1 /* SPDX-License-Identifier: LGPL-2.1-or-later */
3 #if HAVE_VALGRIND_MEMCHECK_H
4 #include <valgrind/memcheck.h>
10 #include <linux/loop.h>
12 #include <sys/ioctl.h>
15 #include "sd-device.h"
18 #include "alloc-util.h"
19 #include "blkid-util.h"
20 #include "blockdev-util.h"
21 #include "btrfs-util.h"
24 #include "conf-files.h"
25 #include "conf-parser.h"
26 #include "constants.h"
27 #include "cryptsetup-util.h"
28 #include "device-util.h"
29 #include "devnum-util.h"
30 #include "dirent-util.h"
32 #include "errno-util.h"
34 #include "fdisk-util.h"
36 #include "format-table.h"
37 #include "format-util.h"
39 #include "glyph-util.h"
41 #include "hexdecoct.h"
43 #include "id128-util.h"
44 #include "initrd-util.h"
48 #include "loop-util.h"
49 #include "main-func.h"
51 #include "mkfs-util.h"
52 #include "mount-util.h"
53 #include "mountpoint-util.h"
54 #include "nulstr-util.h"
55 #include "openssl-util.h"
56 #include "parse-argument.h"
57 #include "parse-helpers.h"
58 #include "pretty-print.h"
59 #include "proc-cmdline.h"
60 #include "process-util.h"
61 #include "random-util.h"
62 #include "resize-fs.h"
64 #include "sort-util.h"
65 #include "specifier.h"
66 #include "stdio-util.h"
67 #include "string-table.h"
68 #include "string-util.h"
70 #include "sync-util.h"
71 #include "terminal-util.h"
72 #include "tmpfile-util.h"
74 #include "tpm2-util.h"
75 #include "user-util.h"
78 /* If not configured otherwise use a minimal partition size of 10M */
79 #define DEFAULT_MIN_SIZE (10ULL*1024ULL*1024ULL)
81 /* Hard lower limit for new partition sizes */
82 #define HARD_MIN_SIZE 4096ULL
84 /* We know up front we're never going to put more than this in a verity sig partition. */
85 #define VERITY_SIG_SIZE (HARD_MIN_SIZE*4ULL)
87 /* libfdisk takes off slightly more than 1M of the disk size when creating a GPT disk label */
88 #define GPT_METADATA_SIZE (1044ULL*1024ULL)
90 /* LUKS2 takes off 16M of the partition size with its metadata by default */
91 #define LUKS2_METADATA_SIZE (16ULL*1024ULL*1024ULL)
93 /* To do LUKS2 offline encryption, we need to keep some extra free space at the end of the partition. */
94 #define LUKS2_METADATA_KEEP_FREE (LUKS2_METADATA_SIZE*2ULL)
96 /* LUKS2 volume key size. */
97 #define VOLUME_KEY_SIZE (512ULL/8ULL)
99 /* Use 4K as the default filesystem sector size because as long as the partitions are aligned to 4K, the
100 * filesystems will then also be compatible with sector sizes 512, 1024 and 2048. */
101 #define DEFAULT_FILESYSTEM_SECTOR_SIZE 4096ULL
103 #define APIVFS_TMP_DIRS_NULSTR "proc\0sys\0dev\0tmp\0run\0var/tmp\0"
105 /* Note: When growing and placing new partitions we always align to 4K sector size. It's how newer hard disks
106 * are designed, and if everything is aligned to that performance is best. And for older hard disks with 512B
107 * sector size devices were generally assumed to have an even number of sectors, hence at the worst we'll
108 * waste 3K per partition, which is probably fine. */
110 typedef enum EmptyMode
{
111 EMPTY_UNSET
, /* no choice has been made yet */
112 EMPTY_REFUSE
, /* refuse empty disks, never create a partition table */
113 EMPTY_ALLOW
, /* allow empty disks, create partition table if necessary */
114 EMPTY_REQUIRE
, /* require an empty disk, create a partition table */
115 EMPTY_FORCE
, /* make disk empty, erase everything, create a partition table always */
116 EMPTY_CREATE
, /* create disk as loopback file, create a partition table always */
118 _EMPTY_MODE_INVALID
= -EINVAL
,
121 typedef enum FilterPartitionType
{
122 FILTER_PARTITIONS_NONE
,
123 FILTER_PARTITIONS_EXCLUDE
,
124 FILTER_PARTITIONS_INCLUDE
,
125 _FILTER_PARTITIONS_MAX
,
126 _FILTER_PARTITIONS_INVALID
= -EINVAL
,
127 } FilterPartitionsType
;
129 static EmptyMode arg_empty
= EMPTY_UNSET
;
130 static bool arg_dry_run
= true;
131 static const char *arg_node
= NULL
;
132 static char *arg_root
= NULL
;
133 static char *arg_image
= NULL
;
134 static char **arg_definitions
= NULL
;
135 static bool arg_discard
= true;
136 static bool arg_can_factory_reset
= false;
137 static int arg_factory_reset
= -1;
138 static sd_id128_t arg_seed
= SD_ID128_NULL
;
139 static bool arg_randomize
= false;
140 static int arg_pretty
= -1;
141 static uint64_t arg_size
= UINT64_MAX
;
142 static bool arg_size_auto
= false;
143 static JsonFormatFlags arg_json_format_flags
= JSON_FORMAT_OFF
;
144 static PagerFlags arg_pager_flags
= 0;
145 static bool arg_legend
= true;
146 static void *arg_key
= NULL
;
147 static size_t arg_key_size
= 0;
148 static EVP_PKEY
*arg_private_key
= NULL
;
149 static X509
*arg_certificate
= NULL
;
150 static char *arg_tpm2_device
= NULL
;
151 static Tpm2PCRValue
*arg_tpm2_hash_pcr_values
= NULL
;
152 static size_t arg_tpm2_n_hash_pcr_values
= 0;
153 static char *arg_tpm2_public_key
= NULL
;
154 static uint32_t arg_tpm2_public_key_pcr_mask
= 0;
155 static char *arg_tpm2_pcrlock
= NULL
;
156 static bool arg_split
= false;
157 static GptPartitionType
*arg_filter_partitions
= NULL
;
158 static size_t arg_n_filter_partitions
= 0;
159 static FilterPartitionsType arg_filter_partitions_type
= FILTER_PARTITIONS_NONE
;
160 static GptPartitionType
*arg_defer_partitions
= NULL
;
161 static size_t arg_n_defer_partitions
= 0;
162 static uint64_t arg_sector_size
= 0;
163 static ImagePolicy
*arg_image_policy
= NULL
;
164 static Architecture arg_architecture
= _ARCHITECTURE_INVALID
;
165 static int arg_offline
= -1;
166 static char **arg_copy_from
= NULL
;
167 static char *arg_copy_source
= NULL
;
168 static char *arg_make_ddi
= NULL
;
170 STATIC_DESTRUCTOR_REGISTER(arg_root
, freep
);
171 STATIC_DESTRUCTOR_REGISTER(arg_image
, freep
);
172 STATIC_DESTRUCTOR_REGISTER(arg_definitions
, strv_freep
);
173 STATIC_DESTRUCTOR_REGISTER(arg_key
, erase_and_freep
);
174 STATIC_DESTRUCTOR_REGISTER(arg_private_key
, EVP_PKEY_freep
);
175 STATIC_DESTRUCTOR_REGISTER(arg_certificate
, X509_freep
);
176 STATIC_DESTRUCTOR_REGISTER(arg_tpm2_device
, freep
);
177 STATIC_DESTRUCTOR_REGISTER(arg_tpm2_hash_pcr_values
, freep
);
178 STATIC_DESTRUCTOR_REGISTER(arg_tpm2_public_key
, freep
);
179 STATIC_DESTRUCTOR_REGISTER(arg_tpm2_pcrlock
, freep
);
180 STATIC_DESTRUCTOR_REGISTER(arg_filter_partitions
, freep
);
181 STATIC_DESTRUCTOR_REGISTER(arg_image_policy
, image_policy_freep
);
182 STATIC_DESTRUCTOR_REGISTER(arg_copy_from
, strv_freep
);
183 STATIC_DESTRUCTOR_REGISTER(arg_copy_source
, freep
);
184 STATIC_DESTRUCTOR_REGISTER(arg_make_ddi
, freep
);
186 typedef struct FreeArea FreeArea
;
188 typedef enum EncryptMode
{
192 ENCRYPT_KEY_FILE_TPM2
,
194 _ENCRYPT_MODE_INVALID
= -EINVAL
,
197 typedef enum VerityMode
{
203 _VERITY_MODE_INVALID
= -EINVAL
,
206 typedef enum MinimizeMode
{
211 _MINIMIZE_MODE_INVALID
= -EINVAL
,
214 typedef struct Partition
{
215 char *definition_path
;
216 char **drop_in_files
;
218 GptPartitionType type
;
219 sd_id128_t current_uuid
, new_uuid
;
220 bool new_uuid_is_set
;
221 char *current_label
, *new_label
;
222 sd_id128_t fs_uuid
, luks_uuid
, verity_uuid
;
223 uint8_t verity_salt
[SHA256_DIGEST_SIZE
];
229 uint32_t weight
, padding_weight
;
231 uint64_t current_size
, new_size
;
232 uint64_t size_min
, size_max
;
234 uint64_t current_padding
, new_padding
;
235 uint64_t padding_min
, padding_max
;
240 struct fdisk_partition
*current_partition
;
241 struct fdisk_partition
*new_partition
;
242 FreeArea
*padding_area
;
243 FreeArea
*allocated_to_area
;
245 char *copy_blocks_path
;
246 bool copy_blocks_path_is_our_file
;
247 bool copy_blocks_auto
;
248 const char *copy_blocks_root
;
250 uint64_t copy_blocks_offset
;
251 uint64_t copy_blocks_size
;
255 char **exclude_files_source
;
256 char **exclude_files_target
;
257 char **make_directories
;
261 char *verity_match_key
;
262 MinimizeMode minimize
;
263 uint64_t verity_data_block_size
;
264 uint64_t verity_hash_block_size
;
271 struct iovec roothash
;
273 char *split_name_format
;
276 struct Partition
*siblings
[_VERITY_MODE_MAX
];
278 LIST_FIELDS(struct Partition
, partitions
);
281 #define PARTITION_IS_FOREIGN(p) (!(p)->definition_path)
282 #define PARTITION_EXISTS(p) (!!(p)->current_partition)
290 typedef struct Context
{
291 LIST_HEAD(Partition
, partitions
);
294 FreeArea
**free_areas
;
297 uint64_t start
, end
, total
;
299 struct fdisk_context
*fdisk_context
;
300 uint64_t sector_size
, grain_size
, fs_sector_size
;
305 bool node_is_our_file
;
311 static const char *empty_mode_table
[_EMPTY_MODE_MAX
] = {
312 [EMPTY_UNSET
] = "unset",
313 [EMPTY_REFUSE
] = "refuse",
314 [EMPTY_ALLOW
] = "allow",
315 [EMPTY_REQUIRE
] = "require",
316 [EMPTY_FORCE
] = "force",
317 [EMPTY_CREATE
] = "create",
320 static const char *encrypt_mode_table
[_ENCRYPT_MODE_MAX
] = {
321 [ENCRYPT_OFF
] = "off",
322 [ENCRYPT_KEY_FILE
] = "key-file",
323 [ENCRYPT_TPM2
] = "tpm2",
324 [ENCRYPT_KEY_FILE_TPM2
] = "key-file+tpm2",
327 static const char *verity_mode_table
[_VERITY_MODE_MAX
] = {
328 [VERITY_OFF
] = "off",
329 [VERITY_DATA
] = "data",
330 [VERITY_HASH
] = "hash",
331 [VERITY_SIG
] = "signature",
334 static const char *minimize_mode_table
[_MINIMIZE_MODE_MAX
] = {
335 [MINIMIZE_OFF
] = "off",
336 [MINIMIZE_BEST
] = "best",
337 [MINIMIZE_GUESS
] = "guess",
340 DEFINE_PRIVATE_STRING_TABLE_LOOKUP(empty_mode
, EmptyMode
);
341 DEFINE_PRIVATE_STRING_TABLE_LOOKUP_FROM_STRING_WITH_BOOLEAN(encrypt_mode
, EncryptMode
, ENCRYPT_KEY_FILE
);
342 DEFINE_PRIVATE_STRING_TABLE_LOOKUP(verity_mode
, VerityMode
);
343 DEFINE_PRIVATE_STRING_TABLE_LOOKUP_FROM_STRING_WITH_BOOLEAN(minimize_mode
, MinimizeMode
, MINIMIZE_BEST
);
345 static uint64_t round_down_size(uint64_t v
, uint64_t p
) {
349 static uint64_t round_up_size(uint64_t v
, uint64_t p
) {
351 v
= DIV_ROUND_UP(v
, p
);
353 if (v
> UINT64_MAX
/ p
)
354 return UINT64_MAX
; /* overflow */
359 static Partition
*partition_new(void) {
362 p
= new(Partition
, 1);
369 .current_size
= UINT64_MAX
,
370 .new_size
= UINT64_MAX
,
371 .size_min
= UINT64_MAX
,
372 .size_max
= UINT64_MAX
,
373 .current_padding
= UINT64_MAX
,
374 .new_padding
= UINT64_MAX
,
375 .padding_min
= UINT64_MAX
,
376 .padding_max
= UINT64_MAX
,
377 .partno
= UINT64_MAX
,
378 .offset
= UINT64_MAX
,
379 .copy_blocks_fd
= -EBADF
,
380 .copy_blocks_offset
= UINT64_MAX
,
381 .copy_blocks_size
= UINT64_MAX
,
385 .verity_data_block_size
= UINT64_MAX
,
386 .verity_hash_block_size
= UINT64_MAX
,
392 static Partition
* partition_free(Partition
*p
) {
396 free(p
->current_label
);
398 free(p
->definition_path
);
399 strv_free(p
->drop_in_files
);
401 if (p
->current_partition
)
402 fdisk_unref_partition(p
->current_partition
);
403 if (p
->new_partition
)
404 fdisk_unref_partition(p
->new_partition
);
406 if (p
->copy_blocks_path_is_our_file
)
407 unlink_and_free(p
->copy_blocks_path
);
409 free(p
->copy_blocks_path
);
410 safe_close(p
->copy_blocks_fd
);
413 strv_free(p
->copy_files
);
414 strv_free(p
->exclude_files_source
);
415 strv_free(p
->exclude_files_target
);
416 strv_free(p
->make_directories
);
417 strv_free(p
->subvolumes
);
418 free(p
->verity_match_key
);
420 iovec_done(&p
->roothash
);
422 free(p
->split_name_format
);
423 unlink_and_free(p
->split_path
);
428 static void partition_foreignize(Partition
*p
) {
430 assert(PARTITION_EXISTS(p
));
432 /* Reset several parameters set through definition file to make the partition foreign. */
434 p
->definition_path
= mfree(p
->definition_path
);
435 p
->drop_in_files
= strv_free(p
->drop_in_files
);
437 p
->copy_blocks_path
= mfree(p
->copy_blocks_path
);
438 p
->copy_blocks_fd
= safe_close(p
->copy_blocks_fd
);
439 p
->copy_blocks_root
= NULL
;
441 p
->format
= mfree(p
->format
);
442 p
->copy_files
= strv_free(p
->copy_files
);
443 p
->exclude_files_source
= strv_free(p
->exclude_files_source
);
444 p
->exclude_files_target
= strv_free(p
->exclude_files_target
);
445 p
->make_directories
= strv_free(p
->make_directories
);
446 p
->subvolumes
= strv_free(p
->subvolumes
);
447 p
->verity_match_key
= mfree(p
->verity_match_key
);
451 p
->padding_weight
= 0;
452 p
->size_min
= UINT64_MAX
;
453 p
->size_max
= UINT64_MAX
;
454 p
->padding_min
= UINT64_MAX
;
455 p
->padding_max
= UINT64_MAX
;
459 p
->verity
= VERITY_OFF
;
462 static bool partition_type_exclude(const GptPartitionType
*type
) {
463 if (arg_filter_partitions_type
== FILTER_PARTITIONS_NONE
)
466 for (size_t i
= 0; i
< arg_n_filter_partitions
; i
++)
467 if (sd_id128_equal(type
->uuid
, arg_filter_partitions
[i
].uuid
))
468 return arg_filter_partitions_type
== FILTER_PARTITIONS_EXCLUDE
;
470 return arg_filter_partitions_type
== FILTER_PARTITIONS_INCLUDE
;
473 static bool partition_type_defer(const GptPartitionType
*type
) {
474 for (size_t i
= 0; i
< arg_n_defer_partitions
; i
++)
475 if (sd_id128_equal(type
->uuid
, arg_defer_partitions
[i
].uuid
))
481 static Partition
* partition_unlink_and_free(Context
*context
, Partition
*p
) {
485 LIST_REMOVE(partitions
, context
->partitions
, p
);
487 assert(context
->n_partitions
> 0);
488 context
->n_partitions
--;
490 return partition_free(p
);
493 DEFINE_TRIVIAL_CLEANUP_FUNC(Partition
*, partition_free
);
495 static Context
*context_new(sd_id128_t seed
) {
498 context
= new(Context
, 1);
502 *context
= (Context
) {
512 static void context_free_free_areas(Context
*context
) {
515 for (size_t i
= 0; i
< context
->n_free_areas
; i
++)
516 free(context
->free_areas
[i
]);
518 context
->free_areas
= mfree(context
->free_areas
);
519 context
->n_free_areas
= 0;
522 static Context
*context_free(Context
*context
) {
526 while (context
->partitions
)
527 partition_unlink_and_free(context
, context
->partitions
);
528 assert(context
->n_partitions
== 0);
530 context_free_free_areas(context
);
532 if (context
->fdisk_context
)
533 fdisk_unref_context(context
->fdisk_context
);
535 safe_close(context
->backing_fd
);
536 if (context
->node_is_our_file
)
537 unlink_and_free(context
->node
);
541 return mfree(context
);
544 DEFINE_TRIVIAL_CLEANUP_FUNC(Context
*, context_free
);
546 static int context_add_free_area(
554 assert(!after
|| !after
->padding_area
);
556 if (!GREEDY_REALLOC(context
->free_areas
, context
->n_free_areas
+ 1))
559 a
= new(FreeArea
, 1);
568 context
->free_areas
[context
->n_free_areas
++] = a
;
571 after
->padding_area
= a
;
576 static void partition_drop_or_foreignize(Partition
*p
) {
577 if (!p
|| p
->dropped
|| PARTITION_IS_FOREIGN(p
))
580 if (PARTITION_EXISTS(p
)) {
581 log_info("Can't grow existing partition %s of priority %" PRIi32
", ignoring.",
582 strna(p
->current_label
?: p
->new_label
), p
->priority
);
584 /* Handle the partition as foreign. Do not set dropped flag. */
585 partition_foreignize(p
);
587 log_info("Can't fit partition %s of priority %" PRIi32
", dropping.",
588 p
->definition_path
, p
->priority
);
591 p
->allocated_to_area
= NULL
;
595 static bool context_drop_or_foreignize_one_priority(Context
*context
) {
596 int32_t priority
= 0;
598 LIST_FOREACH(partitions
, p
, context
->partitions
) {
602 priority
= MAX(priority
, p
->priority
);
605 /* Refuse to drop partitions with 0 or negative priorities or partitions of priorities that have at
606 * least one existing priority */
610 LIST_FOREACH(partitions
, p
, context
->partitions
) {
611 if (p
->priority
< priority
)
614 partition_drop_or_foreignize(p
);
616 /* We ensure that all verity sibling partitions have the same priority, so it's safe
617 * to drop all siblings here as well. */
619 for (VerityMode mode
= VERITY_OFF
+ 1; mode
< _VERITY_MODE_MAX
; mode
++)
620 partition_drop_or_foreignize(p
->siblings
[mode
]);
626 static uint64_t partition_min_size(const Context
*context
, const Partition
*p
) {
632 /* Calculate the disk space we really need at minimum for this partition. If the partition already
633 * exists the current size is what we really need. If it doesn't exist yet refuse to allocate less
636 * DEFAULT_MIN_SIZE is the default SizeMin= we configure if nothing else is specified. */
638 if (PARTITION_IS_FOREIGN(p
)) {
639 /* Don't allow changing size of partitions not managed by us */
640 assert(p
->current_size
!= UINT64_MAX
);
641 return p
->current_size
;
644 if (p
->verity
== VERITY_SIG
)
645 return VERITY_SIG_SIZE
;
647 sz
= p
->current_size
!= UINT64_MAX
? p
->current_size
: HARD_MIN_SIZE
;
649 if (!PARTITION_EXISTS(p
)) {
652 if (p
->encrypt
!= ENCRYPT_OFF
)
653 d
+= round_up_size(LUKS2_METADATA_KEEP_FREE
, context
->grain_size
);
655 if (p
->copy_blocks_size
!= UINT64_MAX
)
656 d
+= round_up_size(p
->copy_blocks_size
, context
->grain_size
);
657 else if (p
->format
|| p
->encrypt
!= ENCRYPT_OFF
) {
660 /* If we shall synthesize a file system, take minimal fs size into account (assumed to be 4K if not known) */
661 f
= p
->format
? round_up_size(minimal_size_by_fs_name(p
->format
), context
->grain_size
) : UINT64_MAX
;
662 d
+= f
== UINT64_MAX
? context
->grain_size
: f
;
669 return MAX(round_up_size(p
->size_min
!= UINT64_MAX
? p
->size_min
: DEFAULT_MIN_SIZE
, context
->grain_size
), sz
);
672 static uint64_t partition_max_size(const Context
*context
, const Partition
*p
) {
675 /* Calculate how large the partition may become at max. This is generally the configured maximum
676 * size, except when it already exists and is larger than that. In that case it's the existing size,
677 * since we never want to shrink partitions. */
682 if (PARTITION_IS_FOREIGN(p
)) {
683 /* Don't allow changing size of partitions not managed by us */
684 assert(p
->current_size
!= UINT64_MAX
);
685 return p
->current_size
;
688 if (p
->verity
== VERITY_SIG
)
689 return VERITY_SIG_SIZE
;
691 if (p
->size_max
== UINT64_MAX
)
694 sm
= round_down_size(p
->size_max
, context
->grain_size
);
696 if (p
->current_size
!= UINT64_MAX
)
697 sm
= MAX(p
->current_size
, sm
);
699 return MAX(partition_min_size(context
, p
), sm
);
702 static uint64_t partition_min_padding(const Partition
*p
) {
704 return p
->padding_min
!= UINT64_MAX
? p
->padding_min
: 0;
707 static uint64_t partition_max_padding(const Partition
*p
) {
709 return p
->padding_max
;
712 static uint64_t partition_min_size_with_padding(Context
*context
, const Partition
*p
) {
715 /* Calculate the disk space we need for this partition plus any free space coming after it. This
716 * takes user configured padding into account as well as any additional whitespace needed to align
717 * the next partition to 4K again. */
722 sz
= partition_min_size(context
, p
) + partition_min_padding(p
);
724 if (PARTITION_EXISTS(p
)) {
725 /* If the partition wasn't aligned, add extra space so that any we might add will be aligned */
726 assert(p
->offset
!= UINT64_MAX
);
727 return round_up_size(p
->offset
+ sz
, context
->grain_size
) - p
->offset
;
730 /* If this is a new partition we'll place it aligned, hence we just need to round up the required size here */
731 return round_up_size(sz
, context
->grain_size
);
734 static uint64_t free_area_available(const FreeArea
*a
) {
737 /* Determines how much of this free area is not allocated yet */
739 assert(a
->size
>= a
->allocated
);
740 return a
->size
- a
->allocated
;
743 static uint64_t free_area_current_end(Context
*context
, const FreeArea
*a
) {
748 return free_area_available(a
);
750 assert(a
->after
->offset
!= UINT64_MAX
);
751 assert(a
->after
->current_size
!= UINT64_MAX
);
753 /* Calculate where the free area ends, based on the offset of the partition preceding it. */
754 return round_up_size(a
->after
->offset
+ a
->after
->current_size
, context
->grain_size
) + free_area_available(a
);
757 static uint64_t free_area_min_end(Context
*context
, const FreeArea
*a
) {
764 assert(a
->after
->offset
!= UINT64_MAX
);
765 assert(a
->after
->current_size
!= UINT64_MAX
);
767 /* Calculate where the partition would end when we give it as much as it needs. */
768 return round_up_size(a
->after
->offset
+ partition_min_size_with_padding(context
, a
->after
), context
->grain_size
);
771 static uint64_t free_area_available_for_new_partitions(Context
*context
, const FreeArea
*a
) {
775 /* Similar to free_area_available(), but takes into account that the required size and padding of the
776 * preceding partition is honoured. */
778 return LESS_BY(free_area_current_end(context
, a
), free_area_min_end(context
, a
));
781 static int free_area_compare(FreeArea
*const *a
, FreeArea
*const*b
, Context
*context
) {
784 return CMP(free_area_available_for_new_partitions(context
, *a
),
785 free_area_available_for_new_partitions(context
, *b
));
788 static uint64_t charge_size(Context
*context
, uint64_t total
, uint64_t amount
) {
790 /* Subtract the specified amount from total, rounding up to multiple of 4K if there's room */
791 assert(amount
<= total
);
792 return LESS_BY(total
, round_up_size(amount
, context
->grain_size
));
795 static uint64_t charge_weight(uint64_t total
, uint64_t amount
) {
796 assert(amount
<= total
);
797 return total
- amount
;
800 static bool context_allocate_partitions(Context
*context
, uint64_t *ret_largest_free_area
) {
803 /* This may be called multiple times. Reset previous assignments. */
804 for (size_t i
= 0; i
< context
->n_free_areas
; i
++)
805 context
->free_areas
[i
]->allocated
= 0;
807 /* Sort free areas by size, putting smallest first */
808 typesafe_qsort_r(context
->free_areas
, context
->n_free_areas
, free_area_compare
, context
);
810 /* In any case return size of the largest free area (i.e. not the size of all free areas
812 if (ret_largest_free_area
)
813 *ret_largest_free_area
=
814 context
->n_free_areas
== 0 ? 0 :
815 free_area_available_for_new_partitions(context
, context
->free_areas
[context
->n_free_areas
-1]);
817 /* Check that each existing partition can fit its area. */
818 for (size_t i
= 0; i
< context
->n_free_areas
; i
++)
819 if (free_area_current_end(context
, context
->free_areas
[i
]) <
820 free_area_min_end(context
, context
->free_areas
[i
]))
823 /* A simple first-fit algorithm. We return true if we can fit the partitions in, otherwise false. */
824 LIST_FOREACH(partitions
, p
, context
->partitions
) {
829 /* Skip partitions we already dropped or that already exist */
830 if (p
->dropped
|| PARTITION_EXISTS(p
))
833 /* How much do we need to fit? */
834 required
= partition_min_size_with_padding(context
, p
);
835 assert(required
% context
->grain_size
== 0);
837 for (size_t i
= 0; i
< context
->n_free_areas
; i
++) {
838 a
= context
->free_areas
[i
];
840 if (free_area_available_for_new_partitions(context
, a
) >= required
) {
847 return false; /* 😢 Oh no! We can't fit this partition into any free area! */
849 /* Assign the partition to this free area */
850 p
->allocated_to_area
= a
;
852 /* Budget the minimal partition size */
853 a
->allocated
+= required
;
859 static int context_sum_weights(Context
*context
, FreeArea
*a
, uint64_t *ret
) {
860 uint64_t weight_sum
= 0;
866 /* Determine the sum of the weights of all partitions placed in or before the specified free area */
868 LIST_FOREACH(partitions
, p
, context
->partitions
) {
869 if (p
->padding_area
!= a
&& p
->allocated_to_area
!= a
)
872 if (p
->weight
> UINT64_MAX
- weight_sum
)
874 weight_sum
+= p
->weight
;
876 if (p
->padding_weight
> UINT64_MAX
- weight_sum
)
878 weight_sum
+= p
->padding_weight
;
885 return log_error_errno(SYNTHETIC_ERRNO(EOVERFLOW
), "Combined weight of partition exceeds unsigned 64-bit range, refusing.");
888 static uint64_t scale_by_weight(uint64_t value
, uint64_t weight
, uint64_t weight_sum
) {
889 assert(weight_sum
>= weight
);
894 if (weight
== weight_sum
)
896 if (value
<= UINT64_MAX
/ weight
)
897 return value
* weight
/ weight_sum
;
899 /* Rescale weight and weight_sum to make not the calculation overflow. To satisfy the
900 * following conditions, 'weight_sum' is rounded up but 'weight' is rounded down:
901 * - the sum of scale_by_weight() for all weights must not be larger than the input value,
902 * - scale_by_weight() must not be larger than the ideal value (i.e. calculated with uint128_t). */
903 weight_sum
= DIV_ROUND_UP(weight_sum
, 2);
908 typedef enum GrowPartitionPhase
{
909 /* The zeroth phase: do not touch foreign partitions (i.e. those we don't manage). */
912 /* The first phase: we charge partitions which need more (according to constraints) than their weight-based share. */
915 /* The second phase: we charge partitions which need less (according to constraints) than their weight-based share. */
918 /* The third phase: we distribute what remains among the remaining partitions, according to the weights */
921 _GROW_PARTITION_PHASE_MAX
,
922 } GrowPartitionPhase
;
924 static bool context_grow_partitions_phase(
927 GrowPartitionPhase phase
,
929 uint64_t *weight_sum
) {
931 bool try_again
= false;
938 /* Now let's look at the intended weights and adjust them taking the minimum space assignments into
939 * account. i.e. if a partition has a small weight but a high minimum space value set it should not
