src/sysext/sysext.c

   1 /* SPDX-License-Identifier: LGPL-2.1-or-later */
   2
   3 #include <fcntl.h>
   4 #include <getopt.h>
   5 #include <sys/mount.h>
   6 #include <unistd.h>
   7
   8 #include "capability-util.h"
   9 #include "discover-image.h"
  10 #include "dissect-image.h"
  11 #include "env-util.h"
  12 #include "escape.h"
  13 #include "extension-release.h"
  14 #include "fd-util.h"
  15 #include "fileio.h"
  16 #include "format-table.h"
  17 #include "fs-util.h"
  18 #include "hashmap.h"
  19 #include "log.h"
  20 #include "main-func.h"
  21 #include "missing_magic.h"
  22 #include "mkdir.h"
  23 #include "mount-util.h"
  24 #include "mountpoint-util.h"
  25 #include "os-util.h"
  26 #include "pager.h"
  27 #include "parse-argument.h"
  28 #include "parse-util.h"
  29 #include "pretty-print.h"
  30 #include "process-util.h"
  31 #include "sort-util.h"
  32 #include "stat-util.h"
  33 #include "terminal-util.h"
  34 #include "user-util.h"
  35 #include "verbs.h"
  36
  37 static char **arg_hierarchies = NULL; /* "/usr" + "/opt" by default */
  38 static char *arg_root = NULL;
  39 static JsonFormatFlags arg_json_format_flags = JSON_FORMAT_OFF;
  40 static PagerFlags arg_pager_flags = 0;
  41 static bool arg_legend = true;
  42 static bool arg_force = false;
  43
  44 STATIC_DESTRUCTOR_REGISTER(arg_hierarchies, strv_freep);
  45 STATIC_DESTRUCTOR_REGISTER(arg_root, freep);
  46
  47 static int is_our_mount_point(const char *p) {
  48         _cleanup_free_ char *buf = NULL, *f = NULL;
  49         struct stat st;
  50         dev_t dev;
  51         int r;
  52
  53         r = path_is_mount_point(p, NULL, 0);
  54         if (r == -ENOENT) {
  55                 log_debug_errno(r, "Hierarchy '%s' doesn't exist.", p);
  56                 return false;
  57         }
  58         if (r < 0)
  59                 return log_error_errno(r, "Failed to determine whether '%s' is a mount point: %m", p);
  60         if (r == 0) {
  61                 log_debug("Hierarchy '%s' is not a mount point, skipping.", p);
  62                 return false;
  63         }
  64
  65         /* So we know now that it's a mount point. Now let's check if it's one of ours, so that we don't
  66          * accidentally unmount the user's own /usr/ but just the mounts we established ourselves. We do this
  67          * check by looking into the metadata directory we place in merged mounts: if the file
  68          * .systemd-sysext/dev contains the major/minor device pair of the mount we have a good reason to
  69          * believe this is one of our mounts. This thorough check has the benefit that we aren't easily
  70          * confused if people tar up one of our merged trees and untar them elsewhere where we might mistake
  71          * them for a live sysext tree. */
  72
  73         f = path_join(p, ".systemd-sysext/dev");
  74         if (!f)
  75                 return log_oom();
  76
  77         r = read_one_line_file(f, &buf);
  78         if (r == -ENOENT) {
  79                 log_debug("Hierarchy '%s' does not carry a .systemd-sysext/dev file, not a sysext merged tree.", p);
  80                 return false;
  81         }
  82         if (r < 0)
  83                 return log_error_errno(r, "Failed to determine whether hierarchy '%s' contains '.systemd-sysext/dev': %m", p);
  84
  85         r = parse_dev(buf, &dev);
  86         if (r < 0)
  87                 return log_error_errno(r, "Failed to parse device major/minor stored in '.systemd-sysext/dev' file on '%s': %m", p);
  88
  89         if (lstat(p, &st) < 0)
  90                 return log_error_errno(r, "Failed to stat %s: %m", p);
  91
  92         if (st.st_dev != dev) {
  93                 log_debug("Hierarchy '%s' reports a different device major/minor than what we are seeing, assuming offline copy.", p);
  94                 return false;
  95         }
  96
  97         return true;
  98 }
  99
 100 static int unmerge_hierarchy(const char *p) {
 101         int r;
 102
 103         for (;;) {
 104                 /* We only unmount /usr/ if it is a mount point and really one of ours, in order not to break
 105                  * systems where /usr/ is a mount point of its own already. */
 106
 107                 r = is_our_mount_point(p);
 108                 if (r < 0)
 109                         return r;
 110                 if (r == 0)
 111                         break;
 112
 113                 r = umount_verbose(LOG_ERR, p, MNT_DETACH|UMOUNT_NOFOLLOW);
 114                 if (r < 0)
 115                         return log_error_errno(r, "Failed to unmount file system '%s': %m", p);
 116
 117                 log_info("Unmerged '%s'.", p);
 118         }
 119
 120         return 0;
 121 }
 122
 123 static int unmerge(void) {
 124         int r, ret = 0;
 125         char **p;
 126
 127         STRV_FOREACH(p, arg_hierarchies) {
 128                 _cleanup_free_ char *resolved = NULL;
 129
 130                 r = chase_symlinks(*p, arg_root, CHASE_PREFIX_ROOT, &resolved, NULL);
 131                 if (r == -ENOENT) {
 132                         log_debug_errno(r, "Hierarchy '%s%s' does not exist, ignoring.", strempty(arg_root), *p);
 133                         continue;
 134                 }
 135                 if (r < 0) {
