]> git.ipfire.org Git - thirdparty/systemd.git/blame - src/shared/dissect-image.c
projects / thirdparty / systemd.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge pull request #21838 from lnussel/logind-refactor
[thirdparty/systemd.git] / src / shared / dissect-image.c
CommitLineData
db9ecf05 1/* SPDX-License-Identifier: LGPL-2.1-or-later */
8c1be37e 2
10c1b188
LP		 3#if HAVE_VALGRIND_MEMCHECK_H
4#include <valgrind/memcheck.h>
5#endif
6
01234e1f
YW		 7#include <linux/dm-ioctl.h>
8#include <linux/loop.h>
8c1be37e 9#include <sys/mount.h>
3b925504
LP		 10#include <sys/prctl.h>
11#include <sys/wait.h>
f5ea63a5 12#include <sysexits.h>
8c1be37e 13
c2fa92e7
LP		 14#if HAVE_OPENSSL
15#include <openssl/err.h>
16#include <openssl/pem.h>
17#include <openssl/x509.h>
18#endif
19
3c1f2cee 20#include "sd-device.h"
dccca82b
LP		 21#include "sd-id128.h"
22
8c1be37e 23#include "architecture.h"
18b5886e 24#include "ask-password-api.h"
8c1be37e 25#include "blkid-util.h"
18c528e9 26#include "blockdev-util.h"
f4351959 27#include "chase-symlinks.h"
c2fa92e7 28#include "conf-files.h"
3b925504 29#include "copy.h"
1e2f3230 30#include "cryptsetup-util.h"
3b925504 31#include "def.h"
553e15f2 32#include "device-nodes.h"
8437c059 33#include "device-util.h"
7718ac97 34#include "discover-image.h"
8c1be37e 35#include "dissect-image.h"
a709a315 36#include "dm-util.h"
686d13b9 37#include "env-file.h"
88b3300f 38#include "env-util.h"
93f59701 39#include "extension-release.h"
18b5886e 40#include "fd-util.h"
78ebe980 41#include "fileio.h"
2eedfd2d 42#include "fs-util.h"
cf32c486 43#include "fsck-util.h"
8c1be37e 44#include "gpt.h"
78ebe980 45#include "hexdecoct.h"
e2054217 46#include "hostname-setup.h"
3b925504 47#include "id128-util.h"
593fe6c0 48#include "import-util.h"
a4e0d617 49#include "io-util.h"
35cd0ba5 50#include "mkdir-label.h"
8c1be37e 51#include "mount-util.h"
e4de7287 52#include "mountpoint-util.h"
6aa05ebd 53#include "namespace-util.h"
d8b4d14d 54#include "nulstr-util.h"
c2fa92e7 55#include "openssl-util.h"
d58ad743 56#include "os-util.h"
8c1be37e 57#include "path-util.h"
3b925504
LP		 58#include "process-util.h"
59#include "raw-clone.h"
81939d9d 60#include "resize-fs.h"
3b925504 61#include "signal-util.h"
8c1be37e 62#include "stat-util.h"
18b5886e 63#include "stdio-util.h"
8c1be37e
LP		 64#include "string-table.h"
65#include "string-util.h"
2eedfd2d 66#include "strv.h"
e4de7287 67#include "tmpfile-util.h"
a8040b6d 68#include "udev-util.h"
2d3a5a73 69#include "user-util.h"
41488e1f 70#include "xattr-util.h"
8c1be37e 71
28e2641a
FF		 72/* how many times to wait for the device nodes to appear */
73#define N_DEVICE_NODE_LIST_ATTEMPTS 10
74
c34b75a1 75int probe_filesystem(const char *node, char **ret_fstype) {
7cc84b2c 76 /* Try to find device content type and return it in *ret_fstype. If nothing is found,
5238e957 77 * 0/NULL will be returned. -EUCLEAN will be returned for ambiguous results, and an
7cc84b2c
ZJS		 78 * different error otherwise. */
79
349cc4a5 80#if HAVE_BLKID
8e766630 81 _cleanup_(blkid_free_probep) blkid_probe b = NULL;
18b5886e
LP		 82 const char *fstype;
83 int r;
84
995fa2e5 85 errno = 0;
18b5886e
LP		 86 b = blkid_new_probe_from_filename(node);
87 if (!b)
66855de7 88 return errno_or_else(ENOMEM);
18b5886e
LP		 89
90 blkid_probe_enable_superblocks(b, 1);
91 blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE);
92
93 errno = 0;
94 r = blkid_do_safeprobe(b);
7cc84b2c
ZJS		 95 if (r == 1) {
96 log_debug("No type detected on partition %s", node);
18b5886e
LP		 97 goto not_found;
98 }
58dfbfbd
LP		 99 if (r == -2)
100 return log_debug_errno(SYNTHETIC_ERRNO(EUCLEAN),
101 "Results ambiguous for partition %s", node);
b382db9f 102 if (r != 0)
66855de7 103 return errno_or_else(EIO);
18b5886e
LP		 104
105 (void) blkid_probe_lookup_value(b, "TYPE", &fstype, NULL);
106
107 if (fstype) {
108 char *t;
109
110 t = strdup(fstype);
111 if (!t)
112 return -ENOMEM;
113
114 *ret_fstype = t;
115 return 1;
116 }
117
118not_found:
119 *ret_fstype = NULL;
120 return 0;
d1c536f5
ZJS		 121#else
122 return -EOPNOTSUPP;
a75e27eb 123#endif
d1c536f5 124}
18b5886e 125
40c10d3f 126#if HAVE_BLKID
4ba86848
LP		 127static int enumerator_for_parent(sd_device *d, sd_device_enumerator **ret) {
128 _cleanup_(sd_device_enumerator_unrefp) sd_device_enumerator *e = NULL;
129 int r;
aae22eb3 130
f70e7f70 131 assert(d);
4ba86848 132 assert(ret);
f70e7f70 133
4ba86848
LP		 134 r = sd_device_enumerator_new(&e);
135 if (r < 0)
136 return r;
3c1f2cee 137
210e1cd6
YW		 138 r = sd_device_enumerator_add_match_subsystem(e, "block", true);
139 if (r < 0)
140 return r;
141
4ba86848
LP		 142 r = sd_device_enumerator_add_match_parent(e, d);
143 if (r < 0)
144 return r;
145
210e1cd6
YW		 146 r = sd_device_enumerator_add_match_sysattr(e, "partition", NULL, true);
147 if (r < 0)
148 return r;
149
4ba86848
LP		 150 *ret = TAKE_PTR(e);
151 return 0;
cde942f6
JPRV		 152}
153
c1737506
LP		 154static int device_is_partition(
155 sd_device *d,
156 sd_device *expected_parent,
157 blkid_partition pp) {
158
0a8f9bc6 159 const char *v, *parent_syspath, *expected_parent_syspath;
4ba86848
LP		 160 blkid_loff_t bsize, bstart;
161 uint64_t size, start;
162 int partno, bpartno, r;
0a8f9bc6 163 sd_device *parent;
aae22eb3 164
f70e7f70 165 assert(d);
0a8f9bc6 166 assert(expected_parent);
4ba86848 167 assert(pp);
f70e7f70 168
11368b69 169 r = sd_device_get_subsystem(d, &v);
4ba86848
LP		 170 if (r < 0)
171 return r;
11368b69
YW		 172 if (!streq(v, "block"))
173 return false;
174
175 if (sd_device_get_devtype(d, &v) < 0 || !streq(v, "partition"))
aae22eb3
LP		 176 return false;
177
0a8f9bc6
YW		 178 r = sd_device_get_parent(d, &parent);
179 if (r < 0)
180 return false; /* Doesn't have a parent? No relevant to us */
181
182 r = sd_device_get_syspath(parent, &parent_syspath); /* Check parent of device of this action */
183 if (r < 0)
184 return r;
185
186 r = sd_device_get_syspath(expected_parent, &expected_parent_syspath); /* Check parent of device we are looking for */
187 if (r < 0)
188 return r;
189
190 if (!path_equal(parent_syspath, expected_parent_syspath))
191 return false; /* Has a different parent than what we need, not interesting to us */
192
7d25c246
LP		 193 /* On kernel uevents we may find the partition number in the PARTN= field. Let's use that preferably,
194 * since it's cheaper and more importantly: the sysfs attribute "partition" appears to become
c1737506
LP		 195 * available late, hence let's use the property instead, which is available at the moment we see the
196 * uevent. */
197 r = sd_device_get_property_value(d, "PARTN", &v);
198 if (r == -ENOENT)
199 r = sd_device_get_sysattr_value(d, "partition", &v);
4ba86848
LP		 200 if (r < 0)
201 return r;
c1737506 202
4ba86848
LP		 203 r = safe_atoi(v, &partno);
204 if (r < 0)
205 return r;
ea887be0 206
4ba86848
LP		 207 errno = 0;
208 bpartno = blkid_partition_get_partno(pp);
209 if (bpartno < 0)
210 return errno_or_else(EIO);
ea887be0 211
4ba86848
LP		 212 if (partno != bpartno)
213 return false;
f70e7f70 214
4ba86848 215 r = sd_device_get_sysattr_value(d, "start", &v);
ea887be0
ZJS		 216 if (r < 0)
217 return r;
4ba86848 218 r = safe_atou64(v, &start);
ea887be0
ZJS		 219 if (r < 0)
220 return r;
221
4ba86848
LP		 222 errno = 0;
223 bstart = blkid_partition_get_start(pp);
224 if (bstart < 0)
225 return errno_or_else(EIO);
226
227 if (start != (uint64_t) bstart)
228 return false;
229
230 r = sd_device_get_sysattr_value(d, "size", &v);
231 if (r < 0)
232 return r;
233 r = safe_atou64(v, &size);
ea887be0
ZJS		 234 if (r < 0)
235 return r;
236
4ba86848
LP		 237 errno = 0;
238 bsize = blkid_partition_get_size(pp);
239 if (bsize < 0)
240 return errno_or_else(EIO);
241
242 if (size != (uint64_t) bsize)
243 return false;
244
245 return true;
ea887be0
ZJS		 246}
247
4ba86848
LP		 248static int find_partition(
249 sd_device *parent,
250 blkid_partition pp,
4a62257d 251 usec_t timestamp_not_before,
3fe398ce 252 DissectImageFlags flags,
4ba86848 253 sd_device **ret) {
ea887be0
ZJS		 254
255 _cleanup_(sd_device_enumerator_unrefp) sd_device_enumerator *e = NULL;
256 sd_device *q;
ea887be0
ZJS		 257 int r;
258
4ba86848
LP		 259 assert(parent);
260 assert(pp);
261 assert(ret);
f70e7f70 262
4ba86848 263 r = enumerator_for_parent(parent, &e);
ea887be0
ZJS		 264 if (r < 0)
265 return r;
266
ea887be0 267 FOREACH_DEVICE(e, q) {
4a62257d
LP		 268 uint64_t usec;
269
3fe398ce
LP		 270 if (!FLAGS_SET(flags, DISSECT_IMAGE_NO_UDEV)) {
271 r = sd_device_get_usec_initialized(q, &usec);
272 if (r == -EBUSY) /* Not initialized yet */
273 continue;
274 if (r < 0)
275 return r;
4a62257d 276
3fe398ce
LP		 277 if (timestamp_not_before != USEC_INFINITY &&
278 usec < timestamp_not_before) /* udev database entry older than our attachment? Then it's not ours */
279 continue;
280 }
4a62257d 281
0a8f9bc6 282 r = device_is_partition(q, parent, pp);
4ba86848
LP		 283 if (r < 0)
284 return r;
285 if (r > 0) {
286 *ret = sd_device_ref(q);
287 return 0;
052eaf5c 288 }
ea887be0
ZJS		 289 }
290
4ba86848
LP		 291 return -ENXIO;
292}
10c1b188 293
4ba86848
LP		 294struct wait_data {
295 sd_device *parent_device;
296 blkid_partition blkidp;
297 sd_device *found;
a3642997 298 uint64_t diskseq;
75dc190d 299 uint64_t uevent_seqnum_not_before;
05c3c620
YW		 300 usec_t timestamp_not_before;
301 DissectImageFlags flags;
4ba86848 302};
ea887be0 303
4ba86848
LP		 304static inline void wait_data_done(struct wait_data *d) {
305 sd_device_unref(d->found);
306}
ea887be0 307
4ba86848 308static int device_monitor_handler(sd_device_monitor *monitor, sd_device *device, void *userdata) {
4ba86848 309 struct wait_data *w = userdata;
4ba86848
LP		 310 int r;
311
312 assert(w);
313
a1130022 314 if (device_for_action(device, SD_DEVICE_REMOVE))
4ba86848
LP		 315 return 0;
316
a3642997
LB		 317 if (w->diskseq != 0) {
318 uint64_t diskseq;
319
320 /* If w->diskseq is non-zero, then we must have a disk seqnum */
321 r = sd_device_get_diskseq(device, &diskseq);
322 if (r < 0) {
323 log_debug_errno(r, "Dropping event because it has no diskseq, but waiting for %" PRIu64, w->diskseq);
324 return 0;
325 }
326 if (diskseq < w->diskseq) {
327 log_debug("Dropping event because diskseq too old (%" PRIu64 " < %" PRIu64 ")",
328 diskseq, w->diskseq);
329 return 0;
330 }
331 if (diskseq > w->diskseq) {
332 r = -EBUSY;
333 goto finish; /* Newer than what we were expecting, so we missed it, stop waiting */
334 }
335 } else if (w->uevent_seqnum_not_before != UINT64_MAX) {
75dc190d
LP		 336 uint64_t seqnum;
337
338 r = sd_device_get_seqnum(device, &seqnum);
339 if (r < 0)
340 goto finish;
341
342 if (seqnum <= w->uevent_seqnum_not_before) { /* From an older use of this loop device */
343 log_debug("Dropping event because seqnum too old (%" PRIu64 " <= %" PRIu64 ")",
344 seqnum, w->uevent_seqnum_not_before);
345 return 0;
346 }
347 }
348
0a8f9bc6 349 r = device_is_partition(device, w->parent_device, w->blkidp);
4ba86848
LP		 350 if (r < 0)
351 goto finish;
352 if (r == 0) /* Not the one we need */
353 return 0;
354
355 /* It's the one we need! Yay! */
356 assert(!w->found);
357 w->found = sd_device_ref(device);
358 r = 0;
359
360finish:
361 return sd_event_exit(sd_device_monitor_get_event(monitor), r);
ea887be0
ZJS		 362}
363
05c3c620
YW		 364static int timeout_handler(sd_event_source *s, uint64_t usec, void *userdata) {
365 struct wait_data *w = userdata;
366 int r;
367
368 assert(w);
369
370 /* Why partition not appeared within the timeout? We may lost some uevent, as some properties
371 * were not ready when we received uevent... Not sure, but anyway, let's try to find the
372 * partition again before give up. */
373
374 r = find_partition(w->parent_device, w->blkidp, w->timestamp_not_before, w->flags, &w->found);
375 if (r == -ENXIO)
376 return log_debug_errno(SYNTHETIC_ERRNO(ETIMEDOUT),
377 "Partition still not appeared after timeout reached.");
378 if (r < 0)
379 return log_debug_errno(r, "Failed to find partition: %m");
380
381 log_debug("Partition appeared after timeout reached.");
382 return sd_event_exit(sd_event_source_get_event(s), 0);
383}
384
61730746
YW		 385static int retry_handler(sd_event_source *s, uint64_t usec, void *userdata) {
386 struct wait_data *w = userdata;
387 int r;
388
389 assert(w);
390
391 r = find_partition(w->parent_device, w->blkidp, w->timestamp_not_before, w->flags, &w->found);
392 if (r != -ENXIO) {
393 if (r < 0)
394 return log_debug_errno(r, "Failed to find partition: %m");
395
396 log_debug("Partition found by a periodic search.");
397 return sd_event_exit(sd_event_source_get_event(s), 0);
398 }
399
400 r = sd_event_source_set_time_relative(s, 500 * USEC_PER_MSEC);
401 if (r < 0)
402 return r;
403
404 return sd_event_source_set_enabled(s, SD_EVENT_ONESHOT);
405}
406
4ba86848
LP		 407static int wait_for_partition_device(
408 sd_device *parent,
409 blkid_partition pp,
410 usec_t deadline,
a3642997 411 uint64_t diskseq,
75dc190d 412 uint64_t uevent_seqnum_not_before,
4a62257d 413 usec_t timestamp_not_before,
3fe398ce 414 DissectImageFlags flags,
4ba86848
LP		 415 sd_device **ret) {
416
61730746 417 _cleanup_(sd_event_source_unrefp) sd_event_source *timeout_source = NULL, *retry_source = NULL;
4ba86848
LP		 418 _cleanup_(sd_device_monitor_unrefp) sd_device_monitor *monitor = NULL;
419 _cleanup_(sd_event_unrefp) sd_event *event = NULL;
ea887be0
ZJS		 420 int r;
421
4ba86848
LP		 422 assert(parent);
423 assert(pp);
424 assert(ret);
425
3fe398ce 426 r = find_partition(parent, pp, timestamp_not_before, flags, ret);
4ba86848
LP		 427 if (r != -ENXIO)
428 return r;
429
430 r = sd_event_new(&event);
431 if (r < 0)
432 return r;
433
434 r = sd_device_monitor_new(&monitor);
435 if (r < 0)
436 return r;
437
438 r = sd_device_monitor_filter_add_match_subsystem_devtype(monitor, "block", "partition");
439 if (r < 0)
440 return r;
441
210e1cd6
YW		 442 r = sd_device_monitor_filter_add_match_parent(monitor, parent, true);
443 if (r < 0)
444 return r;
445
446 r = sd_device_monitor_filter_add_match_sysattr(monitor, "partition", NULL, true);
447 if (r < 0)
448 return r;
449
4ba86848
LP		 450 r = sd_device_monitor_attach_event(monitor, event);
451 if (r < 0)
452 return r;
453
454 _cleanup_(wait_data_done) struct wait_data w = {
455 .parent_device = parent,
456 .blkidp = pp,
a3642997 457 .diskseq = diskseq,
75dc190d 458 .uevent_seqnum_not_before = uevent_seqnum_not_before,
05c3c620
YW		 459 .timestamp_not_before = timestamp_not_before,
460 .flags = flags,
4ba86848 461 };
f70e7f70 462
4ba86848
LP		 463 r = sd_device_monitor_start(monitor, device_monitor_handler, &w);
464 if (r < 0)
465 return r;
a8040b6d 466
4ba86848 467 /* Check again, the partition might have appeared in the meantime */
3fe398ce 468 r = find_partition(parent, pp, timestamp_not_before, flags, ret);
4ba86848
LP		 469 if (r != -ENXIO)
470 return r;
471
472 if (deadline != USEC_INFINITY) {
473 r = sd_event_add_time(
474 event, &timeout_source,
475 CLOCK_MONOTONIC, deadline, 0,
05c3c620
YW		 476 timeout_handler, &w);
477 if (r < 0)
478 return r;
479
480 r = sd_event_source_set_exit_on_failure(timeout_source, true);
4ba86848 481 if (r < 0)
ea887be0
ZJS		 482 return r;
483 }
484
a3642997
LB		 485 /* If we don't have a disk sequence number then we cannot do exact matching,
486 * and we cannot know if we missed it or if it has not been sent yet, so set
487 * up additional retries to increase the chances of receiving the event. */
488 if (diskseq == 0) {
489 r = sd_event_add_time_relative(
490 event, &retry_source,
491 CLOCK_MONOTONIC, 500 * USEC_PER_MSEC, 0,
492 retry_handler, &w);
493 if (r < 0)
494 return r;
61730746 495
a3642997
LB		 496 r = sd_event_source_set_exit_on_failure(retry_source, true);
497 if (r < 0)
498 return r;
499 }
61730746 500
4ba86848
LP		 501 r = sd_event_loop(event);
502 if (r < 0)
503 return r;
504
505 assert(w.found);
506 *ret = TAKE_PTR(w.found);
507 return 0;
ea887be0
ZJS		 508}
509
0f7c9a3d
LP		 510static void check_partition_flags(
511 const char *node,
512 unsigned long long pflags,
513 unsigned long long supported) {
514
515 assert(node);
516
517 /* Mask away all flags supported by this partition's type and the three flags the UEFI spec defines generically */
518 pflags &= ~(supported | GPT_FLAG_REQUIRED_PARTITION | GPT_FLAG_NO_BLOCK_IO_PROTOCOL | GPT_FLAG_LEGACY_BIOS_BOOTABLE);
519
520 if (pflags == 0)
521 return;
522
523 /* If there are other bits set, then log about it, to make things discoverable */
524 for (unsigned i = 0; i < sizeof(pflags) * 8; i++) {
525 unsigned long long bit = 1ULL << i;
526 if (!FLAGS_SET(pflags, bit))
527 continue;
528
529 log_debug("Unexpected partition flag %llu set on %s!", bit, node);
530 }
531}
532
786e3a52
LP		 533static int device_wait_for_initialization_harder(
534 sd_device *device,
535 const char *subsystem,
536 usec_t deadline,
537 sd_device **ret) {
538
786e3a52
LP		 539 usec_t start, left, retrigger_timeout;
540 int r;
541
542 start = now(CLOCK_MONOTONIC);
543 left = usec_sub_unsigned(deadline, start);
544
545 if (DEBUG_LOGGING) {
786e3a52
LP		 546 const char *sn = NULL;
547
548 (void) sd_device_get_sysname(device, &sn);
b64c4ece 549 log_device_debug(device,
cc7ac73d 550 "Will wait up to %s for '%s' to initialize…", FORMAT_TIMESPAN(left, 0), strna(sn));
786e3a52
LP		 551 }
552
553 if (left != USEC_INFINITY)
554 retrigger_timeout = CLAMP(left / 4, 1 * USEC_PER_SEC, 5 * USEC_PER_SEC); /* A fourth of the total timeout, but let's clamp to 1s…5s range */
555 else
556 retrigger_timeout = 2 * USEC_PER_SEC;
557
558 for (;;) {
559 usec_t local_deadline, n;
560 bool last_try;
561
562 n = now(CLOCK_MONOTONIC);
563 assert(n >= start);
564
565 /* Find next deadline, when we'll retrigger */
566 local_deadline = start +
567 DIV_ROUND_UP(n - start, retrigger_timeout) * retrigger_timeout;
568
569 if (deadline != USEC_INFINITY && deadline <= local_deadline) {
570 local_deadline = deadline;
571 last_try = true;
572 } else
573 last_try = false;
574
575 r = device_wait_for_initialization(device, subsystem, local_deadline, ret);
576 if (r >= 0 && DEBUG_LOGGING) {
786e3a52
LP		 577 const char *sn = NULL;
578
579 (void) sd_device_get_sysname(device, &sn);
b64c4ece
LP		 580 log_device_debug(device,
581 "Successfully waited for device '%s' to initialize for %s.",
582 strna(sn),
5291f26d 583 FORMAT_TIMESPAN(usec_sub_unsigned(now(CLOCK_MONOTONIC), start), 0));
786e3a52
LP		 584
585 }
586 if (r != -ETIMEDOUT || last_try)
587 return r;
588
5291f26d 589 if (DEBUG_LOGGING)
b64c4ece
LP		 590 log_device_debug(device,
591 "Device didn't initialize within %s, assuming lost event. Retriggering device.",
5291f26d 592 FORMAT_TIMESPAN(usec_sub_unsigned(now(CLOCK_MONOTONIC), start), 0));
786e3a52 593
b64c4ece 594 r = sd_device_trigger(device, SD_DEVICE_CHANGE);
786e3a52
LP		 595 if (r < 0)
596 return r;
597 }
598}
40c10d3f 599#endif
aae22eb3 600
4ba86848
LP		 601#define DEVICE_TIMEOUT_USEC (45 * USEC_PER_SEC)
602
08fe0a53
LP		 603static void dissected_partition_done(DissectedPartition *p) {
604 assert(p);
605
606 free(p->fstype);
607 free(p->node);
608 free(p->label);
609 free(p->decrypted_fstype);
610 free(p->decrypted_node);
611 free(p->mount_options);
612
613 *p = (DissectedPartition) {
614 .partno = -1,
615 .architecture = -1
616 };
617}
618
4526113f
LP		 619int dissect_image(
620 int fd,
89e62e0b 621 const VeritySettings *verity,
18d73705 622 const MountOptions *mount_options,
a3642997 623 uint64_t diskseq,
75dc190d 624 uint64_t uevent_seqnum_not_before,
4a62257d 625 usec_t timestamp_not_before,
4526113f
LP		 626 DissectImageFlags flags,
627 DissectedImage **ret) {
8c1be37e 628
349cc4a5 629#if HAVE_BLKID
62ea0ed0 630 sd_id128_t root_uuid = SD_ID128_NULL, root_verity_uuid = SD_ID128_NULL;
62ea0ed0 631 sd_id128_t usr_uuid = SD_ID128_NULL, usr_verity_uuid = SD_ID128_NULL;
1f8fb21c 632 bool is_gpt, is_mbr, multiple_generic = false,
de98f631
LP		 633 generic_rw = false, /* initialize to appease gcc */
634 generic_growfs = false;
3c1f2cee 635 _cleanup_(sd_device_unrefp) sd_device *d = NULL;
8c1be37e 636 _cleanup_(dissected_image_unrefp) DissectedImage *m = NULL;
8e766630 637 _cleanup_(blkid_free_probep) blkid_probe b = NULL;
8c1be37e 638 _cleanup_free_ char *generic_node = NULL;
be30ad41 639 sd_id128_t generic_uuid = SD_ID128_NULL;
593fe6c0 640 const char *pttype = NULL, *sysname = NULL;
8c1be37e 641 blkid_partlist pl;
1f8fb21c 642 int r, generic_nr = -1, n_partitions;
8c1be37e 643 struct stat st;
4ba86848 644 usec_t deadline;
8c1be37e
LP		 645
646 assert(fd >= 0);
647 assert(ret);
a0bff7ea 648 assert(!verity || verity->designator < 0 || IN_SET(verity->designator, PARTITION_ROOT, PARTITION_USR));
89e62e0b 649 assert(!verity || verity->root_hash || verity->root_hash_size == 0);
a0bff7ea
LP		 650 assert(!verity || verity->root_hash_sig || verity->root_hash_sig_size == 0);
651 assert(!verity || (verity->root_hash || !verity->root_hash_sig));
e7cbe5cb 652 assert(!((flags & DISSECT_IMAGE_GPT_ONLY) && (flags & DISSECT_IMAGE_NO_PARTITION_TABLE)));
8c1be37e
LP		 653
654 /* Probes a disk image, and returns information about what it found in *ret.
