1 /* SPDX-License-Identifier: LGPL-2.1-or-later */
3 #if HAVE_VALGRIND_MEMCHECK_H
4 #include <valgrind/memcheck.h>
7 #include <linux/dm-ioctl.h>
8 #include <linux/loop.h>
10 #include <sys/prctl.h>
14 #include "sd-device.h"
17 #include "architecture.h"
18 #include "ask-password-api.h"
19 #include "blkid-util.h"
20 #include "blockdev-util.h"
22 #include "cryptsetup-util.h"
24 #include "device-nodes.h"
25 #include "device-util.h"
26 #include "discover-image.h"
27 #include "dissect-image.h"
30 #include "extension-release.h"
34 #include "fsck-util.h"
36 #include "hexdecoct.h"
37 #include "hostname-setup.h"
38 #include "id128-util.h"
39 #include "import-util.h"
41 #include "mount-util.h"
42 #include "mountpoint-util.h"
43 #include "namespace-util.h"
44 #include "nulstr-util.h"
46 #include "path-util.h"
47 #include "process-util.h"
48 #include "raw-clone.h"
49 #include "resize-fs.h"
50 #include "signal-util.h"
51 #include "stat-util.h"
52 #include "stdio-util.h"
53 #include "string-table.h"
54 #include "string-util.h"
56 #include "tmpfile-util.h"
57 #include "udev-util.h"
58 #include "user-util.h"
59 #include "xattr-util.h"
61 /* how many times to wait for the device nodes to appear */
62 #define N_DEVICE_NODE_LIST_ATTEMPTS 10
64 int probe_filesystem(const char *node
, char **ret_fstype
) {
65 /* Try to find device content type and return it in *ret_fstype. If nothing is found,
66 * 0/NULL will be returned. -EUCLEAN will be returned for ambiguous results, and an
67 * different error otherwise. */
70 _cleanup_(blkid_free_probep
) blkid_probe b
= NULL
;
75 b
= blkid_new_probe_from_filename(node
);
77 return errno_or_else(ENOMEM
);
79 blkid_probe_enable_superblocks(b
, 1);
80 blkid_probe_set_superblocks_flags(b
, BLKID_SUBLKS_TYPE
);
83 r
= blkid_do_safeprobe(b
);
85 log_debug("No type detected on partition %s", node
);
89 return log_debug_errno(SYNTHETIC_ERRNO(EUCLEAN
),
90 "Results ambiguous for partition %s", node
);
92 return errno_or_else(EIO
);
94 (void) blkid_probe_lookup_value(b
, "TYPE", &fstype
, NULL
);
116 static int enumerator_for_parent(sd_device
*d
, sd_device_enumerator
**ret
) {
117 _cleanup_(sd_device_enumerator_unrefp
) sd_device_enumerator
*e
= NULL
;
123 r
= sd_device_enumerator_new(&e
);
127 r
= sd_device_enumerator_add_match_subsystem(e
, "block", true);
131 r
= sd_device_enumerator_add_match_parent(e
, d
);
135 r
= sd_device_enumerator_add_match_sysattr(e
, "partition", NULL
, true);
143 static int device_is_partition(
145 sd_device
*expected_parent
,
146 blkid_partition pp
) {
148 const char *v
, *parent_syspath
, *expected_parent_syspath
;
149 blkid_loff_t bsize
, bstart
;
150 uint64_t size
, start
;
151 int partno
, bpartno
, r
;
155 assert(expected_parent
);
158 r
= sd_device_get_subsystem(d
, &v
);
161 if (!streq(v
, "block"))
164 if (sd_device_get_devtype(d
, &v
) < 0 || !streq(v
, "partition"))
167 r
= sd_device_get_parent(d
, &parent
);
169 return false; /* Doesn't have a parent? No relevant to us */
171 r
= sd_device_get_syspath(parent
, &parent_syspath
); /* Check parent of device of this action */
175 r
= sd_device_get_syspath(expected_parent
, &expected_parent_syspath
); /* Check parent of device we are looking for */
179 if (!path_equal(parent_syspath
, expected_parent_syspath
))
180 return false; /* Has a different parent than what we need, not interesting to us */
182 /* On kernel uevents we may find the partition number in the PARTN= field. Let's use that preferably,
183 * since it's cheaper and more importantly: the sysfs attribute "partition" appears to become
184 * available late, hence let's use the property instead, which is available at the moment we see the
186 r
= sd_device_get_property_value(d
, "PARTN", &v
);
188 r
= sd_device_get_sysattr_value(d
, "partition", &v
);
192 r
= safe_atoi(v
, &partno
);
197 bpartno
= blkid_partition_get_partno(pp
);
199 return errno_or_else(EIO
);
201 if (partno
!= bpartno
)
204 r
= sd_device_get_sysattr_value(d
, "start", &v
);
207 r
= safe_atou64(v
, &start
);
212 bstart
= blkid_partition_get_start(pp
);
214 return errno_or_else(EIO
);
216 if (start
!= (uint64_t) bstart
)
219 r
= sd_device_get_sysattr_value(d
, "size", &v
);
222 r
= safe_atou64(v
, &size
);
227 bsize
= blkid_partition_get_size(pp
);
229 return errno_or_else(EIO
);
231 if (size
!= (uint64_t) bsize
)
237 static int find_partition(
240 usec_t timestamp_not_before
,
241 DissectImageFlags flags
,
244 _cleanup_(sd_device_enumerator_unrefp
) sd_device_enumerator
*e
= NULL
;
252 r
= enumerator_for_parent(parent
, &e
);
256 FOREACH_DEVICE(e
, q
) {
259 if (!FLAGS_SET(flags
, DISSECT_IMAGE_NO_UDEV
)) {
260 r
= sd_device_get_usec_initialized(q
, &usec
);
261 if (r
== -EBUSY
) /* Not initialized yet */
266 if (timestamp_not_before
!= USEC_INFINITY
&&
267 usec
< timestamp_not_before
) /* udev database entry older than our attachment? Then it's not ours */
271 r
= device_is_partition(q
, parent
, pp
);
275 *ret
= sd_device_ref(q
);
284 sd_device
*parent_device
;
285 blkid_partition blkidp
;
288 uint64_t uevent_seqnum_not_before
;
289 usec_t timestamp_not_before
;
290 DissectImageFlags flags
;
293 static inline void wait_data_done(struct wait_data
*d
) {
294 sd_device_unref(d
->found
);
297 static int device_monitor_handler(sd_device_monitor
*monitor
, sd_device
*device
, void *userdata
) {
298 struct wait_data
*w
= userdata
;
303 if (device_for_action(device
, SD_DEVICE_REMOVE
))
306 if (w
->diskseq
!= 0) {
309 /* If w->diskseq is non-zero, then we must have a disk seqnum */
310 r
= sd_device_get_diskseq(device
, &diskseq
);
312 log_debug_errno(r
, "Dropping event because it has no diskseq, but waiting for %" PRIu64
, w
->diskseq
);
315 if (diskseq
< w
->diskseq
) {
316 log_debug("Dropping event because diskseq too old (%" PRIu64
" < %" PRIu64
")",
317 diskseq
, w
->diskseq
);
320 if (diskseq
> w
->diskseq
) {
322 goto finish
; /* Newer than what we were expecting, so we missed it, stop waiting */
324 } else if (w
->uevent_seqnum_not_before
!= UINT64_MAX
) {
327 r
= sd_device_get_seqnum(device
, &seqnum
);
331 if (seqnum
<= w
->uevent_seqnum_not_before
) { /* From an older use of this loop device */
332 log_debug("Dropping event because seqnum too old (%" PRIu64
" <= %" PRIu64
")",
333 seqnum
, w
->uevent_seqnum_not_before
);
338 r
= device_is_partition(device
, w
->parent_device
, w
->blkidp
);
341 if (r
== 0) /* Not the one we need */
344 /* It's the one we need! Yay! */
346 w
->found
= sd_device_ref(device
);
350 return sd_event_exit(sd_device_monitor_get_event(monitor
), r
);
353 static int timeout_handler(sd_event_source
*s
, uint64_t usec
, void *userdata
) {
354 struct wait_data
*w
= userdata
;
359 /* Why partition not appeared within the timeout? We may lost some uevent, as some properties
360 * were not ready when we received uevent... Not sure, but anyway, let's try to find the
361 * partition again before give up. */
363 r
= find_partition(w
->parent_device
, w
->blkidp
, w
->timestamp_not_before
, w
->flags
, &w
->found
);
365 return log_debug_errno(SYNTHETIC_ERRNO(ETIMEDOUT
),
366 "Partition still not appeared after timeout reached.");
368 return log_debug_errno(r
, "Failed to find partition: %m");
370 log_debug("Partition appeared after timeout reached.");
371 return sd_event_exit(sd_event_source_get_event(s
), 0);
374 static int retry_handler(sd_event_source
*s
, uint64_t usec
, void *userdata
) {
375 struct wait_data
*w
= userdata
;
380 r
= find_partition(w
->parent_device
, w
->blkidp
, w
->timestamp_not_before
, w
->flags
, &w
->found
);
383 return log_debug_errno(r
, "Failed to find partition: %m");
385 log_debug("Partition found by a periodic search.");
386 return sd_event_exit(sd_event_source_get_event(s
), 0);
389 r
= sd_event_source_set_time_relative(s
, 500 * USEC_PER_MSEC
);
393 return sd_event_source_set_enabled(s
, SD_EVENT_ONESHOT
);
396 static int wait_for_partition_device(
401 uint64_t uevent_seqnum_not_before
,
402 usec_t timestamp_not_before
,
403 DissectImageFlags flags
,
406 _cleanup_(sd_event_source_unrefp
) sd_event_source
*timeout_source
= NULL
, *retry_source
= NULL
;
407 _cleanup_(sd_device_monitor_unrefp
) sd_device_monitor
*monitor
= NULL
;
408 _cleanup_(sd_event_unrefp
) sd_event
*event
= NULL
;
415 r
= find_partition(parent
, pp
, timestamp_not_before
, flags
, ret
);
419 r
= sd_event_new(&event
);
423 r
= sd_device_monitor_new(&monitor
);
427 r
= sd_device_monitor_filter_add_match_subsystem_devtype(monitor
, "block", "partition");
431 r
= sd_device_monitor_filter_add_match_parent(monitor
, parent
, true);
435 r
= sd_device_monitor_filter_add_match_sysattr(monitor
, "partition", NULL
, true);
439 r
= sd_device_monitor_attach_event(monitor
, event
);
443 _cleanup_(wait_data_done
) struct wait_data w
= {
444 .parent_device
= parent
,
447 .uevent_seqnum_not_before
= uevent_seqnum_not_before
,
448 .timestamp_not_before
= timestamp_not_before
,
452 r
= sd_device_monitor_start(monitor
, device_monitor_handler
, &w
);
456 /* Check again, the partition might have appeared in the meantime */
457 r
= find_partition(parent
, pp
, timestamp_not_before
, flags
, ret
);
461 if (deadline
!= USEC_INFINITY
) {
462 r
= sd_event_add_time(
463 event
, &timeout_source
,
464 CLOCK_MONOTONIC
, deadline
, 0,
465 timeout_handler
, &w
);
469 r
= sd_event_source_set_exit_on_failure(timeout_source
, true);
474 /* If we don't have a disk sequence number then we cannot do exact matching,
475 * and we cannot know if we missed it or if it has not been sent yet, so set
476 * up additional retries to increase the chances of receiving the event. */
478 r
= sd_event_add_time_relative(
479 event
, &retry_source
,
480 CLOCK_MONOTONIC
, 500 * USEC_PER_MSEC
, 0,
485 r
= sd_event_source_set_exit_on_failure(retry_source
, true);
490 r
= sd_event_loop(event
);
495 *ret
= TAKE_PTR(w
.found
);
499 static void check_partition_flags(
501 unsigned long long pflags
,
502 unsigned long long supported
) {
506 /* Mask away all flags supported by this partition's type and the three flags the UEFI spec defines generically */
507 pflags
&= ~(supported
| GPT_FLAG_REQUIRED_PARTITION
| GPT_FLAG_NO_BLOCK_IO_PROTOCOL
| GPT_FLAG_LEGACY_BIOS_BOOTABLE
);
512 /* If there are other bits set, then log about it, to make things discoverable */
513 for (unsigned i
= 0; i
< sizeof(pflags
) * 8; i
++) {