 136                         log_error_errno(r, "Failed to resolve path to hierarchy '%s%s': %m", strempty(arg_root), *p);
 137                         if (ret == 0)
 138                                 ret = r;
 139
 140                         continue;
 141                 }
 142
 143                 r = unmerge_hierarchy(resolved);
 144                 if (r < 0 && ret == 0)
 145                         ret = r;
 146         }
 147
 148         return ret;
 149 }
 150
 151 static int verb_unmerge(int argc, char **argv, void *userdata) {
 152
 153         if (!have_effective_cap(CAP_SYS_ADMIN))
 154                 return log_error_errno(SYNTHETIC_ERRNO(EPERM), "Need to be privileged.");
 155
 156         return unmerge();
 157 }
 158
 159 static int verb_status(int argc, char **argv, void *userdata) {
 160         _cleanup_(table_unrefp) Table *t = NULL;
 161         int r, ret = 0;
 162         char **p;
 163
 164         t = table_new("hierarchy", "extensions", "since");
 165         if (!t)
 166                 return log_oom();
 167
 168         (void) table_set_empty_string(t, "-");
 169
 170         STRV_FOREACH(p, arg_hierarchies) {
 171                 _cleanup_free_ char *resolved = NULL, *f = NULL, *buf = NULL;
 172                 _cleanup_strv_free_ char **l = NULL;
 173                 struct stat st;
 174
 175                 r = chase_symlinks(*p, arg_root, CHASE_PREFIX_ROOT, &resolved, NULL);
 176                 if (r == -ENOENT) {
 177                         log_debug_errno(r, "Hierarchy '%s%s' does not exist, ignoring.", strempty(arg_root), *p);
 178                         continue;
 179                 }
 180                 if (r < 0) {
 181                         log_error_errno(r, "Failed to resolve path to hierarchy '%s%s': %m", strempty(arg_root), *p);
 182                         goto inner_fail;
 183                 }
 184
 185                 r = is_our_mount_point(resolved);
 186                 if (r < 0)
 187                         goto inner_fail;
 188                 if (r == 0) {
 189                         r = table_add_many(
 190                                         t,
 191                                         TABLE_PATH, *p,
 192                                         TABLE_STRING, "none",
 193                                         TABLE_SET_COLOR, ansi_grey(),
 194                                         TABLE_EMPTY);
 195                         if (r < 0)
 196                                 return table_log_add_error(r);
 197
 198                         continue;
 199                 }
 200
 201                 f = path_join(*p, ".systemd-sysext/extensions");
 202                 if (!f)
 203                         return log_oom();
 204
 205                 r = read_full_file(f, &buf, NULL);
 206                 if (r < 0)
 207                         return log_error_errno(r, "Failed to open '%s': %m", f);
 208
 209                 l = strv_split_newlines(buf);
 210                 if (!l)
 211                         return log_oom();
 212
 213                 if (stat(*p, &st) < 0)
 214                         return log_error_errno(r, "Failed to stat() '%s': %m", *p);
 215
 216                 r = table_add_many(
 217                                 t,
 218                                 TABLE_PATH, *p,
 219                                 TABLE_STRV, l,
 220                                 TABLE_TIMESTAMP, timespec_load(&st.st_mtim));
 221                 if (r < 0)
 222                         return table_log_add_error(r);
 223
 224                 continue;
 225
 226         inner_fail:
 227                 if (ret == 0)
 228                         ret = r;
 229         }
 230
 231         (void) table_set_sort(t, (size_t) 0);
 232
 233         r = table_print_with_pager(t, arg_json_format_flags, arg_pager_flags, arg_legend);
 234         if (r < 0)
 235                 return r;
 236
 237         return ret;
 238 }
 239
 240 static int mount_overlayfs(
 241                 const char *where,
 242                 char **layers) {
 243
 244         _cleanup_free_ char *options = NULL;
 245         bool separator = false;
 246         char **l;
 247         int r;
 248
 249         assert(where);
 250
 251         options = strdup("lowerdir=");
 252         if (!options)
 253                 return log_oom();
 254
 255         STRV_FOREACH(l, layers) {
 256                 _cleanup_free_ char *escaped = NULL;
 257
 258                 escaped = shell_escape(*l, ",:");
 259                 if (!escaped)
 260                         return log_oom();
 261
 262                 if (!strextend(&options, separator ? ":" : "", escaped))
 263                         return log_oom();
 264
 265                 separator = true;
 266         }
 267
 268         /* Now mount the actual overlayfs */
 269         r = mount_nofollow_verbose(LOG_ERR, "sysext", where, "overlay", MS_RDONLY, options);
 270         if (r < 0)
 271                 return r;
 272
 273         return 0;
 274 }
 275
 276 static int merge_hierarchy(
 277                 const char *hierarchy,
 278                 char **extensions,
 279                 char **paths,
 280                 const char *meta_path,
 281                 const char *overlay_path) {
 282
 283         _cleanup_free_ char *resolved_hierarchy = NULL, *f = NULL, *buf = NULL;