655 *
4623e8e6 656 * Returns -ENOPKG if no suitable partition table or file system could be found.
2679f407
LP		 657 * Returns -EADDRNOTAVAIL if a root hash was specified but no matching root/verity partitions found.
658 * Returns -ENXIO if we couldn't find any partition suitable as root or /usr partition
659 * Returns -ENOTUNIQ if we only found multiple generic partitions and thus don't know what to do with that */
4623e8e6 660
89e62e0b 661 if (verity && verity->root_hash) {
aee36b4e
LP		 662 sd_id128_t fsuuid, vuuid;
663
664 /* If a root hash is supplied, then we use the root partition that has a UUID that match the
665 * first 128bit of the root hash. And we use the verity partition that has a UUID that match
666 * the final 128bit. */
4623e8e6 667
89e62e0b 668 if (verity->root_hash_size < sizeof(sd_id128_t))
4623e8e6
LP		 669 return -EINVAL;
670
aee36b4e
LP		 671 memcpy(&fsuuid, verity->root_hash, sizeof(sd_id128_t));
672 memcpy(&vuuid, (const uint8_t*) verity->root_hash + verity->root_hash_size - sizeof(sd_id128_t), sizeof(sd_id128_t));
4623e8e6 673
aee36b4e 674 if (sd_id128_is_null(fsuuid))
4623e8e6 675 return -EINVAL;
aee36b4e 676 if (sd_id128_is_null(vuuid))
4623e8e6 677 return -EINVAL;
aee36b4e
LP		 678
679 /* If the verity data declares it's for the /usr partition, then search for that, in all
680 * other cases assume it's for the root partition. */
681 if (verity->designator == PARTITION_USR) {
682 usr_uuid = fsuuid;
683 usr_verity_uuid = vuuid;
684 } else {
685 root_uuid = fsuuid;
686 root_verity_uuid = vuuid;
687 }
4623e8e6 688 }
8c1be37e
LP		 689
690 if (fstat(fd, &st) < 0)
691 return -errno;
692
693 if (!S_ISBLK(st.st_mode))
694 return -ENOTBLK;
695
930aa88f 696 r = sd_device_new_from_stat_rdev(&d, &st);
6c544d14
LP		 697 if (r < 0)
698 return r;
699
700 if (!FLAGS_SET(flags, DISSECT_IMAGE_NO_UDEV)) {
701 _cleanup_(sd_device_unrefp) sd_device *initialized = NULL;
702
703 /* If udev support is enabled, then let's wait for the device to be initialized before we doing anything. */
704
786e3a52
LP		 705 r = device_wait_for_initialization_harder(
706 d,
707 "block",
708 usec_add(now(CLOCK_MONOTONIC), DEVICE_TIMEOUT_USEC),
709 &initialized);
6c544d14
LP		 710 if (r < 0)
711 return r;
712
713 sd_device_unref(d);
714 d = TAKE_PTR(initialized);
715 }
716
8c1be37e
LP		 717 b = blkid_new_probe();
718 if (!b)
719 return -ENOMEM;
720
721 errno = 0;
722 r = blkid_probe_set_device(b, fd, 0, 0);
b382db9f 723 if (r != 0)
66855de7 724 return errno_or_else(ENOMEM);
8c1be37e 725
9b6deb03
LP		 726 if ((flags & DISSECT_IMAGE_GPT_ONLY) == 0) {
727 /* Look for file system superblocks, unless we only shall look for GPT partition tables */
728 blkid_probe_enable_superblocks(b, 1);
729 blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE|BLKID_SUBLKS_USAGE);
730 }
731
8c1be37e
LP		 732 blkid_probe_enable_partitions(b, 1);
733 blkid_probe_set_partitions_flags(b, BLKID_PARTS_ENTRY_DETAILS);
734
735 errno = 0;
736 r = blkid_do_safeprobe(b);
59ba6d0c
LP		 737 if (IN_SET(r, -2, 1))
738 return log_debug_errno(SYNTHETIC_ERRNO(ENOPKG), "Failed to identify any partition table.");
b382db9f 739 if (r != 0)
66855de7 740 return errno_or_else(EIO);
8c1be37e 741
a4e0d617 742 m = new(DissectedImage, 1);
8c1be37e
LP		 743 if (!m)
744 return -ENOMEM;
745
a4e0d617
LP		 746 *m = (DissectedImage) {
747 .has_init_system = -1,
748 };
749
593fe6c0
LB		 750 r = sd_device_get_sysname(d, &sysname);
751 if (r < 0)
752 return log_debug_errno(r, "Failed to get device sysname: %m");
753 if (startswith(sysname, "loop")) {
754 _cleanup_free_ char *name_stripped = NULL;
755 const char *full_path;
756
757 r = sd_device_get_sysattr_value(d, "loop/backing_file", &full_path);
758 if (r < 0)
759 log_debug_errno(r, "Failed to lookup image name via loop device backing file sysattr, ignoring: %m");
760 else {
761 r = raw_strip_suffixes(basename(full_path), &name_stripped);
762 if (r < 0)
763 return r;
764 }
765
766 free_and_replace(m->image_name, name_stripped);
767 } else {
768 r = free_and_strdup(&m->image_name, sysname);
769 if (r < 0)
770 return r;
771 }
772
773 if (!image_name_is_valid(m->image_name)) {
774 log_debug("Image name %s is not valid, ignoring", strempty(m->image_name));
775 m->image_name = mfree(m->image_name);
776 }
777
e7cbe5cb 778 if ((!(flags & DISSECT_IMAGE_GPT_ONLY) &&
4b5de5dd 779 (flags & DISSECT_IMAGE_GENERIC_ROOT)) ||
e7cbe5cb 780 (flags & DISSECT_IMAGE_NO_PARTITION_TABLE)) {
9b6deb03 781 const char *usage = NULL;
8c1be37e 782
aee36b4e
LP		 783 /* If flags permit this, also allow using non-partitioned single-filesystem images */
784
9b6deb03
LP		 785 (void) blkid_probe_lookup_value(b, "USAGE", &usage, NULL);
786 if (STRPTR_IN_SET(usage, "filesystem", "crypto")) {
6c544d14 787 const char *fstype = NULL, *options = NULL, *devname = NULL;
18d73705 788 _cleanup_free_ char *t = NULL, *n = NULL, *o = NULL;
8c1be37e 789
9b6deb03
LP		 790 /* OK, we have found a file system, that's our root partition then. */
791 (void) blkid_probe_lookup_value(b, "TYPE", &fstype, NULL);
8c1be37e 792
9b6deb03
LP		 793 if (fstype) {
794 t = strdup(fstype);
795 if (!t)
796 return -ENOMEM;
797 }
798
6c544d14 799 r = sd_device_get_devname(d, &devname);
54b22b26
LP		 800 if (r < 0)
801 return r;
8c1be37e 802
6c544d14
LP		 803 n = strdup(devname);
804 if (!n)
805 return -ENOMEM;
806
e7cbe5cb 807 m->single_file_system = true;
c3c88d67
LP		 808 m->encrypted = streq_ptr(fstype, "crypto_LUKS");
809
810 m->has_verity = verity && verity->data_path;
811 m->verity_ready = m->has_verity &&
812 verity->root_hash &&
813 (verity->designator < 0 || verity->designator == PARTITION_ROOT);
e7cbe5cb 814
8ee9615e
LP		 815 m->has_verity_sig = false; /* signature not embedded, must be specified */
816 m->verity_sig_ready = m->verity_ready &&
817 verity->root_hash_sig;
818
f5215bc8 819 options = mount_options_from_designator(mount_options, PARTITION_ROOT);
18d73705
LB		 820 if (options) {
821 o = strdup(options);
822 if (!o)
823 return -ENOMEM;
824 }
825
9b6deb03
LP		 826 m->partitions[PARTITION_ROOT] = (DissectedPartition) {
827 .found = true,
e0d53d52 828 .rw = !m->verity_ready && !fstype_is_ro(fstype),
9b6deb03
LP		 829 .partno = -1,
830 .architecture = _ARCHITECTURE_INVALID,
1cc6c93a
YW		 831 .fstype = TAKE_PTR(t),
832 .node = TAKE_PTR(n),
18d73705 833 .mount_options = TAKE_PTR(o),
88b3300f
LP		 834 .offset = 0,
835 .size = UINT64_MAX,
9b6deb03 836 };
8c1be37e 837
1cc6c93a 838 *ret = TAKE_PTR(m);
9b6deb03
LP		 839 return 0;
840 }
8c1be37e
LP		 841 }
842
843 (void) blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL);
844 if (!pttype)
845 return -ENOPKG;
846
847 is_gpt = streq_ptr(pttype, "gpt");
848 is_mbr = streq_ptr(pttype, "dos");
849
9b6deb03 850 if (!is_gpt && ((flags & DISSECT_IMAGE_GPT_ONLY) || !is_mbr))
8c1be37e
LP		 851 return -ENOPKG;
852
0903fd26
LP		 853 /* We support external verity data partitions only if the image has no partition table */
854 if (verity && verity->data_path)
855 return -EBADR;
856
4ba86848
LP		 857 /* Safety check: refuse block devices that carry a partition table but for which the kernel doesn't
858 * do partition scanning. */
859 r = blockdev_partscan_enabled(fd);
860 if (r < 0)
861 return r;
862 if (r == 0)
863 return -EPROTONOSUPPORT;
864
8c1be37e
LP		 865 errno = 0;
866 pl = blkid_probe_get_partitions(b);
b382db9f 867 if (!pl)
66855de7 868 return errno_or_else(ENOMEM);
8c1be37e 869
4ba86848
LP		 870 errno = 0;
871 n_partitions = blkid_partlist_numof_partitions(pl);
872 if (n_partitions < 0)
873 return errno_or_else(EIO);
8c1be37e 874
4ba86848
LP		 875 deadline = usec_add(now(CLOCK_MONOTONIC), DEVICE_TIMEOUT_USEC);
876 for (int i = 0; i < n_partitions; i++) {
877 _cleanup_(sd_device_unrefp) sd_device *q = NULL;
9b6deb03 878 unsigned long long pflags;
88b3300f 879 blkid_loff_t start, size;
8c1be37e 880 blkid_partition pp;
cde942f6 881 const char *node;
8c1be37e
LP		 882 int nr;
883
4ba86848
LP		 884 errno = 0;
885 pp = blkid_partlist_get_partition(pl, i);
886 if (!pp)
887 return errno_or_else(EIO);
aae22eb3 888
a3642997 889 r = wait_for_partition_device(d, pp, deadline, diskseq, uevent_seqnum_not_before, timestamp_not_before, flags, &q);
4ba86848
LP		 890 if (r < 0)
891 return r;
7be1420f 892
3c1f2cee
YW		 893 r = sd_device_get_devname(q, &node);
894 if (r < 0)
4ba86848 895 return r;
8c1be37e 896
9b6deb03 897 pflags = blkid_partition_get_flags(pp);
8c1be37e 898
4ba86848 899 errno = 0;
8c1be37e
LP		 900 nr = blkid_partition_get_partno(pp);
901 if (nr < 0)
4ba86848 902 return errno_or_else(EIO);
8c1be37e 903
88b3300f
LP		 904 errno = 0;
905 start = blkid_partition_get_start(pp);
906 if (start < 0)
907 return errno_or_else(EIO);
908
909 assert((uint64_t) start < UINT64_MAX/512);
910
911 errno = 0;
912 size = blkid_partition_get_size(pp);
913 if (size < 0)
914 return errno_or_else(EIO);
915
916 assert((uint64_t) size < UINT64_MAX/512);
917
8c1be37e 918 if (is_gpt) {
569a0e42
LP		 919 PartitionDesignator designator = _PARTITION_DESIGNATOR_INVALID;
920 int architecture = _ARCHITECTURE_INVALID;
08fe0a53 921 const char *stype, *sid, *fstype = NULL, *label;
4623e8e6 922 sd_id128_t type_id, id;
de98f631 923 bool rw = true, growfs = false;
8c1be37e 924
4623e8e6
LP		 925 sid = blkid_partition_get_uuid(pp);
926 if (!sid)
927 continue;
928 if (sd_id128_from_string(sid, &id) < 0)
929 continue;
930
8c1be37e
LP		 931 stype = blkid_partition_get_type_string(pp);
932 if (!stype)
933 continue;
8c1be37e
LP		 934 if (sd_id128_from_string(stype, &type_id) < 0)
935 continue;
936
08fe0a53
LP		 937 label = blkid_partition_get_name(pp); /* libblkid returns NULL here if empty */
938
8c1be37e 939 if (sd_id128_equal(type_id, GPT_HOME)) {
a48dd347 940
de98f631 941 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO|GPT_FLAG_READ_ONLY|GPT_FLAG_GROWFS);
0f7c9a3d 942
a48dd347
LP		 943 if (pflags & GPT_FLAG_NO_AUTO)
944 continue;
945
8c1be37e 946 designator = PARTITION_HOME;
9b6deb03 947 rw = !(pflags & GPT_FLAG_READ_ONLY);
de98f631 948 growfs = FLAGS_SET(pflags, GPT_FLAG_GROWFS);
aee36b4e 949
8c1be37e 950 } else if (sd_id128_equal(type_id, GPT_SRV)) {
a48dd347 951
de98f631 952 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO|GPT_FLAG_READ_ONLY|GPT_FLAG_GROWFS);
0f7c9a3d 953
a48dd347
LP		 954 if (pflags & GPT_FLAG_NO_AUTO)
955 continue;
956
8c1be37e 957 designator = PARTITION_SRV;
9b6deb03 958 rw = !(pflags & GPT_FLAG_READ_ONLY);
de98f631 959 growfs = FLAGS_SET(pflags, GPT_FLAG_GROWFS);
aee36b4e 960
8c1be37e 961 } else if (sd_id128_equal(type_id, GPT_ESP)) {
a48dd347 962
aee36b4e
LP		 963 /* Note that we don't check the GPT_FLAG_NO_AUTO flag for the ESP, as it is
964 * not defined there. We instead check the GPT_FLAG_NO_BLOCK_IO_PROTOCOL, as
965 * recommended by the UEFI spec (See "12.3.3 Number and Location of System
966 * Partitions"). */
a48dd347
LP		 967
968 if (pflags & GPT_FLAG_NO_BLOCK_IO_PROTOCOL)
969 continue;
970
8c1be37e
LP		 971 designator = PARTITION_ESP;
972 fstype = "vfat";
a8c47660
LP		 973
974 } else if (sd_id128_equal(type_id, GPT_XBOOTLDR)) {
975
de98f631 976 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO|GPT_FLAG_READ_ONLY|GPT_FLAG_GROWFS);
0f7c9a3d 977
a8c47660
LP		 978 if (pflags & GPT_FLAG_NO_AUTO)
979 continue;
980
981 designator = PARTITION_XBOOTLDR;
982 rw = !(pflags & GPT_FLAG_READ_ONLY);
de98f631 983 growfs = FLAGS_SET(pflags, GPT_FLAG_GROWFS);
49ae9d91
DDM		 984
985 } else if (gpt_partition_type_is_root(type_id)) {
4623e8e6 986
de98f631 987 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO|GPT_FLAG_READ_ONLY|GPT_FLAG_GROWFS);
0f7c9a3d 988
a48dd347
LP		 989 if (pflags & GPT_FLAG_NO_AUTO)
990 continue;
991
4623e8e6
LP		 992 /* If a root ID is specified, ignore everything but the root id */
993 if (!sd_id128_is_null(root_uuid) && !sd_id128_equal(root_uuid, id))
994 continue;
995
49ae9d91
DDM		 996 assert_se((architecture = gpt_partition_type_uuid_to_arch(type_id)) >= 0);
997 designator = PARTITION_ROOT_OF_ARCH(architecture);
9b6deb03 998 rw = !(pflags & GPT_FLAG_READ_ONLY);
de98f631 999 growfs = FLAGS_SET(pflags, GPT_FLAG_GROWFS);
aee36b4e 1000
49ae9d91 1001 } else if (gpt_partition_type_is_root_verity(type_id)) {
4623e8e6 1002
0f7c9a3d
LP		 1003 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO|GPT_FLAG_READ_ONLY);
1004
a48dd347
LP		 1005 if (pflags & GPT_FLAG_NO_AUTO)
1006 continue;
1007
c3c88d67 1008 m->has_verity = true;
4623e8e6 1009
8ee9615e
LP		 1010 /* If no verity configuration is specified, then don't do verity */
1011 if (!verity)
1012 continue;
1013 if (verity->designator >= 0 && verity->designator != PARTITION_ROOT)
1014 continue;
1015
1016 /* If root hash is specified, then ignore everything but the root id */
1017 if (!sd_id128_is_null(root_verity_uuid) && !sd_id128_equal(root_verity_uuid, id))
4623e8e6
LP		 1018 continue;
1019
49ae9d91
DDM		 1020 assert_se((architecture = gpt_partition_type_uuid_to_arch(type_id)) >= 0);
1021 designator = PARTITION_VERITY_OF(PARTITION_ROOT_OF_ARCH(architecture));
4623e8e6 1022 fstype = "DM_verity_hash";
4623e8e6 1023 rw = false;
8ee9615e 1024
49ae9d91 1025 } else if (gpt_partition_type_is_root_verity_sig(type_id)) {
8ee9615e
LP		 1026
1027 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO|GPT_FLAG_READ_ONLY);
1028
1029 if (pflags & GPT_FLAG_NO_AUTO)
1030 continue;
1031
1032 m->has_verity_sig = true;
1033
1034 /* If root hash is specified explicitly, then ignore any embedded signature */
1035 if (!verity)
1036 continue;
1037 if (verity->designator >= 0 && verity->designator != PARTITION_ROOT)
1038 continue;
1039 if (verity->root_hash)
1040 continue;
1041
49ae9d91
DDM		 1042 assert_se((architecture = gpt_partition_type_uuid_to_arch(type_id)) >= 0);
1043 designator = PARTITION_VERITY_SIG_OF(PARTITION_ROOT_OF_ARCH(architecture));
8ee9615e 1044 fstype = "verity_hash_signature";
4623e8e6 1045 rw = false;
8ee9615e 1046
49ae9d91 1047 } else if (gpt_partition_type_is_usr(type_id)) {
aee36b4e 1048
de98f631 1049 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO|GPT_FLAG_READ_ONLY|GPT_FLAG_GROWFS);
aee36b4e
LP		 1050
1051 if (pflags & GPT_FLAG_NO_AUTO)
1052 continue;
1053
1054 /* If a usr ID is specified, ignore everything but the usr id */
1055 if (!sd_id128_is_null(usr_uuid) && !sd_id128_equal(usr_uuid, id))
1056 continue;
1057
49ae9d91
DDM		 1058 assert_se((architecture = gpt_partition_type_uuid_to_arch(type_id)) >= 0);
1059 designator = PARTITION_USR_OF_ARCH(architecture);
aee36b4e 1060 rw = !(pflags & GPT_FLAG_READ_ONLY);
de98f631 1061 growfs = FLAGS_SET(pflags, GPT_FLAG_GROWFS);
aee36b4e 1062
49ae9d91 1063 } else if (gpt_partition_type_is_usr_verity(type_id)) {
aee36b4e