514 unsigned long long bit
= 1ULL << i
;
515 if (!FLAGS_SET(pflags
, bit
))
518 log_debug("Unexpected partition flag %llu set on %s!", bit
, node
);
522 static int device_wait_for_initialization_harder(
524 const char *subsystem
,
528 usec_t start
, left
, retrigger_timeout
;
531 start
= now(CLOCK_MONOTONIC
);
532 left
= usec_sub_unsigned(deadline
, start
);
535 const char *sn
= NULL
;
537 (void) sd_device_get_sysname(device
, &sn
);
538 log_device_debug(device
,
539 "Waiting for device '%s' to initialize for %s.", strna(sn
), FORMAT_TIMESPAN(left
, 0));
542 if (left
!= USEC_INFINITY
)
543 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 */
545 retrigger_timeout
= 2 * USEC_PER_SEC
;
548 usec_t local_deadline
, n
;
551 n
= now(CLOCK_MONOTONIC
);
554 /* Find next deadline, when we'll retrigger */
555 local_deadline
= start
+
556 DIV_ROUND_UP(n
- start
, retrigger_timeout
) * retrigger_timeout
;
558 if (deadline
!= USEC_INFINITY
&& deadline
<= local_deadline
) {
559 local_deadline
= deadline
;
564 r
= device_wait_for_initialization(device
, subsystem
, local_deadline
, ret
);
565 if (r
>= 0 && DEBUG_LOGGING
) {
566 const char *sn
= NULL
;
568 (void) sd_device_get_sysname(device
, &sn
);
569 log_device_debug(device
,
570 "Successfully waited for device '%s' to initialize for %s.",
572 FORMAT_TIMESPAN(usec_sub_unsigned(now(CLOCK_MONOTONIC
), start
), 0));
575 if (r
!= -ETIMEDOUT
|| last_try
)
579 log_device_debug(device
,
580 "Device didn't initialize within %s, assuming lost event. Retriggering device.",
581 FORMAT_TIMESPAN(usec_sub_unsigned(now(CLOCK_MONOTONIC
), start
), 0));
583 r
= sd_device_trigger(device
, SD_DEVICE_CHANGE
);
590 #define DEVICE_TIMEOUT_USEC (45 * USEC_PER_SEC)
592 static void dissected_partition_done(DissectedPartition
*p
) {
598 free(p
->decrypted_fstype
);
599 free(p
->decrypted_node
);
600 free(p
->mount_options
);
602 *p
= (DissectedPartition
) {
610 const VeritySettings
*verity
,
611 const MountOptions
*mount_options
,
613 uint64_t uevent_seqnum_not_before
,
614 usec_t timestamp_not_before
,
615 DissectImageFlags flags
,
616 DissectedImage
**ret
) {
619 #ifdef GPT_ROOT_NATIVE
620 sd_id128_t root_uuid
= SD_ID128_NULL
, root_verity_uuid
= SD_ID128_NULL
;
622 #ifdef GPT_USR_NATIVE
623 sd_id128_t usr_uuid
= SD_ID128_NULL
, usr_verity_uuid
= SD_ID128_NULL
;
625 bool is_gpt
, is_mbr
, multiple_generic
= false,
626 generic_rw
= false, /* initialize to appease gcc */
627 generic_growfs
= false;
628 _cleanup_(sd_device_unrefp
) sd_device
*d
= NULL
;
629 _cleanup_(dissected_image_unrefp
) DissectedImage
*m
= NULL
;
630 _cleanup_(blkid_free_probep
) blkid_probe b
= NULL
;
631 _cleanup_free_
char *generic_node
= NULL
;
632 sd_id128_t generic_uuid
= SD_ID128_NULL
;
633 const char *pttype
= NULL
, *sysname
= NULL
;
635 int r
, generic_nr
= -1, n_partitions
;
641 assert(!verity
|| verity
->designator
< 0 || IN_SET(verity
->designator
, PARTITION_ROOT
, PARTITION_USR
));
642 assert(!verity
|| verity
->root_hash
|| verity
->root_hash_size
== 0);
643 assert(!verity
|| verity
->root_hash_sig
|| verity
->root_hash_sig_size
== 0);
644 assert(!verity
|| (verity
->root_hash
|| !verity
->root_hash_sig
));
645 assert(!((flags
& DISSECT_IMAGE_GPT_ONLY
) && (flags
& DISSECT_IMAGE_NO_PARTITION_TABLE
)));
647 /* Probes a disk image, and returns information about what it found in *ret.
649 * Returns -ENOPKG if no suitable partition table or file system could be found.
650 * Returns -EADDRNOTAVAIL if a root hash was specified but no matching root/verity partitions found.
651 * Returns -ENXIO if we couldn't find any partition suitable as root or /usr partition
652 * Returns -ENOTUNIQ if we only found multiple generic partitions and thus don't know what to do with that */
654 if (verity
&& verity
->root_hash
) {
655 sd_id128_t fsuuid
, vuuid
;
657 /* If a root hash is supplied, then we use the root partition that has a UUID that match the
658 * first 128bit of the root hash. And we use the verity partition that has a UUID that match
659 * the final 128bit. */
661 if (verity
->root_hash_size
< sizeof(sd_id128_t
))
664 memcpy(&fsuuid
, verity
->root_hash
, sizeof(sd_id128_t
));
665 memcpy(&vuuid
, (const uint8_t*) verity
->root_hash
+ verity
->root_hash_size
- sizeof(sd_id128_t
), sizeof(sd_id128_t
));
667 if (sd_id128_is_null(fsuuid
))
669 if (sd_id128_is_null(vuuid
))
672 /* If the verity data declares it's for the /usr partition, then search for that, in all
673 * other cases assume it's for the root partition. */
674 #ifdef GPT_USR_NATIVE
675 if (verity
->designator
== PARTITION_USR
) {
677 usr_verity_uuid
= vuuid
;
680 #ifdef GPT_ROOT_NATIVE
682 root_verity_uuid
= vuuid
;
684 #ifdef GPT_USR_NATIVE
689 if (fstat(fd
, &st
) < 0)
692 if (!S_ISBLK(st
.st_mode
))
695 r
= sd_device_new_from_stat_rdev(&d
, &st
);
699 if (!FLAGS_SET(flags
, DISSECT_IMAGE_NO_UDEV
)) {
700 _cleanup_(sd_device_unrefp
) sd_device
*initialized
= NULL
;
702 /* If udev support is enabled, then let's wait for the device to be initialized before we doing anything. */
704 r
= device_wait_for_initialization_harder(
707 usec_add(now(CLOCK_MONOTONIC
), DEVICE_TIMEOUT_USEC
),
713 d
= TAKE_PTR(initialized
);
716 b
= blkid_new_probe();
721 r
= blkid_probe_set_device(b
, fd
, 0, 0);
723 return errno_or_else(ENOMEM
);
725 if ((flags
& DISSECT_IMAGE_GPT_ONLY
) == 0) {
726 /* Look for file system superblocks, unless we only shall look for GPT partition tables */
727 blkid_probe_enable_superblocks(b
, 1);
728 blkid_probe_set_superblocks_flags(b
, BLKID_SUBLKS_TYPE
|BLKID_SUBLKS_USAGE
);
731 blkid_probe_enable_partitions(b
, 1);
732 blkid_probe_set_partitions_flags(b
, BLKID_PARTS_ENTRY_DETAILS
);
735 r
= blkid_do_safeprobe(b
);
736 if (IN_SET(r
, -2, 1))
737 return log_debug_errno(SYNTHETIC_ERRNO(ENOPKG
), "Failed to identify any partition table.");
739 return errno_or_else(EIO
);
741 m
= new0(DissectedImage
, 1);
745 r
= sd_device_get_sysname(d
, &sysname
);
747 return log_debug_errno(r
, "Failed to get device sysname: %m");
748 if (startswith(sysname
, "loop")) {
749 _cleanup_free_
char *name_stripped
= NULL
;
750 const char *full_path
;
752 r
= sd_device_get_sysattr_value(d
, "loop/backing_file", &full_path
);
754 log_debug_errno(r
, "Failed to lookup image name via loop device backing file sysattr, ignoring: %m");
756 r
= raw_strip_suffixes(basename(full_path
), &name_stripped
);
761 free_and_replace(m
->image_name
, name_stripped
);
763 r
= free_and_strdup(&m
->image_name
, sysname
);
768 if (!image_name_is_valid(m
->image_name
)) {
769 log_debug("Image name %s is not valid, ignoring", strempty(m
->image_name
));
770 m
->image_name
= mfree(m
->image_name
);
773 if ((!(flags
& DISSECT_IMAGE_GPT_ONLY
) &&
774 (flags
& DISSECT_IMAGE_GENERIC_ROOT
)) ||
775 (flags
& DISSECT_IMAGE_NO_PARTITION_TABLE
)) {
776 const char *usage
= NULL
;
778 /* If flags permit this, also allow using non-partitioned single-filesystem images */
780 (void) blkid_probe_lookup_value(b
, "USAGE", &usage
, NULL
);
781 if (STRPTR_IN_SET(usage
, "filesystem", "crypto")) {
782 const char *fstype
= NULL
, *options
= NULL
, *devname
= NULL
;
783 _cleanup_free_
char *t
= NULL
, *n
= NULL
, *o
= NULL
;
785 /* OK, we have found a file system, that's our root partition then. */
786 (void) blkid_probe_lookup_value(b
, "TYPE", &fstype
, NULL
);
794 r
= sd_device_get_devname(d
, &devname
);
802 m
->single_file_system
= true;
803 m
->encrypted
= streq_ptr(fstype
, "crypto_LUKS");
805 m
->has_verity
= verity
&& verity
->data_path
;
806 m
->verity_ready
= m
->has_verity
&&
808 (verity
->designator
< 0 || verity
->designator
== PARTITION_ROOT
);
810 options
= mount_options_from_designator(mount_options
, PARTITION_ROOT
);
817 m
->partitions
[PARTITION_ROOT
] = (DissectedPartition
) {
819 .rw
= !m
->verity_ready
&& !fstype_is_ro(fstype
),
821 .architecture
= _ARCHITECTURE_INVALID
,
822 .fstype
= TAKE_PTR(t
),
824 .mount_options
= TAKE_PTR(o
),
832 (void) blkid_probe_lookup_value(b
, "PTTYPE", &pttype
, NULL
);
836 is_gpt
= streq_ptr(pttype
, "gpt");
837 is_mbr
= streq_ptr(pttype
, "dos");
839 if (!is_gpt
&& ((flags
& DISSECT_IMAGE_GPT_ONLY
) || !is_mbr
))
842 /* We support external verity data partitions only if the image has no partition table */
843 if (verity
&& verity
->data_path
)
846 /* Safety check: refuse block devices that carry a partition table but for which the kernel doesn't
847 * do partition scanning. */
848 r
= blockdev_partscan_enabled(fd
);
852 return -EPROTONOSUPPORT
;
855 pl
= blkid_probe_get_partitions(b
);
857 return errno_or_else(ENOMEM
);
860 n_partitions
= blkid_partlist_numof_partitions(pl
);
861 if (n_partitions
< 0)
862 return errno_or_else(EIO
);
864 deadline
= usec_add(now(CLOCK_MONOTONIC
), DEVICE_TIMEOUT_USEC
);
865 for (int i
= 0; i
< n_partitions
; i
++) {
866 _cleanup_(sd_device_unrefp
) sd_device
*q
= NULL
;
867 unsigned long long pflags
;
873 pp
= blkid_partlist_get_partition(pl
, i
);
875 return errno_or_else(EIO
);
877 r
= wait_for_partition_device(d
, pp
, deadline
, diskseq
, uevent_seqnum_not_before
, timestamp_not_before
, flags
, &q
);
881 r
= sd_device_get_devname(q
, &node
);
885 pflags
= blkid_partition_get_flags(pp
);
888 nr
= blkid_partition_get_partno(pp
);
890 return errno_or_else(EIO
);
893 PartitionDesignator designator
= _PARTITION_DESIGNATOR_INVALID
;
894 int architecture
= _ARCHITECTURE_INVALID
;
895 const char *stype
, *sid
, *fstype
= NULL
, *label
;
896 sd_id128_t type_id
, id
;
897 bool rw
= true, growfs
= false;
899 sid
= blkid_partition_get_uuid(pp
);
902 if (sd_id128_from_string(sid
, &id
) < 0)
905 stype
= blkid_partition_get_type_string(pp
);
908 if (sd_id128_from_string(stype
, &type_id
) < 0)
911 label
= blkid_partition_get_name(pp
); /* libblkid returns NULL here if empty */
913 if (sd_id128_equal(type_id
, GPT_HOME
)) {
915 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
|GPT_FLAG_GROWFS
);
917 if (pflags
& GPT_FLAG_NO_AUTO
)
920 designator
= PARTITION_HOME
;
921 rw
= !(pflags
& GPT_FLAG_READ_ONLY
);
922 growfs
= FLAGS_SET(pflags
, GPT_FLAG_GROWFS
);
924 } else if (sd_id128_equal(type_id
, GPT_SRV
)) {
926 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
|GPT_FLAG_GROWFS
);
928 if (pflags
& GPT_FLAG_NO_AUTO
)
931 designator
= PARTITION_SRV
;
932 rw
= !(pflags
& GPT_FLAG_READ_ONLY
);
933 growfs
= FLAGS_SET(pflags
