]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/partition/repart.c
projects / thirdparty / systemd.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
tree-wide: port things over to FORMAT_PROC_FD_PATH()
[thirdparty/systemd.git] / src / partition / repart.c
1 /* SPDX-License-Identifier: LGPL-2.1-or-later */
2
3 #if HAVE_VALGRIND_MEMCHECK_H
4 #include <valgrind/memcheck.h>
5 #endif
6
7 #include <fcntl.h>
8 #include <getopt.h>
9 #include <libfdisk.h>
10 #include <linux/fs.h>
11 #include <linux/loop.h>
12 #include <sys/file.h>
13 #include <sys/ioctl.h>
14 #include <sys/stat.h>
15
16 #include <openssl/hmac.h>
17 #include <openssl/sha.h>
18
19 #include "sd-id128.h"
20
21 #include "alloc-util.h"
22 #include "blkid-util.h"
23 #include "blockdev-util.h"
24 #include "btrfs-util.h"
25 #include "conf-files.h"
26 #include "conf-parser.h"
27 #include "cryptsetup-util.h"
28 #include "def.h"
29 #include "dirent-util.h"
30 #include "efivars.h"
31 #include "errno-util.h"
32 #include "fd-util.h"
33 #include "fileio.h"
34 #include "format-table.h"
35 #include "format-util.h"
36 #include "fs-util.h"
37 #include "gpt.h"
38 #include "hexdecoct.h"
39 #include "id128-util.h"
40 #include "json.h"
41 #include "list.h"
42 #include "locale-util.h"
43 #include "loop-util.h"
44 #include "main-func.h"
45 #include "mkdir.h"
46 #include "mkfs-util.h"
47 #include "mount-util.h"
48 #include "mountpoint-util.h"
49 #include "parse-argument.h"
50 #include "parse-util.h"
51 #include "path-util.h"
52 #include "pretty-print.h"
53 #include "proc-cmdline.h"
54 #include "process-util.h"
55 #include "random-util.h"
56 #include "resize-fs.h"
57 #include "sort-util.h"
58 #include "specifier.h"
59 #include "stat-util.h"
60 #include "stdio-util.h"
61 #include "string-table.h"
62 #include "string-util.h"
63 #include "strv.h"
64 #include "terminal-util.h"
65 #include "tpm2-util.h"
66 #include "user-util.h"
67 #include "utf8.h"
68
69 /* If not configured otherwise use a minimal partition size of 10M */
70 #define DEFAULT_MIN_SIZE (10*1024*1024)
71
72 /* Hard lower limit for new partition sizes */
73 #define HARD_MIN_SIZE 4096
74
75 /* libfdisk takes off slightly more than 1M of the disk size when creating a GPT disk label */
76 #define GPT_METADATA_SIZE (1044*1024)
77
78 /* LUKS2 takes off 16M of the partition size with its metadata by default */
79 #define LUKS2_METADATA_SIZE (16*1024*1024)
80
81 /* Note: When growing and placing new partitions we always align to 4K sector size. It's how newer hard disks
82 * are designed, and if everything is aligned to that performance is best. And for older hard disks with 512B
83 * sector size devices were generally assumed to have an even number of sectors, hence at the worst we'll
84 * waste 3K per partition, which is probably fine. */
85
86 static enum {
87 EMPTY_REFUSE, /* refuse empty disks, never create a partition table */
88 EMPTY_ALLOW, /* allow empty disks, create partition table if necessary */
89 EMPTY_REQUIRE, /* require an empty disk, create a partition table */
90 EMPTY_FORCE, /* make disk empty, erase everything, create a partition table always */
91 EMPTY_CREATE, /* create disk as loopback file, create a partition table always */
92 } arg_empty = EMPTY_REFUSE;
93
94 static bool arg_dry_run = true;
95 static const char *arg_node = NULL;
96 static char *arg_root = NULL;
97 static char *arg_image = NULL;
98 static char *arg_definitions = NULL;
99 static bool arg_discard = true;
100 static bool arg_can_factory_reset = false;
101 static int arg_factory_reset = -1;
102 static sd_id128_t arg_seed = SD_ID128_NULL;
103 static bool arg_randomize = false;
104 static int arg_pretty = -1;
105 static uint64_t arg_size = UINT64_MAX;
106 static bool arg_size_auto = false;
107 static JsonFormatFlags arg_json_format_flags = JSON_FORMAT_OFF;
108 static PagerFlags arg_pager_flags = 0;
109 static bool arg_legend = true;
110 static void *arg_key = NULL;
111 static size_t arg_key_size = 0;
112 static char *arg_tpm2_device = NULL;
113 static uint32_t arg_tpm2_pcr_mask = UINT32_MAX;
114
115 STATIC_DESTRUCTOR_REGISTER(arg_root, freep);
116 STATIC_DESTRUCTOR_REGISTER(arg_image, freep);
117 STATIC_DESTRUCTOR_REGISTER(arg_definitions, freep);
118 STATIC_DESTRUCTOR_REGISTER(arg_key, erase_and_freep);
119 STATIC_DESTRUCTOR_REGISTER(arg_tpm2_device, freep);
120
121 typedef struct Partition Partition;
122 typedef struct FreeArea FreeArea;
123 typedef struct Context Context;
124
125 typedef enum EncryptMode {
126 ENCRYPT_OFF,
127 ENCRYPT_KEY_FILE,
128 ENCRYPT_TPM2,
129 ENCRYPT_KEY_FILE_TPM2,
130 _ENCRYPT_MODE_MAX,
131 _ENCRYPT_MODE_INVALID = -EINVAL,
132 } EncryptMode;
133
134 struct Partition {
135 char *definition_path;
136
137 sd_id128_t type_uuid;
138 sd_id128_t current_uuid, new_uuid;
139 char *current_label, *new_label;
140
141 bool dropped;
142 bool factory_reset;
143 int32_t priority;
144
145 uint32_t weight, padding_weight;
146
147 uint64_t current_size, new_size;
148 uint64_t size_min, size_max;
149
150 uint64_t current_padding, new_padding;
151 uint64_t padding_min, padding_max;
152
153 uint64_t partno;
154 uint64_t offset;
155
156 struct fdisk_partition *current_partition;
157 struct fdisk_partition *new_partition;
158 FreeArea *padding_area;
159 FreeArea *allocated_to_area;
160
161 char *copy_blocks_path;
162 bool copy_blocks_auto;
163 int copy_blocks_fd;
164 uint64_t copy_blocks_size;
165
166 char *format;
167 char **copy_files;
168 char **make_directories;
169 EncryptMode encrypt;
170
171 uint64_t gpt_flags;
172 int no_auto;
173 int read_only;
174 int growfs;
175
176 LIST_FIELDS(Partition, partitions);
177 };
178
179 #define PARTITION_IS_FOREIGN(p) (!(p)->definition_path)
180 #define PARTITION_EXISTS(p) (!!(p)->current_partition)
181
182 struct FreeArea {
183 Partition *after;
184 uint64_t size;
185 uint64_t allocated;
186 };
187
188 struct Context {
189 LIST_HEAD(Partition, partitions);
190 size_t n_partitions;
191
192 FreeArea **free_areas;
193 size_t n_free_areas;
194
195 uint64_t start, end, total;
196
197 struct fdisk_context *fdisk_context;
198
199 sd_id128_t seed;
200 };
201
202 static const char *encrypt_mode_table[_ENCRYPT_MODE_MAX] = {
203 [ENCRYPT_OFF] = "off",
204 [ENCRYPT_KEY_FILE] = "key-file",
205 [ENCRYPT_TPM2] = "tpm2",
206 [ENCRYPT_KEY_FILE_TPM2] = "key-file+tpm2",
207 };
208
209 DEFINE_PRIVATE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(encrypt_mode, EncryptMode, ENCRYPT_KEY_FILE);
210
211 static uint64_t round_down_size(uint64_t v, uint64_t p) {
212 return (v / p) * p;
213 }
214
215 static uint64_t round_up_size(uint64_t v, uint64_t p) {
216
217 v = DIV_ROUND_UP(v, p);
218
219 if (v > UINT64_MAX / p)
220 return UINT64_MAX; /* overflow */
221
222 return v * p;
223 }
224
225 static Partition *partition_new(void) {
226 Partition *p;
227
228 p = new(Partition, 1);
229 if (!p)
230 return NULL;
231
232 *p = (Partition) {
233 .weight = 1000,
234 .padding_weight = 0,
235 .current_size = UINT64_MAX,
236 .new_size = UINT64_MAX,
237 .size_min = UINT64_MAX,
238 .size_max = UINT64_MAX,
239 .current_padding = UINT64_MAX,
240 .new_padding = UINT64_MAX,
241 .padding_min = UINT64_MAX,
242 .padding_max = UINT64_MAX,
243 .partno = UINT64_MAX,
244 .offset = UINT64_MAX,
245 .copy_blocks_fd = -1,
246 .copy_blocks_size = UINT64_MAX,
247 .no_auto = -1,
248 .read_only = -1,
249 .growfs = -1,
250 };
251
252 return p;
253 }
254
255 static Partition* partition_free(Partition *p) {
256 if (!p)
257 return NULL;
258
259 free(p->current_label);
260 free(p->new_label);
261 free(p->definition_path);
262
263 if (p->current_partition)
264 fdisk_unref_partition(p->current_partition);
265 if (p->new_partition)
266 fdisk_unref_partition(p->new_partition);
267
268 free(p->copy_blocks_path);
269 safe_close(p->copy_blocks_fd);
270
271 free(p->format);
272 strv_free(p->copy_files);
273 strv_free(p->make_directories);
274
275 return mfree(p);
276 }
277
278 static Partition* partition_unlink_and_free(Context *context, Partition *p) {
279 if (!p)
280 return NULL;
281
282 LIST_REMOVE(partitions, context->partitions, p);
283
284 assert(context->n_partitions > 0);
285 context->n_partitions--;
286
287 return partition_free(p);
288 }
289
290 DEFINE_TRIVIAL_CLEANUP_FUNC(Partition*, partition_free);
291
292 static Context *context_new(sd_id128_t seed) {
293 Context *context;
294
295 context = new(Context, 1);
296 if (!context)
297 return NULL;
298
299 *context = (Context) {
300 .start = UINT64_MAX,
301 .end = UINT64_MAX,
302 .total = UINT64_MAX,
303 .seed = seed,
304 };
305
306 return context;
307 }
308
309 static void context_free_free_areas(Context *context) {
310 assert(context);
311
312 for (size_t i = 0; i < context->n_free_areas; i++)
313 free(context->free_areas[i]);
314
315 context->free_areas = mfree(context->free_areas);
316 context->n_free_areas = 0;
317 }
318
319 static Context *context_free(Context *context) {
320 if (!context)
321 return NULL;
322
323 while (context->partitions)
324 partition_unlink_and_free(context, context->partitions);
325 assert(context->n_partitions == 0);
326
327 context_free_free_areas(context);
328
329 if (context->fdisk_context)
330 fdisk_unref_context(context->fdisk_context);
331
332 return mfree(context);
333 }
334
335 DEFINE_TRIVIAL_CLEANUP_FUNC(Context*, context_free);
336
337 static int context_add_free_area(
338 Context *context,
339 uint64_t size,
340 Partition *after) {
341
342 FreeArea *a;
343
344 assert(context);
345 assert(!after || !after->padding_area);
346
347 if (!GREEDY_REALLOC(context->free_areas, context->n_free_areas + 1))
348 return -ENOMEM;
349
350 a = new(FreeArea, 1);
351 if (!a)
352 return -ENOMEM;
353
354 *a = (FreeArea) {
355 .size = size,
356 .after = after,
357 };
358
359 context->free_areas[context->n_free_areas++] = a;
360
361 if (after)
362 after->padding_area = a;
363
364 return 0;
365 }
366
367 static bool context_drop_one_priority(Context *context) {
368 int32_t priority = 0;
369 Partition *p;
370 bool exists = false;
371
372 LIST_FOREACH(partitions, p, context->partitions) {
373 if (p->dropped)
374 continue;
375 if (p->priority < priority)
376 continue;
377 if (p->priority == priority) {
378 exists = exists || PARTITION_EXISTS(p);
379 continue;
380 }
381
382 priority = p->priority;
383 exists = PARTITION_EXISTS(p);
384 }
385
386 /* Refuse to drop partitions with 0 or negative priorities or partitions of priorities that have at
387 * least one existing priority */
388 if (priority <= 0 || exists)
389 return false;
390
391 LIST_FOREACH(partitions, p, context->partitions) {
392 if (p->priority < priority)
393 continue;
394
395 if (p->dropped)
396 continue;
397
398 p->dropped = true;
399 log_info("Can't fit partition %s of priority %" PRIi32 ", dropping.", p->definition_path, p->priority);
400 }
401
402 return true;
403 }
404
405 static uint64_t partition_min_size(const Partition *p) {
406 uint64_t sz;
407
408 /* Calculate the disk space we really need at minimum for this partition. If the partition already
409 * exists the current size is what we really need. If it doesn't exist yet refuse to allocate less
410 * than 4K.
411 *
412 * DEFAULT_MIN_SIZE is the default SizeMin= we configure if nothing else is specified. */
413
414 if (PARTITION_IS_FOREIGN(p)) {
415 /* Don't allow changing size of partitions not managed by us */
416 assert(p->current_size != UINT64_MAX);
417 return p->current_size;
418 }
419
420 sz = p->current_size != UINT64_MAX ? p->current_size : HARD_MIN_SIZE;
421
422 if (!PARTITION_EXISTS(p)) {
423 uint64_t d = 0;
424
425 if (p->encrypt != ENCRYPT_OFF)
426 d += round_up_size(LUKS2_METADATA_SIZE, 4096);
427
428 if (p->copy_blocks_size != UINT64_MAX)
429 d += round_up_size(p->copy_blocks_size, 4096);
430 else if (p->format || p->encrypt != ENCRYPT_OFF) {
431 uint64_t f;
432
433 /* If we shall synthesize a file system, take minimal fs size into account (assumed to be 4K if not known) */
434 f = p->format ? minimal_size_by_fs_name(p->format) : UINT64_MAX;
435 d += f == UINT64_MAX ? 4096 : f;
436 }
437
438 if (d > sz)
439 sz = d;
440 }
441
442 return MAX(p->size_min != UINT64_MAX ? p->size_min : DEFAULT_MIN_SIZE, sz);
443 }
444
445 static uint64_t partition_max_size(const Partition *p) {
446 /* Calculate how large the partition may become at max. This is generally the configured maximum
447 * size, except when it already exists and is larger than that. In that case it's the existing size,
448 * since we never want to shrink partitions. */
449
450 if (PARTITION_IS_FOREIGN(p)) {
451 /* Don't allow changing size of partitions not managed by us */
452 assert(p->current_size != UINT64_MAX);
453 return p->current_size;
454 }
455
456 if (p->current_size != UINT64_MAX)
457 return MAX(p->current_size, p->size_max);
458
459 return p->size_max;
460 }
461
462 static uint64_t partition_min_size_with_padding(const Partition *p) {
463 uint64_t sz;
464
465 /* Calculate the disk space we need for this partition plus any free space coming after it. This
466 * takes user configured padding into account as well as any additional whitespace needed to align
467 * the next partition to 4K again. */
468
469 sz = partition_min_size(p);
470
471 if (p->padding_min != UINT64_MAX)
472 sz += p->padding_min;
473
474 if (PARTITION_EXISTS(p)) {
475 /* If the partition wasn't aligned, add extra space so that any we might add will be aligned */
476 assert(p->offset != UINT64_MAX);
477 return round_up_size(p->offset + sz, 4096) - p->offset;
478 }
479
480 /* If this is a new partition we'll place it aligned, hence we just need to round up the required size here */
481 return round_up_size(sz, 4096);
482 }
483
484 static uint64_t free_area_available(const FreeArea *a) {
485 assert(a);
486
487 /* Determines how much of this free area is not allocated yet */
488
489 assert(a->size >= a->allocated);
490 return a->size - a->allocated;
491 }
492
493 static uint64_t free_area_available_for_new_partitions(const FreeArea *a) {
494 uint64_t avail;
495
496 /* Similar to free_area_available(), but takes into account that the required size and padding of the
497 * preceding partition is honoured. */
498
499 avail = free_area_available(a);
500 if (a->after) {
501 uint64_t need, space;
502
503 need = partition_min_size_with_padding(a->after);
504
505 assert(a->after->offset != UINT64_MAX);
506 assert(a->after->current_size != UINT64_MAX);
507
508 space = round_up_size(a->after->offset + a->after->current_size, 4096) - a->after->offset + avail;
509 if (need >= space)
510 return 0;
511
512 return space - need;
513 }
514
515 return avail;
516 }
517
518 static int free_area_compare(FreeArea *const *a, FreeArea *const*b) {
519 return CMP(free_area_available_for_new_partitions(*a),
520 free_area_available_for_new_partitions(*b));
521 }
522
523 static uint64_t charge_size(uint64_t total, uint64_t amount) {
524 uint64_t rounded;
525
526 assert(amount <= total);
527
528 /* Subtract the specified amount from total, rounding up to multiple of 4K if there's room */
529 rounded = round_up_size(amount, 4096);
530 if (rounded >= total)
531 return 0;
532
533 return total - rounded;
534 }
535
536 static uint64_t charge_weight(uint64_t total, uint64_t amount) {
537 assert(amount <= total);
538 return total - amount;
539 }
540
541 static bool context_allocate_partitions(Context *context, uint64_t *ret_largest_free_area) {
542 Partition *p;
543
544 assert(context);
545
546 /* Sort free areas by size, putting smallest first */
547 typesafe_qsort(context->free_areas, context->n_free_areas, free_area_compare);
548
549 /* In any case return size of the largest free area (i.e. not the size of all free areas
550 * combined!) */
551 if (ret_largest_free_area)
552 *ret_largest_free_area =
553 context->n_free_areas == 0 ? 0 :
554 free_area_available_for_new_partitions(context->free_areas[context->n_free_areas-1]);
555
556 /* A simple first-fit algorithm. We return true if we can fit the partitions in, otherwise false. */
557 LIST_FOREACH(partitions, p, context->partitions) {
558 bool fits = false;
559 uint64_t required;
560 FreeArea *a = NULL;
561
562 /* Skip partitions we already dropped or that already exist */
563 if (p->dropped || PARTITION_EXISTS(p))
564 continue;
565
566 /* How much do we need to fit? */
567 required = partition_min_size_with_padding(p);
568 assert(required % 4096 == 0);
569
570 for (size_t i = 0; i < context->n_free_areas; i++) {
571 a = context->free_areas[i];
572
573 if (free_area_available_for_new_partitions(a) >= required) {
574 fits = true;
575 break;
576 }
577 }
578
579 if (!fits)
580 return false; /* 😢 Oh no! We can't fit this partition into any free area! */
581
582 /* Assign the partition to this free area */
583 p->allocated_to_area = a;
584
585 /* Budget the minimal partition size */
586 a->allocated += required;
587 }
588
589 return true;
590 }
591
592 static int context_sum_weights(Context *context, FreeArea *a, uint64_t *ret) {
593 uint64_t weight_sum = 0;
594 Partition *p;
595
596 assert(context);
597 assert(a);
598 assert(ret);
599
600 /* Determine the sum of the weights of all partitions placed in or before the specified free area */
601
602 LIST_FOREACH(partitions, p, context->partitions) {
603 if (p->padding_area != a && p->allocated_to_area != a)
604 continue;
605
606 if (p->weight > UINT64_MAX - weight_sum)
607 goto overflow_sum;
608 weight_sum += p->weight;
609
610 if (p->padding_weight > UINT64_MAX - weight_sum)
611 goto overflow_sum;
612 weight_sum += p->padding_weight;
613 }
614
615 *ret = weight_sum;
616 return 0;
617
618 overflow_sum:
619 return log_error_errno(SYNTHETIC_ERRNO(EOVERFLOW), "Combined weight of partition exceeds unsigned 64bit range, refusing.");
620 }
621
622 static int scale_by_weight(uint64_t value, uint64_t weight, uint64_t weight_sum, uint64_t *ret) {
623 assert(weight_sum >= weight);
624 assert(ret);
625
626 if (weight == 0) {
627 *ret = 0;
628 return 0;
629 }
630
631 if (value > UINT64_MAX / weight)
632 return log_error_errno(SYNTHETIC_ERRNO(EOVERFLOW), "Scaling by weight of partition exceeds unsigned 64bit range, refusing.");
633
634 *ret = value * weight / weight_sum;
635 return 0;
636 }
637
638 typedef enum GrowPartitionPhase {
639 /* The first phase: we charge partitions which need more (according to constraints) than their weight-based share. */
640 PHASE_OVERCHARGE,
641
642 /* The second phase: we charge partitions which need less (according to constraints) than their weight-based share. */
643 PHASE_UNDERCHARGE,
644
645 /* The third phase: we distribute what remains among the remaining partitions, according to the weights */
646 PHASE_DISTRIBUTE,
647 } GrowPartitionPhase;
648
649 static int context_grow_partitions_phase(
650 Context *context,
651 FreeArea *a,
652 GrowPartitionPhase phase,
653 uint64_t *span,
654 uint64_t *weight_sum) {
655
656 Partition *p;
657 int r;
658
659 assert(context);
660 assert(a);
661
662 /* Now let's look at the intended weights and adjust them taking the minimum space assignments into
663 * account. i.e. if a partition has a small weight but a high minimum space value set it should not
664 * get any additional room from the left-overs. Similar, if two partitions have the same weight they
665 * should get the same space if possible, even if one has a smaller minimum size than the other. */