940 * get any additional room from the left-overs. Similar, if two partitions have the same weight they
941 * should get the same space if possible, even if one has a smaller minimum size than the other. */
942 LIST_FOREACH(partitions
, p
, context
->partitions
) {
944 /* Look only at partitions associated with this free area, i.e. immediately
945 * preceding it, or allocated into it */
946 if (p
->allocated_to_area
!= a
&& p
->padding_area
!= a
)
949 if (p
->new_size
== UINT64_MAX
) {
950 uint64_t share
, rsz
, xsz
;
953 /* Calculate how much this space this partition needs if everyone would get
954 * the weight based share */
955 share
= scale_by_weight(*span
, p
->weight
, *weight_sum
);
957 rsz
= partition_min_size(context
, p
);
958 xsz
= partition_max_size(context
, p
);
960 if (phase
== PHASE_FOREIGN
&& PARTITION_IS_FOREIGN(p
)) {
961 /* Never change of foreign partitions (i.e. those we don't manage) */
963 p
->new_size
= p
->current_size
;
966 } else if (phase
== PHASE_OVERCHARGE
&& rsz
> share
) {
967 /* This partition needs more than its calculated share. Let's assign
968 * it that, and take this partition out of all calculations and start
972 charge
= try_again
= true;
974 } else if (phase
== PHASE_UNDERCHARGE
&& xsz
< share
) {
975 /* This partition accepts less than its calculated
976 * share. Let's assign it that, and take this partition out
977 * of all calculations and start again. */
980 charge
= try_again
= true;
982 } else if (phase
== PHASE_DISTRIBUTE
) {
983 /* This partition can accept its calculated share. Let's
984 * assign it. There's no need to restart things here since
985 * assigning this shouldn't impact the shares of the other
988 assert(share
>= rsz
);
989 p
->new_size
= CLAMP(round_down_size(share
, context
->grain_size
), rsz
, xsz
);
994 *span
= charge_size(context
, *span
, p
->new_size
);
995 *weight_sum
= charge_weight(*weight_sum
, p
->weight
);
999 if (p
->new_padding
== UINT64_MAX
) {
1000 uint64_t share
, rsz
, xsz
;
1001 bool charge
= false;
1003 share
= scale_by_weight(*span
, p
->padding_weight
, *weight_sum
);
1005 rsz
= partition_min_padding(p
);
1006 xsz
= partition_max_padding(p
);
1008 if (phase
== PHASE_OVERCHARGE
&& rsz
> share
) {
1009 p
->new_padding
= rsz
;
1010 charge
= try_again
= true;
1011 } else if (phase
== PHASE_UNDERCHARGE
&& xsz
< share
) {
1012 p
->new_padding
= xsz
;
1013 charge
= try_again
= true;
1014 } else if (phase
== PHASE_DISTRIBUTE
) {
1015 assert(share
>= rsz
);
1016 p
->new_padding
= CLAMP(round_down_size(share
, context
->grain_size
), rsz
, xsz
);
1021 *span
= charge_size(context
, *span
, p
->new_padding
);
1022 *weight_sum
= charge_weight(*weight_sum
, p
->padding_weight
);
1030 static void context_grow_partition_one(Context
*context
, FreeArea
*a
, Partition
*p
, uint64_t *span
) {
1041 if (p
->allocated_to_area
!= a
)
1044 if (PARTITION_IS_FOREIGN(p
))
1047 assert(p
->new_size
!= UINT64_MAX
);
1049 /* Calculate new size and align. */
1050 m
= round_down_size(p
->new_size
+ *span
, context
->grain_size
);
1051 /* But ensure this doesn't shrink the size. */
1052 m
= MAX(m
, p
->new_size
);
1053 /* And ensure this doesn't exceed the maximum size. */
1054 m
= MIN(m
, partition_max_size(context
, p
));
1056 assert(m
>= p
->new_size
);
1058 *span
= charge_size(context
, *span
, m
- p
->new_size
);
1062 static int context_grow_partitions_on_free_area(Context
*context
, FreeArea
*a
) {
1063 uint64_t weight_sum
= 0, span
;
1069 r
= context_sum_weights(context
, a
, &weight_sum
);
1073 /* Let's calculate the total area covered by this free area and the partition before it */
1076 assert(a
->after
->offset
!= UINT64_MAX
);
1077 assert(a
->after
->current_size
!= UINT64_MAX
);
1079 span
+= round_up_size(a
->after
->offset
+ a
->after
->current_size
, context
->grain_size
) - a
->after
->offset
;
1082 for (GrowPartitionPhase phase
= 0; phase
< _GROW_PARTITION_PHASE_MAX
;)
1083 if (context_grow_partitions_phase(context
, a
, phase
, &span
, &weight_sum
))
1084 phase
++; /* go to the next phase */
1086 /* We still have space left over? Donate to preceding partition if we have one */
1087 if (span
> 0 && a
->after
)
1088 context_grow_partition_one(context
, a
, a
->after
, &span
);
1090 /* What? Even still some space left (maybe because there was no preceding partition, or it had a
1091 * size limit), then let's donate it to whoever wants it. */
1093 LIST_FOREACH(partitions
, p
, context
->partitions
) {
1094 context_grow_partition_one(context
, a
, p
, &span
);
1099 /* Yuck, still no one? Then make it padding */
1100 if (span
> 0 && a
->after
) {
1101 assert(a
->after
->new_padding
!= UINT64_MAX
);
1102 a
->after
->new_padding
+= span
;
1108 static int context_grow_partitions(Context
*context
) {
1113 for (size_t i
= 0; i
< context
->n_free_areas
; i
++) {
1114 r
= context_grow_partitions_on_free_area(context
, context
->free_areas
[i
]);
1119 /* All existing partitions that have no free space after them can't change size */
1120 LIST_FOREACH(partitions
, p
, context
->partitions
) {
1124 if (!PARTITION_EXISTS(p
) || p
->padding_area
) {
1125 /* The algorithm above must have initialized this already */
1126 assert(p
->new_size
!= UINT64_MAX
);
1130 assert(p
->new_size
== UINT64_MAX
);
1131 p
->new_size
= p
->current_size
;
1133 assert(p
->new_padding
== UINT64_MAX
);
1134 p
->new_padding
= p
->current_padding
;
1140 static uint64_t find_first_unused_partno(Context
*context
) {
1141 uint64_t partno
= 0;
1145 for (partno
= 0;; partno
++) {
1147 LIST_FOREACH(partitions
, p
, context
->partitions
)
1148 if (p
->partno
!= UINT64_MAX
&& p
->partno
== partno
)
1157 static void context_place_partitions(Context
*context
) {
1161 for (size_t i
= 0; i
< context
->n_free_areas
; i
++) {
1162 FreeArea
*a
= context
->free_areas
[i
];
1163 _unused_
uint64_t left
;
1167 assert(a
->after
->offset
!= UINT64_MAX
);
1168 assert(a
->after
->new_size
!= UINT64_MAX
);
1169 assert(a
->after
->new_padding
!= UINT64_MAX
);
1171 start
= a
->after
->offset
+ a
->after
->new_size
+ a
->after
->new_padding
;
1173 start
= context
->start
;
1175 start
= round_up_size(start
, context
->grain_size
);
1178 LIST_FOREACH(partitions
, p
, context
->partitions
) {
1179 if (p
->allocated_to_area
!= a
)
1183 p
->partno
= find_first_unused_partno(context
);
1185 assert(left
>= p
->new_size
);
1186 start
+= p
->new_size
;
1187 left
-= p
->new_size
;
1189 assert(left
>= p
->new_padding
);
1190 start
+= p
->new_padding
;
1191 left
-= p
->new_padding
;
1196 static int config_parse_type(
1198 const char *filename
,
1200 const char *section
,
1201 unsigned section_line
,
1208 GptPartitionType
*type
= ASSERT_PTR(data
);
1213 r
= gpt_partition_type_from_string(rvalue
, type
);
1215 return log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse partition type: %s", rvalue
);
1217 if (arg_architecture
>= 0)
1218 *type
= gpt_partition_type_override_architecture(*type
, arg_architecture
);
1223 static int config_parse_label(
1225 const char *filename
,
1227 const char *section
,
1228 unsigned section_line
,
1235 _cleanup_free_
char *resolved
= NULL
;
1236 char **label
= ASSERT_PTR(data
);
1241 /* Nota bene: the empty label is a totally valid one. Let's hence not follow our usual rule of
1242 * assigning the empty string to reset to default here, but really accept it as label to set. */
1244 r
= specifier_printf(rvalue
, GPT_LABEL_MAX
, system_and_tmp_specifier_table
, arg_root
, NULL
, &resolved
);
1246 log_syntax(unit
, LOG_WARNING
, filename
, line
, r
,
1247 "Failed to expand specifiers in Label=, ignoring: %s", rvalue
);
1251 if (!utf8_is_valid(resolved
)) {
1252 log_syntax(unit
, LOG_WARNING
, filename
, line
, 0,
1253 "Partition label not valid UTF-8, ignoring: %s", rvalue
);
1257 r
= gpt_partition_label_valid(resolved
);
1259 log_syntax(unit
, LOG_WARNING
, filename
, line
, r
,
1260 "Failed to check if string is valid as GPT partition label, ignoring: \"%s\" (from \"%s\")",
1265 log_syntax(unit
, LOG_WARNING
, filename
, line
, 0,
1266 "Partition label too long for GPT table, ignoring: \"%s\" (from \"%s\")",
1271 free_and_replace(*label
, resolved
);
1275 static int config_parse_weight(
1277 const char *filename
,
1279 const char *section
,
1280 unsigned section_line
,
1287 uint32_t *w
= ASSERT_PTR(data
), v
;
1292 r
= safe_atou32(rvalue
, &v
);
1294 log_syntax(unit
, LOG_WARNING
, filename
, line
, r
,
1295 "Failed to parse weight value, ignoring: %s", rvalue
);
1299 if (v
> 1000U*1000U) {
1300 log_syntax(unit
, LOG_WARNING
, filename
, line
, 0,
1301 "Weight needs to be in range 0…10000000, ignoring: %" PRIu32
, v
);
1309 static int config_parse_size4096(
1311 const char *filename
,
1313 const char *section
,
1314 unsigned section_line
,
1321 uint64_t *sz
= data
, parsed
;
1327 r
= parse_size(rvalue
, 1024, &parsed
);
1329 return log_syntax(unit
, LOG_ERR
, filename
, line
, r
,
1330 "Failed to parse size value: %s", rvalue
);
1333 *sz
= round_up_size(parsed
, 4096);
1335 *sz
= round_down_size(parsed
, 4096);
1340 log_syntax(unit
, LOG_NOTICE
, filename
, line
, r
, "Rounded %s= size %" PRIu64
" %s %" PRIu64
", a multiple of 4096.",
1341 lvalue
, parsed
, special_glyph(SPECIAL_GLYPH_ARROW_RIGHT
), *sz
);
1346 static int config_parse_block_size(
1348 const char *filename
,
1350 const char *section
,
1351 unsigned section_line
,
1358 uint64_t *blksz
= ASSERT_PTR(data
), parsed
;
1363 r
= parse_size(rvalue
, 1024, &parsed
);
1365 return log_syntax(unit
, LOG_ERR
, filename
, line
, r
,
1366 "Failed to parse size value: %s", rvalue
);
1368 if (parsed
< 512 || parsed
> 4096)
1369 return log_syntax(unit
, LOG_ERR
, filename
, line
, SYNTHETIC_ERRNO(EINVAL
),
1370 "Value not between 512 and 4096: %s", rvalue
);
1372 if (!ISPOWEROF2(parsed
))
1373 return log_syntax(unit
, LOG_ERR
, filename
, line
, SYNTHETIC_ERRNO(EINVAL
),
1374 "Value not a power of 2: %s", rvalue
);
1380 static int config_parse_fstype(
1382 const char *filename
,
1384 const char *section
,
1385 unsigned section_line
,
1392 char **fstype
= ASSERT_PTR(data
);
1397 /* Let's provide an easy way to override the chosen fstype for file system partitions */
1398 e
= secure_getenv("SYSTEMD_REPART_OVERRIDE_FSTYPE");
1399 if (e
&& !streq(rvalue
, e
)) {
1400 log_syntax(unit
, LOG_NOTICE
, filename
, line
, 0,
1401 "Overriding defined file system type '%s' with '%s'.", rvalue
, e
);
1405 if (!filename_is_valid(rvalue
))
1406 return log_syntax(unit
, LOG_ERR
, filename
, line
, 0,
1407 "File system type is not valid, refusing: %s", rvalue
);
1409 return free_and_strdup_warn(fstype
, rvalue
);
1412 static int config_parse_copy_files(
1414 const char *filename
,
1416 const char *section
,
1417 unsigned section_line
,
1424 _cleanup_free_
char *source
= NULL
, *buffer
= NULL
, *resolved_source
= NULL
, *resolved_target
= NULL
;
1425 const char *p
= rvalue
, *target
;
1426 char ***copy_files
= ASSERT_PTR(data
);
1431 r
= extract_first_word(&p
, &source
, ":", EXTRACT_CUNESCAPE
|EXTRACT_DONT_COALESCE_SEPARATORS
);
1433 return log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to extract source path: %s", rvalue
);
1435 log_syntax(unit
, LOG_WARNING
, filename
, line
, 0, "No argument specified: %s", rvalue
);
1439 r
= extract_first_word(&p
, &buffer
, ":", EXTRACT_CUNESCAPE
|EXTRACT_DONT_COALESCE_SEPARATORS
);
1441 return log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to extract target path: %s", rvalue
);
1443 target
= source
; /* No target, then it's the same as the source */
1448 return log_syntax(unit
, LOG_ERR
, filename
, line
, SYNTHETIC_ERRNO(EINVAL
), "Too many arguments: %s", rvalue
);
1450 r
= specifier_printf(source
, PATH_MAX
-1, system_and_tmp_specifier_table
, arg_root
, NULL
, &resolved_source
);
1452 log_syntax(unit
, LOG_WARNING
, filename
, line
, r
,
1453 "Failed to expand specifiers in CopyFiles= source, ignoring: %s", rvalue
);
1457 r
= path_simplify_and_warn(resolved_source
, PATH_CHECK_ABSOLUTE
, unit
, filename
, line
, lvalue
);
1461 r
= specifier_printf(target
, PATH_MAX
-1, system_and_tmp_specifier_table
, arg_root
, NULL
, &resolved_target
);
1463 log_syntax(unit
, LOG_WARNING
, filename
, line
, r
,
1464 "Failed to expand specifiers in CopyFiles= target, ignoring: %s", resolved_target
);
1468 r
= path_simplify_and_warn(resolved_target
, PATH_CHECK_ABSOLUTE
, unit
, filename
, line
, lvalue
);
1472 r
= strv_consume_pair(copy_files
, TAKE_PTR(resolved_source
), TAKE_PTR(resolved_target
));
1479 static int config_parse_exclude_files(
1481 const char *filename
,
1483 const char *section
,
1484 unsigned section_line
,
1490 _cleanup_free_
char *resolved
= NULL
;
1491 char ***exclude_files
= ASSERT_PTR(data
);
1494 if (isempty(rvalue
)) {
1495 *exclude_files
= strv_free(*exclude_files
);
1499 r
= specifier_printf(rvalue
, PATH_MAX
-1, system_and_tmp_specifier_table
, arg_root
, NULL
, &resolved
);
1501 log_syntax(unit
, LOG_WARNING
, filename
, line
, r
,
1502 "Failed to expand specifiers in ExcludeFiles= path, ignoring: %s", rvalue
);
1506 r
= path_simplify_and_warn(resolved
, PATH_CHECK_ABSOLUTE
|PATH_KEEP_TRAILING_SLASH
, unit
, filename
, line
, lvalue
);
1510 if (strv_consume(exclude_files
, TAKE_PTR(resolved
)) < 0)
1516 static int config_parse_copy_blocks(
1518 const char *filename
,
1520 const char *section
,
1521 unsigned section_line
,
1528 _cleanup_free_
char *d
= NULL
;
1529 Partition
*partition
= ASSERT_PTR(data
);
1534 if (isempty(rvalue
)) {
1535 partition
->copy_blocks_path
= mfree(partition
->copy_blocks_path
);
1536 partition
->copy_blocks_auto
= false;
1540 if (streq(rvalue
, "auto")) {
1541 partition
->copy_blocks_path
= mfree(partition
->copy_blocks_path
);
1542 partition
->copy_blocks_auto
= true;
1543 partition
->copy_blocks_root
= arg_root
;
1547 r
= specifier_printf(rvalue
, PATH_MAX
-1, system_and_tmp_specifier_table
, arg_root
, NULL
, &d
);
1549 log_syntax(unit
, LOG_WARNING
, filename
, line
, r
,
1550 "Failed to expand specifiers in CopyBlocks= source path, ignoring: %s", rvalue
);
1554 r
= path_simplify_and_warn(d
, PATH_CHECK_ABSOLUTE
, unit
, filename
, line
, lvalue
);
1558 free_and_replace(partition
->copy_blocks_path
, d
);
1559 partition
->copy_blocks_auto
= false;
1560 partition
->copy_blocks_root
= arg_root
;
1564 static int config_parse_make_dirs(
1566 const char *filename
,
1568 const char *section
,
1569 unsigned section_line
,
1576 char ***sv
= ASSERT_PTR(data
);
1577 const char *p
= ASSERT_PTR(rvalue
);
1581 _cleanup_free_
char *word
= NULL
, *d
= NULL
;
1583 r
= extract_first_word(&p
, &word
, NULL
, EXTRACT_UNQUOTE
);
1587 log_syntax(unit
, LOG_WARNING
, filename
, line
, r
, "Invalid syntax, ignoring: %s", rvalue
);
1593 r
= specifier_printf(word
, PATH_MAX
-1, system_and_tmp_specifier_table
, arg_root
, NULL
, &d
);
1595 log_syntax(unit
, LOG_WARNING
, filename
, line
, r
,
1596 "Failed to expand specifiers in MakeDirectories= parameter, ignoring: %s", word
);
1600 r
= path_simplify_and_warn(d
, PATH_CHECK_ABSOLUTE
, unit
, filename
, line
, lvalue
);
1604 r
= strv_consume(sv
, TAKE_PTR(d
));
1610 static DEFINE_CONFIG_PARSE_ENUM_WITH_DEFAULT(config_parse_encrypt
, encrypt_mode
, EncryptMode
, ENCRYPT_OFF
, "Invalid encryption mode");
1612 static int config_parse_gpt_flags(
1614 const char *filename
,
1616 const char *section
,
1617 unsigned section_line
,
1624 uint64_t *gpt_flags
= ASSERT_PTR(data
);
1629 r
= safe_atou64(rvalue
, gpt_flags
);
1631 log_syntax(unit
, LOG_WARNING
, filename
, line
, r
,
1632 "Failed to parse Flags= value, ignoring: %s", rvalue
);
1639 static int config_parse_uuid(
1641 const char *filename
,
1643 const char *section
,
1644 unsigned section_line
,
1651 Partition
*partition
= ASSERT_PTR(data
);
1654 if (isempty(rvalue
)) {
1655 partition
->new_uuid
= SD_ID128_NULL
;
1656 partition
->new_uuid_is_set
= false;
1660 if (streq(rvalue
, "null")) {
1661 partition
->new_uuid
= SD_ID128_NULL
;
1662 partition
->new_uuid_is_set
= true;
1666 r
= sd_id128_from_string(rvalue
, &partition
->new_uuid
);
1668 log_syntax(unit
, LOG_WARNING
, filename
, line
, r
, "Failed to parse 128-bit ID/UUID, ignoring: %s", rvalue
);
1672 partition
->new_uuid_is_set
= true;
1677 static DEFINE_CONFIG_PARSE_ENUM_WITH_DEFAULT(config_parse_verity
, verity_mode
, VerityMode
, VERITY_OFF
, "Invalid verity mode");
1678 static DEFINE_CONFIG_PARSE_ENUM_WITH_DEFAULT(config_parse_minimize
, minimize_mode
, MinimizeMode
, MINIMIZE_OFF
, "Invalid minimize mode");
1680 static int partition_read_definition(Partition
*p
, const char *path
, const char *const *conf_file_dirs
) {
1682 ConfigTableItem table
[] = {
1683 { "Partition", "Type", config_parse_type
, 0, &p
->type
},
1684 { "Partition", "Label", config_parse_label
, 0, &p
->new_label
},
1685 { "Partition", "UUID", config_parse_uuid
, 0, p
},
1686 { "Partition", "Priority", config_parse_int32
, 0, &p
->priority
},
1687 { "Partition", "Weight", config_parse_weight
, 0, &p
->weight
},
1688 { "Partition", "PaddingWeight", config_parse_weight
, 0, &p
->padding_weight
},
1689 { "Partition", "SizeMinBytes", config_parse_size4096
, 1, &p
->size_min
},
1690 { "Partition", "SizeMaxBytes", config_parse_size4096
, -1, &p
->size_max
},
1691 { "Partition", "PaddingMinBytes", config_parse_size4096
, 1, &p
->padding_min
},
1692 { "Partition", "PaddingMaxBytes", config_parse_size4096
, -1, &p
->padding_max
},
1693 { "Partition", "FactoryReset", config_parse_bool
, 0, &p
->factory_reset
},
1694 { "Partition", "CopyBlocks", config_parse_copy_blocks
, 0, p
},
1695 { "Partition", "Format", config_parse_fstype
, 0, &p
->format
},
1696 { "Partition", "CopyFiles", config_parse_copy_files
, 0, &p
->copy_files
},
1697 { "Partition", "ExcludeFiles", config_parse_exclude_files
, 0, &p
->exclude_files_source
},
1698 { "Partition", "ExcludeFilesTarget", config_parse_exclude_files
, 0, &p
->exclude_files_target
},
1699 { "Partition", "MakeDirectories", config_parse_make_dirs
, 0, &p
->make_directories
},
1700 { "Partition", "Encrypt", config_parse_encrypt
, 0, &p
->encrypt
},
1701 { "Partition", "Verity", config_parse_verity
, 0, &p
->verity
},
1702 { "Partition", "VerityMatchKey", config_parse_string
, 0, &p
->verity_match_key
},
1703 { "Partition", "Flags", config_parse_gpt_flags
, 0, &p
->gpt_flags
},
1704 { "Partition", "ReadOnly", config_parse_tristate
, 0, &p
->read_only
},
1705 { "Partition", "NoAuto", config_parse_tristate
, 0, &p
->no_auto
},
1706 { "Partition", "GrowFileSystem", config_parse_tristate
, 0, &p
->growfs
},
1707 { "Partition", "SplitName", config_parse_string
, 0, &p
->split_name_format
},
1708 { "Partition", "Minimize", config_parse_minimize
, 0, &p
->minimize
},
1709 { "Partition", "Subvolumes", config_parse_make_dirs
, 0, &p
->subvolumes
},
1710 { "Partition", "VerityDataBlockSizeBytes", config_parse_block_size
, 0, &p
->verity_data_block_size
},
1711 { "Partition", "VerityHashBlockSizeBytes", config_parse_block_size
, 0, &p
->verity_hash_block_size
},
1715 _cleanup_free_
char *filename
= NULL
;
1716 const char* dropin_dirname
;
1718 r
= path_extract_filename(path
, &filename
);
1720 return log_error_errno(r
, "Failed to extract filename from path '%s': %m", path
);
1722 dropin_dirname
= strjoina(filename
, ".d");
1724 r
= config_parse_many(
1725 STRV_MAKE_CONST(path
),
1728 arg_definitions
? NULL
: arg_root
,
1730 config_item_table_lookup
, table
,
1738 if (partition_type_exclude(&p
->type
))
1741 if (p
->size_min
!= UINT64_MAX
&& p
->size_max
!= UINT64_MAX
&& p
->size_min
> p
->size_max
)
1742 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1743 "SizeMinBytes= larger than SizeMaxBytes=, refusing.");
1745 if (p
->padding_min
!= UINT64_MAX
&& p
->padding_max
!= UINT64_MAX
&& p
->padding_min
> p
->padding_max
)
1746 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1747 "PaddingMinBytes= larger than PaddingMaxBytes=, refusing.");
1749 if (sd_id128_is_null(p
->type
.uuid
))
1750 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1751 "Type= not defined, refusing.");
1753 if ((p
->copy_blocks_path
|| p
->copy_blocks_auto
) &&
1754 (p
->format
|| !strv_isempty(p
->copy_files
) || !strv_isempty(p
->make_directories
)))
1755 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1756 "Format=/CopyFiles=/MakeDirectories= and CopyBlocks= cannot be combined, refusing.");
1758 if ((!strv_isempty(p
->copy_files
) || !strv_isempty(p
->make_directories
)) && streq_ptr(p
->format
, "swap"))
1759 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1760 "Format=swap and CopyFiles= cannot be combined, refusing.");
1763 const char *format
= NULL
;
1765 if (!strv_isempty(p
->copy_files
) || !strv_isempty(p
->make_directories
) || (p
->encrypt
!= ENCRYPT_OFF
&& !(p
->copy_blocks_path
|| p
->copy_blocks_auto
)))
1766 /* Pick "vfat" as file system for esp and xbootldr partitions, otherwise default to "ext4". */
1767 format
= IN_SET(p
->type
.designator
, PARTITION_ESP
, PARTITION_XBOOTLDR
) ? "vfat" : "ext4";
1768 else if (p
->type
.designator
== PARTITION_SWAP
)
1772 p
->format
= strdup(format
);
1778 if (p
->minimize
!= MINIMIZE_OFF
&& !p
->format
&& p
->verity
!= VERITY_HASH
)
1779 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1780 "Minimize= can only be enabled if Format= or Verity=hash are set");
1782 if (p
->minimize
== MINIMIZE_BEST
&& (p
->format
&& !fstype_is_ro(p
->format
)) && p
->verity
!= VERITY_HASH
)
1783 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1784 "Minimize=best can only be used with read-only filesystems or Verity=hash");
1786 if ((!strv_isempty(p
->copy_files
) || !strv_isempty(p
->make_directories
)) && !mkfs_supports_root_option(p
->format
) && geteuid() != 0)
1787 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EPERM
),
1788 "Need to be root to populate %s filesystems with CopyFiles=/MakeDirectories=",
1791 if (p
->format
&& fstype_is_ro(p
->format
) && strv_isempty(p
->copy_files
) && strv_isempty(p
->make_directories
))
1792 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1793 "Cannot format %s filesystem without source files, refusing", p
->format
);
1795 if (p
->verity
!= VERITY_OFF
|| p
->encrypt
!= ENCRYPT_OFF
) {
1796 r
= dlopen_cryptsetup();
1798 return log_syntax(NULL
, LOG_ERR
, path
, 1, r
,
1799 "libcryptsetup not found, Verity=/Encrypt= are not supported: %m");
1802 if (p
->verity
!= VERITY_OFF
&& !p
->verity_match_key
)
1803 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1804 "VerityMatchKey= must be set if Verity=%s", verity_mode_to_string(p
->verity
));
1806 if (p
->verity
== VERITY_OFF
&& p
->verity_match_key
)
1807 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1808 "VerityMatchKey= can only be set if Verity= is not \"%s\"",
1809 verity_mode_to_string(p
->verity
));
1811 if (IN_SET(p
->verity
, VERITY_HASH
, VERITY_SIG
) &&
1812 (p
->copy_files
|| p
->copy_blocks_path
|| p
->copy_blocks_auto
|| p
->format
|| p
->make_directories
))
1813 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1814 "CopyBlocks=/CopyFiles=/Format=/MakeDirectories= cannot be used with Verity=%s",
1815 verity_mode_to_string(p
->verity
));
1817 if (p
->verity
!= VERITY_OFF
&& p
->encrypt
!= ENCRYPT_OFF
)
1818 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1819 "Encrypting verity hash/data partitions is not supported");
1821 if (p
->verity
== VERITY_SIG
&& !arg_private_key
)
1822 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1823 "Verity signature partition requested but no private key provided (--private-key=)");
1825 if (p
->verity
== VERITY_SIG
&& !arg_certificate
)
1826 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1827 "Verity signature partition requested but no PEM certificate provided (--certificate=)");
1829 if (p
->verity
== VERITY_SIG
&& (p
->size_min
!= UINT64_MAX
|| p
->size_max
!= UINT64_MAX
))
1830 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EINVAL
),
1831 "SizeMinBytes=/SizeMaxBytes= cannot be used with Verity=%s",
1832 verity_mode_to_string(p
->verity
));
1834 if (!strv_isempty(p
->subvolumes
) && arg_offline
> 0)
1835 return log_syntax(NULL
, LOG_ERR
, path
, 1, SYNTHETIC_ERRNO(EOPNOTSUPP
),
1836 "Subvolumes= cannot be used with --offline=yes");
1838 /* Verity partitions are read only, let's imply the RO flag hence, unless explicitly configured otherwise. */
1839 if ((IN_SET(p
->type
.designator
,
1840 PARTITION_ROOT_VERITY
,
1841 PARTITION_USR_VERITY
) || p
->verity
== VERITY_DATA
) && p
->read_only
< 0)
1842 p
->read_only
= true;
1844 /* Default to "growfs" on, unless read-only */
1845 if (gpt_partition_type_knows_growfs(p
->type
) &&
1849 if (!p
->split_name_format
) {
1850 char *s
= strdup("%t");
1854 p
->split_name_format
= s
;
1855 } else if (streq(p
->split_name_format
, "-"))
1856 p
->split_name_format
= mfree(p
->split_name_format
);
1861 static int find_verity_sibling(Context
*context
, Partition