 284         _cleanup_strv_free_ char **layers = NULL;
 285         struct stat st;
 286         char **p;
 287         int r;
 288
 289         assert(hierarchy);
 290         assert(meta_path);
 291         assert(overlay_path);
 292
 293         /* Resolve the path of the host's version of the hierarchy, i.e. what we want to use as lowest layer
 294          * in the overlayfs stack. */
 295         r = chase_symlinks(hierarchy, arg_root, CHASE_PREFIX_ROOT, &resolved_hierarchy, NULL);
 296         if (r == -ENOENT)
 297                 log_debug_errno(r, "Hierarchy '%s' on host doesn't exist, not merging.", hierarchy);
 298         else if (r < 0)
 299                 return log_error_errno(r, "Failed to resolve host hierarchy '%s': %m", hierarchy);
 300         else {
 301                 r = dir_is_empty(resolved_hierarchy);
 302                 if (r < 0)
 303                         return log_error_errno(r, "Failed to check if host hierarchy '%s' is empty: %m", resolved_hierarchy);
 304                 if (r > 0) {
 305                         log_debug("Host hierarchy '%s' is empty, not merging.", resolved_hierarchy);
 306                         resolved_hierarchy = mfree(resolved_hierarchy);
 307                 }
 308         }
 309
 310         /* Let's generate a metadata file that lists all extensions we took into account for this
 311          * hierarchy. We include this in the final fs, to make things nicely discoverable and
 312          * recognizable. */
 313         f = path_join(meta_path, ".systemd-sysext/extensions");
 314         if (!f)
 315                 return log_oom();
 316
 317         buf = strv_join(extensions, "\n");
 318         if (!buf)
 319                 return log_oom();
 320
 321         r = write_string_file(f, buf, WRITE_STRING_FILE_CREATE|WRITE_STRING_FILE_MKDIR_0755);
 322         if (r < 0)
 323                 return log_error_errno(r, "Failed to write extension meta file '%s': %m", f);
 324
 325         /* Put the meta path (i.e. our synthesized stuff) at the top of the layer stack */
 326         layers = strv_new(meta_path);
 327         if (!layers)
 328                 return log_oom();
 329
 330         /* Put the extensions in the middle */
 331         STRV_FOREACH(p, paths) {
 332                 _cleanup_free_ char *resolved = NULL;
 333
 334                 r = chase_symlinks(hierarchy, *p, CHASE_PREFIX_ROOT, &resolved, NULL);
 335                 if (r == -ENOENT) {
 336                         log_debug_errno(r, "Hierarchy '%s' in extension '%s' doesn't exist, not merging.", hierarchy, *p);
 337                         continue;
 338                 }
 339                 if (r < 0)
 340                         return log_error_errno(r, "Failed to resolve hierarchy '%s' in extension '%s': %m", hierarchy, *p);
 341
 342                 r = dir_is_empty(resolved);
 343                 if (r < 0)
 344                         return log_error_errno(r, "Failed to check if hierarchy '%s' in extension '%s' is empty: %m", resolved, *p);
 345                 if (r > 0) {
 346                         log_debug("Hierarchy '%s' in extension '%s' is empty, not merging.", hierarchy, *p);
 347                         continue;
 348                 }
 349
 350                 r = strv_consume(&layers, TAKE_PTR(resolved));
 351                 if (r < 0)
 352                         return log_oom();
 353         }
 354
 355         if (!layers[1]) /* No extension with files in this hierarchy? Then don't do anything. */
 356                 return 0;
 357
 358         if (resolved_hierarchy) {
 359                 /* Add the host hierarchy as last (lowest) layer in the stack */
 360                 r = strv_consume(&layers, TAKE_PTR(resolved_hierarchy));
 361                 if (r < 0)
 362                         return log_oom();
 363         }
 364
 365         r = mkdir_p(overlay_path, 0700);
 366         if (r < 0)
 367                 return log_error_errno(r, "Failed to make directory '%s': %m", overlay_path);
 368
 369         r = mount_overlayfs(overlay_path, layers);
 370         if (r < 0)
 371                 return r;
 372
 373         /* The overlayfs superblock is read-only. Let's also mark the bind mount read-only. Extra turbo safety 😎 */
 374         r = bind_remount_recursive(overlay_path, MS_RDONLY, MS_RDONLY, NULL);
 375         if (r < 0)
 376                 return log_error_errno(r, "Failed to make bind mount '%s' read-only: %m", overlay_path);
 377
 378         /* Now we have mounted the new file system. Let's now figure out its .st_dev field, and make that
 379          * available in the metadata directory. This is useful to detect whether the metadata dir actually
 380          * belongs to the fs it is found on: if .st_dev of the top-level mount matches it, it's pretty likely
 381          * we are looking at a live sysext tree, and not an unpacked tar or so of one. */
 382         if (stat(overlay_path, &st) < 0)
 383                 return log_error_errno(r, "Failed to stat mount '%s': %m", overlay_path);
 384
 385         free(f);
 386         f = path_join(meta_path, ".systemd-sysext/dev");
 387         if (!f)
 388                 return log_oom();
 389
 390         r = write_string_filef(f, WRITE_STRING_FILE_CREATE, "%u:%u", major(st.st_dev), minor(st.st_dev));
 391         if (r < 0)