666 LIST_FOREACH(partitions, p, context->partitions) {
667
668 /* Look only at partitions associated with this free area, i.e. immediately
669 * preceding it, or allocated into it */
670 if (p->allocated_to_area != a && p->padding_area != a)
671 continue;
672
673 if (p->new_size == UINT64_MAX) {
674 bool charge = false, try_again = false;
675 uint64_t share, rsz, xsz;
676
677 /* Calculate how much this space this partition needs if everyone would get
678 * the weight based share */
679 r = scale_by_weight(*span, p->weight, *weight_sum, &share);
680 if (r < 0)
681 return r;
682
683 rsz = partition_min_size(p);
684 xsz = partition_max_size(p);
685
686 if (phase == PHASE_OVERCHARGE && rsz > share) {
687 /* This partition needs more than its calculated share. Let's assign
688 * it that, and take this partition out of all calculations and start
689 * again. */
690
691 p->new_size = rsz;
692 charge = try_again = true;
693
694 } else if (phase == PHASE_UNDERCHARGE && xsz != UINT64_MAX && xsz < share) {
695 /* This partition accepts less than its calculated
696 * share. Let's assign it that, and take this partition out
697 * of all calculations and start again. */
698
699 p->new_size = xsz;
700 charge = try_again = true;
701
702 } else if (phase == PHASE_DISTRIBUTE) {
703 /* This partition can accept its calculated share. Let's
704 * assign it. There's no need to restart things here since
705 * assigning this shouldn't impact the shares of the other
706 * partitions. */
707
708 if (PARTITION_IS_FOREIGN(p))
709 /* Never change of foreign partitions (i.e. those we don't manage) */
710 p->new_size = p->current_size;
711 else
712 p->new_size = MAX(round_down_size(share, 4096), rsz);
713
714 charge = true;
715 }
716
717 if (charge) {
718 *span = charge_size(*span, p->new_size);
719 *weight_sum = charge_weight(*weight_sum, p->weight);
720 }
721
722 if (try_again)
723 return 0; /* try again */
724 }
725
726 if (p->new_padding == UINT64_MAX) {
727 bool charge = false, try_again = false;
728 uint64_t share;
729
730 r = scale_by_weight(*span, p->padding_weight, *weight_sum, &share);
731 if (r < 0)
732 return r;
733
734 if (phase == PHASE_OVERCHARGE && p->padding_min != UINT64_MAX && p->padding_min > share) {
735 p->new_padding = p->padding_min;
736 charge = try_again = true;
737 } else if (phase == PHASE_UNDERCHARGE && p->padding_max != UINT64_MAX && p->padding_max < share) {
738 p->new_padding = p->padding_max;
739 charge = try_again = true;
740 } else if (phase == PHASE_DISTRIBUTE) {
741
742 p->new_padding = round_down_size(share, 4096);
743 if (p->padding_min != UINT64_MAX && p->new_padding < p->padding_min)
744 p->new_padding = p->padding_min;
745
746 charge = true;
747 }
748
749 if (charge) {
750 *span = charge_size(*span, p->new_padding);
751 *weight_sum = charge_weight(*weight_sum, p->padding_weight);
752 }
753
754 if (try_again)
755 return 0; /* try again */
756 }
757 }
758
759 return 1; /* done */
760 }
761
762 static int context_grow_partitions_on_free_area(Context *context, FreeArea *a) {
763 uint64_t weight_sum = 0, span;
764 int r;
765
766 assert(context);
767 assert(a);
768
769 r = context_sum_weights(context, a, &weight_sum);
770 if (r < 0)
771 return r;
772
773 /* Let's calculate the total area covered by this free area and the partition before it */
774 span = a->size;
775 if (a->after) {
776 assert(a->after->offset != UINT64_MAX);
777 assert(a->after->current_size != UINT64_MAX);
778
779 span += round_up_size(a->after->offset + a->after->current_size, 4096) - a->after->offset;
780 }
781
782 GrowPartitionPhase phase = PHASE_OVERCHARGE;
783 for (;;) {
784 r = context_grow_partitions_phase(context, a, phase, &span, &weight_sum);
785 if (r < 0)
786 return r;
787 if (r == 0) /* not done yet, re-run this phase */
788 continue;
789
790 if (phase == PHASE_OVERCHARGE)
791 phase = PHASE_UNDERCHARGE;
792 else if (phase == PHASE_UNDERCHARGE)
793 phase = PHASE_DISTRIBUTE;
794 else if (phase == PHASE_DISTRIBUTE)
795 break;
796 }
797
798 /* We still have space left over? Donate to preceding partition if we have one */
799 if (span > 0 && a->after && !PARTITION_IS_FOREIGN(a->after)) {
800 uint64_t m, xsz;
801
802 assert(a->after->new_size != UINT64_MAX);
803 m = a->after->new_size + span;
804
805 xsz = partition_max_size(a->after);
806 if (xsz != UINT64_MAX && m > xsz)
807 m = xsz;
808
809 span = charge_size(span, m - a->after->new_size);
810 a->after->new_size = m;
811 }
812
813 /* What? Even still some space left (maybe because there was no preceding partition, or it had a
814 * size limit), then let's donate it to whoever wants it. */
815 if (span > 0) {
816 Partition *p;
817
818 LIST_FOREACH(partitions, p, context->partitions) {
819 uint64_t m, xsz;
820
821 if (p->allocated_to_area != a)
822 continue;
823
824 if (PARTITION_IS_FOREIGN(p))
825 continue;
826
827 assert(p->new_size != UINT64_MAX);
828 m = p->new_size + span;
829
830 xsz = partition_max_size(p);
831 if (xsz != UINT64_MAX && m > xsz)
832 m = xsz;
833
834 span = charge_size(span, m - p->new_size);
835 p->new_size = m;
836
837 if (span == 0)
838 break;
839 }
840 }
841
842 /* Yuck, still no one? Then make it padding */
843 if (span > 0 && a->after) {
844 assert(a->after->new_padding != UINT64_MAX);
845 a->after->new_padding += span;
846 }
847
848 return 0;
849 }
850
851 static int context_grow_partitions(Context *context) {
852 Partition *p;
853 int r;
854
855 assert(context);
856
857 for (size_t i = 0; i < context->n_free_areas; i++) {
858 r = context_grow_partitions_on_free_area(context, context->free_areas[i]);
859 if (r < 0)
860 return r;
861 }
862
863 /* All existing partitions that have no free space after them can't change size */
864 LIST_FOREACH(partitions, p, context->partitions) {
865 if (p->dropped)
866 continue;
867
868 if (!PARTITION_EXISTS(p) || p->padding_area) {
869 /* The algorithm above must have initialized this already */
870 assert(p->new_size != UINT64_MAX);
871 continue;
872 }
873
874 assert(p->new_size == UINT64_MAX);
875 p->new_size = p->current_size;
876
877 assert(p->new_padding == UINT64_MAX);
878 p->new_padding = p->current_padding;
879 }
880
881 return 0;
882 }
883
884 static void context_place_partitions(Context *context) {
885 uint64_t partno = 0;
886 Partition *p;
887
888 assert(context);
889
890 /* Determine next partition number to assign */
891 LIST_FOREACH(partitions, p, context->partitions) {
892 if (!PARTITION_EXISTS(p))
893 continue;
894
895 assert(p->partno != UINT64_MAX);
896 if (p->partno >= partno)
897 partno = p->partno + 1;
898 }
899
900 for (size_t i = 0; i < context->n_free_areas; i++) {
901 FreeArea *a = context->free_areas[i];
902 uint64_t start, left;
903
904 if (a->after) {
905 assert(a->after->offset != UINT64_MAX);
906 assert(a->after->new_size != UINT64_MAX);
907 assert(a->after->new_padding != UINT64_MAX);
908
909 start = a->after->offset + a->after->new_size + a->after->new_padding;
910 } else
911 start = context->start;
912
913 start = round_up_size(start, 4096);
914 left = a->size;
915
916 LIST_FOREACH(partitions, p, context->partitions) {
917 if (p->allocated_to_area != a)
918 continue;
919
920 p->offset = start;
921 p->partno = partno++;
922
923 assert(left >= p->new_size);
924 start += p->new_size;
925 left -= p->new_size;
926
927 assert(left >= p->new_padding);
928 start += p->new_padding;
929 left -= p->new_padding;
930 }
931 }
932 }
933
934 static int config_parse_type(
935 const char *unit,
936 const char *filename,
937 unsigned line,
938 const char *section,
939 unsigned section_line,
940 const char *lvalue,
941 int ltype,
942 const char *rvalue,
943 void *data,
944 void *userdata) {
945
946 sd_id128_t *type_uuid = data;
947 int r;
948
949 assert(rvalue);
950 assert(type_uuid);
951
952 r = gpt_partition_type_uuid_from_string(rvalue, type_uuid);
953 if (r < 0)
954 return log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse partition type: %s", rvalue);
955
956 return 0;
957 }
958
959 static int config_parse_label(
960 const char *unit,
961 const char *filename,
962 unsigned line,
963 const char *section,
964 unsigned section_line,
965 const char *lvalue,
966 int ltype,
967 const char *rvalue,
968 void *data,
969 void *userdata) {
970
971 _cleanup_free_ char *resolved = NULL;
972 char **label = data;
973 int r;
974
975 assert(rvalue);
976 assert(label);
977
978 /* Nota bene: the empty label is a totally valid one. Let's hence not follow our usual rule of
979 * assigning the empty string to reset to default here, but really accept it as label to set. */
980
981 r = specifier_printf(rvalue, GPT_LABEL_MAX, system_and_tmp_specifier_table, arg_root, NULL, &resolved);
982 if (r < 0) {
983 log_syntax(unit, LOG_WARNING, filename, line, r,
984 "Failed to expand specifiers in Label=, ignoring: %s", rvalue);
985 return 0;
986 }
987
988 if (!utf8_is_valid(resolved)) {
989 log_syntax(unit, LOG_WARNING, filename, line, 0,
990 "Partition label not valid UTF-8, ignoring: %s", rvalue);
991 return 0;
992 }
993
994 r = gpt_partition_label_valid(resolved);
995 if (r < 0) {
996 log_syntax(unit, LOG_WARNING, filename, line, r,
997 "Failed to check if string is valid as GPT partition label, ignoring: \"%s\" (from \"%s\")",
998 resolved, rvalue);
999 return 0;
1000 }
1001 if (!r) {
1002 log_syntax(unit, LOG_WARNING, filename, line, 0,
1003 "Partition label too long for GPT table, ignoring: \"%s\" (from \"%s\")",
1004 resolved, rvalue);
1005 return 0;
1006 }
1007
1008 free_and_replace(*label, resolved);
1009 return 0;
1010 }
1011
1012 static int config_parse_weight(
1013 const char *unit,
1014 const char *filename,
1015 unsigned line,
1016 const char *section,
1017 unsigned section_line,
1018 const char *lvalue,
1019 int ltype,
1020 const char *rvalue,
1021 void *data,
1022 void *userdata) {
1023
1024 uint32_t *priority = data, v;
1025 int r;
1026
1027 assert(rvalue);
1028 assert(priority);
1029
1030 r = safe_atou32(rvalue, &v);
1031 if (r < 0) {
1032 log_syntax(unit, LOG_WARNING, filename, line, r,
1033 "Failed to parse weight value, ignoring: %s", rvalue);
1034 return 0;
1035 }
1036
1037 if (v > 1000U*1000U) {
1038 log_syntax(unit, LOG_WARNING, filename, line, 0,
1039 "Weight needs to be in range 0…10000000, ignoring: %" PRIu32, v);
1040 return 0;
1041 }
1042
1043 *priority = v;
1044 return 0;
1045 }
1046
1047 static int config_parse_size4096(
1048 const char *unit,
1049 const char *filename,
1050 unsigned line,
1051 const char *section,
1052 unsigned section_line,
1053 const char *lvalue,
1054 int ltype,
1055 const char *rvalue,
1056 void *data,
1057 void *userdata) {
1058
1059 uint64_t *sz = data, parsed;
1060 int r;
1061
1062 assert(rvalue);
1063 assert(data);
1064
1065 r = parse_size(rvalue, 1024, &parsed);
1066 if (r < 0)
1067 return log_syntax(unit, LOG_ERR, filename, line, r,
1068 "Failed to parse size value: %s", rvalue);
1069
1070 if (ltype > 0)
1071 *sz = round_up_size(parsed, 4096);
1072 else if (ltype < 0)
1073 *sz = round_down_size(parsed, 4096);
1074 else
1075 *sz = parsed;
1076
1077 if (*sz != parsed)
1078 log_syntax(unit, LOG_NOTICE, filename, line, r, "Rounded %s= size %" PRIu64 " → %" PRIu64 ", a multiple of 4096.", lvalue, parsed, *sz);
1079
1080 return 0;
1081 }
1082
1083 static int config_parse_fstype(
1084 const char *unit,
1085 const char *filename,
1086 unsigned line,
1087 const char *section,
1088 unsigned section_line,
1089 const char *lvalue,
1090 int ltype,
1091 const char *rvalue,
1092 void *data,
1093 void *userdata) {
1094
1095 char **fstype = data;
1096
1097 assert(rvalue);
1098 assert(data);
1099
1100 if (!filename_is_valid(rvalue))
1101 return log_syntax(unit, LOG_ERR, filename, line, 0,
1102 "File system type is not valid, refusing: %s", rvalue);
1103
1104 return free_and_strdup_warn(fstype, rvalue);
1105 }
1106
1107 static int config_parse_copy_files(
1108 const char *unit,
1109 const char *filename,
1110 unsigned line,
1111 const char *section,
1112 unsigned section_line,
1113 const char *lvalue,
1114 int ltype,
1115 const char *rvalue,
1116 void *data,
1117 void *userdata) {
1118
1119 _cleanup_free_ char *source = NULL, *buffer = NULL, *resolved_source = NULL, *resolved_target = NULL;
1120 const char *p = rvalue, *target;
1121 Partition *partition = data;
1122 int r;
1123
1124 assert(rvalue);
1125 assert(partition);
1126
1127 r = extract_first_word(&p, &source, ":", EXTRACT_CUNESCAPE|EXTRACT_DONT_COALESCE_SEPARATORS);
1128 if (r < 0)
1129 return log_syntax(unit, LOG_ERR, filename, line, r, "Failed to extract source path: %s", rvalue);
1130 if (r == 0) {
1131 log_syntax(unit, LOG_WARNING, filename, line, 0, "No argument specified: %s", rvalue);
1132 return 0;
1133 }
1134
1135 r = extract_first_word(&p, &buffer, ":", EXTRACT_CUNESCAPE|EXTRACT_DONT_COALESCE_SEPARATORS);
1136 if (r < 0)
1137 return log_syntax(unit, LOG_ERR, filename, line, r, "Failed to extract target path: %s", rvalue);
1138 if (r == 0)
1139 target = source; /* No target, then it's the same as the source */
1140 else
1141 target = buffer;
1142
1143 if (!isempty(p))
1144 return log_syntax(unit, LOG_ERR, filename, line, SYNTHETIC_ERRNO(EINVAL), "Too many arguments: %s", rvalue);
1145
1146 r = specifier_printf(source, PATH_MAX-1, system_and_tmp_specifier_table, arg_root, NULL, &resolved_source);
1147 if (r < 0) {
1148 log_syntax(unit, LOG_WARNING, filename, line, r,
1149 "Failed to expand specifiers in CopyFiles= source, ignoring: %s", rvalue);
1150 return 0;
1151 }
1152
1153 r = path_simplify_and_warn(resolved_source, PATH_CHECK_ABSOLUTE, unit, filename, line, lvalue);
1154 if (r < 0)
1155 return 0;
1156
1157 r = specifier_printf(target, PATH_MAX-1, system_and_tmp_specifier_table, arg_root, NULL, &resolved_target);
1158 if (r < 0) {
1159 log_syntax(unit, LOG_WARNING, filename, line, r,
1160 "Failed to expand specifiers in CopyFiles= target, ignoring: %s", resolved_target);
1161 return 0;
1162 }
1163
1164 r = path_simplify_and_warn(resolved_target, PATH_CHECK_ABSOLUTE, unit, filename, line, lvalue);
1165 if (r < 0)
1166 return 0;
1167
1168 r = strv_consume_pair(&partition->copy_files, TAKE_PTR(resolved_source), TAKE_PTR(resolved_target));
1169 if (r < 0)
1170 return log_oom();
1171
1172 return 0;
1173 }
1174
1175 static int config_parse_copy_blocks(
1176 const char *unit,
1177 const char *filename,
1178 unsigned line,
1179 const char *section,
1180 unsigned section_line,
1181 const char *lvalue,
1182 int ltype,
1183 const char *rvalue,
1184 void *data,
1185 void *userdata) {
1186
1187 _cleanup_free_ char *d = NULL;
1188 Partition *partition = data;
1189 int r;
1190
1191 assert(rvalue);
1192 assert(partition);
1193
1194 if (isempty(rvalue)) {
1195 partition->copy_blocks_path = mfree(partition->copy_blocks_path);
1196 partition->copy_blocks_auto = false;
1197 return 0;
1198 }
1199
1200 if (streq(rvalue, "auto")) {
1201 partition->copy_blocks_path = mfree(partition->copy_blocks_path);
1202 partition->copy_blocks_auto = true;
1203 return 0;
1204 }
1205
1206 r = specifier_printf(rvalue, PATH_MAX-1, system_and_tmp_specifier_table, arg_root, NULL, &d);
1207 if (r < 0) {
1208 log_syntax(unit, LOG_WARNING, filename, line, r,
1209 "Failed to expand specifiers in CopyBlocks= source path, ignoring: %s", rvalue);
1210 return 0;
1211 }
1212
1213 r = path_simplify_and_warn(d, PATH_CHECK_ABSOLUTE, unit, filename, line, lvalue);
1214 if (r < 0)
1215 return 0;
1216
1217 free_and_replace(partition->copy_blocks_path, d);
1218 partition->copy_blocks_auto = false;
1219 return 0;
1220 }
1221
1222 static int config_parse_make_dirs(
1223 const char *unit,
1224 const char *filename,
1225 unsigned line,
1226 const char *section,
1227 unsigned section_line,
1228 const char *lvalue,
1229 int ltype,
1230 const char *rvalue,
1231 void *data,
1232 void *userdata) {
1233
1234 Partition *partition = data;
1235 const char *p = rvalue;
1236 int r;
1237
1238 assert(rvalue);
1239 assert(partition);
1240
1241 for (;;) {
1242 _cleanup_free_ char *word = NULL, *d = NULL;
1243
1244 r = extract_first_word(&p, &word, NULL, EXTRACT_UNQUOTE);
1245 if (r == -ENOMEM)
1246 return log_oom();
1247 if (r < 0) {
1248 log_syntax(unit, LOG_WARNING, filename, line, r, "Invalid syntax, ignoring: %s", rvalue);
1249 return 0;
1250 }
1251 if (r == 0)
1252 return 0;
1253
1254 r = specifier_printf(word, PATH_MAX-1, system_and_tmp_specifier_table, arg_root, NULL, &d);
1255 if (r < 0) {
1256 log_syntax(unit, LOG_WARNING, filename, line, r,
1257 "Failed to expand specifiers in MakeDirectories= parameter, ignoring: %s", word);
1258 continue;
1259 }
1260
1261 r = path_simplify_and_warn(d, PATH_CHECK_ABSOLUTE, unit, filename, line, lvalue);
1262 if (r < 0)
1263 continue;
1264
1265 r = strv_consume(&partition->make_directories, TAKE_PTR(d));
1266 if (r < 0)
1267 return log_oom();
1268 }
1269 }
1270
1271 static DEFINE_CONFIG_PARSE_ENUM_WITH_DEFAULT(config_parse_encrypt, encrypt_mode, EncryptMode, ENCRYPT_OFF, "Invalid encryption mode");
1272
1273 static int config_parse_gpt_flags(
1274 const char *unit,
1275 const char *filename,
1276 unsigned line,
1277 const char *section,
1278 unsigned section_line,
1279 const char *lvalue,
1280 int ltype,
1281 const char *rvalue,
1282 void *data,
1283 void *userdata) {
1284
1285 uint64_t *gpt_flags = data;
1286 int r;
1287
1288 assert(rvalue);
1289 assert(gpt_flags);
1290
1291 r = safe_atou64(rvalue, gpt_flags);
1292 if (r < 0) {
1293 log_syntax(unit, LOG_WARNING, filename, line, r,
1294 "Failed to parse Flags= value, ignoring: %s", rvalue);
1295 return 0;
1296 }
1297
1298 return 0;
1299 }
1300
1301 static int partition_read_definition(Partition *p, const char *path) {
1302
1303 ConfigTableItem table[] = {
1304 { "Partition", "Type", config_parse_type, 0, &p->type_uuid },
1305 { "Partition", "Label", config_parse_label, 0, &p->new_label },
1306 { "Partition", "UUID", config_parse_id128, 0, &p->new_uuid },
1307 { "Partition", "Priority", config_parse_int32, 0, &p->priority },
1308 { "Partition", "Weight", config_parse_weight, 0, &p->weight },
1309 { "Partition", "PaddingWeight", config_parse_weight, 0, &p->padding_weight },
1310 { "Partition", "SizeMinBytes", config_parse_size4096, 1, &p->size_min },
1311 { "Partition", "SizeMaxBytes", config_parse_size4096, -1, &p->size_max },
1312 { "Partition", "PaddingMinBytes", config_parse_size4096, 1, &p->padding_min },
1313 { "Partition", "PaddingMaxBytes", config_parse_size4096, -1, &p->padding_max },
1314 { "Partition", "FactoryReset", config_parse_bool, 0, &p->factory_reset },
1315 { "Partition", "CopyBlocks", config_parse_copy_blocks, 0, p },
1316 { "Partition", "Format", config_parse_fstype, 0, &p->format },
1317 { "Partition", "CopyFiles", config_parse_copy_files, 0, p },
1318 { "Partition", "MakeDirectories", config_parse_make_dirs, 0, p },
1319 { "Partition", "Encrypt", config_parse_encrypt, 0, &p->encrypt },
1320 { "Partition", "Flags", config_parse_gpt_flags, 0, &p->gpt_flags },
1321 { "Partition", "ReadOnly", config_parse_tristate, 0, &p->read_only },
1322 { "Partition", "NoAuto", config_parse_tristate, 0, &p->no_auto },
1323 { "Partition", "GrowFileSystem", config_parse_tristate, 0, &p->growfs },
1324 {}
1325 };
1326 int r;
1327
1328 r = config_parse(NULL, path, NULL,
1329 "Partition\0",
1330 config_item_table_lookup, table,