LP		 1064
1065 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO|GPT_FLAG_READ_ONLY);
1066
1067 if (pflags & GPT_FLAG_NO_AUTO)
1068 continue;
1069
c3c88d67 1070 m->has_verity = true;
aee36b4e 1071
8ee9615e
LP		 1072 if (!verity)
1073 continue;
1074 if (verity->designator >= 0 && verity->designator != PARTITION_USR)
1075 continue;
1076
1077 /* If usr hash is specified, then ignore everything but the usr id */
1078 if (!sd_id128_is_null(usr_verity_uuid) && !sd_id128_equal(usr_verity_uuid, id))
aee36b4e
LP		 1079 continue;
1080
49ae9d91
DDM		 1081 assert_se((architecture = gpt_partition_type_uuid_to_arch(type_id)) >= 0);
1082 designator = PARTITION_VERITY_OF(PARTITION_USR_OF_ARCH(architecture));
aee36b4e 1083 fstype = "DM_verity_hash";
aee36b4e 1084 rw = false;
8ee9615e 1085
49ae9d91 1086 } else if (gpt_partition_type_is_usr_verity_sig(type_id)) {
8ee9615e
LP		 1087
1088 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO|GPT_FLAG_READ_ONLY);
1089
1090 if (pflags & GPT_FLAG_NO_AUTO)
1091 continue;
1092
1093 m->has_verity_sig = true;
1094
1095 /* If usr hash is specified explicitly, then ignore any embedded signature */
1096 if (!verity)
1097 continue;
1098 if (verity->designator >= 0 && verity->designator != PARTITION_USR)
1099 continue;
1100 if (verity->root_hash)
1101 continue;
1102
49ae9d91
DDM		 1103 assert_se((architecture = gpt_partition_type_uuid_to_arch(type_id)) >= 0);
1104 designator = PARTITION_VERITY_SIG_OF(PARTITION_USR_OF_ARCH(architecture));
8ee9615e 1105 fstype = "verity_hash_signature";
aee36b4e 1106 rw = false;
8ee9615e 1107
49ae9d91 1108 } else if (sd_id128_equal(type_id, GPT_SWAP)) {
a48dd347 1109
0f7c9a3d
LP		 1110 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO);
1111
a48dd347
LP		 1112 if (pflags & GPT_FLAG_NO_AUTO)
1113 continue;
1114
8c1be37e 1115 designator = PARTITION_SWAP;
aee36b4e 1116
8c1be37e
LP		 1117 } else if (sd_id128_equal(type_id, GPT_LINUX_GENERIC)) {
1118
de98f631 1119 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO|GPT_FLAG_READ_ONLY|GPT_FLAG_GROWFS);
0f7c9a3d 1120
a48dd347
LP		 1121 if (pflags & GPT_FLAG_NO_AUTO)
1122 continue;
1123
8c1be37e
LP		 1124 if (generic_node)
1125 multiple_generic = true;
1126 else {
1127 generic_nr = nr;
9b6deb03 1128 generic_rw = !(pflags & GPT_FLAG_READ_ONLY);
de98f631 1129 generic_growfs = FLAGS_SET(pflags, GPT_FLAG_GROWFS);
be30ad41 1130 generic_uuid = id;
8c1be37e
LP		 1131 generic_node = strdup(node);
1132 if (!generic_node)
1133 return -ENOMEM;
1134 }
d4dffb85
LP		 1135
1136 } else if (sd_id128_equal(type_id, GPT_TMP)) {
1137
de98f631 1138 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO|GPT_FLAG_READ_ONLY|GPT_FLAG_GROWFS);
0f7c9a3d 1139
d4dffb85
LP		 1140 if (pflags & GPT_FLAG_NO_AUTO)
1141 continue;
1142
1143 designator = PARTITION_TMP;
1144 rw = !(pflags & GPT_FLAG_READ_ONLY);
de98f631 1145 growfs = FLAGS_SET(pflags, GPT_FLAG_GROWFS);
d4dffb85
LP		 1146
1147 } else if (sd_id128_equal(type_id, GPT_VAR)) {
1148
de98f631 1149 check_partition_flags(node, pflags, GPT_FLAG_NO_AUTO|GPT_FLAG_READ_ONLY|GPT_FLAG_GROWFS);
0f7c9a3d 1150
d4dffb85
LP		 1151 if (pflags & GPT_FLAG_NO_AUTO)
1152 continue;
1153
1154 if (!FLAGS_SET(flags, DISSECT_IMAGE_RELAX_VAR_CHECK)) {
1155 sd_id128_t var_uuid;
1156
1157 /* For /var we insist that the uuid of the partition matches the
1158 * HMAC-SHA256 of the /var GPT partition type uuid, keyed by machine
1159 * ID. Why? Unlike the other partitions /var is inherently
1160 * installation specific, hence we need to be careful not to mount it
1161 * in the wrong installation. By hashing the partition UUID from
1162 * /etc/machine-id we can securely bind the partition to the
1163 * installation. */
1164
1165 r = sd_id128_get_machine_app_specific(GPT_VAR, &var_uuid);
1166 if (r < 0)
1167 return r;
1168
1169 if (!sd_id128_equal(var_uuid, id)) {
1170 log_debug("Found a /var/ partition, but its UUID didn't match our expectations, ignoring.");
1171 continue;
1172 }
1173 }
1174
1175 designator = PARTITION_VAR;
1176 rw = !(pflags & GPT_FLAG_READ_ONLY);
de98f631 1177 growfs = FLAGS_SET(pflags, GPT_FLAG_GROWFS);
8c1be37e
LP		 1178 }
1179
1180 if (designator != _PARTITION_DESIGNATOR_INVALID) {
08fe0a53 1181 _cleanup_free_ char *t = NULL, *n = NULL, *o = NULL, *l = NULL;
18d73705 1182 const char *options = NULL;
8c1be37e 1183
08fe0a53
LP		 1184 if (m->partitions[designator].found) {
1185 /* For most partition types the first one we see wins. Except for the
1186 * rootfs and /usr, where we do a version compare of the label, and
1187 * let the newest version win. This permits a simple A/B versioning
1188 * scheme in OS images. */
1189
1190 if (!PARTITION_DESIGNATOR_VERSIONED(designator) ||
1191 strverscmp_improved(m->partitions[designator].label, label) >= 0)
1192 continue;
1193
1194 dissected_partition_done(m->partitions + designator);
1195 }
8c1be37e
LP		 1196
1197 if (fstype) {
1198 t = strdup(fstype);
1199 if (!t)
1200 return -ENOMEM;
1201 }
1202
1203 n = strdup(node);
1204 if (!n)
1205 return -ENOMEM;
1206
08fe0a53
LP		 1207 if (label) {
1208 l = strdup(label);
1209 if (!l)
1210 return -ENOMEM;
1211 }
1212
f5215bc8 1213 options = mount_options_from_designator(mount_options, designator);
18d73705
LB		 1214 if (options) {
1215 o = strdup(options);
1216 if (!o)
1217 return -ENOMEM;
1218 }
1219
8c1be37e
LP		 1220 m->partitions[designator] = (DissectedPartition) {
1221 .found = true,
1222 .partno = nr,
1223 .rw = rw,
de98f631 1224 .growfs = growfs,
8c1be37e 1225 .architecture = architecture,
1cc6c93a
YW		 1226 .node = TAKE_PTR(n),
1227 .fstype = TAKE_PTR(t),
08fe0a53 1228 .label = TAKE_PTR(l),
be30ad41 1229 .uuid = id,
18d73705 1230 .mount_options = TAKE_PTR(o),
88b3300f
LP		 1231 .offset = (uint64_t) start * 512,
1232 .size = (uint64_t) size * 512,
8c1be37e 1233 };
8c1be37e
LP		 1234 }
1235
1236 } else if (is_mbr) {
1237
a8c47660 1238 switch (blkid_partition_get_type(pp)) {
8c1be37e 1239
a8c47660
LP		 1240 case 0x83: /* Linux partition */
1241
1242 if (pflags != 0x80) /* Bootable flag */
1243 continue;
8c1be37e 1244
a8c47660
LP		 1245 if (generic_node)
1246 multiple_generic = true;
1247 else {
1248 generic_nr = nr;
1249 generic_rw = true;
de98f631 1250 generic_growfs = false;
a8c47660
LP		 1251 generic_node = strdup(node);
1252 if (!generic_node)
1253 return -ENOMEM;
1254 }
1255
1256 break;
1257
1258 case 0xEA: { /* Boot Loader Spec extended $BOOT partition */
18d73705 1259 _cleanup_free_ char *n = NULL, *o = NULL;
a8c47660 1260 sd_id128_t id = SD_ID128_NULL;
18d73705 1261 const char *sid, *options = NULL;
a8c47660
LP		 1262
1263 /* First one wins */
1264 if (m->partitions[PARTITION_XBOOTLDR].found)
1265 continue;
1266
1267 sid = blkid_partition_get_uuid(pp);
1268 if (sid)
1269 (void) sd_id128_from_string(sid, &id);
1270
1271 n = strdup(node);
1272 if (!n)
8c1be37e 1273 return -ENOMEM;
a8c47660 1274
f5215bc8 1275 options = mount_options_from_designator(mount_options, PARTITION_XBOOTLDR);
18d73705
LB		 1276 if (options) {
1277 o = strdup(options);
1278 if (!o)
1279 return -ENOMEM;
1280 }
1281
a8c47660
LP		 1282 m->partitions[PARTITION_XBOOTLDR] = (DissectedPartition) {
1283 .found = true,
1284 .partno = nr,
1285 .rw = true,
de98f631 1286 .growfs = false,
a8c47660
LP		 1287 .architecture = _ARCHITECTURE_INVALID,
1288 .node = TAKE_PTR(n),
1289 .uuid = id,
18d73705 1290 .mount_options = TAKE_PTR(o),
88b3300f
LP		 1291 .offset = (uint64_t) start * 512,
1292 .size = (uint64_t) size * 512,
a8c47660
LP		 1293 };
1294
1295 break;
1296 }}
8c1be37e
LP		 1297 }
1298 }
1299
74cb2db9 1300 if (m->partitions[PARTITION_ROOT].found) {
49ae9d91
DDM		 1301 /* If we found the primary arch, then invalidate the secondary and other arch to avoid any
1302 * ambiguities, since we never want to mount the secondary or other arch in this case. */
74cb2db9
LP		 1303 m->partitions[PARTITION_ROOT_SECONDARY].found = false;
1304 m->partitions[PARTITION_ROOT_SECONDARY_VERITY].found = false;
8ee9615e 1305 m->partitions[PARTITION_ROOT_SECONDARY_VERITY_SIG].found = false;
aee36b4e
LP		 1306 m->partitions[PARTITION_USR_SECONDARY].found = false;
1307 m->partitions[PARTITION_USR_SECONDARY_VERITY].found = false;
8ee9615e 1308 m->partitions[PARTITION_USR_SECONDARY_VERITY_SIG].found = false;
8c1be37e 1309
49ae9d91
DDM		 1310 m->partitions[PARTITION_ROOT_OTHER].found = false;
1311 m->partitions[PARTITION_ROOT_OTHER_VERITY].found = false;
1312 m->partitions[PARTITION_ROOT_OTHER_VERITY_SIG].found = false;
1313 m->partitions[PARTITION_USR_OTHER].found = false;
1314 m->partitions[PARTITION_USR_OTHER_VERITY].found = false;
1315 m->partitions[PARTITION_USR_OTHER_VERITY_SIG].found = false;
1316
8ee9615e
LP		 1317 } else if (m->partitions[PARTITION_ROOT_VERITY].found ||
1318 m->partitions[PARTITION_ROOT_VERITY_SIG].found)
7cf66030 1319 return -EADDRNOTAVAIL; /* Verity found but no matching rootfs? Something is off, refuse. */
4623e8e6 1320
7cf66030 1321 else if (m->partitions[PARTITION_ROOT_SECONDARY].found) {
aee36b4e 1322
7cf66030 1323 /* No root partition found but there's one for the secondary architecture? Then upgrade
49ae9d91
DDM		 1324 * secondary arch to first and invalidate the other arch. */
1325
1326 log_debug("No root partition found of the native architecture, falling back to a root "
1327 "partition of the secondary architecture.");
4623e8e6 1328
7cf66030
LP		 1329 m->partitions[PARTITION_ROOT] = m->partitions[PARTITION_ROOT_SECONDARY];
1330 zero(m->partitions[PARTITION_ROOT_SECONDARY]);
1331 m->partitions[PARTITION_ROOT_VERITY] = m->partitions[PARTITION_ROOT_SECONDARY_VERITY];
1332 zero(m->partitions[PARTITION_ROOT_SECONDARY_VERITY]);
8ee9615e
LP		 1333 m->partitions[PARTITION_ROOT_VERITY_SIG] = m->partitions[PARTITION_ROOT_SECONDARY_VERITY_SIG];
1334 zero(m->partitions[PARTITION_ROOT_SECONDARY_VERITY_SIG]);
aee36b4e 1335
7cf66030
LP		 1336 m->partitions[PARTITION_USR] = m->partitions[PARTITION_USR_SECONDARY];
1337 zero(m->partitions[PARTITION_USR_SECONDARY]);
1338 m->partitions[PARTITION_USR_VERITY] = m->partitions[PARTITION_USR_SECONDARY_VERITY];
1339 zero(m->partitions[PARTITION_USR_SECONDARY_VERITY]);
8ee9615e
LP		 1340 m->partitions[PARTITION_USR_VERITY_SIG] = m->partitions[PARTITION_USR_SECONDARY_VERITY_SIG];
1341 zero(m->partitions[PARTITION_USR_SECONDARY_VERITY_SIG]);
e0f9e7bd 1342
49ae9d91
DDM		 1343 m->partitions[PARTITION_ROOT_OTHER].found = false;
1344 m->partitions[PARTITION_ROOT_OTHER_VERITY].found = false;
1345 m->partitions[PARTITION_ROOT_OTHER_VERITY_SIG].found = false;
1346 m->partitions[PARTITION_USR_OTHER].found = false;
1347 m->partitions[PARTITION_USR_OTHER_VERITY].found = false;
1348 m->partitions[PARTITION_USR_OTHER_VERITY_SIG].found = false;
1349
8ee9615e
LP		 1350 } else if (m->partitions[PARTITION_ROOT_SECONDARY_VERITY].found ||
1351 m->partitions[PARTITION_ROOT_SECONDARY_VERITY_SIG].found)
7cf66030 1352 return -EADDRNOTAVAIL; /* as above */
18d73705 1353
49ae9d91
DDM		 1354 else if (m->partitions[PARTITION_ROOT_OTHER].found) {
1355
1356 /* No root or secondary partition found but there's one for another architecture? Then
1357 * upgrade the other architecture to first. */
1358
1359 log_debug("No root partition found of the native architecture or the secondary architecture, "
1360 "falling back to a root partition of a non-native architecture (%s).",
1361 architecture_to_string(m->partitions[PARTITION_ROOT_OTHER].architecture));
1362
1363 m->partitions[PARTITION_ROOT] = m->partitions[PARTITION_ROOT_OTHER];
1364 zero(m->partitions[PARTITION_ROOT_OTHER]);
1365 m->partitions[PARTITION_ROOT_VERITY] = m->partitions[PARTITION_ROOT_OTHER_VERITY];
1366 zero(m->partitions[PARTITION_ROOT_OTHER_VERITY]);
1367 m->partitions[PARTITION_ROOT_VERITY_SIG] = m->partitions[PARTITION_ROOT_OTHER_VERITY_SIG];
1368 zero(m->partitions[PARTITION_ROOT_OTHER_VERITY_SIG]);
1369
1370 m->partitions[PARTITION_USR] = m->partitions[PARTITION_USR_OTHER];
1371 zero(m->partitions[PARTITION_USR_OTHER]);
1372 m->partitions[PARTITION_USR_VERITY] = m->partitions[PARTITION_USR_OTHER_VERITY];
1373 zero(m->partitions[PARTITION_USR_OTHER_VERITY]);
1374 m->partitions[PARTITION_USR_VERITY_SIG] = m->partitions[PARTITION_USR_OTHER_VERITY_SIG];
1375 zero(m->partitions[PARTITION_USR_OTHER_VERITY_SIG]);
1376 }
1377
8ee9615e
LP		 1378 /* Hmm, we found a signature partition but no Verity data? Something is off. */
1379 if (m->partitions[PARTITION_ROOT_VERITY_SIG].found && !m->partitions[PARTITION_ROOT_VERITY].found)
1380 return -EADDRNOTAVAIL;
7cf66030 1381
8ee9615e 1382 if (m->partitions[PARTITION_USR].found) {
49ae9d91 1383 /* Invalidate secondary and other arch /usr/ if we found the primary arch */
7cf66030
LP		 1384 m->partitions[PARTITION_USR_SECONDARY].found = false;
1385 m->partitions[PARTITION_USR_SECONDARY_VERITY].found = false;
8ee9615e 1386 m->partitions[PARTITION_USR_SECONDARY_VERITY_SIG].found = false;
7cf66030 1387
49ae9d91
DDM		 1388 m->partitions[PARTITION_USR_OTHER].found = false;
1389 m->partitions[PARTITION_USR_OTHER_VERITY].found = false;
1390 m->partitions[PARTITION_USR_OTHER_VERITY_SIG].found = false;
1391
8ee9615e
LP		 1392 } else if (m->partitions[PARTITION_USR_VERITY].found ||
1393 m->partitions[PARTITION_USR_VERITY_SIG].found)
7cf66030 1394 return -EADDRNOTAVAIL; /* as above */
8c1be37e 1395
7cf66030 1396 else if (m->partitions[PARTITION_USR_SECONDARY].found) {
e0f9e7bd 1397
49ae9d91
DDM		 1398 log_debug("No usr partition found of the native architecture, falling back to a usr "
1399 "partition of the secondary architecture.");
1400
7cf66030
LP		 1401 /* Upgrade secondary arch to primary */
1402 m->partitions[PARTITION_USR] = m->partitions[PARTITION_USR_SECONDARY];
1403 zero(m->partitions[PARTITION_USR_SECONDARY]);
1404 m->partitions[PARTITION_USR_VERITY] = m->partitions[PARTITION_USR_SECONDARY_VERITY];
1405 zero(m->partitions[PARTITION_USR_SECONDARY_VERITY]);
8ee9615e
LP		 1406 m->partitions[PARTITION_USR_VERITY_SIG] = m->partitions[PARTITION_USR_SECONDARY_VERITY_SIG];
1407 zero(m->partitions[PARTITION_USR_SECONDARY_VERITY_SIG]);
7cf66030 1408
49ae9d91
DDM		 1409 m->partitions[PARTITION_USR_OTHER].found = false;
1410 m->partitions[PARTITION_USR_OTHER_VERITY].found = false;
1411 m->partitions[PARTITION_USR_OTHER_VERITY_SIG].found = false;
1412
8ee9615e
LP		 1413 } else if (m->partitions[PARTITION_USR_SECONDARY_VERITY].found ||
1414 m->partitions[PARTITION_USR_SECONDARY_VERITY_SIG].found)
7cf66030
LP		 1415 return -EADDRNOTAVAIL; /* as above */
1416
49ae9d91
DDM		 1417 else if (m->partitions[PARTITION_USR_OTHER].found) {
1418
1419 log_debug("No usr partition found of the native architecture or the secondary architecture, "