 392                 return log_error_errno(r, "Failed to write '%s': %m", f);
 393
 394         /* Make sure the top-level dir has an mtime marking the point we established the merge */
 395         if (utimensat(AT_FDCWD, meta_path, NULL, AT_SYMLINK_NOFOLLOW) < 0)
 396                 return log_error_errno(r, "Failed fix mtime of '%s': %m", meta_path);
 397
 398         return 1;
 399 }
 400
 401 static int strverscmp_improvedp(char *const* a, char *const* b) {
 402         /* usable in qsort() for sorting a string array with strverscmp_improved() */
 403         return strverscmp_improved(*a, *b);
 404 }
 405
 406 static int validate_version(
 407                 const char *root,
 408                 const Image *img,
 409                 const char *host_os_release_id,
 410                 const char *host_os_release_version_id,
 411                 const char *host_os_release_sysext_level) {
 412
 413         int r;
 414
 415         assert(root);
 416         assert(img);
 417
 418         if (arg_force) {
 419                 log_debug("Force mode enabled, skipping version validation.");
 420                 return 1;
 421         }
 422
 423         /* Insist that extension images do not overwrite the underlying OS release file (it's fine if
 424          * they place one in /etc/os-release, i.e. where things don't matter, as they aren't
 425          * merged.) */
 426         r = chase_symlinks("/usr/lib/os-release", root, CHASE_PREFIX_ROOT, NULL, NULL);
 427         if (r < 0) {
 428                 if (r != -ENOENT)
 429                         return log_error_errno(r, "Failed to determine whether /usr/lib/os-release exists in the extension image: %m");
 430         } else
 431                 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
 432                                        "Extension image contains /usr/lib/os-release file, which is not allowed (it may carry /etc/os-release), refusing.");
 433
 434         return extension_release_validate(
 435                         img->name,
 436                         host_os_release_id,
 437                         host_os_release_version_id,
 438                         host_os_release_sysext_level,
 439                         img->extension_release);
 440 }
 441
 442 static int merge_subprocess(Hashmap *images, const char *workspace) {
 443         _cleanup_free_ char *host_os_release_id = NULL, *host_os_release_version_id = NULL, *host_os_release_sysext_level = NULL,
 444                 *buf = NULL;
 445         _cleanup_strv_free_ char **extensions = NULL, **paths = NULL;
 446         size_t n_extensions = 0;
 447         unsigned n_ignored = 0;
 448         Image *img;
 449         char **h;
 450         int r;
 451
 452         /* Mark the whole of /run as MS_SLAVE, so that we can mount stuff below it that doesn't show up on
 453          * the host otherwise. */
 454         r = mount_nofollow_verbose(LOG_ERR, NULL, "/run", NULL, MS_SLAVE|MS_REC, NULL);
 455         if (r < 0)
 456                 return log_error_errno(r, "Failed to remount /run/ MS_SLAVE: %m");
 457
 458         /* Let's create the workspace if it's missing */
 459         r = mkdir_p(workspace, 0700);
 460         if (r < 0)
 461                 return log_error_errno(r, "Failed to create /run/systemd/sysext: %m");
 462
 463         /* Let's mount a tmpfs to our workspace. This way we don't need to clean up the inodes we mount over,
 464          * but let the kernel do that entirely automatically, once our namespace dies. Note that this file
 465          * system won't be visible to anyone but us, since we opened our own namespace and then made the
 466          * /run/ hierarchy (which our workspace is contained in) MS_SLAVE, see above. */
 467         r = mount_nofollow_verbose(LOG_ERR, "sysexit", workspace, "tmpfs", 0, "mode=0700");
 468         if (r < 0)
 469                 return r;
 470
 471         /* Acquire host OS release info, so that we can compare it with the extension's data */
 472         r = parse_os_release(
 473                         arg_root,
 474                         "ID", &host_os_release_id,
 475                         "VERSION_ID", &host_os_release_version_id,
 476                         "SYSEXT_LEVEL", &host_os_release_sysext_level);
 477         if (r < 0)
 478                 return log_error_errno(r, "Failed to acquire 'os-release' data of OS tree '%s': %m", empty_to_root(arg_root));
 479
 480         /* Let's now mount all images */
 481         HASHMAP_FOREACH(img, images) {
 482                 _cleanup_free_ char *p = NULL;
 483
 484                 p = path_join(workspace, "extensions", img->name);
 485                 if (!p)
 486                         return log_oom();
 487
 488                 r = mkdir_p(p, 0700);
 489                 if (r < 0)
 490                         return log_error_errno(r, "Failed to create %s: %m", p);
 491
 492                 switch (img->type) {
 493                 case IMAGE_DIRECTORY:
 494                 case IMAGE_SUBVOLUME:
 495                         r = mount_nofollow_verbose(LOG_ERR, img->path, p, NULL, MS_BIND, NULL);
 496                         if (r < 0)
 497                                 return r;
 498
 499                         /* Make this a read-only bind mount */