, GPT_FLAG_GROWFS
);
935 } else if (sd_id128_equal(type_id
, GPT_ESP
)) {
937 /* Note that we don't check the GPT_FLAG_NO_AUTO flag for the ESP, as it is
938 * not defined there. We instead check the GPT_FLAG_NO_BLOCK_IO_PROTOCOL, as
939 * recommended by the UEFI spec (See "12.3.3 Number and Location of System
942 if (pflags
& GPT_FLAG_NO_BLOCK_IO_PROTOCOL
)
945 designator
= PARTITION_ESP
;
948 } else if (sd_id128_equal(type_id
, GPT_XBOOTLDR
)) {
950 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
|GPT_FLAG_GROWFS
);
952 if (pflags
& GPT_FLAG_NO_AUTO
)
955 designator
= PARTITION_XBOOTLDR
;
956 rw
= !(pflags
& GPT_FLAG_READ_ONLY
);
957 growfs
= FLAGS_SET(pflags
, GPT_FLAG_GROWFS
);
959 #ifdef GPT_ROOT_NATIVE
960 else if (sd_id128_equal(type_id
, GPT_ROOT_NATIVE
)) {
962 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
|GPT_FLAG_GROWFS
);
964 if (pflags
& GPT_FLAG_NO_AUTO
)
967 /* If a root ID is specified, ignore everything but the root id */
968 if (!sd_id128_is_null(root_uuid
) && !sd_id128_equal(root_uuid
, id
))
971 designator
= PARTITION_ROOT
;
972 architecture
= native_architecture();
973 rw
= !(pflags
& GPT_FLAG_READ_ONLY
);
974 growfs
= FLAGS_SET(pflags
, GPT_FLAG_GROWFS
);
976 } else if (sd_id128_equal(type_id
, GPT_ROOT_NATIVE_VERITY
)) {
978 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
);
980 if (pflags
& GPT_FLAG_NO_AUTO
)
983 m
->has_verity
= true;
985 /* Ignore verity unless a root hash is specified */
986 if (sd_id128_is_null(root_verity_uuid
) || !sd_id128_equal(root_verity_uuid
, id
))
989 designator
= PARTITION_ROOT_VERITY
;
990 fstype
= "DM_verity_hash";
991 architecture
= native_architecture();
995 #ifdef GPT_ROOT_SECONDARY
996 else if (sd_id128_equal(type_id
, GPT_ROOT_SECONDARY
)) {
998 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
|GPT_FLAG_GROWFS
);
1000 if (pflags
& GPT_FLAG_NO_AUTO
)
1003 /* If a root ID is specified, ignore everything but the root id */
1004 if (!sd_id128_is_null(root_uuid
) && !sd_id128_equal(root_uuid
, id
))
1007 designator
= PARTITION_ROOT_SECONDARY
;
1008 architecture
= SECONDARY_ARCHITECTURE
;
1009 rw
= !(pflags
& GPT_FLAG_READ_ONLY
);
1010 growfs
= FLAGS_SET(pflags
, GPT_FLAG_GROWFS
);
1012 } else if (sd_id128_equal(type_id
, GPT_ROOT_SECONDARY_VERITY
)) {
1014 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
);
1016 if (pflags
& GPT_FLAG_NO_AUTO
)
1019 m
->has_verity
= true;
1021 /* Ignore verity unless root has is specified */
1022 if (sd_id128_is_null(root_verity_uuid
) || !sd_id128_equal(root_verity_uuid
, id
))
1025 designator
= PARTITION_ROOT_SECONDARY_VERITY
;
1026 fstype
= "DM_verity_hash";
1027 architecture
= SECONDARY_ARCHITECTURE
;
1031 #ifdef GPT_USR_NATIVE
1032 else if (sd_id128_equal(type_id
, GPT_USR_NATIVE
)) {
1034 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
|GPT_FLAG_GROWFS
);
1036 if (pflags
& GPT_FLAG_NO_AUTO
)
1039 /* If a usr ID is specified, ignore everything but the usr id */
1040 if (!sd_id128_is_null(usr_uuid
) && !sd_id128_equal(usr_uuid
, id
))
1043 designator
= PARTITION_USR
;
1044 architecture
= native_architecture();
1045 rw
= !(pflags
& GPT_FLAG_READ_ONLY
);
1046 growfs
= FLAGS_SET(pflags
, GPT_FLAG_GROWFS
);
1048 } else if (sd_id128_equal(type_id
, GPT_USR_NATIVE_VERITY
)) {
1050 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
);
1052 if (pflags
& GPT_FLAG_NO_AUTO
)
1055 m
->has_verity
= true;
1057 /* Ignore verity unless a usr hash is specified */
1058 if (sd_id128_is_null(usr_verity_uuid
) || !sd_id128_equal(usr_verity_uuid
, id
))
1061 designator
= PARTITION_USR_VERITY
;
1062 fstype
= "DM_verity_hash";
1063 architecture
= native_architecture();
1067 #ifdef GPT_USR_SECONDARY
1068 else if (sd_id128_equal(type_id
, GPT_USR_SECONDARY
)) {
1070 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
|GPT_FLAG_GROWFS
);
1072 if (pflags
& GPT_FLAG_NO_AUTO
)
1075 /* If a usr ID is specified, ignore everything but the usr id */
1076 if (!sd_id128_is_null(usr_uuid
) && !sd_id128_equal(usr_uuid
, id
))
1079 designator
= PARTITION_USR_SECONDARY
;
1080 architecture
= SECONDARY_ARCHITECTURE
;
1081 rw
= !(pflags
& GPT_FLAG_READ_ONLY
);
1082 growfs
= FLAGS_SET(pflags
, GPT_FLAG_GROWFS
);
1084 } else if (sd_id128_equal(type_id
, GPT_USR_SECONDARY_VERITY
)) {
1086 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
);
1088 if (pflags
& GPT_FLAG_NO_AUTO
)
1091 m
->has_verity
= true;
1093 /* Ignore verity unless usr has is specified */
1094 if (sd_id128_is_null(usr_verity_uuid
) || !sd_id128_equal(usr_verity_uuid
, id
))
1097 designator
= PARTITION_USR_SECONDARY_VERITY
;
1098 fstype
= "DM_verity_hash";
1099 architecture
= SECONDARY_ARCHITECTURE
;
1103 else if (sd_id128_equal(type_id
, GPT_SWAP
)) {
1105 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
);
1107 if (pflags
& GPT_FLAG_NO_AUTO
)
1110 designator
= PARTITION_SWAP
;
1112 } else if (sd_id128_equal(type_id
, GPT_LINUX_GENERIC
)) {
1114 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
|GPT_FLAG_GROWFS
);
1116 if (pflags
& GPT_FLAG_NO_AUTO
)
1120 multiple_generic
= true;
1123 generic_rw
= !(pflags
& GPT_FLAG_READ_ONLY
);
1124 generic_growfs
= FLAGS_SET(pflags
, GPT_FLAG_GROWFS
);
1126 generic_node
= strdup(node
);
1131 } else if (sd_id128_equal(type_id
, GPT_TMP
)) {
1133 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
|GPT_FLAG_GROWFS
);
1135 if (pflags
& GPT_FLAG_NO_AUTO
)
1138 designator
= PARTITION_TMP
;
1139 rw
= !(pflags
& GPT_FLAG_READ_ONLY
);
1140 growfs
= FLAGS_SET(pflags
, GPT_FLAG_GROWFS
);
1142 } else if (sd_id128_equal(type_id
, GPT_VAR
)) {
1144 check_partition_flags(node
, pflags
, GPT_FLAG_NO_AUTO
|GPT_FLAG_READ_ONLY
|GPT_FLAG_GROWFS
);
1146 if (pflags
& GPT_FLAG_NO_AUTO
)
1149 if (!FLAGS_SET(flags
, DISSECT_IMAGE_RELAX_VAR_CHECK
)) {
1150 sd_id128_t var_uuid
;
1152 /* For /var we insist that the uuid of the partition matches the
1153 * HMAC-SHA256 of the /var GPT partition type uuid, keyed by machine
1154 * ID. Why? Unlike the other partitions /var is inherently
1155 * installation specific, hence we need to be careful not to mount it
1156 * in the wrong installation. By hashing the partition UUID from
1157 * /etc/machine-id we can securely bind the partition to the
1160 r
= sd_id128_get_machine_app_specific(GPT_VAR
, &var_uuid
);
1164 if (!sd_id128_equal(var_uuid
, id
)) {
1165 log_debug("Found a /var/ partition, but its UUID didn't match our expectations, ignoring.");
1170 designator
= PARTITION_VAR
;
1171 rw
= !(pflags
& GPT_FLAG_READ_ONLY
);
1172 growfs
= FLAGS_SET(pflags
, GPT_FLAG_GROWFS
);
1175 if (designator
!= _PARTITION_DESIGNATOR_INVALID
) {
1176 _cleanup_free_
char *t
= NULL
, *n
= NULL
, *o
= NULL
, *l
= NULL
;
1177 const char *options
= NULL
;
1179 if (m
->partitions
[designator
].found
) {
1180 /* For most partition types the first one we see wins. Except for the
1181 * rootfs and /usr, where we do a version compare of the label, and
1182 * let the newest version win. This permits a simple A/B versioning
1183 * scheme in OS images. */
1185 if (!PARTITION_DESIGNATOR_VERSIONED(designator
) ||
1186 strverscmp_improved(m
->partitions
[designator
].label
, label
) >= 0)
1189 dissected_partition_done(m
->partitions
+ designator
);
1208 options
= mount_options_from_designator(mount_options
, designator
);
1210 o
= strdup(options
);
1215 m
->partitions
[designator
] = (DissectedPartition
) {
1220 .architecture
= architecture
,
1221 .node
= TAKE_PTR(n
),
1222 .fstype
= TAKE_PTR(t
),
1223 .label
= TAKE_PTR(l
),
1225 .mount_options
= TAKE_PTR(o
),
1229 } else if (is_mbr
) {
1231 switch (blkid_partition_get_type(pp
)) {
1233 case 0x83: /* Linux partition */
1235 if (pflags
!= 0x80) /* Bootable flag */
1239 multiple_generic
= true;
1243 generic_growfs
= false;
1244 generic_node
= strdup(node
);
1251 case 0xEA: { /* Boot Loader Spec extended $BOOT partition */
1252 _cleanup_free_
char *n
= NULL
, *o
= NULL
;
1253 sd_id128_t id
= SD_ID128_NULL
;
1254 const char *sid
, *options
= NULL
;
1256 /* First one wins */
1257 if (m
->partitions
[PARTITION_XBOOTLDR
].found
)
1260 sid
= blkid_partition_get_uuid(pp
);
1262 (void) sd_id128_from_string(sid
, &id
);
1268 options
= mount_options_from_designator(mount_options
, PARTITION_XBOOTLDR
);
1270 o
= strdup(options
);
1275 m
->partitions
[PARTITION_XBOOTLDR
] = (DissectedPartition
) {
1280 .architecture
= _ARCHITECTURE_INVALID
,
1281 .node
= TAKE_PTR(n
),
1283 .mount_options
= TAKE_PTR(o
),
1291 if (m
->partitions
[PARTITION_ROOT
].found
) {
1292 /* If we found the primary arch, then invalidate the secondary arch to avoid any ambiguities,
1293 * since we never want to mount the secondary arch in this case. */
1294 m
->partitions
[PARTITION_ROOT_SECONDARY
].found
= false;
1295 m
->partitions
[PARTITION_ROOT_SECONDARY_VERITY
].found
= false;
1296 m
->partitions
[PARTITION_USR_SECONDARY
].found
= false;
1297 m
->partitions
[PARTITION_USR_SECONDARY_VERITY
].found
= false;
1299 } else if (m
->partitions
[PARTITION_ROOT_VERITY
].found
)
1300 return -EADDRNOTAVAIL
; /* Verity found but no matching rootfs? Something is off, refuse. */
1302 else if (m
->partitions
[PARTITION_ROOT_SECONDARY
].found
) {
1304 /* No root partition found but there's one for the secondary architecture? Then upgrade
1305 * secondary arch to first */
1307 m
->partitions
[PARTITION_ROOT
] = m
->partitions
[PARTITION_ROOT_SECONDARY
];
1308 zero(m
->partitions
[PARTITION_ROOT_SECONDARY
]);
1309 m
->partitions
[PARTITION_ROOT_VERITY
] = m
->partitions
[PARTITION_ROOT_SECONDARY_VERITY
];
1310 zero(m
->partitions
[PARTITION_ROOT_SECONDARY_VERITY
]);
1312 m
->partitions
[PARTITION_USR
] = m
->partitions
[PARTITION_USR_SECONDARY
];
1313 zero(m
->partitions
[PARTITION_USR_SECONDARY
]);
1314 m
->partitions
[PARTITION_USR_VERITY
] = m
->partitions
[PARTITION_USR_SECONDARY_VERITY
];
1315 zero(m
->partitions
[PARTITION_USR_SECONDARY_VERITY
]);
1317 } else if (m
->partitions
[PARTITION_ROOT_SECONDARY_VERITY
].found
)
1318 return -EADDRNOTAVAIL
; /* as above */
1320 if (m
->partitions
[PARTITION_USR
].found
) {
1322 /* Invalidate secondary arch /usr/ if we found the primary arch */
1323 m
->partitions
[PARTITION_USR_SECONDARY
].found
= false;
1324 m
->partitions
[PARTITION_USR_SECONDARY_VERITY
].found
= false;
1326 } else if (m
->partitions
[PARTITION_USR_VERITY
].found
)
1327 return -EADDRNOTAVAIL
; /* as above */