1331 CONFIG_PARSE_WARN,
1332 p,
1333 NULL);
1334 if (r < 0)
1335 return r;
1336
1337 if (p->size_min != UINT64_MAX && p->size_max != UINT64_MAX && p->size_min > p->size_max)
1338 return log_syntax(NULL, LOG_ERR, path, 1, SYNTHETIC_ERRNO(EINVAL),
1339 "SizeMinBytes= larger than SizeMaxBytes=, refusing.");
1340
1341 if (p->padding_min != UINT64_MAX && p->padding_max != UINT64_MAX && p->padding_min > p->padding_max)
1342 return log_syntax(NULL, LOG_ERR, path, 1, SYNTHETIC_ERRNO(EINVAL),
1343 "PaddingMinBytes= larger than PaddingMaxBytes=, refusing.");
1344
1345 if (sd_id128_is_null(p->type_uuid))
1346 return log_syntax(NULL, LOG_ERR, path, 1, SYNTHETIC_ERRNO(EINVAL),
1347 "Type= not defined, refusing.");
1348
1349 if ((p->copy_blocks_path || p->copy_blocks_auto) &&
1350 (p->format || !strv_isempty(p->copy_files) || !strv_isempty(p->make_directories)))
1351 return log_syntax(NULL, LOG_ERR, path, 1, SYNTHETIC_ERRNO(EINVAL),
1352 "Format=/CopyFiles=/MakeDirectories= and CopyBlocks= cannot be combined, refusing.");
1353
1354 if ((!strv_isempty(p->copy_files) || !strv_isempty(p->make_directories)) && streq_ptr(p->format, "swap"))
1355 return log_syntax(NULL, LOG_ERR, path, 1, SYNTHETIC_ERRNO(EINVAL),
1356 "Format=swap and CopyFiles= cannot be combined, refusing.");
1357
1358 if (!p->format && (!strv_isempty(p->copy_files) || !strv_isempty(p->make_directories) || (p->encrypt != ENCRYPT_OFF && !(p->copy_blocks_path || p->copy_blocks_auto)))) {
1359 /* Pick "ext4" as file system if we are configured to copy files or encrypt the device */
1360 p->format = strdup("ext4");
1361 if (!p->format)
1362 return log_oom();
1363 }
1364
1365 /* Verity partitions are read only, let's imply the RO flag hence, unless explicitly configured otherwise. */
1366 if ((gpt_partition_type_is_root_verity(p->type_uuid) ||
1367 gpt_partition_type_is_usr_verity(p->type_uuid)) &&
1368 p->read_only < 0)
1369 p->read_only = true;
1370
1371 /* Default to "growfs" on, unless read-only */
1372 if (gpt_partition_type_knows_growfs(p->type_uuid) &&
1373 p->read_only <= 0)
1374 p->growfs = true;
1375
1376 return 0;
1377 }
1378
1379 static int context_read_definitions(
1380 Context *context,
1381 const char *directory,
1382 const char *root) {
1383
1384 _cleanup_strv_free_ char **files = NULL;
1385 Partition *last = NULL;
1386 char **f;
1387 int r;
1388
1389 assert(context);
1390
1391 if (directory)
1392 r = conf_files_list_strv(&files, ".conf", NULL, CONF_FILES_REGULAR|CONF_FILES_FILTER_MASKED, (const char**) STRV_MAKE(directory));
1393 else
1394 r = conf_files_list_strv(&files, ".conf", root, CONF_FILES_REGULAR|CONF_FILES_FILTER_MASKED, (const char**) CONF_PATHS_STRV("repart.d"));
1395 if (r < 0)
1396 return log_error_errno(r, "Failed to enumerate *.conf files: %m");
1397
1398 STRV_FOREACH(f, files) {
1399 _cleanup_(partition_freep) Partition *p = NULL;
1400
1401 p = partition_new();
1402 if (!p)
1403 return log_oom();
1404
1405 p->definition_path = strdup(*f);
1406 if (!p->definition_path)
1407 return log_oom();
1408
1409 r = partition_read_definition(p, *f);
1410 if (r < 0)
1411 return r;
1412
1413 LIST_INSERT_AFTER(partitions, context->partitions, last, p);
1414 last = TAKE_PTR(p);
1415 context->n_partitions++;
1416 }
1417
1418 return 0;
1419 }
1420
1421 DEFINE_TRIVIAL_CLEANUP_FUNC_FULL(struct fdisk_context*, fdisk_unref_context, NULL);
1422 DEFINE_TRIVIAL_CLEANUP_FUNC_FULL(struct fdisk_partition*, fdisk_unref_partition, NULL);
1423 DEFINE_TRIVIAL_CLEANUP_FUNC_FULL(struct fdisk_parttype*, fdisk_unref_parttype, NULL);
1424 DEFINE_TRIVIAL_CLEANUP_FUNC_FULL(struct fdisk_table*, fdisk_unref_table, NULL);
1425
1426 static int determine_current_padding(
1427 struct fdisk_context *c,
1428 struct fdisk_table *t,
1429 struct fdisk_partition *p,
1430 uint64_t *ret) {
1431
1432 size_t n_partitions;
1433 uint64_t offset, next = UINT64_MAX;
1434
1435 assert(c);
1436 assert(t);
1437 assert(p);
1438
1439 if (!fdisk_partition_has_end(p))
1440 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Partition has no end!");
1441
1442 offset = fdisk_partition_get_end(p);
1443 assert(offset < UINT64_MAX / 512);
1444 offset *= 512;
1445
1446 n_partitions = fdisk_table_get_nents(t);
1447 for (size_t i = 0; i < n_partitions; i++) {
1448 struct fdisk_partition *q;
1449 uint64_t start;
1450
1451 q = fdisk_table_get_partition(t, i);
1452 if (!q)
1453 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Failed to read partition metadata: %m");
1454
1455 if (fdisk_partition_is_used(q) <= 0)
1456 continue;
1457
1458 if (!fdisk_partition_has_start(q))
1459 continue;
1460
1461 start = fdisk_partition_get_start(q);
1462 assert(start < UINT64_MAX / 512);
1463 start *= 512;
1464
1465 if (start >= offset && (next == UINT64_MAX || next > start))
1466 next = start;
1467 }
1468
1469 if (next == UINT64_MAX) {
1470 /* No later partition? In that case check the end of the usable area */
1471 next = fdisk_get_last_lba(c);
1472 assert(next < UINT64_MAX);
1473 next++; /* The last LBA is one sector before the end */
1474
1475 assert(next < UINT64_MAX / 512);
1476 next *= 512;
1477
1478 if (offset > next)
1479 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Partition end beyond disk end.");
1480 }
1481
1482 assert(next >= offset);
1483 offset = round_up_size(offset, 4096);
1484 next = round_down_size(next, 4096);
1485
1486 if (next >= offset) /* Check again, rounding might have fucked things up */
1487 *ret = next - offset;
1488 else
1489 *ret = 0;
1490
1491 return 0;
1492 }
1493
1494 static int fdisk_ask_cb(struct fdisk_context *c, struct fdisk_ask *ask, void *data) {
1495 _cleanup_free_ char *ids = NULL;
1496 int r;
1497
1498 if (fdisk_ask_get_type(ask) != FDISK_ASKTYPE_STRING)
1499 return -EINVAL;
1500
1501 ids = new(char, ID128_UUID_STRING_MAX);
1502 if (!ids)
1503 return -ENOMEM;
1504
1505 r = fdisk_ask_string_set_result(ask, id128_to_uuid_string(*(sd_id128_t*) data, ids));
1506 if (r < 0)
1507 return r;
1508
1509 TAKE_PTR(ids);
1510 return 0;
1511 }
1512
1513 static int fdisk_set_disklabel_id_by_uuid(struct fdisk_context *c, sd_id128_t id) {
1514 int r;
1515
1516 r = fdisk_set_ask(c, fdisk_ask_cb, &id);
1517 if (r < 0)
1518 return r;
1519
1520 r = fdisk_set_disklabel_id(c);
1521 if (r < 0)
1522 return r;
1523
1524 return fdisk_set_ask(c, NULL, NULL);
1525 }
1526
1527 static int derive_uuid(sd_id128_t base, const char *token, sd_id128_t *ret) {
1528 union {
1529 unsigned char md[SHA256_DIGEST_LENGTH];
1530 sd_id128_t id;
1531 } result;
1532
1533 assert(token);
1534 assert(ret);
1535
1536 /* Derive a new UUID from the specified UUID in a stable and reasonably safe way. Specifically, we
1537 * calculate the HMAC-SHA256 of the specified token string, keyed by the supplied base (typically the
1538 * machine ID). We use the machine ID as key (and not as cleartext!) of the HMAC operation since it's
1539 * the machine ID we don't want to leak. */
1540
1541 if (!HMAC(EVP_sha256(),
1542 &base, sizeof(base),
1543 (const unsigned char*) token, strlen(token),
1544 result.md, NULL))
1545 return log_error_errno(SYNTHETIC_ERRNO(ENOTRECOVERABLE), "HMAC-SHA256 calculation failed.");
1546
1547 /* Take the first half, mark it as v4 UUID */
1548 assert_cc(sizeof(result.md) == sizeof(result.id) * 2);
1549 *ret = id128_make_v4_uuid(result.id);
1550 return 0;
1551 }
1552
1553 static int context_load_partition_table(
1554 Context *context,
1555 const char *node,
1556 int *backing_fd) {
1557
1558 _cleanup_(fdisk_unref_contextp) struct fdisk_context *c = NULL;
1559 _cleanup_(fdisk_unref_tablep) struct fdisk_table *t = NULL;
1560 uint64_t left_boundary = UINT64_MAX, first_lba, last_lba, nsectors;
1561 _cleanup_free_ char *disk_uuid_string = NULL;
1562 bool from_scratch = false;
1563 sd_id128_t disk_uuid;
1564 size_t n_partitions;
1565 int r;
1566
1567 assert(context);
1568 assert(node);
1569 assert(backing_fd);
1570 assert(!context->fdisk_context);
1571 assert(!context->free_areas);
1572 assert(context->start == UINT64_MAX);
1573 assert(context->end == UINT64_MAX);
1574 assert(context->total == UINT64_MAX);
1575
1576 c = fdisk_new_context();
1577 if (!c)
1578 return log_oom();
1579
1580 /* libfdisk doesn't have an API to operate on arbitrary fds, hence reopen the fd going via the
1581 * /proc/self/fd/ magic path if we have an existing fd. Open the original file otherwise. */
1582 if (*backing_fd < 0)
1583 r = fdisk_assign_device(c, node, arg_dry_run);
1584 else
1585 r = fdisk_assign_device(c, FORMAT_PROC_FD_PATH(*backing_fd), arg_dry_run);
1586 if (r == -EINVAL && arg_size_auto) {
1587 struct stat st;
1588
1589 /* libfdisk returns EINVAL if opening a file of size zero. Let's check for that, and accept
1590 * it if automatic sizing is requested. */
1591
1592 if (*backing_fd < 0)
1593 r = stat(node, &st);
1594 else
1595 r = fstat(*backing_fd, &st);
1596 if (r < 0)
1597 return log_error_errno(errno, "Failed to stat block device '%s': %m", node);
1598
1599 if (S_ISREG(st.st_mode) && st.st_size == 0)
1600 return /* from_scratch = */ true;
1601
1602 r = -EINVAL;
1603 }
1604 if (r < 0)
1605 return log_error_errno(r, "Failed to open device '%s': %m", node);
1606
1607 if (*backing_fd < 0) {
1608 /* If we have no fd referencing the device yet, make a copy of the fd now, so that we have one */
1609 *backing_fd = fcntl(fdisk_get_devfd(c), F_DUPFD_CLOEXEC, 3);
1610 if (*backing_fd < 0)
1611 return log_error_errno(errno, "Failed to duplicate fdisk fd: %m");
1612 }
1613
1614 /* Tell udev not to interfere while we are processing the device */
1615 if (flock(fdisk_get_devfd(c), arg_dry_run ? LOCK_SH : LOCK_EX) < 0)
1616 return log_error_errno(errno, "Failed to lock block device: %m");
1617
1618 switch (arg_empty) {
1619
1620 case EMPTY_REFUSE:
1621 /* Refuse empty disks, insist on an existing GPT partition table */
1622 if (!fdisk_is_labeltype(c, FDISK_DISKLABEL_GPT))
1623 return log_notice_errno(SYNTHETIC_ERRNO(EHWPOISON), "Disk %s has no GPT disk label, not repartitioning.", node);
1624
1625 break;
1626
1627 case EMPTY_REQUIRE:
1628 /* Require an empty disk, refuse any existing partition table */
1629 r = fdisk_has_label(c);
1630 if (r < 0)
1631 return log_error_errno(r, "Failed to determine whether disk %s has a disk label: %m", node);
1632 if (r > 0)
1633 return log_notice_errno(SYNTHETIC_ERRNO(EHWPOISON), "Disk %s already has a disk label, refusing.", node);
1634
1635 from_scratch = true;
1636 break;
1637
1638 case EMPTY_ALLOW:
1639 /* Allow both an empty disk and an existing partition table, but only GPT */
1640 r = fdisk_has_label(c);
1641 if (r < 0)
1642 return log_error_errno(r, "Failed to determine whether disk %s has a disk label: %m", node);
1643 if (r > 0) {
1644 if (!fdisk_is_labeltype(c, FDISK_DISKLABEL_GPT))
1645 return log_notice_errno(SYNTHETIC_ERRNO(EHWPOISON), "Disk %s has non-GPT disk label, not repartitioning.", node);
1646 } else
1647 from_scratch = true;
1648
1649 break;
1650
1651 case EMPTY_FORCE:
1652 case EMPTY_CREATE:
1653 /* Always reinitiaize the disk, don't consider what there was on the disk before */
1654 from_scratch = true;
1655 break;
1656 }
1657
1658 if (from_scratch) {
1659 r = fdisk_create_disklabel(c, "gpt");
1660 if (r < 0)
1661 return log_error_errno(r, "Failed to create GPT disk label: %m");
1662
1663 r = derive_uuid(context->seed, "disk-uuid", &disk_uuid);
1664 if (r < 0)
1665 return log_error_errno(r, "Failed to acquire disk GPT uuid: %m");
1666
1667 r = fdisk_set_disklabel_id_by_uuid(c, disk_uuid);
1668 if (r < 0)
1669 return log_error_errno(r, "Failed to set GPT disk label: %m");
1670
1671 goto add_initial_free_area;
1672 }
1673
1674 r = fdisk_get_disklabel_id(c, &disk_uuid_string);
1675 if (r < 0)
1676 return log_error_errno(r, "Failed to get current GPT disk label UUID: %m");
1677
1678 r = sd_id128_from_string(disk_uuid_string, &disk_uuid);
1679 if (r < 0)
1680 return log_error_errno(r, "Failed to parse current GPT disk label UUID: %m");
1681
1682 if (sd_id128_is_null(disk_uuid)) {
1683 r = derive_uuid(context->seed, "disk-uuid", &disk_uuid);
1684 if (r < 0)
1685 return log_error_errno(r, "Failed to acquire disk GPT uuid: %m");
1686
1687 r = fdisk_set_disklabel_id(c);
1688 if (r < 0)
1689 return log_error_errno(r, "Failed to set GPT disk label: %m");
1690 }
1691
1692 r = fdisk_get_partitions(c, &t);
1693 if (r < 0)
1694 return log_error_errno(r, "Failed to acquire partition table: %m");
1695
1696 n_partitions = fdisk_table_get_nents(t);
1697 for (size_t i = 0; i < n_partitions; i++) {
1698 _cleanup_free_ char *label_copy = NULL;
1699 Partition *pp, *last = NULL;
1700 struct fdisk_partition *p;
1701 struct fdisk_parttype *pt;
1702 const char *pts, *ids, *label;
1703 uint64_t sz, start;
1704 bool found = false;
1705 sd_id128_t ptid, id;
1706 size_t partno;
1707
1708 p = fdisk_table_get_partition(t, i);
1709 if (!p)
1710 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Failed to read partition metadata: %m");
1711
1712 if (fdisk_partition_is_used(p) <= 0)
1713 continue;
1714
1715 if (fdisk_partition_has_start(p) <= 0 ||
1716 fdisk_partition_has_size(p) <= 0 ||
1717 fdisk_partition_has_partno(p) <= 0)
1718 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Found a partition without a position, size or number.");
1719
1720 pt = fdisk_partition_get_type(p);
1721 if (!pt)
1722 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Failed to acquire type of partition: %m");
1723
1724 pts = fdisk_parttype_get_string(pt);
1725 if (!pts)
1726 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Failed to acquire type of partition as string: %m");
1727
1728 r = sd_id128_from_string(pts, &ptid);
1729 if (r < 0)
1730 return log_error_errno(r, "Failed to parse partition type UUID %s: %m", pts);
1731
1732 ids = fdisk_partition_get_uuid(p);
1733 if (!ids)
1734 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Found a partition without a UUID.");
1735
1736 r = sd_id128_from_string(ids, &id);
1737 if (r < 0)
1738 return log_error_errno(r, "Failed to parse partition UUID %s: %m", ids);
1739
1740 label = fdisk_partition_get_name(p);
1741 if (!isempty(label)) {
1742 label_copy = strdup(label);
1743 if (!label_copy)
1744 return log_oom();
1745 }
1746
1747 sz = fdisk_partition_get_size(p);
1748 assert_se(sz <= UINT64_MAX/512);
1749 sz *= 512;
1750
1751 start = fdisk_partition_get_start(p);
1752 assert_se(start <= UINT64_MAX/512);
1753 start *= 512;
1754
1755 partno = fdisk_partition_get_partno(p);
1756
1757 if (left_boundary == UINT64_MAX || left_boundary > start)
1758 left_boundary = start;
1759
1760 /* Assign this existing partition to the first partition of the right type that doesn't have
1761 * an existing one assigned yet. */
1762 LIST_FOREACH(partitions, pp, context->partitions) {
1763 last = pp;
1764
1765 if (!sd_id128_equal(pp->type_uuid, ptid))
1766 continue;
1767
1768 if (!pp->current_partition) {
1769 pp->current_uuid = id;
1770 pp->current_size = sz;
1771 pp->offset = start;
1772 pp->partno = partno;
1773 pp->current_label = TAKE_PTR(label_copy);
1774
1775 pp->current_partition = p;
1776 fdisk_ref_partition(p);
1777
1778 r = determine_current_padding(c, t, p, &pp->current_padding);
1779 if (r < 0)
1780 return r;
1781
1782 if (pp->current_padding > 0) {
1783 r = context_add_free_area(context, pp->current_padding, pp);
1784 if (r < 0)
1785 return r;
1786 }
1787
1788 found = true;
1789 break;
1790 }
1791 }
1792
1793 /* If we have no matching definition, create a new one. */
1794 if (!found) {
1795 _cleanup_(partition_freep) Partition *np = NULL;
1796
1797 np = partition_new();
1798 if (!np)
1799 return log_oom();
1800
1801 np->current_uuid = id;
1802 np->type_uuid = ptid;
1803 np->current_size = sz;
1804 np->offset = start;
1805 np->partno = partno;
1806 np->current_label = TAKE_PTR(label_copy);
1807
1808 np->current_partition = p;
1809 fdisk_ref_partition(p);
1810
1811 r = determine_current_padding(c, t, p, &np->current_padding);
1812 if (r < 0)
1813 return r;
1814
1815 if (np->current_padding > 0) {
1816 r = context_add_free_area(context, np->current_padding, np);
1817 if (r < 0)
1818 return r;
1819 }
1820
1821 LIST_INSERT_AFTER(partitions, context->partitions, last, TAKE_PTR(np));
1822 context->n_partitions++;
1823 }
1824 }
1825
1826 add_initial_free_area:
1827 nsectors = fdisk_get_nsectors(c);
1828 assert(nsectors <= UINT64_MAX/512);
1829 nsectors *= 512;
1830
1831 first_lba = fdisk_get_first_lba(c);
1832 assert(first_lba <= UINT64_MAX/512);
1833 first_lba *= 512;
1834
1835 last_lba = fdisk_get_last_lba(c);
1836 assert(last_lba < UINT64_MAX);
1837 last_lba++;
1838 assert(last_lba <= UINT64_MAX/512);
1839 last_lba *= 512;
1840
1841 assert(last_lba >= first_lba);
1842
1843 if (left_boundary == UINT64_MAX) {
1844 /* No partitions at all? Then the whole disk is up for grabs. */
1845
1846 first_lba = round_up_size(first_lba, 4096);
1847 last_lba = round_down_size(last_lba, 4096);
1848
1849 if (last_lba > first_lba) {
1850 r = context_add_free_area(context, last_lba - first_lba, NULL);
1851 if (r < 0)
1852 return r;
1853 }
1854 } else {
1855 /* Add space left of first partition */
1856 assert(left_boundary >= first_lba);
1857
1858 first_lba = round_up_size(first_lba, 4096);
1859 left_boundary = round_down_size(left_boundary, 4096);
1860 last_lba = round_down_size(last_lba, 4096);
1861
1862 if (left_boundary > first_lba) {
1863 r = context_add_free_area(context, left_boundary - first_lba, NULL);
1864 if (r < 0)
1865 return r;
1866 }
1867 }
1868
1869 context->start = first_lba;
1870 context->end = last_lba;
1871 context->total = nsectors;
1872 context->fdisk_context = TAKE_PTR(c);
1873
1874 return from_scratch;
1875 }
1876
1877 static void context_unload_partition_table(Context *context) {
1878 Partition *p, *next;
1879
1880 assert(context);
1881
1882 LIST_FOREACH_SAFE(partitions, p, next, context->partitions) {
1883
1884 /* Entirely remove partitions that have no configuration */
1885 if (PARTITION_IS_FOREIGN(p)) {
1886 partition_unlink_and_free(context, p);
1887 continue;
1888 }
1889
1890 /* Otherwise drop all data we read off the block device and everything we might have
1891 * calculated based on it */
1892
1893 p->dropped = false;
1894 p->current_size = UINT64_MAX;
1895 p->new_size = UINT64_MAX;
1896 p->current_padding = UINT64_MAX;
1897 p->new_padding = UINT64_MAX;
1898 p->partno = UINT64_MAX;
1899 p->offset = UINT64_MAX;
1900
1901 if (p->current_partition) {
1902 fdisk_unref_partition(p->current_partition);
1903 p->current_partition = NULL;
1904 }
1905
1906 if (p->new_partition) {
1907 fdisk_unref_partition(p->new_partition);
1908 p->new_partition = NULL;
1909 }
1910
1911 p->padding_area = NULL;
1912 p->allocated_to_area = NULL;
1913
1914 p->current_uuid = SD_ID128_NULL;
1915 p->current_label = mfree(p->current_label);
1916 }
1917
1918 context->start = UINT64_MAX;
1919 context->end = UINT64_MAX;
1920 context->total = UINT64_MAX;
1921
1922 if (context->fdisk_context) {
1923 fdisk_unref_context(context->fdisk_context);
1924 context->fdisk_context = NULL;
1925 }
1926
1927 context_free_free_areas(context);
1928 }
1929
1930 static int format_size_change(uint64_t from, uint64_t to, char **ret) {
1931 char *t;
1932
1933 if (from != UINT64_MAX) {
1934 if (from == to || to == UINT64_MAX)
1935 t = strdup(FORMAT_BYTES(from));