*p
, VerityMode mode
, Partition
**ret
) {
1862 Partition
*s
= NULL
;
1865 assert(p
->verity
!= VERITY_OFF
);
1866 assert(p
->verity_match_key
);
1867 assert(mode
!= VERITY_OFF
);
1868 assert(p
->verity
!= mode
);
1871 /* Try to find the matching sibling partition of the given type for a verity partition. For a data
1872 * partition, this is the corresponding hash partition with the same verity name (and vice versa for
1873 * the hash partition). */
1875 LIST_FOREACH(partitions
, q
, context
->partitions
) {
1879 if (q
->verity
!= mode
)
1882 assert(q
->verity_match_key
);
1884 if (!streq(p
->verity_match_key
, q
->verity_match_key
))
1901 static int context_open_and_lock_backing_fd(const char *node
, int operation
, int *backing_fd
) {
1902 _cleanup_close_
int fd
= -EBADF
;
1907 if (*backing_fd
>= 0)
1910 fd
= open(node
, O_RDONLY
|O_CLOEXEC
);
1912 return log_error_errno(errno
, "Failed to open device '%s': %m", node
);
1914 /* Tell udev not to interfere while we are processing the device */
1915 if (flock(fd
, operation
) < 0)
1916 return log_error_errno(errno
, "Failed to lock device '%s': %m", node
);
1918 log_debug("Device %s opened and locked.", node
);
1919 *backing_fd
= TAKE_FD(fd
);
1923 static int determine_current_padding(
1924 struct fdisk_context
*c
,
1925 struct fdisk_table
*t
,
1926 struct fdisk_partition
*p
,
1931 size_t n_partitions
;
1932 uint64_t offset
, next
= UINT64_MAX
;
1939 if (!fdisk_partition_has_end(p
))
1940 return log_error_errno(SYNTHETIC_ERRNO(EIO
), "Partition has no end!");
1942 offset
= fdisk_partition_get_end(p
);
1943 assert(offset
< UINT64_MAX
);
1944 offset
++; /* The end is one sector before the next partition or padding. */
1945 assert(offset
< UINT64_MAX
/ secsz
);
1948 n_partitions
= fdisk_table_get_nents(t
);
1949 for (size_t i
= 0; i
< n_partitions
; i
++) {
1950 struct fdisk_partition
*q
;
1953 q
= fdisk_table_get_partition(t
, i
);
1955 return log_error_errno(SYNTHETIC_ERRNO(EIO
), "Failed to read partition metadata: %m");
1957 if (fdisk_partition_is_used(q
) <= 0)
1960 if (!fdisk_partition_has_start(q
))
1963 start
= fdisk_partition_get_start(q
);
1964 assert(start
< UINT64_MAX
/ secsz
);
1967 if (start
>= offset
&& (next
== UINT64_MAX
|| next
> start
))
1971 if (next
== UINT64_MAX
) {
1972 /* No later partition? In that case check the end of the usable area */
1973 next
= fdisk_get_last_lba(c
);
1974 assert(next
< UINT64_MAX
);
1975 next
++; /* The last LBA is one sector before the end */
1977 assert(next
< UINT64_MAX
/ secsz
);
1981 return log_error_errno(SYNTHETIC_ERRNO(EIO
), "Partition end beyond disk end.");
1984 assert(next
>= offset
);
1985 offset
= round_up_size(offset
, grainsz
);
1986 next
= round_down_size(next
, grainsz
);
1988 *ret
= LESS_BY(next
, offset
); /* Saturated subtraction, rounding might have fucked things up */
1992 static int context_copy_from_one(Context
*context
, const char *src
) {
1993 _cleanup_close_
int fd
= -EBADF
;
1994 _cleanup_(fdisk_unref_contextp
) struct fdisk_context
*c
= NULL
;
1995 _cleanup_(fdisk_unref_tablep
) struct fdisk_table
*t
= NULL
;
1996 Partition
*last
= NULL
;
1997 unsigned long secsz
, grainsz
;
1998 size_t n_partitions
;
2003 r
= context_open_and_lock_backing_fd(src
, LOCK_SH
, &fd
);
2007 r
= fd_verify_regular(fd
);
2009 return log_error_errno(r
, "%s is not a file: %m", src
);
2011 r
= fdisk_new_context_at(fd
, /* path = */ NULL
, /* read_only = */ true, /* sector_size = */ UINT32_MAX
, &c
);
2013 return log_error_errno(r
, "Failed to create fdisk context: %m");
2015 secsz
= fdisk_get_sector_size(c
);
2016 grainsz
= fdisk_get_grain_size(c
);
2018 /* Insist on a power of two, and that it's a multiple of 512, i.e. the traditional sector size. */
2019 if (secsz
< 512 || !ISPOWEROF2(secsz
))
2020 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "Sector size %lu is not a power of two larger than 512? Refusing.", secsz
);
2022 if (!fdisk_is_labeltype(c
, FDISK_DISKLABEL_GPT
))
2023 return log_error_errno(SYNTHETIC_ERRNO(EHWPOISON
), "Cannot copy from disk %s with no GPT disk label.", src
);
2025 r
= fdisk_get_partitions(c
, &t
);
2027 return log_error_errno(r
, "Failed to acquire partition table: %m");
2029 n_partitions
= fdisk_table_get_nents(t
);
2030 for (size_t i
= 0; i
< n_partitions
; i
++) {
2031 _cleanup_(partition_freep
) Partition
*np
= NULL
;
2032 _cleanup_free_
char *label_copy
= NULL
;
2033 struct fdisk_partition
*p
;
2035 uint64_t sz
, start
, padding
;
2036 sd_id128_t ptid
, id
;
2037 GptPartitionType type
;
2039 p
= fdisk_table_get_partition(t
, i
);
2041 return log_error_errno(SYNTHETIC_ERRNO(EIO
), "Failed to read partition metadata: %m");
2043 if (fdisk_partition_is_used(p
) <= 0)
2046 if (fdisk_partition_has_start(p
) <= 0 ||
2047 fdisk_partition_has_size(p
) <= 0 ||
2048 fdisk_partition_has_partno(p
) <= 0)
2049 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "Found a partition without a position, size or number.");
2051 r
= fdisk_partition_get_type_as_id128(p
, &ptid
);
2053 return log_error_errno(r
, "Failed to query partition type UUID: %m");
2055 type
= gpt_partition_type_from_uuid(ptid
);
2057 r
= fdisk_partition_get_uuid_as_id128(p
, &id
);
2059 return log_error_errno(r
, "Failed to query partition UUID: %m");
2061 label
= fdisk_partition_get_name(p
);
2062 if (!isempty(label
)) {
2063 label_copy
= strdup(label
);
2068 sz
= fdisk_partition_get_size(p
);
2069 assert(sz
<= UINT64_MAX
/secsz
);
2072 start
= fdisk_partition_get_start(p
);
2073 assert(start
<= UINT64_MAX
/secsz
);
2076 if (partition_type_exclude(&type
))
2079 np
= partition_new();
2085 np
->new_uuid_is_set
= true;
2086 np
->size_min
= np
->size_max
= sz
;
2087 np
->new_label
= TAKE_PTR(label_copy
);
2089 np
->definition_path
= strdup(src
);
2090 if (!np
->definition_path
)
2093 r
= determine_current_padding(c
, t
, p
, secsz
, grainsz
, &padding
);
2097 np
->padding_min
= np
->padding_max
= padding
;
2099 np
->copy_blocks_path
= strdup(src
);
2100 if (!np
->copy_blocks_path
)
2103 np
->copy_blocks_fd
= fcntl(fd
, F_DUPFD_CLOEXEC
, 3);
2104 if (np
->copy_blocks_fd
< 0)
2105 return log_error_errno(r
, "Failed to duplicate file descriptor of %s: %m", src
);
2107 np
->copy_blocks_offset
= start
;
2108 np
->copy_blocks_size
= sz
;
2110 r
= fdisk_partition_get_attrs_as_uint64(p
, &np
->gpt_flags
);
2112 return log_error_errno(r
, "Failed to get partition flags: %m");
2114 LIST_INSERT_AFTER(partitions
, context
->partitions
, last
, np
);
2115 last
= TAKE_PTR(np
);
2116 context
->n_partitions
++;
2122 static int context_copy_from(Context
*context
) {
2127 STRV_FOREACH(src
, arg_copy_from
) {
2128 r
= context_copy_from_one(context
, *src
);
2136 static int context_read_definitions(Context
*context
) {
2137 _cleanup_strv_free_
char **files
= NULL
;
2138 Partition
*last
= LIST_FIND_TAIL(partitions
, context
->partitions
);
2139 const char *const *dirs
;
2144 dirs
= (const char* const*) (arg_definitions
?: CONF_PATHS_STRV("repart.d"));
2146 r
= conf_files_list_strv(&files
, ".conf", arg_definitions
? NULL
: arg_root
, CONF_FILES_REGULAR
|CONF_FILES_FILTER_MASKED
, dirs
);
2148 return log_error_errno(r
, "Failed to enumerate *.conf files: %m");
2150 STRV_FOREACH(f
, files
) {
2151 _cleanup_(partition_freep
) Partition
*p
= NULL
;
2153 p
= partition_new();
2157 p
->definition_path
= strdup(*f
);
2158 if (!p
->definition_path
)
2161 r
= partition_read_definition(p
, *f
, dirs
);
2167 LIST_INSERT_AFTER(partitions
, context
->partitions
, last
, p
);
2169 context
->n_partitions
++;
2172 /* Check that each configured verity hash/data partition has a matching verity data/hash partition. */
2174 LIST_FOREACH(partitions
, p
, context
->partitions
) {
2175 if (p
->verity
== VERITY_OFF
)
2178 for (VerityMode mode
= VERITY_OFF
+ 1; mode
< _VERITY_MODE_MAX
; mode
++) {
2179 Partition
*q
= NULL
;
2181 if (p
->verity
== mode
)
2184 if (p
->siblings
[mode
])
2187 r
= find_verity_sibling(context
, p
, mode
, &q
);
2189 if (mode
!= VERITY_SIG
)
2190 return log_syntax(NULL
, LOG_ERR
, p
->definition_path
, 1, SYNTHETIC_ERRNO(EINVAL
),
2191 "Missing verity %s partition for verity %s partition with VerityMatchKey=%s",
2192 verity_mode_to_string(mode
), verity_mode_to_string(p
->verity
), p
->verity_match_key
);
2193 } else if (r
== -ENOTUNIQ
)
2194 return log_syntax(NULL
, LOG_ERR
, p
->definition_path
, 1, SYNTHETIC_ERRNO(EINVAL
),
2195 "Multiple verity %s partitions found for verity %s partition with VerityMatchKey=%s",
2196 verity_mode_to_string(mode
), verity_mode_to_string(p
->verity
), p
->verity_match_key
);
2198 return log_syntax(NULL
, LOG_ERR
, p
->definition_path
, 1, r
,
2199 "Failed to find verity %s partition for verity %s partition with VerityMatchKey=%s",
2200 verity_mode_to_string(mode
), verity_mode_to_string(p
->verity
), p
->verity_match_key
);
2203 if (q
->priority
!= p
->priority
)
2204 return log_syntax(NULL
, LOG_ERR
, p
->definition_path
, 1, SYNTHETIC_ERRNO(EINVAL
),
2205 "Priority mismatch (%i != %i) for verity sibling partitions with VerityMatchKey=%s",
2206 p
->priority
, q
->priority
, p
->verity_match_key
);
2208 p
->siblings
[mode
] = q
;
2213 LIST_FOREACH(partitions
, p
, context
->partitions
) {
2216 if (p
->verity
!= VERITY_HASH
)
2219 if (p
->minimize
== MINIMIZE_OFF
)
2222 assert_se(dp
= p
->siblings
[VERITY_DATA
]);
2224 if (dp
->minimize
== MINIMIZE_OFF
&& !(dp
->copy_blocks_path
|| dp
->copy_blocks_auto
))
2225 return log_syntax(NULL
, LOG_ERR
, p
->definition_path
, 1, SYNTHETIC_ERRNO(EINVAL
),
2226 "Minimize= set for verity hash partition but data partition does "
2227 "not set CopyBlocks= or Minimize=");
2234 static int fdisk_ask_cb(struct fdisk_context
*c
, struct fdisk_ask
*ask
, void *data
) {
2235 _cleanup_free_
char *ids
= NULL
;
2238 if (fdisk_ask_get_type(ask
) != FDISK_ASKTYPE_STRING
)
2241 ids
= new(char, SD_ID128_UUID_STRING_MAX
);
2245 r
= fdisk_ask_string_set_result(ask
, sd_id128_to_uuid_string(*(sd_id128_t
*) data
, ids
));
2253 static int fdisk_set_disklabel_id_by_uuid(struct fdisk_context
*c
, sd_id128_t id
) {
2256 r
= fdisk_set_ask(c
, fdisk_ask_cb
, &id
);
2260 r
= fdisk_set_disklabel_id(c
);
2264 return fdisk_set_ask(c
, NULL
, NULL
);
2267 static int derive_uuid(sd_id128_t base
, const char *token
, sd_id128_t
*ret
) {
2269 uint8_t md
[SHA256_DIGEST_SIZE
];
2276 /* Derive a new UUID from the specified UUID in a stable and reasonably safe way. Specifically, we
2277 * calculate the HMAC-SHA256 of the specified token string, keyed by the supplied base (typically the
2278 * machine ID). We use the machine ID as key (and not as cleartext!) of the HMAC operation since it's
2279 * the machine ID we don't want to leak. */
2281 hmac_sha256(base
.bytes
, sizeof(base
.bytes
), token
, strlen(token
), result
.md
);
2283 /* Take the first half, mark it as v4 UUID */
2284 assert_cc(sizeof(result
.md
) == sizeof(result
.id
) * 2);
2285 *ret
= id128_make_v4_uuid(result
.id
);
2289 static void derive_salt(sd_id128_t base
, const char *token
, uint8_t ret
[static SHA256_DIGEST_SIZE
]) {
2292 hmac_sha256(base
.bytes
, sizeof(base
.bytes
), token
, strlen(token
), ret
);
2295 static int context_load_partition_table(Context
*context
) {
2296 _cleanup_(fdisk_unref_contextp
) struct fdisk_context
*c
= NULL
;
2297 _cleanup_(fdisk_unref_tablep
) struct fdisk_table
*t
= NULL
;
2298 uint64_t left_boundary
= UINT64_MAX
, first_lba
, last_lba
, nsectors
;
2299 _cleanup_free_
char *disk_uuid_string
= NULL
;
2300 bool from_scratch
= false;
2301 sd_id128_t disk_uuid
;
2302 size_t n_partitions
;
2303 unsigned long secsz
;
2304 uint64_t grainsz
, fs_secsz
= DEFAULT_FILESYSTEM_SECTOR_SIZE
;
2308 assert(!context
->fdisk_context
);
2309 assert(!context
->free_areas
);
2310 assert(context
->start
== UINT64_MAX
);
2311 assert(context
->end
== UINT64_MAX
);
2312 assert(context
->total
== UINT64_MAX
);
2314 c
= fdisk_new_context();
2318 if (arg_sector_size
> 0) {
2319 fs_secsz
= arg_sector_size
;
2320 r
= fdisk_save_user_sector_size(c
, /* phy= */ 0, arg_sector_size
);
2325 r
= context_open_and_lock_backing_fd(context
->node
, arg_dry_run
? LOCK_SH
: LOCK_EX
,
2326 &context
->backing_fd
);
2330 if (fstat(context
->backing_fd
, &st
) < 0)
2331 return log_error_errno(r
, "Failed to stat %s: %m", context
->node
);
2333 /* Auto-detect sector size if not specified. */
2334 r
= probe_sector_size_prefer_ioctl(context
->backing_fd
, &ssz
);
2336 return log_error_errno(r
, "Failed to probe sector size of '%s': %m", context
->node
);
2338 /* If we found the sector size and we're operating on a block device, use it as the file
2339 * system sector size as well, as we know its the sector size of the actual block device and
2340 * not just the offset at which we found the GPT header. */
2341 if (r
> 0 && S_ISBLK(st
.st_mode
))
2344 r
= fdisk_save_user_sector_size(c
, /* phy= */ 0, ssz
);
2347 return log_error_errno(r
, "Failed to set sector size: %m");
2349 /* libfdisk doesn't have an API to operate on arbitrary fds, hence reopen the fd going via the
2350 * /proc/self/fd/ magic path if we have an existing fd. Open the original file otherwise. */
2351 r
= fdisk_assign_device(
2353 context
->backing_fd
>= 0 ? FORMAT_PROC_FD_PATH(context
->backing_fd
) : context
->node
,
2355 if (r
== -EINVAL
&& arg_size_auto
) {
2358 /* libfdisk returns EINVAL if opening a file of size zero. Let's check for that, and accept
2359 * it if automatic sizing is requested. */
2361 if (context
->backing_fd
< 0)
2362 r
= stat(context
->node
, &st
);
2364 r
= fstat(context
->backing_fd
, &st
);
2366 return log_error_errno(errno
, "Failed to stat block device '%s': %m", context
->node
);
2368 if (S_ISREG(st
.st_mode
) && st
.st_size
== 0) {
2369 /* Use the fallback values if we have no better idea */
2370 context
->sector_size
= fdisk_get_sector_size(c
);
2371 context
->fs_sector_size
= fs_secsz
;
2372 context
->grain_size
= 4096;
2373 return /* from_scratch = */ true;
2379 return log_error_errno(r
, "Failed to open device '%s': %m", context
->node
);
2381 if (context
->backing_fd
< 0) {
2382 /* If we have no fd referencing the device yet, make a copy of the fd now, so that we have one */
2383 r
= context_open_and_lock_backing_fd(FORMAT_PROC_FD_PATH(fdisk_get_devfd(c
)),
2384 arg_dry_run
? LOCK_SH
: LOCK_EX
,
2385 &context
->backing_fd
);
2390 /* The offsets/sizes libfdisk returns to us will be in multiple of the sector size of the
2391 * device. This is typically 512, and sometimes 4096. Let's query libfdisk once for it, and then use
2392 * it for all our needs. Note that the values we use ourselves always are in bytes though, thus mean
2393 * the same thing universally. Also note that regardless what kind of sector size is in use we'll
2394 * place partitions at multiples of 4K. */
2395 secsz
= fdisk_get_sector_size(c
);
2397 /* Insist on a power of two, and that it's a multiple of 512, i.e. the traditional sector size. */
2398 if (secsz
< 512 || !ISPOWEROF2(secsz
))
2399 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "Sector size %lu is not a power of two larger than 512? Refusing.", secsz
);
2401 /* Use at least 4K, and ensure it's a multiple of the sector size, regardless if that is smaller or
2403 grainsz
= secsz
< 4096 ? 4096 : secsz
;
2405 log_debug("Sector size of device is %lu bytes. Using grain size of %" PRIu64
".", secsz
, grainsz
);
2407 switch (arg_empty
) {
2410 /* Refuse empty disks, insist on an existing GPT partition table */
2411 if (!fdisk_is_labeltype(c
, FDISK_DISKLABEL_GPT
))
2412 return log_notice_errno(SYNTHETIC_ERRNO(EHWPOISON
), "Disk %s has no GPT disk label, not repartitioning.", context
->node
);
2417 /* Require an empty disk, refuse any existing partition table */
2418 r
= fdisk_has_label(c
);
2420 return log_error_errno(r
, "Failed to determine whether disk %s has a disk label: %m", context
->node
);
2422 return log_notice_errno(SYNTHETIC_ERRNO(EHWPOISON
), "Disk %s already has a disk label, refusing.", context
->node
);
2424 from_scratch
= true;
2428 /* Allow both an empty disk and an existing partition table, but only GPT */
2429 r
= fdisk_has_label(c
);
2431 return log_error_errno(r
, "Failed to determine whether disk %s has a disk label: %m", context
->node
);
2433 if (!fdisk_is_labeltype(c
, FDISK_DISKLABEL_GPT
))
2434 return log_notice_errno(SYNTHETIC_ERRNO(EHWPOISON
), "Disk %s has non-GPT disk label, not repartitioning.", context
->node
);
2436 from_scratch
= true;
2442 /* Always reinitiaize the disk, don't consider what there was on the disk before */
2443 from_scratch
= true;
2447 assert_not_reached();
2451 r
= fdisk_create_disklabel(c
, "gpt");
2453 return log_error_errno(r
, "Failed to create GPT disk label: %m");
2455 r
= derive_uuid(context
->seed
, "disk-uuid", &disk_uuid
);
2457 return log_error_errno(r
, "Failed to acquire disk GPT uuid: %m");
2459 r
= fdisk_set_disklabel_id_by_uuid(c
, disk_uuid
);
2461 return log_error_errno(r
, "Failed to set GPT disk label: %m");
2463 goto add_initial_free_area
;
2466 r
= fdisk_get_disklabel_id(c
, &disk_uuid_string
);
2468 return log_error_errno(r
, "Failed to get current GPT disk label UUID: %m");
2470 r
= id128_from_string_nonzero(disk_uuid_string
, &disk_uuid
);
2472 r
= derive_uuid(context
->seed
, "disk-uuid", &disk_uuid
);
2474 return log_error_errno(r
, "Failed to acquire disk GPT uuid: %m");
2476 r
= fdisk_set_disklabel_id(c
);
2478 return log_error_errno(r
, "Failed to set GPT disk label: %m");
2480 return log_error_errno(r
, "Failed to parse current GPT disk label UUID: %m");
2482 r
= fdisk_get_partitions(c
, &t
);
2484 return log_error_errno(r
, "Failed to acquire partition table: %m");
2486 n_partitions
= fdisk_table_get_nents(t
);
2487 for (size_t i
= 0; i
< n_partitions
; i
++) {
2488 _cleanup_free_
char *label_copy
= NULL
;
2489 Partition
*last
= NULL
;
2490 struct fdisk_partition
*p
;
2494 sd_id128_t ptid
, id
;
2497 p
= fdisk_table_get_partition(t
, i
);
2499 return log_error_errno(SYNTHETIC_ERRNO(EIO
), "Failed to read partition metadata: %m");
2501 if (fdisk_partition_is_used(p
) <= 0)
2504 if (fdisk_partition_has_start(p
) <= 0 ||
2505 fdisk_partition_has_size(p
) <= 0 ||
2506 fdisk_partition_has_partno(p
) <= 0)
2507 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "Found a partition without a position, size or number.");
2509 r
= fdisk_partition_get_type_as_id128(p
, &ptid
);
2511 return log_error_errno(r
, "Failed to query partition type UUID: %m");
2513 r
= fdisk_partition_get_uuid_as_id128(p
, &id
);
2515 return log_error_errno(r
, "Failed to query partition UUID: %m");
2517 label
= fdisk_partition_get_name(p
);
2518 if (!isempty(label
)) {
2519 label_copy
= strdup(label
);
2524 sz
= fdisk_partition_get_size(p
);
2525 assert(sz
<= UINT64_MAX
/secsz
);
2528 start
= fdisk_partition_get_start(p
);
2529 assert(start
<= UINT64_MAX
/secsz
);
2532 partno
= fdisk_partition_get_partno(p
);
2534 if (left_boundary
== UINT64_MAX
|| left_boundary
> start
)
2535 left_boundary
= start
;
2537 /* Assign this existing partition to the first partition of the right type that doesn't have
2538 * an existing one assigned yet. */
2539 LIST_FOREACH(partitions
, pp
, context
->partitions
) {
2542 if (!sd_id128_equal(pp
->type
.uuid
, ptid
))
2545 if (!pp
->current_partition
) {
2546 pp
->current_uuid
= id
;
2547 pp
->current_size
= sz
;
2549 pp
->partno
= partno
;
2550 pp
->current_label
= TAKE_PTR(label_copy
);
2552 pp
->current_partition
= p
;
2553 fdisk_ref_partition(p
);
2555 r
= determine_current_padding(c
, t
, p
, secsz
, grainsz
, &pp
->current_padding
);
2559 if (pp
->current_padding
> 0) {
2560 r
= context_add_free_area(context
, pp
->current_padding
, pp
);
2570 /* If we have no matching definition, create a new one. */
2572 _cleanup_(partition_freep
) Partition
*np
= NULL
;
2574 np
= partition_new();
2578 np
->current_uuid
= id
;
2579 np
->type
= gpt_partition_type_from_uuid(ptid
);
2580 np
->current_size
= sz
;
2582 np
->partno
= partno
;
2583 np
->current_label
= TAKE_PTR(label_copy
);
2585 np
->current_partition
= p
;
2586 fdisk_ref_partition(p
);
2588 r
= determine_current_padding(c
, t
, p
, secsz
, grainsz
, &np
->current_padding
);
2592 if (np
->current_padding
> 0) {
2593 r
= context_add_free_area(context
, np
->current_padding
, np
);
2598 LIST_INSERT_AFTER(partitions
, context
->partitions
, last
, TAKE_PTR(np
));
2599 context
->n_partitions
++;
2603 add_initial_free_area
:
2604 nsectors
= fdisk_get_nsectors(c
);
2605 assert(nsectors
<= UINT64_MAX
/secsz
);
2608 first_lba
= fdisk_get_first_lba(c
);
2609 assert(first_lba
<= UINT64_MAX
/secsz
);
2612 last_lba
= fdisk_get_last_lba(c
);
2613 assert(last_lba
< UINT64_MAX
);
2615 assert(last_lba
<= UINT64_MAX
/secsz
);
2618 assert(last_lba
>= first_lba
);
2620 if (left_boundary
== UINT64_MAX
) {
2621 /* No partitions at all? Then the whole disk is up for grabs. */
2623 first_lba
= round_up_size(first_lba
, grainsz
);
2624 last_lba
= round_down_size(last_lba
, grainsz
);
2626 if (last_lba
> first_lba
) {
2627 r
= context_add_free_area(context
, last_lba
- first_lba
, NULL
);
2632 /* Add space left of first partition */
2633 assert(left_boundary
>= first_lba
);
2635 first_lba
= round_up_size(first_lba
, grainsz
);
2636 left_boundary
= round_down_size(left_boundary
, grainsz
);
2637 last_lba
= round_down_size(last_lba
, grainsz
);
2639 if (left_boundary
> first_lba
) {
2640 r
= context_add_free_area(context
, left_boundary
- first_lba
, NULL
);
2646 context
->start
= first_lba
;
2647 context
->end
= last_lba
;
2648 context
->total
= nsectors
;
2649 context
->sector_size
= secsz
;
2650 context
->fs_sector_size
= fs_secsz
;
2651 context
->grain_size
= grainsz
;
2652 context
->fdisk_context
= TAKE_PTR(c
);
2654 return from_scratch
;
2657 static void context_unload_partition_table(Context
*context
) {
2660 LIST_FOREACH(partitions
, p
, context
->partitions
) {
2662 /* Entirely remove partitions that have no configuration */
2663 if (PARTITION_IS_FOREIGN(p
)) {
2664 partition_unlink_and_free(context
, p
);
2668 /* Otherwise drop all data we read off the block device and everything we might have
2669 * calculated based on it */
2672 p
->current_size
= UINT64_MAX
;
2673 p
->new_size
= UINT64_MAX
;
2674 p
->current_padding
= UINT64_MAX
;
2675 p
->new_padding
= UINT64_MAX
;
2676 p
->partno
= UINT64_MAX
;
2677 p
->offset
= UINT64_MAX
;
2679 if (p
->current_partition
) {
2680 fdisk_unref_partition(p
->current_partition
);
2681 p
->current_partition
= NULL
;
2684 if (p
->new_partition
) {
2685 fdisk_unref_partition(p
->new_partition
);
2686 p
->new_partition
= NULL
;
2689 p
->padding_area
= NULL
;
2690 p
->allocated_to_area
= NULL
;
2692 p
->current_uuid
= SD_ID128_NULL
;
2693 p
->current_label
= mfree(p
->current_label
);
2696 context
->start
= UINT64_MAX
;
2697 context
->end
= UINT64_MAX
;
2698 context
->total
= UINT64_MAX
;
2700 if (context
->fdisk_context
) {
2701 fdisk_unref_context(context
->fdisk_context
);
2702 context
->fdisk_context
= NULL
;
2705 context_free_free_areas(context
);
2708 static int format_size_change(uint64_t from
, uint64_t to
, char **ret
) {
2711 if (from
!= UINT64_MAX
) {
2712 if (from
== to
|| to
== UINT64_MAX
)
2713 t
= strdup(FORMAT_BYTES(from
));
2715 t
= strjoin(FORMAT_BYTES(from
), " ", special_glyph(SPECIAL_GLYPH_ARROW_RIGHT
), " ", FORMAT_BYTES(to
));
2716 } else if (to
!= UINT64_MAX
)
2717 t
= strjoin(special_glyph(SPECIAL_GLYPH_ARROW_RIGHT
), " ", FORMAT_BYTES(to
));
2730 static const char *partition_label(const Partition
*p
) {
2734 return p
->new_label
;
2736 if (p
->current_label
)
2737 return p
->current_label
;
2739 return gpt_partition_type_uuid_to_string(p
->type
.uuid
);
2742 static int context_dump_partitions(Context
*context
) {
2743 _cleanup_(table_unrefp
) Table
*t
= NULL
;
2744 uint64_t sum_padding
= 0, sum_size
= 0;
2746 const size_t roothash_col
= 14, dropin_files_col
= 15, split_path_col
= 16;
2747 bool has_roothash
= false, has_dropin_files
= false, has_split_path
= false;
2749 if ((arg_json_format_flags
& JSON_FORMAT_OFF
) && context
->n_partitions
== 0) {
2750 log_info("Empty partition table.");
2754 t
= table_new("type",
2774 if (!DEBUG_LOGGING
) {
2775 if (arg_json_format_flags
& JSON_FORMAT_OFF
)
2776 (void) table_set_display(t
, (size_t) 0, (size_t) 1, (size_t) 2, (size_t) 3, (size_t) 4,
2777 (size_t) 8, (size_t) 9, (size_t) 12, roothash_col
, dropin_files_col
,
2780 (void) table_set_display(t
, (size_t) 0, (size_t) 1, (size_t) 2, (size_t) 3, (size_t) 4,
2781 (size_t) 5, (size_t) 6, (size_t) 7, (size_t) 8, (size_t) 10,
2782 (size_t) 11, (size_t) 13, roothash_col
, dropin_files_col
,
2786 (void) table_set_align_percent(t
, table_get_cell(t
, 0, 5), 100);
2787 (void) table_set_align_percent(t
, table_get_cell(t
, 0, 6), 100);
2788 (void) table_set_align_percent(t
, table_get_cell(t
, 0, 7), 100);
2789 (void) table_set_align_percent(t
, table_get_cell(t