1329 else if (m
->partitions
[PARTITION_USR_SECONDARY
].found
) {
1331 /* Upgrade secondary arch to primary */
1332 m
->partitions
[PARTITION_USR
] = m
->partitions
[PARTITION_USR_SECONDARY
];
1333 zero(m
->partitions
[PARTITION_USR_SECONDARY
]);
1334 m
->partitions
[PARTITION_USR_VERITY
] = m
->partitions
[PARTITION_USR_SECONDARY_VERITY
];
1335 zero(m
->partitions
[PARTITION_USR_SECONDARY_VERITY
]);
1337 } else if (m
->partitions
[PARTITION_USR_SECONDARY_VERITY
].found
)
1338 return -EADDRNOTAVAIL
; /* as above */
1340 /* If root and /usr are combined then insist that the architecture matches */
1341 if (m
->partitions
[PARTITION_ROOT
].found
&&
1342 m
->partitions
[PARTITION_USR
].found
&&
1343 (m
->partitions
[PARTITION_ROOT
].architecture
>= 0 &&
1344 m
->partitions
[PARTITION_USR
].architecture
>= 0 &&
1345 m
->partitions
[PARTITION_ROOT
].architecture
!= m
->partitions
[PARTITION_USR
].architecture
))
1346 return -EADDRNOTAVAIL
;
1348 if (!m
->partitions
[PARTITION_ROOT
].found
&&
1349 !m
->partitions
[PARTITION_USR
].found
&&
1350 (flags
& DISSECT_IMAGE_GENERIC_ROOT
) &&
1351 (!verity
|| !verity
->root_hash
|| verity
->designator
!= PARTITION_USR
)) {
1353 /* OK, we found nothing usable, then check if there's a single generic one distro, and use
1354 * that. If the root hash was set however, then we won't fall back to a generic node, because
1355 * the root hash decides. */
1357 /* If we didn't find a properly marked root partition, but we did find a single suitable
1358 * generic Linux partition, then use this as root partition, if the caller asked for it. */
1359 if (multiple_generic
)
1362 /* If we didn't find a generic node, then we can't fix this up either */
1364 _cleanup_free_
char *o
= NULL
;
1365 const char *options
;
1367 options
= mount_options_from_designator(mount_options
, PARTITION_ROOT
);
1369 o
= strdup(options
);
1374 assert(generic_nr
>= 0);
1375 m
->partitions
[PARTITION_ROOT
] = (DissectedPartition
) {
1378 .growfs
= generic_growfs
,
1379 .partno
= generic_nr
,
1380 .architecture
= _ARCHITECTURE_INVALID
,
1381 .node
= TAKE_PTR(generic_node
),
1382 .uuid
= generic_uuid
,
1383 .mount_options
= TAKE_PTR(o
),
1388 /* 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 */
1389 if (FLAGS_SET(flags
, DISSECT_IMAGE_REQUIRE_ROOT
) &&
1390 !(m
->partitions
[PARTITION_ROOT
].found
|| (m
->partitions
[PARTITION_USR
].found
&& FLAGS_SET(flags
, DISSECT_IMAGE_USR_NO_ROOT
))))
1393 if (m
->partitions
[PARTITION_ROOT_VERITY
].found
) {
1394 /* We only support one verity partition per image, i.e. can't do for both /usr and root fs */
1395 if (m
->partitions
[PARTITION_USR_VERITY
].found
)
1398 /* We don't support verity enabled root with a split out /usr. Neither with nor without
1399 * verity there. (Note that we do support verity-less root with verity-full /usr, though.) */
1400 if (m
->partitions
[PARTITION_USR
].found
)
1401 return -EADDRNOTAVAIL
;
1405 /* If a verity designator is specified, then insist that the matching partition exists */
1406 if (verity
->designator
>= 0 && !m
->partitions
[verity
->designator
].found
)
1407 return -EADDRNOTAVAIL
;
1409 if (verity
->root_hash
) {
1410 if (verity
->designator
< 0 || verity
->designator
== PARTITION_ROOT
) {
1411 if (!m
->partitions
[PARTITION_ROOT_VERITY
].found
|| !m
->partitions
[PARTITION_ROOT
].found
)
1412 return -EADDRNOTAVAIL
;
1414 /* If we found a verity setup, then the root partition is necessarily read-only. */
1415 m
->partitions
[PARTITION_ROOT
].rw
= false;
1416 m
->verity_ready
= true;
1419 assert(verity
->designator
== PARTITION_USR
);
1421 if (!m
->partitions
[PARTITION_USR_VERITY
].found
|| !m
->partitions
[PARTITION_USR
].found
)
1422 return -EADDRNOTAVAIL
;
1424 m
->partitions
[PARTITION_USR
].rw
= false;
1425 m
->verity_ready
= true;
1430 blkid_free_probe(b
);
1433 /* Fill in file system types if we don't know them yet. */
1434 for (PartitionDesignator i
= 0; i
< _PARTITION_DESIGNATOR_MAX
; i
++) {
1435 DissectedPartition
*p
= m
->partitions
+ i
;
1440 if (!p
->fstype
&& p
->node
) {
1441 r
= probe_filesystem(p
->node
, &p
->fstype
);
1442 if (r
< 0 && r
!= -EUCLEAN
)
1446 if (streq_ptr(p
->fstype
, "crypto_LUKS"))
1447 m
->encrypted
= true;
1449 if (p
->fstype
&& fstype_is_ro(p
->fstype
))
1463 DissectedImage
* dissected_image_unref(DissectedImage
*m
) {
1467 for (PartitionDesignator i
= 0; i
< _PARTITION_DESIGNATOR_MAX
; i
++)
1468 dissected_partition_done(m
->partitions
+ i
);
1470 free(m
->image_name
);
1472 strv_free(m
->machine_info
);
1473 strv_free(m
->os_release
);
1474 strv_free(m
->extension_release
);
1479 static int is_loop_device(const char *path
) {
1480 char s
[SYS_BLOCK_PATH_MAX("/../loop/")];
1485 if (stat(path
, &st
) < 0)
1488 if (!S_ISBLK(st
.st_mode
))
1491 xsprintf_sys_block_path(s
, "/loop/", st
.st_dev
);
1492 if (access(s
, F_OK
) < 0) {
1493 if (errno
!= ENOENT
)
1496 /* The device itself isn't a loop device, but maybe it's a partition and its parent is? */
1497 xsprintf_sys_block_path(s
, "/../loop/", st
.st_dev
);
1498 if (access(s
, F_OK
) < 0)
1499 return errno
== ENOENT
? false : -errno
;
1505 static int run_fsck(const char *node
, const char *fstype
) {
1512 r
= fsck_exists(fstype
);
1514 log_debug_errno(r
, "Couldn't determine whether fsck for %s exists, proceeding anyway.", fstype
);
1518 log_debug("Not checking partition %s, as fsck for %s does not exist.", node
, fstype
);
1522 r
= safe_fork("(fsck)", FORK_RESET_SIGNALS
|FORK_CLOSE_ALL_FDS
|FORK_RLIMIT_NOFILE_SAFE
|FORK_DEATHSIG
|FORK_NULL_STDIO
, &pid
);
1524 return log_debug_errno(r
, "Failed to fork off fsck: %m");
1527 execl("/sbin/fsck", "/sbin/fsck", "-aT", node
, NULL
);
1529 log_debug_errno(errno
, "Failed to execl() fsck: %m");
1530 _exit(FSCK_OPERATIONAL_ERROR
);
1533 exit_status
= wait_for_terminate_and_check("fsck", pid
, 0);
1534 if (exit_status
< 0)
1535 return log_debug_errno(exit_status
, "Failed to fork off /sbin/fsck: %m");
1537 if ((exit_status
& ~FSCK_ERROR_CORRECTED
) != FSCK_SUCCESS
) {
1538 log_debug("fsck failed with exit status %i.", exit_status
);
1540 if ((exit_status
& (FSCK_SYSTEM_SHOULD_REBOOT
|FSCK_ERRORS_LEFT_UNCORRECTED
)) != 0)
1541 return log_debug_errno(SYNTHETIC_ERRNO(EUCLEAN
), "File system is corrupted, refusing.");
1543 log_debug("Ignoring fsck error.");
1549 static int fs_grow(const char *node_path
, const char *mount_path
) {
1550 _cleanup_close_
int mount_fd
= -1, node_fd
= -1;
1551 uint64_t size
, newsize
;
1554 node_fd
= open(node_path
, O_RDONLY
|O_CLOEXEC
|O_NONBLOCK
|O_NOCTTY
);
1556 return log_debug_errno(errno
, "Failed to open node device %s: %m", node_path
);
1558 if (ioctl(node_fd
, BLKGETSIZE64
, &size
) != 0)
1559 return log_debug_errno(errno
, "Failed to get block device size of %s: %m", node_path
);
1561 mount_fd
= open(mount_path
, O_RDONLY
|O_DIRECTORY
|O_CLOEXEC
);
1563 return log_debug_errno(errno
, "Failed to open mountd file system %s: %m", mount_path
);
1565 log_debug("Resizing \"%s\" to %"PRIu64
" bytes...", mount_path
, size
);
1566 r
= resize_fs(mount_fd
, size
, &newsize
);
1568 return log_debug_errno(r
, "Failed to resize \"%s\" to %"PRIu64
" bytes: %m", mount_path
, size
);
1570 if (newsize
== size
)
1571 log_debug("Successfully resized \"%s\" to %s bytes.",
1572 mount_path
, FORMAT_BYTES(newsize
));
1574 assert(newsize
< size
);
1575 log_debug("Successfully resized \"%s\" to %s bytes (%"PRIu64
" bytes lost due to blocksize).",
1576 mount_path
, FORMAT_BYTES(newsize
), size
- newsize
);
1582 static int mount_partition(
1583 DissectedPartition
*m
,
1585 const char *directory
,
1588 DissectImageFlags flags
) {
1590 _cleanup_free_
char *chased
= NULL
, *options
= NULL
;
1591 const char *p
, *node
, *fstype
;
1592 bool rw
, remap_uid_gid
= false;
1598 /* Use decrypted node and matching fstype if available, otherwise use the original device */
1599 node
= m
->decrypted_node
?: m
->node
;
1600 fstype
= m
->decrypted_node
? m
->decrypted_fstype
: m
->fstype
;
1602 if (!m
->found
|| !node
)
1605 return -EAFNOSUPPORT
;
1607 /* 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. */
1608 if (streq(fstype
, "crypto_LUKS"))
1611 rw
= m
->rw
&& !(flags
& DISSECT_IMAGE_MOUNT_READ_ONLY
);
1613 if (FLAGS_SET(flags
, DISSECT_IMAGE_FSCK
) && rw
) {
1614 r
= run_fsck(node
, fstype
);
1620 /* Automatically create missing mount points inside the image, if necessary. */
1621 r
= mkdir_p_root(where
, directory
, uid_shift
, (gid_t
) uid_shift
, 0755);
1622 if (r
< 0 && r
!= -EROFS
)
1625 r
= chase_symlinks(directory
, where
, CHASE_PREFIX_ROOT
, &chased
, NULL
);
1631 /* Create top-level mount if missing – but only if this is asked for. This won't modify the
1632 * image (as the branch above does) but the host hierarchy, and the created directory might
1633 * survive our mount in the host hierarchy hence. */
1634 if (FLAGS_SET(flags
, DISSECT_IMAGE_MKDIR
)) {
1635 r
= mkdir_p(where
, 0755);
1643 /* If requested, turn on discard support. */
1644 if (fstype_can_discard(fstype
) &&
1645 ((flags
& DISSECT_IMAGE_DISCARD
) ||
1646 ((flags
& DISSECT_IMAGE_DISCARD_ON_LOOP
) && is_loop_device(m
->node
) > 0))) {
1647 options
= strdup("discard");
1652 if (uid_is_valid(uid_shift
) && uid_shift
!= 0) {
1654 if (fstype_can_uid_gid(fstype
)) {
1655 _cleanup_free_
char *uid_option
= NULL
;
1657 if (asprintf(&uid_option
, "uid=" UID_FMT
",gid=" GID_FMT
, uid_shift
, (gid_t
) uid_shift
) < 0)
1660 if (!strextend_with_separator(&options
, ",", uid_option
))
1662 } else if (FLAGS_SET(flags
, DISSECT_IMAGE_MOUNT_IDMAPPED
))
1663 remap_uid_gid
= true;
1666 if (!isempty(m
->mount_options
))
1667 if (!strextend_with_separator(&options
, ",", m
->mount_options
))
1670 /* So, when you request MS_RDONLY from ext4, then this means nothing. It happily still writes to the
1671 * backing storage. What's worse, the BLKRO[GS]ET flag and (in case of loopback devices)
1672 * LO_FLAGS_READ_ONLY don't mean anything, they affect userspace accesses only, and write accesses
1673 * from the upper file system still get propagated through to the underlying file system,
1674 * unrestricted. To actually get ext4/xfs/btrfs to stop writing to the device we need to specify
1675 * "norecovery" as mount option, in addition to MS_RDONLY. Yes, this sucks, since it means we need to
1676 * carry a per file system table here.
