]> git.ipfire.org Git - thirdparty/xfsprogs-dev.git/blob - mkfs/xfs_mkfs.c
77eb5f0fcfa9388d14a6def21dbc552aeff9cc0a
[thirdparty/xfsprogs-dev.git] / mkfs / xfs_mkfs.c
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
4 * All Rights Reserved.
5 */
6 #include "libfrog.h"
7 #include "libxfs.h"
8 #include <ctype.h>
9 #include "xfs_multidisk.h"
10 #include "libxcmd.h"
11 #include "fsgeom.h"
12
13
14 #define TERABYTES(count, blog) ((uint64_t)(count) << (40 - (blog)))
15 #define GIGABYTES(count, blog) ((uint64_t)(count) << (30 - (blog)))
16 #define MEGABYTES(count, blog) ((uint64_t)(count) << (20 - (blog)))
17
18 /*
19 * Use this macro before we have superblock and mount structure to
20 * convert from basic blocks to filesystem blocks.
21 */
22 #define DTOBT(d, bl) ((xfs_rfsblock_t)((d) >> ((bl) - BBSHIFT)))
23
24 /*
25 * amount (in bytes) we zero at the beginning and end of the device to
26 * remove traces of other filesystems, raid superblocks, etc.
27 */
28 #define WHACK_SIZE (128 * 1024)
29
30 /*
31 * XXX: The configured block and sector sizes are defined as global variables so
32 * that they don't need to be passed to getnum/cvtnum().
33 */
34 unsigned int blocksize;
35 unsigned int sectorsize;
36
37 /*
38 * Enums for each CLI parameter type are declared first so we can calculate the
39 * maximum array size needed to hold them automatically.
40 */
41 enum {
42 B_SIZE = 0,
43 B_MAX_OPTS,
44 };
45
46 enum {
47 D_AGCOUNT = 0,
48 D_FILE,
49 D_NAME,
50 D_SIZE,
51 D_SUNIT,
52 D_SWIDTH,
53 D_AGSIZE,
54 D_SU,
55 D_SW,
56 D_SECTSIZE,
57 D_NOALIGN,
58 D_RTINHERIT,
59 D_PROJINHERIT,
60 D_EXTSZINHERIT,
61 D_COWEXTSIZE,
62 D_MAX_OPTS,
63 };
64
65 enum {
66 I_ALIGN = 0,
67 I_MAXPCT,
68 I_PERBLOCK,
69 I_SIZE,
70 I_ATTR,
71 I_PROJID32BIT,
72 I_SPINODES,
73 I_MAX_OPTS,
74 };
75
76 enum {
77 L_AGNUM = 0,
78 L_INTERNAL,
79 L_SIZE,
80 L_VERSION,
81 L_SUNIT,
82 L_SU,
83 L_DEV,
84 L_SECTSIZE,
85 L_FILE,
86 L_NAME,
87 L_LAZYSBCNTR,
88 L_MAX_OPTS,
89 };
90
91 enum {
92 N_SIZE = 0,
93 N_VERSION,
94 N_FTYPE,
95 N_MAX_OPTS,
96 };
97
98 enum {
99 R_EXTSIZE = 0,
100 R_SIZE,
101 R_DEV,
102 R_FILE,
103 R_NAME,
104 R_NOALIGN,
105 R_MAX_OPTS,
106 };
107
108 enum {
109 S_SIZE = 0,
110 S_SECTSIZE,
111 S_MAX_OPTS,
112 };
113
114 enum {
115 M_CRC = 0,
116 M_FINOBT,
117 M_UUID,
118 M_RMAPBT,
119 M_REFLINK,
120 M_MAX_OPTS,
121 };
122
123 /* Just define the max options array size manually right now */
124 #define MAX_SUBOPTS D_MAX_OPTS
125
126 #define SUBOPT_NEEDS_VAL (-1LL)
127 #define MAX_CONFLICTS 8
128 #define LAST_CONFLICT (-1)
129
130 /*
131 * Table for parsing mkfs parameters.
132 *
133 * Description of the structure members follows:
134 *
135 * name MANDATORY
136 * Name is a single char, e.g., for '-d file', name is 'd'.
137 *
138 * subopts MANDATORY
139 * Subopts is a list of strings naming suboptions. In the example above,
140 * it would contain "file". The last entry of this list has to be NULL.