 500                         r = bind_remount_recursive(p, MS_RDONLY, MS_RDONLY, NULL);
 501                         if (r < 0)
 502                                 return log_error_errno(r, "Failed to make bind mount '%s' read-only: %m", p);
 503
 504                         break;
 505
 506                 case IMAGE_RAW:
 507                 case IMAGE_BLOCK: {
 508                         _cleanup_(dissected_image_unrefp) DissectedImage *m = NULL;
 509                         _cleanup_(loop_device_unrefp) LoopDevice *d = NULL;
 510                         _cleanup_(decrypted_image_unrefp) DecryptedImage *di = NULL;
 511                         _cleanup_(verity_settings_done) VeritySettings verity_settings = VERITY_SETTINGS_DEFAULT;
 512                         DissectImageFlags flags =
 513                                 DISSECT_IMAGE_READ_ONLY |
 514                                 DISSECT_IMAGE_GENERIC_ROOT |
 515                                 DISSECT_IMAGE_REQUIRE_ROOT |
 516                                 DISSECT_IMAGE_MOUNT_ROOT_ONLY |
 517                                 DISSECT_IMAGE_USR_NO_ROOT;
 518
 519                         r = verity_settings_load(&verity_settings, img->path, NULL, NULL);
 520                         if (r < 0)
 521                                 return log_error_errno(r, "Failed to read verity artifacts for %s: %m", img->path);
 522
 523                         if (verity_settings.data_path)
 524                                 flags |= DISSECT_IMAGE_NO_PARTITION_TABLE;
 525
 526                         r = loop_device_make_by_path(img->path, O_RDONLY, 0, &d);
 527                         if (r < 0)
 528                                 return log_error_errno(r, "Failed to set up loopback device: %m");
 529
 530                         r = dissect_image_and_warn(
 531                                         d->fd,
 532                                         img->path,
 533                                         &verity_settings,
 534                                         NULL,
 535                                         flags,
 536                                         &m);
 537                         if (r < 0)
 538                                 return r;
 539
 540                         r = dissected_image_decrypt_interactively(
 541                                         m, NULL,
 542                                         &verity_settings,
 543                                         flags,
 544                                         &di);
 545                         if (r < 0)
 546                                 return r;
 547
 548                         r = dissected_image_mount_and_warn(
 549                                         m,
 550                                         p,
 551                                         UID_INVALID,
 552                                         flags);
 553                         if (r < 0)
 554                                 return r;
 555
 556                         if (di) {
 557                                 r = decrypted_image_relinquish(di);
 558                                 if (r < 0)
 559                                         return log_error_errno(r, "Failed to relinquish DM devices: %m");
 560                         }
 561
 562                         loop_device_relinquish(d);
 563                         break;
 564                 }
 565                 default:
 566                         assert_not_reached("Unsupported image type");
 567                 }
 568
 569                 r = validate_version(
 570                                 p,
 571                                 img,
 572                                 host_os_release_id,
 573                                 host_os_release_version_id,
 574                                 host_os_release_sysext_level);
 575                 if (r < 0)
 576                         return r;
 577                 if (r == 0) {
 578                         n_ignored++;
 579                         continue;
 580                 }
 581
 582                 /* Noice! This one is an extension we want. */
 583                 r = strv_extend(&extensions, img->name);
 584                 if (r < 0)
 585                         return log_oom();
 586
 587                 n_extensions ++;
 588         }
 589
 590         /* Nothing left? Then shortcut things */
 591         if (n_extensions == 0) {
 592                 if (n_ignored > 0)
 593                         log_info("No suitable extensions found (%u ignored due to incompatible version).", n_ignored);
 594                 else
 595                         log_info("No extensions found.");
 596                 return 0;
 597         }
 598
 599         /* Order by version sort with strverscmp_improved() */
 600         typesafe_qsort(extensions, n_extensions, strverscmp_improvedp);
 601
 602         buf = strv_join(extensions, "', '");
 603         if (!buf)
 604                 return log_oom();
 605
 606         log_info("Using extensions '%s'.", buf);
 607
 608         /* Build table of extension paths (in reverse order) */
 609         paths = new0(char*, n_extensions + 1);
 610         if (!paths)
 611                 return log_oom();
 612
 613         for (size_t k = 0; k < n_extensions; k++) {
 614                 _cleanup_free_ char *p = NULL;
 615