1678 * Note that this means that we might not be able to mount corrupted file systems as read-only
1679 * anymore (since in some cases the kernel implementations will refuse mounting when corrupted,
1680 * read-only and "norecovery" is specified). But I think for the case of automatically determined
1681 * mount options for loopback devices this is the right choice, since otherwise using the same
1682 * loopback file twice even in read-only mode, is going to fail badly sooner or later. The usecase of
1683 * making reuse of the immutable images "just work" is more relevant to us than having read-only
1684 * access that actually modifies stuff work on such image files. Or to say this differently: if
1685 * people want their file systems to be fixed up they should just open them in writable mode, where
1686 * all these problems don't exist. */
1687 if (!rw
&& STRPTR_IN_SET(fstype
, "ext3", "ext4", "xfs", "btrfs"))
1688 if (!strextend_with_separator(&options
, ",", "norecovery"))
1691 r
= mount_nofollow_verbose(LOG_DEBUG
, node
, p
, fstype
, MS_NODEV
|(rw
? 0 : MS_RDONLY
), options
);
1695 if (rw
&& m
->growfs
&& FLAGS_SET(flags
, DISSECT_IMAGE_GROWFS
))
1696 (void) fs_grow(node
, p
);
1698 if (remap_uid_gid
) {
1699 r
= remount_idmap(p
, uid_shift
, uid_range
);
1707 static int mount_root_tmpfs(const char *where
, uid_t uid_shift
, DissectImageFlags flags
) {
1708 _cleanup_free_
char *options
= NULL
;
1713 /* For images that contain /usr/ but no rootfs, let's mount rootfs as tmpfs */
1715 if (FLAGS_SET(flags
, DISSECT_IMAGE_MKDIR
)) {
1716 r
= mkdir_p(where
, 0755);
1721 if (uid_is_valid(uid_shift
)) {
1722 if (asprintf(&options
, "uid=" UID_FMT
",gid=" GID_FMT
, uid_shift
, (gid_t
) uid_shift
) < 0)
1726 r
= mount_nofollow_verbose(LOG_DEBUG
, "rootfs", where
, "tmpfs", MS_NODEV
, options
);
1733 int dissected_image_mount(
1738 DissectImageFlags flags
) {
1740 int r
, xbootldr_mounted
;
1747 * -ENXIO → No root partition found
1748 * -EMEDIUMTYPE → DISSECT_IMAGE_VALIDATE_OS set but no os-release/extension-release file found
1749 * -EUNATCH → Encrypted partition found for which no dm-crypt was set up yet
1750 * -EUCLEAN → fsck for file system failed
1751 * -EBUSY → File system already mounted/used elsewhere (kernel)
1752 * -EAFNOSUPPORT → File system type not supported or not known
1755 if (!(m
->partitions
[PARTITION_ROOT
].found
||
1756 (m
->partitions
[PARTITION_USR
].found
&& FLAGS_SET(flags
, DISSECT_IMAGE_USR_NO_ROOT
))))
1757 return -ENXIO
; /* Require a root fs or at least a /usr/ fs (the latter is subject to a flag of its own) */
1759 if ((flags
& DISSECT_IMAGE_MOUNT_NON_ROOT_ONLY
) == 0) {
1761 /* First mount the root fs. If there's none we use a tmpfs. */
1762 if (m
->partitions
[PARTITION_ROOT
].found
)
1763 r
= mount_partition(m
->partitions
+ PARTITION_ROOT
, where
, NULL
, uid_shift
, uid_range
, flags
);
1765 r
= mount_root_tmpfs(where
, uid_shift
, flags
);
1769 /* For us mounting root always means mounting /usr as well */
1770 r
= mount_partition(m
->partitions
+ PARTITION_USR
, where
, "/usr", uid_shift
, uid_range
, flags
);
1774 if ((flags
& (DISSECT_IMAGE_VALIDATE_OS
|DISSECT_IMAGE_VALIDATE_OS_EXT
)) != 0) {
1775 /* If either one of the validation flags are set, ensure that the image qualifies
1776 * as one or the other (or both). */
1779 if (FLAGS_SET(flags
, DISSECT_IMAGE_VALIDATE_OS
)) {
1780 r
= path_is_os_tree(where
);
1786 if (!ok
&& FLAGS_SET(flags
, DISSECT_IMAGE_VALIDATE_OS_EXT
)) {
1787 r
= path_is_extension_tree(where
, m
->image_name
);
1799 if (flags
& DISSECT_IMAGE_MOUNT_ROOT_ONLY
)
1802 r
= mount_partition(m
->partitions
+ PARTITION_HOME
, where
, "/home", uid_shift
, uid_range
, flags
);
1806 r
= mount_partition(m
->partitions
+ PARTITION_SRV
, where
, "/srv", uid_shift
, uid_range
, flags
);
1810 r
= mount_partition(m
->partitions
+ PARTITION_VAR
, where
, "/var", uid_shift
, uid_range
, flags
);
1814 r
= mount_partition(m
->partitions
+ PARTITION_TMP
, where
, "/var/tmp", uid_shift
, uid_range
, flags
);
1818 xbootldr_mounted
= mount_partition(m
->partitions
+ PARTITION_XBOOTLDR
, where
, "/boot", uid_shift
, uid_range
, flags
);
1819 if (xbootldr_mounted
< 0)
1820 return xbootldr_mounted
;
1822 if (m
->partitions
[PARTITION_ESP
].found
) {
1823 int esp_done
= false;
1825 /* Mount the ESP to /efi if it exists. If it doesn't exist, use /boot instead, but only if it
1826 * exists and is empty, and we didn't already mount the XBOOTLDR partition into it. */
1828 r
= chase_symlinks("/efi", where
, CHASE_PREFIX_ROOT
, NULL
, NULL
);
1833 /* /efi doesn't exist. Let's see if /boot is suitable then */
1835 if (!xbootldr_mounted
) {
1836 _cleanup_free_
char *p
= NULL
;
1838 r
= chase_symlinks("/boot", where
, CHASE_PREFIX_ROOT
, &p
, NULL
);
1842 } else if (dir_is_empty(p
) > 0) {
1843 /* It exists and is an empty directory. Let's mount the ESP there. */
1844 r
= mount_partition(m
->partitions
+ PARTITION_ESP
, where
, "/boot", uid_shift
, uid_range
, flags
);
1854 /* OK, let's mount the ESP now to /efi (possibly creating the dir if missing) */
1856 r
= mount_partition(m
->partitions
+ PARTITION_ESP
, where
, "/efi", uid_shift
, uid_range
, flags
);
1865 int dissected_image_mount_and_warn(
1870 DissectImageFlags flags
) {
1877 r
= dissected_image_mount(m
, where
, uid_shift
, uid_range
, flags
);
1879 return log_error_errno(r
, "Not root file system found in image.");
1880 if (r
== -EMEDIUMTYPE
)
1881 return log_error_errno(r
, "No suitable os-release/extension-release file in image found.");
1883 return log_error_errno(r
, "Encrypted file system discovered, but decryption not requested.");
1885 return log_error_errno(r
, "File system check on image failed.");
1887 return log_error_errno(r
, "File system already mounted elsewhere.");
1888 if (r
== -EAFNOSUPPORT
)
1889 return log_error_errno(r
, "File system type not supported or not known.");
1891 return log_error_errno(r
, "Failed to mount image: %m");
1896 #if HAVE_LIBCRYPTSETUP
1897 typedef struct DecryptedPartition
{
1898 struct crypt_device
*device
;
1901 } DecryptedPartition
;
1903 struct DecryptedImage
{
1904 DecryptedPartition
*decrypted
;
1909 DecryptedImage
* decrypted_image_unref(DecryptedImage
* d
) {
1910 #if HAVE_LIBCRYPTSETUP
1916 for (size_t i
= 0; i
< d
->n_decrypted
; i
++) {
1917 DecryptedPartition
*p
= d
->decrypted
+ i
;
1919 if (p
->device
&& p
->name
&& !p
->relinquished
) {
1920 r
= sym_crypt_deactivate_by_name(p
->device
, p
->name
, 0);
1922 log_debug_errno(r
, "Failed to deactivate encrypted partition %s", p
->name
);
1926 sym_crypt_free(p
->device
);
1936 #if HAVE_LIBCRYPTSETUP
1938 static int make_dm_name_and_node(const void *original_node
, const char *suffix
, char **ret_name
, char **ret_node
) {
1939 _cleanup_free_
char *name
= NULL
, *node
= NULL
;
1942 assert(original_node
);
1947 base
= strrchr(original_node
, '/');
1949 base
= original_node
;
1955 name
= strjoin(base
, suffix
);
1958 if (!filename_is_valid(name
))
1961 node
= path_join(sym_crypt_get_dir(), name
);
1965 *ret_name
= TAKE_PTR(name
);
1966 *ret_node
= TAKE_PTR(node
);
1971 static int decrypt_partition(
1972 DissectedPartition
*m
,
1973 const char *passphrase
,
1974 DissectImageFlags flags
,
1975 DecryptedImage
*d
) {
1977 _cleanup_free_
char *node
= NULL
, *name
= NULL
;
1978 _cleanup_(sym_crypt_freep
) struct crypt_device
*cd
= NULL
;
1984 if (!m
->found
|| !m
->node
|| !m
->fstype
)
1987 if (!streq(m
->fstype
, "crypto_LUKS"))
1993 r
= dlopen_cryptsetup();
1997 r
= make_dm_name_and_node(m
->node
, "-decrypted", &name
, &node
);
2001 if (!GREEDY_REALLOC0(d
->decrypted
, d
->n_decrypted
+ 1))
2004 r
= sym_crypt_init(&cd
, m
->node
);
2006 return log_debug_errno(r
, "Failed to initialize dm-crypt: %m");
2008 cryptsetup_enable_logging(cd
);
2010 r
= sym_crypt_load(cd
, CRYPT_LUKS
, NULL
);
2012 return log_debug_errno(r
, "Failed to load LUKS metadata: %m");