1936 else
1937 t = strjoin(FORMAT_BYTES(from), " ", special_glyph(SPECIAL_GLYPH_ARROW), " ", FORMAT_BYTES(to));
1938 } else if (to != UINT64_MAX)
1939 t = strjoin(special_glyph(SPECIAL_GLYPH_ARROW), " ", FORMAT_BYTES(to));
1940 else {
1941 *ret = NULL;
1942 return 0;
1943 }
1944
1945 if (!t)
1946 return log_oom();
1947
1948 *ret = t;
1949 return 1;
1950 }
1951
1952 static const char *partition_label(const Partition *p) {
1953 assert(p);
1954
1955 if (p->new_label)
1956 return p->new_label;
1957
1958 if (p->current_label)
1959 return p->current_label;
1960
1961 return gpt_partition_type_uuid_to_string(p->type_uuid);
1962 }
1963
1964 static int context_dump_partitions(Context *context, const char *node) {
1965 _cleanup_(table_unrefp) Table *t = NULL;
1966 uint64_t sum_padding = 0, sum_size = 0;
1967 Partition *p;
1968 int r;
1969
1970 if ((arg_json_format_flags & JSON_FORMAT_OFF) && context->n_partitions == 0) {
1971 log_info("Empty partition table.");
1972 return 0;
1973 }
1974
1975 t = table_new("type", "label", "uuid", "file", "node", "offset", "old size", "raw size", "size", "old padding", "raw padding", "padding", "activity");
1976 if (!t)
1977 return log_oom();
1978
1979 if (!DEBUG_LOGGING) {
1980 if (arg_json_format_flags & JSON_FORMAT_OFF)
1981 (void) table_set_display(t, (size_t) 0, (size_t) 1, (size_t) 2, (size_t) 3, (size_t) 4,
1982 (size_t) 8, (size_t) 11);
1983 else
1984 (void) table_set_display(t, (size_t) 0, (size_t) 1, (size_t) 2, (size_t) 3, (size_t) 4,
1985 (size_t) 5, (size_t) 6, (size_t) 7, (size_t) 9, (size_t) 10, (size_t) 12);
1986 }
1987
1988 (void) table_set_align_percent(t, table_get_cell(t, 0, 5), 100);
1989 (void) table_set_align_percent(t, table_get_cell(t, 0, 6), 100);
1990 (void) table_set_align_percent(t, table_get_cell(t, 0, 7), 100);
1991 (void) table_set_align_percent(t, table_get_cell(t, 0, 8), 100);
1992 (void) table_set_align_percent(t, table_get_cell(t, 0, 9), 100);
1993 (void) table_set_align_percent(t, table_get_cell(t, 0, 10), 100);
1994 (void) table_set_align_percent(t, table_get_cell(t, 0, 11), 100);
1995
1996 LIST_FOREACH(partitions, p, context->partitions) {
1997 _cleanup_free_ char *size_change = NULL, *padding_change = NULL, *partname = NULL;
1998 char uuid_buffer[ID128_UUID_STRING_MAX];
1999 const char *label, *activity = NULL;
2000
2001 if (p->dropped)
2002 continue;
2003
2004 if (p->current_size == UINT64_MAX)
2005 activity = "create";
2006 else if (p->current_size != p->new_size)
2007 activity = "resize";
2008
2009 label = partition_label(p);
2010 partname = p->partno != UINT64_MAX ? fdisk_partname(node, p->partno+1) : NULL;
2011
2012 r = format_size_change(p->current_size, p->new_size, &size_change);
2013 if (r < 0)
2014 return r;
2015
2016 r = format_size_change(p->current_padding, p->new_padding, &padding_change);
2017 if (r < 0)
2018 return r;
2019
2020 if (p->new_size != UINT64_MAX)
2021 sum_size += p->new_size;
2022 if (p->new_padding != UINT64_MAX)
2023 sum_padding += p->new_padding;
2024
2025 r = table_add_many(
2026 t,
2027 TABLE_STRING, gpt_partition_type_uuid_to_string_harder(p->type_uuid, uuid_buffer),
2028 TABLE_STRING, empty_to_null(label) ?: "-", TABLE_SET_COLOR, empty_to_null(label) ? NULL : ansi_grey(),
2029 TABLE_UUID, sd_id128_is_null(p->new_uuid) ? p->current_uuid : p->new_uuid,
2030 TABLE_STRING, p->definition_path ? basename(p->definition_path) : "-", TABLE_SET_COLOR, p->definition_path ? NULL : ansi_grey(),
2031 TABLE_STRING, partname ?: "-", TABLE_SET_COLOR, partname ? NULL : ansi_highlight(),
2032 TABLE_UINT64, p->offset,
2033 TABLE_UINT64, p->current_size == UINT64_MAX ? 0 : p->current_size,
2034 TABLE_UINT64, p->new_size,
2035 TABLE_STRING, size_change, TABLE_SET_COLOR, !p->partitions_next && sum_size > 0 ? ansi_underline() : NULL,
2036 TABLE_UINT64, p->current_padding == UINT64_MAX ? 0 : p->current_padding,
2037 TABLE_UINT64, p->new_padding,
2038 TABLE_STRING, padding_change, TABLE_SET_COLOR, !p->partitions_next && sum_padding > 0 ? ansi_underline() : NULL,
2039 TABLE_STRING, activity ?: "unchanged");
2040 if (r < 0)
2041 return table_log_add_error(r);
2042 }
2043
2044 if ((arg_json_format_flags & JSON_FORMAT_OFF) && (sum_padding > 0 || sum_size > 0)) {
2045 const char *a, *b;
2046
2047 a = strjoina(special_glyph(SPECIAL_GLYPH_SIGMA), " = ", FORMAT_BYTES(sum_size));
2048 b = strjoina(special_glyph(SPECIAL_GLYPH_SIGMA), " = ", FORMAT_BYTES(sum_padding));
2049
2050 r = table_add_many(
2051 t,
2052 TABLE_EMPTY,
2053 TABLE_EMPTY,
2054 TABLE_EMPTY,
2055 TABLE_EMPTY,
2056 TABLE_EMPTY,
2057 TABLE_EMPTY,
2058 TABLE_EMPTY,
2059 TABLE_EMPTY,
2060 TABLE_STRING, a,
2061 TABLE_EMPTY,
2062 TABLE_EMPTY,
2063 TABLE_STRING, b,
2064 TABLE_EMPTY);
2065 if (r < 0)
2066 return table_log_add_error(r);
2067 }
2068
2069 return table_print_with_pager(t, arg_json_format_flags, arg_pager_flags, arg_legend);
2070 }
2071
2072 static void context_bar_char_process_partition(
2073 Context *context,
2074 Partition *bar[],
2075 size_t n,
2076 Partition *p,
2077 size_t *ret_start) {
2078
2079 uint64_t from, to, total;
2080 size_t x, y;
2081
2082 assert(context);
2083 assert(bar);
2084 assert(n > 0);
2085 assert(p);
2086
2087 if (p->dropped)
2088 return;
2089
2090 assert(p->offset != UINT64_MAX);
2091 assert(p->new_size != UINT64_MAX);
2092
2093 from = p->offset;
2094 to = from + p->new_size;
2095
2096 assert(context->end >= context->start);
2097 total = context->end - context->start;
2098
2099 assert(from >= context->start);
2100 assert(from <= context->end);
2101 x = (from - context->start) * n / total;
2102
2103 assert(to >= context->start);
2104 assert(to <= context->end);
2105 y = (to - context->start) * n / total;
2106
2107 assert(x <= y);
2108 assert(y <= n);
2109
2110 for (size_t i = x; i < y; i++)
2111 bar[i] = p;
2112
2113 *ret_start = x;
2114 }
2115
2116 static int partition_hint(const Partition *p, const char *node, char **ret) {
2117 _cleanup_free_ char *buf = NULL;
2118 char ids[ID128_UUID_STRING_MAX];
2119 const char *label;
2120 sd_id128_t id;
2121
2122 /* Tries really hard to find a suitable description for this partition */
2123
2124 if (p->definition_path) {
2125 buf = strdup(basename(p->definition_path));
2126 goto done;
2127 }
2128
2129 label = partition_label(p);
2130 if (!isempty(label)) {
2131 buf = strdup(label);
2132 goto done;
2133 }
2134
2135 if (p->partno != UINT64_MAX) {
2136 buf = fdisk_partname(node, p->partno+1);
2137 goto done;
2138 }
2139
2140 if (!sd_id128_is_null(p->new_uuid))
2141 id = p->new_uuid;
2142 else if (!sd_id128_is_null(p->current_uuid))
2143 id = p->current_uuid;
2144 else
2145 id = p->type_uuid;
2146
2147 buf = strdup(id128_to_uuid_string(id, ids));
2148
2149 done:
2150 if (!buf)
2151 return -ENOMEM;
2152
2153 *ret = TAKE_PTR(buf);
2154 return 0;
2155 }
2156
2157 static int context_dump_partition_bar(Context *context, const char *node) {
2158 _cleanup_free_ Partition **bar = NULL;
2159 _cleanup_free_ size_t *start_array = NULL;
2160 Partition *p, *last = NULL;
2161 bool z = false;
2162 size_t c, j = 0;
2163
2164 assert_se((c = columns()) >= 2);
2165 c -= 2; /* We do not use the leftmost and rightmost character cell */
2166
2167 bar = new0(Partition*, c);
2168 if (!bar)
2169 return log_oom();
2170
2171 start_array = new(size_t, context->n_partitions);
2172 if (!start_array)
2173 return log_oom();
2174
2175 LIST_FOREACH(partitions, p, context->partitions)
2176 context_bar_char_process_partition(context, bar, c, p, start_array + j++);
2177
2178 putc(' ', stdout);
2179
2180 for (size_t i = 0; i < c; i++) {
2181 if (bar[i]) {
2182 if (last != bar[i])
2183 z = !z;
2184
2185 fputs(z ? ansi_green() : ansi_yellow(), stdout);
2186 fputs(special_glyph(SPECIAL_GLYPH_DARK_SHADE), stdout);
2187 } else {
2188 fputs(ansi_normal(), stdout);
2189 fputs(special_glyph(SPECIAL_GLYPH_LIGHT_SHADE), stdout);
2190 }
2191
2192 last = bar[i];
2193 }
2194
2195 fputs(ansi_normal(), stdout);
2196 putc('\n', stdout);
2197
2198 for (size_t i = 0; i < context->n_partitions; i++) {
2199 _cleanup_free_ char **line = NULL;
2200
2201 line = new0(char*, c);
2202 if (!line)
2203 return log_oom();
2204
2205 j = 0;
2206 LIST_FOREACH(partitions, p, context->partitions) {
2207 _cleanup_free_ char *d = NULL;
2208 j++;
2209
2210 if (i < context->n_partitions - j) {
2211
2212 if (line[start_array[j-1]]) {
2213 const char *e;
2214
2215 /* Upgrade final corner to the right with a branch to the right */
2216 e = startswith(line[start_array[j-1]], special_glyph(SPECIAL_GLYPH_TREE_RIGHT));
2217 if (e) {
2218 d = strjoin(special_glyph(SPECIAL_GLYPH_TREE_BRANCH), e);
2219 if (!d)
2220 return log_oom();
2221 }
2222 }
2223
2224 if (!d) {
2225 d = strdup(special_glyph(SPECIAL_GLYPH_TREE_VERTICAL));
2226 if (!d)
2227 return log_oom();
2228 }
2229
2230 } else if (i == context->n_partitions - j) {
2231 _cleanup_free_ char *hint = NULL;
2232
2233 (void) partition_hint(p, node, &hint);
2234
2235 if (streq_ptr(line[start_array[j-1]], special_glyph(SPECIAL_GLYPH_TREE_VERTICAL)))
2236 d = strjoin(special_glyph(SPECIAL_GLYPH_TREE_BRANCH), " ", strna(hint));
2237 else
2238 d = strjoin(special_glyph(SPECIAL_GLYPH_TREE_RIGHT), " ", strna(hint));
2239
2240 if (!d)
2241 return log_oom();
2242 }
2243
2244 if (d)
2245 free_and_replace(line[start_array[j-1]], d);
2246 }
2247
2248 putc(' ', stdout);
2249
2250 j = 0;
2251 while (j < c) {
2252 if (line[j]) {
2253 fputs(line[j], stdout);
2254 j += utf8_console_width(line[j]);
2255 } else {
2256 putc(' ', stdout);
2257 j++;
2258 }
2259 }
2260
2261 putc('\n', stdout);
2262
2263 for (j = 0; j < c; j++)
2264 free(line[j]);
2265 }
2266
2267 return 0;
2268 }
2269
2270 static bool context_changed(const Context *context) {
2271 Partition *p;
2272
2273 LIST_FOREACH(partitions, p, context->partitions) {
2274 if (p->dropped)
2275 continue;
2276
2277 if (p->allocated_to_area)
2278 return true;
2279
2280 if (p->new_size != p->current_size)
2281 return true;
2282 }
2283
2284 return false;
2285 }
2286
2287 static int context_wipe_range(Context *context, uint64_t offset, uint64_t size) {
2288 _cleanup_(blkid_free_probep) blkid_probe probe = NULL;
2289 int r;
2290
2291 assert(context);
2292 assert(offset != UINT64_MAX);
2293 assert(size != UINT64_MAX);
2294
2295 probe = blkid_new_probe();
2296 if (!probe)
2297 return log_oom();
2298
2299 errno = 0;
2300 r = blkid_probe_set_device(probe, fdisk_get_devfd(context->fdisk_context), offset, size);
2301 if (r < 0)
2302 return log_error_errno(errno ?: SYNTHETIC_ERRNO(EIO), "Failed to allocate device probe for wiping.");
2303
2304 errno = 0;
2305 if (blkid_probe_enable_superblocks(probe, true) < 0 ||
2306 blkid_probe_set_superblocks_flags(probe, BLKID_SUBLKS_MAGIC|BLKID_SUBLKS_BADCSUM) < 0 ||
2307 blkid_probe_enable_partitions(probe, true) < 0 ||
2308 blkid_probe_set_partitions_flags(probe, BLKID_PARTS_MAGIC) < 0)
2309 return log_error_errno(errno ?: SYNTHETIC_ERRNO(EIO), "Failed to enable superblock and partition probing.");
2310
2311 for (;;) {
2312 errno = 0;
2313 r = blkid_do_probe(probe);
2314 if (r < 0)
2315 return log_error_errno(errno ?: SYNTHETIC_ERRNO(EIO), "Failed to probe for file systems.");
2316 if (r > 0)
2317 break;
2318
2319 errno = 0;
2320 if (blkid_do_wipe(probe, false) < 0)
2321 return log_error_errno(errno ?: SYNTHETIC_ERRNO(EIO), "Failed to wipe file system signature.");
2322 }
2323
2324 return 0;
2325 }
2326
2327 static int context_wipe_partition(Context *context, Partition *p) {
2328 int r;
2329
2330 assert(context);
2331 assert(p);
2332 assert(!PARTITION_EXISTS(p)); /* Safety check: never wipe existing partitions */
2333
2334 assert(p->offset != UINT64_MAX);
2335 assert(p->new_size != UINT64_MAX);
2336
2337 r = context_wipe_range(context, p->offset, p->new_size);
2338 if (r < 0)
2339 return r;
2340
2341 log_info("Successfully wiped file system signatures from future partition %" PRIu64 ".", p->partno);
2342 return 0;
2343 }
2344
2345 static int context_discard_range(
2346 Context *context,
2347 uint64_t offset,
2348 uint64_t size) {
2349
2350 struct stat st;
2351 int fd;
2352
2353 assert(context);
2354 assert(offset != UINT64_MAX);
2355 assert(size != UINT64_MAX);
2356
2357 if (size <= 0)
2358 return 0;
2359
2360 assert_se((fd = fdisk_get_devfd(context->fdisk_context)) >= 0);
2361
2362 if (fstat(fd, &st) < 0)
2363 return -errno;
2364
2365 if (S_ISREG(st.st_mode)) {
2366 if (fallocate(fd, FALLOC_FL_PUNCH_HOLE|FALLOC_FL_KEEP_SIZE, offset, size) < 0) {
2367 if (ERRNO_IS_NOT_SUPPORTED(errno))
2368 return -EOPNOTSUPP;
2369
2370 return -errno;
2371 }
2372
2373 return 1;
2374 }
2375
2376 if (S_ISBLK(st.st_mode)) {
2377 uint64_t range[2], end;
2378
2379 range[0] = round_up_size(offset, 512);
2380
2381 if (offset > UINT64_MAX - size)
2382 return -ERANGE;
2383
2384 end = offset + size;
2385 if (end <= range[0])
2386 return 0;
2387
2388 range[1] = round_down_size(end - range[0], 512);
2389 if (range[1] <= 0)
2390 return 0;
2391
2392 if (ioctl(fd, BLKDISCARD, range) < 0) {
2393 if (ERRNO_IS_NOT_SUPPORTED(errno))
2394 return -EOPNOTSUPP;
2395
2396 return -errno;
2397 }
2398
2399 return 1;
2400 }
2401
2402 return -EOPNOTSUPP;
2403 }
2404
2405 static int context_discard_partition(Context *context, Partition *p) {
2406 int r;
2407
2408 assert(context);
2409 assert(p);
2410
2411 assert(p->offset != UINT64_MAX);
2412 assert(p->new_size != UINT64_MAX);
2413 assert(!PARTITION_EXISTS(p)); /* Safety check: never discard existing partitions */
2414
2415 if (!arg_discard)
2416 return 0;
2417
2418 r = context_discard_range(context, p->offset, p->new_size);
2419 if (r == -EOPNOTSUPP) {
2420 log_info("Storage does not support discard, not discarding data in future partition %" PRIu64 ".", p->partno);
2421 return 0;
2422 }
2423 if (r == -EBUSY) {
2424 /* Let's handle this gracefully: https://bugzilla.kernel.org/show_bug.cgi?id=211167 */
2425 log_info("Block device is busy, not discarding partition %" PRIu64 " because it probably is mounted.", p->partno);
2426 return 0;
2427 }
2428 if (r == 0) {
2429 log_info("Partition %" PRIu64 " too short for discard, skipping.", p->partno);
2430 return 0;
2431 }
2432 if (r < 0)
2433 return log_error_errno(r, "Failed to discard data for future partition %" PRIu64 ".", p->partno);
2434
2435 log_info("Successfully discarded data from future partition %" PRIu64 ".", p->partno);
2436 return 1;
2437 }
2438
2439 static int context_discard_gap_after(Context *context, Partition *p) {
2440 uint64_t gap, next = UINT64_MAX;
2441 Partition *q;
2442 int r;
2443
2444 assert(context);
2445 assert(!p || (p->offset != UINT64_MAX && p->new_size != UINT64_MAX));
2446
2447 if (p)
2448 gap = p->offset + p->new_size;
2449 else
2450 gap = context->start;
2451
2452 LIST_FOREACH(partitions, q, context->partitions) {
2453 if (q->dropped)
2454 continue;
2455
2456 assert(q->offset != UINT64_MAX);
2457 assert(q->new_size != UINT64_MAX);
2458
2459 if (q->offset < gap)
2460 continue;
2461
2462 if (next == UINT64_MAX || q->offset < next)
2463 next = q->offset;
2464 }
2465
2466 if (next == UINT64_MAX) {
2467 next = context->end;
2468 if (gap > next)
2469 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Partition end beyond disk end.");
2470 }
2471
2472 assert(next >= gap);
2473 r = context_discard_range(context, gap, next - gap);
2474 if (r == -EOPNOTSUPP) {
2475 if (p)
2476 log_info("Storage does not support discard, not discarding gap after partition %" PRIu64 ".", p->partno);
2477 else
2478 log_info("Storage does not support discard, not discarding gap at beginning of disk.");
2479 return 0;
2480 }
2481 if (r == 0) /* Too short */
2482 return 0;
2483 if (r < 0) {
2484 if (p)
2485 return log_error_errno(r, "Failed to discard gap after partition %" PRIu64 ".", p->partno);
2486 else
2487 return log_error_errno(r, "Failed to discard gap at beginning of disk.");
2488 }
2489
2490 if (p)
2491 log_info("Successfully discarded gap after partition %" PRIu64 ".", p->partno);
2492 else
2493 log_info("Successfully discarded gap at beginning of disk.");
2494
2495 return 0;
2496 }
2497
2498 static int context_wipe_and_discard(Context *context, bool from_scratch) {
2499 Partition *p;
2500 int r;
2501
2502 assert(context);
2503
2504 /* Wipe and discard the contents of all partitions we are about to create. We skip the discarding if
2505 * we were supposed to start from scratch anyway, as in that case we just discard the whole block
2506 * device in one go early on. */
2507
2508 LIST_FOREACH(partitions, p, context->partitions) {
2509
2510 if (!p->allocated_to_area)
2511 continue;
2512
2513 r = context_wipe_partition(context, p);
2514 if (r < 0)
2515 return r;
2516
2517 if (!from_scratch) {
2518 r = context_discard_partition(context, p);
2519 if (r < 0)
2520 return r;
2521
2522 r = context_discard_gap_after(context, p);
2523 if (r < 0)
2524 return r;
2525 }
2526 }
2527
2528 if (!from_scratch) {
2529 r = context_discard_gap_after(context, NULL);
2530 if (r < 0)
2531 return r;
2532 }
2533
2534 return 0;
2535 }
2536
2537 static int partition_encrypt(
2538 Partition *p,
2539 const char *node,
2540 struct crypt_device **ret_cd,
2541 char **ret_volume,
2542 int *ret_fd) {
2543 #if HAVE_LIBCRYPTSETUP
2544 _cleanup_(sym_crypt_freep) struct crypt_device *cd = NULL;
2545 _cleanup_(erase_and_freep) void *volume_key = NULL;
2546 _cleanup_free_ char *dm_name = NULL, *vol = NULL;
2547 char suuid[ID128_UUID_STRING_MAX];
2548 size_t volume_key_size = 256 / 8;
2549 sd_id128_t uuid;
2550 int r;
2551
2552 assert(p);
2553 assert(p->encrypt != ENCRYPT_OFF);
2554
2555 log_debug("Encryption mode for partition %" PRIu64 ": %s", p->partno, encrypt_mode_to_string(p->encrypt));
2556
2557 r = dlopen_cryptsetup();
2558 if (r < 0)
2559 return log_error_errno(r, "libcryptsetup not found, cannot encrypt: %m");
2560
2561 if (asprintf(&dm_name, "luks-repart-%08" PRIx64, random_u64()) < 0)
2562 return log_oom();
2563
2564 if (ret_volume) {
2565 vol = path_join("/dev/mapper/", dm_name);
2566 if (!vol)
2567 return log_oom();
2568 }
2569
2570 r = derive_uuid(p->new_uuid, "luks-uuid", &uuid);
2571 if (r < 0)
2572 return r;
2573
2574 log_info("Encrypting future partition %" PRIu64 "...", p->partno);
2575
2576 volume_key = malloc(volume_key_size);
2577 if (!volume_key)
2578 return log_oom();
2579
2580 r = genuine_random_bytes(volume_key, volume_key_size, RANDOM_BLOCK);
2581 if (r < 0)
2582 return log_error_errno(r, "Failed to generate volume key: %m");
2583
2584 r = sym_crypt_init(&cd, node);
2585 if (r < 0)
2586 return log_error_errno(r, "Failed to allocate libcryptsetup context: %m");
2587
2588 cryptsetup_enable_logging(cd);
2589
2590 r = sym_crypt_format(cd,
2591 CRYPT_LUKS2,
2592 "aes",
2593 "xts-plain64",
2594 id128_to_uuid_string(uuid, suuid),
2595 volume_key,
2596 volume_key_size,
2597 &(struct crypt_params_luks2) {
2598 .label = strempty(p->new_label),
2599 .sector_size = 512U,
2600 });
2601 if (r < 0)
2602 return log_error_errno(r, "Failed to LUKS2 format future partition: %m");
2603
2604 if (IN_SET(p->encrypt, ENCRYPT_KEY_FILE, ENCRYPT_KEY_FILE_TPM2)) {
2605 r = sym_crypt_keyslot_add_by_volume_key(
2606 cd,
2607 CRYPT_ANY_SLOT,
2608 volume_key,
2609 volume_key_size,
2610 strempty(arg_key),
2611 arg_key_size);
2612 if (r < 0)
2613 return log_error_errno(r, "Failed to add LUKS2 key: %m");
2614 }
2615