141 *
142 * subopt_params MANDATORY
143 * This is a list of structs tied with subopts. For each entry in subopts,
144 * a corresponding entry has to be defined:
145 *
146 * subopt_params struct:
147 * index MANDATORY
148 * This number, starting from zero, denotes which item in subopt_params
149 * it is. The index has to be the same as is the order in subopts list,
150 * so we can access the right item both in subopt_param and subopts.
151 *
152 * seen INTERNAL
153 * Do not set this flag when definning a subopt. It is used to remeber that
154 * this subopt was already seen, for example for conflicts detection.
155 *
156 * str_seen INTERNAL
157 * Do not set. It is used internally for respecification, when some options
158 * has to be parsed twice - at first as a string, then later as a number.
159 *
160 * convert OPTIONAL
161 * A flag signalling whether the user-given value can use suffixes.
162 * If you want to allow the use of user-friendly values like 13k, 42G,
163 * set it to true.
164 *
165 * is_power_2 OPTIONAL
166 * An optional flag for subopts where the given value has to be a power
167 * of two.
168 *
169 * conflicts MANDATORY
170 * If your subopt is in a conflict with some other option, specify it.
171 * Accepts the .index values of the conflicting subopts and the last
172 * member of this list has to be LAST_CONFLICT.
173 *
174 * minval, maxval OPTIONAL
175 * These options are used for automatic range check and they have to be
176 * always used together in pair. If you don't want to limit the max value,
177 * use something like UINT_MAX. If no value is given, then you must either
178 * supply your own validation, or refuse any value in the 'case
179 * X_SOMETHING' block. If you forget to define the min and max value, but
180 * call a standard function for validating user's value, it will cause an
181 * error message notifying you about this issue.
182 *
183 * (Said in another way, you can't have minval and maxval both equal
184 * to zero. But if one value is different: minval=0 and maxval=1,
185 * then it is OK.)
186 *
187 * defaultval MANDATORY
188 * The value used if user specifies the subopt, but no value.
189 * If the subopt accepts some values (-d file=[1|0]), then this
190 * sets what is used with simple specifying the subopt (-d file).
191 * A special SUBOPT_NEEDS_VAL can be used to require a user-given
192 * value in any case.
193 */
194 struct opt_params {
195 const char name;
196 const char *subopts[MAX_SUBOPTS];
197
198 struct subopt_param {
199 int index;
200 bool seen;
201 bool str_seen;
202 bool convert;
203 bool is_power_2;
204 struct _conflict {
205 struct opt_params *opts;
206 int subopt;
207 } conflicts[MAX_CONFLICTS];
208 long long minval;
209 long long maxval;
210 long long defaultval;
211 } subopt_params[MAX_SUBOPTS];
212 };
213
214 /*
215 * The two dimensional conflict array requires some initialisations to know
216 * about tables that haven't yet been defined. Work around this ordering
217 * issue with extern definitions here.
218 */
219 extern struct opt_params sopts;