2014 r
= sym_crypt_activate_by_passphrase(cd
, name
, CRYPT_ANY_SLOT
, passphrase
, strlen(passphrase
),
2015 ((flags
& DISSECT_IMAGE_DEVICE_READ_ONLY
) ? CRYPT_ACTIVATE_READONLY
: 0) |
2016 ((flags
& DISSECT_IMAGE_DISCARD_ON_CRYPTO
) ? CRYPT_ACTIVATE_ALLOW_DISCARDS
: 0));
2018 log_debug_errno(r
, "Failed to activate LUKS device: %m");
2019 return r
== -EPERM
? -EKEYREJECTED
: r
;
2022 d
->decrypted
[d
->n_decrypted
++] = (DecryptedPartition
) {
2023 .name
= TAKE_PTR(name
),
2024 .device
= TAKE_PTR(cd
),
2027 m
->decrypted_node
= TAKE_PTR(node
);
2032 static int verity_can_reuse(
2033 const VeritySettings
*verity
,
2035 struct crypt_device
**ret_cd
) {
2037 /* If the same volume was already open, check that the root hashes match, and reuse it if they do */
2038 _cleanup_free_
char *root_hash_existing
= NULL
;
2039 _cleanup_(sym_crypt_freep
) struct crypt_device
*cd
= NULL
;
2040 struct crypt_params_verity crypt_params
= {};
2041 size_t root_hash_existing_size
;
2048 r
= sym_crypt_init_by_name(&cd
, name
);
2050 return log_debug_errno(r
, "Error opening verity device, crypt_init_by_name failed: %m");
2052 cryptsetup_enable_logging(cd
);
2054 r
= sym_crypt_get_verity_info(cd
, &crypt_params
);
2056 return log_debug_errno(r
, "Error opening verity device, crypt_get_verity_info failed: %m");
2058 root_hash_existing_size
= verity
->root_hash_size
;
2059 root_hash_existing
= malloc0(root_hash_existing_size
);
2060 if (!root_hash_existing
)
2063 r
= sym_crypt_volume_key_get(cd
, CRYPT_ANY_SLOT
, root_hash_existing
, &root_hash_existing_size
, NULL
, 0);
2065 return log_debug_errno(r
, "Error opening verity device, crypt_volume_key_get failed: %m");
2066 if (verity
->root_hash_size
!= root_hash_existing_size
||
2067 memcmp(root_hash_existing
, verity
->root_hash
, verity
->root_hash_size
) != 0)
2068 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL
), "Error opening verity device, it already exists but root hashes are different.");
2070 #if HAVE_CRYPT_ACTIVATE_BY_SIGNED_KEY
2071 /* Ensure that, if signatures are supported, we only reuse the device if the previous mount used the
2072 * same settings, so that a previous unsigned mount will not be reused if the user asks to use
2073 * signing for the new one, and vice versa. */
2074 if (!!verity
->root_hash_sig
!= !!(crypt_params
.flags
& CRYPT_VERITY_ROOT_HASH_SIGNATURE
))
2075 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL
), "Error opening verity device, it already exists but signature settings are not the same.");
2078 *ret_cd
= TAKE_PTR(cd
);
2082 static inline char* dm_deferred_remove_clean(char *name
) {
2086 (void) sym_crypt_deactivate_by_name(NULL
, name
, CRYPT_DEACTIVATE_DEFERRED
);
2089 DEFINE_TRIVIAL_CLEANUP_FUNC(char *, dm_deferred_remove_clean
);
2091 static int verity_partition(
2092 PartitionDesignator designator
,
2093 DissectedPartition
*m
,
2094 DissectedPartition
*v
,
2095 const VeritySettings
*verity
,
2096 DissectImageFlags flags
,
2097 DecryptedImage
*d
) {
2099 _cleanup_(sym_crypt_freep
) struct crypt_device
*cd
= NULL
;
2100 _cleanup_(dm_deferred_remove_cleanp
) char *restore_deferred_remove
= NULL
;
2101 _cleanup_free_
char *node
= NULL
, *name
= NULL
;
2105 assert(v
|| (verity
&& verity
->data_path
));
2107 if (!verity
|| !verity
->root_hash
)
2109 if (!((verity
->designator
< 0 && designator
== PARTITION_ROOT
) ||
2110 (verity
->designator
== designator
)))
2113 if (!m
->found
|| !m
->node
|| !m
->fstype
)
2115 if (!verity
->data_path
) {
2116 if (!v
->found
|| !v
->node
|| !v
->fstype
)
2119 if (!streq(v
->fstype
, "DM_verity_hash"))
2123 r
= dlopen_cryptsetup();
2127 if (FLAGS_SET(flags
, DISSECT_IMAGE_VERITY_SHARE
)) {
2128 /* Use the roothash, which is unique per volume, as the device node name, so that it can be reused */
2129 _cleanup_free_
char *root_hash_encoded
= NULL
;
2131 root_hash_encoded
= hexmem(verity
->root_hash
, verity
->root_hash_size
);
2132 if (!root_hash_encoded
)
2135 r
= make_dm_name_and_node(root_hash_encoded
, "-verity", &name
, &node
);
2137 r
= make_dm_name_and_node(m
->node
, "-verity", &name
, &node
);
2141 r
= sym_crypt_init(&cd
, verity
->data_path
?: v
->node
);
2145 cryptsetup_enable_logging(cd
);
2147 r
= sym_crypt_load(cd
, CRYPT_VERITY
, NULL
);
2151 r
= sym_crypt_set_data_device(cd
, m
->node
);
2155 if (!GREEDY_REALLOC0(d
->decrypted
, d
->n_decrypted
+ 1))
2158 /* If activating fails because the device already exists, check the metadata and reuse it if it matches.
2159 * In case of ENODEV/ENOENT, which can happen if another process is activating at the exact same time,
2160 * retry a few times before giving up. */
2161 for (unsigned i
= 0; i
< N_DEVICE_NODE_LIST_ATTEMPTS
; i
++) {
2162 if (verity
->root_hash_sig
) {
2163 #if HAVE_CRYPT_ACTIVATE_BY_SIGNED_KEY
2164 r
= sym_crypt_activate_by_signed_key(
2168 verity
->root_hash_size
,
2169 verity
->root_hash_sig
,
2170 verity
->root_hash_sig_size
,
2171 CRYPT_ACTIVATE_READONLY
);
2173 r
= log_debug_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
2174 "Activation of verity device with signature requested, but not supported by %s due to missing crypt_activate_by_signed_key().", program_invocation_short_name
);
2177 r
= sym_crypt_activate_by_volume_key(
2181 verity
->root_hash_size
,
2182 CRYPT_ACTIVATE_READONLY
);
2183 /* libdevmapper can return EINVAL when the device is already in the activation stage.
2184 * There's no way to distinguish this situation from a genuine error due to invalid
2185 * parameters, so immediately fall back to activating the device with a unique name.
2186 * Improvements in libcrypsetup can ensure this never happens:
2187 * https://gitlab.com/cryptsetup/cryptsetup/-/merge_requests/96 */
2188 if (r
== -EINVAL
&& FLAGS_SET(flags
, DISSECT_IMAGE_VERITY_SHARE
))
2189 return verity_partition(designator
, m
, v
, verity
, flags
& ~DISSECT_IMAGE_VERITY_SHARE
, d
);
2192 -EEXIST
, /* Volume is already open and ready to be used */
2193 -EBUSY
, /* Volume is being opened but not ready, crypt_init_by_name can fetch details */
2194 -ENODEV
/* Volume is being opened but not ready, crypt_init_by_name would fail, try to open again */))
2196 if (IN_SET(r
, -EEXIST
, -EBUSY
)) {
2197 struct crypt_device
*existing_cd
= NULL
;
2199 if (!restore_deferred_remove
){
2200 /* To avoid races, disable automatic removal on umount while setting up the new device. Restore it on failure. */
2201 r
= dm_deferred_remove_cancel(name
);
2202 /* If activation returns EBUSY there might be no deferred removal to cancel, that's fine */
2203 if (r
< 0 && r
!= -ENXIO
)
2204 return log_debug_errno(r
, "Disabling automated deferred removal for verity device %s failed: %m", node
);
2206 restore_deferred_remove
= strdup(name
);
2207 if (!restore_deferred_remove
)
2212 r
= verity_can_reuse(verity
, name
, &existing_cd
);
2213 /* Same as above, -EINVAL can randomly happen when it actually means -EEXIST */
2214 if (r
== -EINVAL
&& FLAGS_SET(flags
, DISSECT_IMAGE_VERITY_SHARE
))
2215 return verity_partition(designator
, m
, v
, verity
, flags
& ~DISSECT_IMAGE_VERITY_SHARE
, d
);
2216 if (!IN_SET(r
, 0, -ENODEV
, -ENOENT
, -EBUSY
))
2217 return log_debug_errno(r
, "Checking whether existing verity device %s can be reused failed: %m", node
);
2219 /* devmapper might say that the device exists, but the devlink might not yet have been
2220 * created. Check and wait for the udev event in that case. */
2221 r
= device_wait_for_devlink(node
, "block", usec_add(now(CLOCK_MONOTONIC
), 100 * USEC_PER_MSEC
), NULL
);
2222 /* Fallback to activation with a unique device if it's taking too long */
2223 if (r
== -ETIMEDOUT
)
2236 /* Device is being opened by another process, but it has not finished yet, yield for 2ms */
2237 (void) usleep(2 * USEC_PER_MSEC
);
2240 /* An existing verity device was reported by libcryptsetup/libdevmapper, but we can't use it at this time.