1420 "falling back to a usr partition of a non-native architecture (%s).",
1421 architecture_to_string(m->partitions[PARTITION_ROOT_OTHER].architecture));
1422
1423 /* Upgrade other arch to primary */
1424 m->partitions[PARTITION_USR] = m->partitions[PARTITION_USR_OTHER];
1425 zero(m->partitions[PARTITION_USR_OTHER]);
1426 m->partitions[PARTITION_USR_VERITY] = m->partitions[PARTITION_USR_OTHER_VERITY];
1427 zero(m->partitions[PARTITION_USR_OTHER_VERITY]);
1428 m->partitions[PARTITION_USR_VERITY_SIG] = m->partitions[PARTITION_USR_OTHER_VERITY_SIG];
1429 zero(m->partitions[PARTITION_USR_OTHER_VERITY_SIG]);
1430 }
1431
8ee9615e
LP		 1432 /* Hmm, we found a signature partition but no Verity data? Something is off. */
1433 if (m->partitions[PARTITION_USR_VERITY_SIG].found && !m->partitions[PARTITION_USR_VERITY].found)
1434 return -EADDRNOTAVAIL;
1435
cb241a69
LP		 1436 /* If root and /usr are combined then insist that the architecture matches */
1437 if (m->partitions[PARTITION_ROOT].found &&
1438 m->partitions[PARTITION_USR].found &&
1439 (m->partitions[PARTITION_ROOT].architecture >= 0 &&
1440 m->partitions[PARTITION_USR].architecture >= 0 &&
1441 m->partitions[PARTITION_ROOT].architecture != m->partitions[PARTITION_USR].architecture))
1442 return -EADDRNOTAVAIL;
1443
4ab51780
LP		 1444 if (!m->partitions[PARTITION_ROOT].found &&
1445 !m->partitions[PARTITION_USR].found &&
1446 (flags & DISSECT_IMAGE_GENERIC_ROOT) &&
00a8b34f 1447 (!verity || !verity->root_hash || verity->designator != PARTITION_USR)) {
7cf66030
LP		 1448
1449 /* OK, we found nothing usable, then check if there's a single generic one distro, and use
4b5de5dd
LP		 1450 * that. If the root hash was set however, then we won't fall back to a generic node, because
1451 * the root hash decides. */
7cf66030
LP		 1452
1453 /* If we didn't find a properly marked root partition, but we did find a single suitable
1454 * generic Linux partition, then use this as root partition, if the caller asked for it. */
1455 if (multiple_generic)
1456 return -ENOTUNIQ;
1457
4b5de5dd
LP		 1458 /* If we didn't find a generic node, then we can't fix this up either */
1459 if (generic_node) {
1460 _cleanup_free_ char *o = NULL;
1461 const char *options;
8c1be37e 1462
f5215bc8 1463 options = mount_options_from_designator(mount_options, PARTITION_ROOT);
18d73705
LB		 1464 if (options) {
1465 o = strdup(options);
1466 if (!o)
1467 return -ENOMEM;
1468 }
1469
1f8fb21c 1470 assert(generic_nr >= 0);
8c1be37e
LP		 1471 m->partitions[PARTITION_ROOT] = (DissectedPartition) {
1472 .found = true,
1473 .rw = generic_rw,
de98f631 1474 .growfs = generic_growfs,
8c1be37e
LP		 1475 .partno = generic_nr,
1476 .architecture = _ARCHITECTURE_INVALID,
1cc6c93a 1477 .node = TAKE_PTR(generic_node),
be30ad41 1478 .uuid = generic_uuid,
18d73705 1479 .mount_options = TAKE_PTR(o),
88b3300f
LP		 1480 .offset = UINT64_MAX,
1481 .size = UINT64_MAX,
8c1be37e 1482 };
e0f9e7bd 1483 }
8c1be37e
LP		 1484 }
1485
4b5de5dd
LP		 1486 /* Check if we have a root fs if we are told to do check. /usr alone is fine too, but only if appropriate flag for that is set too */
1487 if (FLAGS_SET(flags, DISSECT_IMAGE_REQUIRE_ROOT) &&
1488 !(m->partitions[PARTITION_ROOT].found || (m->partitions[PARTITION_USR].found && FLAGS_SET(flags, DISSECT_IMAGE_USR_NO_ROOT))))
1489 return -ENXIO;
1490
7b32164f
LP		 1491 if (m->partitions[PARTITION_ROOT_VERITY].found) {
1492 /* We only support one verity partition per image, i.e. can't do for both /usr and root fs */
1493 if (m->partitions[PARTITION_USR_VERITY].found)
1494 return -ENOTUNIQ;
1495
1496 /* We don't support verity enabled root with a split out /usr. Neither with nor without
1497 * verity there. (Note that we do support verity-less root with verity-full /usr, though.) */
1498 if (m->partitions[PARTITION_USR].found)
1499 return -EADDRNOTAVAIL;
1500 }
aee36b4e 1501
1903defc
LP		 1502 if (verity) {
1503 /* If a verity designator is specified, then insist that the matching partition exists */
1504 if (verity->designator >= 0 && !m->partitions[verity->designator].found)
1505 return -EADDRNOTAVAIL;
aee36b4e 1506
1903defc 1507 if (verity->root_hash) {
8ee9615e
LP		 1508 /* If we have an explicit root hash and found the partitions for it, then we are ready to use
1509 * Verity, set things up for it */
1510
1903defc
LP		 1511 if (verity->designator < 0 || verity->designator == PARTITION_ROOT) {
1512 if (!m->partitions[PARTITION_ROOT_VERITY].found || !m->partitions[PARTITION_ROOT].found)
1513 return -EADDRNOTAVAIL;
4623e8e6 1514
1903defc
LP		 1515 /* If we found a verity setup, then the root partition is necessarily read-only. */
1516 m->partitions[PARTITION_ROOT].rw = false;
1517 m->verity_ready = true;
1903defc 1518
f9e0bb61
LP		 1519 } else {
1520 assert(verity->designator == PARTITION_USR);
1521
1903defc
LP		 1522 if (!m->partitions[PARTITION_USR_VERITY].found || !m->partitions[PARTITION_USR].found)
1523 return -EADDRNOTAVAIL;
4623e8e6 1524
1903defc
LP		 1525 m->partitions[PARTITION_USR].rw = false;
1526 m->verity_ready = true;
1527 }
8ee9615e
LP		 1528
1529 if (m->verity_ready)
b98416e1 1530 m->verity_sig_ready = verity->root_hash_sig;
8ee9615e
LP		 1531
1532 } else if (m->partitions[verity->designator == PARTITION_USR ? PARTITION_USR_VERITY_SIG : PARTITION_ROOT_VERITY_SIG].found) {
1533
1534 /* If we found an embedded signature partition, we are ready, too. */
1535
1536 m->verity_ready = m->verity_sig_ready = true;
1537 m->partitions[verity->designator == PARTITION_USR ? PARTITION_USR : PARTITION_ROOT].rw = false;
aee36b4e 1538 }
4623e8e6
LP		 1539 }
1540
18b5886e
LP		 1541 blkid_free_probe(b);
1542 b = NULL;
1543
8c1be37e 1544 /* Fill in file system types if we don't know them yet. */
569a0e42 1545 for (PartitionDesignator i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) {
18b5886e 1546 DissectedPartition *p = m->partitions + i;
8c1be37e 1547
18b5886e 1548 if (!p->found)
8c1be37e
LP		 1549 continue;
1550
18b5886e
LP		 1551 if (!p->fstype && p->node) {
1552 r = probe_filesystem(p->node, &p->fstype);
7cc84b2c 1553 if (r < 0 && r != -EUCLEAN)
18b5886e 1554 return r;
8c1be37e
LP		 1555 }
1556
18b5886e
LP		 1557 if (streq_ptr(p->fstype, "crypto_LUKS"))
1558 m->encrypted = true;
896f937f
LP		 1559
1560 if (p->fstype && fstype_is_ro(p->fstype))
1561 p->rw = false;
de98f631
LP		 1562
1563 if (!p->rw)
1564 p->growfs = false;
8c1be37e
LP		 1565 }
1566
1cc6c93a 1567 *ret = TAKE_PTR(m);
8c1be37e
LP		 1568 return 0;
1569#else
1570 return -EOPNOTSUPP;
1571#endif
1572}
1573
1574DissectedImage* dissected_image_unref(DissectedImage *m) {
8c1be37e
LP		 1575 if (!m)
1576 return NULL;
1577
08fe0a53
LP		 1578 for (PartitionDesignator i = 0; i < _PARTITION_DESIGNATOR_MAX; i++)
1579 dissected_partition_done(m->partitions + i);
8c1be37e 1580
593fe6c0 1581 free(m->image_name);
3b925504
LP		 1582 free(m->hostname);
1583 strv_free(m->machine_info);
1584 strv_free(m->os_release);
7718ac97 1585 strv_free(m->extension_release);
3b925504 1586
5fecf46d 1587 return mfree(m);
8c1be37e
LP		 1588}
1589
18b5886e 1590static int is_loop_device(const char *path) {
553e15f2 1591 char s[SYS_BLOCK_PATH_MAX("/../loop/")];
18b5886e
LP		 1592 struct stat st;
1593
1594 assert(path);
1595
1596 if (stat(path, &st) < 0)
1597 return -errno;
1598
1599 if (!S_ISBLK(st.st_mode))
1600 return -ENOTBLK;
1601
553e15f2 1602 xsprintf_sys_block_path(s, "/loop/", st.st_dev);
18b5886e
LP		 1603 if (access(s, F_OK) < 0) {
1604 if (errno != ENOENT)
1605 return -errno;
1606
1607 /* The device itself isn't a loop device, but maybe it's a partition and its parent is? */
553e15f2 1608 xsprintf_sys_block_path(s, "/../loop/", st.st_dev);
18b5886e
LP		 1609 if (access(s, F_OK) < 0)
1610 return errno == ENOENT ? false : -errno;
1611 }
1612
1613 return true;
1614}
1615
cf32c486
LP		 1616static int run_fsck(const char *node, const char *fstype) {
1617 int r, exit_status;
1618 pid_t pid;
1619
1620 assert(node);
1621 assert(fstype);
1622
1623 r = fsck_exists(fstype);
1624 if (r < 0) {
1625 log_debug_errno(r, "Couldn't determine whether fsck for %s exists, proceeding anyway.", fstype);
1626 return 0;
1627 }
1628 if (r == 0) {
1629 log_debug("Not checking partition %s, as fsck for %s does not exist.", node, fstype);
1630 return 0;
1631 }
1632
1633 r = safe_fork("(fsck)", FORK_RESET_SIGNALS|FORK_CLOSE_ALL_FDS|FORK_RLIMIT_NOFILE_SAFE|FORK_DEATHSIG|FORK_NULL_STDIO, &pid);
1634 if (r < 0)
1635 return log_debug_errno(r, "Failed to fork off fsck: %m");
1636 if (r == 0) {
1637 /* Child */
1638 execl("/sbin/fsck", "/sbin/fsck", "-aT", node, NULL);
7e0ed2e9 1639 log_open();
cf32c486
LP		 1640 log_debug_errno(errno, "Failed to execl() fsck: %m");
1641 _exit(FSCK_OPERATIONAL_ERROR);
1642 }
1643
1644 exit_status = wait_for_terminate_and_check("fsck", pid, 0);
1645 if (exit_status < 0)
1646 return log_debug_errno(exit_status, "Failed to fork off /sbin/fsck: %m");
1647
1648 if ((exit_status & ~FSCK_ERROR_CORRECTED) != FSCK_SUCCESS) {
1649 log_debug("fsck failed with exit status %i.", exit_status);
1650
1651 if ((exit_status & (FSCK_SYSTEM_SHOULD_REBOOT|FSCK_ERRORS_LEFT_UNCORRECTED)) != 0)
1652 return log_debug_errno(SYNTHETIC_ERRNO(EUCLEAN), "File system is corrupted, refusing.");
1653
1654 log_debug("Ignoring fsck error.");
1655 }
1656
1657 return 0;
1658}
1659
81939d9d
LP		 1660static int fs_grow(const char *node_path, const char *mount_path) {
1661 _cleanup_close_ int mount_fd = -1, node_fd = -1;
81939d9d
LP		 1662 uint64_t size, newsize;
1663 int r;
1664
1665 node_fd = open(node_path, O_RDONLY|O_CLOEXEC|O_NONBLOCK|O_NOCTTY);
1666 if (node_fd < 0)
1667 return log_debug_errno(errno, "Failed to open node device %s: %m", node_path);
1668
1669 if (ioctl(node_fd, BLKGETSIZE64, &size) != 0)
1670 return log_debug_errno(errno, "Failed to get block device size of %s: %m", node_path);
1671
1672 mount_fd = open(mount_path, O_RDONLY|O_DIRECTORY|O_CLOEXEC);
1673 if (mount_fd < 0)
1674 return log_debug_errno(errno, "Failed to open mountd file system %s: %m", mount_path);
1675
1676 log_debug("Resizing \"%s\" to %"PRIu64" bytes...", mount_path, size);
1677 r = resize_fs(mount_fd, size, &newsize);
1678 if (r < 0)
1679 return log_debug_errno(r, "Failed to resize \"%s\" to %"PRIu64" bytes: %m", mount_path, size);
1680
1681 if (newsize == size)
1682 log_debug("Successfully resized \"%s\" to %s bytes.",
2b59bf51 1683 mount_path, FORMAT_BYTES(newsize));
81939d9d
LP		 1684 else {
1685 assert(newsize < size);
1686 log_debug("Successfully resized \"%s\" to %s bytes (%"PRIu64" bytes lost due to blocksize).",
2b59bf51 1687 mount_path, FORMAT_BYTES(newsize), size - newsize);
81939d9d
LP		 1688 }
1689
1690 return 0;
1691}
1692
18b5886e
LP		 1693static int mount_partition(
1694 DissectedPartition *m,
1695 const char *where,
1696 const char *directory,
2d3a5a73 1697 uid_t uid_shift,
21b61b1d 1698 uid_t uid_range,
18b5886e
LP		 1699 DissectImageFlags flags) {
1700
2d3a5a73
LP		 1701 _cleanup_free_ char *chased = NULL, *options = NULL;
1702 const char *p, *node, *fstype;
21b61b1d 1703 bool rw, remap_uid_gid = false;
2eedfd2d 1704 int r;
8c1be37e
LP		 1705
1706 assert(m);
1707 assert(where);
1708
4dc28665 1709 /* Use decrypted node and matching fstype if available, otherwise use the original device */
18b5886e 1710 node = m->decrypted_node ?: m->node;
4dc28665 1711 fstype = m->decrypted_node ? m->decrypted_fstype: m->fstype;
18b5886e 1712
4dc28665 1713 if (!m->found || !node)
8c1be37e 1714 return 0;
4dc28665
LP		 1715 if (!fstype)
1716 return -EAFNOSUPPORT;
8c1be37e 1717
fa45d12c 1718 /* We are looking at an encrypted partition? This either means stacked encryption, or the caller didn't call dissected_image_decrypt() beforehand. Let's return a recognizable error for this case. */
4dc28665 1719 if (streq(fstype, "crypto_LUKS"))
fa45d12c 1720 return -EUNATCH;
18b5886e 1721
ef9c184d 1722 rw = m->rw && !(flags & DISSECT_IMAGE_MOUNT_READ_ONLY);
8c1be37e 1723
cf32c486
LP		 1724 if (FLAGS_SET(flags, DISSECT_IMAGE_FSCK) && rw) {
1725 r = run_fsck(node, fstype);
1726 if (r < 0)
1727 return r;
1728 }
1729
2eedfd2d 1730 if (directory) {
334eb5b0
LP		 1731 /* Automatically create missing mount points inside the image, if necessary. */
1732 r = mkdir_p_root(where, directory, uid_shift, (gid_t) uid_shift, 0755);
1733 if (r < 0 && r != -EROFS)
1734 return r;
1f0f82f1 1735
a5648b80 1736 r = chase_symlinks(directory, where, CHASE_PREFIX_ROOT, &chased, NULL);
2eedfd2d
LP		 1737 if (r < 0)
1738 return r;
1739
1740 p = chased;
9842905e
LP		 1741 } else {
1742 /* Create top-level mount if missing – but only if this is asked for. This won't modify the
1743 * image (as the branch above does) but the host hierarchy, and the created directory might
1744 * survive our mount in the host hierarchy hence. */
1745 if (FLAGS_SET(flags, DISSECT_IMAGE_MKDIR)) {
1746 r = mkdir_p(where, 0755);
1747 if (r < 0)
1748 return r;
1749 }
1750
8c1be37e 1751 p = where;
9842905e 1752 }
8c1be37e 1753
18b5886e 1754 /* If requested, turn on discard support. */
154d2269 1755 if (fstype_can_discard(fstype) &&
18b5886e 1756 ((flags & DISSECT_IMAGE_DISCARD) ||
3afda7c7 1757 ((flags & DISSECT_IMAGE_DISCARD_ON_LOOP) && is_loop_device(m->node) > 0))) {
2d3a5a73
LP		 1758 options = strdup("discard");
1759 if (!options)
1760 return -ENOMEM;
1761 }
1762
21b61b1d 1763 if (uid_is_valid(uid_shift) && uid_shift != 0) {
2d3a5a73 1764
21b61b1d
LP		 1765 if (fstype_can_uid_gid(fstype)) {
1766 _cleanup_free_ char *uid_option = NULL;
2d3a5a73 1767
21b61b1d
LP		 1768 if (asprintf(&uid_option, "uid=" UID_FMT ",gid=" GID_FMT, uid_shift, (gid_t) uid_shift) < 0)
1769 return -ENOMEM;
1770
1771 if (!strextend_with_separator(&options, ",", uid_option))
1772 return -ENOMEM;
1773 } else if (FLAGS_SET(flags, DISSECT_IMAGE_MOUNT_IDMAPPED))
1774 remap_uid_gid = true;
2d3a5a73 1775 }
8c1be37e 1776
18d73705 1777 if (!isempty(m->mount_options))
c2bc710b 1778 if (!strextend_with_separator(&options, ",", m->mount_options))
18d73705
LB		 1779 return -ENOMEM;
1780
b620bf33
LP		 1781 /* So, when you request MS_RDONLY from ext4, then this means nothing. It happily still writes to the
1782 * backing storage. What's worse, the BLKRO[GS]ET flag and (in case of loopback devices)
1783 * LO_FLAGS_READ_ONLY don't mean anything, they affect userspace accesses only, and write accesses
1784 * from the upper file system still get propagated through to the underlying file system,
1785 * unrestricted. To actually get ext4/xfs/btrfs to stop writing to the device we need to specify
1786 * "norecovery" as mount option, in addition to MS_RDONLY. Yes, this sucks, since it means we need to
1787 * carry a per file system table here.