, 0, 8), 100);
2790 (void) table_set_align_percent(t
, table_get_cell(t
, 0, 9), 100);
2791 (void) table_set_align_percent(t
, table_get_cell(t
, 0, 10), 100);
2792 (void) table_set_align_percent(t
, table_get_cell(t
, 0, 11), 100);
2794 LIST_FOREACH(partitions
, p
, context
->partitions
) {
2795 _cleanup_free_
char *size_change
= NULL
, *padding_change
= NULL
, *partname
= NULL
, *rh
= NULL
;
2796 char uuid_buffer
[SD_ID128_UUID_STRING_MAX
];
2797 const char *label
, *activity
= NULL
;
2802 if (p
->current_size
== UINT64_MAX
)
2803 activity
= "create";
2804 else if (p
->current_size
!= p
->new_size
)
2805 activity
= "resize";
2807 label
= partition_label(p
);
2808 partname
= p
->partno
!= UINT64_MAX
? fdisk_partname(context
->node
, p
->partno
+1) : NULL
;
2810 r
= format_size_change(p
->current_size
, p
->new_size
, &size_change
);
2814 r
= format_size_change(p
->current_padding
, p
->new_padding
, &padding_change
);
2818 if (p
->new_size
!= UINT64_MAX
)
2819 sum_size
+= p
->new_size
;
2820 if (p
->new_padding
!= UINT64_MAX
)
2821 sum_padding
+= p
->new_padding
;
2823 if (p
->verity
!= VERITY_OFF
) {
2824 Partition
*hp
= p
->verity
== VERITY_HASH
? p
: p
->siblings
[VERITY_HASH
];
2826 rh
= iovec_is_set(&hp
->roothash
) ? hexmem(hp
->roothash
.iov_base
, hp
->roothash
.iov_len
) : strdup("TBD");
2833 TABLE_STRING
, gpt_partition_type_uuid_to_string_harder(p
->type
.uuid
, uuid_buffer
),
2834 TABLE_STRING
, empty_to_null(label
) ?: "-", TABLE_SET_COLOR
, empty_to_null(label
) ? NULL
: ansi_grey(),
2835 TABLE_UUID
, p
->new_uuid_is_set
? p
->new_uuid
: p
->current_uuid
,
2836 TABLE_UINT64
, p
->partno
,
2837 TABLE_PATH_BASENAME
, p
->definition_path
, TABLE_SET_COLOR
, p
->definition_path
? NULL
: ansi_grey(),
2838 TABLE_STRING
, partname
?: "-", TABLE_SET_COLOR
, partname
? NULL
: ansi_highlight(),
2839 TABLE_UINT64
, p
->offset
,
2840 TABLE_UINT64
, p
->current_size
== UINT64_MAX
? 0 : p
->current_size
,
2841 TABLE_UINT64
, p
->new_size
,
2842 TABLE_STRING
, size_change
, TABLE_SET_COLOR
, !p
->partitions_next
&& sum_size
> 0 ? ansi_underline() : NULL
,
2843 TABLE_UINT64
, p
->current_padding
== UINT64_MAX
? 0 : p
->current_padding
,
2844 TABLE_UINT64
, p
->new_padding
,
2845 TABLE_STRING
, padding_change
, TABLE_SET_COLOR
, !p
->partitions_next
&& sum_padding
> 0 ? ansi_underline() : NULL
,
2846 TABLE_STRING
, activity
?: "unchanged",
2848 TABLE_STRV
, p
->drop_in_files
,
2849 TABLE_STRING
, empty_to_null(p
->split_path
) ?: "-");
2851 return table_log_add_error(r
);
2853 has_roothash
= has_roothash
|| !isempty(rh
);
2854 has_dropin_files
= has_dropin_files
|| !strv_isempty(p
->drop_in_files
);
2855 has_split_path
= has_split_path
|| !isempty(p
->split_path
);
2858 if ((arg_json_format_flags
& JSON_FORMAT_OFF
) && (sum_padding
> 0 || sum_size
> 0)) {
2861 a
= strjoina(special_glyph(SPECIAL_GLYPH_SIGMA
), " = ", FORMAT_BYTES(sum_size
));
2862 b
= strjoina(special_glyph(SPECIAL_GLYPH_SIGMA
), " = ", FORMAT_BYTES(sum_padding
));
2884 return table_log_add_error(r
);
2887 if (!has_roothash
) {
2888 r
= table_hide_column_from_display(t
, roothash_col
);
2890 return log_error_errno(r
, "Failed to set columns to display: %m");
2893 if (!has_dropin_files
) {
2894 r
= table_hide_column_from_display(t
, dropin_files_col
);
2896 return log_error_errno(r
, "Failed to set columns to display: %m");
2899 if (!has_split_path
) {
2900 r
= table_hide_column_from_display(t
, split_path_col
);
2902 return log_error_errno(r
, "Failed to set columns to display: %m");
2905 return table_print_with_pager(t
, arg_json_format_flags
, arg_pager_flags
, arg_legend
);
2908 static void context_bar_char_process_partition(
2913 size_t *ret_start
) {
2915 uint64_t from
, to
, total
;
2926 assert(p
->offset
!= UINT64_MAX
);
2927 assert(p
->new_size
!= UINT64_MAX
);
2930 to
= from
+ p
->new_size
;
2932 assert(context
->total
> 0);
2933 total
= context
->total
;
2935 assert(from
<= total
);
2936 x
= from
* n
/ total
;
2938 assert(to
<= total
);
2944 for (size_t i
= x
; i
< y
; i
++)
2950 static int partition_hint(const Partition
*p
, const char *node
, char **ret
) {
2951 _cleanup_free_
char *buf
= NULL
;
2955 /* Tries really hard to find a suitable description for this partition */
2957 if (p
->definition_path
)
2958 return path_extract_filename(p
->definition_path
, ret
);
2960 label
= partition_label(p
);
2961 if (!isempty(label
)) {
2962 buf
= strdup(label
);
2966 if (p
->partno
!= UINT64_MAX
) {
2967 buf
= fdisk_partname(node
, p
->partno
+1);
2971 if (p
->new_uuid_is_set
)
2973 else if (!sd_id128_is_null(p
->current_uuid
))
2974 id
= p
->current_uuid
;
2978 buf
= strdup(SD_ID128_TO_UUID_STRING(id
));
2984 *ret
= TAKE_PTR(buf
);
2988 static int context_dump_partition_bar(Context
*context
) {
2989 _cleanup_free_ Partition
**bar
= NULL
;
2990 _cleanup_free_
size_t *start_array
= NULL
;
2991 Partition
*last
= NULL
;
2995 assert_se((c
= columns()) >= 2);
2996 c
-= 2; /* We do not use the leftmost and rightmost character cell */
2998 bar
= new0(Partition
*, c
);
3002 start_array
= new(size_t, context
->n_partitions
);
3006 LIST_FOREACH(partitions
, p
, context
->partitions
)
3007 context_bar_char_process_partition(context
, bar
, c
, p
, start_array
+ j
++);
3011 for (size_t i
= 0; i
< c
; i
++) {
3016 fputs(z
? ansi_green() : ansi_yellow(), stdout
);
3017 fputs(special_glyph(SPECIAL_GLYPH_DARK_SHADE
), stdout
);
3019 fputs(ansi_normal(), stdout
);
3020 fputs(special_glyph(SPECIAL_GLYPH_LIGHT_SHADE
), stdout
);
3026 fputs(ansi_normal(), stdout
);
3029 for (size_t i
= 0; i
< context
->n_partitions
; i
++) {
3030 _cleanup_free_
char **line
= NULL
;
3032 line
= new0(char*, c
);
3037 LIST_FOREACH(partitions
, p
, context
->partitions
) {
3038 _cleanup_free_
char *d
= NULL
;
3041 if (i
< context
->n_partitions
- j
) {
3043 if (line
[start_array
[j
-1]]) {
3046 /* Upgrade final corner to the right with a branch to the right */
3047 e
= startswith(line
[start_array
[j
-1]], special_glyph(SPECIAL_GLYPH_TREE_RIGHT
));
3049 d
= strjoin(special_glyph(SPECIAL_GLYPH_TREE_BRANCH
), e
);
3056 d
= strdup(special_glyph(SPECIAL_GLYPH_TREE_VERTICAL
));
3061 } else if (i
== context
->n_partitions
- j
) {
3062 _cleanup_free_
char *hint
= NULL
;
3064 (void) partition_hint(p
, context
->node
, &hint
);
3066 if (streq_ptr(line
[start_array
[j
-1]], special_glyph(SPECIAL_GLYPH_TREE_VERTICAL
)))
3067 d
= strjoin(special_glyph(SPECIAL_GLYPH_TREE_BRANCH
), " ", strna(hint
));
3069 d
= strjoin(special_glyph(SPECIAL_GLYPH_TREE_RIGHT
), " ", strna(hint
));
3076 free_and_replace(line
[start_array
[j
-1]], d
);
3084 fputs(line
[j
], stdout
);
3085 j
+= utf8_console_width(line
[j
]);
3094 for (j
= 0; j
< c
; j
++)
3101 static bool context_has_roothash(Context
*context
) {
3102 LIST_FOREACH(partitions
, p
, context
->partitions
)
3103 if (iovec_is_set(&p
->roothash
))
3109 static int context_dump(Context
*context
, bool late
) {
3114 if (arg_pretty
== 0 && FLAGS_SET(arg_json_format_flags
, JSON_FORMAT_OFF
))
3117 /* If we're outputting JSON, only dump after doing all operations so we can include the roothashes
3119 if (!late
&& !FLAGS_SET(arg_json_format_flags
, JSON_FORMAT_OFF
))
3122 /* If we're not outputting JSON, only dump again after doing all operations if there are any
3123 * roothashes that we need to communicate to the user. */
3124 if (late
&& FLAGS_SET(arg_json_format_flags
, JSON_FORMAT_OFF
) && !context_has_roothash(context
))
3127 r
= context_dump_partitions(context
);
3131 /* Make sure we only write the partition bar once, even if we're writing the partition table twice to
3132 * communicate roothashes. */
3133 if (FLAGS_SET(arg_json_format_flags
, JSON_FORMAT_OFF
) && !late
) {
3136 r
= context_dump_partition_bar(context
);
3149 static bool context_changed(const Context
*context
) {
3152 LIST_FOREACH(partitions
, p
, context
->partitions
) {
3156 if (p
->allocated_to_area
)
3159 if (p
->new_size
!= p
->current_size
)
3166 static int context_wipe_range(Context
*context
, uint64_t offset
, uint64_t size
) {
3167 _cleanup_(blkid_free_probep
) blkid_probe probe
= NULL
;
3171 assert(offset
!= UINT64_MAX
);
3172 assert(size
!= UINT64_MAX
);
3174 probe
= blkid_new_probe();
3179 r
= blkid_probe_set_device(probe
, fdisk_get_devfd(context
->fdisk_context
), offset
, size
);
3181 return log_error_errno(errno
?: SYNTHETIC_ERRNO(EIO
), "Failed to allocate device probe for wiping.");
3184 if (blkid_probe_enable_superblocks(probe
, true) < 0 ||
3185 blkid_probe_set_superblocks_flags(probe
, BLKID_SUBLKS_MAGIC
|BLKID_SUBLKS_BADCSUM
) < 0 ||
3186 blkid_probe_enable_partitions(probe
, true) < 0 ||
3187 blkid_probe_set_partitions_flags(probe
, BLKID_PARTS_MAGIC
) < 0)
3188 return log_error_errno(errno
?: SYNTHETIC_ERRNO(EIO
), "Failed to enable superblock and partition probing.");
3192 r
= blkid_do_probe(probe
);
3194 return log_error_errno(errno
?: SYNTHETIC_ERRNO(EIO
), "Failed to probe for file systems.");
3199 if (blkid_do_wipe(probe
, false) < 0)
3200 return log_error_errno(errno
?: SYNTHETIC_ERRNO(EIO
), "Failed to wipe file system signature.");
3206 static int context_wipe_partition(Context
*context
, Partition
*p
) {
3211 assert(!PARTITION_EXISTS(p
)); /* Safety check: never wipe existing partitions */
3213 assert(p
->offset
!= UINT64_MAX
);
3214 assert(p
->new_size
!= UINT64_MAX
);
3216 r
= context_wipe_range(context
, p
->offset
, p
->new_size
);
3220 log_info("Successfully wiped file system signatures from future partition %" PRIu64
".", p
->partno
);
3224 static int context_discard_range(
3233 assert(offset
!= UINT64_MAX
);
3234 assert(size
!= UINT64_MAX
);
3239 assert_se((fd
= fdisk_get_devfd(context
->fdisk_context
)) >= 0);
3241 if (fstat(fd
, &st
) < 0)
3244 if (S_ISREG(st
.st_mode
)) {
3245 if (fallocate(fd
, FALLOC_FL_PUNCH_HOLE
|FALLOC_FL_KEEP_SIZE
, offset
, size
) < 0) {
3246 if (ERRNO_IS_NOT_SUPPORTED(errno
))
3255 if (S_ISBLK(st
.st_mode
)) {
3256 uint64_t range
[2], end
;
3258 range
[0] = round_up_size(offset
, context
->sector_size
);
3260 if (offset
> UINT64_MAX
- size
)
3263 end
= offset
+ size
;
3264 if (end
<= range
[0])
3267 range
[1] = round_down_size(end
- range
[0], context
->sector_size
);
3271 if (ioctl(fd
, BLKDISCARD
, range
) < 0) {
3272 if (ERRNO_IS_NOT_SUPPORTED(errno
))
3284 static int context_discard_partition(Context
*context
, Partition
*p
) {
3290 assert(p
->offset
!= UINT64_MAX
);
3291 assert(p
->new_size
!= UINT64_MAX
);
3292 assert(!PARTITION_EXISTS(p
)); /* Safety check: never discard existing partitions */
3297 r
= context_discard_range(context
, p
->offset
, p
->new_size
);
3298 if (r
== -EOPNOTSUPP
) {
3299 log_info("Storage does not support discard, not discarding data in future partition %" PRIu64
".", p
->partno
);
3303 /* Let's handle this gracefully: https://bugzilla.kernel.org/show_bug.cgi?id=211167 */
3304 log_info("Block device is busy, not discarding partition %" PRIu64
" because it probably is mounted.", p
->partno
);
3308 log_info("Partition %" PRIu64
" too short for discard, skipping.", p
->partno
);
3312 return log_error_errno(r
, "Failed to discard data for future partition %" PRIu64
".", p
->partno
);
3314 log_info("Successfully discarded data from future partition %" PRIu64
".", p
->partno
);
3318 static int context_discard_gap_after(Context
*context
, Partition
*p
) {
3319 uint64_t gap
, next
= UINT64_MAX
;
3323 assert(!p
|| (p
->offset
!= UINT64_MAX
&& p
->new_size
!= UINT64_MAX
));
3329 gap
= p
->offset
+ p
->new_size
;
3331 /* The context start gets rounded up to grain_size, however
3332 * existing partitions may be before that so ensure the gap
3333 * starts at the first actually usable lba
3335 gap
= fdisk_get_first_lba(context
->fdisk_context
) * context
->sector_size
;
3337 LIST_FOREACH(partitions
, q
, context
->partitions
) {
3341 assert(q
->offset
!= UINT64_MAX
);
3342 assert(q
->new_size
!= UINT64_MAX
);
3344 if (q
->offset
< gap
)
3347 if (next
== UINT64_MAX
|| q
->offset
< next
)
3351 if (next
== UINT64_MAX
) {
3352 next
= (fdisk_get_last_lba(context
->fdisk_context
) + 1) * context
->sector_size
;
3354 return log_error_errno(SYNTHETIC_ERRNO(EIO
), "Partition end beyond disk end.");
3357 assert(next
>= gap
);
3358 r
= context_discard_range(context
, gap
, next
- gap
);
3359 if (r
== -EOPNOTSUPP
) {
3361 log_info("Storage does not support discard, not discarding gap after partition %" PRIu64
".", p
->partno
);
3363 log_info("Storage does not support discard, not discarding gap at beginning of disk.");
3366 if (r
== 0) /* Too short */
3370 return log_error_errno(r
, "Failed to discard gap after partition %" PRIu64
".", p
->partno
);
3372 return log_error_errno(r
, "Failed to discard gap at beginning of disk.");
3376 log_info("Successfully discarded gap after partition %" PRIu64
".", p
->partno
);
3378 log_info("Successfully discarded gap at beginning of disk.");
3383 static int context_wipe_and_discard(Context
*context
) {
3388 if (arg_empty
== EMPTY_CREATE
) /* If we just created the image, no need to wipe */
3391 /* Wipe and discard the contents of all partitions we are about to create. We skip the discarding if
3392 * we were supposed to start from scratch anyway, as in that case we just discard the whole block
3393 * device in one go early on. */
3395 LIST_FOREACH(partitions
, p
, context
->partitions
) {
3397 if (!p
->allocated_to_area
)
3400 if (partition_type_defer(&p
->type
))
3403 r
= context_wipe_partition(context
, p
);
3407 if (!context
->from_scratch
) {
3408 r
= context_discard_partition(context
, p
);
3412 r
= context_discard_gap_after(context
, p
);
3418 if (!context
->from_scratch
) {
3419 r
= context_discard_gap_after(context
, NULL
);
3427 typedef struct DecryptedPartitionTarget
{
3431 struct crypt_device
*device
;
3432 } DecryptedPartitionTarget
;
3434 static DecryptedPartitionTarget
* decrypted_partition_target_free(DecryptedPartitionTarget
*t
) {
3435 #if HAVE_LIBCRYPTSETUP
3443 /* udev or so might access out block device in the background while we are done. Let's hence
3444 * force detach the volume. We sync'ed before, hence this should be safe. */
3445 r
= sym_crypt_deactivate_by_name(t
->device
, t
->dm_name
, CRYPT_DEACTIVATE_FORCE
);
3447 log_warning_errno(r
, "Failed to deactivate LUKS device, ignoring: %m");
3449 sym_crypt_free(t
->device
);
3462 DecryptedPartitionTarget
*decrypted
;
3465 static int partition_target_fd(PartitionTarget
*t
) {
3467 assert(t
->loop
|| t
->fd
>= 0 || t
->whole_fd
>= 0);
3470 return t
->decrypted
->fd
;
3481 static const char* partition_target_path(PartitionTarget
*t
) {
3483 assert(t
->loop
|| t
->path
);
3486 return t
->decrypted
->volume
;
3489 return t
->loop
->node
;
3494 static PartitionTarget
*partition_target_free(PartitionTarget
*t
) {
3498 decrypted_partition_target_free(t
->decrypted
);
3499 loop_device_unref(t
->loop
);
3501 unlink_and_free(t
->path
);
3506 DEFINE_TRIVIAL_CLEANUP_FUNC(PartitionTarget
*, partition_target_free
);
3508 static int prepare_temporary_file(PartitionTarget
*t
, uint64_t size
) {
3509 _cleanup_(unlink_and_freep
) char *temp
= NULL
;
3510 _cleanup_close_
int fd
= -EBADF
;
3516 r
= var_tmp_dir(&vt
);
3518 return log_error_errno(r
, "Could not determine temporary directory: %m");
3520 temp
= path_join(vt
, "repart-XXXXXX");
3524 fd
= mkostemp_safe(temp
);
3526 return log_error_errno(fd
, "Failed to create temporary file: %m");
3528 if (ftruncate(fd
, size
) < 0)
3529 return log_error_errno(errno
, "Failed to truncate temporary file to %s: %m",
3530 FORMAT_BYTES(size
));
3532 t
->fd
= TAKE_FD(fd
);
3533 t
->path
= TAKE_PTR(temp
);
3538 static int partition_target_prepare(
3543 PartitionTarget
**ret
) {
3545 _cleanup_(partition_target_freep
) PartitionTarget
*t
= NULL
;
3546 _cleanup_(loop_device_unrefp
) LoopDevice
*d
= NULL
;
3553 assert_se((whole_fd
= fdisk_get_devfd(context
->fdisk_context
)) >= 0);
3555 t
= new(PartitionTarget
, 1);
3558 *t
= (PartitionTarget
) {
3564 if (lseek(whole_fd
, p
->offset
, SEEK_SET
) < 0)
3565 return log_error_errno(errno
, "Failed to seek to partition offset: %m");
3567 t
->whole_fd
= whole_fd
;
3572 /* Loopback block devices are not only useful to turn regular files into block devices, but
3573 * also to cut out sections of block devices into new block devices. */
3575 if (arg_offline
<= 0) {
3576 r
= loop_device_make(whole_fd
, O_RDWR
, p
->offset
, size
, context
->sector_size
, 0, LOCK_EX
, &d
);
3577 if (r
< 0 && (arg_offline
== 0 || (r
!= -ENOENT
&& !ERRNO_IS_PRIVILEGE(r
)) || !strv_isempty(p
->subvolumes
)))
3578 return log_error_errno(r
, "Failed to make loopback device of future partition %" PRIu64
": %m", p
->partno
);
3580 t
->loop
= TAKE_PTR(d
);
3585 log_debug_errno(r
, "No access to loop devices, falling back to a regular file");
3588 /* If we can't allocate a loop device, let's write to a regular file that we copy into the final
3589 * image so we can run in containers and without needing root privileges. On filesystems with
3590 * reflinking support, we can take advantage of this and just reflink the result into the image.
3593 r
= prepare_temporary_file(t
, size
);
3602 static int partition_target_grow(PartitionTarget
*t
, uint64_t size
) {
3606 assert(!t
->decrypted
);
3609 r
= loop_device_refresh_size(t
->loop
, UINT64_MAX
, size
);
3611 return log_error_errno(r
, "Failed to refresh loopback device size: %m");
3612 } else if (t
->fd
>= 0) {
3613 if (ftruncate(t
->fd
, size
) < 0)
3614 return log_error_errno(errno
, "Failed to grow '%s' to %s by truncation: %m",
3615 t
->path
, FORMAT_BYTES(size
));
3621 static int partition_target_sync(Context
*context
, Partition
*p
, PartitionTarget
*t
) {
3628 assert_se((whole_fd
= fdisk_get_devfd(context
->fdisk_context
)) >= 0);
3630 if (t
->decrypted
&& fsync(t
->decrypted
->fd
) < 0)
3631 return log_error_errno(errno
, "Failed to sync changes to '%s': %m", t
->decrypted
->volume
);
3634 r
= loop_device_sync(t
->loop
);
3636 return log_error_errno(r
, "Failed to sync loopback device: %m");
3637 } else if (t
->fd
>= 0) {
3640 if (lseek(whole_fd
, p
->offset
, SEEK_SET
) < 0)
3641 return log_error_errno(errno
, "Failed to seek to partition offset: %m");
3643 if (lseek(t
->fd
, 0, SEEK_SET
) < 0)
3644 return log_error_errno(errno
, "Failed to seek to start of temporary file: %m");
3646 if (fstat(t
->fd
, &st
) < 0)
3647 return log_error_errno(errno
, "Failed to stat temporary file: %m");
3649 if (st
.st_size
> (off_t
) p
->new_size
)
3650 return log_error_errno(SYNTHETIC_ERRNO(ENOSPC
),
3651 "Partition %" PRIu64
"'s contents (%s) don't fit in the partition (%s)",
3652 p
->partno
, FORMAT_BYTES(st
.st_size
), FORMAT_BYTES(p
->new_size
));
3654 r
= copy_bytes(t
->fd
, whole_fd
, UINT64_MAX
, COPY_REFLINK
|COPY_HOLES
|COPY_FSYNC
);
3656 return log_error_errno(r
, "Failed to copy bytes to partition: %m");
3658 if (fsync(t
->whole_fd
) < 0)
3659 return log_error_errno(errno
, "Failed to sync changes: %m");
3665 static int partition_encrypt(Context
*context
, Partition
*p
, PartitionTarget
*target
, bool offline
) {
3666 #if HAVE_LIBCRYPTSETUP && HAVE_CRYPT_SET_DATA_OFFSET && HAVE_CRYPT_REENCRYPT_INIT_BY_PASSPHRASE && HAVE_CRYPT_REENCRYPT
3667 const char *node
= partition_target_path(target
);
3668 struct crypt_params_luks2 luks_params
= {
3669 .label
= strempty(ASSERT_PTR(p
)->new_label
),
3670 .sector_size
= ASSERT_PTR(context
)->fs_sector_size
,
3671 .data_device
= offline
? node
: NULL
,
3673 struct crypt_params_reencrypt reencrypt_params
= {
3674 .mode
= CRYPT_REENCRYPT_ENCRYPT
,
3675 .direction
= CRYPT_REENCRYPT_BACKWARD
,
3676 .resilience
= "datashift",
3677 .data_shift
= LUKS2_METADATA_SIZE
/ 512,
3678 .luks2
= &luks_params
,
3679 .flags
= CRYPT_REENCRYPT_INITIALIZE_ONLY
|CRYPT_REENCRYPT_MOVE_FIRST_SEGMENT
,
3681 _cleanup_(sym_crypt_freep
) struct crypt_device
*cd
= NULL
;
3682 _cleanup_(erase_and_freep
) char *base64_encoded
= NULL
;
3683 _cleanup_fclose_
FILE *h
= NULL
;
3684 _cleanup_free_
char *hp
= NULL
, *vol
= NULL
, *dm_name
= NULL
;
3685 const char *passphrase
= NULL
;
3686 size_t passphrase_size
= 0;
3692 assert(p
->encrypt
!= ENCRYPT_OFF
);
3694 r
= dlopen_cryptsetup();
3696 return log_error_errno(r
, "libcryptsetup not found, cannot encrypt: %m");
3698 log_info("Encrypting future partition %" PRIu64
"...", p
->partno
);
3701 r
= var_tmp_dir(&vt
);
3703 return log_error_errno(r
, "Failed to determine temporary files directory: %m");
3705 r
= fopen_temporary_child(vt
, &h
, &hp
);
3707 return log_error_errno(r
, "Failed to create temporary LUKS header file: %m");
3709 /* Weird cryptsetup requirement which requires the header file to be the size of at least one
3711 r
= ftruncate(fileno(h
), luks_params
.sector_size
);
3713 return log_error_errno(r
, "Failed to grow temporary LUKS header file: %m");
3715 if (asprintf(&dm_name
, "luks-repart-%08" PRIx64
, random_u64()) < 0)
3718 vol
= path_join("/dev/mapper/", dm_name
);
3723 r
= sym_crypt_init(&cd
, offline
? hp
: node
);
3725 return log_error_errno(r
, "Failed to allocate libcryptsetup context for %s: %m", hp
);
3727 cryptsetup_enable_logging(cd
);
3730 /* Disable kernel keyring usage by libcryptsetup as a workaround for
3731 * https://gitlab.com/cryptsetup/cryptsetup/-/merge_requests/273. This makes sure that we can
3732 * do offline encryption even when repart is running in a container. */
3733 r
= sym_crypt_volume_key_keyring(cd
, false);
3735 return log_error_errno(r
, "Failed to disable kernel keyring: %m");
3737 r
= sym_crypt_metadata_locking(cd
, false);
3739 return log_error_errno(r
, "Failed to disable metadata locking: %m");
3741 r
= sym_crypt_set_data_offset(cd
, LUKS2_METADATA_SIZE
/ 512);
3743 return log_error_errno(r
, "Failed to set data offset: %m");
3746 r
= sym_crypt_format(cd
,
3750 SD_ID128_TO_UUID_STRING(p
->luks_uuid
),
3755 return log_error_errno(r
, "Failed to LUKS2 format future partition: %m");
3757 if (IN_SET(p
->encrypt
, ENCRYPT_KEY_FILE
, ENCRYPT_KEY_FILE_TPM2
)) {
3758 r
= sym_crypt_keyslot_add_by_volume_key(
3766 return log_error_errno(r
, "Failed to add LUKS2 key: %m");
3768 passphrase
= strempty(arg_key
);
3769 passphrase_size
= arg_key_size
;
3772 if (IN_SET(p
->encrypt
, ENCRYPT_TPM2
, ENCRYPT_KEY_FILE_TPM2
)) {
3774 _cleanup_(json_variant_unrefp
) JsonVariant
*v
= NULL
;
3775 _cleanup_(erase_and_freep
) void *secret
= NULL
;
3776 _cleanup_free_
void *pubkey
= NULL
;
3777 _cleanup_free_
void *blob
= NULL
, *srk_buf
= NULL
;
3778 size_t secret_size
, blob_size
, pubkey_size
= 0, srk_buf_size
= 0;
3779 ssize_t base64_encoded_size
;
3782 if (arg_tpm2_public_key_pcr_mask
!= 0) {
3783 r
= tpm2_load_pcr_public_key(arg_tpm2_public_key
, &pubkey
, &pubkey_size
);
3785 if (arg_tpm2_public_key
|| r
!= -ENOENT
)
3786 return log_error_errno(r
, "Failed to read TPM PCR public key: %m");
3788 log_debug_errno(r
, "Failed to read TPM2 PCR public key, proceeding without: %m");
3789 arg_tpm2_public_key_pcr_mask
= 0;
3793 _cleanup_(tpm2_context_unrefp
) Tpm2Context
*tpm2_context
= NULL
;
3794 r
= tpm2_context_new(arg_tpm2_device
, &tpm2_context
);
3796 return log_error_errno(r
, "Failed to create TPM2 context: %m");
3798 TPM2B_PUBLIC
public;
3800 r
= tpm2_tpm2b_public_from_pem(pubkey
, pubkey_size
, &public);
3802 return log_error_errno(r
, "Could not convert public key to TPM2B_PUBLIC: %m");
3805 r
= tpm2_pcr_read_missing_values(tpm2_context
, arg_tpm2_hash_pcr_values
, arg_tpm2_n_hash_pcr_values
);
3807 return log_error_errno(r
, "Could not read pcr values: %m");
3809 uint16_t hash_pcr_bank
= 0;
3810 uint32_t hash_pcr_mask
= 0;
3811 if (arg_tpm2_n_hash_pcr_values
> 0) {
3813 r
= tpm2_pcr_values_hash_count(arg_tpm2_hash_pcr_values
, arg_tpm2_n_hash_pcr_values
, &hash_count
);
3815 return log_error_errno(r
, "Could not get hash count: %m");
3818 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "Multiple PCR banks selected.");
3820 hash_pcr_bank
= arg_tpm2_hash_pcr_values
[0].hash
;
3821 r
= tpm2_pcr_values_to_mask(arg_tpm2_hash_pcr_values
, arg_tpm2_n_hash_pcr_values
, hash_pcr_bank
, &hash_pcr_mask
);
3823 return log_error_errno(r
, "Could not get hash mask: %m");
3826 _cleanup_(tpm2_pcrlock_policy_done
) Tpm2PCRLockPolicy pcrlock_policy
= {};
3827 if (arg_tpm2_pcrlock
) {
3828 r
= tpm2_pcrlock_policy_load(arg_tpm2_pcrlock
, &pcrlock_policy
);
3833 TPM2B_DIGEST policy
= TPM2B_DIGEST_MAKE(NULL
, TPM2_SHA256_DIGEST_SIZE
);
3834 r
= tpm2_calculate_sealing_policy(
3835 arg_tpm2_hash_pcr_values
,
3836 arg_tpm2_n_hash_pcr_values
,
3837 pubkey
? &public : NULL
,
3838 /* use_pin= */ false,
3839 arg_tpm2_pcrlock
? &pcrlock_policy
: NULL
,
3842 return log_error_errno(r
, "Could not calculate sealing policy digest: %m");
3844 r
= tpm2_seal(tpm2_context
,
3845 /* seal_key_handle= */ 0,
3848 &secret
, &secret_size
,
3850 /* ret_primary_alg= */ NULL
,
3851 &srk_buf
, &srk_buf_size
);
3853 return log_error_errno(r
, "Failed to seal to TPM2: %m");