 616                 assert_se(img = hashmap_get(images, extensions[n_extensions - 1 - k]));
 617
 618                 p = path_join(workspace, "extensions", img->name);
 619                 if (!p)
 620                         return log_oom();
 621
 622                 paths[k] = TAKE_PTR(p);
 623         }
 624
 625         /* Let's now unmerge the status quo ante, since to build the new overlayfs we need a reference to the
 626          * underlying fs. */
 627         STRV_FOREACH(h, arg_hierarchies) {
 628                 _cleanup_free_ char *resolved = NULL;
 629
 630                 r = chase_symlinks(*h, arg_root, CHASE_PREFIX_ROOT|CHASE_NONEXISTENT, &resolved, NULL);
 631                 if (r < 0)
 632                         return log_error_errno(r, "Failed to resolve hierarchy '%s%s': %m", strempty(arg_root), *h);
 633
 634                 r = unmerge_hierarchy(resolved);
 635                 if (r < 0)
 636                         return r;
 637         }
 638
 639         /* Create overlayfs mounts for all hierarchies */
 640         STRV_FOREACH(h, arg_hierarchies) {
 641                 _cleanup_free_ char *meta_path = NULL, *overlay_path = NULL;
 642
 643                 meta_path = path_join(workspace, "meta", *h); /* The place where to store metadata about this instance */
 644                 if (!meta_path)
 645                         return log_oom();
 646
 647                 overlay_path = path_join(workspace, "overlay", *h); /* The resulting overlayfs instance */
 648                 if (!overlay_path)
 649                         return log_oom();
 650
 651                 r = merge_hierarchy(*h, extensions, paths, meta_path, overlay_path);
 652                 if (r < 0)
 653                         return r;
 654         }
 655
 656         /* And move them all into place. This is where things appear in the host namespace */
 657         STRV_FOREACH(h, arg_hierarchies) {
 658                 _cleanup_free_ char *p = NULL, *resolved = NULL;
 659
 660                 p = path_join(workspace, "overlay", *h);
 661                 if (!p)
 662                         return log_oom();
 663
 664                 if (laccess(p, F_OK) < 0) {
 665                         if (errno != ENOENT)
 666                                 return log_error_errno(errno, "Failed to check if '%s' exists: %m", p);
 667
 668                         /* Hierarchy apparently was empty in all extensions, and wasn't mounted, ignoring. */
 669                         continue;
 670                 }
 671
 672                 r = chase_symlinks(*h, arg_root, CHASE_PREFIX_ROOT|CHASE_NONEXISTENT, &resolved, NULL);
 673                 if (r < 0)
 674                         return log_error_errno(r, "Failed to resolve hierarchy '%s%s': %m", strempty(arg_root), *h);
 675
 676                 r = mkdir_p(resolved, 0755);
 677                 if (r < 0)
 678                         return log_error_errno(r, "Failed to create hierarchy mount point '%s': %m", resolved);
 679
 680                 r = mount_nofollow_verbose(LOG_ERR, p, resolved, NULL, MS_BIND, NULL);
 681                 if (r < 0)
 682                         return r;
 683
 684                 log_info("Merged extensions into '%s'.", resolved);
 685         }
 686
 687         return 1;
 688 }
 689
 690 static int merge(Hashmap *images) {
 691         pid_t pid;
 692         int r;
 693
 694         r = safe_fork("(sd-sysext)", FORK_DEATHSIG|FORK_LOG|FORK_NEW_MOUNTNS, &pid);
 695         if (r < 0)
 696                 return log_error_errno(r, "Failed to fork off child: %m");
 697         if (r == 0) {
 698                 /* Child with its own mount namespace */
 699
 700                 r = merge_subprocess(images, "/run/systemd/sysext");
 701                 if (r < 0)
 702                         _exit(EXIT_FAILURE);
 703
 704                 /* Our namespace ceases to exist here, also implicitly detaching all temporary mounts we
 705                  * created below /run. Nice! */
 706
 707                 _exit(r > 0 ? EXIT_SUCCESS : 123); /* 123 means: didn't find any extensions */
 708         }
 709
 710         r = wait_for_terminate_and_check("(sd-sysext)", pid, WAIT_LOG_ABNORMAL);
 711         if (r < 0)
 712                 return r;
 713
 714         return r != 123; /* exit code 123 means: didn't do anything */
 715 }
 716
 717 static int image_discover_and_read_metadata(Hashmap **ret_images) {
 718         _cleanup_(hashmap_freep) Hashmap *images = NULL;
 719         Image *img;
 720         int r;
 721
 722         assert(ret_images);
 723
 724         images = hashmap_new(&image_hash_ops);
 725         if (!images)
 726                 return log_oom();
 727
 728         r = image_discover(IMAGE_EXTENSION, arg_root, images);
 729         if (r < 0)
 730                 return log_error_errno(r, "Failed to discover extension images: %m");
 731
 732         HASHMAP_FOREACH(img, images) {
 733                 r = image_read_metadata(img);
 734                 if (r < 0)
 735                         return log_error_errno(r, "Failed to read metadata for image %s: %m", img->name);
 736         }
 737
 738         *ret_images = TAKE_PTR(images);
 739
 740         return 0;
 741 }
 742
 743 static int verb_merge(int argc, char **argv, void *userdata) {