2616 if (IN_SET(p->encrypt, ENCRYPT_TPM2, ENCRYPT_KEY_FILE_TPM2)) {
2617 #if HAVE_TPM2
2618 _cleanup_(erase_and_freep) char *base64_encoded = NULL;
2619 _cleanup_(json_variant_unrefp) JsonVariant *v = NULL;
2620 _cleanup_(erase_and_freep) void *secret = NULL;
2621 _cleanup_free_ void *blob = NULL, *hash = NULL;
2622 size_t secret_size, blob_size, hash_size;
2623 uint16_t pcr_bank;
2624 int keyslot;
2625
2626 r = tpm2_seal(arg_tpm2_device, arg_tpm2_pcr_mask, &secret, &secret_size, &blob, &blob_size, &hash, &hash_size, &pcr_bank);
2627 if (r < 0)
2628 return log_error_errno(r, "Failed to seal to TPM2: %m");
2629
2630 r = base64mem(secret, secret_size, &base64_encoded);
2631 if (r < 0)
2632 return log_error_errno(r, "Failed to base64 encode secret key: %m");
2633
2634 r = cryptsetup_set_minimal_pbkdf(cd);
2635 if (r < 0)
2636 return log_error_errno(r, "Failed to set minimal PBKDF: %m");
2637
2638 keyslot = sym_crypt_keyslot_add_by_volume_key(
2639 cd,
2640 CRYPT_ANY_SLOT,
2641 volume_key,
2642 volume_key_size,
2643 base64_encoded,
2644 strlen(base64_encoded));
2645 if (keyslot < 0)
2646 return log_error_errno(keyslot, "Failed to add new TPM2 key to %s: %m", node);
2647
2648 r = tpm2_make_luks2_json(keyslot, arg_tpm2_pcr_mask, pcr_bank, blob, blob_size, hash, hash_size, &v);
2649 if (r < 0)
2650 return log_error_errno(r, "Failed to prepare TPM2 JSON token object: %m");
2651
2652 r = cryptsetup_add_token_json(cd, v);
2653 if (r < 0)
2654 return log_error_errno(r, "Failed to add TPM2 JSON token to LUKS2 header: %m");
2655 #else
2656 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2657 "Support for TPM2 enrollment not enabled.");
2658 #endif
2659 }
2660
2661 r = sym_crypt_activate_by_volume_key(
2662 cd,
2663 dm_name,
2664 volume_key,
2665 volume_key_size,
2666 arg_discard ? CRYPT_ACTIVATE_ALLOW_DISCARDS : 0);
2667 if (r < 0)
2668 return log_error_errno(r, "Failed to activate LUKS superblock: %m");
2669
2670 log_info("Successfully encrypted future partition %" PRIu64 ".", p->partno);
2671
2672 if (ret_fd) {
2673 _cleanup_close_ int dev_fd = -1;
2674
2675 dev_fd = open(vol, O_RDWR|O_CLOEXEC|O_NOCTTY);
2676 if (dev_fd < 0)
2677 return log_error_errno(errno, "Failed to open LUKS volume '%s': %m", vol);
2678
2679 *ret_fd = TAKE_FD(dev_fd);
2680 }
2681
2682 if (ret_cd)
2683 *ret_cd = TAKE_PTR(cd);
2684 if (ret_volume)
2685 *ret_volume = TAKE_PTR(vol);
2686
2687 return 0;
2688 #else
2689 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP), "libcryptsetup is not supported, cannot encrypt: %m");
2690 #endif
2691 }
2692
2693 static int deactivate_luks(struct crypt_device *cd, const char *node) {
2694 #if HAVE_LIBCRYPTSETUP
2695 int r;
2696
2697 if (!cd)
2698 return 0;
2699
2700 assert(node);
2701
2702 /* udev or so might access out block device in the background while we are done. Let's hence force
2703 * detach the volume. We sync'ed before, hence this should be safe. */
2704
2705 r = sym_crypt_deactivate_by_name(cd, basename(node), CRYPT_DEACTIVATE_FORCE);
2706 if (r < 0)
2707 return log_error_errno(r, "Failed to deactivate LUKS device: %m");
2708
2709 return 1;
2710 #else
2711 return 0;
2712 #endif
2713 }
2714
2715 static int context_copy_blocks(Context *context) {
2716 Partition *p;
2717 int whole_fd = -1, r;
2718
2719 assert(context);
2720
2721 /* Copy in file systems on the block level */
2722
2723 LIST_FOREACH(partitions, p, context->partitions) {
2724 _cleanup_(sym_crypt_freep) struct crypt_device *cd = NULL;
2725 _cleanup_(loop_device_unrefp) LoopDevice *d = NULL;
2726 _cleanup_free_ char *encrypted = NULL;
2727 _cleanup_close_ int encrypted_dev_fd = -1;
2728 int target_fd;
2729
2730 if (p->copy_blocks_fd < 0)
2731 continue;
2732
2733 if (p->dropped)
2734 continue;
2735
2736 if (PARTITION_EXISTS(p)) /* Never copy over existing partitions */
2737 continue;
2738
2739 assert(p->new_size != UINT64_MAX);
2740 assert(p->copy_blocks_size != UINT64_MAX);
2741 assert(p->new_size >= p->copy_blocks_size);
2742
2743 if (whole_fd < 0)
2744 assert_se((whole_fd = fdisk_get_devfd(context->fdisk_context)) >= 0);
2745
2746 if (p->encrypt != ENCRYPT_OFF) {
2747 r = loop_device_make(whole_fd, O_RDWR, p->offset, p->new_size, 0, &d);
2748 if (r < 0)
2749 return log_error_errno(r, "Failed to make loopback device of future partition %" PRIu64 ": %m", p->partno);
2750
2751 r = loop_device_flock(d, LOCK_EX);
2752 if (r < 0)
2753 return log_error_errno(r, "Failed to lock loopback device: %m");
2754
2755 r = partition_encrypt(p, d->node, &cd, &encrypted, &encrypted_dev_fd);
2756 if (r < 0)
2757 return log_error_errno(r, "Failed to encrypt device: %m");
2758
2759 if (flock(encrypted_dev_fd, LOCK_EX) < 0)
2760 return log_error_errno(errno, "Failed to lock LUKS device: %m");
2761
2762 target_fd = encrypted_dev_fd;
2763 } else {
2764 if (lseek(whole_fd, p->offset, SEEK_SET) == (off_t) -1)
2765 return log_error_errno(errno, "Failed to seek to partition offset: %m");
2766
2767 target_fd = whole_fd;
2768 }
2769
2770 log_info("Copying in '%s' (%s) on block level into future partition %" PRIu64 ".",
2771 p->copy_blocks_path, FORMAT_BYTES(p->copy_blocks_size), p->partno);
2772
2773 r = copy_bytes_full(p->copy_blocks_fd, target_fd, p->copy_blocks_size, 0, NULL, NULL, NULL, NULL);
2774 if (r < 0)
2775 return log_error_errno(r, "Failed to copy in data from '%s': %m", p->copy_blocks_path);
2776
2777 if (fsync(target_fd) < 0)
2778 return log_error_errno(r, "Failed to synchronize copied data blocks: %m");
2779
2780 if (p->encrypt != ENCRYPT_OFF) {
2781 encrypted_dev_fd = safe_close(encrypted_dev_fd);
2782
2783 r = deactivate_luks(cd, encrypted);
2784 if (r < 0)
2785 return r;
2786
2787 sym_crypt_free(cd);
2788 cd = NULL;
2789
2790 r = loop_device_sync(d);
2791 if (r < 0)
2792 return log_error_errno(r, "Failed to sync loopback device: %m");
2793 }
2794
2795 log_info("Copying in of '%s' on block level completed.", p->copy_blocks_path);
2796 }
2797
2798 return 0;
2799 }
2800
2801 static int do_copy_files(Partition *p, const char *fs) {
2802 char **source, **target;
2803 int r;
2804
2805 assert(p);
2806 assert(fs);
2807
2808 STRV_FOREACH_PAIR(source, target, p->copy_files) {
2809 _cleanup_close_ int sfd = -1, pfd = -1, tfd = -1;
2810
2811 sfd = chase_symlinks_and_open(*source, arg_root, CHASE_PREFIX_ROOT|CHASE_WARN, O_CLOEXEC|O_NOCTTY, NULL);
2812 if (sfd < 0)
2813 return log_error_errno(sfd, "Failed to open source file '%s%s': %m", strempty(arg_root), *source);
2814
2815 r = fd_verify_regular(sfd);
2816 if (r < 0) {
2817 if (r != -EISDIR)
2818 return log_error_errno(r, "Failed to check type of source file '%s': %m", *source);
2819
2820 /* We are looking at a directory */
2821 tfd = chase_symlinks_and_open(*target, fs, CHASE_PREFIX_ROOT|CHASE_WARN, O_RDONLY|O_DIRECTORY|O_CLOEXEC, NULL);
2822 if (tfd < 0) {
2823 _cleanup_free_ char *dn = NULL, *fn = NULL;
2824
2825 if (tfd != -ENOENT)
2826 return log_error_errno(tfd, "Failed to open target directory '%s': %m", *target);
2827
2828 r = path_extract_filename(*target, &fn);
2829 if (r < 0)
2830 return log_error_errno(r, "Failed to extract filename from '%s': %m", *target);
2831
2832 r = path_extract_directory(*target, &dn);
2833 if (r < 0)
2834 return log_error_errno(r, "Failed to extract directory from '%s': %m", *target);
2835
2836 r = mkdir_p_root(fs, dn, UID_INVALID, GID_INVALID, 0755);
2837 if (r < 0)
2838 return log_error_errno(r, "Failed to create parent directory '%s': %m", dn);
2839
2840 pfd = chase_symlinks_and_open(dn, fs, CHASE_PREFIX_ROOT|CHASE_WARN, O_RDONLY|O_DIRECTORY|O_CLOEXEC, NULL);
2841 if (pfd < 0)
2842 return log_error_errno(pfd, "Failed to open parent directory of target: %m");
2843
2844 r = copy_tree_at(
2845 sfd, ".",
2846 pfd, fn,
2847 UID_INVALID, GID_INVALID,
2848 COPY_REFLINK|COPY_MERGE|COPY_REPLACE|COPY_SIGINT|COPY_HARDLINKS|COPY_ALL_XATTRS);
2849 } else
2850 r = copy_tree_at(
2851 sfd, ".",
2852 tfd, ".",
2853 UID_INVALID, GID_INVALID,
2854 COPY_REFLINK|COPY_MERGE|COPY_REPLACE|COPY_SIGINT|COPY_HARDLINKS|COPY_ALL_XATTRS);
2855 if (r < 0)
2856 return log_error_errno(r, "Failed to copy '%s' to '%s%s': %m", *source, strempty(arg_root), *target);
2857 } else {
2858 _cleanup_free_ char *dn = NULL, *fn = NULL;
2859
2860 /* We are looking at a regular file */
2861
2862 r = path_extract_filename(*target, &fn);
2863 if (r == -EADDRNOTAVAIL || r == O_DIRECTORY)
2864 return log_error_errno(SYNTHETIC_ERRNO(EISDIR),
2865 "Target path '%s' refers to a directory, but source path '%s' refers to regular file, can't copy.", *target, *source);
2866 if (r < 0)
2867 return log_error_errno(r, "Failed to extract filename from '%s': %m", *target);
2868
2869 r = path_extract_directory(*target, &dn);
2870 if (r < 0)
2871 return log_error_errno(r, "Failed to extract directory from '%s': %m", *target);
2872
2873 r = mkdir_p_root(fs, dn, UID_INVALID, GID_INVALID, 0755);
2874 if (r < 0)
2875 return log_error_errno(r, "Failed to create parent directory: %m");
2876
2877 pfd = chase_symlinks_and_open(dn, fs, CHASE_PREFIX_ROOT|CHASE_WARN, O_RDONLY|O_DIRECTORY|O_CLOEXEC, NULL);
2878 if (pfd < 0)
2879 return log_error_errno(pfd, "Failed to open parent directory of target: %m");
2880
2881 tfd = openat(pfd, fn, O_CREAT|O_EXCL|O_WRONLY|O_CLOEXEC, 0700);
2882 if (tfd < 0)
2883 return log_error_errno(errno, "Failed to create target file '%s': %m", *target);
2884
2885 r = copy_bytes(sfd, tfd, UINT64_MAX, COPY_REFLINK|COPY_SIGINT);
2886 if (r < 0)
2887 return log_error_errno(r, "Failed to copy '%s' to '%s%s': %m", *source, strempty(arg_root), *target);
2888
2889 (void) copy_xattr(sfd, tfd, COPY_ALL_XATTRS);
2890 (void) copy_access(sfd, tfd);
2891 (void) copy_times(sfd, tfd, 0);
2892 }
2893 }
2894
2895 return 0;
2896 }
2897
2898 static int do_make_directories(Partition *p, const char *fs) {
2899 char **d;
2900 int r;
2901
2902 assert(p);
2903 assert(fs);
2904
2905 STRV_FOREACH(d, p->make_directories) {
2906
2907 r = mkdir_p_root(fs, *d, UID_INVALID, GID_INVALID, 0755);
2908 if (r < 0)
2909 return log_error_errno(r, "Failed to create directory '%s' in file system: %m", *d);
2910 }
2911
2912 return 0;
2913 }
2914
2915 static int partition_populate(Partition *p, const char *node) {
2916 int r;
2917
2918 assert(p);
2919 assert(node);
2920
2921 if (strv_isempty(p->copy_files) && strv_isempty(p->make_directories))
2922 return 0;
2923
2924 log_info("Populating partition %" PRIu64 " with files.", p->partno);
2925
2926 /* We copy in a child process, since we have to mount the fs for that, and we don't want that fs to
2927 * appear in the host namespace. Hence we fork a child that has its own file system namespace and
2928 * detached mount propagation. */
2929
2930 r = safe_fork("(sd-copy)", FORK_DEATHSIG|FORK_LOG|FORK_WAIT|FORK_NEW_MOUNTNS|FORK_MOUNTNS_SLAVE, NULL);
2931 if (r < 0)
2932 return r;
2933 if (r == 0) {
2934 static const char fs[] = "/run/systemd/mount-root";
2935 /* This is a child process with its own mount namespace and propagation to host turned off */
2936
2937 r = mkdir_p(fs, 0700);
2938 if (r < 0) {
2939 log_error_errno(r, "Failed to create mount point: %m");
2940 _exit(EXIT_FAILURE);
2941 }
2942
2943 if (mount_nofollow_verbose(LOG_ERR, node, fs, p->format, MS_NOATIME|MS_NODEV|MS_NOEXEC|MS_NOSUID, NULL) < 0)
2944 _exit(EXIT_FAILURE);
2945
2946 if (do_copy_files(p, fs) < 0)
2947 _exit(EXIT_FAILURE);
2948
2949 if (do_make_directories(p, fs) < 0)
2950 _exit(EXIT_FAILURE);
2951
2952 r = syncfs_path(AT_FDCWD, fs);
2953 if (r < 0) {
2954 log_error_errno(r, "Failed to synchronize written files: %m");
2955 _exit(EXIT_FAILURE);
2956 }
2957
2958 _exit(EXIT_SUCCESS);
2959 }
2960
2961 log_info("Successfully populated partition %" PRIu64 " with files.", p->partno);
2962 return 0;
2963 }
2964
2965 static int context_mkfs(Context *context) {
2966 Partition *p;
2967 int fd = -1, r;
2968
2969 assert(context);
2970
2971 /* Make a file system */
2972
2973 LIST_FOREACH(partitions, p, context->partitions) {
2974 _cleanup_(sym_crypt_freep) struct crypt_device *cd = NULL;
2975 _cleanup_(loop_device_unrefp) LoopDevice *d = NULL;
2976 _cleanup_free_ char *encrypted = NULL;
2977 _cleanup_close_ int encrypted_dev_fd = -1;
2978 const char *fsdev;
2979 sd_id128_t fs_uuid;
2980
2981 if (p->dropped)
2982 continue;
2983
2984 if (PARTITION_EXISTS(p)) /* Never format existing partitions */
2985 continue;
2986
2987 if (!p->format)
2988 continue;
2989
2990 assert(p->offset != UINT64_MAX);
2991 assert(p->new_size != UINT64_MAX);
2992
2993 if (fd < 0)
2994 assert_se((fd = fdisk_get_devfd(context->fdisk_context)) >= 0);
2995
2996 /* Loopback block devices are not only useful to turn regular files into block devices, but
2997 * also to cut out sections of block devices into new block devices. */
2998
2999 r = loop_device_make(fd, O_RDWR, p->offset, p->new_size, 0, &d);
3000 if (r < 0)
3001 return log_error_errno(r, "Failed to make loopback device of future partition %" PRIu64 ": %m", p->partno);
3002
3003 r = loop_device_flock(d, LOCK_EX);
3004 if (r < 0)
3005 return log_error_errno(r, "Failed to lock loopback device: %m");
3006
3007 if (p->encrypt != ENCRYPT_OFF) {
3008 r = partition_encrypt(p, d->node, &cd, &encrypted, &encrypted_dev_fd);
3009 if (r < 0)
3010 return log_error_errno(r, "Failed to encrypt device: %m");
3011
3012 if (flock(encrypted_dev_fd, LOCK_EX) < 0)
3013 return log_error_errno(errno, "Failed to lock LUKS device: %m");
3014
3015 fsdev = encrypted;
3016 } else
3017 fsdev = d->node;
3018
3019 log_info("Formatting future partition %" PRIu64 ".", p->partno);
3020
3021 /* Calculate the UUID for the file system as HMAC-SHA256 of the string "file-system-uuid",
3022 * keyed off the partition UUID. */
3023 r = derive_uuid(p->new_uuid, "file-system-uuid", &fs_uuid);
3024 if (r < 0)
3025 return r;
3026
3027 r = make_filesystem(fsdev, p->format, strempty(p->new_label), fs_uuid, arg_discard);
3028 if (r < 0) {
3029 encrypted_dev_fd = safe_close(encrypted_dev_fd);
3030 (void) deactivate_luks(cd, encrypted);
3031 return r;
3032 }
3033
3034 log_info("Successfully formatted future partition %" PRIu64 ".", p->partno);
3035
3036 /* The file system is now created, no need to delay udev further */
3037 if (p->encrypt != ENCRYPT_OFF)
3038 if (flock(encrypted_dev_fd, LOCK_UN) < 0)
3039 return log_error_errno(errno, "Failed to unlock LUKS device: %m");
3040
3041 r = partition_populate(p, fsdev);
3042 if (r < 0) {
3043 encrypted_dev_fd = safe_close(encrypted_dev_fd);
3044 (void) deactivate_luks(cd, encrypted);
3045 return r;
3046 }
3047
3048 /* Note that we always sync explicitly here, since mkfs.fat doesn't do that on its own, and
3049 * if we don't sync before detaching a block device the in-flight sectors possibly won't hit
3050 * the disk. */
3051
3052 if (p->encrypt != ENCRYPT_OFF) {
3053 if (fsync(encrypted_dev_fd) < 0)
3054 return log_error_errno(r, "Failed to synchronize LUKS volume: %m");
3055 encrypted_dev_fd = safe_close(encrypted_dev_fd);
3056
3057 r = deactivate_luks(cd, encrypted);
3058 if (r < 0)
3059 return r;
3060
3061 sym_crypt_free(cd);
3062 cd = NULL;
3063 }
3064
3065 r = loop_device_sync(d);
3066 if (r < 0)
3067 return log_error_errno(r, "Failed to sync loopback device: %m");
3068 }
3069
3070 return 0;
3071 }
3072
3073 static int partition_acquire_uuid(Context *context, Partition *p, sd_id128_t *ret) {
3074 struct {
3075 sd_id128_t type_uuid;
3076 uint64_t counter;
3077 } _packed_ plaintext = {};
3078 union {
3079 unsigned char md[SHA256_DIGEST_LENGTH];
3080 sd_id128_t id;
3081 } result;
3082
3083 uint64_t k = 0;
3084 Partition *q;
3085 int r;
3086
3087 assert(context);
3088 assert(p);
3089 assert(ret);
3090
3091 /* Calculate a good UUID for the indicated partition. We want a certain degree of reproducibility,
3092 * hence we won't generate the UUIDs randomly. Instead we use a cryptographic hash (precisely:
3093 * HMAC-SHA256) to derive them from a single seed. The seed is generally the machine ID of the
3094 * installation we are processing, but if random behaviour is desired can be random, too. We use the
3095 * seed value as key for the HMAC (since the machine ID is something we generally don't want to leak)
3096 * and the partition type as plaintext. The partition type is suffixed with a counter (only for the
3097 * second and later partition of the same type) if we have more than one partition of the same
3098 * time. Or in other words:
3099 *
3100 * With:
3101 * SEED := /etc/machine-id
3102 *
3103 * If first partition instance of type TYPE_UUID:
3104 * PARTITION_UUID := HMAC-SHA256(SEED, TYPE_UUID)
3105 *
3106 * For all later partition instances of type TYPE_UUID with INSTANCE being the LE64 encoded instance number:
3107 * PARTITION_UUID := HMAC-SHA256(SEED, TYPE_UUID || INSTANCE)
3108 */
3109
3110 LIST_FOREACH(partitions, q, context->partitions) {
3111 if (p == q)
3112 break;
3113
3114 if (!sd_id128_equal(p->type_uuid, q->type_uuid))
3115 continue;
3116
3117 k++;
3118 }
3119
3120 plaintext.type_uuid = p->type_uuid;
3121 plaintext.counter = htole64(k);
3122
3123 if (!HMAC(EVP_sha256(),
3124 &context->seed, sizeof(context->seed),
3125 (const unsigned char*) &plaintext, k == 0 ? sizeof(sd_id128_t) : sizeof(plaintext),
3126 result.md, NULL))
3127 return log_error_errno(SYNTHETIC_ERRNO(ENOTRECOVERABLE), "SHA256 calculation failed.");
3128
3129 /* Take the first half, mark it as v4 UUID */
3130 assert_cc(sizeof(result.md) == sizeof(result.id) * 2);
3131 result.id = id128_make_v4_uuid(result.id);
3132
3133 /* Ensure this partition UUID is actually unique, and there's no remaining partition from an earlier run? */
3134 LIST_FOREACH(partitions, q, context->partitions) {
3135 if (p == q)
3136 continue;
3137
3138 if (sd_id128_in_set(result.id, q->current_uuid, q->new_uuid)) {
3139 log_warning("Partition UUID calculated from seed for partition %" PRIu64 " already used, reverting to randomized UUID.", p->partno);
3140
3141 r = sd_id128_randomize(&result.id);
3142 if (r < 0)
3143 return log_error_errno(r, "Failed to generate randomized UUID: %m");
3144
3145 break;
3146 }
3147 }
3148
3149 *ret = result.id;
3150 return 0;
3151 }
3152
3153 static int partition_acquire_label(Context *context, Partition *p, char **ret) {
3154 _cleanup_free_ char *label = NULL;
3155 const char *prefix;
3156 unsigned k = 1;
3157
3158 assert(context);
3159 assert(p);
3160 assert(ret);
3161
3162 prefix = gpt_partition_type_uuid_to_string(p->type_uuid);
3163 if (!prefix)
3164 prefix = "linux";
3165
3166 for (;;) {
3167 const char *ll = label ?: prefix;
3168 bool retry = false;
3169 Partition *q;
3170
3171 LIST_FOREACH(partitions, q, context->partitions) {
3172 if (p == q)
3173 break;
3174
3175 if (streq_ptr(ll, q->current_label) ||
3176 streq_ptr(ll, q->new_label)) {
3177 retry = true;
3178 break;
3179 }
3180 }
3181
3182 if (!retry)
3183 break;
3184
3185 label = mfree(label);
3186 if (asprintf(&label, "%s-%u", prefix, ++k) < 0)
3187 return log_oom();
3188 }
3189
3190 if (!label) {
3191 label = strdup(prefix);
3192 if (!label)
3193 return log_oom();
3194 }
3195
3196 *ret = TAKE_PTR(label);
3197 return 0;
3198 }
3199
3200 static int context_acquire_partition_uuids_and_labels(Context *context) {
3201 Partition *p;
3202 int r;
3203
3204 assert(context);
3205
3206 LIST_FOREACH(partitions, p, context->partitions) {
