
1796         r = setup_one_tmp_dir(id, "/tmp", &a);
1797         if (r < 0)
1798                 return r;
1799
1800         r = setup_one_tmp_dir(id, "/var/tmp", &b);
1801         if (r < 0) {
1802                 char *t;
1803
1804                 t = strjoina(a, "/tmp");
1805                 (void) rmdir(t);
1806                 (void) rmdir(a);
1807
1808                 free(a);
1809                 return r;
1810         }
1811
1812         *tmp_dir = a;
1813         *var_tmp_dir = b;
1814
1815         return 0;
1816 }
1817
1818 int setup_netns(const int netns_storage_socket[static 2]) {
1819         _cleanup_close_ int netns = -1;
1820         int r, q;
1821
1822         assert(netns_storage_socket);
1823         assert(netns_storage_socket[0] >= 0);
1824         assert(netns_storage_socket[1] >= 0);
1825
1826         /* We use the passed socketpair as a storage buffer for our
1827          * namespace reference fd. Whatever process runs this first
1828          * shall create a new namespace, all others should just join
1829          * it. To serialize that we use a file lock on the socket
1830          * pair.
1831          *
1832          * It's a bit crazy, but hey, works great! */
1833
1834         if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0)
1835                 return -errno;
1836
1837         netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT);
1838         if (netns == -EAGAIN) {
1839                 /* Nothing stored yet, so let's create a new namespace. */
1840
1841                 if (unshare(CLONE_NEWNET) < 0) {
1842                         r = -errno;
1843                         goto fail;
1844                 }
1845
1846                 (void) loopback_setup();
1847
1848                 netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY);
1849                 if (netns < 0) {
1850                         r = -errno;
1851                         goto fail;
1852                 }
1853
1854                 r = 1;
1855
1856         } else if (netns < 0) {
1857                 r = netns;
1858                 goto fail;
1859
1860         } else {
1861                 /* Yay, found something, so let's join the namespace */
1862                 if (setns(netns, CLONE_NEWNET) < 0) {
1863                         r = -errno;
1864                         goto fail;
1865                 }
1866
1867                 r = 0;
1868         }
1869
1870         q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT);
1871         if (q < 0) {
1872                 r = q;
1873                 goto fail;
1874         }
1875
1876 fail:
1877         (void) lockf(netns_storage_socket[0], F_ULOCK, 0);
1878         return r;
1879 }
1880
1881 int open_netns_path(const int netns_storage_socket[static 2], const char *path) {
1882         _cleanup_close_ int netns = -1;
1883         int q, r;
1884
1885         assert(netns_storage_socket);
1886         assert(netns_storage_socket[0] >= 0);
1887         assert(netns_storage_socket[1] >= 0);
1888         assert(path);
1889
1890         /* If the storage socket doesn't contain a netns fd yet, open one via the file system and store it in
1891          * it. This is supposed to be called ahead of time, i.e. before setup_netns() which will allocate a
1892          * new anonymous netns if needed. */
1893
1894         if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0)
1895                 return -errno;
1896
1897         netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT);
1898         if (netns == -EAGAIN) {
1899                 /* Nothing stored yet. Open the file from the file system. */
1900
1901                 netns = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC);
1902                 if (netns < 0) {
1903                         r = -errno;
1904                         goto fail;
1905                 }
1906
1907                 r = fd_is_network_ns(netns);
1908                 if (r == 0) { /* Not a netns? Refuse early. */
1909                         r = -EINVAL;
1910                         goto fail;
1911                 }
1912                 if (r < 0 && r != -EUCLEAN) /* EUCLEAN: we don't know */
1913                         goto fail;
1914
1915                 r = 1;
1916
1917         } else if (netns < 0) {
1918                 r = netns;
1919                 goto fail;
1920         } else
1921                 r = 0; /* Already allocated */
1922
1923         q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT);
1924         if (q < 0) {
1925                 r = q;
1926                 goto fail;
1927         }
1928
1929 fail:
1930         (void) lockf(netns_storage_socket[0], F_ULOCK, 0);
1931         return r;
1932 }
1933
1934 bool ns_type_supported(NamespaceType type) {
1935         const char *t, *ns_proc;
1936
1937         t = namespace_type_to_string(type);
1938         if (!t) /* Don't know how to translate this? Then it's not supported */
1939                 return false;
1940
1941         ns_proc = strjoina("/proc/self/ns/", t);
1942         return access(ns_proc, F_OK) == 0;
1943 }
1944
1945 static const char *const protect_home_table[_PROTECT_HOME_MAX] = {
1946         [PROTECT_HOME_NO] = "no",
1947         [PROTECT_HOME_YES] = "yes",
1948         [PROTECT_HOME_READ_ONLY] = "read-only",
1949         [PROTECT_HOME_TMPFS] = "tmpfs",
1950 };
1951
1952 DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_home, ProtectHome, PROTECT_HOME_YES);
1953
1954 static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = {
1955         [PROTECT_SYSTEM_NO] = "no",
1956         [PROTECT_SYSTEM_YES] = "yes",
1957         [PROTECT_SYSTEM_FULL] = "full",
1958         [PROTECT_SYSTEM_STRICT] = "strict",
1959 };
1960
1961 DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_system, ProtectSystem, PROTECT_SYSTEM_YES);
1962
1963 static const char* const namespace_type_table[] = {
1964         [NAMESPACE_MOUNT] = "mnt",
1965         [NAMESPACE_CGROUP] = "cgroup",
1966         [NAMESPACE_UTS] = "uts",
1967         [NAMESPACE_IPC] = "ipc",
1968         [NAMESPACE_USER] = "user",
1969         [NAMESPACE_PID] = "pid",
1970         [NAMESPACE_NET] = "net",
1971 };
1972
1973 DEFINE_STRING_TABLE_LOOKUP(namespace_type, NamespaceType);