 744         _cleanup_(hashmap_freep) Hashmap *images = NULL;
 745         char **p;
 746         int r;
 747
 748         if (!have_effective_cap(CAP_SYS_ADMIN))
 749                 return log_error_errno(SYNTHETIC_ERRNO(EPERM), "Need to be privileged.");
 750
 751         r = image_discover_and_read_metadata(&images);
 752         if (r < 0)
 753                 return r;
 754
 755         /* In merge mode fail if things are already merged. (In --refresh mode below we'll unmerge if we find
 756          * things are already merged...) */
 757         STRV_FOREACH(p, arg_hierarchies) {
 758                 _cleanup_free_ char *resolved = NULL;
 759
 760                 r = chase_symlinks(*p, arg_root, CHASE_PREFIX_ROOT, &resolved, NULL);
 761                 if (r == -ENOENT) {
 762                         log_debug_errno(r, "Hierarchy '%s%s' does not exist, ignoring.", strempty(arg_root), *p);
 763                         continue;
 764                 }
 765                 if (r < 0)
 766                         return log_error_errno(r, "Failed to resolve path to hierarchy '%s%s': %m", strempty(arg_root), *p);
 767
 768                 r = is_our_mount_point(resolved);
 769                 if (r < 0)
 770                         return r;
 771                 if (r > 0)
 772                         return log_error_errno(SYNTHETIC_ERRNO(EBUSY),
 773                                                "Hierarchy '%s' is already merged.", *p);
 774         }
 775
 776         return merge(images);
 777 }
 778
 779 static int verb_refresh(int argc, char **argv, void *userdata) {
 780         _cleanup_(hashmap_freep) Hashmap *images = NULL;
 781         int r;
 782
 783         if (!have_effective_cap(CAP_SYS_ADMIN))
 784                 return log_error_errno(SYNTHETIC_ERRNO(EPERM), "Need to be privileged.");
 785
 786         r = image_discover_and_read_metadata(&images);
 787         if (r < 0)
 788                 return r;
 789
 790         r = merge(images); /* Returns > 0 if it did something, i.e. a new overlayfs is mounted now. When it
 791                             * does so it implicitly unmounts any overlayfs placed there before. Returns == 0
 792                             * if it did nothing, i.e. no extension images found. In this case the old
 793                             * overlayfs remains in place if there was one. */
 794         if (r < 0)
 795                 return r;
 796         if (r == 0) /* No images found? Then unmerge. The goal of --refresh is after all that after having
 797                      * called there's a guarantee that the merge status matches the installed extensions. */
 798                 r = unmerge();
 799
 800         /* Net result here is that:
 801          *
 802          * 1. If an overlayfs was mounted before and no extensions exist anymore, we'll have unmerged things.
 803          *
 804          * 2. If an overlayfs was mounted before, and there are still extensions installed' we'll have
 805          *    unmerged and then merged things again.
 806          *
 807          * 3. If an overlayfs so far wasn't mounted, and there are extensions installed, we'll have it
 808          *    mounted now.
 809          *
 810          * 4. If there was no overlayfs mount so far, and no extensions installed, we implement a NOP.
 811          */
 812
 813         return 0;
 814 }
 815
 816 static int verb_list(int argc, char **argv, void *userdata) {
 817         _cleanup_(hashmap_freep) Hashmap *images = NULL;
 818         _cleanup_(table_unrefp) Table *t = NULL;
 819         Image *img;
 820         int r;
 821
 822         images = hashmap_new(&image_hash_ops);
 823         if (!images)
 824                 return log_oom();
 825
 826         r = image_discover(IMAGE_EXTENSION, arg_root, images);
 827         if (r < 0)
 828                 return log_error_errno(r, "Failed to discover extension images: %m");
 829
 830         if ((arg_json_format_flags & JSON_FORMAT_OFF) && hashmap_isempty(images)) {
 831                 log_info("No OS extensions found.");
 832                 return 0;
 833         }
 834
 835         t = table_new("name", "type", "path", "time");
 836         if (!t)
 837                 return log_oom();
 838
 839         HASHMAP_FOREACH(img, images) {
 840                 r = table_add_many(
 841                                 t,
 842                                 TABLE_STRING, img->name,
 843                                 TABLE_STRING, image_type_to_string(img->type),
 844                                 TABLE_PATH, img->path,
 845                                 TABLE_TIMESTAMP, img->mtime != 0 ? img->mtime : img->crtime);
 846                 if (r < 0)
 847                         return table_log_add_error(r);
 848         }
 849
 850         (void) table_set_sort(t, (size_t) 0);
 851
 852         return table_print_with_pager(t, arg_json_format_flags, arg_pager_flags, arg_legend);
 853 }
 854
 855 static int verb_help(int argc, char **argv, void *userdata) {
 856         _cleanup_free_ char *link = NULL;
 857         int r;
 858
 859         r = terminal_urlify_man("systemd-sysext", "1", &link);
 860         if (r < 0)
 861                 return log_oom();
 862
 863         printf("%1$s [OPTIONS...] [DEVICE]\n"
 864                "\n%5$sMerge extension images into /usr/ and /opt/ hierarchies.%6$s\n"