1788 *
1789 * Note that this means that we might not be able to mount corrupted file systems as read-only
1790 * anymore (since in some cases the kernel implementations will refuse mounting when corrupted,
1791 * read-only and "norecovery" is specified). But I think for the case of automatically determined
1792 * mount options for loopback devices this is the right choice, since otherwise using the same
1793 * loopback file twice even in read-only mode, is going to fail badly sooner or later. The usecase of
1794 * making reuse of the immutable images "just work" is more relevant to us than having read-only
1795 * access that actually modifies stuff work on such image files. Or to say this differently: if
1796 * people want their file systems to be fixed up they should just open them in writable mode, where
1797 * all these problems don't exist. */
1798 if (!rw && STRPTR_IN_SET(fstype, "ext3", "ext4", "xfs", "btrfs"))
1799 if (!strextend_with_separator(&options, ",", "norecovery"))
1800 return -ENOMEM;
1801
511a8cfe 1802 r = mount_nofollow_verbose(LOG_DEBUG, node, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), options);
d9223c07
LP		 1803 if (r < 0)
1804 return r;
1805
81939d9d
LP		 1806 if (rw && m->growfs && FLAGS_SET(flags, DISSECT_IMAGE_GROWFS))
1807 (void) fs_grow(node, p);
1808
21b61b1d
LP		 1809 if (remap_uid_gid) {
1810 r = remount_idmap(p, uid_shift, uid_range);
1811 if (r < 0)
1812 return r;
1813 }
1814
d9223c07 1815 return 1;
8c1be37e
LP		 1816}
1817
7cf66030
LP		 1818static int mount_root_tmpfs(const char *where, uid_t uid_shift, DissectImageFlags flags) {
1819 _cleanup_free_ char *options = NULL;
1820 int r;
1821
1822 assert(where);
1823
1824 /* For images that contain /usr/ but no rootfs, let's mount rootfs as tmpfs */
1825
1826 if (FLAGS_SET(flags, DISSECT_IMAGE_MKDIR)) {
1827 r = mkdir_p(where, 0755);
1828 if (r < 0)
1829 return r;
1830 }
1831
1832 if (uid_is_valid(uid_shift)) {
1833 if (asprintf(&options, "uid=" UID_FMT ",gid=" GID_FMT, uid_shift, (gid_t) uid_shift) < 0)
1834 return -ENOMEM;
1835 }
1836
1837 r = mount_nofollow_verbose(LOG_DEBUG, "rootfs", where, "tmpfs", MS_NODEV, options);
1838 if (r < 0)
1839 return r;
1840
1841 return 1;
1842}
1843
21b61b1d
LP		 1844int dissected_image_mount(
1845 DissectedImage *m,
1846 const char *where,
1847 uid_t uid_shift,
1848 uid_t uid_range,
1849 DissectImageFlags flags) {
1850
1f0f82f1 1851 int r, xbootldr_mounted;
8c1be37e
LP		 1852
1853 assert(m);
1854 assert(where);
1855
fa45d12c
LP		 1856 /* Returns:
1857 *
1858 * -ENXIO → No root partition found
7718ac97 1859 * -EMEDIUMTYPE → DISSECT_IMAGE_VALIDATE_OS set but no os-release/extension-release file found
fa45d12c
LP		 1860 * -EUNATCH → Encrypted partition found for which no dm-crypt was set up yet
1861 * -EUCLEAN → fsck for file system failed
1862 * -EBUSY → File system already mounted/used elsewhere (kernel)
4dc28665 1863 * -EAFNOSUPPORT → File system type not supported or not known
fa45d12c
LP		 1864 */
1865
7cf66030
LP		 1866 if (!(m->partitions[PARTITION_ROOT].found ||
1867 (m->partitions[PARTITION_USR].found && FLAGS_SET(flags, DISSECT_IMAGE_USR_NO_ROOT))))
1868 return -ENXIO; /* Require a root fs or at least a /usr/ fs (the latter is subject to a flag of its own) */
8c1be37e 1869
2d3a5a73 1870 if ((flags & DISSECT_IMAGE_MOUNT_NON_ROOT_ONLY) == 0) {
7cf66030
LP		 1871
1872 /* First mount the root fs. If there's none we use a tmpfs. */
1873 if (m->partitions[PARTITION_ROOT].found)
21b61b1d 1874 r = mount_partition(m->partitions + PARTITION_ROOT, where, NULL, uid_shift, uid_range, flags);
7cf66030
LP		 1875 else
1876 r = mount_root_tmpfs(where, uid_shift, flags);
2d3a5a73
LP		 1877 if (r < 0)
1878 return r;
aee36b4e 1879
aee36b4e 1880 /* For us mounting root always means mounting /usr as well */
21b61b1d 1881 r = mount_partition(m->partitions + PARTITION_USR, where, "/usr", uid_shift, uid_range, flags);
aee36b4e
LP		 1882 if (r < 0)
1883 return r;
03bcb6d4 1884
9ccb531a
LB		 1885 if ((flags & (DISSECT_IMAGE_VALIDATE_OS|DISSECT_IMAGE_VALIDATE_OS_EXT)) != 0) {
1886 /* If either one of the validation flags are set, ensure that the image qualifies
1887 * as one or the other (or both). */
1888 bool ok = false;
1889
1890 if (FLAGS_SET(flags, DISSECT_IMAGE_VALIDATE_OS)) {
1891 r = path_is_os_tree(where);
1892 if (r < 0)
1893 return r;
1894 if (r > 0)
1895 ok = true;
1896 }
1897 if (!ok && FLAGS_SET(flags, DISSECT_IMAGE_VALIDATE_OS_EXT)) {
7718ac97
LB		 1898 r = path_is_extension_tree(where, m->image_name);
1899 if (r < 0)
1900 return r;
9ccb531a
LB		 1901 if (r > 0)
1902 ok = true;
7718ac97 1903 }
9ccb531a
LB		 1904
1905 if (!ok)
1906 return -ENOMEDIUM;
03bcb6d4 1907 }
2d3a5a73
LP		 1908 }
1909
705727fd 1910 if (flags & DISSECT_IMAGE_MOUNT_ROOT_ONLY)
2d3a5a73 1911 return 0;
8c1be37e 1912
21b61b1d 1913 r = mount_partition(m->partitions + PARTITION_HOME, where, "/home", uid_shift, uid_range, flags);
8c1be37e
LP		 1914 if (r < 0)
1915 return r;
1916
21b61b1d 1917 r = mount_partition(m->partitions + PARTITION_SRV, where, "/srv", uid_shift, uid_range, flags);
8c1be37e
LP		 1918 if (r < 0)
1919 return r;
1920
21b61b1d 1921 r = mount_partition(m->partitions + PARTITION_VAR, where, "/var", uid_shift, uid_range, flags);
d4dffb85
LP		 1922 if (r < 0)
1923 return r;
1924
21b61b1d 1925 r = mount_partition(m->partitions + PARTITION_TMP, where, "/var/tmp", uid_shift, uid_range, flags);
d4dffb85
LP		 1926 if (r < 0)
1927 return r;
1928
21b61b1d 1929 xbootldr_mounted = mount_partition(m->partitions + PARTITION_XBOOTLDR, where, "/boot", uid_shift, uid_range, flags);
1f0f82f1
LP		 1930 if (xbootldr_mounted < 0)
1931 return xbootldr_mounted;
d9223c07 1932
8c1be37e 1933 if (m->partitions[PARTITION_ESP].found) {
1f0f82f1
LP		 1934 int esp_done = false;
1935
d9223c07
LP		 1936 /* Mount the ESP to /efi if it exists. If it doesn't exist, use /boot instead, but only if it
1937 * exists and is empty, and we didn't already mount the XBOOTLDR partition into it. */
8c1be37e 1938
a5648b80 1939 r = chase_symlinks("/efi", where, CHASE_PREFIX_ROOT, NULL, NULL);
1f0f82f1
LP		 1940 if (r < 0) {
1941 if (r != -ENOENT)
d9223c07 1942 return r;
8c1be37e 1943
1f0f82f1
LP		 1944 /* /efi doesn't exist. Let's see if /boot is suitable then */
1945
1946 if (!xbootldr_mounted) {
1947 _cleanup_free_ char *p = NULL;
2eedfd2d 1948
1f0f82f1
LP		 1949 r = chase_symlinks("/boot", where, CHASE_PREFIX_ROOT, &p, NULL);
1950 if (r < 0) {
1951 if (r != -ENOENT)
1952 return r;
1953 } else if (dir_is_empty(p) > 0) {
1954 /* It exists and is an empty directory. Let's mount the ESP there. */
21b61b1d 1955 r = mount_partition(m->partitions + PARTITION_ESP, where, "/boot", uid_shift, uid_range, flags);
1f0f82f1
LP		 1956 if (r < 0)
1957 return r;
1958
1959 esp_done = true;
1960 }
2eedfd2d 1961 }
8c1be37e 1962 }
1f0f82f1
LP		 1963
1964 if (!esp_done) {
1965 /* OK, let's mount the ESP now to /efi (possibly creating the dir if missing) */
1966
21b61b1d 1967 r = mount_partition(m->partitions + PARTITION_ESP, where, "/efi", uid_shift, uid_range, flags);
1f0f82f1
LP		 1968 if (r < 0)
1969 return r;
1970 }
8c1be37e
LP		 1971 }
1972
1973 return 0;
1974}
1975
21b61b1d
LP		 1976int dissected_image_mount_and_warn(
1977 DissectedImage *m,
1978 const char *where,
1979 uid_t uid_shift,
1980 uid_t uid_range,
1981 DissectImageFlags flags) {
1982
af187ab2
LP		 1983 int r;
1984
1985 assert(m);
1986 assert(where);
1987
21b61b1d 1988 r = dissected_image_mount(m, where, uid_shift, uid_range, flags);
af187ab2
LP		 1989 if (r == -ENXIO)
1990 return log_error_errno(r, "Not root file system found in image.");
1991 if (r == -EMEDIUMTYPE)
7718ac97 1992 return log_error_errno(r, "No suitable os-release/extension-release file in image found.");
af187ab2
LP		 1993 if (r == -EUNATCH)
1994 return log_error_errno(r, "Encrypted file system discovered, but decryption not requested.");
1995 if (r == -EUCLEAN)
1996 return log_error_errno(r, "File system check on image failed.");
1997 if (r == -EBUSY)
1998 return log_error_errno(r, "File system already mounted elsewhere.");
4dc28665
LP		 1999 if (r == -EAFNOSUPPORT)
2000 return log_error_errno(r, "File system type not supported or not known.");
af187ab2
LP		 2001 if (r < 0)
2002 return log_error_errno(r, "Failed to mount image: %m");
2003
2004 return r;
2005}
2006
349cc4a5 2007#if HAVE_LIBCRYPTSETUP
18b5886e
LP		 2008typedef struct DecryptedPartition {
2009 struct crypt_device *device;
2010 char *name;
2011 bool relinquished;
2012} DecryptedPartition;
2013
2014struct DecryptedImage {
2015 DecryptedPartition *decrypted;
2016 size_t n_decrypted;
18b5886e
LP		 2017};
2018#endif
2019
2020DecryptedImage* decrypted_image_unref(DecryptedImage* d) {
349cc4a5 2021#if HAVE_LIBCRYPTSETUP
18b5886e
LP		 2022 int r;
2023
2024 if (!d)
2025 return NULL;
2026
67f63ee5 2027 for (size_t i = 0; i < d->n_decrypted; i++) {
18b5886e
LP		 2028 DecryptedPartition *p = d->decrypted + i;
2029
2030 if (p->device && p->name && !p->relinquished) {
0d12936d 2031 r = sym_crypt_deactivate_by_name(p->device, p->name, 0);
18b5886e
LP		 2032 if (r < 0)
2033 log_debug_errno(r, "Failed to deactivate encrypted partition %s", p->name);
2034 }
2035
2036 if (p->device)
0d12936d 2037 sym_crypt_free(p->device);
18b5886e
LP		 2038 free(p->name);
2039 }
2040
f91861e4 2041 free(d->decrypted);
18b5886e
LP		 2042 free(d);
2043#endif
2044 return NULL;
2045}
2046
349cc4a5 2047#if HAVE_LIBCRYPTSETUP
4623e8e6
LP		 2048
2049static int make_dm_name_and_node(const void *original_node, const char *suffix, char **ret_name, char **ret_node) {
2050 _cleanup_free_ char *name = NULL, *node = NULL;
2051 const char *base;
2052
2053 assert(original_node);
2054 assert(suffix);
2055 assert(ret_name);
2056 assert(ret_node);
2057
2058 base = strrchr(original_node, '/');
2059 if (!base)
ac1f3ad0
LB		 2060 base = original_node;
2061 else
2062 base++;
4623e8e6
LP		 2063 if (isempty(base))
2064 return -EINVAL;
2065
2066 name = strjoin(base, suffix);
2067 if (!name)
2068 return -ENOMEM;
2069 if (!filename_is_valid(name))
2070 return -EINVAL;
2071
0d12936d 2072 node = path_join(sym_crypt_get_dir(), name);
4623e8e6
LP		 2073 if (!node)
2074 return -ENOMEM;
2075
1cc6c93a
YW		 2076 *ret_name = TAKE_PTR(name);
2077 *ret_node = TAKE_PTR(node);
4623e8e6 2078
4623e8e6
LP		 2079 return 0;
2080}
2081
18b5886e
LP		 2082static int decrypt_partition(
2083 DissectedPartition *m,
2084 const char *passphrase,
2085 DissectImageFlags flags,
2086 DecryptedImage *d) {
2087
2088 _cleanup_free_ char *node = NULL, *name = NULL;
0d12936d 2089 _cleanup_(sym_crypt_freep) struct crypt_device *cd = NULL;
18b5886e
LP		 2090 int r;
2091
2092 assert(m);
2093 assert(d);
2094
2095 if (!m->found || !m->node || !m->fstype)
2096 return 0;
2097
2098 if (!streq(m->fstype, "crypto_LUKS"))
2099 return 0;
2100
bdd73ac5
ZJS		 2101 if (!passphrase)
2102 return -ENOKEY;
2103
0d12936d
LP		 2104 r = dlopen_cryptsetup();
2105 if (r < 0)
2106 return r;
2107
4623e8e6
LP		 2108 r = make_dm_name_and_node(m->node, "-decrypted", &name, &node);
2109 if (r < 0)
2110 return r;
18b5886e 2111
319a4f4b 2112 if (!GREEDY_REALLOC0(d->decrypted, d->n_decrypted + 1))
18b5886e
LP		 2113 return -ENOMEM;
2114
0d12936d 2115 r = sym_crypt_init(&cd, m->node);
18b5886e 2116 if (r < 0)
715cbb81 2117 return log_debug_errno(r, "Failed to initialize dm-crypt: %m");
18b5886e 2118
efc3b12f 2119 cryptsetup_enable_logging(cd);
1887032f 2120
0d12936d 2121 r = sym_crypt_load(cd, CRYPT_LUKS, NULL);
294bd454
ZJS		 2122 if (r < 0)
2123 return log_debug_errno(r, "Failed to load LUKS metadata: %m");
18b5886e 2124
0d12936d 2125 r = sym_crypt_activate_by_passphrase(cd, name, CRYPT_ANY_SLOT, passphrase, strlen(passphrase),
ef9c184d 2126 ((flags & DISSECT_IMAGE_DEVICE_READ_ONLY) ? CRYPT_ACTIVATE_READONLY : 0) |
0d12936d 2127 ((flags & DISSECT_IMAGE_DISCARD_ON_CRYPTO) ? CRYPT_ACTIVATE_ALLOW_DISCARDS : 0));
294bd454 2128 if (r < 0) {
715cbb81 2129 log_debug_errno(r, "Failed to activate LUKS device: %m");
294bd454 2130 return r == -EPERM ? -EKEYREJECTED : r;
18b5886e 2131 }
18b5886e 2132
94344385
LP		 2133 d->decrypted[d->n_decrypted++] = (DecryptedPartition) {
2134 .name = TAKE_PTR(name),
2135 .device = TAKE_PTR(cd),
2136 };
18b5886e 2137
1cc6c93a 2138 m->decrypted_node = TAKE_PTR(node);
18b5886e
LP		 2139
2140 return 0;
4623e8e6
LP		 2141}
2142
89e62e0b
LP		 2143static int verity_can_reuse(
2144 const VeritySettings *verity,
2145 const char *name,
2146 struct crypt_device **ret_cd) {
2147
ac1f3ad0
LB		 2148 /* If the same volume was already open, check that the root hashes match, and reuse it if they do */
2149 _cleanup_free_ char *root_hash_existing = NULL;
0d12936d 2150 _cleanup_(sym_crypt_freep) struct crypt_device *cd = NULL;
ac1f3ad0 2151 struct crypt_params_verity crypt_params = {};
89e62e0b 2152 size_t root_hash_existing_size;
ac1f3ad0
LB		 2153 int r;
2154
89e62e0b
LP		 2155 assert(verity);
2156 assert(name);
ac1f3ad0
LB		 2157 assert(ret_cd);
2158
0d12936d 2159 r = sym_crypt_init_by_name(&cd, name);
ac1f3ad0
LB		 2160 if (r < 0)
2161 return log_debug_errno(r, "Error opening verity device, crypt_init_by_name failed: %m");
0d12936d 2162
c719805e
LP		 2163 cryptsetup_enable_logging(cd);
2164
0d12936d 2165 r = sym_crypt_get_verity_info(cd, &crypt_params);
ac1f3ad0
LB		 2166 if (r < 0)
2167 return log_debug_errno(r, "Error opening verity device, crypt_get_verity_info failed: %m");
0d12936d 2168
89e62e0b
LP		 2169 root_hash_existing_size = verity->root_hash_size;
2170 root_hash_existing = malloc0(root_hash_existing_size);
ac1f3ad0
LB		 2171 if (!root_hash_existing)
2172 return -ENOMEM;
0d12936d
LP		 2173
2174 r = sym_crypt_volume_key_get(cd, CRYPT_ANY_SLOT, root_hash_existing, &root_hash_existing_size, NULL, 0);
ac1f3ad0
LB		 2175 if (r < 0)
2176 return log_debug_errno(r, "Error opening verity device, crypt_volume_key_get failed: %m");
89e62e0b
LP		 2177 if (verity->root_hash_size != root_hash_existing_size ||
2178 memcmp(root_hash_existing, verity->root_hash, verity->root_hash_size) != 0)
ac1f3ad0 2179 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Error opening verity device, it already exists but root hashes are different.");
89e62e0b 2180
ac1f3ad0 2181#if HAVE_CRYPT_ACTIVATE_BY_SIGNED_KEY
89e62e0b
LP		 2182 /* Ensure that, if signatures are supported, we only reuse the device if the previous mount used the
2183 * same settings, so that a previous unsigned mount will not be reused if the user asks to use
28423d9a 2184 * signing for the new one, and vice versa. */
89e62e0b 2185 if (!!verity->root_hash_sig != !!(crypt_params.flags & CRYPT_VERITY_ROOT_HASH_SIGNATURE))
ac1f3ad0
LB		 2186 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Error opening verity device, it already exists but signature settings are not the same.");
2187#endif