3855 base64_encoded_size
= base64mem(secret
, secret_size
, &base64_encoded
);
3856 if (base64_encoded_size
< 0)
3857 return log_error_errno(base64_encoded_size
, "Failed to base64 encode secret key: %m");
3859 r
= cryptsetup_set_minimal_pbkdf(cd
);
3861 return log_error_errno(r
, "Failed to set minimal PBKDF: %m");
3863 keyslot
= sym_crypt_keyslot_add_by_volume_key(
3869 base64_encoded_size
);
3871 return log_error_errno(keyslot
, "Failed to add new TPM2 key: %m");
3873 r
= tpm2_make_luks2_json(
3877 pubkey
, pubkey_size
,
3878 arg_tpm2_public_key_pcr_mask
,
3879 /* primary_alg= */ 0,
3881 policy
.buffer
, policy
.size
,
3882 NULL
, 0, /* no salt because tpm2_seal has no pin */
3883 srk_buf
, srk_buf_size
,
3887 return log_error_errno(r
, "Failed to prepare TPM2 JSON token object: %m");
3889 r
= cryptsetup_add_token_json(cd
, v
);
3891 return log_error_errno(r
, "Failed to add TPM2 JSON token to LUKS2 header: %m");
3893 passphrase
= base64_encoded
;
3894 passphrase_size
= strlen(base64_encoded
);
3896 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
3897 "Support for TPM2 enrollment not enabled.");
3902 r
= sym_crypt_reencrypt_init_by_passphrase(
3909 sym_crypt_get_cipher(cd
),
3910 sym_crypt_get_cipher_mode(cd
),
3913 return log_error_errno(r
, "Failed to prepare for reencryption: %m");
3915 /* crypt_reencrypt_init_by_passphrase() doesn't actually put the LUKS header at the front, we
3916 * have to do that ourselves. */
3921 r
= sym_crypt_init(&cd
, node
);
3923 return log_error_errno(r
, "Failed to allocate libcryptsetup context for %s: %m", node
);
3925 r
= sym_crypt_header_restore(cd
, CRYPT_LUKS2
, hp
);
3927 return log_error_errno(r
, "Failed to place new LUKS header at head of %s: %m", node
);
3929 reencrypt_params
.flags
&= ~CRYPT_REENCRYPT_INITIALIZE_ONLY
;
3931 r
= sym_crypt_reencrypt_init_by_passphrase(
3942 return log_error_errno(r
, "Failed to load reencryption context: %m");
3944 r
= sym_crypt_reencrypt(cd
, NULL
);
3946 return log_error_errno(r
, "Failed to encrypt %s: %m", node
);
3948 _cleanup_free_ DecryptedPartitionTarget
*t
= NULL
;
3949 _cleanup_close_
int dev_fd
= -1;
3951 r
= sym_crypt_activate_by_volume_key(
3956 arg_discard
? CRYPT_ACTIVATE_ALLOW_DISCARDS
: 0);
3958 return log_error_errno(r
, "Failed to activate LUKS superblock: %m");
3960 dev_fd
= open(vol
, O_RDWR
|O_CLOEXEC
|O_NOCTTY
);
3962 return log_error_errno(errno
, "Failed to open LUKS volume '%s': %m", vol
);
3964 if (flock(dev_fd
, LOCK_EX
) < 0)
3965 return log_error_errno(errno
, "Failed to lock '%s': %m", vol
);
3967 t
= new(DecryptedPartitionTarget
, 1);
3971 *t
= (DecryptedPartitionTarget
) {
3972 .fd
= TAKE_FD(dev_fd
),
3973 .dm_name
= TAKE_PTR(dm_name
),
3974 .volume
= TAKE_PTR(vol
),
3975 .device
= TAKE_PTR(cd
),
3978 target
->decrypted
= TAKE_PTR(t
);
3981 log_info("Successfully encrypted future partition %" PRIu64
".", p
->partno
);
3985 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
3986 "libcryptsetup is not supported or is missing required symbols, cannot encrypt: %m");
3990 static int partition_format_verity_hash(
3994 const char *data_node
) {
3996 #if HAVE_LIBCRYPTSETUP
3998 _cleanup_(partition_target_freep
) PartitionTarget
*t
= NULL
;
3999 _cleanup_(sym_crypt_freep
) struct crypt_device
*cd
= NULL
;
4000 _cleanup_free_
char *hint
= NULL
;
4005 assert(p
->verity
== VERITY_HASH
);
4011 if (PARTITION_EXISTS(p
)) /* Never format existing partitions */
4014 /* Minimized partitions will use the copy blocks logic so let's make sure to skip those here. */
4015 if (p
->copy_blocks_fd
>= 0)
4018 assert_se(dp
= p
->siblings
[VERITY_DATA
]);
4019 assert(!dp
->dropped
);
4021 (void) partition_hint(p
, node
, &hint
);
4023 r
= dlopen_cryptsetup();
4025 return log_error_errno(r
, "libcryptsetup not found, cannot setup verity: %m");
4028 r
= partition_target_prepare(context
, p
, p
->new_size
, /*need_path=*/ true, &t
);
4032 node
= partition_target_path(t
);
4035 if (p
->verity_data_block_size
== UINT64_MAX
)
4036 p
->verity_data_block_size
= context
->fs_sector_size
;
4037 if (p
->verity_hash_block_size
== UINT64_MAX
)
4038 p
->verity_hash_block_size
= context
->fs_sector_size
;
4040 r
= sym_crypt_init(&cd
, node
);
4042 return log_error_errno(r
, "Failed to allocate libcryptsetup context for %s: %m", node
);
4044 cryptsetup_enable_logging(cd
);
4046 r
= sym_crypt_format(
4047 cd
, CRYPT_VERITY
, NULL
, NULL
, SD_ID128_TO_UUID_STRING(p
->verity_uuid
), NULL
, 0,
4048 &(struct crypt_params_verity
){
4049 .data_device
= data_node
,
4050 .flags
= CRYPT_VERITY_CREATE_HASH
,
4051 .hash_name
= "sha256",
4053 .data_block_size
= p
->verity_data_block_size
,
4054 .hash_block_size
= p
->verity_hash_block_size
,
4055 .salt_size
= sizeof(p
->verity_salt
),
4056 .salt
= (const char*)p
->verity_salt
,
4059 /* libcryptsetup reports non-descriptive EIO errors for every I/O failure. Luckily, it
4060 * doesn't clobber errno so let's check for ENOSPC so we can report a better error if the
4061 * partition is too small. */
4062 if (r
== -EIO
&& errno
== ENOSPC
)
4063 return log_error_errno(errno
,
4064 "Verity hash data does not fit in partition %s with size %s",
4065 strna(hint
), FORMAT_BYTES(p
->new_size
));
4067 return log_error_errno(r
, "Failed to setup verity hash data of partition %s: %m", strna(hint
));
4071 r
= partition_target_sync(context
, p
, t
);
4076 r
= sym_crypt_get_volume_key_size(cd
);
4078 return log_error_errno(r
, "Failed to determine verity root hash size of partition %s: %m", strna(hint
));
4080 _cleanup_(iovec_done
) struct iovec rh
= {
4081 .iov_base
= malloc(r
),
4087 r
= sym_crypt_volume_key_get(cd
, CRYPT_ANY_SLOT
, (char *) rh
.iov_base
, &rh
.iov_len
, NULL
, 0);
4089 return log_error_errno(r
, "Failed to get verity root hash of partition %s: %m", strna(hint
));
4091 assert(rh
.iov_len
>= sizeof(sd_id128_t
) * 2);
4093 if (!dp
->new_uuid_is_set
) {
4094 memcpy_safe(dp
->new_uuid
.bytes
, rh
.iov_base
, sizeof(sd_id128_t
));
4095 dp
->new_uuid_is_set
= true;
4098 if (!p
->new_uuid_is_set
) {
4099 memcpy_safe(p
->new_uuid
.bytes
, (uint8_t*) rh
.iov_base
+ (rh
.iov_len
- sizeof(sd_id128_t
)), sizeof(sd_id128_t
));
4100 p
->new_uuid_is_set
= true;
4103 p
->roothash
= TAKE_STRUCT(rh
);
4107 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
), "libcryptsetup is not supported, cannot setup verity hashes: %m");
4111 static int sign_verity_roothash(
4112 const struct iovec
*roothash
,
4113 struct iovec
*ret_signature
) {
4116 _cleanup_(BIO_freep
) BIO
*rb
= NULL
;
4117 _cleanup_(PKCS7_freep
) PKCS7
*p7
= NULL
;
4118 _cleanup_free_
char *hex
= NULL
;
4119 _cleanup_free_
uint8_t *sig
= NULL
;
4123 assert(iovec_is_set(roothash
));
4124 assert(ret_signature
);
4126 hex
= hexmem(roothash
->iov_base
, roothash
->iov_len
);
4130 rb
= BIO_new_mem_buf(hex
, -1);
4134 p7
= PKCS7_sign(arg_certificate
, arg_private_key
, NULL
, rb
, PKCS7_DETACHED
|PKCS7_NOATTR
|PKCS7_BINARY
);
4136 return log_error_errno(SYNTHETIC_ERRNO(EIO
), "Failed to calculate PKCS7 signature: %s",
4137 ERR_error_string(ERR_get_error(), NULL
));
4139 sigsz
= i2d_PKCS7(p7
, &sig
);
4141 return log_error_errno(SYNTHETIC_ERRNO(EIO
), "Failed to convert PKCS7 signature to DER: %s",
4142 ERR_error_string(ERR_get_error(), NULL
));
4144 ret_signature
->iov_base
= TAKE_PTR(sig
);
4145 ret_signature
->iov_len
= sigsz
;
4149 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
), "openssl is not supported, cannot setup verity signature: %m");
4153 static int partition_format_verity_sig(Context
*context
, Partition
*p
) {
4154 _cleanup_(json_variant_unrefp
) JsonVariant
*v
= NULL
;
4155 _cleanup_(iovec_done
) struct iovec sig
= {};
4156 _cleanup_free_
char *text
= NULL
, *hint
= NULL
;
4158 uint8_t fp
[X509_FINGERPRINT_SIZE
];
4161 assert(p
->verity
== VERITY_SIG
);
4166 if (PARTITION_EXISTS(p
))
4169 (void) partition_hint(p
, context
->node
, &hint
);
4171 assert_se(hp
= p
->siblings
[VERITY_HASH
]);
4172 assert(!hp
->dropped
);
4174 assert(arg_certificate
);
4176 assert_se((whole_fd
= fdisk_get_devfd(context
->fdisk_context
)) >= 0);
4178 r
= sign_verity_roothash(&hp
->roothash
, &sig
);
4182 r
= x509_fingerprint(arg_certificate
, fp
);
4184 return log_error_errno(r
, "Unable to calculate X509 certificate fingerprint: %m");
4188 JSON_BUILD_PAIR("rootHash", JSON_BUILD_HEX(hp
->roothash
.iov_base
, hp
->roothash
.iov_len
)),
4190 "certificateFingerprint",
4191 JSON_BUILD_HEX(fp
, sizeof(fp
))
4193 JSON_BUILD_PAIR("signature", JSON_BUILD_IOVEC_BASE64(&sig
))
4197 return log_error_errno(r
, "Failed to build verity signature JSON object: %m");
4199 r
= json_variant_format(v
, 0, &text
);
4201 return log_error_errno(r
, "Failed to format verity signature JSON object: %m");
4203 if (strlen(text
)+1 > p
->new_size
)
4204 return log_error_errno(SYNTHETIC_ERRNO(E2BIG
), "Verity signature too long for partition: %m");
4206 r
= strgrowpad0(&text
, p
->new_size
);
4208 return log_error_errno(r
, "Failed to pad string to %s", FORMAT_BYTES(p
->new_size
));
4210 if (lseek(whole_fd
, p
->offset
, SEEK_SET
) < 0)
4211 return log_error_errno(errno
, "Failed to seek to partition %s offset: %m", strna(hint
));
4213 r
= loop_write(whole_fd
, text
, p
->new_size
);
4215 return log_error_errno(r
, "Failed to write verity signature to partition %s: %m", strna(hint
));
4217 if (fsync(whole_fd
) < 0)
4218 return log_error_errno(errno
, "Failed to synchronize partition %s: %m", strna(hint
));
4223 static int context_copy_blocks(Context
*context
) {
4228 /* Copy in file systems on the block level */
4230 LIST_FOREACH(partitions
, p
, context
->partitions
) {
4231 _cleanup_(partition_target_freep
) PartitionTarget
*t
= NULL
;
4233 if (p
->copy_blocks_fd
< 0)
4239 if (PARTITION_EXISTS(p
)) /* Never copy over existing partitions */
4242 if (partition_type_defer(&p
->type
))
4245 assert(p
->new_size
!= UINT64_MAX
);
4246 assert(p
->copy_blocks_size
!= UINT64_MAX
);
4247 assert(p
->new_size
>= p
->copy_blocks_size
+ (p
->encrypt
!= ENCRYPT_OFF
? LUKS2_METADATA_KEEP_FREE
: 0));
4249 r
= partition_target_prepare(context
, p
, p
->new_size
,
4250 /*need_path=*/ p
->encrypt
!= ENCRYPT_OFF
|| p
->siblings
[VERITY_HASH
],
4255 if (p
->encrypt
!= ENCRYPT_OFF
&& t
->loop
) {
4256 r
= partition_encrypt(context
, p
, t
, /* offline = */ false);
4261 log_info("Copying in '%s' (%s) on block level into future partition %" PRIu64
".",
4262 p
->copy_blocks_path
, FORMAT_BYTES(p
->copy_blocks_size
), p
->partno
);
4264 if (p
->copy_blocks_offset
!= UINT64_MAX
&& lseek(p
->copy_blocks_fd
, p
->copy_blocks_offset
, SEEK_SET
) < 0)
4265 return log_error_errno(errno
, "Failed to seek to copy blocks offset in %s: %m", p
->copy_blocks_path
);
4267 r
= copy_bytes(p
->copy_blocks_fd
, partition_target_fd(t
), p
->copy_blocks_size
, COPY_REFLINK
);
4269 return log_error_errno(r
, "Failed to copy in data from '%s': %m", p
->copy_blocks_path
);
4271 log_info("Copying in of '%s' on block level completed.", p
->copy_blocks_path
);
4273 if (p
->encrypt
!= ENCRYPT_OFF
&& !t
->loop
) {
4274 r
= partition_encrypt(context
, p
, t
, /* offline = */ true);
4279 r
= partition_target_sync(context
, p
, t
);
4283 if (p
->siblings
[VERITY_HASH
] && !partition_type_defer(&p
->siblings
[VERITY_HASH
]->type
)) {
4284 r
= partition_format_verity_hash(context
, p
->siblings
[VERITY_HASH
],
4285 /* node = */ NULL
, partition_target_path(t
));
4290 if (p
->siblings
[VERITY_SIG
] && !partition_type_defer(&p
->siblings
[VERITY_SIG
]->type
)) {
4291 r
= partition_format_verity_sig(context
, p
->siblings
[VERITY_SIG
]);
4300 static int add_exclude_path(const char *path
, Hashmap
**denylist
, DenyType type
) {
4301 _cleanup_free_
struct stat
*st
= NULL
;
4307 st
= new(struct stat
, 1);
4311 r
= chase_and_stat(path
, arg_copy_source
, CHASE_PREFIX_ROOT
, NULL
, st
);
4315 return log_error_errno(r
, "Failed to stat source file '%s/%s': %m", strempty(arg_copy_source
), path
);
4317 r
= hashmap_ensure_put(denylist
, &inode_hash_ops
, st
, INT_TO_PTR(type
));
4328 static int make_copy_files_denylist(
4335 _cleanup_hashmap_free_ Hashmap
*denylist
= NULL
;
4344 /* Always exclude the top level APIVFS and temporary directories since the contents of these
4345 * directories are almost certainly not intended to end up in an image. */
4347 NULSTR_FOREACH(s
, APIVFS_TMP_DIRS_NULSTR
) {
4348 r
= add_exclude_path(s
, &denylist
, DENY_CONTENTS
);
4353 /* Add the user configured excludes. */
4355 STRV_FOREACH(e
, p
->exclude_files_source
) {
4356 r
= add_exclude_path(*e
, &denylist
, endswith(*e
, "/") ? DENY_CONTENTS
: DENY_INODE
);
4361 STRV_FOREACH(e
, p
->exclude_files_target
) {
4362 _cleanup_free_
char *path
= NULL
;
4364 const char *s
= path_startswith(*e
, target
);
4368 path
= path_join(source
, s
);
4372 r
= add_exclude_path(path
, &denylist
, endswith(*e
, "/") ? DENY_CONTENTS
: DENY_INODE
);
4377 /* If we're populating a root partition, we don't want any files to end up under the APIVFS mount
4378 * points. While we already exclude <source>/proc, users could still do something such as
4379 * "CopyFiles=/abc:/". Now, if /abc has a proc subdirectory with files in it, those will end up in
4380 * the top level proc directory in the root partition, which we want to avoid. To deal with these
4381 * cases, whenever we're populating a root partition and the target of CopyFiles= is the root
4382 * directory of the root partition, we exclude all directories under the source that are named after
4383 * APIVFS directories or named after mount points of other partitions that are also going to be part
4386 if (p
->type
.designator
== PARTITION_ROOT
&& empty_or_root(target
)) {
4387 LIST_FOREACH(partitions
, q
, context
->partitions
) {
4388 if (q
->type
.designator
== PARTITION_ROOT
)
4391 const char *sources
= gpt_partition_type_mountpoint_nulstr(q
->type
);
4395 NULSTR_FOREACH(s
, sources
) {
4396 _cleanup_free_
char *path
= NULL
;
4398 /* Exclude only the children of partition mount points so that the nested
4399 * partition mount point itself still ends up in the upper partition. */
4401 path
= path_join(source
, s
);
4405 r
= add_exclude_path(path
, &denylist
, DENY_CONTENTS
);
4411 NULSTR_FOREACH(s
, APIVFS_TMP_DIRS_NULSTR
) {
4412 _cleanup_free_
char *path
= NULL
;
4414 path
= path_join(source
, s
);
4418 r
= add_exclude_path(path
, &denylist
, DENY_CONTENTS
);
4424 *ret
= TAKE_PTR(denylist
);
4428 static int add_subvolume_path(const char *path
, Set
**subvolumes
) {
4429 _cleanup_free_
struct stat
*st
= NULL
;
4435 st
= new(struct stat
, 1);
4439 r
= chase_and_stat(path
, arg_copy_source
, CHASE_PREFIX_ROOT
, NULL
, st
);
4443 return log_error_errno(r
, "Failed to stat source file '%s/%s': %m", strempty(arg_copy_source
), path
);
4445 r
= set_ensure_consume(subvolumes
, &inode_hash_ops
, TAKE_PTR(st
));
4452 static int make_subvolumes_set(
4458 _cleanup_set_free_ Set
*subvolumes
= NULL
;
4466 STRV_FOREACH(subvolume
, p
->subvolumes
) {
4467 _cleanup_free_
char *path
= NULL
;
4469 const char *s
= path_startswith(*subvolume
, target
);
4473 path
= path_join(source
, s
);
4477 r
= add_subvolume_path(path
, &subvolumes
);
4482 *ret
= TAKE_PTR(subvolumes
);
4486 static int do_copy_files(Context
*context
, Partition
*p
, const char *root
) {
4492 /* copy_tree_at() automatically copies the permissions of source directories to target directories if
4493 * it created them. However, the root directory is created by us, so we have to manually take care
4494 * that it is initialized. We use the first source directory targeting "/" as the metadata source for
4495 * the root directory. */
4496 STRV_FOREACH_PAIR(source
, target
, p
->copy_files
) {
4497 _cleanup_close_
int rfd
= -EBADF
, sfd
= -EBADF
;
4499 if (!path_equal(*target
, "/"))
4502 rfd
= open(root
, O_DIRECTORY
|O_CLOEXEC
|O_NOFOLLOW
);
4506 sfd
= chase_and_open(*source
, arg_copy_source
, CHASE_PREFIX_ROOT
, O_PATH
|O_DIRECTORY
|O_CLOEXEC
|O_NOCTTY
, NULL
);
4508 return log_error_errno(sfd
, "Failed to open source file '%s%s': %m", strempty(arg_copy_source
), *source
);
4510 (void) copy_xattr(sfd
, NULL
, rfd
, NULL
, COPY_ALL_XATTRS
);
4511 (void) copy_access(sfd
, rfd
);
4512 (void) copy_times(sfd
, rfd
, 0);
4517 STRV_FOREACH_PAIR(source
, target
, p
->copy_files
) {
4518 _cleanup_hashmap_free_ Hashmap
*denylist
= NULL
;
4519 _cleanup_set_free_ Set
*subvolumes_by_source_inode
= NULL
;
4520 _cleanup_close_
int sfd
= -EBADF
, pfd
= -EBADF
, tfd
= -EBADF
;
4522 r
= make_copy_files_denylist(context
, p
, *source
, *target
, &denylist
);
4526 r
= make_subvolumes_set(context
, p
, *source
, *target
, &subvolumes_by_source_inode
);
4530 sfd
= chase_and_open(*source
, arg_copy_source
, CHASE_PREFIX_ROOT
, O_CLOEXEC
|O_NOCTTY
, NULL
);
4531 if (sfd
== -ENOENT
) {
4532 log_notice_errno(sfd
, "Failed to open source file '%s%s', skipping: %m", strempty(arg_copy_source
), *source
);
4536 return log_error_errno(sfd
, "Failed to open source file '%s%s': %m", strempty(arg_copy_source
), *source
);
4538 r
= fd_verify_regular(sfd
);
4541 return log_error_errno(r
, "Failed to check type of source file '%s': %m", *source
);
4543 /* We are looking at a directory */
4544 tfd
= chase_and_open(*target
, root
, CHASE_PREFIX_ROOT
, O_RDONLY
|O_DIRECTORY
|O_CLOEXEC
, NULL
);
4546 _cleanup_free_
char *dn
= NULL
, *fn
= NULL
;
4549 return log_error_errno(tfd
, "Failed to open target directory '%s': %m", *target
);
4551 r
= path_extract_filename(*target
, &fn
);
4553 return log_error_errno(r
, "Failed to extract filename from '%s': %m", *target
);
4555 r
= path_extract_directory(*target
, &dn
);
4557 return log_error_errno(r
, "Failed to extract directory from '%s': %m", *target
);
4559 r
= mkdir_p_root(root
, dn
, UID_INVALID
, GID_INVALID
, 0755, p
->subvolumes
);
4561 return log_error_errno(r
, "Failed to create parent directory '%s': %m", dn
);
4563 pfd
= chase_and_open(dn
, root
, CHASE_PREFIX_ROOT
, O_RDONLY
|O_DIRECTORY
|O_CLOEXEC
, NULL
);
4565 return log_error_errno(pfd
, "Failed to open parent directory of target: %m");
4570 UID_INVALID
, GID_INVALID
,
4571 COPY_REFLINK
|COPY_HOLES
|COPY_MERGE
|COPY_REPLACE
|COPY_SIGINT
|COPY_HARDLINKS
|COPY_ALL_XATTRS
|COPY_GRACEFUL_WARN
|COPY_TRUNCATE
,
4572 denylist
, subvolumes_by_source_inode
);
4577 UID_INVALID
, GID_INVALID
,
4578 COPY_REFLINK
|COPY_HOLES
|COPY_MERGE
|COPY_REPLACE
|COPY_SIGINT
|COPY_HARDLINKS
|COPY_ALL_XATTRS
|COPY_GRACEFUL_WARN
|COPY_TRUNCATE
,
4579 denylist
, subvolumes_by_source_inode
);
4581 return log_error_errno(r
, "Failed to copy '%s%s' to '%s%s': %m",
4582 strempty(arg_copy_source
), *source
, strempty(root
), *target
);
4584 _cleanup_free_
char *dn
= NULL
, *fn
= NULL
;
4586 /* We are looking at a regular file */
4588 r
= path_extract_filename(*target
, &fn
);
4589 if (r
== -EADDRNOTAVAIL
|| r
== O_DIRECTORY
)
4590 return log_error_errno(SYNTHETIC_ERRNO(EISDIR
),
4591 "Target path '%s' refers to a directory, but source path '%s' refers to regular file, can't copy.", *target
, *source
);
4593 return log_error_errno(r
, "Failed to extract filename from '%s': %m", *target
);
4595 r
= path_extract_directory(*target
, &dn
);
4597 return log_error_errno(r
, "Failed to extract directory from '%s': %m", *target
);
4599 r
= mkdir_p_root(root
, dn
, UID_INVALID
, GID_INVALID
, 0755, p
->subvolumes
);
4601 return log_error_errno(r
, "Failed to create parent directory: %m");
4603 pfd
= chase_and_open(dn
, root
, CHASE_PREFIX_ROOT
, O_RDONLY
|O_DIRECTORY
|O_CLOEXEC
, NULL
);
4605 return log_error_errno(pfd
, "Failed to open parent directory of target: %m");
4607 tfd
= openat(pfd
, fn
, O_CREAT
|O_EXCL
|O_WRONLY
|O_CLOEXEC
, 0700);
4609 return log_error_errno(errno
, "Failed to create target file '%s': %m", *target
);
4611 r
= copy_bytes(sfd
, tfd
, UINT64_MAX
, COPY_REFLINK
|COPY_HOLES
|COPY_SIGINT
|COPY_TRUNCATE
);
4613 return log_error_errno(r
, "Failed to copy '%s' to '%s%s': %m", *source
, strempty(arg_copy_source
), *target
);
4615 (void) copy_xattr(sfd
, NULL
, tfd
, NULL
, COPY_ALL_XATTRS
);
4616 (void) copy_access(sfd
, tfd
);
4617 (void) copy_times(sfd
, tfd
, 0);
4624 static int do_make_directories(Partition
*p
, const char *root
) {
4630 STRV_FOREACH(d
, p
->make_directories
) {
4631 r
= mkdir_p_root(root
, *d
, UID_INVALID
, GID_INVALID
, 0755, p
->subvolumes
);
4633 return log_error_errno(r
, "Failed to create directory '%s' in file system: %m", *d
);
4639 static bool partition_needs_populate(Partition
*p
) {
4641 return !strv_isempty(p
->copy_files
) || !strv_isempty(p
->make_directories
);
4644 static int partition_populate_directory(Context
*context
, Partition
*p
, char **ret
) {
4645 _cleanup_(rm_rf_physical_and_freep
) char *root
= NULL
;
4651 log_info("Populating %s filesystem.", p
->format
);
4653 r
= var_tmp_dir(&vt
);
4655 return log_error_errno(r
, "Could not determine temporary directory: %m");
4657 r
= tempfn_random_child(vt
, "repart", &root
);
4659 return log_error_errno(r
, "Failed to generate temporary directory: %m");
4661 r
= mkdir(root
, 0755);
4663 return log_error_errno(errno
, "Failed to create temporary directory: %m");
4665 r
= do_copy_files(context
, p
, root
);
4669 r
= do_make_directories(p
, root
);
4673 log_info("Successfully populated %s filesystem.", p
->format
);
4675 *ret
= TAKE_PTR(root
);
4679 static int partition_populate_filesystem(Context
*context
, Partition
*p
, const char *node
) {
4685 log_info("Populating %s filesystem.", p
->format
);
4687 /* We copy in a child process, since we have to mount the fs for that, and we don't want that fs to
4688 * appear in the host namespace. Hence we fork a child that has its own file system namespace and
4689 * detached mount propagation. */
4691 r
= safe_fork("(sd-copy)", FORK_DEATHSIG_SIGTERM
|FORK_LOG
|FORK_WAIT
|FORK_NEW_MOUNTNS
|FORK_MOUNTNS_SLAVE
, NULL
);
4695 static const char fs
[] = "/run/systemd/mount-root";
4696 /* This is a child process with its own mount namespace and propagation to host turned off */
4698 r
= mkdir_p(fs
, 0700);
4700 log_error_errno(r
, "Failed to create mount point: %m");
4701 _exit(EXIT_FAILURE
);
4704 if (mount_nofollow_verbose(LOG_ERR
, node
, fs
, p
->format
, MS_NOATIME
|MS_NODEV
|MS_NOEXEC
|MS_NOSUID
, NULL
) < 0)
4705 _exit(EXIT_FAILURE
);
4707 if (do_copy_files(context
, p
, fs
) < 0)
4708 _exit(EXIT_FAILURE
);
4710 if (do_make_directories(p
, fs
) < 0)
4711 _exit(EXIT_FAILURE
);
4713 r
= syncfs_path(AT_FDCWD
, fs
);
4715 log_error_errno(r
, "Failed to synchronize written files: %m");
4716 _exit(EXIT_FAILURE
);
4719 _exit(EXIT_SUCCESS
);
4722 log_info("Successfully populated %s filesystem.", p
->format
);
4726 static int context_mkfs(Context
*context
) {
4731 /* Make a file system */
4733 LIST_FOREACH(partitions
, p
, context
->partitions
) {
4734 _cleanup_(rm_rf_physical_and_freep
) char *root
= NULL
;
4735 _cleanup_(partition_target_freep
) PartitionTarget
*t
= NULL
;
4736 _cleanup_strv_free_
char **extra_mkfs_options
= NULL
;
4741 if (PARTITION_EXISTS(p
)) /* Never format existing partitions */
4747 /* Minimized partitions will use the copy blocks logic so let's make sure to skip those here. */
4748 if (p
->copy_blocks_fd
>= 0)
4751 if (partition_type_defer(&p
->type
))
4754 assert(p
->offset
!= UINT64_MAX
);
4755 assert(p
->new_size
!= UINT64_MAX
);
4756 assert(p
->new_size
>= (p
->encrypt
!= ENCRYPT_OFF
? LUKS2_METADATA_KEEP_FREE
: 0));
4758 /* If we're doing encryption, we make sure we keep free space at the end which is required
4759 * for cryptsetup's offline encryption. */
4760 r
= partition_target_prepare(context
, p
,
4761 p
->new_size
- (p
->encrypt
!= ENCRYPT_OFF
? LUKS2_METADATA_KEEP_FREE
: 0),
4762 /*need_path=*/ true,
4767 if (p
->encrypt
!= ENCRYPT_OFF
&& t
->loop
) {
4768 r
= partition_target_grow(t
, p
->new_size
);
4772 r
= partition_encrypt(context
, p
, t
, /* offline = */ false);
4774 return log_error_errno(r
, "Failed to encrypt device: %m");
4777 log_info("Formatting future partition %" PRIu64
".", p
->partno
);
4779 /* If we're not writing to a loop device or if we're populating a read-only filesystem, we
4780 * have to populate using the filesystem's mkfs's --root (or equivalent) option. To do that,
4781 * we need to set up the final directory tree beforehand. */
4783 if (partition_needs_populate(p
) && (!t
->loop
|| fstype_is_ro(p
->format
))) {
4784 if (!mkfs_supports_root_option(p
->format
))
4785 return log_error_errno(SYNTHETIC_ERRNO(ENODEV
),
4786 "Loop device access is required to populate %s filesystems.",
4789 r
= partition_populate_directory(context
, p
, &root
);
4794 r
= mkfs_options_from_env("REPART", p
->format
, &extra_mkfs_options
);
4796 return log_error_errno(r
,
4797 "Failed to determine mkfs command line options for '%s': %m",
4800 r
= make_filesystem(partition_target_path(t
), p
->format
, strempty(p
->new_label
), root
,
4801 p
->fs_uuid
, arg_discard
, /* quiet = */ false,