3207 /* Never touch foreign partitions */
3208 if (PARTITION_IS_FOREIGN(p)) {
3209 p->new_uuid = p->current_uuid;
3210
3211 if (p->current_label) {
3212 r = free_and_strdup_warn(&p->new_label, strempty(p->current_label));
3213 if (r < 0)
3214 return r;
3215 }
3216
3217 continue;
3218 }
3219
3220 if (!sd_id128_is_null(p->current_uuid))
3221 p->new_uuid = p->current_uuid; /* Never change initialized UUIDs */
3222 else if (sd_id128_is_null(p->new_uuid)) {
3223 /* Not explicitly set by user! */
3224 r = partition_acquire_uuid(context, p, &p->new_uuid);
3225 if (r < 0)
3226 return r;
3227 }
3228
3229 if (!isempty(p->current_label)) {
3230 /* never change initialized labels */
3231 r = free_and_strdup_warn(&p->new_label, p->current_label);
3232 if (r < 0)
3233 return r;
3234 } else if (!p->new_label) {
3235 /* Not explicitly set by user! */
3236
3237 r = partition_acquire_label(context, p, &p->new_label);
3238 if (r < 0)
3239 return r;
3240 }
3241 }
3242
3243 return 0;
3244 }
3245
3246 static int set_gpt_flags(struct fdisk_partition *q, uint64_t flags) {
3247 _cleanup_free_ char *a = NULL;
3248
3249 for (unsigned i = 0; i < sizeof(flags) * 8; i++) {
3250 uint64_t bit = UINT64_C(1) << i;
3251 char buf[DECIMAL_STR_MAX(unsigned)+1];
3252
3253 if (!FLAGS_SET(flags, bit))
3254 continue;
3255
3256 xsprintf(buf, "%u", i);
3257 if (!strextend_with_separator(&a, ",", buf))
3258 return -ENOMEM;
3259 }
3260
3261 return fdisk_partition_set_attrs(q, a);
3262 }
3263
3264 static uint64_t partition_merge_flags(Partition *p) {
3265 uint64_t f;
3266
3267 assert(p);
3268
3269 f = p->gpt_flags;
3270
3271 if (p->no_auto >= 0) {
3272 if (gpt_partition_type_knows_no_auto(p->type_uuid))
3273 SET_FLAG(f, GPT_FLAG_NO_AUTO, p->no_auto);
3274 else {
3275 char buffer[ID128_UUID_STRING_MAX];
3276 log_warning("Configured NoAuto=%s for partition type '%s' that doesn't support it, ignoring.",
3277 yes_no(p->no_auto),
3278 gpt_partition_type_uuid_to_string_harder(p->type_uuid, buffer));
3279 }
3280 }
3281
3282 if (p->read_only >= 0) {
3283 if (gpt_partition_type_knows_read_only(p->type_uuid))
3284 SET_FLAG(f, GPT_FLAG_READ_ONLY, p->read_only);
3285 else {
3286 char buffer[ID128_UUID_STRING_MAX];
3287 log_warning("Configured ReadOnly=%s for partition type '%s' that doesn't support it, ignoring.",
3288 yes_no(p->read_only),
3289 gpt_partition_type_uuid_to_string_harder(p->type_uuid, buffer));
3290 }
3291 }
3292
3293 if (p->growfs >= 0) {
3294 if (gpt_partition_type_knows_growfs(p->type_uuid))
3295 SET_FLAG(f, GPT_FLAG_GROWFS, p->growfs);
3296 else {
3297 char buffer[ID128_UUID_STRING_MAX];
3298 log_warning("Configured GrowFileSystem=%s for partition type '%s' that doesn't support it, ignoring.",
3299 yes_no(p->growfs),
3300 gpt_partition_type_uuid_to_string_harder(p->type_uuid, buffer));
3301 }
3302 }
3303
3304 return f;
3305 }
3306
3307 static int context_mangle_partitions(Context *context) {
3308 Partition *p;
3309 int r;
3310
3311 assert(context);
3312
3313 LIST_FOREACH(partitions, p, context->partitions) {
3314 if (p->dropped)
3315 continue;
3316
3317 assert(p->new_size != UINT64_MAX);
3318 assert(p->offset != UINT64_MAX);
3319 assert(p->partno != UINT64_MAX);
3320
3321 if (PARTITION_EXISTS(p)) {
3322 bool changed = false;
3323
3324 assert(p->current_partition);
3325
3326 if (p->new_size != p->current_size) {
3327 assert(p->new_size >= p->current_size);
3328 assert(p->new_size % 512 == 0);
3329
3330 r = fdisk_partition_size_explicit(p->current_partition, true);
3331 if (r < 0)
3332 return log_error_errno(r, "Failed to enable explicit sizing: %m");
3333
3334 r = fdisk_partition_set_size(p->current_partition, p->new_size / 512);
3335 if (r < 0)
3336 return log_error_errno(r, "Failed to grow partition: %m");
3337
3338 log_info("Growing existing partition %" PRIu64 ".", p->partno);
3339 changed = true;
3340 }
3341
3342 if (!sd_id128_equal(p->new_uuid, p->current_uuid)) {
3343 char buf[ID128_UUID_STRING_MAX];
3344
3345 assert(!sd_id128_is_null(p->new_uuid));
3346
3347 r = fdisk_partition_set_uuid(p->current_partition, id128_to_uuid_string(p->new_uuid, buf));
3348 if (r < 0)
3349 return log_error_errno(r, "Failed to set partition UUID: %m");
3350
3351 log_info("Initializing UUID of existing partition %" PRIu64 ".", p->partno);
3352 changed = true;
3353 }
3354
3355 if (!streq_ptr(p->new_label, p->current_label)) {
3356 r = fdisk_partition_set_name(p->current_partition, strempty(p->new_label));
3357 if (r < 0)
3358 return log_error_errno(r, "Failed to set partition label: %m");
3359
3360 log_info("Setting partition label of existing partition %" PRIu64 ".", p->partno);
3361 changed = true;
3362 }
3363
3364 if (changed) {
3365 assert(!PARTITION_IS_FOREIGN(p)); /* never touch foreign partitions */
3366
3367 r = fdisk_set_partition(context->fdisk_context, p->partno, p->current_partition);
3368 if (r < 0)
3369 return log_error_errno(r, "Failed to update partition: %m");
3370 }
3371 } else {
3372 _cleanup_(fdisk_unref_partitionp) struct fdisk_partition *q = NULL;
3373 _cleanup_(fdisk_unref_parttypep) struct fdisk_parttype *t = NULL;
3374 char ids[ID128_UUID_STRING_MAX];
3375
3376 assert(!p->new_partition);
3377 assert(p->offset % 512 == 0);
3378 assert(p->new_size % 512 == 0);
3379 assert(!sd_id128_is_null(p->new_uuid));
3380 assert(p->new_label);
3381
3382 t = fdisk_new_parttype();
3383 if (!t)
3384 return log_oom();
3385
3386 r = fdisk_parttype_set_typestr(t, id128_to_uuid_string(p->type_uuid, ids));
3387 if (r < 0)
3388 return log_error_errno(r, "Failed to initialize partition type: %m");
3389
3390 q = fdisk_new_partition();
3391 if (!q)
3392 return log_oom();
3393
3394 r = fdisk_partition_set_type(q, t);
3395 if (r < 0)
3396 return log_error_errno(r, "Failed to set partition type: %m");
3397
3398 r = fdisk_partition_size_explicit(q, true);
3399 if (r < 0)
3400 return log_error_errno(r, "Failed to enable explicit sizing: %m");
3401
3402 r = fdisk_partition_set_start(q, p->offset / 512);
3403 if (r < 0)
3404 return log_error_errno(r, "Failed to position partition: %m");
3405
3406 r = fdisk_partition_set_size(q, p->new_size / 512);
3407 if (r < 0)
3408 return log_error_errno(r, "Failed to grow partition: %m");
3409
3410 r = fdisk_partition_set_partno(q, p->partno);
3411 if (r < 0)
3412 return log_error_errno(r, "Failed to set partition number: %m");
3413
3414 r = fdisk_partition_set_uuid(q, id128_to_uuid_string(p->new_uuid, ids));
3415 if (r < 0)
3416 return log_error_errno(r, "Failed to set partition UUID: %m");
3417
3418 r = fdisk_partition_set_name(q, strempty(p->new_label));
3419 if (r < 0)
3420 return log_error_errno(r, "Failed to set partition label: %m");
3421
3422 /* Merge the no auto + read only + growfs setting with the literal flags, and set them for the partition */
3423 r = set_gpt_flags(q, partition_merge_flags(p));
3424 if (r < 0)
3425 return log_error_errno(r, "Failed to set GPT partition flags: %m");
3426
3427 log_info("Adding new partition %" PRIu64 " to partition table.", p->partno);
3428
3429 r = fdisk_add_partition(context->fdisk_context, q, NULL);
3430 if (r < 0)
3431 return log_error_errno(r, "Failed to add partition: %m");
3432
3433 assert(!p->new_partition);
3434 p->new_partition = TAKE_PTR(q);
3435 }
3436 }
3437
3438 return 0;
3439 }
3440
3441 static int context_write_partition_table(
3442 Context *context,
3443 const char *node,
3444 bool from_scratch) {
3445
3446 _cleanup_(fdisk_unref_tablep) struct fdisk_table *original_table = NULL;
3447 int capable, r;
3448
3449 assert(context);
3450
3451 if (arg_pretty > 0 ||
3452 (arg_pretty < 0 && isatty(STDOUT_FILENO) > 0) ||
3453 !FLAGS_SET(arg_json_format_flags, JSON_FORMAT_OFF)) {
3454
3455 (void) context_dump_partitions(context, node);
3456
3457 putc('\n', stdout);
3458
3459 if (arg_json_format_flags & JSON_FORMAT_OFF)
3460 (void) context_dump_partition_bar(context, node);
3461 putc('\n', stdout);
3462 fflush(stdout);
3463 }
3464
3465 if (!from_scratch && !context_changed(context)) {
3466 log_info("No changes.");
3467 return 0;
3468 }
3469
3470 if (arg_dry_run) {
3471 log_notice("Refusing to repartition, please re-run with --dry-run=no.");
3472 return 0;
3473 }
3474
3475 log_info("Applying changes.");
3476
3477 if (from_scratch) {
3478 r = context_wipe_range(context, 0, context->total);
3479 if (r < 0)
3480 return r;
3481
3482 log_info("Wiped block device.");
3483
3484 r = context_discard_range(context, 0, context->total);
3485 if (r == -EOPNOTSUPP)
3486 log_info("Storage does not support discard, not discarding entire block device data.");
3487 else if (r < 0)
3488 return log_error_errno(r, "Failed to discard entire block device: %m");
3489 else if (r > 0)
3490 log_info("Discarded entire block device.");
3491 }
3492
3493 r = fdisk_get_partitions(context->fdisk_context, &original_table);
3494 if (r < 0)
3495 return log_error_errno(r, "Failed to acquire partition table: %m");
3496
3497 /* Wipe fs signatures and discard sectors where the new partitions are going to be placed and in the
3498 * gaps between partitions, just to be sure. */
3499 r = context_wipe_and_discard(context, from_scratch);
3500 if (r < 0)
3501 return r;
3502
3503 r = context_copy_blocks(context);
3504 if (r < 0)
3505 return r;
3506
3507 r = context_mkfs(context);
3508 if (r < 0)
3509 return r;
3510
3511 r = context_mangle_partitions(context);
3512 if (r < 0)
3513 return r;
3514
3515 log_info("Writing new partition table.");
3516
3517 r = fdisk_write_disklabel(context->fdisk_context);
3518 if (r < 0)
3519 return log_error_errno(r, "Failed to write partition table: %m");
3520
3521 capable = blockdev_partscan_enabled(fdisk_get_devfd(context->fdisk_context));
3522 if (capable == -ENOTBLK)
3523 log_debug("Not telling kernel to reread partition table, since we are not operating on a block device.");
3524 else if (capable < 0)
3525 return log_error_errno(capable, "Failed to check if block device supports partition scanning: %m");
3526 else if (capable > 0) {
3527 log_info("Telling kernel to reread partition table.");
3528
3529 if (from_scratch)
3530 r = fdisk_reread_partition_table(context->fdisk_context);
3531 else
3532 r = fdisk_reread_changes(context->fdisk_context, original_table);
3533 if (r < 0)
3534 return log_error_errno(r, "Failed to reread partition table: %m");
3535 } else
3536 log_notice("Not telling kernel to reread partition table, because selected image does not support kernel partition block devices.");
3537
3538 log_info("All done.");
3539
3540 return 0;
3541 }
3542
3543 static int context_read_seed(Context *context, const char *root) {
3544 int r;
3545
3546 assert(context);
3547
3548 if (!sd_id128_is_null(context->seed))
3549 return 0;
3550
3551 if (!arg_randomize) {
3552 _cleanup_close_ int fd = -1;
3553
3554 fd = chase_symlinks_and_open("/etc/machine-id", root, CHASE_PREFIX_ROOT, O_RDONLY|O_CLOEXEC, NULL);
3555 if (fd == -ENOENT)
3556 log_info("No machine ID set, using randomized partition UUIDs.");
3557 else if (fd < 0)
3558 return log_error_errno(fd, "Failed to determine machine ID of image: %m");
3559 else {
3560 r = id128_read_fd(fd, ID128_PLAIN_OR_UNINIT, &context->seed);
3561 if (r == -ENOMEDIUM)
3562 log_info("No machine ID set, using randomized partition UUIDs.");
3563 else if (r < 0)
3564 return log_error_errno(r, "Failed to parse machine ID of image: %m");
3565
3566 return 0;
3567 }
3568 }
3569
3570 r = sd_id128_randomize(&context->seed);
3571 if (r < 0)
3572 return log_error_errno(r, "Failed to generate randomized seed: %m");
3573
3574 return 0;
3575 }
3576
3577 static int context_factory_reset(Context *context, bool from_scratch) {
3578 Partition *p;
3579 size_t n = 0;
3580 int r;
3581
3582 assert(context);
3583
3584 if (arg_factory_reset <= 0)
3585 return 0;
3586
3587 if (from_scratch) /* Nothing to reset if we start from scratch */
3588 return 0;
3589
3590 if (arg_dry_run) {
3591 log_notice("Refusing to factory reset, please re-run with --dry-run=no.");
3592 return 0;
3593 }
3594
3595 log_info("Applying factory reset.");
3596
3597 LIST_FOREACH(partitions, p, context->partitions) {
3598
3599 if (!p->factory_reset || !PARTITION_EXISTS(p))
3600 continue;
3601
3602 assert(p->partno != UINT64_MAX);
3603
3604 log_info("Removing partition %" PRIu64 " for factory reset.", p->partno);
3605
3606 r = fdisk_delete_partition(context->fdisk_context, p->partno);
3607 if (r < 0)
3608 return log_error_errno(r, "Failed to remove partition %" PRIu64 ": %m", p->partno);
3609
3610 n++;
3611 }
3612
3613 if (n == 0) {
3614 log_info("Factory reset requested, but no partitions to delete found.");
3615 return 0;
3616 }
3617
3618 r = fdisk_write_disklabel(context->fdisk_context);
3619 if (r < 0)
3620 return log_error_errno(r, "Failed to write disk label: %m");
3621
3622 log_info("Successfully deleted %zu partitions.", n);
3623 return 1;
3624 }
3625
3626 static int context_can_factory_reset(Context *context) {
3627 Partition *p;
3628
3629 assert(context);
3630
3631 LIST_FOREACH(partitions, p, context->partitions)
3632 if (p->factory_reset && PARTITION_EXISTS(p))
3633 return true;
3634
3635 return false;
3636 }
3637
3638 static int resolve_copy_blocks_auto_candidate(
3639 dev_t partition_devno,
3640 sd_id128_t partition_type_uuid,
3641 dev_t restrict_devno,
3642 sd_id128_t *ret_uuid) {
3643
3644 _cleanup_(blkid_free_probep) blkid_probe b = NULL;
3645 _cleanup_free_ char *p = NULL;
3646 _cleanup_close_ int fd = -1;
3647 const char *pttype, *t;
3648 sd_id128_t pt_parsed, u;
3649 blkid_partition pp;
3650 dev_t whole_devno;
3651 blkid_partlist pl;
3652 struct stat st;
3653 int r;
3654
3655 /* Checks if the specified partition has the specified GPT type UUID, and is located on the specified
3656 * 'restrict_devno' device. The type check is particularly relevant if we have Verity volume which is
3657 * backed by two separate partitions: the data and the hash partitions, and we need to find the right
3658 * one of the two. */
3659
3660 r = block_get_whole_disk(partition_devno, &whole_devno);
3661 if (r < 0)
3662 return log_error_errno(
3663 r,
3664 "Unable to determine containing block device of partition %u:%u: %m",
3665 major(partition_devno), minor(partition_devno));
3666
3667 if (restrict_devno != (dev_t) -1 &&
3668 restrict_devno != whole_devno)
3669 return log_error_errno(
3670 SYNTHETIC_ERRNO(EPERM),
3671 "Partition %u:%u is located outside of block device %u:%u, refusing.",
3672 major(partition_devno), minor(partition_devno),
3673 major(restrict_devno), minor(restrict_devno));
3674
3675 r = device_path_make_major_minor(S_IFBLK, whole_devno, &p);
3676 if (r < 0)
3677 return log_error_errno(r, "Failed to convert block device to device node path: %m");
3678
3679 fd = open(p, O_RDONLY|O_CLOEXEC|O_NONBLOCK);
3680 if (fd < 0)
3681 return log_error_errno(r, "Failed to open '%s': %m", p);
3682
3683 if (fstat(fd, &st) < 0)
3684 return log_error_errno(r, "Failed to stat '%s': %m", p);
3685
3686 if (!S_ISBLK(st.st_mode) || st.st_rdev != whole_devno)
3687 return log_error_errno(
3688 SYNTHETIC_ERRNO(EPERM),
3689 "Opened and determined block device don't match, refusing.");
3690
3691 b = blkid_new_probe();
3692 if (!b)
3693 return log_oom();
3694
3695 errno = 0;
3696 r = blkid_probe_set_device(b, fd, 0, 0);
3697 if (r != 0)
3698 return log_error_errno(errno_or_else(ENOMEM), "Failed to open block device '%s': %m", p);
3699
3700 (void) blkid_probe_enable_partitions(b, 1);
3701 (void) blkid_probe_set_partitions_flags(b, BLKID_PARTS_ENTRY_DETAILS);
3702
3703 errno = 0;
3704 r = blkid_do_safeprobe(b);
3705 if (IN_SET(r, -2, 1)) { /* nothing found or ambiguous result */
3706 log_debug("Didn't find partition table on block device '%s'.", p);
3707 return false;
3708 }
3709 if (r != 0)
3710 return log_error_errno(errno_or_else(EIO), "Unable to probe for partition table of '%s': %m", p);
3711
3712 (void) blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL);
3713 if (!streq_ptr(pttype, "gpt")) {
3714 log_debug("Didn't find a GPT partition table on '%s'.", p);
3715 return false;
3716 }
3717
3718 errno = 0;
3719 pl = blkid_probe_get_partitions(b);
3720 if (!pl)
3721 return log_error_errno(errno_or_else(EIO), "Unable read partition table of '%s': %m", p);
3722 errno = 0;
3723
3724 pp = blkid_partlist_devno_to_partition(pl, partition_devno);
3725 if (!pp) {
3726 log_debug("Partition %u:%u has no matching partition table entry on '%s'.",
3727 major(partition_devno), minor(partition_devno), p);
3728 return false;
3729 }
3730
3731 t = blkid_partition_get_type_string(pp);
3732 if (isempty(t)) {
3733 log_debug("Partition %u:%u has no type on '%s'.",
3734 major(partition_devno), minor(partition_devno), p);
3735 return false;
3736 }
3737
3738 r = sd_id128_from_string(t, &pt_parsed);
3739 if (r < 0) {
3740 log_debug_errno(r, "Failed to parse partition type \"%s\": %m", t);
3741 return false;
3742 }
3743
3744 if (!sd_id128_equal(pt_parsed, partition_type_uuid)) {
3745 log_debug("Partition %u:%u has non-matching partition type " SD_ID128_FORMAT_STR " (needed: " SD_ID128_FORMAT_STR "), ignoring.",
3746 major(partition_devno), minor(partition_devno),
3747 SD_ID128_FORMAT_VAL(pt_parsed), SD_ID128_FORMAT_VAL(partition_type_uuid));
3748 return false;
3749 }
3750
3751 t = blkid_partition_get_uuid(pp);
3752 if (isempty(t)) {
3753 log_debug("Partition %u:%u has no UUID.",
3754 major(partition_devno), minor(partition_devno));
3755 return false;
3756 }
3757
3758 r = sd_id128_from_string(t, &u);
3759 if (r < 0) {
3760 log_debug_errno(r, "Failed to parse partition UUID \"%s\": %m", t);
3761 return false;
3762 }
3763
3764 log_debug("Automatically found partition %u:%u of right type " SD_ID128_FORMAT_STR ".",
3765 major(partition_devno), minor(partition_devno),
3766 SD_ID128_FORMAT_VAL(pt_parsed));
3767
3768 if (ret_uuid)
3769 *ret_uuid = u;
3770
3771 return true;
3772 }
3773
3774 static int find_backing_devno(
3775 const char *path,
3776 const char *root,
3777 dev_t *ret) {
3778
3779 _cleanup_free_ char *resolved = NULL;
3780 int r;
3781
3782 assert(path);
3783
3784 r = chase_symlinks(path, root, CHASE_PREFIX_ROOT, &resolved, NULL);
3785 if (r < 0)
3786 return r;
3787
3788 r = path_is_mount_point(resolved, NULL, 0);
3789 if (r < 0)
3790 return r;
3791 if (r == 0) /* Not a mount point, then it's not a partition of its own, let's not automatically use it. */
3792 return -ENOENT;
3793
3794 r = get_block_device(resolved, ret);
3795 if (r < 0)
3796 return r;
3797 if (r == 0) /* Not backed by physical file system, we can't use this */
3798 return -ENOENT;
3799
3800 return 0;
3801 }
3802
3803 static int resolve_copy_blocks_auto(
3804 sd_id128_t type_uuid,
3805 const char *root,
3806 dev_t restrict_devno,
3807 char **ret_path,
3808 sd_id128_t *ret_uuid) {
3809
3810 const char *try1 = NULL, *try2 = NULL;
3811 char p[SYS_BLOCK_PATH_MAX("/slaves")];
3812 _cleanup_(closedirp) DIR *d = NULL;
3813 sd_id128_t found_uuid = SD_ID128_NULL;
3814 dev_t devno, found = 0;
3815 int r;
3816
3817 assert(ret_path);
3818
3819 /* Enforce some security restrictions: CopyBlocks=auto should not be an avenue to get outside of the
3820 * --root=/--image= confinement. Specifically, refuse CopyBlocks= in combination with --root= at all,
3821 * and restrict block device references in the --image= case to loopback block device we set up.