2188
2189 *ret_cd = TAKE_PTR(cd);
2190 return 0;
2191}
2192
75db809a 2193static inline char* dm_deferred_remove_clean(char *name) {
ac1f3ad0 2194 if (!name)
75db809a 2195 return NULL;
0d12936d
LP		 2196
2197 (void) sym_crypt_deactivate_by_name(NULL, name, CRYPT_DEACTIVATE_DEFERRED);
75db809a 2198 return mfree(name);
ac1f3ad0
LB		 2199}
2200DEFINE_TRIVIAL_CLEANUP_FUNC(char *, dm_deferred_remove_clean);
2201
c2fa92e7
LP		 2202static int validate_signature_userspace(const VeritySettings *verity) {
2203#if HAVE_OPENSSL
2204 _cleanup_(sk_X509_free_allp) STACK_OF(X509) *sk = NULL;
2205 _cleanup_strv_free_ char **certs = NULL;
2206 _cleanup_(PKCS7_freep) PKCS7 *p7 = NULL;
2207 _cleanup_free_ char *s = NULL;
2208 _cleanup_(BIO_freep) BIO *bio = NULL; /* 'bio' must be freed first, 's' second, hence keep this order
2209 * of declaration in place, please */
2210 const unsigned char *d;
2211 char **i;
2212 int r;
2213
2214 assert(verity);
2215 assert(verity->root_hash);
2216 assert(verity->root_hash_sig);
2217
2218 /* Because installing a signature certificate into the kernel chain is so messy, let's optionally do
2219 * userspace validation. */
2220
2221 r = conf_files_list_nulstr(&certs, ".crt", NULL, CONF_FILES_REGULAR|CONF_FILES_FILTER_MASKED, CONF_PATHS_NULSTR("verity.d"));
2222 if (r < 0)
2223 return log_debug_errno(r, "Failed to enumerate certificates: %m");
2224 if (strv_isempty(certs)) {
2225 log_debug("No userspace dm-verity certificates found.");
2226 return 0;
2227 }
2228
2229 d = verity->root_hash_sig;
2230 p7 = d2i_PKCS7(NULL, &d, (long) verity->root_hash_sig_size);
2231 if (!p7)
2232 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Failed to parse PKCS7 DER signature data.");
2233
2234 s = hexmem(verity->root_hash, verity->root_hash_size);
2235 if (!s)
2236 return log_oom_debug();
2237
2238 bio = BIO_new_mem_buf(s, strlen(s));
2239 if (!bio)
2240 return log_oom_debug();
2241
2242 sk = sk_X509_new_null();
2243 if (!sk)
2244 return log_oom_debug();
2245
2246 STRV_FOREACH(i, certs) {
2247 _cleanup_(X509_freep) X509 *c = NULL;
2248 _cleanup_fclose_ FILE *f = NULL;
2249
2250 f = fopen(*i, "re");
2251 if (!f) {
2252 log_debug_errno(errno, "Failed to open '%s', ignoring: %m", *i);
2253 continue;
2254 }
2255
2256 c = PEM_read_X509(f, NULL, NULL, NULL);
2257 if (!c) {
2258 log_debug("Failed to load X509 certificate '%s', ignoring.", *i);
2259 continue;
2260 }
2261
2262 if (sk_X509_push(sk, c) == 0)
2263 return log_oom_debug();
2264
2265 TAKE_PTR(c);
2266 }
2267
2268 r = PKCS7_verify(p7, sk, NULL, bio, NULL, PKCS7_NOINTERN|PKCS7_NOVERIFY);
2269 if (r)
2270 log_debug("Userspace PKCS#7 validation succeeded.");
2271 else
2272 log_debug("Userspace PKCS#7 validation failed: %s", ERR_error_string(ERR_get_error(), NULL));
2273
2274 return r;
2275#else
2276 log_debug("Not doing client-side validation of dm-verity root hash signatures, OpenSSL support disabled.");
2277 return 0;
2278#endif
2279}
2280
2281static int do_crypt_activate_verity(
2282 struct crypt_device *cd,
2283 const char *name,
2284 const VeritySettings *verity) {
2285
2286 bool check_signature;
2287 int r;
2288
2289 assert(cd);
2290 assert(name);
2291 assert(verity);
2292
2293 if (verity->root_hash_sig) {
2294 r = getenv_bool_secure("SYSTEMD_DISSECT_VERITY_SIGNATURE");
2295 if (r < 0 && r != -ENXIO)
2296 log_debug_errno(r, "Failed to parse $SYSTEMD_DISSECT_VERITY_SIGNATURE");
2297
2298 check_signature = r != 0;
2299 } else
2300 check_signature = false;
2301
2302 if (check_signature) {
2303
2304#if HAVE_CRYPT_ACTIVATE_BY_SIGNED_KEY
2305 /* First, if we have support for signed keys in the kernel, then try that first. */
2306 r = sym_crypt_activate_by_signed_key(
2307 cd,
2308 name,
2309 verity->root_hash,
2310 verity->root_hash_size,
2311 verity->root_hash_sig,
2312 verity->root_hash_sig_size,
2313 CRYPT_ACTIVATE_READONLY);
2314 if (r >= 0)
2315 return r;
2316
2317 log_debug("Validation of dm-verity signature failed via the kernel, trying userspace validation instead.");
2318#else
2319 log_debug("Activation of verity device with signature requested, but not supported via the kernel by %s due to missing crypt_activate_by_signed_key(), trying userspace validation instead.",
2320 program_invocation_short_name);
2321#endif
2322
2323 /* So this didn't work via the kernel, then let's try userspace validation instead. If that
2324 * works we'll try to activate without telling the kernel the signature. */
2325
2326 r = validate_signature_userspace(verity);
2327 if (r < 0)
2328 return r;
2329 if (r == 0)
2330 return log_debug_errno(SYNTHETIC_ERRNO(ENOKEY),
2331 "Activation of signed Verity volume worked neither via the kernel nor in userspace, can't activate.");
2332 }
2333
2334 return sym_crypt_activate_by_volume_key(
2335 cd,
2336 name,
2337 verity->root_hash,
2338 verity->root_hash_size,
2339 CRYPT_ACTIVATE_READONLY);
2340}
2341
4623e8e6 2342static int verity_partition(
aee36b4e 2343 PartitionDesignator designator,
4623e8e6
LP		 2344 DissectedPartition *m,
2345 DissectedPartition *v,
89e62e0b 2346 const VeritySettings *verity,
4623e8e6
LP		 2347 DissectImageFlags flags,
2348 DecryptedImage *d) {
2349
0d12936d 2350 _cleanup_(sym_crypt_freep) struct crypt_device *cd = NULL;
ac1f3ad0 2351 _cleanup_(dm_deferred_remove_cleanp) char *restore_deferred_remove = NULL;
89e62e0b 2352 _cleanup_free_ char *node = NULL, *name = NULL;
4623e8e6
LP		 2353 int r;
2354
2355 assert(m);
89e62e0b 2356 assert(v || (verity && verity->data_path));
4623e8e6 2357
89e62e0b 2358 if (!verity || !verity->root_hash)
4623e8e6 2359 return 0;
aee36b4e
LP		 2360 if (!((verity->designator < 0 && designator == PARTITION_ROOT) ||
2361 (verity->designator == designator)))
2362 return 0;
4623e8e6
LP		 2363
2364 if (!m->found || !m->node || !m->fstype)
2365 return 0;
89e62e0b 2366 if (!verity->data_path) {
e7cbe5cb
LB		 2367 if (!v->found || !v->node || !v->fstype)
2368 return 0;
4623e8e6 2369
e7cbe5cb
LB		 2370 if (!streq(v->fstype, "DM_verity_hash"))
2371 return 0;
2372 }
4623e8e6 2373
0d12936d
LP		 2374 r = dlopen_cryptsetup();
2375 if (r < 0)
2376 return r;
2377
ac1f3ad0
LB		 2378 if (FLAGS_SET(flags, DISSECT_IMAGE_VERITY_SHARE)) {
2379 /* Use the roothash, which is unique per volume, as the device node name, so that it can be reused */
2380 _cleanup_free_ char *root_hash_encoded = NULL;
0d12936d 2381
89e62e0b 2382 root_hash_encoded = hexmem(verity->root_hash, verity->root_hash_size);
ac1f3ad0
LB		 2383 if (!root_hash_encoded)
2384 return -ENOMEM;
aee36b4e 2385
ac1f3ad0
LB		 2386 r = make_dm_name_and_node(root_hash_encoded, "-verity", &name, &node);
2387 } else
2388 r = make_dm_name_and_node(m->node, "-verity", &name, &node);
4623e8e6
LP		 2389 if (r < 0)
2390 return r;
2391
89e62e0b 2392 r = sym_crypt_init(&cd, verity->data_path ?: v->node);
4623e8e6
LP		 2393 if (r < 0)
2394 return r;
2395
efc3b12f 2396 cryptsetup_enable_logging(cd);
1887032f 2397
0d12936d 2398 r = sym_crypt_load(cd, CRYPT_VERITY, NULL);
4623e8e6 2399 if (r < 0)
294bd454 2400 return r;
4623e8e6 2401
0d12936d 2402 r = sym_crypt_set_data_device(cd, m->node);
4623e8e6 2403 if (r < 0)
294bd454 2404 return r;
4623e8e6 2405
319a4f4b 2406 if (!GREEDY_REALLOC0(d->decrypted, d->n_decrypted + 1))
ac1f3ad0
LB		 2407 return -ENOMEM;
2408
2409 /* If activating fails because the device already exists, check the metadata and reuse it if it matches.
2410 * In case of ENODEV/ENOENT, which can happen if another process is activating at the exact same time,
2411 * retry a few times before giving up. */
2412 for (unsigned i = 0; i < N_DEVICE_NODE_LIST_ATTEMPTS; i++) {
c2fa92e7
LP		 2413
2414 r = do_crypt_activate_verity(cd, name, verity);
ac1f3ad0
LB		 2415 /* libdevmapper can return EINVAL when the device is already in the activation stage.
2416 * There's no way to distinguish this situation from a genuine error due to invalid
2aed63f4 2417 * parameters, so immediately fall back to activating the device with a unique name.
89e62e0b
LP		 2418 * Improvements in libcrypsetup can ensure this never happens:
2419 * https://gitlab.com/cryptsetup/cryptsetup/-/merge_requests/96 */
ac1f3ad0 2420 if (r == -EINVAL && FLAGS_SET(flags, DISSECT_IMAGE_VERITY_SHARE))
aee36b4e 2421 return verity_partition(designator, m, v, verity, flags & ~DISSECT_IMAGE_VERITY_SHARE, d);
9ecb5c10 2422 if (!IN_SET(r,
22043172 2423 0, /* Success */
9ecb5c10 2424 -EEXIST, /* Volume is already open and ready to be used */
22043172
LP		 2425 -EBUSY, /* Volume is being opened but not ready, crypt_init_by_name can fetch details */
2426 -ENODEV /* Volume is being opened but not ready, crypt_init_by_name would fail, try to open again */))
ac1f3ad0 2427 return r;
9ecb5c10 2428 if (IN_SET(r, -EEXIST, -EBUSY)) {
ac1f3ad0 2429 struct crypt_device *existing_cd = NULL;
c2923fdc 2430
ac1f3ad0
LB		 2431 if (!restore_deferred_remove){
2432 /* To avoid races, disable automatic removal on umount while setting up the new device. Restore it on failure. */
2433 r = dm_deferred_remove_cancel(name);
9ecb5c10
LB		 2434 /* If activation returns EBUSY there might be no deferred removal to cancel, that's fine */
2435 if (r < 0 && r != -ENXIO)
ac1f3ad0 2436 return log_debug_errno(r, "Disabling automated deferred removal for verity device %s failed: %m", node);
9ecb5c10
LB		 2437 if (r == 0) {
2438 restore_deferred_remove = strdup(name);
2439 if (!restore_deferred_remove)
2440 return -ENOMEM;
2441 }
ac1f3ad0 2442 }
c2923fdc 2443
89e62e0b 2444 r = verity_can_reuse(verity, name, &existing_cd);
ac1f3ad0
LB		 2445 /* Same as above, -EINVAL can randomly happen when it actually means -EEXIST */
2446 if (r == -EINVAL && FLAGS_SET(flags, DISSECT_IMAGE_VERITY_SHARE))
aee36b4e 2447 return verity_partition(designator, m, v, verity, flags & ~DISSECT_IMAGE_VERITY_SHARE, d);
9ecb5c10 2448 if (!IN_SET(r, 0, -ENODEV, -ENOENT, -EBUSY))
ac1f3ad0
LB		 2449 return log_debug_errno(r, "Checking whether existing verity device %s can be reused failed: %m", node);
2450 if (r == 0) {
c419b6f0
LB		 2451 /* devmapper might say that the device exists, but the devlink might not yet have been
2452 * created. Check and wait for the udev event in that case. */
9e3d9067 2453 r = device_wait_for_devlink(node, "block", usec_add(now(CLOCK_MONOTONIC), 100 * USEC_PER_MSEC), NULL);
c419b6f0
LB		 2454 /* Fallback to activation with a unique device if it's taking too long */
2455 if (r == -ETIMEDOUT)
2456 break;
2457 if (r < 0)
2458 return r;
2459
ac1f3ad0 2460 if (cd)
0d12936d 2461 sym_crypt_free(cd);
ac1f3ad0
LB		 2462 cd = existing_cd;
2463 }
c2923fdc 2464 }
ac1f3ad0
LB		 2465 if (r == 0)
2466 break;
ecab4c47
LB		 2467
2468 /* Device is being opened by another process, but it has not finished yet, yield for 2ms */
2469 (void) usleep(2 * USEC_PER_MSEC);
ac1f3ad0
LB		 2470 }
2471
ac1f3ad0
LB		 2472 /* An existing verity device was reported by libcryptsetup/libdevmapper, but we can't use it at this time.
2473 * Fall back to activating it with a unique device name. */
2474 if (r != 0 && FLAGS_SET(flags, DISSECT_IMAGE_VERITY_SHARE))
aee36b4e 2475 return verity_partition(designator, m, v, verity, flags & ~DISSECT_IMAGE_VERITY_SHARE, d);
ac1f3ad0
LB		 2476
2477 /* Everything looks good and we'll be able to mount the device, so deferred remove will be re-enabled at that point. */
2478 restore_deferred_remove = mfree(restore_deferred_remove);
4623e8e6 2479
94344385
LP		 2480 d->decrypted[d->n_decrypted++] = (DecryptedPartition) {
2481 .name = TAKE_PTR(name),
2482 .device = TAKE_PTR(cd),
2483 };
4623e8e6 2484
1cc6c93a 2485 m->decrypted_node = TAKE_PTR(node);
4623e8e6
LP		 2486
2487 return 0;
18b5886e
LP		 2488}
2489#endif
2490
2491int dissected_image_decrypt(
2492 DissectedImage *m,
2493 const char *passphrase,
89e62e0b 2494 const VeritySettings *verity,
18b5886e
LP		 2495 DissectImageFlags flags,
2496 DecryptedImage **ret) {
2497
349cc4a5 2498#if HAVE_LIBCRYPTSETUP
49b5b3b4 2499 _cleanup_(decrypted_image_unrefp) DecryptedImage *d = NULL;
18b5886e
LP		 2500 int r;
2501#endif
2502
2503 assert(m);
89e62e0b 2504 assert(!verity || verity->root_hash || verity->root_hash_size == 0);
18b5886e
LP		 2505
2506 /* Returns:
2507 *
2508 * = 0 → There was nothing to decrypt
2509 * > 0 → Decrypted successfully
d1c536f5 2510 * -ENOKEY → There's something to decrypt but no key was supplied
18b5886e
LP		 2511 * -EKEYREJECTED → Passed key was not correct
2512 */
2513
89e62e0b 2514 if (verity && verity->root_hash && verity->root_hash_size < sizeof(sd_id128_t))
4623e8e6
LP		 2515 return -EINVAL;
2516
c3c88d67 2517 if (!m->encrypted && !m->verity_ready) {
18b5886e
LP		 2518 *ret = NULL;
2519 return 0;
2520 }
2521
349cc4a5 2522#if HAVE_LIBCRYPTSETUP
18b5886e
LP		 2523 d = new0(DecryptedImage, 1);
2524 if (!d)
2525 return -ENOMEM;
2526
569a0e42 2527 for (PartitionDesignator i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) {
18b5886e 2528 DissectedPartition *p = m->partitions + i;
22043172 2529 PartitionDesignator k;
18b5886e
LP		 2530
2531 if (!p->found)
2532 continue;
2533
2534 r = decrypt_partition(p, passphrase, flags, d);
2535 if (r < 0)
2536 return r;
2537
4623e8e6
LP		 2538 k = PARTITION_VERITY_OF(i);
2539 if (k >= 0) {
aee36b4e 2540 r = verity_partition(i, p, m->partitions + k, verity, flags | DISSECT_IMAGE_VERITY_SHARE, d);
4623e8e6
LP		 2541 if (r < 0)
2542 return r;
2543 }
2544
18b5886e
LP		 2545 if (!p->decrypted_fstype && p->decrypted_node) {
2546 r = probe_filesystem(p->decrypted_node, &p->decrypted_fstype);
7cc84b2c 2547 if (r < 0 && r != -EUCLEAN)
18b5886e
LP		 2548 return r;
2549 }
2550 }
2551
1cc6c93a 2552 *ret = TAKE_PTR(d);
18b5886e
LP		 2553
2554 return 1;
2555#else
2556 return -EOPNOTSUPP;
2557#endif
2558}
2559
2560int dissected_image_decrypt_interactively(
2561 DissectedImage *m,
2562 const char *passphrase,
89e62e0b 2563 const VeritySettings *verity,
18b5886e
LP		 2564 DissectImageFlags flags,
2565 DecryptedImage **ret) {
2566
2567 _cleanup_strv_free_erase_ char **z = NULL;
2568 int n = 3, r;
2569
2570 if (passphrase)
2571 n--;
2572
2573 for (;;) {
89e62e0b 2574 r = dissected_image_decrypt(m, passphrase, verity, flags, ret);
18b5886e
LP		 2575 if (r >= 0)
2576 return r;
2577 if (r == -EKEYREJECTED)
2578 log_error_errno(r, "Incorrect passphrase, try again!");
fc95c359
YW		 2579 else if (r != -ENOKEY)
2580 return log_error_errno(r, "Failed to decrypt image: %m");
18b5886e 2581
baaa35ad
ZJS		 2582 if (--n < 0)
2583 return log_error_errno(SYNTHETIC_ERRNO(EKEYREJECTED),
2584 "Too many retries.");
18b5886e
LP		 2585
2586 z = strv_free(z);
2587
8806bb4b 2588 r = ask_password_auto("Please enter image passphrase:", NULL, "dissect", "dissect", "dissect.passphrase", USEC_INFINITY, 0, &z);
18b5886e
LP		 2589 if (r < 0)
2590 return log_error_errno(r, "Failed to query for passphrase: %m");
2591
2592 passphrase = z[0];