220
221 struct opt_params bopts = {
222 .name = 'b',
223 .subopts = {
224 [B_SIZE] = "size",
225 },
226 .subopt_params = {
227 { .index = B_SIZE,
228 .convert = true,
229 .is_power_2 = true,
230 .conflicts = { { NULL, LAST_CONFLICT } },
231 .minval = XFS_MIN_BLOCKSIZE,
232 .maxval = XFS_MAX_BLOCKSIZE,
233 .defaultval = SUBOPT_NEEDS_VAL,
234 },
235 },
236 };
237
238 struct opt_params dopts = {
239 .name = 'd',
240 .subopts = {
241 [D_AGCOUNT] = "agcount",
242 [D_FILE] = "file",
243 [D_NAME] = "name",
244 [D_SIZE] = "size",
245 [D_SUNIT] = "sunit",
246 [D_SWIDTH] = "swidth",
247 [D_AGSIZE] = "agsize",
248 [D_SU] = "su",
249 [D_SW] = "sw",
250 [D_SECTSIZE] = "sectsize",
251 [D_NOALIGN] = "noalign",
252 [D_RTINHERIT] = "rtinherit",
253 [D_PROJINHERIT] = "projinherit",
254 [D_EXTSZINHERIT] = "extszinherit",
255 [D_COWEXTSIZE] = "cowextsize",
256 },
257 .subopt_params = {
258 { .index = D_AGCOUNT,
259 .conflicts = { { &dopts, D_AGSIZE },
260 { NULL, LAST_CONFLICT } },
261 .minval = 1,
262 .maxval = XFS_MAX_AGNUMBER,
263 .defaultval = SUBOPT_NEEDS_VAL,
264 },
265 { .index = D_FILE,
266 .conflicts = { { NULL, LAST_CONFLICT } },
267 .minval = 0,
268 .maxval = 1,
269 .defaultval = 1,
270 },
271 { .index = D_NAME,
272 .conflicts = { { NULL, LAST_CONFLICT } },
273 .defaultval = SUBOPT_NEEDS_VAL,
274 },
275 { .index = D_SIZE,
276 .conflicts = { { NULL, LAST_CONFLICT } },
277 .convert = true,
278 .minval = XFS_AG_MIN_BYTES,
279 .maxval = LLONG_MAX,
280 .defaultval = SUBOPT_NEEDS_VAL,
281 },
282 { .index = D_SUNIT,
283 .conflicts = { { &dopts, D_NOALIGN },
284 { &dopts, D_SU },
285 { &dopts, D_SW },
286 { NULL, LAST_CONFLICT } },
287 .minval = 0,
288 .maxval = UINT_MAX,
289 .defaultval = SUBOPT_NEEDS_VAL,
290 },
291 { .index = D_SWIDTH,
292 .conflicts = { { &dopts, D_NOALIGN },
293 { &dopts, D_SU },
294 { &dopts, D_SW },
295 { NULL, LAST_CONFLICT } },
296 .minval = 0,
297 .maxval = UINT_MAX,
298 .defaultval = SUBOPT_NEEDS_VAL,
299 },
300 { .index = D_AGSIZE,
301 .conflicts = { { &dopts, D_AGCOUNT },
302 { NULL, LAST_CONFLICT } },
303 .convert = true,
304 .minval = XFS_AG_MIN_BYTES,
305 .maxval = XFS_AG_MAX_BYTES,
306 .defaultval = SUBOPT_NEEDS_VAL,
307 },
308 { .index = D_SU,
309 .conflicts = { { &dopts, D_NOALIGN },
310 { &dopts, D_SUNIT },
311 { &dopts, D_SWIDTH },
312 { NULL, LAST_CONFLICT } },
313 .convert = true,
314 .minval = 0,
315 .maxval = UINT_MAX,
316 .defaultval = SUBOPT_NEEDS_VAL,
317 },
318 { .index = D_SW,
319 .conflicts = { { &dopts, D_NOALIGN },
320 { &dopts, D_SUNIT },
321 { &dopts, D_SWIDTH },
322 { NULL, LAST_CONFLICT } },
323 .minval = 0,
324 .maxval = UINT_MAX,
325 .defaultval = SUBOPT_NEEDS_VAL,
326 },
327 { .index = D_SECTSIZE,
328 .conflicts = { { &sopts, S_SIZE },
329 { &sopts, S_SECTSIZE },
330 { NULL, LAST_CONFLICT } },
331 .convert = true,
332 .is_power_2 = true,
333 .minval = XFS_MIN_SECTORSIZE,
334 .maxval = XFS_MAX_SECTORSIZE,
335 .defaultval = SUBOPT_NEEDS_VAL,
336 },
337 { .index = D_NOALIGN,
338 .conflicts = { { &dopts, D_SU },
339 { &dopts, D_SW },