4802 context
->fs_sector_size
, extra_mkfs_options
);
4806 log_info("Successfully formatted future partition %" PRIu64
".", p
->partno
);
4808 /* If we're writing to a loop device, we can now mount the empty filesystem and populate it. */
4809 if (partition_needs_populate(p
) && !root
) {
4812 r
= partition_populate_filesystem(context
, p
, partition_target_path(t
));
4817 if (p
->encrypt
!= ENCRYPT_OFF
&& !t
->loop
) {
4818 r
= partition_target_grow(t
, p
->new_size
);
4822 r
= partition_encrypt(context
, p
, t
, /* offline = */ true);
4824 return log_error_errno(r
, "Failed to encrypt device: %m");
4827 /* Note that we always sync explicitly here, since mkfs.fat doesn't do that on its own, and
4828 * if we don't sync before detaching a block device the in-flight sectors possibly won't hit
4831 r
= partition_target_sync(context
, p
, t
);
4835 if (p
->siblings
[VERITY_HASH
] && !partition_type_defer(&p
->siblings
[VERITY_HASH
]->type
)) {
4836 r
= partition_format_verity_hash(context
, p
->siblings
[VERITY_HASH
],
4837 /* node = */ NULL
, partition_target_path(t
));
4842 if (p
->siblings
[VERITY_SIG
] && !partition_type_defer(&p
->siblings
[VERITY_SIG
]->type
)) {
4843 r
= partition_format_verity_sig(context
, p
->siblings
[VERITY_SIG
]);
4852 static int parse_x509_certificate(const char *certificate
, size_t certificate_size
, X509
**ret
) {
4854 _cleanup_(X509_freep
) X509
*cert
= NULL
;
4855 _cleanup_(BIO_freep
) BIO
*cb
= NULL
;
4857 assert(certificate
);
4858 assert(certificate_size
> 0);
4861 cb
= BIO_new_mem_buf(certificate
, certificate_size
);
4865 cert
= PEM_read_bio_X509(cb
, NULL
, NULL
, NULL
);
4867 return log_error_errno(SYNTHETIC_ERRNO(EBADMSG
), "Failed to parse X.509 certificate: %s",
4868 ERR_error_string(ERR_get_error(), NULL
));
4871 *ret
= TAKE_PTR(cert
);
4875 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
), "openssl is not supported, cannot parse X509 certificate.");
4879 static int parse_private_key(const char *key
, size_t key_size
, EVP_PKEY
**ret
) {
4881 _cleanup_(BIO_freep
) BIO
*kb
= NULL
;
4882 _cleanup_(EVP_PKEY_freep
) EVP_PKEY
*pk
= NULL
;
4885 assert(key_size
> 0);
4888 kb
= BIO_new_mem_buf(key
, key_size
);
4892 pk
= PEM_read_bio_PrivateKey(kb
, NULL
, NULL
, NULL
);
4894 return log_error_errno(SYNTHETIC_ERRNO(EIO
), "Failed to parse PEM private key: %s",
4895 ERR_error_string(ERR_get_error(), NULL
));
4898 *ret
= TAKE_PTR(pk
);
4902 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
), "openssl is not supported, cannot parse private key.");
4906 static int partition_acquire_uuid(Context
*context
, Partition
*p
, sd_id128_t
*ret
) {
4908 sd_id128_t type_uuid
;
4910 } _packed_ plaintext
= {};
4912 uint8_t md
[SHA256_DIGEST_SIZE
];
4923 /* Calculate a good UUID for the indicated partition. We want a certain degree of reproducibility,
4924 * hence we won't generate the UUIDs randomly. Instead we use a cryptographic hash (precisely:
4925 * HMAC-SHA256) to derive them from a single seed. The seed is generally the machine ID of the
4926 * installation we are processing, but if random behaviour is desired can be random, too. We use the
4927 * seed value as key for the HMAC (since the machine ID is something we generally don't want to leak)
4928 * and the partition type as plaintext. The partition type is suffixed with a counter (only for the
4929 * second and later partition of the same type) if we have more than one partition of the same
4930 * time. Or in other words:
4933 * SEED := /etc/machine-id
4935 * If first partition instance of type TYPE_UUID:
4936 * PARTITION_UUID := HMAC-SHA256(SEED, TYPE_UUID)
4938 * For all later partition instances of type TYPE_UUID with INSTANCE being the LE64 encoded instance number:
4939 * PARTITION_UUID := HMAC-SHA256(SEED, TYPE_UUID || INSTANCE)
4942 LIST_FOREACH(partitions
, q
, context
->partitions
) {
4946 if (!sd_id128_equal(p
->type
.uuid
, q
->type
.uuid
))
4952 plaintext
.type_uuid
= p
->type
.uuid
;
4953 plaintext
.counter
= htole64(k
);
4955 hmac_sha256(context
->seed
.bytes
, sizeof(context
->seed
.bytes
),
4957 k
== 0 ? sizeof(sd_id128_t
) : sizeof(plaintext
),
4960 /* Take the first half, mark it as v4 UUID */
4961 assert_cc(sizeof(result
.md
) == sizeof(result
.id
) * 2);
4962 result
.id
= id128_make_v4_uuid(result
.id
);
4964 /* Ensure this partition UUID is actually unique, and there's no remaining partition from an earlier run? */
4965 LIST_FOREACH(partitions
, q
, context
->partitions
) {
4969 if (sd_id128_in_set(result
.id
, q
->current_uuid
, q
->new_uuid
)) {
4970 log_warning("Partition UUID calculated from seed for partition %" PRIu64
" already used, reverting to randomized UUID.", p
->partno
);
4972 r
= sd_id128_randomize(&result
.id
);
4974 return log_error_errno(r
, "Failed to generate randomized UUID: %m");
4984 static int partition_acquire_label(Context
*context
, Partition
*p
, char **ret
) {
4985 _cleanup_free_
char *label
= NULL
;
4993 prefix
= gpt_partition_type_uuid_to_string(p
->type
.uuid
);
4998 const char *ll
= label
?: prefix
;
5001 LIST_FOREACH(partitions
, q
, context
->partitions
) {
5005 if (streq_ptr(ll
, q
->current_label
) ||
5006 streq_ptr(ll
, q
->new_label
)) {
5015 label
= mfree(label
);
5016 if (asprintf(&label
, "%s-%u", prefix
, ++k
) < 0)
5021 label
= strdup(prefix
);
5026 *ret
= TAKE_PTR(label
);
5030 static int context_acquire_partition_uuids_and_labels(Context
*context
) {
5035 LIST_FOREACH(partitions
, p
, context
->partitions
) {
5038 /* Never touch foreign partitions */
5039 if (PARTITION_IS_FOREIGN(p
)) {
5040 p
->new_uuid
= p
->current_uuid
;
5042 if (p
->current_label
) {
5043 r
= free_and_strdup_warn(&p
->new_label
, strempty(p
->current_label
));
5051 if (!sd_id128_is_null(p
->current_uuid
))
5052 p
->new_uuid
= uuid
= p
->current_uuid
; /* Never change initialized UUIDs */
5053 else if (p
->new_uuid_is_set
)
5056 /* Not explicitly set by user! */
5057 r
= partition_acquire_uuid(context
, p
, &uuid
);
5061 /* The final verity hash/data UUIDs can only be determined after formatting the
5062 * verity hash partition. However, we still want to use the generated partition UUID
5063 * to derive other UUIDs to keep things unique and reproducible, so we always
5064 * generate a UUID if none is set, but we only use it as the actual partition UUID if
5065 * verity is not configured. */
5066 if (!IN_SET(p
->verity
, VERITY_DATA
, VERITY_HASH
)) {
5068 p
->new_uuid_is_set
= true;
5072 /* Calculate the UUID for the file system as HMAC-SHA256 of the string "file-system-uuid",
5073 * keyed off the partition UUID. */
5074 r
= derive_uuid(uuid
, "file-system-uuid", &p
->fs_uuid
);
5078 if (p
->encrypt
!= ENCRYPT_OFF
) {
5079 r
= derive_uuid(uuid
, "luks-uuid", &p
->luks_uuid
);
5084 /* Derive the verity salt and verity superblock UUID from the seed to keep them reproducible */
5085 if (p
->verity
== VERITY_HASH
) {
5086 derive_salt(context
->seed
, "verity-salt", p
->verity_salt
);
5088 r
= derive_uuid(context
->seed
, "verity-uuid", &p
->verity_uuid
);
5090 return log_error_errno(r
, "Failed to acquire verity uuid: %m");
5093 if (!isempty(p
->current_label
)) {
5094 /* never change initialized labels */
5095 r
= free_and_strdup_warn(&p
->new_label
, p
->current_label
);
5098 } else if (!p
->new_label
) {
5099 /* Not explicitly set by user! */
5101 r
= partition_acquire_label(context
, p
, &p
->new_label
);
5110 static int set_gpt_flags(struct fdisk_partition
*q
, uint64_t flags
) {
5111 _cleanup_free_
char *a
= NULL
;
5113 for (unsigned i
= 0; i
< sizeof(flags
) * 8; i
++) {
5114 uint64_t bit
= UINT64_C(1) << i
;
5115 char buf
[DECIMAL_STR_MAX(unsigned)+1];
5117 if (!FLAGS_SET(flags
, bit
))
5120 xsprintf(buf
, "%u", i
);
5121 if (!strextend_with_separator(&a
, ",", buf
))
5125 return fdisk_partition_set_attrs(q
, a
);
5128 static uint64_t partition_merge_flags(Partition
*p
) {
5135 if (p
->no_auto
>= 0) {
5136 if (gpt_partition_type_knows_no_auto(p
->type
))
5137 SET_FLAG(f
, SD_GPT_FLAG_NO_AUTO
, p
->no_auto
);
5139 char buffer
[SD_ID128_UUID_STRING_MAX
];
5140 log_warning("Configured NoAuto=%s for partition type '%s' that doesn't support it, ignoring.",
5142 gpt_partition_type_uuid_to_string_harder(p
->type
.uuid
, buffer
));
5146 if (p
->read_only
>= 0) {
5147 if (gpt_partition_type_knows_read_only(p
->type
))
5148 SET_FLAG(f
, SD_GPT_FLAG_READ_ONLY
, p
->read_only
);
5150 char buffer
[SD_ID128_UUID_STRING_MAX
];
5151 log_warning("Configured ReadOnly=%s for partition type '%s' that doesn't support it, ignoring.",
5152 yes_no(p
->read_only
),
5153 gpt_partition_type_uuid_to_string_harder(p
->type
.uuid
, buffer
));
5157 if (p
->growfs
>= 0) {
5158 if (gpt_partition_type_knows_growfs(p
->type
))
5159 SET_FLAG(f
, SD_GPT_FLAG_GROWFS
, p
->growfs
);
5161 char buffer
[SD_ID128_UUID_STRING_MAX
];
5162 log_warning("Configured GrowFileSystem=%s for partition type '%s' that doesn't support it, ignoring.",
5164 gpt_partition_type_uuid_to_string_harder(p
->type
.uuid
, buffer
));
5171 static int context_mangle_partitions(Context
*context
) {
5176 LIST_FOREACH(partitions
, p
, context
->partitions
) {
5180 if (partition_type_defer(&p
->type
))
5183 assert(p
->new_size
!= UINT64_MAX
);
5184 assert(p
->offset
!= UINT64_MAX
);
5185 assert(p
->partno
!= UINT64_MAX
);
5187 if (PARTITION_EXISTS(p
)) {
5188 bool changed
= false;
5190 assert(p
->current_partition
);
5192 if (p
->new_size
!= p
->current_size
) {
5193 assert(p
->new_size
>= p
->current_size
);
5194 assert(p
->new_size
% context
->sector_size
== 0);
5196 r
= fdisk_partition_size_explicit(p
->current_partition
, true);
5198 return log_error_errno(r
, "Failed to enable explicit sizing: %m");
5200 r
= fdisk_partition_set_size(p
->current_partition
, p
->new_size
/ context
->sector_size
);
5202 return log_error_errno(r
, "Failed to grow partition: %m");
5204 log_info("Growing existing partition %" PRIu64
".", p
->partno
);
5208 if (!sd_id128_equal(p
->new_uuid
, p
->current_uuid
)) {
5209 r
= fdisk_partition_set_uuid(p
->current_partition
, SD_ID128_TO_UUID_STRING(p
->new_uuid
));
5211 return log_error_errno(r
, "Failed to set partition UUID: %m");
5213 log_info("Initializing UUID of existing partition %" PRIu64
".", p
->partno
);
5217 if (!streq_ptr(p
->new_label
, p
->current_label
)) {
5218 r
= fdisk_partition_set_name(p
->current_partition
, strempty(p
->new_label
));
5220 return log_error_errno(r
, "Failed to set partition label: %m");
5222 log_info("Setting partition label of existing partition %" PRIu64
".", p
->partno
);
5227 assert(!PARTITION_IS_FOREIGN(p
)); /* never touch foreign partitions */
5229 r
= fdisk_set_partition(context
->fdisk_context
, p
->partno
, p
->current_partition
);
5231 return log_error_errno(r
, "Failed to update partition: %m");
5234 _cleanup_(fdisk_unref_partitionp
) struct fdisk_partition
*q
= NULL
;
5235 _cleanup_(fdisk_unref_parttypep
) struct fdisk_parttype
*t
= NULL
;
5237 assert(!p
->new_partition
);
5238 assert(p
->offset
% context
->sector_size
== 0);
5239 assert(p
->new_size
% context
->sector_size
== 0);
5240 assert(p
->new_label
);
5242 t
= fdisk_new_parttype();
5246 r
= fdisk_parttype_set_typestr(t
, SD_ID128_TO_UUID_STRING(p
->type
.uuid
));
5248 return log_error_errno(r
, "Failed to initialize partition type: %m");
5250 q
= fdisk_new_partition();
5254 r
= fdisk_partition_set_type(q
, t
);
5256 return log_error_errno(r
, "Failed to set partition type: %m");
5258 r
= fdisk_partition_size_explicit(q
, true);
5260 return log_error_errno(r
, "Failed to enable explicit sizing: %m");
5262 r
= fdisk_partition_set_start(q
, p
->offset
/ context
->sector_size
);
5264 return log_error_errno(r
, "Failed to position partition: %m");
5266 r
= fdisk_partition_set_size(q
, p
->new_size
/ context
->sector_size
);
5268 return log_error_errno(r
, "Failed to grow partition: %m");
5270 r
= fdisk_partition_set_partno(q
, p
->partno
);
5272 return log_error_errno(r
, "Failed to set partition number: %m");
5274 r
= fdisk_partition_set_uuid(q
, SD_ID128_TO_UUID_STRING(p
->new_uuid
));
5276 return log_error_errno(r
, "Failed to set partition UUID: %m");
5278 r
= fdisk_partition_set_name(q
, strempty(p
->new_label
));
5280 return log_error_errno(r
, "Failed to set partition label: %m");
5282 /* Merge the no auto + read only + growfs setting with the literal flags, and set them for the partition */
5283 r
= set_gpt_flags(q
, partition_merge_flags(p
));
5285 return log_error_errno(r
, "Failed to set GPT partition flags: %m");
5287 log_info("Adding new partition %" PRIu64
" to partition table.", p
->partno
);
5289 r
= fdisk_add_partition(context
->fdisk_context
, q
, NULL
);
5291 return log_error_errno(r
, "Failed to add partition: %m");
5293 assert(!p
->new_partition
);
5294 p
->new_partition
= TAKE_PTR(q
);
5301 static int split_name_printf(Partition
*p
, char **ret
) {
5304 const Specifier table
[] = {
5305 { 't', specifier_string
, GPT_PARTITION_TYPE_UUID_TO_STRING_HARDER(p
->type
.uuid
) },
5306 { 'T', specifier_id128
, &p
->type
.uuid
},
5307 { 'U', specifier_id128
, &p
->new_uuid
},
5308 { 'n', specifier_uint64
, &p
->partno
},
5310 COMMON_SYSTEM_SPECIFIERS
,
5314 return specifier_printf(p
->split_name_format
, NAME_MAX
, table
, arg_root
, p
, ret
);
5317 static int split_node(const char *node
, char **ret_base
, char **ret_ext
) {
5318 _cleanup_free_
char *base
= NULL
, *ext
= NULL
;
5326 r
= path_extract_filename(node
, &base
);
5327 if (r
== O_DIRECTORY
|| r
== -EADDRNOTAVAIL
)
5328 return log_error_errno(r
, "Device node %s cannot be a directory", node
);
5330 return log_error_errno(r
, "Failed to extract filename from %s: %m", node
);
5332 e
= endswith(base
, ".raw");
5341 *ret_base
= TAKE_PTR(base
);
5342 *ret_ext
= TAKE_PTR(ext
);
5347 static int split_name_resolve(Context
*context
) {
5348 _cleanup_free_
char *parent
= NULL
, *base
= NULL
, *ext
= NULL
;
5353 r
= path_extract_directory(context
->node
, &parent
);
5354 if (r
< 0 && r
!= -EDESTADDRREQ
)
5355 return log_error_errno(r
, "Failed to extract directory from %s: %m", context
->node
);
5357 r
= split_node(context
->node
, &base
, &ext
);
5361 LIST_FOREACH(partitions
, p
, context
->partitions
) {
5362 _cleanup_free_
char *resolved
= NULL
;
5367 if (!p
->split_name_format
)
5370 r
= split_name_printf(p
, &resolved
);
5372 return log_error_errno(r
, "Failed to resolve specifiers in %s: %m", p
->split_name_format
);
5375 p
->split_path
= strjoin(parent
, "/", base
, ".", resolved
, ext
);
5377 p
->split_path
= strjoin(base
, ".", resolved
, ext
);
5382 LIST_FOREACH(partitions
, p
, context
->partitions
) {
5386 LIST_FOREACH(partitions
, q
, context
->partitions
) {
5393 if (!streq(p
->split_path
, q
->split_path
))
5396 return log_error_errno(SYNTHETIC_ERRNO(ENOTUNIQ
),
5397 "%s and %s have the same resolved split name \"%s\", refusing",
5398 p
->definition_path
, q
->definition_path
, p
->split_path
);
5405 static int context_split(Context
*context
) {
5413 /* We can't do resolution earlier because the partition UUIDs for verity partitions are only filled
5414 * in after they've been generated. */
5416 r
= split_name_resolve(context
);
5420 LIST_FOREACH(partitions
, p
, context
->partitions
) {
5421 _cleanup_close_
int fdt
= -EBADF
;
5429 if (partition_type_defer(&p
->type
))
5432 fdt
= open(p
->split_path
, O_WRONLY
|O_NOCTTY
|O_CLOEXEC
|O_NOFOLLOW
|O_CREAT
|O_EXCL
, 0666);
5434 return log_error_errno(fdt
, "Failed to open split partition file %s: %m", p
->split_path
);
5437 assert_se((fd
= fdisk_get_devfd(context
->fdisk_context
)) >= 0);
5439 if (lseek(fd
, p
->offset
, SEEK_SET
) < 0)
5440 return log_error_errno(errno
, "Failed to seek to partition offset: %m");
5442 r
= copy_bytes(fd
, fdt
, p
->new_size
, COPY_REFLINK
|COPY_HOLES
|COPY_TRUNCATE
);
5444 return log_error_errno(r
, "Failed to copy to split partition %s: %m", p
->split_path
);
5450 static int context_write_partition_table(Context
*context
) {
5451 _cleanup_(fdisk_unref_tablep
) struct fdisk_table
*original_table
= NULL
;
5456 if (!context
->from_scratch
&& !context_changed(context
)) {
5457 log_info("No changes.");
5462 log_notice("Refusing to repartition, please re-run with --dry-run=no.");
5466 log_info("Applying changes.");
5468 if (context
->from_scratch
&& arg_empty
!= EMPTY_CREATE
) {
5469 /* Erase everything if we operate from scratch, except if the image was just created anyway, and thus is definitely empty. */
5470 r
= context_wipe_range(context
, 0, context
->total
);
5474 log_info("Wiped block device.");
5477 r
= context_discard_range(context
, 0, context
->total
);
5478 if (r
== -EOPNOTSUPP
)
5479 log_info("Storage does not support discard, not discarding entire block device data.");
5481 return log_error_errno(r
, "Failed to discard entire block device: %m");
5483 log_info("Discarded entire block device.");
5487 r
= fdisk_get_partitions(context
->fdisk_context
, &original_table
);
5489 return log_error_errno(r
, "Failed to acquire partition table: %m");
5491 /* Wipe fs signatures and discard sectors where the new partitions are going to be placed and in the
5492 * gaps between partitions, just to be sure. */
5493 r
= context_wipe_and_discard(context
);
5497 r
= context_copy_blocks(context
);
5501 r
= context_mkfs(context
);
5505 r
= context_mangle_partitions(context
);
5509 log_info("Writing new partition table.");
5511 r
= fdisk_write_disklabel(context
->fdisk_context
);
5513 return log_error_errno(r
, "Failed to write partition table: %m");
5515 capable
= blockdev_partscan_enabled(fdisk_get_devfd(context
->fdisk_context
));
5516 if (capable
== -ENOTBLK
)
5517 log_debug("Not telling kernel to reread partition table, since we are not operating on a block device.");
5518 else if (capable
< 0)
5519 return log_error_errno(capable
, "Failed to check if block device supports partition scanning: %m");
5520 else if (capable
> 0) {
5521 log_info("Telling kernel to reread partition table.");
5523 if (context
->from_scratch
)
5524 r
= fdisk_reread_partition_table(context
->fdisk_context
);
5526 r
= fdisk_reread_changes(context
->fdisk_context
, original_table
);
5528 return log_error_errno(r
, "Failed to reread partition table: %m");
5530 log_notice("Not telling kernel to reread partition table, because selected image does not support kernel partition block devices.");
5532 log_info("All done.");
5537 static int context_read_seed(Context
*context
, const char *root
) {
5542 if (!sd_id128_is_null(context
->seed
))
5545 if (!arg_randomize
) {
5546 r
= id128_get_machine(root
, &context
->seed
);
5550 if (!ERRNO_IS_MACHINE_ID_UNSET(r
))
5551 return log_error_errno(r
, "Failed to parse machine ID of image: %m");
5553 log_info("No machine ID set, using randomized partition UUIDs.");
5556 r
= sd_id128_randomize(&context
->seed
);
5558 return log_error_errno(r
, "Failed to generate randomized seed: %m");
5563 static int context_factory_reset(Context
*context
) {
5569 if (arg_factory_reset
<= 0)
5572 if (context
->from_scratch
) /* Nothing to reset if we start from scratch */
5576 log_notice("Refusing to factory reset, please re-run with --dry-run=no.");
5580 log_info("Applying factory reset.");
5582 LIST_FOREACH(partitions
, p
, context
->partitions
) {
5584 if (!p
->factory_reset
|| !PARTITION_EXISTS(p
))
5587 assert(p
->partno
!= UINT64_MAX
);
5589 log_info("Removing partition %" PRIu64
" for factory reset.", p
->partno
);
5591 r
= fdisk_delete_partition(context
->fdisk_context
, p
->partno
);
5593 return log_error_errno(r
, "Failed to remove partition %" PRIu64
": %m", p
->partno
);
5599 log_info("Factory reset requested, but no partitions to delete found.");
5603 r
= fdisk_write_disklabel(context
->fdisk_context
);
5605 return log_error_errno(r
, "Failed to write disk label: %m");
5607 log_info("Successfully deleted %zu partitions.", n
);
5611 static int context_can_factory_reset(Context
*context
) {
5614 LIST_FOREACH(partitions
, p
, context
->partitions
)
5615 if (p
->factory_reset
&& PARTITION_EXISTS(p
))
5621 static int resolve_copy_blocks_auto_candidate(
5622 dev_t partition_devno
,
5623 GptPartitionType partition_type
,
5624 dev_t restrict_devno
,
5625 sd_id128_t
*ret_uuid
) {
5627 _cleanup_(blkid_free_probep
) blkid_probe b
= NULL
;
5628 _cleanup_close_
int fd
= -EBADF
;
5629 _cleanup_free_
char *p
= NULL
;
5630 const char *pttype
, *t
;
5631 sd_id128_t pt_parsed
, u
;
5637 /* Checks if the specified partition has the specified GPT type UUID, and is located on the specified
5638 * 'restrict_devno' device. The type check is particularly relevant if we have Verity volume which is
5639 * backed by two separate partitions: the data and the hash partitions, and we need to find the right
5640 * one of the two. */
5642 r
= block_get_whole_disk(partition_devno
, &whole_devno
);
5644 return log_error_errno(
5646 "Unable to determine containing block device of partition %u:%u: %m",
5647 major(partition_devno
), minor(partition_devno
));
5649 if (restrict_devno
!= (dev_t
) -1 &&
5650 restrict_devno
!= whole_devno
)
5651 return log_error_errno(
5652 SYNTHETIC_ERRNO(EPERM
),
5653 "Partition %u:%u is located outside of block device %u:%u, refusing.",
5654 major(partition_devno
), minor(partition_devno
),
5655 major(restrict_devno
), minor(restrict_devno
));
5657 fd
= r
= device_open_from_devnum(S_IFBLK
, whole_devno
, O_RDONLY
|O_CLOEXEC
|O_NONBLOCK
, &p
);
5659 return log_error_errno(r
, "Failed to open block device " DEVNUM_FORMAT_STR
": %m",
5660 DEVNUM_FORMAT_VAL(whole_devno
));
5662 b
= blkid_new_probe();
5667 r
= blkid_probe_set_device(b
, fd
, 0, 0);
5669 return log_error_errno(errno_or_else(ENOMEM
), "Failed to open block device '%s': %m", p
);
5671 (void) blkid_probe_enable_partitions(b
, 1);
5672 (void) blkid_probe_set_partitions_flags(b
, BLKID_PARTS_ENTRY_DETAILS
);
5675 r
= blkid_do_safeprobe(b
);
5676 if (r
== _BLKID_SAFEPROBE_ERROR
)
5677 return log_error_errno(errno_or_else(EIO
), "Unable to probe for partition table of '%s': %m", p
);
5678 if (IN_SET(r
, _BLKID_SAFEPROBE_AMBIGUOUS
, _BLKID_SAFEPROBE_NOT_FOUND
)) {
5679 log_debug("Didn't find partition table on block device '%s'.", p
);
5683 assert(r
== _BLKID_SAFEPROBE_FOUND
);
5685 (void) blkid_probe_lookup_value(b
, "PTTYPE", &pttype
, NULL
);
5686 if (!streq_ptr(pttype
, "gpt")) {
5687 log_debug("Didn't find a GPT partition table on '%s'.", p
);
5692 pl
= blkid_probe_get_partitions(b
);
5694 return log_error_errno(errno_or_else(EIO
), "Unable read partition table of '%s': %m", p
);
5696 pp
= blkid_partlist_devno_to_partition(pl
, partition_devno
);
5698 log_debug("Partition %u:%u has no matching partition table entry on '%s'.",
5699 major(partition_devno
), minor(partition_devno
), p
);
5703 t
= blkid_partition_get_type_string(pp
);
5705 log_debug("Partition %u:%u has no type on '%s'.",
5706 major(partition_devno
), minor(partition_devno
), p
);
5710 r
= sd_id128_from_string(t
, &pt_parsed
);
5712 log_debug_errno(r
, "Failed to parse partition type \"%s\": %m", t
);
5716 if (!sd_id128_equal(pt_parsed
, partition_type
.uuid
)) {
5717 log_debug("Partition %u:%u has non-matching partition type " SD_ID128_FORMAT_STR
" (needed: " SD_ID128_FORMAT_STR
"), ignoring.",
5718 major(partition_devno
), minor(partition_devno
),
5719 SD_ID128_FORMAT_VAL(pt_parsed
), SD_ID128_FORMAT_VAL(partition_type
.uuid
));
5723 r
= blkid_partition_get_uuid_id128(pp
, &u
);
5725 log_debug_errno(r
, "Partition " DEVNUM_FORMAT_STR
" has no UUID.", DEVNUM_FORMAT_VAL(partition_devno
));
5729 log_debug_errno(r
, "Failed to read partition UUID of " DEVNUM_FORMAT_STR
": %m", DEVNUM_FORMAT_VAL(partition_devno
));
5733 log_debug("Automatically found partition " DEVNUM_FORMAT_STR
" of right type " SD_ID128_FORMAT_STR
".",
5734 DEVNUM_FORMAT_VAL(partition_devno
),
5735 SD_ID128_FORMAT_VAL(pt_parsed
));
5743 static int find_backing_devno(
5748 _cleanup_free_
char *resolved
= NULL
;
5753 r
= chase(path
, root
, CHASE_PREFIX_ROOT
, &resolved
, NULL
);
5757 r
= path_is_mount_point(resolved
, NULL
, 0);
5760 if (r
== 0) /* Not a mount point, then it's not a partition of its own, let's not automatically use it. */
5763 r
= get_block_device(resolved
, ret
);
5766 if (r
== 0) /* Not backed by physical file system, we can't use this */
5772 static int resolve_copy_blocks_auto(
5773 GptPartitionType type
,
5775 dev_t restrict_devno
,
5777 sd_id128_t
*ret_uuid
) {
5779 const char *try1
= NULL
, *try2
= NULL
;
5780 char p
[SYS_BLOCK_PATH_MAX("/slaves")];
5781 _cleanup_closedir_
DIR *d
= NULL
;
5782 sd_id128_t found_uuid
= SD_ID128_NULL
;
5783 dev_t devno
, found
= 0;
5786 /* Enforce some security restrictions: CopyBlocks=auto should not be an avenue to get outside of the
5787 * --root=/--image= confinement. Specifically, refuse CopyBlocks= in combination with --root= at all,
5788 * and restrict block device references in the --image= case to loopback block device we set up.