2241 * Fall back to activating it with a unique device name. */
2242 if (r
!= 0 && FLAGS_SET(flags
, DISSECT_IMAGE_VERITY_SHARE
))
2243 return verity_partition(designator
, m
, v
, verity
, flags
& ~DISSECT_IMAGE_VERITY_SHARE
, d
);
2245 /* Everything looks good and we'll be able to mount the device, so deferred remove will be re-enabled at that point. */
2246 restore_deferred_remove
= mfree(restore_deferred_remove
);
2248 d
->decrypted
[d
->n_decrypted
++] = (DecryptedPartition
) {
2249 .name
= TAKE_PTR(name
),
2250 .device
= TAKE_PTR(cd
),
2253 m
->decrypted_node
= TAKE_PTR(node
);
2259 int dissected_image_decrypt(
2261 const char *passphrase
,
2262 const VeritySettings
*verity
,
2263 DissectImageFlags flags
,
2264 DecryptedImage
**ret
) {
2266 #if HAVE_LIBCRYPTSETUP
2267 _cleanup_(decrypted_image_unrefp
) DecryptedImage
*d
= NULL
;
2272 assert(!verity
|| verity
->root_hash
|| verity
->root_hash_size
== 0);
2276 * = 0 → There was nothing to decrypt
2277 * > 0 → Decrypted successfully
2278 * -ENOKEY → There's something to decrypt but no key was supplied
2279 * -EKEYREJECTED → Passed key was not correct
2282 if (verity
&& verity
->root_hash
&& verity
->root_hash_size
< sizeof(sd_id128_t
))
2285 if (!m
->encrypted
&& !m
->verity_ready
) {
2290 #if HAVE_LIBCRYPTSETUP
2291 d
= new0(DecryptedImage
, 1);
2295 for (PartitionDesignator i
= 0; i
< _PARTITION_DESIGNATOR_MAX
; i
++) {
2296 DissectedPartition
*p
= m
->partitions
+ i
;
2297 PartitionDesignator k
;
2302 r
= decrypt_partition(p
, passphrase
, flags
, d
);
2306 k
= PARTITION_VERITY_OF(i
);
2308 r
= verity_partition(i
, p
, m
->partitions
+ k
, verity
, flags
| DISSECT_IMAGE_VERITY_SHARE
, d
);
2313 if (!p
->decrypted_fstype
&& p
->decrypted_node
) {
2314 r
= probe_filesystem(p
->decrypted_node
, &p
->decrypted_fstype
);
2315 if (r
< 0 && r
!= -EUCLEAN
)
2328 int dissected_image_decrypt_interactively(
2330 const char *passphrase
,
2331 const VeritySettings
*verity
,
2332 DissectImageFlags flags
,
2333 DecryptedImage
**ret
) {
2335 _cleanup_strv_free_erase_
char **z
= NULL
;
2342 r
= dissected_image_decrypt(m
, passphrase
, verity
, flags
, ret
);
2345 if (r
== -EKEYREJECTED
)
2346 log_error_errno(r
, "Incorrect passphrase, try again!");
2347 else if (r
!= -ENOKEY
)
2348 return log_error_errno(r
, "Failed to decrypt image: %m");
2351 return log_error_errno(SYNTHETIC_ERRNO(EKEYREJECTED
),
2352 "Too many retries.");
2356 r
= ask_password_auto("Please enter image passphrase:", NULL
, "dissect", "dissect", "dissect.passphrase", USEC_INFINITY
, 0, &z
);
2358 return log_error_errno(r
, "Failed to query for passphrase: %m");
2364 int decrypted_image_relinquish(DecryptedImage
*d
) {
2367 /* Turns on automatic removal after the last use ended for all DM devices of this image, and sets a
2368 * boolean so that we don't clean it up ourselves either anymore */
2370 #if HAVE_LIBCRYPTSETUP
2373 for (size_t i
= 0; i
< d
->n_decrypted
; i
++) {
2374 DecryptedPartition
*p
= d
->decrypted
+ i
;
2376 if (p
->relinquished
)
2379 r
= sym_crypt_deactivate_by_name(NULL
, p
->name
, CRYPT_DEACTIVATE_DEFERRED
);
2381 return log_debug_errno(r
, "Failed to mark %s for auto-removal: %m", p
->name
);
2383 p
->relinquished
= true;
2390 static char *build_auxiliary_path(const char *image
, const char *suffix
) {
2397 e
= endswith(image
, ".raw");
2399 return strjoin(e
, suffix
);
2401 n
= new(char, e
- image
+ strlen(suffix
) + 1);
2405 strcpy(mempcpy(n
, image
, e
- image
), suffix
);
2409 void verity_settings_done(VeritySettings
*v
) {
2412 v
->root_hash
= mfree(v
->root_hash
);
2413 v
->root_hash_size
= 0;
2415 v
->root_hash_sig
= mfree(v
->root_hash_sig
);
2416 v
->root_hash_sig_size
= 0;
2418 v
->data_path
= mfree(v
->data_path
);
2421 int verity_settings_load(
2422 VeritySettings
*verity
,
2424 const char *root_hash_path
,
2425 const char *root_hash_sig_path
) {
2427 _cleanup_free_
void *root_hash
= NULL
, *root_hash_sig
= NULL
;
2428 size_t root_hash_size
= 0, root_hash_sig_size
= 0;
2429 _cleanup_free_
char *verity_data_path
= NULL
;
2430 PartitionDesignator designator
;
2435 assert(verity
->designator
< 0 || IN_SET(verity
->designator
, PARTITION_ROOT
, PARTITION_USR
));
2437 /* If we are asked to load the root hash for a device node, exit early */
2438 if (is_device_path(image
))
2441 designator
= verity
->designator
;
2443 /* We only fill in what isn't already filled in */
2445 if (!verity
->root_hash
) {
2446 _cleanup_free_
char *text
= NULL
;
2448 if (root_hash_path
) {
2449 /* If explicitly specified it takes precedence */
2450 r
= read_one_line_file(root_hash_path
, &text
);
2455 designator
= PARTITION_ROOT
;
2457 /* Otherwise look for xattr and separate file, and first for the data for root and if
2458 * that doesn't exist for /usr */
2460 if (designator
< 0 || designator
== PARTITION_ROOT
) {
2461 r
= getxattr_malloc(image
, "user.verity.roothash", &text
, true);
2463 _cleanup_free_
char *p
= NULL
;
2465 if (!IN_SET(r
, -ENODATA
, -ENOENT
) && !ERRNO_IS_NOT_SUPPORTED(r
))
2468 p
= build_auxiliary_path(image
, ".roothash");
2472 r
= read_one_line_file(p
, &text
);
2473 if (r
< 0 && r
!= -ENOENT
)
2478 designator
= PARTITION_ROOT
;
2481 if (!text
&& (designator
< 0 || designator
== PARTITION_USR
)) {
2482 /* So in the "roothash" xattr/file name above the "root" of course primarily
2483 * refers to the root of the Verity Merkle tree. But coincidentally it also
2484 * is the hash for the *root* file system, i.e. the "root" neatly refers to
2485 * two distinct concepts called "root". Taking benefit of this happy
2486 * coincidence we call the file with the root hash for the /usr/ file system
2487 * `usrhash`, because `usrroothash` or `rootusrhash` would just be too
2488 * confusing. We thus drop the reference to the root of the Merkle tree, and
2489 * just indicate which file system it's about. */
2490 r
= getxattr_malloc(image
, "user.verity.usrhash", &text
, true);
2492 _cleanup_free_
char *p
= NULL
;
2494 if (!IN_SET(r
, -ENODATA
, -ENOENT
) && !ERRNO_IS_NOT_SUPPORTED(r
))
2497 p
= build_auxiliary_path(image
, ".usrhash");
2501 r
= read_one_line_file(p
, &text
);
2502 if (r
< 0 && r
!= -ENOENT
)
2507 designator
= PARTITION_USR
;
2512 r
= unhexmem(text
, strlen(text
), &root_hash
, &root_hash_size
);
2515 if (root_hash_size
< sizeof(sd_id128_t
))
2520 if ((root_hash
|| verity
->root_hash
) && !verity
->root_hash_sig
) {
2521 if (root_hash_sig_path
) {
2522 r
= read_full_file(root_hash_sig_path
, (char**) &root_hash_sig
, &root_hash_sig_size
);
2523 if (r
< 0 && r
!= -ENOENT
)
2527 designator
= PARTITION_ROOT
;
2529 if (designator
< 0 || designator
== PARTITION_ROOT
) {
2530 _cleanup_free_
char *p
= NULL
;
2532 /* Follow naming convention recommended by the relevant RFC:
2533 * https://tools.ietf.org/html/rfc5751#section-3.2.1 */
2534 p
= build_auxiliary_path(image
, ".roothash.p7s");
2538 r
= read_full_file(p
, (char**) &root_hash_sig
, &root_hash_sig_size
);
2539 if (r
< 0 && r
!= -ENOENT
)
2542 designator
= PARTITION_ROOT
;
2545 if (!root_hash_sig
&& (designator
< 0 || designator
== PARTITION_USR
)) {
2546 _cleanup_free_
char *p
= NULL
;
2548 p
= build_auxiliary_path(image
, ".usrhash.p7s");
2552 r
= read_full_file(p
, (char**) &root_hash_sig
, &root_hash_sig_size
);
2553 if (r
< 0 && r
!= -ENOENT
)
2556 designator
= PARTITION_USR
;
2560 if (root_hash_sig
&& root_hash_sig_size
== 0) /* refuse empty size signatures */
2564 if (!verity
->data_path
) {
2565 _cleanup_free_
char *p
= NULL
;
2567 p
= build_auxiliary_path(image
, ".verity");
2571 if (access(p
, F_OK
) < 0) {
2572 if (errno
!= ENOENT
)
2575 verity_data_path
= TAKE_PTR(p
);
2579 verity
->root_hash
= TAKE_PTR(root_hash
);
2580 verity
->root_hash_size
= root_hash_size
;
2583 if (root_hash_sig
) {
2584 verity
->root_hash_sig
= TAKE_PTR(root_hash_sig
);
2585 verity
->root_hash_sig_size
= root_hash_sig_size
;
2588 if (verity_data_path
)
2589 verity
->data_path
= TAKE_PTR(verity_data_path
);
2591 if (verity
->designator
< 0)
2592 verity
->designator
= designator
;
2597 int dissected_image_acquire_metadata(DissectedImage
*m
) {
2604 META_EXTENSION_RELEASE
,
2608 static const char *const paths
[_META_MAX
] = {
2609 [META_HOSTNAME
] = "/etc/hostname\0",
2610 [META_MACHINE_ID
] = "/etc/machine-id\0",
2611 [META_MACHINE_INFO
] = "/etc/machine-info\0",
2612 [META_OS_RELEASE
] = ("/etc/os-release\0"
2613 "/usr/lib/os-release\0"),
2614 [META_EXTENSION_RELEASE
] = "extension-release\0", /* Used only for logging. */
2617 _cleanup_strv_free_
char **machine_info
= NULL
, **os_release
= NULL
, **extension_release
= NULL
;
2618 _cleanup_close_pair_
int error_pipe
[2] = { -1, -1 };
2619 _cleanup_(rmdir_and_freep
) char *t
= NULL
;
2620 _cleanup_(sigkill_waitp
) pid_t child
= 0;
2621 sd_id128_t machine_id
= SD_ID128_NULL
;
2622 _cleanup_free_
char *hostname
= NULL
;
2623 unsigned n_meta_initialized
= 0;
2624 int fds
[2 * _META_MAX
], r
, v
;
2627 BLOCK_SIGNALS(SIGCHLD
);
2631 for (; n_meta_initialized
< _META_MAX
; n_meta_initialized
++) {
2632 if (!paths
[n_meta_initialized
]) {
2633 fds
[2*n_meta_initialized
] = fds
[2*n_meta_initialized
+1] = -1;
2637 if (pipe2(fds
+ 2*n_meta_initialized
, O_CLOEXEC
) < 0) {
2643 r
= mkdtemp_malloc("/tmp/dissect-XXXXXX", &t
);
2647 if (pipe2(error_pipe
, O_CLOEXEC
) < 0) {
2652 r
= safe_fork("(sd-dissect)", FORK_RESET_SIGNALS
|FORK_DEATHSIG
|FORK_NEW_MOUNTNS
|FORK_MOUNTNS_SLAVE
, &child
);
2656 error_pipe
[0] = safe_close(error_pipe
[0]);
2658 r
= dissected_image_mount(
2663 DISSECT_IMAGE_READ_ONLY
|
2664 DISSECT_IMAGE_MOUNT_ROOT_ONLY
|
2665 DISSECT_IMAGE_VALIDATE_OS
|
2666 DISSECT_IMAGE_VALIDATE_OS_EXT
|
2667 DISSECT_IMAGE_USR_NO_ROOT
);
2669 /* Let parent know the error */
2670 (void) write(error_pipe
[1], &r
, sizeof(r
));
2672 log_debug_errno(r
, "Failed to mount dissected image: %m");
2673 _exit(EXIT_FAILURE
);
2676 for (unsigned k
= 0; k
< _META_MAX
; k
++) {
2677 _cleanup_close_
int fd
= -ENOENT
;
2683 fds
[2*k
] = safe_close(fds
[2*k
]);
2685 if (k
== META_EXTENSION_RELEASE
) {
2686 /* As per the os-release spec, if the image is an extension it will have a file
2687 * named after the image name in extension-release.d/ - we use the image name
2688 * and try to resolve it with the extension-release helpers, as sometimes
2689 * the image names are mangled on deployment and do not match anymore.