340 { &dopts, D_SUNIT },
341 { &dopts, D_SWIDTH },
342 { NULL, LAST_CONFLICT } },
343 .minval = 0,
344 .maxval = 1,
345 .defaultval = 1,
346 },
347 { .index = D_RTINHERIT,
348 .conflicts = { { NULL, LAST_CONFLICT } },
349 .minval = 1,
350 .maxval = 1,
351 .defaultval = 1,
352 },
353 { .index = D_PROJINHERIT,
354 .conflicts = { { NULL, LAST_CONFLICT } },
355 .minval = 0,
356 .maxval = UINT_MAX,
357 .defaultval = SUBOPT_NEEDS_VAL,
358 },
359 { .index = D_EXTSZINHERIT,
360 .conflicts = { { NULL, LAST_CONFLICT } },
361 .minval = 0,
362 .maxval = UINT_MAX,
363 .defaultval = SUBOPT_NEEDS_VAL,
364 },
365 { .index = D_COWEXTSIZE,
366 .conflicts = { { NULL, LAST_CONFLICT } },
367 .minval = 0,
368 .maxval = UINT_MAX,
369 .defaultval = SUBOPT_NEEDS_VAL,
370 },
371 },
372 };
373
374
375 struct opt_params iopts = {
376 .name = 'i',
377 .subopts = {
378 [I_ALIGN] = "align",
379 [I_MAXPCT] = "maxpct",
380 [I_PERBLOCK] = "perblock",
381 [I_SIZE] = "size",
382 [I_ATTR] = "attr",
383 [I_PROJID32BIT] = "projid32bit",
384 [I_SPINODES] = "sparse",
385 },
386 .subopt_params = {
387 { .index = I_ALIGN,
388 .conflicts = { { NULL, LAST_CONFLICT } },
389 .minval = 0,
390 .maxval = 1,
391 .defaultval = 1,
392 },
393 { .index = I_MAXPCT,
394 .conflicts = { { NULL, LAST_CONFLICT } },
395 .minval = 0,
396 .maxval = 100,
397 .defaultval = SUBOPT_NEEDS_VAL,
398 },
399 { .index = I_PERBLOCK,
400 .conflicts = { { &iopts, I_SIZE },
401 { NULL, LAST_CONFLICT } },
402 .is_power_2 = true,
403 .minval = XFS_MIN_INODE_PERBLOCK,
404 .maxval = XFS_MAX_BLOCKSIZE / XFS_DINODE_MIN_SIZE,
405 .defaultval = SUBOPT_NEEDS_VAL,
406 },
407 { .index = I_SIZE,
408 .conflicts = { { &iopts, I_PERBLOCK },
409 { NULL, LAST_CONFLICT } },
410 .is_power_2 = true,
411 .minval = XFS_DINODE_MIN_SIZE,
412 .maxval = XFS_DINODE_MAX_SIZE,
413 .defaultval = SUBOPT_NEEDS_VAL,
414 },
415 { .index = I_ATTR,
416 .conflicts = { { NULL, LAST_CONFLICT } },
417 .minval = 0,
418 .maxval = 2,
419 .defaultval = SUBOPT_NEEDS_VAL,
420 },
421 { .index = I_PROJID32BIT,
422 .conflicts = { { NULL, LAST_CONFLICT } },
423 .minval = 0,
424 .maxval = 1,
425 .defaultval = 1,
426 },
427 { .index = I_SPINODES,
428 .conflicts = { { NULL, LAST_CONFLICT } },
429 .minval = 0,
430 .maxval = 1,
431 .defaultval = 1,
432 },
433 },
434 };
435
436 struct opt_params lopts = {
437 .name = 'l',
438 .subopts = {
439 [L_AGNUM] = "agnum",
440 [L_INTERNAL] = "internal",
441 [L_SIZE] = "size",
442 [L_VERSION] = "version",
443 [L_SUNIT] = "sunit",
444 [L_SU] = "su",
445 [L_DEV] = "logdev",
446 [L_SECTSIZE] = "sectsize",
447 [L_FILE] = "file",
448 [L_NAME] = "name",
449 [L_LAZYSBCNTR] = "lazy-count",
450 },
451 .subopt_params = {
452 { .index = L_AGNUM,
453 .conflicts = { { &lopts, L_DEV },
454 { NULL, LAST_CONFLICT } },
455 .minval = 0,
456 .maxval = UINT_MAX,
457 .defaultval = SUBOPT_NEEDS_VAL,
458 },
459 { .index = L_INTERNAL,
460 .conflicts = { { &lopts, L_FILE },
461 { &lopts, L_DEV },
462 { &lopts, L_SECTSIZE },
463 { NULL, LAST_CONFLICT } },
464 .minval = 0,
465 .maxval = 1,
466 .defaultval = 1,
467 },