5790 * restrict_devno contain the dev_t of the loop back device we operate on in case of --image=, and
5791 * thus declares which device (and its partition subdevices) we shall limit access to. If
5792 * restrict_devno is zero no device probing access shall be allowed at all (used for --root=) and if
5793 * it is (dev_t) -1 then free access shall be allowed (if neither switch is used). */
5795 if (restrict_devno
== 0)
5796 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
5797 "Automatic discovery of backing block devices not permitted in --root= mode, refusing.");
5799 /* Handles CopyBlocks=auto, and finds the right source partition to copy from. We look for matching
5800 * partitions in the host, using the appropriate directory as key and ensuring that the partition
5803 if (type
.designator
== PARTITION_ROOT
)
5805 else if (type
.designator
== PARTITION_USR
)
5807 else if (type
.designator
== PARTITION_ROOT_VERITY
)
5809 else if (type
.designator
== PARTITION_USR_VERITY
)
5811 else if (type
.designator
== PARTITION_ESP
) {
5814 } else if (type
.designator
== PARTITION_XBOOTLDR
)
5817 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
5818 "Partition type " SD_ID128_FORMAT_STR
" not supported from automatic source block device discovery.",
5819 SD_ID128_FORMAT_VAL(type
.uuid
));
5821 r
= find_backing_devno(try1
, root
, &devno
);
5822 if (r
== -ENOENT
&& try2
)
5823 r
= find_backing_devno(try2
, root
, &devno
);
5825 return log_error_errno(r
, "Failed to resolve automatic CopyBlocks= path for partition type " SD_ID128_FORMAT_STR
", sorry: %m",
5826 SD_ID128_FORMAT_VAL(type
.uuid
));
5828 xsprintf_sys_block_path(p
, "/slaves", devno
);
5834 _cleanup_free_
char *q
= NULL
, *t
= NULL
;
5839 de
= readdir_no_dot(d
);
5842 return log_error_errno(errno
, "Failed to read directory '%s': %m", p
);
5847 if (!IN_SET(de
->d_type
, DT_LNK
, DT_UNKNOWN
))
5850 q
= path_join(p
, de
->d_name
, "/dev");
5854 r
= read_one_line_file(q
, &t
);
5856 return log_error_errno(r
, "Failed to read %s: %m", q
);
5858 r
= parse_devnum(t
, &sl
);
5860 log_debug_errno(r
, "Failed to parse %s, ignoring: %m", q
);
5863 if (major(sl
) == 0) {
5864 log_debug("Device backing %s is special, ignoring.", q
);
5868 r
= resolve_copy_blocks_auto_candidate(sl
, type
, restrict_devno
, &u
);
5872 /* We found a matching one! */
5874 return log_error_errno(SYNTHETIC_ERRNO(ENOTUNIQ
),
5875 "Multiple matching partitions found, refusing.");
5881 } else if (errno
!= ENOENT
)
5882 return log_error_errno(errno
, "Failed open %s: %m", p
);
5884 r
= resolve_copy_blocks_auto_candidate(devno
, type
, restrict_devno
, &found_uuid
);
5892 return log_error_errno(SYNTHETIC_ERRNO(ENXIO
),
5893 "Unable to automatically discover suitable partition to copy blocks from.");
5899 *ret_uuid
= found_uuid
;
5904 static int context_open_copy_block_paths(
5906 dev_t restrict_devno
) {
5912 LIST_FOREACH(partitions
, p
, context
->partitions
) {
5913 _cleanup_close_
int source_fd
= -EBADF
;
5914 _cleanup_free_
char *opened
= NULL
;
5915 sd_id128_t uuid
= SD_ID128_NULL
;
5919 if (p
->copy_blocks_fd
>= 0)
5922 assert(p
->copy_blocks_size
== UINT64_MAX
);
5924 if (PARTITION_EXISTS(p
)) /* Never copy over partitions that already exist! */
5927 if (p
->copy_blocks_path
) {
5929 source_fd
= chase_and_open(p
->copy_blocks_path
, p
->copy_blocks_root
, CHASE_PREFIX_ROOT
, O_RDONLY
|O_CLOEXEC
|O_NONBLOCK
, &opened
);
5931 return log_error_errno(source_fd
, "Failed to open '%s': %m", p
->copy_blocks_path
);
5933 if (fstat(source_fd
, &st
) < 0)
5934 return log_error_errno(errno
, "Failed to stat block copy file '%s': %m", opened
);
5936 if (!S_ISREG(st
.st_mode
) && restrict_devno
!= (dev_t
) -1)
5937 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
5938 "Copying from block device node is not permitted in --image=/--root= mode, refusing.");
5940 } else if (p
->copy_blocks_auto
) {
5941 dev_t devno
= 0; /* Fake initialization to appease gcc. */
5943 r
= resolve_copy_blocks_auto(p
->type
, p
->copy_blocks_root
, restrict_devno
, &devno
, &uuid
);
5948 source_fd
= r
= device_open_from_devnum(S_IFBLK
, devno
, O_RDONLY
|O_CLOEXEC
|O_NONBLOCK
, &opened
);
5950 return log_error_errno(r
, "Failed to open automatically determined source block copy device " DEVNUM_FORMAT_STR
": %m",
5951 DEVNUM_FORMAT_VAL(devno
));
5953 if (fstat(source_fd
, &st
) < 0)
5954 return log_error_errno(errno
, "Failed to stat block copy file '%s': %m", opened
);
5958 if (S_ISDIR(st
.st_mode
)) {
5959 _cleanup_free_
char *bdev
= NULL
;
5962 /* If the file is a directory, automatically find the backing block device */
5964 if (major(st
.st_dev
) != 0)
5967 /* Special support for btrfs */
5968 r
= btrfs_get_block_device_fd(source_fd
, &devt
);
5970 return btrfs_log_dev_root(LOG_ERR
, r
, opened
);
5972 return log_error_errno(r
, "Unable to determine backing block device of '%s': %m", opened
);
5975 safe_close(source_fd
);
5977 source_fd
= r
= device_open_from_devnum(S_IFBLK
, devt
, O_RDONLY
|O_CLOEXEC
|O_NONBLOCK
, &bdev
);
5979 return log_error_errno(r
, "Failed to open block device backing '%s': %m", opened
);
5981 if (fstat(source_fd
, &st
) < 0)
5982 return log_error_errno(errno
, "Failed to stat block device '%s': %m", bdev
);
5985 if (S_ISREG(st
.st_mode
))
5987 else if (S_ISBLK(st
.st_mode
)) {
5988 if (ioctl(source_fd
, BLKGETSIZE64
, &size
) != 0)
5989 return log_error_errno(errno
, "Failed to determine size of block device to copy from: %m");
5991 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "Specified path to copy blocks from '%s' is not a regular file, block device or directory, refusing: %m", opened
);
5994 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "File to copy bytes from '%s' has zero size, refusing.", opened
);
5995 if (size
% 512 != 0)
5996 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "File to copy bytes from '%s' has size that is not multiple of 512, refusing.", opened
);
5998 p
->copy_blocks_fd
= TAKE_FD(source_fd
);
5999 p
->copy_blocks_size
= size
;
6001 free_and_replace(p
->copy_blocks_path
, opened
);
6003 /* When copying from an existing partition copy that partitions UUID if none is configured explicitly */
6004 if (!p
->new_uuid_is_set
&& !sd_id128_is_null(uuid
)) {
6006 p
->new_uuid_is_set
= true;
6013 static int fd_apparent_size(int fd
, uint64_t *ret
) {
6020 initial
= lseek(fd
, 0, SEEK_CUR
);
6022 return log_error_errno(errno
, "Failed to get file offset: %m");
6024 for (off_t off
= 0;;) {
6027 r
= lseek(fd
, off
, SEEK_DATA
);
6028 if (r
< 0 && errno
== ENXIO
)
6029 /* If errno == ENXIO, that means we've reached the final hole of the file and
6030 * that hole isn't followed by more data. */
6033 return log_error_errno(errno
, "Failed to seek data in file from offset %"PRIi64
": %m", off
);
6035 off
= r
; /* Set the offset to the start of the data segment. */
6037 /* After copying a potential hole, find the end of the data segment by looking for
6038 * the next hole. If we get ENXIO, we're at EOF. */
6039 r
= lseek(fd
, off
, SEEK_HOLE
);
6043 return log_error_errno(errno
, "Failed to seek hole in file from offset %"PRIi64
": %m", off
);
6050 if (lseek(fd
, initial
, SEEK_SET
) < 0)
6051 return log_error_errno(errno
, "Failed to reset file offset: %m");
6058 static int context_minimize(Context
*context
) {
6059 const char *vt
= NULL
;
6064 LIST_FOREACH(partitions
, p
, context
->partitions
) {
6065 _cleanup_(rm_rf_physical_and_freep
) char *root
= NULL
;
6066 _cleanup_(unlink_and_freep
) char *temp
= NULL
;
6067 _cleanup_(loop_device_unrefp
) LoopDevice
*d
= NULL
;
6068 _cleanup_strv_free_
char **extra_mkfs_options
= NULL
;
6069 _cleanup_close_
int fd
= -EBADF
;
6070 _cleanup_free_
char *hint
= NULL
;
6078 if (PARTITION_EXISTS(p
)) /* Never format existing partitions */
6084 if (p
->copy_blocks_fd
>= 0)
6087 if (p
->minimize
== MINIMIZE_OFF
)
6090 if (!partition_needs_populate(p
))
6093 assert(!p
->copy_blocks_path
);
6095 (void) partition_hint(p
, context
->node
, &hint
);
6097 log_info("Pre-populating %s filesystem of partition %s twice to calculate minimal partition size",
6098 p
->format
, strna(hint
));
6101 r
= var_tmp_dir(&vt
);
6103 return log_error_errno(r
, "Could not determine temporary directory: %m");
6106 r
= tempfn_random_child(vt
, "repart", &temp
);
6108 return log_error_errno(r
, "Failed to generate temporary file path: %m");
6110 if (fstype_is_ro(p
->format
))
6111 fs_uuid
= p
->fs_uuid
;
6113 fd
= open(temp
, O_CREAT
|O_EXCL
|O_CLOEXEC
|O_RDWR
|O_NOCTTY
, 0600);
6115 return log_error_errno(errno
, "Failed to open temporary file %s: %m", temp
);
6117 /* This may seem huge but it will be created sparse so it doesn't take up any space
6118 * on disk until written to. */
6119 if (ftruncate(fd
, 1024ULL * 1024ULL * 1024ULL * 1024ULL) < 0)
6120 return log_error_errno(errno
, "Failed to truncate temporary file to %s: %m",
6121 FORMAT_BYTES(1024ULL * 1024ULL * 1024ULL * 1024ULL));
6123 if (arg_offline
<= 0) {
6124 r
= loop_device_make(fd
, O_RDWR
, 0, UINT64_MAX
, context
->sector_size
, 0, LOCK_EX
, &d
);
6125 if (r
< 0 && (arg_offline
== 0 || (r
!= -ENOENT
&& !ERRNO_IS_PRIVILEGE(r
)) || !strv_isempty(p
->subvolumes
)))
6126 return log_error_errno(r
, "Failed to make loopback device of %s: %m", temp
);
6129 /* We're going to populate this filesystem twice so use a random UUID the first time
6130 * to avoid UUID conflicts. */
6131 r
= sd_id128_randomize(&fs_uuid
);
6136 if (!d
|| fstype_is_ro(p
->format
)) {
6137 if (!mkfs_supports_root_option(p
->format
))
6138 return log_error_errno(SYNTHETIC_ERRNO(ENODEV
),
6139 "Loop device access is required to populate %s filesystems",
6142 r
= partition_populate_directory(context
, p
, &root
);
6147 r
= mkfs_options_from_env("REPART", p
->format
, &extra_mkfs_options
);
6149 return log_error_errno(r
,
6150 "Failed to determine mkfs command line options for '%s': %m",
6153 r
= make_filesystem(d
? d
->node
: temp
,
6155 strempty(p
->new_label
),
6158 arg_discard
, /* quiet = */ false,
6159 context
->fs_sector_size
,
6160 extra_mkfs_options
);
6164 /* Read-only filesystems are minimal from the first try because they create and size the
6165 * loopback file for us. */
6166 if (fstype_is_ro(p
->format
)) {
6169 fd
= open(temp
, O_RDONLY
|O_CLOEXEC
|O_NONBLOCK
);
6171 return log_error_errno(errno
, "Failed to open temporary file %s: %m", temp
);
6173 if (fstat(fd
, &st
) < 0)
6174 return log_error_errno(errno
, "Failed to stat temporary file: %m");
6176 log_info("Minimal partition size of %s filesystem of partition %s is %s",
6177 p
->format
, strna(hint
), FORMAT_BYTES(st
.st_size
));
6179 p
->copy_blocks_path
= TAKE_PTR(temp
);
6180 p
->copy_blocks_path_is_our_file
= true;
6181 p
->copy_blocks_fd
= TAKE_FD(fd
);
6182 p
->copy_blocks_size
= st
.st_size
;
6189 r
= partition_populate_filesystem(context
, p
, d
->node
);
6194 /* Other filesystems need to be provided with a pre-sized loopback file and will adapt to
6195 * fully occupy it. Because we gave the filesystem a 1T sparse file, we need to shrink the
6196 * filesystem down to a reasonable size again to fit it in the disk image. While there are
6197 * some filesystems that support shrinking, it doesn't always work properly (e.g. shrinking
6198 * btrfs gives us a 2.0G filesystem regardless of what we put in it). Instead, let's populate
6199 * the filesystem again, but this time, instead of providing the filesystem with a 1T sparse
6200 * loopback file, let's size the loopback file based on the actual data used by the
6201 * filesystem in the sparse file after the first attempt. This should be a good guess of the
6202 * minimal amount of space needed in the filesystem to fit all the required data.
6204 r
= fd_apparent_size(fd
, &fsz
);
6208 /* Massage the size a bit because just going by actual data used in the sparse file isn't
6210 uint64_t heuristic
= streq(p
->format
, "xfs") ? fsz
: fsz
/ 2;
6211 fsz
= round_up_size(fsz
+ heuristic
, context
->grain_size
);
6212 if (minimal_size_by_fs_name(p
->format
) != UINT64_MAX
)
6213 fsz
= MAX(minimal_size_by_fs_name(p
->format
), fsz
);
6215 log_info("Minimal partition size of %s filesystem of partition %s is %s",
6216 p
->format
, strna(hint
), FORMAT_BYTES(fsz
));
6218 d
= loop_device_unref(d
);
6220 /* Erase the previous filesystem first. */
6221 if (ftruncate(fd
, 0))
6222 return log_error_errno(errno
, "Failed to erase temporary file: %m");
6224 if (ftruncate(fd
, fsz
))
6225 return log_error_errno(errno
, "Failed to truncate temporary file to %s: %m", FORMAT_BYTES(fsz
));
6227 if (arg_offline
<= 0) {
6228 r
= loop_device_make(fd
, O_RDWR
, 0, UINT64_MAX
, context
->sector_size
, 0, LOCK_EX
, &d
);
6229 if (r
< 0 && (arg_offline
== 0 || (r
!= -ENOENT
&& !ERRNO_IS_PRIVILEGE(r
)) || !strv_isempty(p
->subvolumes
)))
6230 return log_error_errno(r
, "Failed to make loopback device of %s: %m", temp
);
6233 r
= make_filesystem(d
? d
->node
: temp
,
6235 strempty(p
->new_label
),
6239 /* quiet = */ false,
6240 context
->fs_sector_size
,
6241 extra_mkfs_options
);
6248 r
= partition_populate_filesystem(context
, p
, d
->node
);
6253 if (fstat(fd
, &st
) < 0)
6254 return log_error_errno(errno
, "Failed to stat temporary file: %m");
6256 p
->copy_blocks_path
= TAKE_PTR(temp
);
6257 p
->copy_blocks_path_is_our_file
= true;
6258 p
->copy_blocks_fd
= TAKE_FD(fd
);
6259 p
->copy_blocks_size
= st
.st_size
;
6262 /* Now that we've done the data partitions, do the verity hash partitions. We do these in a separate
6263 * step because they might depend on data generated in the previous step. */
6265 LIST_FOREACH(partitions
, p
, context
->partitions
) {
6266 _cleanup_(unlink_and_freep
) char *temp
= NULL
;
6267 _cleanup_free_
char *hint
= NULL
;
6268 _cleanup_close_
int fd
= -EBADF
;
6275 if (PARTITION_EXISTS(p
)) /* Never format existing partitions */
6278 if (p
->minimize
== MINIMIZE_OFF
)
6281 if (p
->verity
!= VERITY_HASH
)
6284 assert_se(dp
= p
->siblings
[VERITY_DATA
]);
6285 assert(!dp
->dropped
);
6286 assert(dp
->copy_blocks_path
);
6288 (void) partition_hint(p
, context
->node
, &hint
);
6290 log_info("Pre-populating verity hash data of partition %s to calculate minimal partition size",
6294 r
= var_tmp_dir(&vt
);
6296 return log_error_errno(r
, "Could not determine temporary directory: %m");
6299 r
= tempfn_random_child(vt
, "repart", &temp
);
6301 return log_error_errno(r
, "Failed to generate temporary file path: %m");
6305 return log_error_errno(r
, "Failed to create temporary file: %m");
6307 r
= partition_format_verity_hash(context
, p
, temp
, dp
->copy_blocks_path
);
6311 fd
= open(temp
, O_RDONLY
|O_CLOEXEC
|O_NONBLOCK
);
6313 return log_error_errno(errno
, "Failed to open temporary file %s: %m", temp
);
6315 if (fstat(fd
, &st
) < 0)
6316 return log_error_errno(r
, "Failed to stat temporary file: %m");
6318 log_info("Minimal partition size of verity hash partition %s is %s",
6319 strna(hint
), FORMAT_BYTES(st
.st_size
));
6321 p
->copy_blocks_path
= TAKE_PTR(temp
);
6322 p
->copy_blocks_path_is_our_file
= true;
6323 p
->copy_blocks_fd
= TAKE_FD(fd
);
6324 p
->copy_blocks_size
= st
.st_size
;
6330 static int parse_partition_types(const char *p
, GptPartitionType
**partitions
, size_t *n_partitions
) {
6334 assert(n_partitions
);
6337 _cleanup_free_
char *name
= NULL
;
6338 GptPartitionType type
;
6340 r
= extract_first_word(&p
, &name
, ",", EXTRACT_CUNESCAPE
|EXTRACT_DONT_COALESCE_SEPARATORS
);
6344 return log_error_errno(r
, "Failed to extract partition type identifier or GUID: %s", p
);
6346 r
= gpt_partition_type_from_string(name
, &type
);
6348 return log_error_errno(r
, "'%s' is not a valid partition type identifier or GUID", name
);
6350 if (!GREEDY_REALLOC(*partitions
, *n_partitions
+ 1))
6353 (*partitions
)[(*n_partitions
)++] = type
;
6359 static int help(void) {
6360 _cleanup_free_
char *link
= NULL
;
6363 r
= terminal_urlify_man("systemd-repart", "8", &link
);
6367 printf("%s [OPTIONS...] [DEVICE]\n"
6368 "\n%sGrow and add partitions to partition table.%s\n\n"
6369 " -h --help Show this help\n"
6370 " --version Show package version\n"
6371 " --no-pager Do not pipe output into a pager\n"
6372 " --no-legend Do not show the headers and footers\n"
6373 " --dry-run=BOOL Whether to run dry-run operation\n"
6374 " --empty=MODE One of refuse, allow, require, force, create; controls\n"
6375 " how to handle empty disks lacking partition tables\n"
6376 " --discard=BOOL Whether to discard backing blocks for new partitions\n"
6377 " --pretty=BOOL Whether to show pretty summary before doing changes\n"
6378 " --factory-reset=BOOL Whether to remove data partitions before recreating\n"
6380 " --can-factory-reset Test whether factory reset is defined\n"
6381 " --root=PATH Operate relative to root path\n"
6382 " --image=PATH Operate relative to image file\n"
6383 " --image-policy=POLICY\n"
6384 " Specify disk image dissection policy\n"
6385 " --definitions=DIR Find partition definitions in specified directory\n"
6386 " --key-file=PATH Key to use when encrypting partitions\n"
6387 " --private-key=PATH Private key to use when generating verity roothash\n"
6389 " --certificate=PATH PEM certificate to use when generating verity\n"
6390 " roothash signatures\n"
6391 " --tpm2-device=PATH Path to TPM2 device node to use\n"
6392 " --tpm2-pcrs=PCR1+PCR2+PCR3+…\n"
6393 " TPM2 PCR indexes to use for TPM2 enrollment\n"
6394 " --tpm2-public-key=PATH\n"
6395 " Enroll signed TPM2 PCR policy against PEM public key\n"
6396 " --tpm2-public-key-pcrs=PCR1+PCR2+PCR3+…\n"
6397 " Enroll signed TPM2 PCR policy for specified TPM2 PCRs\n"
6398 " --tpm2-pcrlock=PATH\n"
6399 " Specify pcrlock policy to lock against\n"
6400 " --seed=UUID 128-bit seed UUID to derive all UUIDs from\n"
6401 " --size=BYTES Grow loopback file to specified size\n"
6402 " --json=pretty|short|off\n"
6403 " Generate JSON output\n"
6404 " --split=BOOL Whether to generate split artifacts\n"
6405 " --include-partitions=PARTITION1,PARTITION2,PARTITION3,…\n"
6406 " Ignore partitions not of the specified types\n"
6407 " --exclude-partitions=PARTITION1,PARTITION2,PARTITION3,…\n"
6408 " Ignore partitions of the specified types\n"
6409 " --defer-partitions=PARTITION1,PARTITION2,PARTITION3,…\n"
6410 " Take partitions of the specified types into account\n"
6411 " but don't populate them yet\n"
6412 " --sector-size=SIZE Set the logical sector size for the image\n"
6413 " --architecture=ARCH Set the generic architecture for the image\n"
6414 " --offline=BOOL Whether to build the image offline\n"
6415 " -s --copy-source=PATH Specify the primary source tree to copy files from\n"
6416 " --copy-from=IMAGE Copy partitions from the given image(s)\n"
6417 " -S --make-ddi=sysext Make a system extension DDI\n"
6418 " -C --make-ddi=confext Make a configuration extension DDI\n"
6419 " -P --make-ddi=portable Make a portable service DDI\n"
6420 "\nSee the %s for details.\n",
6421 program_invocation_short_name
,
6429 static int parse_argv(int argc
, char *argv
[]) {
6432 ARG_VERSION
= 0x100,
6439 ARG_CAN_FACTORY_RESET
,
6453 ARG_TPM2_PUBLIC_KEY
,
6454 ARG_TPM2_PUBLIC_KEY_PCRS
,
6457 ARG_INCLUDE_PARTITIONS
,
6458 ARG_EXCLUDE_PARTITIONS
,
6459 ARG_DEFER_PARTITIONS
,
6461 ARG_SKIP_PARTITIONS
,
6468 static const struct option options
[] = {
6469 { "help", no_argument
, NULL
, 'h' },
6470 { "version", no_argument
, NULL
, ARG_VERSION
},
6471 { "no-pager", no_argument
, NULL
, ARG_NO_PAGER
},
6472 { "no-legend", no_argument
, NULL
, ARG_NO_LEGEND
},
6473 { "dry-run", required_argument
, NULL
, ARG_DRY_RUN
},
6474 { "empty", required_argument
, NULL
, ARG_EMPTY
},
6475 { "discard", required_argument
, NULL
, ARG_DISCARD
},
6476 { "factory-reset", required_argument
, NULL
, ARG_FACTORY_RESET
},
6477 { "can-factory-reset", no_argument
, NULL
, ARG_CAN_FACTORY_RESET
},
6478 { "root", required_argument
, NULL
, ARG_ROOT
},
6479 { "image", required_argument
, NULL
, ARG_IMAGE
},
6480 { "image-policy", required_argument
, NULL
, ARG_IMAGE_POLICY
},
6481 { "seed", required_argument
, NULL
, ARG_SEED
},
6482 { "pretty", required_argument
, NULL
, ARG_PRETTY
},
6483 { "definitions", required_argument
, NULL
, ARG_DEFINITIONS
},
6484 { "size", required_argument
, NULL
, ARG_SIZE
},
6485 { "json", required_argument
, NULL
, ARG_JSON
},
6486 { "key-file", required_argument
, NULL
, ARG_KEY_FILE
},
6487 { "private-key", required_argument
, NULL
, ARG_PRIVATE_KEY
},
6488 { "certificate", required_argument
, NULL
, ARG_CERTIFICATE
},
6489 { "tpm2-device", required_argument
, NULL
, ARG_TPM2_DEVICE
},
6490 { "tpm2-pcrs", required_argument
, NULL
, ARG_TPM2_PCRS
},
6491 { "tpm2-public-key", required_argument
, NULL
, ARG_TPM2_PUBLIC_KEY
},
6492 { "tpm2-public-key-pcrs", required_argument
, NULL
, ARG_TPM2_PUBLIC_KEY_PCRS
},
6493 { "tpm2-pcrlock", required_argument
, NULL
, ARG_TPM2_PCRLOCK
},
6494 { "split", required_argument
, NULL
, ARG_SPLIT
},
6495 { "include-partitions", required_argument
, NULL
, ARG_INCLUDE_PARTITIONS
},
6496 { "exclude-partitions", required_argument
, NULL
, ARG_EXCLUDE_PARTITIONS
},
6497 { "defer-partitions", required_argument
, NULL
, ARG_DEFER_PARTITIONS
},
6498 { "sector-size", required_argument
, NULL
, ARG_SECTOR_SIZE
},
6499 { "architecture", required_argument
, NULL
, ARG_ARCHITECTURE
},
6500 { "offline", required_argument
, NULL
, ARG_OFFLINE
},
6501 { "copy-from", required_argument
, NULL
, ARG_COPY_FROM
},
6502 { "copy-source", required_argument
, NULL
, 's' },
6503 { "make-ddi", required_argument
, NULL
, ARG_MAKE_DDI
},
6507 bool auto_hash_pcr_values
= true, auto_public_key_pcr_mask
= true, auto_pcrlock
= true;
6513 while ((c
= getopt_long(argc
, argv
, "hs:SCP", options
, NULL
)) >= 0)
6524 arg_pager_flags
|= PAGER_DISABLE
;
6532 r
= parse_boolean_argument("--dry-run=", optarg
, &arg_dry_run
);
6538 if (isempty(optarg
)) {
6539 arg_empty
= EMPTY_UNSET
;
6543 arg_empty
= empty_mode_from_string(optarg
);
6545 return log_error_errno(arg_empty
, "Failed to parse --empty= parameter: %s", optarg
);
6550 r
= parse_boolean_argument("--discard=", optarg
, &arg_discard
);
6555 case ARG_FACTORY_RESET
:
6556 r
= parse_boolean_argument("--factory-reset=", optarg
, NULL
);
6559 arg_factory_reset
= r
;
6562 case ARG_CAN_FACTORY_RESET
:
6563 arg_can_factory_reset
= true;
6567 r
= parse_path_argument(optarg
, /* suppress_root= */ false, &arg_root
);
6573 r
= parse_path_argument(optarg
, /* suppress_root= */ false, &arg_image
);
6578 case ARG_IMAGE_POLICY
:
6579 r
= parse_image_policy_argument(optarg
, &arg_image_policy
);
6585 if (isempty(optarg
)) {
6586 arg_seed
= SD_ID128_NULL
;
6587 arg_randomize
= false;
6588 } else if (streq(optarg
, "random"))
6589 arg_randomize
= true;
6591 r
= sd_id128_from_string(optarg
, &arg_seed
);
6593 return log_error_errno(r
, "Failed to parse seed: %s", optarg
);
6595 arg_randomize
= false;
6601 r
= parse_boolean_argument("--pretty=", optarg
, NULL
);
6607 case ARG_DEFINITIONS
: {
6608 _cleanup_free_
char *path
= NULL
;
6609 r
= parse_path_argument(optarg
, false, &path
);
6612 if (strv_consume(&arg_definitions
, TAKE_PTR(path
)) < 0)
6618 uint64_t parsed
, rounded
;
6620 if (streq(optarg
, "auto")) {
6621 arg_size
= UINT64_MAX
;
6622 arg_size_auto
= true;
6626 r
= parse_size(optarg
, 1024, &parsed
);
6628 return log_error_errno(r
, "Failed to parse --size= parameter: %s", optarg
);
6630 rounded
= round_up_size(parsed
, 4096);
6632 return log_error_errno(SYNTHETIC_ERRNO(ERANGE
), "Specified image size too small, refusing.");
6633 if (rounded
== UINT64_MAX
)
6634 return log_error_errno(SYNTHETIC_ERRNO(ERANGE
), "Specified image size too large, refusing.");
6636 if (rounded
!= parsed
)
6637 log_warning("Specified size is not a multiple of 4096, rounding up automatically. (%" PRIu64
" %s %" PRIu64
")",
6638 parsed
, special_glyph(SPECIAL_GLYPH_ARROW_RIGHT
), rounded
);
6641 arg_size_auto
= false;
6646 r
= parse_json_argument(optarg
, &arg_json_format_flags
);
6652 case ARG_KEY_FILE
: {
6653 _cleanup_(erase_and_freep
) char *k
= NULL
;
6656 r
= read_full_file_full(
6657 AT_FDCWD
, optarg
, UINT64_MAX
, SIZE_MAX
,
6658 READ_FULL_FILE_SECURE
|READ_FULL_FILE_WARN_WORLD_READABLE
|READ_FULL_FILE_CONNECT_SOCKET
,
6662 return log_error_errno(r
, "Failed to read key file '%s': %m", optarg
);
6664 erase_and_free(arg_key
);
6665 arg_key
= TAKE_PTR(k
);
6670 case ARG_PRIVATE_KEY
: {
6671 _cleanup_(erase_and_freep
) char *k
= NULL
;
6674 r
= read_full_file_full(
6675 AT_FDCWD
, optarg
, UINT64_MAX
, SIZE_MAX
,
6676 READ_FULL_FILE_SECURE
|READ_FULL_FILE_WARN_WORLD_READABLE
|READ_FULL_FILE_CONNECT_SOCKET
,
6680 return log_error_errno(r
, "Failed to read key file '%s': %m", optarg
);
6682 EVP_PKEY_free(arg_private_key
);
6683 arg_private_key
= NULL
;
6684 r
= parse_private_key(k
, n
, &arg_private_key
);
6690 case ARG_CERTIFICATE
: {
6691 _cleanup_free_
char *cert
= NULL
;
6694 r
= read_full_file_full(
6695 AT_FDCWD
, optarg
, UINT64_MAX
, SIZE_MAX
,
6696 READ_FULL_FILE_CONNECT_SOCKET
,
6700 return log_error_errno(r
, "Failed to read certificate file '%s': %m", optarg
);
6702 X509_free(arg_certificate
);
6703 arg_certificate
= NULL
;
6704 r
= parse_x509_certificate(cert
, n
, &arg_certificate
);
6710 case ARG_TPM2_DEVICE
: {
6711 _cleanup_free_
char *device
= NULL
;