2690 * Unlike other paths this is not fixed, and the image name
2691 * can be mangled on deployment, so by calling into the helper
2692 * we allow a fallback that matches on the first extension-release
2693 * file found in the directory, if one named after the image cannot
2694 * be found first. */
2695 r
= open_extension_release(t
, m
->image_name
, NULL
, &fd
);
2697 fd
= r
; /* Propagate the error. */
2699 NULSTR_FOREACH(p
, paths
[k
]) {
2700 fd
= chase_symlinks_and_open(p
, t
, CHASE_PREFIX_ROOT
, O_RDONLY
|O_CLOEXEC
|O_NOCTTY
, NULL
);
2705 log_debug_errno(fd
, "Failed to read %s file of image, ignoring: %m", paths
[k
]);
2706 fds
[2*k
+1] = safe_close(fds
[2*k
+1]);
2710 r
= copy_bytes(fd
, fds
[2*k
+1], UINT64_MAX
, 0);
2712 (void) write(error_pipe
[1], &r
, sizeof(r
));
2713 _exit(EXIT_FAILURE
);
2716 fds
[2*k
+1] = safe_close(fds
[2*k
+1]);
2719 _exit(EXIT_SUCCESS
);
2722 error_pipe
[1] = safe_close(error_pipe
[1]);
2724 for (unsigned k
= 0; k
< _META_MAX
; k
++) {
2725 _cleanup_fclose_
FILE *f
= NULL
;
2730 fds
[2*k
+1] = safe_close(fds
[2*k
+1]);
2732 f
= take_fdopen(&fds
[2*k
], "r");
2741 r
= read_etc_hostname_stream(f
, &hostname
);
2743 log_debug_errno(r
, "Failed to read /etc/hostname: %m");
2747 case META_MACHINE_ID
: {
2748 _cleanup_free_
char *line
= NULL
;
2750 r
= read_line(f
, LONG_LINE_MAX
, &line
);
2752 log_debug_errno(r
, "Failed to read /etc/machine-id: %m");
2754 r
= sd_id128_from_string(line
, &machine_id
);
2756 log_debug_errno(r
, "Image contains invalid /etc/machine-id: %s", line
);
2758 log_debug("/etc/machine-id file is empty.");
2759 else if (streq(line
, "uninitialized"))
2760 log_debug("/etc/machine-id file is uninitialized (likely aborted first boot).");
2762 log_debug("/etc/machine-id has unexpected length %i.", r
);
2767 case META_MACHINE_INFO
:
2768 r
= load_env_file_pairs(f
, "machine-info", &machine_info
);
2770 log_debug_errno(r
, "Failed to read /etc/machine-info: %m");
2774 case META_OS_RELEASE
:
2775 r
= load_env_file_pairs(f
, "os-release", &os_release
);
2777 log_debug_errno(r
, "Failed to read OS release file: %m");
2781 case META_EXTENSION_RELEASE
:
2782 r
= load_env_file_pairs(f
, "extension-release", &extension_release
);
2784 log_debug_errno(r
, "Failed to read extension release file: %m");
2790 r
= wait_for_terminate_and_check("(sd-dissect)", child
, 0);
2795 n
= read(error_pipe
[0], &v
, sizeof(v
));
2799 return v
; /* propagate error sent to us from child */
2803 if (r
!= EXIT_SUCCESS
)
2806 free_and_replace(m
->hostname
, hostname
);
2807 m
->machine_id
= machine_id
;
2808 strv_free_and_replace(m
->machine_info
, machine_info
);
2809 strv_free_and_replace(m
->os_release
, os_release
);
2810 strv_free_and_replace(m
->extension_release
, extension_release
);
2813 for (unsigned k
= 0; k
< n_meta_initialized
; k
++)
2814 safe_close_pair(fds
+ 2*k
);
2819 int dissect_image_and_warn(
2822 const VeritySettings
*verity
,
2823 const MountOptions
*mount_options
,
2825 uint64_t uevent_seqnum_not_before
,
2826 usec_t timestamp_not_before
,
2827 DissectImageFlags flags
,
2828 DissectedImage
**ret
) {
2830 _cleanup_free_
char *buffer
= NULL
;
2834 r
= fd_get_path(fd
, &buffer
);
2841 r
= dissect_image(fd
, verity
, mount_options
, diskseq
, uevent_seqnum_not_before
, timestamp_not_before
, flags
, ret
);
2845 return log_error_errno(r
, "Dissecting images is not supported, compiled without blkid support.");
2848 return log_error_errno(r
, "Couldn't identify a suitable partition table or file system in '%s'.", name
);
2850 case -EADDRNOTAVAIL
:
2851 return log_error_errno(r
, "No root partition for specified root hash found in '%s'.", name
);
2854 return log_error_errno(r
, "Multiple suitable root partitions found in image '%s'.", name
);
2857 return log_error_errno(r
, "No suitable root partition found in image '%s'.", name
);
2859 case -EPROTONOSUPPORT
:
2860 return log_error_errno(r
, "Device '%s' is loopback block device with partition scanning turned off, please turn it on.", name
);
2864 return log_error_errno(r
, "Failed to dissect image '%s': %m", name
);
2870 bool dissected_image_verity_candidate(const DissectedImage
*image
, PartitionDesignator partition_designator
) {
2873 /* Checks if this partition could theoretically do Verity. For non-partitioned images this only works
2874 * if there's an external verity file supplied, for which we can consult .has_verity. For partitioned
2875 * images we only check the partition type.
2877 * This call is used to decide whether to suppress or show a verity column in tabular output of the
2880 if (image
->single_file_system
)
2881 return partition_designator
== PARTITION_ROOT
&& image
->has_verity
;
2883 return PARTITION_VERITY_OF(partition_designator
) >= 0;
2886 bool dissected_image_verity_ready(const DissectedImage
*image
, PartitionDesignator partition_designator
) {
2887 PartitionDesignator k
;
2891 /* Checks if this partition has verity data available that we can activate. For non-partitioned this
2892 * works for the root partition, for others only if the associated verity partition was found. */
2894 if (!image
->verity_ready
)
2897 if (image
->single_file_system
)
2898 return partition_designator
== PARTITION_ROOT
;
2900 k
= PARTITION_VERITY_OF(partition_designator
);
2901 return k
>= 0 && image
->partitions
[k
].found
;
2904 MountOptions
* mount_options_free_all(MountOptions
*options
) {
2907 while ((m
= options
)) {
2908 LIST_REMOVE(mount_options
, options
, m
);
2916 const char* mount_options_from_designator(const MountOptions
*options
, PartitionDesignator designator
) {
2917 const MountOptions
*m
;
2919 LIST_FOREACH(mount_options
, m
, options
)
2920 if (designator
== m
->partition_designator
&& !isempty(m
->options
))
2926 int mount_image_privately_interactively(
2928 DissectImageFlags flags
,
2929 char **ret_directory
,
2930 LoopDevice
**ret_loop_device
,
2931 DecryptedImage
**ret_decrypted_image
) {
2933 _cleanup_(verity_settings_done
) VeritySettings verity
= VERITY_SETTINGS_DEFAULT
;
2934 _cleanup_(loop_device_unrefp
) LoopDevice
*d
= NULL
;
2935 _cleanup_(decrypted_image_unrefp
) DecryptedImage
*decrypted_image
= NULL
;
2936 _cleanup_(dissected_image_unrefp
) DissectedImage
*dissected_image
= NULL
;
2937 _cleanup_(rmdir_and_freep
) char *created_dir
= NULL
;
2938 _cleanup_free_
char *temp
= NULL
;
2941 /* Mounts an OS image at a temporary place, inside a newly created mount namespace of our own. This
2942 * is used by tools such as systemd-tmpfiles or systemd-firstboot to operate on some disk image
2946 assert(ret_directory
);
2947 assert(ret_loop_device
);
2948 assert(ret_decrypted_image
);
2950 r
= verity_settings_load(&verity
, image
, NULL
, NULL
);
2952 return log_error_errno(r
, "Failed to load root hash data: %m");
2954 r
= tempfn_random_child(NULL
, program_invocation_short_name
, &temp
);
2956 return log_error_errno(r
, "Failed to generate temporary mount directory: %m");
2958 r
= loop_device_make_by_path(
2960 FLAGS_SET(flags
, DISSECT_IMAGE_DEVICE_READ_ONLY
) ? O_RDONLY
: O_RDWR
,
2961 FLAGS_SET(flags
, DISSECT_IMAGE_NO_PARTITION_TABLE
) ? 0 : LO_FLAGS_PARTSCAN
,
2964 return log_error_errno(r
, "Failed to set up loopback device for %s: %m", image
);
2966 r
= dissect_image_and_warn(d
->fd
, image
, &verity
, NULL
, d
->diskseq
, d
->uevent_seqnum_not_before
, d
->timestamp_not_before
, flags
, &dissected_image
);
2970 r
= dissected_image_decrypt_interactively(dissected_image
, NULL
, &verity
, flags
, &decrypted_image
);
2974 r
= detach_mount_namespace();
2976 return log_error_errno(r
, "Failed to detach mount namespace: %m");
2978 r
= mkdir_p(temp
, 0700);
2980 return log_error_errno(r
, "Failed to create mount point: %m");
2982 created_dir
= TAKE_PTR(temp
);
2984 r
= dissected_image_mount_and_warn(dissected_image
, created_dir
, UID_INVALID
, UID_INVALID
, flags
);
2988 if (decrypted_image
) {
2989 r
= decrypted_image_relinquish(decrypted_image
);
2991 return log_error_errno(r
, "Failed to relinquish DM devices: %m");
2994 loop_device_relinquish(d
);
2996 *ret_directory
= TAKE_PTR(created_dir
);
2997 *ret_loop_device
= TAKE_PTR(d
);
2998 *ret_decrypted_image
= TAKE_PTR(decrypted_image
);
3003 static const char *const partition_designator_table
[] = {
3004 [PARTITION_ROOT
] = "root",
3005 [PARTITION_ROOT_SECONDARY
] = "root-secondary",
3006 [PARTITION_USR
] = "usr",
3007 [PARTITION_USR_SECONDARY
] = "usr-secondary",
3008 [PARTITION_HOME
] = "home",
3009 [PARTITION_SRV
] = "srv",
3010 [PARTITION_ESP
] = "esp",
3011 [PARTITION_XBOOTLDR
] = "xbootldr",
3012 [PARTITION_SWAP
] = "swap",
3013 [PARTITION_ROOT_VERITY
] = "root-verity",
3014 [PARTITION_ROOT_SECONDARY_VERITY
] = "root-secondary-verity",
3015 [PARTITION_USR_VERITY
] = "usr-verity",
3016 [PARTITION_USR_SECONDARY_VERITY
] = "usr-secondary-verity",
3017 [PARTITION_TMP
] = "tmp",
3018 [PARTITION_VAR
] = "var",
3021 int verity_dissect_and_mount(
3024 const MountOptions
*options
,
3025 const char *required_host_os_release_id
,
3026 const char *required_host_os_release_version_id
,
3027 const char *required_host_os_release_sysext_level
) {
3029 _cleanup_(loop_device_unrefp
) LoopDevice
*loop_device
= NULL
;
3030 _cleanup_(decrypted_image_unrefp
) DecryptedImage
*decrypted_image
= NULL
;
3031 _cleanup_(dissected_image_unrefp
) DissectedImage
*dissected_image
= NULL
;
3032 _cleanup_(verity_settings_done
) VeritySettings verity
= VERITY_SETTINGS_DEFAULT
;
3033 DissectImageFlags dissect_image_flags
;
3039 r
= verity_settings_load(&verity
, src
, NULL
, NULL
);
3041 return log_debug_errno(r
, "Failed to load root hash: %m");
3043 dissect_image_flags
= verity
.data_path
? DISSECT_IMAGE_NO_PARTITION_TABLE
: 0;
3045 r
= loop_device_make_by_path(
3048 verity
.data_path
? 0 : LO_FLAGS_PARTSCAN
,
3051 return log_debug_errno(r
, "Failed to create loop device for image: %m");
3057 loop_device
->diskseq
,
3058 loop_device
->uevent_seqnum_not_before
,
3059 loop_device
->timestamp_not_before
,
3060 dissect_image_flags
,
3062 /* No partition table? Might be a single-filesystem image, try again */
3063 if (!verity
.data_path
&& r
== -ENOPKG
)
3068 loop_device
->diskseq
,
3069 loop_device
->uevent_seqnum_not_before
,
3070 loop_device
->timestamp_not_before
,
3071 dissect_image_flags
| DISSECT_IMAGE_NO_PARTITION_TABLE
,
3074 return log_debug_errno(r
, "Failed to dissect image: %m");
3076 r
= dissected_image_decrypt(
3080 dissect_image_flags
,
3083 return log_debug_errno(r
, "Failed to decrypt dissected image: %m");
3085 r
= mkdir_p_label(dest
, 0755);
3087 return log_debug_errno(r
, "Failed to create destination directory %s: %m", dest
);
3088 r
= umount_recursive(dest
, 0);
3090 return log_debug_errno(r
, "Failed to umount under destination directory %s: %m", dest
);
3092 r
= dissected_image_mount(dissected_image
, dest
, UID_INVALID
, UID_INVALID
, dissect_image_flags
);
3094 return log_debug_errno(r
, "Failed to mount image: %m");
3096 /* If we got os-release values from the caller, then we need to match them with the image's
3097 * extension-release.d/ content. Return -EINVAL if there's any mismatch.
3098 * First, check the distro ID. If that matches, then check the new SYSEXT_LEVEL value if
3099 * available, or else fallback to VERSION_ID. */
3100 if (required_host_os_release_id
&&
3101 (required_host_os_release_version_id
|| required_host_os_release_sysext_level
)) {
3102 _cleanup_strv_free_
char **extension_release
= NULL
;
3104 r
= load_extension_release_pairs(dest
, dissected_image
->image_name
, &extension_release
);
3106 return log_debug_errno(r
, "Failed to parse image %s extension-release metadata: %m", dissected_image
->image_name
);
3108 r
= extension_release_validate(
3109 dissected_image
->image_name
,
3110 required_host_os_release_id
,
3111 required_host_os_release_version_id
,
3112 required_host_os_release_sysext_level
,
3115 return log_debug_errno(SYNTHETIC_ERRNO(ESTALE
), "Image %s extension-release metadata does not match the root's", dissected_image
->image_name
);
3117 return log_debug_errno(r
, "Failed to compare image %s extension-release metadata with the root's os-release: %m", dissected_image
->image_name
);
3120 if (decrypted_image
) {
3121 r
= decrypted_image_relinquish(decrypted_image
);
3123 return log_debug_errno(r
, "Failed to relinquish decrypted image: %m");
3126 loop_device_relinquish(loop_device
);
3131 DEFINE_STRING_TABLE_LOOKUP(partition_designator
, PartitionDesignator
);