3822 *
3823 * restrict_devno contain the dev_t of the loop back device we operate on in case of --image=, and
3824 * thus declares which device (and its partition subdevices) we shall limit access to. If
3825 * restrict_devno is zero no device probing access shall be allowed at all (used for --root=) and if
3826 * it is (dev_t) -1 then free access shall be allowed (if neither switch is used). */
3827
3828 if (restrict_devno == 0)
3829 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
3830 "Automatic discovery of backing block devices not permitted in --root= mode, refusing.");
3831
3832 /* Handles CopyBlocks=auto, and finds the right source partition to copy from. We look for matching
3833 * partitions in the host, using the appropriate directory as key and ensuring that the partition
3834 * type matches. */
3835
3836 if (gpt_partition_type_is_root(type_uuid))
3837 try1 = "/";
3838 else if (gpt_partition_type_is_usr(type_uuid))
3839 try1 = "/usr/";
3840 else if (gpt_partition_type_is_root_verity(type_uuid))
3841 try1 = "/";
3842 else if (gpt_partition_type_is_usr_verity(type_uuid))
3843 try1 = "/usr/";
3844 else if (sd_id128_equal(type_uuid, GPT_ESP)) {
3845 try1 = "/efi/";
3846 try2 = "/boot/";
3847 } else if (sd_id128_equal(type_uuid, GPT_XBOOTLDR))
3848 try1 = "/boot/";
3849 else
3850 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
3851 "Partition type " SD_ID128_FORMAT_STR " not supported from automatic source block device discovery.",
3852 SD_ID128_FORMAT_VAL(type_uuid));
3853
3854 r = find_backing_devno(try1, root, &devno);
3855 if (r == -ENOENT && try2)
3856 r = find_backing_devno(try2, root, &devno);
3857 if (r < 0)
3858 return log_error_errno(r, "Failed to resolve automatic CopyBlocks= path for partition type " SD_ID128_FORMAT_STR ", sorry: %m",
3859 SD_ID128_FORMAT_VAL(type_uuid));
3860
3861 xsprintf_sys_block_path(p, "/slaves", devno);
3862 d = opendir(p);
3863 if (d) {
3864 struct dirent *de;
3865
3866 for (;;) {
3867 _cleanup_free_ char *q = NULL, *t = NULL;
3868 sd_id128_t u;
3869 dev_t sl;
3870
3871 errno = 0;
3872 de = readdir_no_dot(d);
3873 if (!de) {
3874 if (errno != 0)
3875 return log_error_errno(errno, "Failed to read directory '%s': %m", p);
3876
3877 break;
3878 }
3879
3880 if (!IN_SET(de->d_type, DT_LNK, DT_UNKNOWN))
3881 continue;
3882
3883 q = path_join(p, de->d_name, "/dev");
3884 if (!q)
3885 return log_oom();
3886
3887 r = read_one_line_file(q, &t);
3888 if (r < 0)
3889 return log_error_errno(r, "Failed to read %s: %m", q);
3890
3891 r = parse_dev(t, &sl);
3892 if (r < 0) {
3893 log_debug_errno(r, "Failed to parse %s, ignoring: %m", q);
3894 continue;
3895 }
3896 if (major(sl) == 0) {
3897 log_debug_errno(r, "Device backing %s is special, ignoring: %m", q);
3898 continue;
3899 }
3900
3901 r = resolve_copy_blocks_auto_candidate(sl, type_uuid, restrict_devno, &u);
3902 if (r < 0)
3903 return r;
3904 if (r > 0) {
3905 /* We found a matching one! */
3906 if (found != 0)
3907 return log_error_errno(SYNTHETIC_ERRNO(ENOTUNIQ),
3908 "Multiple matching partitions found, refusing.");
3909
3910 found = sl;
3911 found_uuid = u;
3912 }
3913 }
3914 } else if (errno != ENOENT)
3915 return log_error_errno(errno, "Failed open %s: %m", p);
3916 else {
3917 r = resolve_copy_blocks_auto_candidate(devno, type_uuid, restrict_devno, &found_uuid);
3918 if (r < 0)
3919 return r;
3920 if (r > 0)
3921 found = devno;
3922 }
3923
3924 if (found == 0)
3925 return log_error_errno(SYNTHETIC_ERRNO(ENXIO),
3926 "Unable to automatically discover suitable partition to copy blocks from.");
3927
3928 r = device_path_make_major_minor(S_IFBLK, found, ret_path);
3929 if (r < 0)
3930 return log_error_errno(r, "Failed to convert dev_t to device node path: %m");
3931
3932 if (ret_uuid)
3933 *ret_uuid = found_uuid;
3934
3935 return 0;
3936 }
3937
3938 static int context_open_copy_block_paths(
3939 Context *context,
3940 const char *root,
3941 dev_t restrict_devno) {
3942
3943 Partition *p;
3944 int r;
3945
3946 assert(context);
3947
3948 LIST_FOREACH(partitions, p, context->partitions) {
3949 _cleanup_close_ int source_fd = -1;
3950 _cleanup_free_ char *opened = NULL;
3951 sd_id128_t uuid = SD_ID128_NULL;
3952 uint64_t size;
3953 struct stat st;
3954
3955 assert(p->copy_blocks_fd < 0);
3956 assert(p->copy_blocks_size == UINT64_MAX);
3957
3958 if (PARTITION_EXISTS(p)) /* Never copy over partitions that already exist! */
3959 continue;
3960
3961 if (p->copy_blocks_path) {
3962
3963 source_fd = chase_symlinks_and_open(p->copy_blocks_path, root, CHASE_PREFIX_ROOT, O_RDONLY|O_CLOEXEC|O_NONBLOCK, &opened);
3964 if (source_fd < 0)
3965 return log_error_errno(source_fd, "Failed to open '%s': %m", p->copy_blocks_path);
3966
3967 if (fstat(source_fd, &st) < 0)
3968 return log_error_errno(errno, "Failed to stat block copy file '%s': %m", opened);
3969
3970 if (!S_ISREG(st.st_mode) && restrict_devno != (dev_t) -1)
3971 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
3972 "Copying from block device node is not permitted in --image=/--root= mode, refusing.");
3973
3974 } else if (p->copy_blocks_auto) {
3975
3976 r = resolve_copy_blocks_auto(p->type_uuid, root, restrict_devno, &opened, &uuid);
3977 if (r < 0)
3978 return r;
3979
3980 source_fd = open(opened, O_RDONLY|O_CLOEXEC|O_NOCTTY);
3981 if (source_fd < 0)
3982 return log_error_errno(errno, "Failed to open automatically determined source block copy device '%s': %m", opened);
3983
3984 if (fstat(source_fd, &st) < 0)
3985 return log_error_errno(errno, "Failed to stat block copy file '%s': %m", opened);
3986
3987 /* If we found it automatically, it must be a block device, let's enforce that */
3988 if (!S_ISBLK(st.st_mode))
3989 return log_error_errno(SYNTHETIC_ERRNO(EBADF),
3990 "Automatically detected source block copy device '%s' is not a block device, refusing: %m", opened);
3991 } else
3992 continue;
3993
3994 if (S_ISDIR(st.st_mode)) {
3995 _cleanup_free_ char *bdev = NULL;
3996
3997 /* If the file is a directory, automatically find the backing block device */
3998
3999 if (major(st.st_dev) != 0)
4000 r = device_path_make_major_minor(S_IFBLK, st.st_dev, &bdev);
4001 else {
4002 dev_t devt;
4003
4004 /* Special support for btrfs */
4005
4006 r = btrfs_get_block_device_fd(source_fd, &devt);
4007 if (r == -EUCLEAN)
4008 return btrfs_log_dev_root(LOG_ERR, r, opened);
4009 if (r < 0)
4010 return log_error_errno(r, "Unable to determine backing block device of '%s': %m", opened);
4011
4012 r = device_path_make_major_minor(S_IFBLK, devt, &bdev);
4013 }
4014 if (r < 0)
4015 return log_error_errno(r, "Failed to determine block device path for block device backing '%s': %m", opened);
4016
4017 safe_close(source_fd);
4018
4019 source_fd = open(bdev, O_RDONLY|O_CLOEXEC|O_NOCTTY);
4020 if (source_fd < 0)
4021 return log_error_errno(errno, "Failed to open block device '%s': %m", bdev);
4022
4023 if (fstat(source_fd, &st) < 0)
4024 return log_error_errno(errno, "Failed to stat block device '%s': %m", bdev);
4025
4026 if (!S_ISBLK(st.st_mode))
4027 return log_error_errno(SYNTHETIC_ERRNO(ENOTBLK), "Block device '%s' is not actually a block device, refusing.", bdev);
4028 }
4029
4030 if (S_ISREG(st.st_mode))
4031 size = st.st_size;
4032 else if (S_ISBLK(st.st_mode)) {
4033 if (ioctl(source_fd, BLKGETSIZE64, &size) != 0)
4034 return log_error_errno(errno, "Failed to determine size of block device to copy from: %m");
4035 } else
4036 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);
4037
4038 if (size <= 0)
4039 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "File to copy bytes from '%s' has zero size, refusing.", opened);
4040 if (size % 512 != 0)
4041 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "File to copy bytes from '%s' has size that is not multiple of 512, refusing.", opened);
4042
4043 p->copy_blocks_fd = TAKE_FD(source_fd);
4044 p->copy_blocks_size = size;
4045
4046 free_and_replace(p->copy_blocks_path, opened);
4047
4048 /* When copying from an existing partition copy that partitions UUID if none is configured explicitly */
4049 if (sd_id128_is_null(p->new_uuid) && !sd_id128_is_null(uuid))
4050 p->new_uuid = uuid;
4051 }
4052
4053 return 0;
4054 }
4055
4056 static int help(void) {
4057 _cleanup_free_ char *link = NULL;
4058 int r;
4059
4060 r = terminal_urlify_man("systemd-repart", "1", &link);
4061 if (r < 0)
4062 return log_oom();
4063
4064 printf("%s [OPTIONS...] [DEVICE]\n"
4065 "\n%sGrow and add partitions to partition table.%s\n\n"
4066 " -h --help Show this help\n"
4067 " --version Show package version\n"
4068 " --no-pager Do not pipe output into a pager\n"
4069 " --no-legend Do not show the headers and footers\n"
4070 " --dry-run=BOOL Whether to run dry-run operation\n"
4071 " --empty=MODE One of refuse, allow, require, force, create; controls\n"
4072 " how to handle empty disks lacking partition tables\n"
4073 " --discard=BOOL Whether to discard backing blocks for new partitions\n"
4074 " --pretty=BOOL Whether to show pretty summary before doing changes\n"
4075 " --factory-reset=BOOL Whether to remove data partitions before recreating\n"
4076 " them\n"
4077 " --can-factory-reset Test whether factory reset is defined\n"
4078 " --root=PATH Operate relative to root path\n"
4079 " --image=PATH Operate relative to image file\n"
4080 " --definitions=DIR Find partition definitions in specified directory\n"
4081 " --key-file=PATH Key to use when encrypting partitions\n"
4082 " --tpm2-device=PATH Path to TPM2 device node to use\n"
4083 " --tpm2-pcrs=PCR1+PCR2+PCR3+…\n"
4084 " TPM2 PCR indexes to use for TPM2 enrollment\n"
4085 " --seed=UUID 128bit seed UUID to derive all UUIDs from\n"
4086 " --size=BYTES Grow loopback file to specified size\n"
4087 " --json=pretty|short|off\n"
4088 " Generate JSON output\n"
4089 "\nSee the %s for details.\n",
4090 program_invocation_short_name,
4091 ansi_highlight(),
4092 ansi_normal(),
4093 link);
4094
4095 return 0;
4096 }
4097
4098 static int parse_argv(int argc, char *argv[]) {
4099
4100 enum {
4101 ARG_VERSION = 0x100,
4102 ARG_NO_PAGER,
4103 ARG_NO_LEGEND,
4104 ARG_DRY_RUN,
4105 ARG_EMPTY,
4106 ARG_DISCARD,
4107 ARG_FACTORY_RESET,
4108 ARG_CAN_FACTORY_RESET,
4109 ARG_ROOT,
4110 ARG_IMAGE,
4111 ARG_SEED,
4112 ARG_PRETTY,
4113 ARG_DEFINITIONS,
4114 ARG_SIZE,
4115 ARG_JSON,
4116 ARG_KEY_FILE,
4117 ARG_TPM2_DEVICE,
4118 ARG_TPM2_PCRS,
4119 };
4120
4121 static const struct option options[] = {
4122 { "help", no_argument, NULL, 'h' },
4123 { "version", no_argument, NULL, ARG_VERSION },
4124 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
4125 { "no-legend", no_argument, NULL, ARG_NO_LEGEND },
4126 { "dry-run", required_argument, NULL, ARG_DRY_RUN },
4127 { "empty", required_argument, NULL, ARG_EMPTY },
4128 { "discard", required_argument, NULL, ARG_DISCARD },
4129 { "factory-reset", required_argument, NULL, ARG_FACTORY_RESET },
4130 { "can-factory-reset", no_argument, NULL, ARG_CAN_FACTORY_RESET },
4131 { "root", required_argument, NULL, ARG_ROOT },
4132 { "image", required_argument, NULL, ARG_IMAGE },
4133 { "seed", required_argument, NULL, ARG_SEED },
4134 { "pretty", required_argument, NULL, ARG_PRETTY },
4135 { "definitions", required_argument, NULL, ARG_DEFINITIONS },
4136 { "size", required_argument, NULL, ARG_SIZE },
4137 { "json", required_argument, NULL, ARG_JSON },
4138 { "key-file", required_argument, NULL, ARG_KEY_FILE },
4139 { "tpm2-device", required_argument, NULL, ARG_TPM2_DEVICE },
4140 { "tpm2-pcrs", required_argument, NULL, ARG_TPM2_PCRS },
4141 {}
4142 };
4143
4144 int c, r, dry_run = -1;
4145
4146 assert(argc >= 0);
4147 assert(argv);
4148
4149 while ((c = getopt_long(argc, argv, "h", options, NULL)) >= 0)
4150
4151 switch (c) {
4152
4153 case 'h':
4154 return help();
4155
4156 case ARG_VERSION:
4157 return version();
4158
4159 case ARG_NO_PAGER:
4160 arg_pager_flags |= PAGER_DISABLE;
4161 break;
4162
4163 case ARG_NO_LEGEND:
4164 arg_legend = false;
4165 break;
4166
4167 case ARG_DRY_RUN:
4168 r = parse_boolean_argument("--dry-run=", optarg, &arg_dry_run);
4169 if (r < 0)
4170 return r;
4171 break;
4172
4173 case ARG_EMPTY:
4174 if (isempty(optarg) || streq(optarg, "refuse"))
4175 arg_empty = EMPTY_REFUSE;
4176 else if (streq(optarg, "allow"))
4177 arg_empty = EMPTY_ALLOW;
4178 else if (streq(optarg, "require"))
4179 arg_empty = EMPTY_REQUIRE;
4180 else if (streq(optarg, "force"))
4181 arg_empty = EMPTY_FORCE;
4182 else if (streq(optarg, "create")) {
4183 arg_empty = EMPTY_CREATE;
4184
4185 if (dry_run < 0)
4186 dry_run = false; /* Imply --dry-run=no if we create the loopback file
4187 * anew. After all we cannot really break anyone's
4188 * partition tables that way. */
4189 } else
4190 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
4191 "Failed to parse --empty= parameter: %s", optarg);
4192 break;
4193
4194 case ARG_DISCARD:
4195 r = parse_boolean_argument("--discard=", optarg, &arg_discard);
4196 if (r < 0)
4197 return r;
4198 break;
4199
4200 case ARG_FACTORY_RESET:
4201 r = parse_boolean_argument("--factory-reset=", optarg, NULL);
4202 if (r < 0)
4203 return r;
4204 arg_factory_reset = r;
4205 break;
4206
4207 case ARG_CAN_FACTORY_RESET:
4208 arg_can_factory_reset = true;
4209 break;
4210
4211 case ARG_ROOT:
4212 r = parse_path_argument(optarg, /* suppress_root= */ false, &arg_root);
4213 if (r < 0)
4214 return r;
4215 break;
4216
4217 case ARG_IMAGE:
4218 r = parse_path_argument(optarg, /* suppress_root= */ false, &arg_image);
4219 if (r < 0)
4220 return r;
4221 break;
4222
4223 case ARG_SEED:
4224 if (isempty(optarg)) {
4225 arg_seed = SD_ID128_NULL;
4226 arg_randomize = false;
4227 } else if (streq(optarg, "random"))
4228 arg_randomize = true;
4229 else {
4230 r = sd_id128_from_string(optarg, &arg_seed);
4231 if (r < 0)
4232 return log_error_errno(r, "Failed to parse seed: %s", optarg);
4233
4234 arg_randomize = false;
4235 }
4236
4237 break;
4238
4239 case ARG_PRETTY:
4240 r = parse_boolean_argument("--pretty=", optarg, NULL);
4241 if (r < 0)
4242 return r;
4243 arg_pretty = r;
4244 break;
4245
4246 case ARG_DEFINITIONS:
4247 r = parse_path_argument(optarg, false, &arg_definitions);
4248 if (r < 0)
4249 return r;
4250 break;
4251
4252 case ARG_SIZE: {
4253 uint64_t parsed, rounded;
4254
4255 if (streq(optarg, "auto")) {
4256 arg_size = UINT64_MAX;
4257 arg_size_auto = true;
4258 break;
4259 }
4260
4261 r = parse_size(optarg, 1024, &parsed);
4262 if (r < 0)
4263 return log_error_errno(r, "Failed to parse --size= parameter: %s", optarg);
4264
4265 rounded = round_up_size(parsed, 4096);
4266 if (rounded == 0)
4267 return log_error_errno(SYNTHETIC_ERRNO(ERANGE), "Specified image size too small, refusing.");
4268 if (rounded == UINT64_MAX)
4269 return log_error_errno(SYNTHETIC_ERRNO(ERANGE), "Specified image size too large, refusing.");
4270
4271 if (rounded != parsed)
4272 log_warning("Specified size is not a multiple of 4096, rounding up automatically. (%" PRIu64 " → %" PRIu64 ")",
4273 parsed, rounded);
4274
4275 arg_size = rounded;
4276 arg_size_auto = false;
4277 break;
4278 }
4279
4280 case ARG_JSON:
4281 r = parse_json_argument(optarg, &arg_json_format_flags);
4282 if (r <= 0)
4283 return r;
4284
4285 break;
4286
4287 case ARG_KEY_FILE: {
4288 _cleanup_(erase_and_freep) char *k = NULL;
4289 size_t n = 0;
4290
4291 r = read_full_file_full(
4292 AT_FDCWD, optarg, UINT64_MAX, SIZE_MAX,
4293 READ_FULL_FILE_SECURE|READ_FULL_FILE_WARN_WORLD_READABLE|READ_FULL_FILE_CONNECT_SOCKET,
4294 NULL,
4295 &k, &n);
4296 if (r < 0)
4297 return log_error_errno(r, "Failed to read key file '%s': %m", optarg);
4298
4299 erase_and_free(arg_key);
4300 arg_key = TAKE_PTR(k);
4301 arg_key_size = n;
4302 break;
4303 }
4304
4305 case ARG_TPM2_DEVICE: {
4306 _cleanup_free_ char *device = NULL;
4307
4308 if (streq(optarg, "list"))
4309 return tpm2_list_devices();
4310
4311 if (!streq(optarg, "auto")) {
4312 device = strdup(optarg);
4313 if (!device)
4314 return log_oom();
4315 }
4316
4317 free(arg_tpm2_device);
4318 arg_tpm2_device = TAKE_PTR(device);
4319 break;
4320 }
4321
4322 case ARG_TPM2_PCRS: {
4323 uint32_t mask;
4324
4325 if (isempty(optarg)) {
4326 arg_tpm2_pcr_mask = 0;
4327 break;
4328 }
4329
4330 r = tpm2_parse_pcrs(optarg, &mask);
4331 if (r < 0)
4332 return r;
4333
4334 if (arg_tpm2_pcr_mask == UINT32_MAX)
4335 arg_tpm2_pcr_mask = mask;
4336 else
4337 arg_tpm2_pcr_mask |= mask;
4338
4339 break;
4340 }
4341
4342 case '?':
4343 return -EINVAL;
4344
4345 default:
4346 assert_not_reached();
4347 }
4348
4349 if (argc - optind > 1)
4350 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
4351 "Expected at most one argument, the path to the block device.");
4352
4353 if (arg_factory_reset > 0 && IN_SET(arg_empty, EMPTY_FORCE, EMPTY_REQUIRE, EMPTY_CREATE))
4354 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
4355 "Combination of --factory-reset=yes and --empty=force/--empty=require/--empty=create is invalid.");
4356
4357 if (arg_can_factory_reset)
4358 arg_dry_run = true; /* When --can-factory-reset is specified we don't make changes, hence
4359 * non-dry-run mode makes no sense. Thus, imply dry run mode so that we
4360 * open things strictly read-only. */
4361 else if (dry_run >= 0)
4362 arg_dry_run = dry_run;
4363
4364 if (arg_empty == EMPTY_CREATE && (arg_size == UINT64_MAX && !arg_size_auto))
4365 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
4366 "If --empty=create is specified, --size= must be specified, too.");
4367
4368 if (arg_image && arg_root)
4369 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Please specify either --root= or --image=, the combination of both is not supported.");
4370 else if (!arg_image && !arg_root && in_initrd()) {
4371
4372 /* By default operate on /sysusr/ or /sysroot/ when invoked in the initrd. We prefer the
4373 * former, if it is mounted, so that we have deterministic behaviour on systems where /usr/
4374 * is vendor-supplied but the root fs formatted on first boot. */
4375 r = path_is_mount_point("/sysusr/usr", NULL, 0);
4376 if (r <= 0) {
4377 if (r < 0 && r != -ENOENT)
4378 log_debug_errno(r, "Unable to determine whether /sysusr/usr is a mount point, assuming it is not: %m");
4379
4380 arg_root = strdup("/sysroot");
4381 } else
4382 arg_root = strdup("/sysusr");
4383 if (!arg_root)
4384 return log_oom();
4385 }
4386
4387 arg_node = argc > optind ? argv[optind] : NULL;
4388
4389 if (IN_SET(arg_empty, EMPTY_FORCE, EMPTY_REQUIRE, EMPTY_CREATE) && !arg_node && !arg_image)
4390 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
4391 "A path to a device node or loopback file must be specified when --empty=force, --empty=require or --empty=create are used.");
4392
4393 if (arg_tpm2_pcr_mask == UINT32_MAX)
4394 arg_tpm2_pcr_mask = TPM2_PCR_MASK_DEFAULT;
4395
4396 return 1;
4397 }
4398