6713 if (streq(optarg
, "list"))
6714 return tpm2_list_devices();
6716 if (!streq(optarg
, "auto")) {
6717 device
= strdup(optarg
);
6722 free(arg_tpm2_device
);
6723 arg_tpm2_device
= TAKE_PTR(device
);
6728 auto_hash_pcr_values
= false;
6729 r
= tpm2_parse_pcr_argument_append(optarg
, &arg_tpm2_hash_pcr_values
, &arg_tpm2_n_hash_pcr_values
);
6735 case ARG_TPM2_PUBLIC_KEY
:
6736 r
= parse_path_argument(optarg
, /* suppress_root= */ false, &arg_tpm2_public_key
);
6742 case ARG_TPM2_PUBLIC_KEY_PCRS
:
6743 auto_public_key_pcr_mask
= false;
6744 r
= tpm2_parse_pcr_argument_to_mask(optarg
, &arg_tpm2_public_key_pcr_mask
);
6750 case ARG_TPM2_PCRLOCK
:
6751 r
= parse_path_argument(optarg
, /* suppress_root= */ false, &arg_tpm2_pcrlock
);
6755 auto_pcrlock
= false;
6759 r
= parse_boolean_argument("--split=", optarg
, NULL
);
6766 case ARG_INCLUDE_PARTITIONS
:
6767 if (arg_filter_partitions_type
== FILTER_PARTITIONS_EXCLUDE
)
6768 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
6769 "Combination of --include-partitions= and --exclude-partitions= is invalid.");
6771 r
= parse_partition_types(optarg
, &arg_filter_partitions
, &arg_n_filter_partitions
);
6775 arg_filter_partitions_type
= FILTER_PARTITIONS_INCLUDE
;
6779 case ARG_EXCLUDE_PARTITIONS
:
6780 if (arg_filter_partitions_type
== FILTER_PARTITIONS_INCLUDE
)
6781 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
6782 "Combination of --include-partitions= and --exclude-partitions= is invalid.");
6784 r
= parse_partition_types(optarg
, &arg_filter_partitions
, &arg_n_filter_partitions
);
6788 arg_filter_partitions_type
= FILTER_PARTITIONS_EXCLUDE
;
6792 case ARG_DEFER_PARTITIONS
:
6793 r
= parse_partition_types(optarg
, &arg_defer_partitions
, &arg_n_defer_partitions
);
6799 case ARG_SECTOR_SIZE
:
6800 r
= parse_sector_size(optarg
, &arg_sector_size
);
6806 case ARG_ARCHITECTURE
:
6807 r
= architecture_from_string(optarg
);
6809 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "Invalid architecture '%s'", optarg
);
6811 arg_architecture
= r
;
6815 if (streq(optarg
, "auto"))
6818 r
= parse_boolean_argument("--offline=", optarg
, NULL
);
6827 case ARG_COPY_FROM
: {
6828 _cleanup_free_
char *p
= NULL
;
6830 r
= parse_path_argument(optarg
, /* suppress_root= */ false, &p
);
6834 if (strv_consume(&arg_copy_from
, TAKE_PTR(p
)) < 0)
6841 r
= parse_path_argument(optarg
, /* suppress_root= */ false, &arg_copy_source
);
6847 if (!filename_is_valid(optarg
))
6848 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "Invalid DDI type: %s", optarg
);
6850 r
= free_and_strdup_warn(&arg_make_ddi
, optarg
);
6856 r
= free_and_strdup_warn(&arg_make_ddi
, "sysext");
6862 r
= free_and_strdup_warn(&arg_make_ddi
, "confext");
6868 r
= free_and_strdup_warn(&arg_make_ddi
, "portable");
6877 assert_not_reached();
6880 if (argc
- optind
> 1)
6881 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
6882 "Expected at most one argument, the path to the block device or image file.");
6885 if (arg_definitions
)
6886 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "Combination of --make-ddi= and --definitions= is not supported.");
6887 if (!IN_SET(arg_empty
, EMPTY_UNSET
, EMPTY_CREATE
))
6888 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "Combination of --make-ddi= and --empty=%s is not supported.", empty_mode_to_string(arg_empty
));
6890 /* Imply automatic sizing in DDI mode */
6891 if (arg_size
== UINT64_MAX
)
6892 arg_size_auto
= true;
6894 if (!arg_copy_source
)
6895 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "No --copy-source= specified, refusing.");
6897 r
= dir_is_empty(arg_copy_source
, /* ignore_hidden_or_backup= */ false);
6899 return log_error_errno(r
, "Failed to determine if '%s' is empty: %m", arg_copy_source
);
6901 return log_error_errno(SYNTHETIC_ERRNO(ENOENT
), "Source directory '%s' is empty, refusing to create empty image.", arg_copy_source
);
6903 if (sd_id128_is_null(arg_seed
) && !arg_randomize
) {
6904 /* We don't want that /etc/machine-id leaks into any image built this way, hence
6905 * let's randomize the seed if not specified explicitly */
6906 log_notice("No seed value specified, randomizing generated UUIDs, resulting image will not be reproducible.");
6907 arg_randomize
= true;
6910 arg_empty
= EMPTY_CREATE
;
6913 if (arg_empty
== EMPTY_UNSET
) /* default to refuse mode, if not otherwise specified */
6914 arg_empty
= EMPTY_REFUSE
;
6916 if (arg_factory_reset
> 0 && IN_SET(arg_empty
, EMPTY_FORCE
, EMPTY_REQUIRE
, EMPTY_CREATE
))
6917 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
6918 "Combination of --factory-reset=yes and --empty=force/--empty=require/--empty=create is invalid.");
6920 if (arg_can_factory_reset
)
6921 arg_dry_run
= true; /* When --can-factory-reset is specified we don't make changes, hence
6922 * non-dry-run mode makes no sense. Thus, imply dry run mode so that we
6923 * open things strictly read-only. */
6924 else if (arg_empty
== EMPTY_CREATE
)
6925 arg_dry_run
= false; /* Imply --dry-run=no if we create the loopback file anew. After all we
6926 * cannot really break anyone's partition tables that way. */
6928 /* Disable pager once we are not just reviewing, but doing things. */
6930 arg_pager_flags
|= PAGER_DISABLE
;
6932 if (arg_empty
== EMPTY_CREATE
&& arg_size
== UINT64_MAX
&& !arg_size_auto
)
6933 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
6934 "If --empty=create is specified, --size= must be specified, too.");
6936 if (arg_image
&& arg_root
)
6937 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
), "Please specify either --root= or --image=, the combination of both is not supported.");
6938 else if (!arg_image
&& !arg_root
&& in_initrd()) {
6940 /* By default operate on /sysusr/ or /sysroot/ when invoked in the initrd. We prefer the
6941 * former, if it is mounted, so that we have deterministic behaviour on systems where /usr/
6942 * is vendor-supplied but the root fs formatted on first boot. */
6943 r
= path_is_mount_point("/sysusr/usr", NULL
, 0);
6945 if (r
< 0 && r
!= -ENOENT
)
6946 log_debug_errno(r
, "Unable to determine whether /sysusr/usr is a mount point, assuming it is not: %m");
6948 arg_root
= strdup("/sysroot");
6950 arg_root
= strdup("/sysusr");
6955 arg_node
= argc
> optind
? argv
[optind
] : NULL
;
6957 if (IN_SET(arg_empty
, EMPTY_FORCE
, EMPTY_REQUIRE
, EMPTY_CREATE
) && !arg_node
&& !arg_image
)
6958 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
6959 "A path to a device node or image file must be specified when --make-ddi=, --empty=force, --empty=require or --empty=create are used.");
6961 if (arg_split
&& !arg_node
)
6962 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
6963 "A path to an image file must be specified when --split is used.");
6966 assert(!arg_tpm2_pcrlock
);
6968 r
= tpm2_pcrlock_search_file(NULL
, NULL
, &arg_tpm2_pcrlock
);
6971 log_warning_errno(r
, "Search for pcrlock.json failed, assuming it does not exist: %m");
6973 log_info("Automatically using pcrlock policy '%s'.", arg_tpm2_pcrlock
);
6976 if (auto_public_key_pcr_mask
) {
6977 assert(arg_tpm2_public_key_pcr_mask
== 0);
6978 arg_tpm2_public_key_pcr_mask
= INDEX_TO_MASK(uint32_t, TPM2_PCR_KERNEL_BOOT
);
6981 if (auto_hash_pcr_values
&& !arg_tpm2_pcrlock
) { /* Only lock to PCR 7 if no pcr policy is specified. */
6982 assert(arg_tpm2_n_hash_pcr_values
== 0);
6984 if (!GREEDY_REALLOC_APPEND(
6985 arg_tpm2_hash_pcr_values
,
6986 arg_tpm2_n_hash_pcr_values
,
6987 &TPM2_PCR_VALUE_MAKE(TPM2_PCR_INDEX_DEFAULT
, /* hash= */ 0, /* value= */ {}),
6992 if (arg_pretty
< 0 && isatty(STDOUT_FILENO
))
6995 if (arg_architecture
>= 0) {
6996 FOREACH_ARRAY(p
, arg_filter_partitions
, arg_n_filter_partitions
)
6997 *p
= gpt_partition_type_override_architecture(*p
, arg_architecture
);
6999 FOREACH_ARRAY(p
, arg_defer_partitions
, arg_n_defer_partitions
)
7000 *p
= gpt_partition_type_override_architecture(*p
, arg_architecture
);
7006 static int parse_proc_cmdline_factory_reset(void) {
7010 if (arg_factory_reset
>= 0) /* Never override what is specified on the process command line */
7013 if (!in_initrd()) /* Never honour kernel command line factory reset request outside of the initrd */
7016 r
= proc_cmdline_get_bool("systemd.factory_reset", /* flags = */ 0, &b
);
7018 return log_error_errno(r
, "Failed to parse systemd.factory_reset kernel command line argument: %m");
7020 arg_factory_reset
= b
;
7023 log_notice("Honouring factory reset requested via kernel command line.");
7029 static int parse_efi_variable_factory_reset(void) {
7030 _cleanup_free_
char *value
= NULL
;
7033 if (arg_factory_reset
>= 0) /* Never override what is specified on the process command line */
7036 if (!in_initrd()) /* Never honour EFI variable factory reset request outside of the initrd */
7039 r
= efi_get_variable_string(EFI_SYSTEMD_VARIABLE(FactoryReset
), &value
);
7041 if (r
== -ENOENT
|| ERRNO_IS_NOT_SUPPORTED(r
))
7043 return log_error_errno(r
, "Failed to read EFI variable FactoryReset: %m");
7046 r
= parse_boolean(value
);
7048 return log_error_errno(r
, "Failed to parse EFI variable FactoryReset: %m");
7050 arg_factory_reset
= r
;
7052 log_notice("Factory reset requested via EFI variable FactoryReset.");
7057 static int remove_efi_variable_factory_reset(void) {
7060 r
= efi_set_variable(EFI_SYSTEMD_VARIABLE(FactoryReset
), NULL
, 0);
7062 if (r
== -ENOENT
|| ERRNO_IS_NOT_SUPPORTED(r
))
7064 return log_error_errno(r
, "Failed to remove EFI variable FactoryReset: %m");
7067 log_info("Successfully unset EFI variable FactoryReset.");
7071 static int acquire_root_devno(
7078 _cleanup_free_
char *found_path
= NULL
, *node
= NULL
;
7079 dev_t devno
, fd_devno
= MODE_INVALID
;
7080 _cleanup_close_
int fd
= -EBADF
;
7088 fd
= chase_and_open(p
, root
, CHASE_PREFIX_ROOT
, mode
, &found_path
);
7092 if (fstat(fd
, &st
) < 0)
7095 if (S_ISREG(st
.st_mode
)) {
7096 *ret
= TAKE_PTR(found_path
);
7097 *ret_fd
= TAKE_FD(fd
);
7101 if (S_ISBLK(st
.st_mode
)) {
7102 /* Refuse referencing explicit block devices if a root dir is specified, after all we should
7103 * not be able to leave the image the root path constrains us to. */
7107 fd_devno
= devno
= st
.st_rdev
;
7108 } else if (S_ISDIR(st
.st_mode
)) {
7111 if (major(devno
) == 0) {
7112 r
= btrfs_get_block_device_fd(fd
, &devno
);
7113 if (r
== -ENOTTY
) /* not btrfs */
7121 /* From dm-crypt to backing partition */
7122 r
= block_get_originating(devno
, &devno
);
7124 log_debug_errno(r
, "Device '%s' has no dm-crypt/dm-verity device, no need to look for underlying block device.", p
);
7126 log_debug_errno(r
, "Failed to find underlying block device for '%s', ignoring: %m", p
);
7128 /* From partition to whole disk containing it */
7129 r
= block_get_whole_disk(devno
, &devno
);
7131 log_debug_errno(r
, "Failed to find whole disk block device for '%s', ignoring: %m", p
);
7133 r
= devname_from_devnum(S_IFBLK
, devno
, &node
);
7135 return log_debug_errno(r
, "Failed to determine canonical path for '%s': %m", p
);
7137 /* Only if we still look at the same block device we can reuse the fd. Otherwise return an
7138 * invalidated fd. */
7139 if (fd_devno
!= MODE_INVALID
&& fd_devno
== devno
) {
7140 /* Tell udev not to interfere while we are processing the device */
7141 if (flock(fd
, arg_dry_run
? LOCK_SH
: LOCK_EX
) < 0)
7142 return log_error_errno(errno
, "Failed to lock device '%s': %m", node
);
7144 *ret_fd
= TAKE_FD(fd
);
7148 *ret
= TAKE_PTR(node
);
7152 static int find_root(Context
*context
) {
7153 _cleanup_free_
char *device
= NULL
;
7159 if (arg_empty
== EMPTY_CREATE
) {
7160 _cleanup_close_
int fd
= -EBADF
;
7161 _cleanup_free_
char *s
= NULL
;
7163 s
= strdup(arg_node
);
7167 fd
= open(arg_node
, O_RDONLY
|O_CREAT
|O_EXCL
|O_CLOEXEC
|O_NOFOLLOW
, 0666);
7169 return log_error_errno(errno
, "Failed to create '%s': %m", arg_node
);
7171 context
->node
= TAKE_PTR(s
);
7172 context
->node_is_our_file
= true;
7173 context
->backing_fd
= TAKE_FD(fd
);
7177 /* Note that we don't specify a root argument here: if the user explicitly configured a node
7178 * we'll take it relative to the host, not the image */
7179 r
= acquire_root_devno(arg_node
, NULL
, O_RDONLY
|O_CLOEXEC
, &context
->node
, &context
->backing_fd
);
7181 return btrfs_log_dev_root(LOG_ERR
, r
, arg_node
);
7183 return log_error_errno(r
, "Failed to open file or determine backing device of %s: %m", arg_node
);
7188 assert(IN_SET(arg_empty
, EMPTY_REFUSE
, EMPTY_ALLOW
));
7190 /* If the root mount has been replaced by some form of volatile file system (overlayfs), the
7191 * original root block device node is symlinked in /run/systemd/volatile-root. Let's read that
7193 r
= readlink_malloc("/run/systemd/volatile-root", &device
);
7194 if (r
== -ENOENT
) { /* volatile-root not found */
7195 /* Let's search for the root device. We look for two cases here: first in /, and then in /usr. The
7196 * latter we check for cases where / is a tmpfs and only /usr is an actual persistent block device
7197 * (think: volatile setups) */
7199 FOREACH_STRING(p
, "/", "/usr") {
7201 r
= acquire_root_devno(p
, arg_root
, O_RDONLY
|O_DIRECTORY
|O_CLOEXEC
, &context
->node
,
7202 &context
->backing_fd
);
7205 return btrfs_log_dev_root(LOG_ERR
, r
, p
);
7207 return log_error_errno(r
, "Failed to determine backing device of %s: %m", p
);
7212 return log_error_errno(r
, "Failed to read symlink /run/systemd/volatile-root: %m");
7214 r
= acquire_root_devno(device
, NULL
, O_RDONLY
|O_CLOEXEC
, &context
->node
, &context
->backing_fd
);
7216 return btrfs_log_dev_root(LOG_ERR
, r
, device
);
7218 return log_error_errno(r
, "Failed to open file or determine backing device of %s: %m", device
);
7223 return log_error_errno(SYNTHETIC_ERRNO(ENODEV
), "Failed to discover root block device.");
7226 static int resize_pt(int fd
, uint64_t sector_size
) {
7227 _cleanup_(fdisk_unref_contextp
) struct fdisk_context
*c
= NULL
;
7230 /* After resizing the backing file we need to resize the partition table itself too, so that it takes
7231 * possession of the enlarged backing file. For this it suffices to open the device with libfdisk and
7232 * immediately write it again, with no changes. */
7234 r
= fdisk_new_context_at(fd
, /* path= */ NULL
, /* read_only= */ false, sector_size
, &c
);
7236 return log_error_errno(r
, "Failed to open device '%s': %m", FORMAT_PROC_FD_PATH(fd
));
7238 r
= fdisk_has_label(c
);
7240 return log_error_errno(r
, "Failed to determine whether disk '%s' has a disk label: %m", FORMAT_PROC_FD_PATH(fd
));
7242 log_debug("Not resizing partition table, as there currently is none.");
7246 r
= fdisk_write_disklabel(c
);
7248 return log_error_errno(r
, "Failed to write resized partition table: %m");
7250 log_info("Resized partition table.");
7254 static int resize_backing_fd(
7255 const char *node
, /* The primary way we access the disk image to operate on */
7256 int *fd
, /* An O_RDONLY fd referring to that inode */
7257 const char *backing_file
, /* If the above refers to a loopback device, the backing regular file for that, which we can grow */
7258 LoopDevice
*loop_device
,
7259 uint64_t sector_size
) {
7261 _cleanup_close_
int writable_fd
= -EBADF
;
7262 uint64_t current_size
;
7269 if (arg_size
== UINT64_MAX
) /* Nothing to do */
7273 /* Open the file if we haven't opened it yet. Note that we open it read-only here, just to
7274 * keep a reference to the file we can pass around. */
7275 *fd
= open(node
, O_RDONLY
|O_CLOEXEC
);
7277 return log_error_errno(errno
, "Failed to open '%s' in order to adjust size: %m", node
);
7280 if (fstat(*fd
, &st
) < 0)
7281 return log_error_errno(errno
, "Failed to stat '%s': %m", node
);
7283 if (S_ISBLK(st
.st_mode
)) {
7285 return log_error_errno(SYNTHETIC_ERRNO(EBADF
), "Cannot resize block device '%s'.", node
);
7287 assert(loop_device
);
7289 if (ioctl(*fd
, BLKGETSIZE64
, ¤t_size
) < 0)
7290 return log_error_errno(errno
, "Failed to determine size of block device %s: %m", node
);
7292 r
= stat_verify_regular(&st
);
7294 return log_error_errno(r
, "Specified path '%s' is not a regular file or loopback block device, cannot resize: %m", node
);
7296 assert(!backing_file
);
7297 assert(!loop_device
);
7298 current_size
= st
.st_size
;
7301 if (current_size
>= arg_size
) {
7302 log_info("File '%s' already is of requested size or larger, not growing. (%s >= %s)",
7303 node
, FORMAT_BYTES(current_size
), FORMAT_BYTES(arg_size
));
7307 if (S_ISBLK(st
.st_mode
)) {
7308 assert(backing_file
);
7310 /* This is a loopback device. We can't really grow those directly, but we can grow the
7311 * backing file, hence let's do that. */
7313 writable_fd
= open(backing_file
, O_WRONLY
|O_CLOEXEC
|O_NONBLOCK
);
7314 if (writable_fd
< 0)
7315 return log_error_errno(errno
, "Failed to open backing file '%s': %m", backing_file
);
7317 if (fstat(writable_fd
, &st
) < 0)
7318 return log_error_errno(errno
, "Failed to stat() backing file '%s': %m", backing_file
);
7320 r
= stat_verify_regular(&st
);
7322 return log_error_errno(r
, "Backing file '%s' of block device is not a regular file: %m", backing_file
);
7324 if ((uint64_t) st
.st_size
!= current_size
)
7325 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
7326 "Size of backing file '%s' of loopback block device '%s' don't match, refusing.",
7327 node
, backing_file
);
7329 assert(S_ISREG(st
.st_mode
));
7330 assert(!backing_file
);
7332 /* The file descriptor is read-only. In order to grow the file we need to have a writable fd. We
7333 * reopen the file for that temporarily. We keep the writable fd only open for this operation though,
7334 * as fdisk can't accept it anyway. */
7336 writable_fd
= fd_reopen(*fd
, O_WRONLY
|O_CLOEXEC
);
7337 if (writable_fd
< 0)
7338 return log_error_errno(writable_fd
, "Failed to reopen backing file '%s' writable: %m", node
);
7342 if (fallocate(writable_fd
, 0, 0, arg_size
) < 0) {
7343 if (!ERRNO_IS_NOT_SUPPORTED(errno
))
7344 return log_error_errno(errno
, "Failed to grow '%s' from %s to %s by allocation: %m",
7345 node
, FORMAT_BYTES(current_size
), FORMAT_BYTES(arg_size
));
7347 /* Fallback to truncation, if fallocate() is not supported. */
7348 log_debug("Backing file system does not support fallocate(), falling back to ftruncate().");
7350 if (current_size
== 0) /* Likely regular file just created by us */
7351 log_info("Allocated %s for '%s'.", FORMAT_BYTES(arg_size
), node
);
7353 log_info("File '%s' grown from %s to %s by allocation.",
7354 node
, FORMAT_BYTES(current_size
), FORMAT_BYTES(arg_size
));
7360 if (ftruncate(writable_fd
, arg_size
) < 0)
7361 return log_error_errno(errno
, "Failed to grow '%s' from %s to %s by truncation: %m",
7362 node
, FORMAT_BYTES(current_size
), FORMAT_BYTES(arg_size
));
7364 if (current_size
== 0) /* Likely regular file just created by us */
7365 log_info("Sized '%s' to %s.", node
, FORMAT_BYTES(arg_size
));
7367 log_info("File '%s' grown from %s to %s by truncation.",
7368 node
, FORMAT_BYTES(current_size
), FORMAT_BYTES(arg_size
));
7371 r
= resize_pt(writable_fd
, sector_size
);
7376 r
= loop_device_refresh_size(loop_device
, UINT64_MAX
, arg_size
);
7378 return log_error_errno(r
, "Failed to update loop device size: %m");
7384 static int determine_auto_size(Context
*c
) {
7389 sum
= round_up_size(GPT_METADATA_SIZE
, 4096);
7391 LIST_FOREACH(partitions
, p
, c
->partitions
) {
7397 m
= partition_min_size_with_padding(c
, p
);
7398 if (m
> UINT64_MAX
- sum
)
7399 return log_error_errno(SYNTHETIC_ERRNO(EOVERFLOW
), "Image would grow too large, refusing.");
7404 if (c
->total
!= UINT64_MAX
)
7405 /* Image already allocated? Then show its size. */
7406 log_info("Automatically determined minimal disk image size as %s, current image size is %s.",
7407 FORMAT_BYTES(sum
), FORMAT_BYTES(c
->total
));
7409 /* If the image is being created right now, then it has no previous size, suppress any comment about it hence. */
7410 log_info("Automatically determined minimal disk image size as %s.",
7417 static int run(int argc
, char *argv
[]) {
7418 _cleanup_(loop_device_unrefp
) LoopDevice
*loop_device
= NULL
;
7419 _cleanup_(umount_and_freep
) char *mounted_dir
= NULL
;
7420 _cleanup_(context_freep
) Context
* context
= NULL
;
7421 bool node_is_our_loop
= false;
7424 log_show_color(true);
7425 log_parse_environment();
7428 r
= parse_argv(argc
, argv
);
7432 r
= parse_proc_cmdline_factory_reset();
7436 r
= parse_efi_variable_factory_reset();
7440 #if HAVE_LIBCRYPTSETUP
7441 cryptsetup_enable_logging(NULL
);
7447 /* Mount this strictly read-only: we shall modify the partition table, not the file
7449 r
= mount_image_privately_interactively(
7452 DISSECT_IMAGE_MOUNT_READ_ONLY
|
7453 (arg_node
? DISSECT_IMAGE_DEVICE_READ_ONLY
: 0) | /* If a different node to make changes to is specified let's open the device in read-only mode) */
7454 DISSECT_IMAGE_GPT_ONLY
|
7455 DISSECT_IMAGE_RELAX_VAR_CHECK
|
7456 DISSECT_IMAGE_USR_NO_ROOT
|
7457 DISSECT_IMAGE_REQUIRE_ROOT
,
7459 /* ret_dir_fd= */ NULL
,
7464 arg_root
= strdup(mounted_dir
);
7469 arg_node
= strdup(loop_device
->node
);
7473 /* Remember that the device we are about to manipulate is actually the one we
7474 * allocated here, and thus to increase its backing file we know what to do */
7475 node_is_our_loop
= true;
7479 if (!arg_copy_source
&& arg_root
) {
7480 /* If no explicit copy source is specified, then use --root=/--image= */
7481 arg_copy_source
= strdup(arg_root
);
7482 if (!arg_copy_source
)
7486 context
= context_new(arg_seed
);
7490 r
= context_copy_from(context
);
7495 _cleanup_free_
char *d
= NULL
, *dp
= NULL
;
7496 assert(!arg_definitions
);
7498 d
= strjoin(arg_make_ddi
, ".repart.d/");
7502 r
= search_and_access(d
, F_OK
, arg_root
, CONF_PATHS_USR_STRV("systemd/repart/definitions"), &dp
);
7504 return log_error_errno(errno
, "DDI type '%s' is not defined: %m", arg_make_ddi
);
7506 if (strv_consume(&arg_definitions
, TAKE_PTR(dp
)) < 0)
7509 strv_uniq(arg_definitions
);
7511 r
= context_read_definitions(context
);
7515 r
= find_root(context
);
7517 return 76; /* Special return value which means "Root block device not found, so not doing
7518 * anything". This isn't really an error when called at boot. */
7522 if (arg_size
!= UINT64_MAX
) {
7523 r
= resize_backing_fd(
7525 &context
->backing_fd
,
7526 node_is_our_loop
? arg_image
: NULL
,
7527 node_is_our_loop
? loop_device
: NULL
,
7528 context
->sector_size
);
7533 r
= context_load_partition_table(context
);
7534 if (r
== -EHWPOISON
)
7535 return 77; /* Special return value which means "Not GPT, so not doing anything". This isn't
7536 * really an error when called at boot. */
7539 context
->from_scratch
= r
> 0; /* Starting from scratch */
7541 if (arg_can_factory_reset
) {
7542 r
= context_can_factory_reset(context
);
7546 return EXIT_FAILURE
;
7551 r
= context_factory_reset(context
);
7555 /* We actually did a factory reset! */
7556 r
= remove_efi_variable_factory_reset();
7560 /* Reload the reduced partition table */
7561 context_unload_partition_table(context
);
7562 r
= context_load_partition_table(context
);
7567 r
= context_read_seed(context
, arg_root
);
7571 /* Make sure each partition has a unique UUID and unique label */
7572 r
= context_acquire_partition_uuids_and_labels(context
);
7576 /* Open all files to copy blocks from now, since we want to take their size into consideration */
7577 r
= context_open_copy_block_paths(
7579 loop_device
? loop_device
->devno
: /* if --image= is specified, only allow partitions on the loopback device */
7580 arg_root
&& !arg_image
? 0 : /* if --root= is specified, don't accept any block device */
7581 (dev_t
) -1); /* if neither is specified, make no restrictions */
7585 r
= context_minimize(context
);
7589 if (arg_size_auto
) {
7590 r
= determine_auto_size(context
);
7594 /* Flush out everything again, and let's grow the file first, then start fresh */
7595 context_unload_partition_table(context
);
7597 assert(arg_size
!= UINT64_MAX
);
7598 r
= resize_backing_fd(
7600 &context
->backing_fd
,
7601 node_is_our_loop
? arg_image
: NULL
,
7602 node_is_our_loop
? loop_device
: NULL
,
7603 context
->sector_size
);
7607 r
= context_load_partition_table(context
);
7612 /* First try to fit new partitions in, dropping by priority until it fits */
7614 uint64_t largest_free_area
;
7616 if (context_allocate_partitions(context
, &largest_free_area
))
7617 break; /* Success! */
7619 if (!context_drop_or_foreignize_one_priority(context
)) {
7620 r
= log_error_errno(SYNTHETIC_ERRNO(ENOSPC
),
7621 "Can't fit requested partitions into available free space (%s), refusing.",
7622 FORMAT_BYTES(largest_free_area
));
7623 determine_auto_size(context
);
7628 /* Now assign free space according to the weight logic */
7629 r
= context_grow_partitions(context
);
7633 /* Now calculate where each new partition gets placed */
7634 context_place_partitions(context
);
7636 (void) context_dump(context
, /*late=*/ false);
7638 r
= context_write_partition_table(context
);
7642 r
= context_split(context
);
7646 (void) context_dump(context
, /*late=*/ true);
7648 context
->node
= mfree(context
->node
);
7650 LIST_FOREACH(partitions
, p
, context
->partitions
)
7651 p
->split_path
= mfree(p
->split_path
);
7656 DEFINE_MAIN_FUNCTION_WITH_POSITIVE_FAILURE(run
);