468 { .index = L_SIZE,
469 .conflicts = { { NULL, LAST_CONFLICT } },
470 .convert = true,
471 .minval = 2 * 1024 * 1024LL, /* XXX: XFS_MIN_LOG_BYTES */
472 .maxval = XFS_MAX_LOG_BYTES,
473 .defaultval = SUBOPT_NEEDS_VAL,
474 },
475 { .index = L_VERSION,
476 .conflicts = { { NULL, LAST_CONFLICT } },
477 .minval = 1,
478 .maxval = 2,
479 .defaultval = SUBOPT_NEEDS_VAL,
480 },
481 { .index = L_SUNIT,
482 .conflicts = { { &lopts, L_SU },
483 { NULL, LAST_CONFLICT } },
484 .minval = 1,
485 .maxval = BTOBB(XLOG_MAX_RECORD_BSIZE),
486 .defaultval = SUBOPT_NEEDS_VAL,
487 },
488 { .index = L_SU,
489 .conflicts = { { &lopts, L_SUNIT },
490 { NULL, LAST_CONFLICT } },
491 .convert = true,
492 .minval = BBTOB(1),
493 .maxval = XLOG_MAX_RECORD_BSIZE,
494 .defaultval = SUBOPT_NEEDS_VAL,
495 },
496 { .index = L_DEV,
497 .conflicts = { { &lopts, L_AGNUM },
498 { &lopts, L_NAME },
499 { &lopts, L_INTERNAL },
500 { NULL, LAST_CONFLICT } },
501 .defaultval = SUBOPT_NEEDS_VAL,
502 },
503 { .index = L_SECTSIZE,
504 .conflicts = { { &lopts, L_INTERNAL },
505 { NULL, LAST_CONFLICT } },
506 .convert = true,
507 .is_power_2 = true,
508 .minval = XFS_MIN_SECTORSIZE,
509 .maxval = XFS_MAX_SECTORSIZE,
510 .defaultval = SUBOPT_NEEDS_VAL,
511 },
512 { .index = L_FILE,
513 .conflicts = { { &lopts, L_INTERNAL },
514 { NULL, LAST_CONFLICT } },
515 .minval = 0,
516 .maxval = 1,
517 .defaultval = 1,
518 },
519 { .index = L_NAME,
520 .conflicts = { { &lopts, L_AGNUM },
521 { &lopts, L_DEV },
522 { &lopts, L_INTERNAL },
523 { NULL, LAST_CONFLICT } },
524 .defaultval = SUBOPT_NEEDS_VAL,
525 },
526 { .index = L_LAZYSBCNTR,
527 .conflicts = { { NULL, LAST_CONFLICT } },
528 .minval = 0,
529 .maxval = 1,
530 .defaultval = 1,
531 },
532 },
533 };
534
535 struct opt_params nopts = {
536 .name = 'n',
537 .subopts = {
538 [N_SIZE] = "size",
539 [N_VERSION] = "version",
540 [N_FTYPE] = "ftype",
541 },
542 .subopt_params = {
543 { .index = N_SIZE,
544 .conflicts = { { NULL, LAST_CONFLICT } },
545 .convert = true,
546 .is_power_2 = true,
547 .minval = 1 << XFS_MIN_REC_DIRSIZE,
548 .maxval = XFS_MAX_BLOCKSIZE,
549 .defaultval = SUBOPT_NEEDS_VAL,
550 },
551 { .index = N_VERSION,
552 .conflicts = { { NULL, LAST_CONFLICT } },
553 .minval = 2,
554 .maxval = 2,
555 .defaultval = SUBOPT_NEEDS_VAL,
556 },
557 { .index = N_FTYPE,
558 .conflicts = { { NULL, LAST_CONFLICT } },
559 .minval = 0,
560 .maxval = 1,
561 .defaultval = 1,
562 },
563 },
564 };
565
566 struct opt_params ropts = {
567 .name = 'r',
568 .subopts = {
569 [R_EXTSIZE] = "extsize",
570 [R_SIZE] = "size",
571 [R_DEV] = "rtdev",
572 [R_FILE] = "file",
573 [R_NAME] = "name",
574 [R_NOALIGN] = "noalign",
575 },
576 .subopt_params = {
577 { .index = R_EXTSIZE,
578 .conflicts = { { NULL, LAST_CONFLICT } },
579 .convert = true,
580 .minval = XFS_MIN_RTEXTSIZE,
581 .maxval = XFS_MAX_RTEXTSIZE,
582 .defaultval = SUBOPT_NEEDS_VAL,
583 },
584 { .index = R_SIZE,
585 .conflicts = { { NULL, LAST_CONFLICT } },
586 .convert = true,
587 .minval = 0,
588 .maxval = LLONG_MAX,
589 .defaultval = SUBOPT_NEEDS_VAL,