4399 static int parse_proc_cmdline_factory_reset(void) {
4400 bool b;
4401 int r;
4402
4403 if (arg_factory_reset >= 0) /* Never override what is specified on the process command line */
4404 return 0;
4405
4406 if (!in_initrd()) /* Never honour kernel command line factory reset request outside of the initrd */
4407 return 0;
4408
4409 r = proc_cmdline_get_bool("systemd.factory_reset", &b);
4410 if (r < 0)
4411 return log_error_errno(r, "Failed to parse systemd.factory_reset kernel command line argument: %m");
4412 if (r > 0) {
4413 arg_factory_reset = b;
4414
4415 if (b)
4416 log_notice("Honouring factory reset requested via kernel command line.");
4417 }
4418
4419 return 0;
4420 }
4421
4422 static int parse_efi_variable_factory_reset(void) {
4423 _cleanup_free_ char *value = NULL;
4424 int r;
4425
4426 if (arg_factory_reset >= 0) /* Never override what is specified on the process command line */
4427 return 0;
4428
4429 if (!in_initrd()) /* Never honour EFI variable factory reset request outside of the initrd */
4430 return 0;
4431
4432 r = efi_get_variable_string(EFI_SYSTEMD_VARIABLE(FactoryReset), &value);
4433 if (r == -ENOENT || ERRNO_IS_NOT_SUPPORTED(r))
4434 return 0;
4435 if (r < 0)
4436 return log_error_errno(r, "Failed to read EFI variable FactoryReset: %m");
4437
4438 r = parse_boolean(value);
4439 if (r < 0)
4440 return log_error_errno(r, "Failed to parse EFI variable FactoryReset: %m");
4441
4442 arg_factory_reset = r;
4443 if (r)
4444 log_notice("Factory reset requested via EFI variable FactoryReset.");
4445
4446 return 0;
4447 }
4448
4449 static int remove_efi_variable_factory_reset(void) {
4450 int r;
4451
4452 r = efi_set_variable(EFI_SYSTEMD_VARIABLE(FactoryReset), NULL, 0);
4453 if (r == -ENOENT || ERRNO_IS_NOT_SUPPORTED(r))
4454 return 0;
4455 if (r < 0)
4456 return log_error_errno(r, "Failed to remove EFI variable FactoryReset: %m");
4457
4458 log_info("Successfully unset EFI variable FactoryReset.");
4459 return 0;
4460 }
4461
4462 static int acquire_root_devno(
4463 const char *p,
4464 const char *root,
4465 int mode,
4466 char **ret,
4467 int *ret_fd) {
4468
4469 _cleanup_free_ char *found_path = NULL;
4470 dev_t devno, fd_devno = MODE_INVALID;
4471 _cleanup_close_ int fd = -1;
4472 struct stat st;
4473 int r;
4474
4475 assert(p);
4476 assert(ret);
4477 assert(ret_fd);
4478
4479 fd = chase_symlinks_and_open(p, root, CHASE_PREFIX_ROOT, mode, &found_path);
4480 if (fd < 0)
4481 return fd;
4482
4483 if (fstat(fd, &st) < 0)
4484 return -errno;
4485
4486 if (S_ISREG(st.st_mode)) {
4487 *ret = TAKE_PTR(found_path);
4488 *ret_fd = TAKE_FD(fd);
4489 return 0;
4490 }
4491
4492 if (S_ISBLK(st.st_mode)) {
4493 /* Refuse referencing explicit block devices if a root dir is specified, after all we should
4494 * not be able to leave the image the root path constrains us to. */
4495 if (root)
4496 return -EPERM;
4497
4498 fd_devno = devno = st.st_rdev;
4499 } else if (S_ISDIR(st.st_mode)) {
4500
4501 devno = st.st_dev;
4502 if (major(devno) == 0) {
4503 r = btrfs_get_block_device_fd(fd, &devno);
4504 if (r == -ENOTTY) /* not btrfs */
4505 return -ENODEV;
4506 if (r < 0)
4507 return r;
4508 }
4509 } else
4510 return -ENOTBLK;
4511
4512 /* From dm-crypt to backing partition */
4513 r = block_get_originating(devno, &devno);
4514 if (r == -ENOENT)
4515 log_debug_errno(r, "Device '%s' has no dm-crypt/dm-verity device, no need to look for underlying block device.", p);
4516 else if (r < 0)
4517 log_debug_errno(r, "Failed to find underlying block device for '%s', ignoring: %m", p);
4518
4519 /* From partition to whole disk containing it */
4520 r = block_get_whole_disk(devno, &devno);
4521 if (r < 0)
4522 log_debug_errno(r, "Failed to find whole disk block device for '%s', ignoring: %m", p);
4523
4524 r = device_path_make_canonical(S_IFBLK, devno, ret);
4525 if (r < 0)
4526 return log_debug_errno(r, "Failed to determine canonical path for '%s': %m", p);
4527
4528 /* Only if we still look at the same block device we can reuse the fd. Otherwise return an
4529 * invalidated fd. */
4530 *ret_fd = fd_devno != MODE_INVALID && fd_devno == devno ? TAKE_FD(fd) : -1;
4531 return 0;
4532 }
4533
4534 static int find_root(char **ret, int *ret_fd) {
4535 const char *p;
4536 int r;
4537
4538 assert(ret);
4539 assert(ret_fd);
4540
4541 if (arg_node) {
4542 if (arg_empty == EMPTY_CREATE) {
4543 _cleanup_close_ int fd = -1;
4544 _cleanup_free_ char *s = NULL;
4545
4546 s = strdup(arg_node);
4547 if (!s)
4548 return log_oom();
4549
4550 fd = open(arg_node, O_RDONLY|O_CREAT|O_EXCL|O_CLOEXEC|O_NOFOLLOW, 0666);
4551 if (fd < 0)
4552 return log_error_errno(errno, "Failed to create '%s': %m", arg_node);
4553
4554 *ret = TAKE_PTR(s);
4555 *ret_fd = TAKE_FD(fd);
4556 return 0;
4557 }
4558
4559 /* Note that we don't specify a root argument here: if the user explicitly configured a node
4560 * we'll take it relative to the host, not the image */
4561 r = acquire_root_devno(arg_node, NULL, O_RDONLY|O_CLOEXEC, ret, ret_fd);
4562 if (r == -EUCLEAN)
4563 return btrfs_log_dev_root(LOG_ERR, r, arg_node);
4564 if (r < 0)
4565 return log_error_errno(r, "Failed to open file or determine backing device of %s: %m", arg_node);
4566
4567 return 0;
4568 }
4569
4570 assert(IN_SET(arg_empty, EMPTY_REFUSE, EMPTY_ALLOW));
4571
4572 /* Let's search for the root device. We look for two cases here: first in /, and then in /usr. The
4573 * latter we check for cases where / is a tmpfs and only /usr is an actual persistent block device
4574 * (think: volatile setups) */
4575
4576 FOREACH_STRING(p, "/", "/usr") {
4577
4578 r = acquire_root_devno(p, arg_root, O_RDONLY|O_DIRECTORY|O_CLOEXEC, ret, ret_fd);
4579 if (r < 0) {
4580 if (r == -EUCLEAN)
4581 return btrfs_log_dev_root(LOG_ERR, r, p);
4582 if (r != -ENODEV)
4583 return log_error_errno(r, "Failed to determine backing device of %s: %m", p);
4584 } else
4585 return 0;
4586 }
4587
4588 return log_error_errno(SYNTHETIC_ERRNO(ENODEV), "Failed to discover root block device.");
4589 }
4590
4591 static int resize_pt(int fd) {
4592 _cleanup_(fdisk_unref_contextp) struct fdisk_context *c = NULL;
4593 int r;
4594
4595 /* After resizing the backing file we need to resize the partition table itself too, so that it takes
4596 * possession of the enlarged backing file. For this it suffices to open the device with libfdisk and
4597 * immediately write it again, with no changes. */
4598
4599 c = fdisk_new_context();
4600 if (!c)
4601 return log_oom();
4602
4603 r = fdisk_assign_device(c, FORMAT_PROC_FD_PATH(fd), 0);
4604 if (r < 0)
4605 return log_error_errno(r, "Failed to open device '%s': %m", FORMAT_PROC_FD_PATH(fd));
4606
4607 r = fdisk_has_label(c);
4608 if (r < 0)
4609 return log_error_errno(r, "Failed to determine whether disk '%s' has a disk label: %m", FORMAT_PROC_FD_PATH(fd));
4610 if (r == 0) {
4611 log_debug("Not resizing partition table, as there currently is none.");
4612 return 0;
4613 }
4614
4615 r = fdisk_write_disklabel(c);
4616 if (r < 0)
4617 return log_error_errno(r, "Failed to write resized partition table: %m");
4618
4619 log_info("Resized partition table.");
4620 return 1;
4621 }
4622
4623 static int resize_backing_fd(
4624 const char *node, /* The primary way we access the disk image to operate on */
4625 int *fd, /* An O_RDONLY fd referring to that inode */
4626 const char *backing_file, /* If the above refers to a loopback device, the backing regular file for that, which we can grow */
4627 LoopDevice *loop_device) {
4628
4629 _cleanup_close_ int writable_fd = -1;
4630 uint64_t current_size;
4631 struct stat st;
4632 int r;
4633
4634 assert(node);
4635 assert(fd);
4636
4637 if (arg_size == UINT64_MAX) /* Nothing to do */
4638 return 0;
4639
4640 if (*fd < 0) {
4641 /* Open the file if we haven't opened it yet. Note that we open it read-only here, just to
4642 * keep a reference to the file we can pass around. */
4643 *fd = open(node, O_RDONLY|O_CLOEXEC);
4644 if (*fd < 0)
4645 return log_error_errno(errno, "Failed to open '%s' in order to adjust size: %m", node);
4646 }
4647
4648 if (fstat(*fd, &st) < 0)
4649 return log_error_errno(errno, "Failed to stat '%s': %m", node);
4650
4651 if (S_ISBLK(st.st_mode)) {
4652 if (!backing_file)
4653 return log_error_errno(SYNTHETIC_ERRNO(EBADF), "Cannot resize block device '%s'.", node);
4654
4655 assert(loop_device);
4656
4657 if (ioctl(*fd, BLKGETSIZE64, &current_size) < 0)
4658 return log_error_errno(errno, "Failed to determine size of block device %s: %m", node);
4659 } else {
4660 r = stat_verify_regular(&st);
4661 if (r < 0)
4662 return log_error_errno(r, "Specified path '%s' is not a regular file or loopback block device, cannot resize: %m", node);
4663
4664 assert(!backing_file);
4665 assert(!loop_device);
4666 current_size = st.st_size;
4667 }
4668
4669 if (current_size >= arg_size) {
4670 log_info("File '%s' already is of requested size or larger, not growing. (%s >= %s)",
4671 node, FORMAT_BYTES(current_size), FORMAT_BYTES(arg_size));
4672 return 0;
4673 }
4674
4675 if (S_ISBLK(st.st_mode)) {
4676 assert(backing_file);
4677
4678 /* This is a loopback device. We can't really grow those directly, but we can grow the
4679 * backing file, hence let's do that. */
4680
4681 writable_fd = open(backing_file, O_WRONLY|O_CLOEXEC|O_NONBLOCK);
4682 if (writable_fd < 0)
4683 return log_error_errno(errno, "Failed to open backing file '%s': %m", backing_file);
4684
4685 if (fstat(writable_fd, &st) < 0)
4686 return log_error_errno(errno, "Failed to stat() backing file '%s': %m", backing_file);
4687
4688 r = stat_verify_regular(&st);
4689 if (r < 0)
4690 return log_error_errno(r, "Backing file '%s' of block device is not a regular file: %m", backing_file);
4691
4692 if ((uint64_t) st.st_size != current_size)
4693 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
4694 "Size of backing file '%s' of loopback block device '%s' don't match, refusing.",
4695 node, backing_file);
4696 } else {
4697 assert(S_ISREG(st.st_mode));
4698 assert(!backing_file);
4699
4700 /* The file descriptor is read-only. In order to grow the file we need to have a writable fd. We
4701 * reopen the file for that temporarily. We keep the writable fd only open for this operation though,
4702 * as fdisk can't accept it anyway. */
4703
4704 writable_fd = fd_reopen(*fd, O_WRONLY|O_CLOEXEC);
4705 if (writable_fd < 0)
4706 return log_error_errno(writable_fd, "Failed to reopen backing file '%s' writable: %m", node);
4707 }
4708
4709 if (!arg_discard) {
4710 if (fallocate(writable_fd, 0, 0, arg_size) < 0) {
4711 if (!ERRNO_IS_NOT_SUPPORTED(errno))
4712 return log_error_errno(errno, "Failed to grow '%s' from %s to %s by allocation: %m",
4713 node, FORMAT_BYTES(current_size), FORMAT_BYTES(arg_size));
4714
4715 /* Fallback to truncation, if fallocate() is not supported. */
4716 log_debug("Backing file system does not support fallocate(), falling back to ftruncate().");
4717 } else {
4718 if (current_size == 0) /* Likely regular file just created by us */
4719 log_info("Allocated %s for '%s'.", FORMAT_BYTES(arg_size), node);
4720 else
4721 log_info("File '%s' grown from %s to %s by allocation.",
4722 node, FORMAT_BYTES(current_size), FORMAT_BYTES(arg_size));
4723
4724 goto done;
4725 }
4726 }
4727
4728 if (ftruncate(writable_fd, arg_size) < 0)
4729 return log_error_errno(errno, "Failed to grow '%s' from %s to %s by truncation: %m",
4730 node, FORMAT_BYTES(current_size), FORMAT_BYTES(arg_size));
4731
4732 if (current_size == 0) /* Likely regular file just created by us */
4733 log_info("Sized '%s' to %s.", node, FORMAT_BYTES(arg_size));
4734 else
4735 log_info("File '%s' grown from %s to %s by truncation.",
4736 node, FORMAT_BYTES(current_size), FORMAT_BYTES(arg_size));
4737
4738 done:
4739 r = resize_pt(writable_fd);
4740 if (r < 0)
4741 return r;
4742
4743 if (loop_device) {
4744 r = loop_device_refresh_size(loop_device, UINT64_MAX, arg_size);
4745 if (r < 0)
4746 return log_error_errno(r, "Failed to update loop device size: %m");
4747 }
4748
4749 return 1;
4750 }
4751
4752 static int determine_auto_size(Context *c) {
4753 uint64_t sum = round_up_size(GPT_METADATA_SIZE, 4096);
4754 Partition *p;
4755
4756 assert_se(c);
4757
4758 LIST_FOREACH(partitions, p, c->partitions) {
4759 uint64_t m;
4760
4761 if (p->dropped)
4762 continue;
4763
4764 m = partition_min_size_with_padding(p);
4765 if (m > UINT64_MAX - sum)
4766 return log_error_errno(SYNTHETIC_ERRNO(EOVERFLOW), "Image would grow too large, refusing.");
4767
4768 sum += m;
4769 }
4770
4771 if (c->total != UINT64_MAX)
4772 /* Image already allocated? Then show its size. */
4773 log_info("Automatically determined minimal disk image size as %s, current image size is %s.",
4774 FORMAT_BYTES(sum), FORMAT_BYTES(c->total));
4775 else
4776 /* If the image is being created right now, then it has no previous size, suppress any comment about it hence. */
4777 log_info("Automatically determined minimal disk image size as %s.",
4778 FORMAT_BYTES(sum));
4779
4780 arg_size = sum;
4781 return 0;
4782 }
4783
4784 static int run(int argc, char *argv[]) {
4785 _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL;
4786 _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL;
4787 _cleanup_(umount_and_rmdir_and_freep) char *mounted_dir = NULL;
4788 _cleanup_(context_freep) Context* context = NULL;
4789 _cleanup_free_ char *node = NULL;
4790 _cleanup_close_ int backing_fd = -1;
4791 bool from_scratch, node_is_our_loop = false;
4792 int r;
4793
4794 log_show_color(true);
4795 log_parse_environment();
4796 log_open();
4797
4798 r = parse_argv(argc, argv);
4799 if (r <= 0)
4800 return r;
4801
4802 r = parse_proc_cmdline_factory_reset();
4803 if (r < 0)
4804 return r;
4805
4806 r = parse_efi_variable_factory_reset();
4807 if (r < 0)
4808 return r;
4809
4810 if (arg_image) {
4811 assert(!arg_root);
4812
4813 /* Mount this strictly read-only: we shall modify the partition table, not the file
4814 * systems */
4815 r = mount_image_privately_interactively(
4816 arg_image,
4817 DISSECT_IMAGE_MOUNT_READ_ONLY |
4818 (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) */
4819 DISSECT_IMAGE_GPT_ONLY |
4820 DISSECT_IMAGE_RELAX_VAR_CHECK |
4821 DISSECT_IMAGE_USR_NO_ROOT |
4822 DISSECT_IMAGE_REQUIRE_ROOT,
4823 &mounted_dir,
4824 &loop_device,
4825 &decrypted_image);
4826 if (r < 0)
4827 return r;
4828
4829 arg_root = strdup(mounted_dir);
4830 if (!arg_root)
4831 return log_oom();
4832
4833 if (!arg_node) {
4834 arg_node = strdup(loop_device->node);
4835 if (!arg_node)
4836 return log_oom();
4837
4838 /* Remember that the device we are about to manipulate is actually the one we
4839 * allocated here, and thus to increase its backing file we know what to do */
4840 node_is_our_loop = true;
4841 }
4842 }
4843
4844 context = context_new(arg_seed);
4845 if (!context)
4846 return log_oom();
4847
4848 r = context_read_definitions(context, arg_definitions, arg_root);
4849 if (r < 0)
4850 return r;
4851
4852 if (context->n_partitions <= 0 && arg_empty == EMPTY_REFUSE) {
4853 log_info("Didn't find any partition definition files, nothing to do.");
4854 return 0;
4855 }
4856
4857 r = find_root(&node, &backing_fd);
4858 if (r < 0)
4859 return r;
4860
4861 if (arg_size != UINT64_MAX) {
4862 r = resize_backing_fd(
4863 node,
4864 &backing_fd,
4865 node_is_our_loop ? arg_image : NULL,
4866 node_is_our_loop ? loop_device : NULL);
4867 if (r < 0)
4868 return r;
4869 }
4870
4871 r = context_load_partition_table(context, node, &backing_fd);
4872 if (r == -EHWPOISON)
4873 return 77; /* Special return value which means "Not GPT, so not doing anything". This isn't
4874 * really an error when called at boot. */
4875 if (r < 0)
4876 return r;
4877 from_scratch = r > 0; /* Starting from scratch */
4878
4879 if (arg_can_factory_reset) {
4880 r = context_can_factory_reset(context);
4881 if (r < 0)
4882 return r;
4883 if (r == 0)
4884 return EXIT_FAILURE;
4885
4886 return 0;
4887 }
4888
4889 r = context_factory_reset(context, from_scratch);
4890 if (r < 0)
4891 return r;
4892 if (r > 0) {
4893 /* We actually did a factory reset! */
4894 r = remove_efi_variable_factory_reset();
4895 if (r < 0)
4896 return r;
4897
4898 /* Reload the reduced partition table */
4899 context_unload_partition_table(context);
4900 r = context_load_partition_table(context, node, &backing_fd);
4901 if (r < 0)
4902 return r;
4903 }
4904
4905 #if 0
4906 (void) context_dump_partitions(context, node);
4907 putchar('\n');
4908 #endif
4909
4910 r = context_read_seed(context, arg_root);
4911 if (r < 0)
4912 return r;
4913
4914 /* Open all files to copy blocks from now, since we want to take their size into consideration */
4915 r = context_open_copy_block_paths(
4916 context,
4917 arg_root,
4918 loop_device ? loop_device->devno : /* if --image= is specified, only allow partitions on the loopback device */
4919 arg_root && !arg_image ? 0 : /* if --root= is specified, don't accept any block device */
4920 (dev_t) -1); /* if neither is specified, make no restrictions */
4921 if (r < 0)
4922 return r;
4923
4924 if (arg_size_auto) {
4925 r = determine_auto_size(context);
4926 if (r < 0)
4927 return r;
4928
4929 /* Flush out everything again, and let's grow the file first, then start fresh */
4930 context_unload_partition_table(context);
4931
4932 assert_se(arg_size != UINT64_MAX);
4933 r = resize_backing_fd(
4934 node,
4935 &backing_fd,
4936 node_is_our_loop ? arg_image : NULL,
4937 node_is_our_loop ? loop_device : NULL);
4938 if (r < 0)
4939 return r;
4940
4941 r = context_load_partition_table(context, node, &backing_fd);
4942 if (r < 0)
4943 return r;
4944 }
4945
4946 /* First try to fit new partitions in, dropping by priority until it fits */
4947 for (;;) {
4948 uint64_t largest_free_area;
4949
4950 if (context_allocate_partitions(context, &largest_free_area))
4951 break; /* Success! */
4952
4953 if (!context_drop_one_priority(context)) {
4954 r = log_error_errno(SYNTHETIC_ERRNO(ENOSPC),
4955 "Can't fit requested partitions into available free space (%s), refusing.",
4956 FORMAT_BYTES(largest_free_area));
4957 determine_auto_size(context);
4958 return r;
4959 }
4960 }
4961
4962 /* Now assign free space according to the weight logic */
4963 r = context_grow_partitions(context);
4964 if (r < 0)
4965 return r;
4966
4967 /* Now calculate where each partition gets placed */
4968 context_place_partitions(context);
4969
4970 /* Make sure each partition has a unique UUID and unique label */
4971 r = context_acquire_partition_uuids_and_labels(context);
4972 if (r < 0)
4973 return r;
4974
4975 r = context_write_partition_table(context, node, from_scratch);
4976 if (r < 0)
4977 return r;
4978
4979 return 0;
4980 }
4981
4982 DEFINE_MAIN_FUNCTION_WITH_POSITIVE_FAILURE(run);
Unnamed repository; edit this file 'description' to name the repository.