2593 }
2594}
2595
18b5886e 2596int decrypted_image_relinquish(DecryptedImage *d) {
18b5886e
LP		 2597 assert(d);
2598
67f63ee5
ZJS		 2599 /* Turns on automatic removal after the last use ended for all DM devices of this image, and sets a
2600 * boolean so that we don't clean it up ourselves either anymore */
18b5886e 2601
349cc4a5 2602#if HAVE_LIBCRYPTSETUP
67f63ee5
ZJS		 2603 int r;
2604
2605 for (size_t i = 0; i < d->n_decrypted; i++) {
18b5886e
LP		 2606 DecryptedPartition *p = d->decrypted + i;
2607
2608 if (p->relinquished)
2609 continue;
2610
0d12936d 2611 r = sym_crypt_deactivate_by_name(NULL, p->name, CRYPT_DEACTIVATE_DEFERRED);
18b5886e
LP		 2612 if (r < 0)
2613 return log_debug_errno(r, "Failed to mark %s for auto-removal: %m", p->name);
2614
2615 p->relinquished = true;
2616 }
2617#endif
2618
2619 return 0;
2620}
2621
89e62e0b
LP		 2622static char *build_auxiliary_path(const char *image, const char *suffix) {
2623 const char *e;
2624 char *n;
2625
2626 assert(image);
2627 assert(suffix);
2628
2629 e = endswith(image, ".raw");
2630 if (!e)
2631 return strjoin(e, suffix);
2632
2633 n = new(char, e - image + strlen(suffix) + 1);
2634 if (!n)
2635 return NULL;
2636
2637 strcpy(mempcpy(n, image, e - image), suffix);
2638 return n;
2639}
2640
2641void verity_settings_done(VeritySettings *v) {
2642 assert(v);
2643
2644 v->root_hash = mfree(v->root_hash);
2645 v->root_hash_size = 0;
2646
2647 v->root_hash_sig = mfree(v->root_hash_sig);
2648 v->root_hash_sig_size = 0;
2649
2650 v->data_path = mfree(v->data_path);
2651}
2652
2653int verity_settings_load(
2654 VeritySettings *verity,
f5ea63a5
LP		 2655 const char *image,
2656 const char *root_hash_path,
89e62e0b
LP		 2657 const char *root_hash_sig_path) {
2658
2659 _cleanup_free_ void *root_hash = NULL, *root_hash_sig = NULL;
2660 size_t root_hash_size = 0, root_hash_sig_size = 0;
2661 _cleanup_free_ char *verity_data_path = NULL;
aee36b4e 2662 PartitionDesignator designator;
78ebe980
LP		 2663 int r;
2664
89e62e0b 2665 assert(verity);
78ebe980 2666 assert(image);
aee36b4e 2667 assert(verity->designator < 0 || IN_SET(verity->designator, PARTITION_ROOT, PARTITION_USR));
78ebe980 2668
89e62e0b
LP		 2669 /* If we are asked to load the root hash for a device node, exit early */
2670 if (is_device_path(image))
78ebe980 2671 return 0;
78ebe980 2672
d5fcc5b0
LP		 2673 r = getenv_bool_secure("SYSTEMD_DISSECT_VERITY_SIDECAR");
2674 if (r < 0 && r != -ENXIO)
2675 log_debug_errno(r, "Failed to parse $SYSTEMD_DISSECT_VERITY_SIDECAR, ignoring: %m");
2676 if (r == 0)
2677 return 0;
2678
aee36b4e
LP		 2679 designator = verity->designator;
2680
89e62e0b 2681 /* We only fill in what isn't already filled in */
c2923fdc 2682
89e62e0b 2683 if (!verity->root_hash) {
e7cbe5cb 2684 _cleanup_free_ char *text = NULL;
e7cbe5cb 2685
0389f4fa 2686 if (root_hash_path) {
aee36b4e 2687 /* If explicitly specified it takes precedence */
0389f4fa
LB		 2688 r = read_one_line_file(root_hash_path, &text);
2689 if (r < 0)
e7cbe5cb 2690 return r;
aee36b4e
LP		 2691
2692 if (designator < 0)
2693 designator = PARTITION_ROOT;
0389f4fa 2694 } else {
aee36b4e
LP		 2695 /* Otherwise look for xattr and separate file, and first for the data for root and if
2696 * that doesn't exist for /usr */
0389f4fa 2697
aee36b4e 2698 if (designator < 0 || designator == PARTITION_ROOT) {
c53e07e2 2699 r = getxattr_malloc(image, "user.verity.roothash", &text);
aee36b4e
LP		 2700 if (r < 0) {
2701 _cleanup_free_ char *p = NULL;
78ebe980 2702
aee36b4e
LP		 2703 if (!IN_SET(r, -ENODATA, -ENOENT) && !ERRNO_IS_NOT_SUPPORTED(r))
2704 return r;
e7cbe5cb 2705
aee36b4e
LP		 2706 p = build_auxiliary_path(image, ".roothash");
2707 if (!p)
2708 return -ENOMEM;
2709
2710 r = read_one_line_file(p, &text);
2711 if (r < 0 && r != -ENOENT)
2712 return r;
2713 }
2714
2715 if (text)
2716 designator = PARTITION_ROOT;
2717 }
2718
2719 if (!text && (designator < 0 || designator == PARTITION_USR)) {
2720 /* So in the "roothash" xattr/file name above the "root" of course primarily
2721 * refers to the root of the Verity Merkle tree. But coincidentally it also
2722 * is the hash for the *root* file system, i.e. the "root" neatly refers to
2723 * two distinct concepts called "root". Taking benefit of this happy
2724 * coincidence we call the file with the root hash for the /usr/ file system
2725 * `usrhash`, because `usrroothash` or `rootusrhash` would just be too
2726 * confusing. We thus drop the reference to the root of the Merkle tree, and
2727 * just indicate which file system it's about. */
c53e07e2 2728 r = getxattr_malloc(image, "user.verity.usrhash", &text);
aee36b4e
LP		 2729 if (r < 0) {
2730 _cleanup_free_ char *p = NULL;
2731
2732 if (!IN_SET(r, -ENODATA, -ENOENT) && !ERRNO_IS_NOT_SUPPORTED(r))
2733 return r;
2734
2735 p = build_auxiliary_path(image, ".usrhash");
2736 if (!p)
2737 return -ENOMEM;
2738
2739 r = read_one_line_file(p, &text);
2740 if (r < 0 && r != -ENOENT)
2741 return r;
2742 }
2743
2744 if (text)
2745 designator = PARTITION_USR;
0389f4fa 2746 }
e7cbe5cb
LB		 2747 }
2748
2749 if (text) {
89e62e0b 2750 r = unhexmem(text, strlen(text), &root_hash, &root_hash_size);
e7cbe5cb
LB		 2751 if (r < 0)
2752 return r;
89e62e0b 2753 if (root_hash_size < sizeof(sd_id128_t))
e7cbe5cb
LB		 2754 return -EINVAL;
2755 }
2756 }
2757
90f98986 2758 if ((root_hash || verity->root_hash) && !verity->root_hash_sig) {
aee36b4e 2759 if (root_hash_sig_path) {
ae9cf30b 2760 r = read_full_file(root_hash_sig_path, (char**) &root_hash_sig, &root_hash_sig_size);
aee36b4e
LP		 2761 if (r < 0 && r != -ENOENT)
2762 return r;
2763
2764 if (designator < 0)
2765 designator = PARTITION_ROOT;
2766 } else {
2767 if (designator < 0 || designator == PARTITION_ROOT) {
2768 _cleanup_free_ char *p = NULL;
2769
2770 /* Follow naming convention recommended by the relevant RFC:
2771 * https://tools.ietf.org/html/rfc5751#section-3.2.1 */
2772 p = build_auxiliary_path(image, ".roothash.p7s");
2773 if (!p)
2774 return -ENOMEM;
89e62e0b 2775
ae9cf30b 2776 r = read_full_file(p, (char**) &root_hash_sig, &root_hash_sig_size);
aee36b4e
LP		 2777 if (r < 0 && r != -ENOENT)
2778 return r;
2779 if (r >= 0)
2780 designator = PARTITION_ROOT;
2781 }
2782
2783 if (!root_hash_sig && (designator < 0 || designator == PARTITION_USR)) {
2784 _cleanup_free_ char *p = NULL;
2785
2786 p = build_auxiliary_path(image, ".usrhash.p7s");
2787 if (!p)
2788 return -ENOMEM;
89e62e0b 2789
ae9cf30b 2790 r = read_full_file(p, (char**) &root_hash_sig, &root_hash_sig_size);
aee36b4e
LP		 2791 if (r < 0 && r != -ENOENT)
2792 return r;
2793 if (r >= 0)
2794 designator = PARTITION_USR;
2795 }
89e62e0b
LP		 2796 }
2797
aee36b4e 2798 if (root_hash_sig && root_hash_sig_size == 0) /* refuse empty size signatures */
89e62e0b
LP		 2799 return -EINVAL;
2800 }
2801
2802 if (!verity->data_path) {
2803 _cleanup_free_ char *p = NULL;
2804
2805 p = build_auxiliary_path(image, ".verity");
2806 if (!p)
2807 return -ENOMEM;
2808
2809 if (access(p, F_OK) < 0) {
2810 if (errno != ENOENT)
2811 return -errno;
2812 } else
2813 verity_data_path = TAKE_PTR(p);
2814 }
2815
2816 if (root_hash) {
2817 verity->root_hash = TAKE_PTR(root_hash);
2818 verity->root_hash_size = root_hash_size;
2819 }
2820
2821 if (root_hash_sig) {
2822 verity->root_hash_sig = TAKE_PTR(root_hash_sig);
2823 verity->root_hash_sig_size = root_hash_sig_size;
e7cbe5cb 2824 }
89e62e0b
LP		 2825
2826 if (verity_data_path)
2827 verity->data_path = TAKE_PTR(verity_data_path);
78ebe980 2828
aee36b4e
LP		 2829 if (verity->designator < 0)
2830 verity->designator = designator;
2831
78ebe980
LP		 2832 return 1;
2833}
2834
88b3300f
LP		 2835int dissected_image_load_verity_sig_partition(
2836 DissectedImage *m,
2837 int fd,
2838 VeritySettings *verity) {
2839
2840 _cleanup_free_ void *root_hash = NULL, *root_hash_sig = NULL;
2841 _cleanup_(json_variant_unrefp) JsonVariant *v = NULL;
2842 size_t root_hash_size, root_hash_sig_size;
2843 _cleanup_free_ char *buf = NULL;
2844 PartitionDesignator d;
2845 DissectedPartition *p;
2846 JsonVariant *rh, *sig;
2847 ssize_t n;
2848 char *e;
2849 int r;
2850
2851 assert(m);
2852 assert(fd >= 0);
2853 assert(verity);
2854
2855 if (verity->root_hash && verity->root_hash_sig) /* Already loaded? */
2856 return 0;
2857
2858 r = getenv_bool_secure("SYSTEMD_DISSECT_VERITY_EMBEDDED");
2859 if (r < 0 && r != -ENXIO)
2860 log_debug_errno(r, "Failed to parse $SYSTEMD_DISSECT_VERITY_EMBEDDED, ignoring: %m");
2861 if (r == 0)
2862 return 0;
2863
2864 d = PARTITION_VERITY_SIG_OF(verity->designator < 0 ? PARTITION_ROOT : verity->designator);
2865 assert(d >= 0);
2866
2867 p = m->partitions + d;
2868 if (!p->found)
2869 return 0;
2870 if (p->offset == UINT64_MAX || p->size == UINT64_MAX)
2871 return -EINVAL;
2872
2873 if (p->size > 4*1024*1024) /* Signature data cannot possible be larger than 4M, refuse that */
2874 return -EFBIG;
2875
2876 buf = new(char, p->size+1);
2877 if (!buf)
2878 return -ENOMEM;
2879
2880 n = pread(fd, buf, p->size, p->offset);
2881 if (n < 0)
2882 return -ENOMEM;
2883 if ((uint64_t) n != p->size)
2884 return -EIO;
2885
2886 e = memchr(buf, 0, p->size);
2887 if (e) {
2888 /* If we found a NUL byte then the rest of the data must be NUL too */
2889 if (!memeqzero(e, p->size - (e - buf)))
2890 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Signature data contains embedded NUL byte.");
2891 } else
2892 buf[p->size] = 0;
2893
2894 r = json_parse(buf, 0, &v, NULL, NULL);
2895 if (r < 0)
2896 return log_debug_errno(r, "Failed to parse signature JSON data: %m");
2897
2898 rh = json_variant_by_key(v, "rootHash");
2899 if (!rh)
2900 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Signature JSON object lacks 'rootHash' field.");
2901 if (!json_variant_is_string(rh))
2902 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "'rootHash' field of signature JSON object is not a string.");
2903
2904 r = unhexmem(json_variant_string(rh), SIZE_MAX, &root_hash, &root_hash_size);
2905 if (r < 0)
2906 return log_debug_errno(r, "Failed to parse root hash field: %m");
2907
2908 /* Check if specified root hash matches if it is specified */
2909 if (verity->root_hash &&
2910 memcmp_nn(verity->root_hash, verity->root_hash_size, root_hash, root_hash_size) != 0) {
2911 _cleanup_free_ char *a = NULL, *b = NULL;
2912
2913 a = hexmem(root_hash, root_hash_size);
2914 b = hexmem(verity->root_hash, verity->root_hash_size);
2915
2916 return log_debug_errno(r, "Root hash in signature JSON data (%s) doesn't match configured hash (%s).", strna(a), strna(b));
2917 }
2918
2919 sig = json_variant_by_key(v, "signature");
2920 if (!sig)
2921 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Signature JSON object lacks 'signature' field.");
2922 if (!json_variant_is_string(sig))
2923 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "'signature' field of signature JSON object is not a string.");
2924
2925 r = unbase64mem(json_variant_string(sig), SIZE_MAX, &root_hash_sig, &root_hash_sig_size);
2926 if (r < 0)
2927 return log_debug_errno(r, "Failed to parse signature field: %m");
2928
2929 free_and_replace(verity->root_hash, root_hash);
2930 verity->root_hash_size = root_hash_size;
2931
2932 free_and_replace(verity->root_hash_sig, root_hash_sig);
2933 verity->root_hash_sig_size = root_hash_sig_size;
2934
2935 return 1;
2936}
2937
22847508 2938int dissected_image_acquire_metadata(DissectedImage *m, DissectImageFlags extra_flags) {
3b925504
LP		 2939
2940 enum {
2941 META_HOSTNAME,
2942 META_MACHINE_ID,
2943 META_MACHINE_INFO,
2944 META_OS_RELEASE,
7718ac97 2945 META_EXTENSION_RELEASE,
a4e0d617 2946 META_HAS_INIT_SYSTEM,
3b925504
LP		 2947 _META_MAX,
2948 };
2949
9a4b883b 2950 static const char *const paths[_META_MAX] = {
7718ac97
LB		 2951 [META_HOSTNAME] = "/etc/hostname\0",
2952 [META_MACHINE_ID] = "/etc/machine-id\0",
2953 [META_MACHINE_INFO] = "/etc/machine-info\0",
9a4b883b 2954 [META_OS_RELEASE] = ("/etc/os-release\0"
22847508 2955 "/usr/lib/os-release\0"),
a4e0d617
LP		 2956 [META_EXTENSION_RELEASE] = "extension-release\0", /* Used only for logging. */
2957 [META_HAS_INIT_SYSTEM] = "has-init-system\0", /* ditto */
3b925504
LP		 2958 };
2959
7718ac97 2960 _cleanup_strv_free_ char **machine_info = NULL, **os_release = NULL, **extension_release = NULL;
af8219d5 2961 _cleanup_close_pair_ int error_pipe[2] = { -1, -1 };
3b925504
LP		 2962 _cleanup_(rmdir_and_freep) char *t = NULL;
2963 _cleanup_(sigkill_waitp) pid_t child = 0;
2964 sd_id128_t machine_id = SD_ID128_NULL;
2965 _cleanup_free_ char *hostname = NULL;
67f63ee5 2966 unsigned n_meta_initialized = 0;
af8219d5 2967 int fds[2 * _META_MAX], r, v;
a4e0d617 2968 int has_init_system = -1;
af8219d5 2969 ssize_t n;
3b925504
LP		 2970
2971 BLOCK_SIGNALS(SIGCHLD);
2972
2973 assert(m);
2974
7718ac97 2975 for (; n_meta_initialized < _META_MAX; n_meta_initialized ++) {
d9119c00
LP		 2976 if (!paths[n_meta_initialized]) {
2977 fds[2*n_meta_initialized] = fds[2*n_meta_initialized+1] = -1;
7718ac97 2978 continue;
d9119c00
LP		 2979 }
2980
3b925504
LP		 2981 if (pipe2(fds + 2*n_meta_initialized, O_CLOEXEC) < 0) {
2982 r = -errno;
2983 goto finish;
2984 }
7718ac97 2985 }
3b925504
LP		 2986
2987 r = mkdtemp_malloc("/tmp/dissect-XXXXXX", &t);
2988 if (r < 0)
2989 goto finish;
2990
af8219d5
LP		 2991 if (pipe2(error_pipe, O_CLOEXEC) < 0) {
2992 r = -errno;
2993 goto finish;
2994 }
2995
e2047ba9 2996 r = safe_fork("(sd-dissect)", FORK_RESET_SIGNALS|FORK_DEATHSIG|FORK_NEW_MOUNTNS|FORK_MOUNTNS_SLAVE, &child);
be39f6ee 2997 if (r < 0)
3b925504 2998 goto finish;
be39f6ee 2999 if (r == 0) {
a4e0d617 3000 /* Child in a new mount namespace */
af8219d5
LP		 3001 error_pipe[0] = safe_close(error_pipe[0]);
3002
7cf66030
LP		 3003 r = dissected_image_mount(
3004 m,
3005 t,
3006 UID_INVALID,
21b61b1d 3007 UID_INVALID,
22847508
ZJS		 3008 extra_flags |
3009 DISSECT_IMAGE_READ_ONLY |
3010 DISSECT_IMAGE_MOUNT_ROOT_ONLY |
7cf66030 3011 DISSECT_IMAGE_USR_NO_ROOT);
429d4e41
LP		 3012 if (r < 0) {
3013 log_debug_errno(r, "Failed to mount dissected image: %m");
03ae68f4 3014 goto inner_fail;
429d4e41 3015 }
3b925504 3016
67f63ee5 3017 for (unsigned k = 0; k < _META_MAX; k++) {
37e44c3f 3018 _cleanup_close_ int fd = -ENOENT;
3b925504
LP		 3019 const char *p;
3020
7718ac97
LB		 3021 if (!paths[k])
3022 continue;
3023
3b925504
LP		 3024 fds[2*k] = safe_close(fds[2*k]);
3025
a4e0d617
LP		 3026 switch (k) {
3027
3028 case META_EXTENSION_RELEASE:
9a4b883b
LB		 3029 /* As per the os-release spec, if the image is an extension it will have a file
3030 * named after the image name in extension-release.d/ - we use the image name
3031 * and try to resolve it with the extension-release helpers, as sometimes
3032 * the image names are mangled on deployment and do not match anymore.