 865                "\n%3$sCommands:%4$s\n"
 866                "  status                  Show current merge status (default)\n"
 867                "  merge                   Merge extensions into /usr/ and /opt/\n"
 868                "  unmerge                 Unmerge extensions from /usr/ and /opt/\n"
 869                "  refresh                 Unmerge/merge extensions again\n"
 870                "  list                    List installed extensions\n"
 871                "  -h --help               Show this help\n"
 872                "     --version            Show package version\n"
 873                "\n%3$sOptions:%4$s\n"
 874                "     --no-pager           Do not pipe output into a pager\n"
 875                "     --no-legend          Do not show the headers and footers\n"
 876                "     --root=PATH          Operate relative to root path\n"
 877                "     --json=pretty|short|off\n"
 878                "                          Generate JSON output\n"
 879                "     --force              Ignore version incompatibilities\n"
 880                "\nSee the %2$s for details.\n",
 881                program_invocation_short_name,
 882                link,
 883                ansi_underline(),
 884                ansi_normal(),
 885                ansi_highlight(),
 886                ansi_normal());
 887
 888         return 0;
 889 }
 890
 891 static int parse_argv(int argc, char *argv[]) {
 892
 893         enum {
 894                 ARG_VERSION = 0x100,
 895                 ARG_NO_PAGER,
 896                 ARG_NO_LEGEND,
 897                 ARG_ROOT,
 898                 ARG_JSON,
 899                 ARG_FORCE,
 900         };
 901
 902         static const struct option options[] = {
 903                 { "help",      no_argument,       NULL, 'h'           },
 904                 { "version",   no_argument,       NULL, ARG_VERSION   },
 905                 { "no-pager",  no_argument,       NULL, ARG_NO_PAGER  },
 906                 { "no-legend", no_argument,       NULL, ARG_NO_LEGEND },
 907                 { "root",      required_argument, NULL, ARG_ROOT      },
 908                 { "json",      required_argument, NULL, ARG_JSON      },
 909                 { "force",     no_argument,       NULL, ARG_FORCE     },
 910                 {}
 911         };
 912
 913         int c, r;
 914
 915         assert(argc >= 0);
 916         assert(argv);
 917
 918         while ((c = getopt_long(argc, argv, "h", options, NULL)) >= 0)
 919
 920                 switch (c) {
 921
 922                 case 'h':
 923                         return verb_help(argc, argv, NULL);
 924
 925                 case ARG_VERSION:
 926                         return version();
 927
 928                 case ARG_NO_PAGER:
 929                         arg_pager_flags |= PAGER_DISABLE;
 930                         break;
 931
 932                 case ARG_NO_LEGEND:
 933                         arg_legend = false;
 934                         break;
 935
 936                 case ARG_ROOT:
 937                         r = parse_path_argument(optarg, false, &arg_root);
 938                         if (r < 0)
 939                                 return r;
 940                         break;
 941
 942                 case ARG_JSON:
 943                         r = parse_json_argument(optarg, &arg_json_format_flags);
 944                         if (r <= 0)
 945                                 return r;
 946
 947                         break;
 948
 949                 case ARG_FORCE:
 950                         arg_force = true;
 951                         break;
 952
 953                 case '?':
 954                         return -EINVAL;
 955
 956                 default:
 957                         assert_not_reached("Unhandled option");
 958                 }
 959
 960         return 1;
 961 }
 962
 963 static int sysext_main(int argc, char *argv[]) {
 964
 965         static const Verb verbs[] = {
 966                 { "status",   VERB_ANY, 1, VERB_DEFAULT, verb_status  },
 967                 { "merge",    VERB_ANY, 1, 0,            verb_merge   },
 968                 { "unmerge",  VERB_ANY, 1, 0,            verb_unmerge },
 969                 { "refresh",  VERB_ANY, 1, 0,            verb_refresh },
 970                 { "list",     VERB_ANY, 1, 0,            verb_list    },
 971                 { "help",     VERB_ANY, 1, 0,            verb_help    },
 972                 {}
 973         };
 974
 975         return dispatch_verb(argc, argv, verbs, NULL);
 976 }
 977
 978 static int run(int argc, char *argv[]) {
 979         int r;
 980
 981         log_setup();
 982
 983         r = parse_argv(argc, argv);
 984         if (r <= 0)
 985                 return r;
 986
 987         /* For debugging purposes it might make sense to do this for other hierarchies than /usr/ and
 988          * /opt/, but let's make that a hacker/debugging feature, i.e. env var instead of cmdline
 989          * switch. */
 990         r = parse_env_extension_hierarchies(&arg_hierarchies);
 991         if (r < 0)
 992                 return log_error_errno(r, "Failed to parse $SYSTEMD_SYSEXT_HIERARCHIES environment variable: %m");
 993
 994         return sysext_main(argc, argv);
 995 }
 996
 997 DEFINE_MAIN_FUNCTION(run);