3033 * Unlike other paths this is not fixed, and the image name
3034 * can be mangled on deployment, so by calling into the helper
3035 * we allow a fallback that matches on the first extension-release
3036 * file found in the directory, if one named after the image cannot
3037 * be found first. */
3038 r = open_extension_release(t, m->image_name, NULL, &fd);
3039 if (r < 0)
3040 fd = r; /* Propagate the error. */
a4e0d617
LP		 3041 break;
3042
3043 case META_HAS_INIT_SYSTEM: {
3044 bool found = false;
3045 const char *init;
3046
3047 FOREACH_STRING(init,
3048 "/usr/lib/systemd/systemd", /* systemd on /usr merged system */
3049 "/lib/systemd/systemd", /* systemd on /usr non-merged systems */
3050 "/sbin/init") { /* traditional path the Linux kernel invokes */
3051
3052 r = chase_symlinks(init, t, CHASE_PREFIX_ROOT, NULL, NULL);
3053 if (r < 0) {
3054 if (r != -ENOENT)
3055 log_debug_errno(r, "Failed to resolve %s, ignoring: %m", init);
3056 } else {
3057 found = true;
3058 break;
3059 }
3060 }
3061
3062 r = loop_write(fds[2*k+1], &found, sizeof(found), false);
3063 if (r < 0)
3064 goto inner_fail;
3065
3066 continue;
3067 }
3068
3069 default:
9a4b883b
LB		 3070 NULSTR_FOREACH(p, paths[k]) {
3071 fd = chase_symlinks_and_open(p, t, CHASE_PREFIX_ROOT, O_RDONLY|O_CLOEXEC|O_NOCTTY, NULL);
3072 if (fd >= 0)
3073 break;
3074 }
a4e0d617
LP		 3075 }
3076
36952d19
LP		 3077 if (fd < 0) {
3078 log_debug_errno(fd, "Failed to read %s file of image, ignoring: %m", paths[k]);
37e44c3f 3079 fds[2*k+1] = safe_close(fds[2*k+1]);
3b925504 3080 continue;
36952d19 3081 }
3b925504 3082
f5fbe71d 3083 r = copy_bytes(fd, fds[2*k+1], UINT64_MAX, 0);
a4e0d617
LP		 3084 if (r < 0)
3085 goto inner_fail;
3b925504
LP		 3086
3087 fds[2*k+1] = safe_close(fds[2*k+1]);
3088 }
3089
3090 _exit(EXIT_SUCCESS);
a4e0d617
LP		 3091
3092 inner_fail:
03ae68f4 3093 /* Let parent know the error */
a4e0d617
LP		 3094 (void) write(error_pipe[1], &r, sizeof(r));
3095 _exit(EXIT_FAILURE);
3b925504
LP		 3096 }
3097
af8219d5
LP		 3098 error_pipe[1] = safe_close(error_pipe[1]);
3099
67f63ee5 3100 for (unsigned k = 0; k < _META_MAX; k++) {
3b925504
LP		 3101 _cleanup_fclose_ FILE *f = NULL;
3102
7718ac97
LB		 3103 if (!paths[k])
3104 continue;
3105
3b925504
LP		 3106 fds[2*k+1] = safe_close(fds[2*k+1]);
3107
4fa744a3 3108 f = take_fdopen(&fds[2*k], "r");
3b925504
LP		 3109 if (!f) {
3110 r = -errno;
3111 goto finish;
3112 }
3113
3b925504
LP		 3114 switch (k) {
3115
3116 case META_HOSTNAME:
3117 r = read_etc_hostname_stream(f, &hostname);
3118 if (r < 0)
f6048e5e 3119 log_debug_errno(r, "Failed to read /etc/hostname of image: %m");
3b925504
LP		 3120
3121 break;
3122
3123 case META_MACHINE_ID: {
3124 _cleanup_free_ char *line = NULL;
3125
3126 r = read_line(f, LONG_LINE_MAX, &line);
3127 if (r < 0)
f6048e5e 3128 log_debug_errno(r, "Failed to read /etc/machine-id of image: %m");
3b925504
LP		 3129 else if (r == 33) {
3130 r = sd_id128_from_string(line, &machine_id);
3131 if (r < 0)
3132 log_debug_errno(r, "Image contains invalid /etc/machine-id: %s", line);
3133 } else if (r == 0)
f6048e5e 3134 log_debug("/etc/machine-id file of image is empty.");
ab763cb2 3135 else if (streq(line, "uninitialized"))
f6048e5e 3136 log_debug("/etc/machine-id file of image is uninitialized (likely aborted first boot).");
3b925504 3137 else
f6048e5e 3138 log_debug("/etc/machine-id file of image has unexpected length %i.", r);
3b925504
LP		 3139
3140 break;
3141 }
3142
3143 case META_MACHINE_INFO:
aa8fbc74 3144 r = load_env_file_pairs(f, "machine-info", &machine_info);
3b925504 3145 if (r < 0)
f6048e5e 3146 log_debug_errno(r, "Failed to read /etc/machine-info of image: %m");
3b925504
LP		 3147
3148 break;
3149
3150 case META_OS_RELEASE:
aa8fbc74 3151 r = load_env_file_pairs(f, "os-release", &os_release);
3b925504 3152 if (r < 0)
f6048e5e 3153 log_debug_errno(r, "Failed to read OS release file of image: %m");
3b925504
LP		 3154
3155 break;
7718ac97
LB		 3156
3157 case META_EXTENSION_RELEASE:
3158 r = load_env_file_pairs(f, "extension-release", &extension_release);
3159 if (r < 0)
f6048e5e 3160 log_debug_errno(r, "Failed to read extension release file of image: %m");
7718ac97
LB		 3161
3162 break;
a4e0d617
LP		 3163
3164 case META_HAS_INIT_SYSTEM: {
3165 bool b = false;
3166 size_t nr;
3167
3168 errno = 0;
3169 nr = fread(&b, 1, sizeof(b), f);
3170 if (nr != sizeof(b))
3171 log_debug_errno(errno_or_else(EIO), "Failed to read has-init-system boolean: %m");
3172 else
3173 has_init_system = b;
3174
3175 break;
3176 }}
3b925504
LP		 3177 }
3178
2e87a1fd 3179 r = wait_for_terminate_and_check("(sd-dissect)", child, 0);
3b925504 3180 child = 0;
2e87a1fd 3181 if (r < 0)
af8219d5
LP		 3182 return r;
3183
3184 n = read(error_pipe[0], &v, sizeof(v));
3185 if (n < 0)
3186 return -errno;
3187 if (n == sizeof(v))
3188 return v; /* propagate error sent to us from child */
3189 if (n != 0)
3190 return -EIO;
3191
2e87a1fd
LP		 3192 if (r != EXIT_SUCCESS)
3193 return -EPROTO;
3b925504
LP		 3194
3195 free_and_replace(m->hostname, hostname);
3196 m->machine_id = machine_id;
3197 strv_free_and_replace(m->machine_info, machine_info);
3198 strv_free_and_replace(m->os_release, os_release);
7718ac97 3199 strv_free_and_replace(m->extension_release, extension_release);
a4e0d617 3200 m->has_init_system = has_init_system;
3b925504
LP		 3201
3202finish:
67f63ee5 3203 for (unsigned k = 0; k < n_meta_initialized; k++)
3b925504
LP		 3204 safe_close_pair(fds + 2*k);
3205
3206 return r;
3207}
3208
4526113f
LP		 3209int dissect_image_and_warn(
3210 int fd,
3211 const char *name,
89e62e0b 3212 const VeritySettings *verity,
18d73705 3213 const MountOptions *mount_options,
a3642997 3214 uint64_t diskseq,
75dc190d 3215 uint64_t uevent_seqnum_not_before,
4a62257d 3216 usec_t timestamp_not_before,
4526113f
LP		 3217 DissectImageFlags flags,
3218 DissectedImage **ret) {
3219
3220 _cleanup_free_ char *buffer = NULL;
3221 int r;
3222
3223 if (!name) {
3224 r = fd_get_path(fd, &buffer);
3225 if (r < 0)
3226 return r;
3227
3228 name = buffer;
3229 }
3230
a3642997 3231 r = dissect_image(fd, verity, mount_options, diskseq, uevent_seqnum_not_before, timestamp_not_before, flags, ret);
4526113f
LP		 3232 switch (r) {
3233
3234 case -EOPNOTSUPP:
3235 return log_error_errno(r, "Dissecting images is not supported, compiled without blkid support.");
3236
3237 case -ENOPKG:
48084df6
ZJS		 3238 return log_error_errno(r, "%s: Couldn't identify a suitable partition table or file system.", name);
3239
3240 case -ENOMEDIUM:
3241 return log_error_errno(r, "%s: The image does not pass validation.", name);
4526113f
LP		 3242
3243 case -EADDRNOTAVAIL:
48084df6 3244 return log_error_errno(r, "%s: No root partition for specified root hash found.", name);
4526113f
LP		 3245
3246 case -ENOTUNIQ:
48084df6 3247 return log_error_errno(r, "%s: Multiple suitable root partitions found in image.", name);
4526113f
LP		 3248
3249 case -ENXIO:
48084df6 3250 return log_error_errno(r, "%s: No suitable root partition found in image.", name);
4526113f
LP		 3251
3252 case -EPROTONOSUPPORT:
3253 return log_error_errno(r, "Device '%s' is loopback block device with partition scanning turned off, please turn it on.", name);
3254
48084df6
ZJS		 3255 case -ENOTBLK:
3256 return log_error_errno(r, "%s: Image is not a block device.", name);
3257
a94aa2b9
LP		 3258 case -EBADR:
3259 return log_error_errno(r,
3260 "Combining partitioned images (such as '%s') with external Verity data (such as '%s') not supported. "
3261 "(Consider setting $SYSTEMD_DISSECT_VERITY_SIDECAR=0 to disable automatic discovery of external Verity data.)",
3262 name, strna(verity ? verity->data_path : NULL));
3263
4526113f
LP		 3264 default:
3265 if (r < 0)
3266 return log_error_errno(r, "Failed to dissect image '%s': %m", name);
3267
3268 return r;
3269 }
3270}
3271
49536766
LP		 3272bool dissected_image_verity_candidate(const DissectedImage *image, PartitionDesignator partition_designator) {
3273 assert(image);
3274
3275 /* Checks if this partition could theoretically do Verity. For non-partitioned images this only works
3276 * if there's an external verity file supplied, for which we can consult .has_verity. For partitioned
3277 * images we only check the partition type.
3278 *
3279 * This call is used to decide whether to suppress or show a verity column in tabular output of the
3280 * image. */
3281
e7cbe5cb 3282 if (image->single_file_system)
c3c88d67 3283 return partition_designator == PARTITION_ROOT && image->has_verity;
e7cbe5cb
LB		 3284
3285 return PARTITION_VERITY_OF(partition_designator) >= 0;
3286}
3287
49536766
LP		 3288bool dissected_image_verity_ready(const DissectedImage *image, PartitionDesignator partition_designator) {
3289 PartitionDesignator k;
3290
3291 assert(image);
3292
3293 /* Checks if this partition has verity data available that we can activate. For non-partitioned this
3294 * works for the root partition, for others only if the associated verity partition was found. */
3295
3296 if (!image->verity_ready)
3297 return false;
e7cbe5cb
LB		 3298
3299 if (image->single_file_system)
49536766 3300 return partition_designator == PARTITION_ROOT;
e7cbe5cb
LB		 3301
3302 k = PARTITION_VERITY_OF(partition_designator);
3303 return k >= 0 && image->partitions[k].found;
3304}
3305
8ee9615e
LP		 3306bool dissected_image_verity_sig_ready(const DissectedImage *image, PartitionDesignator partition_designator) {
3307 PartitionDesignator k;
3308
3309 assert(image);
3310
3311 /* Checks if this partition has verity signature data available that we can use. */
3312
3313 if (!image->verity_sig_ready)
3314 return false;
3315
3316 if (image->single_file_system)
3317 return partition_designator == PARTITION_ROOT;
3318
3319 k = PARTITION_VERITY_SIG_OF(partition_designator);
3320 return k >= 0 && image->partitions[k].found;
3321}
3322
18d73705
LB		 3323MountOptions* mount_options_free_all(MountOptions *options) {
3324 MountOptions *m;
3325
3326 while ((m = options)) {
3327 LIST_REMOVE(mount_options, options, m);
3328 free(m->options);
3329 free(m);
3330 }
3331
3332 return NULL;
3333}
3334
569a0e42 3335const char* mount_options_from_designator(const MountOptions *options, PartitionDesignator designator) {
f5215bc8 3336 const MountOptions *m;
18d73705 3337
f5215bc8 3338 LIST_FOREACH(mount_options, m, options)
9ece6444 3339 if (designator == m->partition_designator && !isempty(m->options))
18d73705 3340 return m->options;
6aa05ebd 3341
18d73705
LB		 3342 return NULL;
3343}
3344
6aa05ebd
LP		 3345int mount_image_privately_interactively(
3346 const char *image,
3347 DissectImageFlags flags,
3348 char **ret_directory,
3349 LoopDevice **ret_loop_device,
3350 DecryptedImage **ret_decrypted_image) {
3351
27ec815e 3352 _cleanup_(verity_settings_done) VeritySettings verity = VERITY_SETTINGS_DEFAULT;
6aa05ebd
LP		 3353 _cleanup_(loop_device_unrefp) LoopDevice *d = NULL;
3354 _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL;
3355 _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL;
3356 _cleanup_(rmdir_and_freep) char *created_dir = NULL;
3357 _cleanup_free_ char *temp = NULL;
3358 int r;
3359
3360 /* Mounts an OS image at a temporary place, inside a newly created mount namespace of our own. This
3361 * is used by tools such as systemd-tmpfiles or systemd-firstboot to operate on some disk image
3362 * easily. */
3363
3364 assert(image);
3365 assert(ret_directory);
3366 assert(ret_loop_device);
3367 assert(ret_decrypted_image);
3368
27ec815e
LP		 3369 r = verity_settings_load(&verity, image, NULL, NULL);
3370 if (r < 0)
3371 return log_error_errno(r, "Failed to load root hash data: %m");
3372
6aa05ebd
LP		 3373 r = tempfn_random_child(NULL, program_invocation_short_name, &temp);
3374 if (r < 0)
3375 return log_error_errno(r, "Failed to generate temporary mount directory: %m");
3376
3377 r = loop_device_make_by_path(
3378 image,
ef9c184d 3379 FLAGS_SET(flags, DISSECT_IMAGE_DEVICE_READ_ONLY) ? O_RDONLY : O_RDWR,
6aa05ebd
LP		 3380 FLAGS_SET(flags, DISSECT_IMAGE_NO_PARTITION_TABLE) ? 0 : LO_FLAGS_PARTSCAN,
3381 &d);
3382 if (r < 0)
7b87fe4c 3383 return log_error_errno(r, "Failed to set up loopback device for %s: %m", image);
6aa05ebd 3384
a3642997 3385 r = dissect_image_and_warn(d->fd, image, &verity, NULL, d->diskseq, d->uevent_seqnum_not_before, d->timestamp_not_before, flags, &dissected_image);
6aa05ebd
LP		 3386 if (r < 0)
3387 return r;
3388
88b3300f
LP		 3389 r = dissected_image_load_verity_sig_partition(dissected_image, d->fd, &verity);
3390 if (r < 0)
3391 return r;
3392
27ec815e 3393 r = dissected_image_decrypt_interactively(dissected_image, NULL, &verity, flags, &decrypted_image);
6aa05ebd
LP		 3394 if (r < 0)
3395 return r;
3396
3397 r = detach_mount_namespace();
3398 if (r < 0)
3399 return log_error_errno(r, "Failed to detach mount namespace: %m");
3400
3401 r = mkdir_p(temp, 0700);
3402 if (r < 0)
3403 return log_error_errno(r, "Failed to create mount point: %m");
3404
3405 created_dir = TAKE_PTR(temp);
3406
21b61b1d 3407 r = dissected_image_mount_and_warn(dissected_image, created_dir, UID_INVALID, UID_INVALID, flags);
6aa05ebd 3408 if (r < 0)
af187ab2 3409 return r;
6aa05ebd
LP		 3410
3411 if (decrypted_image) {
3412 r = decrypted_image_relinquish(decrypted_image);
3413 if (r < 0)
3414 return log_error_errno(r, "Failed to relinquish DM devices: %m");
3415 }
3416
3417 loop_device_relinquish(d);
3418
3419 *ret_directory = TAKE_PTR(created_dir);
3420 *ret_loop_device = TAKE_PTR(d);
3421 *ret_decrypted_image = TAKE_PTR(decrypted_image);
3422
3423 return 0;
3424}
3425
8c1be37e
LP		 3426static const char *const partition_designator_table[] = {
3427 [PARTITION_ROOT] = "root",
3428 [PARTITION_ROOT_SECONDARY] = "root-secondary",
49ae9d91 3429 [PARTITION_ROOT_OTHER] = "root-other",
aee36b4e
LP		 3430 [PARTITION_USR] = "usr",
3431 [PARTITION_USR_SECONDARY] = "usr-secondary",
49ae9d91 3432 [PARTITION_USR_OTHER] = "usr-other",
8c1be37e
LP		 3433 [PARTITION_HOME] = "home",
3434 [PARTITION_SRV] = "srv",
3435 [PARTITION_ESP] = "esp",
a8c47660 3436 [PARTITION_XBOOTLDR] = "xbootldr",
8c1be37e 3437 [PARTITION_SWAP] = "swap",
4623e8e6
LP		 3438 [PARTITION_ROOT_VERITY] = "root-verity",
3439 [PARTITION_ROOT_SECONDARY_VERITY] = "root-secondary-verity",
49ae9d91 3440 [PARTITION_ROOT_OTHER_VERITY] = "root-other-verity",
aee36b4e
LP		 3441 [PARTITION_USR_VERITY] = "usr-verity",
3442 [PARTITION_USR_SECONDARY_VERITY] = "usr-secondary-verity",
49ae9d91 3443 [PARTITION_USR_OTHER_VERITY] = "usr-other-verity",
8ee9615e
LP		 3444 [PARTITION_ROOT_VERITY_SIG] = "root-verity-sig",
3445 [PARTITION_ROOT_SECONDARY_VERITY_SIG] = "root-secondary-verity-sig",
49ae9d91 3446 [PARTITION_ROOT_OTHER_VERITY_SIG] = "root-other-verity-sig",
8ee9615e
LP		 3447 [PARTITION_USR_VERITY_SIG] = "usr-verity-sig",
3448 [PARTITION_USR_SECONDARY_VERITY_SIG] = "usr-secondary-verity-sig",
49ae9d91 3449 [PARTITION_USR_OTHER_VERITY_SIG] = "usr-other-verity-sig",
d4dffb85
LP		 3450 [PARTITION_TMP] = "tmp",
3451 [PARTITION_VAR] = "var",
8c1be37e
LP		 3452};
3453
93f59701
LB		 3454int verity_dissect_and_mount(
3455 const char *src,
3456 const char *dest,
3457 const MountOptions *options,
3458 const char *required_host_os_release_id,
3459 const char *required_host_os_release_version_id,
60c5f700
LP		 3460 const char *required_host_os_release_sysext_level,
3461 const char *required_sysext_scope) {
93f59701 3462
4beda316
LB		 3463 _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL;
3464 _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL;
3465 _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL;
3466 _cleanup_(verity_settings_done) VeritySettings verity = VERITY_SETTINGS_DEFAULT;
3467 DissectImageFlags dissect_image_flags;
3468 int r;
3469
3470 assert(src);
3471 assert(dest);
3472
3473 r = verity_settings_load(&verity, src, NULL, NULL);
3474 if (r < 0)
3475 return log_debug_errno(r, "Failed to load root hash: %m");
3476
3477 dissect_image_flags = verity.data_path ? DISSECT_IMAGE_NO_PARTITION_TABLE : 0;
3478
3479 r = loop_device_make_by_path(
3480 src,
3481 -1,
3482 verity.data_path ? 0 : LO_FLAGS_PARTSCAN,
3483 &loop_device);
3484 if (r < 0)
3485 return log_debug_errno(r, "Failed to create loop device for image: %m");
3486
3487 r = dissect_image(
3488 loop_device->fd,
3489 &verity,
3490 options,
a3642997 3491 loop_device->diskseq,
75dc190d 3492 loop_device->uevent_seqnum_not_before,
4a62257d 3493 loop_device->timestamp_not_before,
4beda316
LB		 3494 dissect_image_flags,
3495 &dissected_image);
3496 /* No partition table? Might be a single-filesystem image, try again */
3497 if (!verity.data_path && r == -ENOPKG)
3498 r = dissect_image(
3499 loop_device->fd,
3500 &verity,
3501 options,
a3642997 3502 loop_device->diskseq,
75dc190d 3503 loop_device->uevent_seqnum_not_before,
4a62257d 3504 loop_device->timestamp_not_before,
75dc190d 3505 dissect_image_flags | DISSECT_IMAGE_NO_PARTITION_TABLE,
4beda316
LB		 3506 &dissected_image);
3507 if (r < 0)
3508 return log_debug_errno(r, "Failed to dissect image: %m");
3509
88b3300f
LP		 3510 r = dissected_image_load_verity_sig_partition(dissected_image, loop_device->fd, &verity);
3511 if (r < 0)
3512 return r;
3513
4beda316
LB		 3514 r = dissected_image_decrypt(
3515 dissected_image,
3516 NULL,
3517 &verity,
3518 dissect_image_flags,
3519 &decrypted_image);
3520 if (r < 0)
3521 return log_debug_errno(r, "Failed to decrypt dissected image: %m");
3522
3523 r = mkdir_p_label(dest, 0755);
3524 if (r < 0)
3525 return log_debug_errno(r, "Failed to create destination directory %s: %m", dest);
3526 r = umount_recursive(dest, 0);
3527 if (r < 0)
3528 return log_debug_errno(r, "Failed to umount under destination directory %s: %m", dest);
3529
21b61b1d 3530 r = dissected_image_mount(dissected_image, dest, UID_INVALID, UID_INVALID, dissect_image_flags);
4beda316
LB		 3531 if (r < 0)
3532 return log_debug_errno(r, "Failed to mount image: %m");
3533
93f59701
LB		 3534 /* If we got os-release values from the caller, then we need to match them with the image's
3535 * extension-release.d/ content. Return -EINVAL if there's any mismatch.
3536 * First, check the distro ID. If that matches, then check the new SYSEXT_LEVEL value if
37361f46
LB		 3537 * available, or else fallback to VERSION_ID. If neither is present (eg: rolling release),
3538 * then a simple match on the ID will be performed. */
3539 if (required_host_os_release_id) {
93f59701
LB		 3540 _cleanup_strv_free_ char **extension_release = NULL;
3541
3542 r = load_extension_release_pairs(dest, dissected_image->image_name, &extension_release);
3543 if (r < 0)
3544 return log_debug_errno(r, "Failed to parse image %s extension-release metadata: %m", dissected_image->image_name);
3545
3546 r = extension_release_validate(
60c5f700
LP		 3547 dissected_image->image_name,
3548 required_host_os_release_id,
3549 required_host_os_release_version_id,
3550 required_host_os_release_sysext_level,
3551 required_sysext_scope,
3552 extension_release);
93f59701
LB		 3553 if (r == 0)
3554 return log_debug_errno(SYNTHETIC_ERRNO(ESTALE), "Image %s extension-release metadata does not match the root's", dissected_image->image_name);
3555 if (r < 0)
3556 return log_debug_errno(r, "Failed to compare image %s extension-release metadata with the root's os-release: %m", dissected_image->image_name);
3557 }
3558
4beda316
LB		 3559 if (decrypted_image) {
3560 r = decrypted_image_relinquish(decrypted_image);
3561 if (r < 0)
3562 return log_debug_errno(r, "Failed to relinquish decrypted image: %m");
3563 }
3564
3565 loop_device_relinquish(loop_device);
3566
3567 return 0;
3568}
3569
569a0e42 3570DEFINE_STRING_TABLE_LOOKUP(partition_designator, PartitionDesignator);
Unnamed repository; edit this file 'description' to name the repository.