590 },
591 { .index = R_DEV,
592 .conflicts = { { &ropts, R_NAME },
593 { NULL, LAST_CONFLICT } },
594 .defaultval = SUBOPT_NEEDS_VAL,
595 },
596 { .index = R_FILE,
597 .minval = 0,
598 .maxval = 1,
599 .defaultval = 1,
600 .conflicts = { { NULL, LAST_CONFLICT } },
601 },
602 { .index = R_NAME,
603 .conflicts = { { &ropts, R_DEV },
604 { NULL, LAST_CONFLICT } },
605 .defaultval = SUBOPT_NEEDS_VAL,
606 },
607 { .index = R_NOALIGN,
608 .minval = 0,
609 .maxval = 1,
610 .defaultval = 1,
611 .conflicts = { { NULL, LAST_CONFLICT } },
612 },
613 },
614 };
615
616 struct opt_params sopts = {
617 .name = 's',
618 .subopts = {
619 [S_SIZE] = "size",
620 [S_SECTSIZE] = "sectsize",
621 },
622 .subopt_params = {
623 { .index = S_SIZE,
624 .conflicts = { { &sopts, S_SECTSIZE },
625 { &dopts, D_SECTSIZE },
626 { NULL, LAST_CONFLICT } },
627 .convert = true,
628 .is_power_2 = true,
629 .minval = XFS_MIN_SECTORSIZE,
630 .maxval = XFS_MAX_SECTORSIZE,
631 .defaultval = SUBOPT_NEEDS_VAL,
632 },
633 { .index = S_SECTSIZE,
634 .conflicts = { { &sopts, S_SIZE },
635 { &dopts, D_SECTSIZE },
636 { NULL, LAST_CONFLICT } },
637 .convert = true,
638 .is_power_2 = true,
639 .minval = XFS_MIN_SECTORSIZE,
640 .maxval = XFS_MAX_SECTORSIZE,
641 .defaultval = SUBOPT_NEEDS_VAL,
642 },
643 },
644 };
645
646 struct opt_params mopts = {
647 .name = 'm',
648 .subopts = {
649 [M_CRC] = "crc",
650 [M_FINOBT] = "finobt",
651 [M_UUID] = "uuid",
652 [M_RMAPBT] = "rmapbt",
653 [M_REFLINK] = "reflink",
654 },
655 .subopt_params = {
656 { .index = M_CRC,
657 .conflicts = { { NULL, LAST_CONFLICT } },
658 .minval = 0,
659 .maxval = 1,
660 .defaultval = 1,
661 },
662 { .index = M_FINOBT,
663 .conflicts = { { NULL, LAST_CONFLICT } },
664 .minval = 0,
665 .maxval = 1,
666 .defaultval = 1,
667 },
668 { .index = M_UUID,
669 .conflicts = { { NULL, LAST_CONFLICT } },
670 .defaultval = SUBOPT_NEEDS_VAL,
671 },
672 { .index = M_RMAPBT,
673 .conflicts = { { NULL, LAST_CONFLICT } },
674 .minval = 0,
675 .maxval = 1,
676 .defaultval = 1,
677 },
678 { .index = M_REFLINK,
679 .conflicts = { { NULL, LAST_CONFLICT } },
680 .minval = 0,
681 .maxval = 1,
682 .defaultval = 1,
683 },
684 },
685 };
686
687 /* quick way of checking if a parameter was set on the CLI */
688 static bool
689 cli_opt_set(
690 struct opt_params *opts,
691 int subopt)
692 {
693 return opts->subopt_params[subopt].seen ||
694 opts->subopt_params[subopt].str_seen;
695 }
696
697 /*
698 * Options configured on the command line.
699 *
700 * This stores all the specific config parameters the user sets on the command
701 * line. We do not use these values directly - they are inputs to the mkfs
702 * geometry validation and override any default configuration value we have.
703 *
704 * We don't keep flags to indicate what parameters are set - if we need to check
705 * if an option was set on the command line, we check the relevant entry in the
706 * option table which records whether it was specified in the .seen and
707 * .str_seen variables in the table.
708 *
709 * Some parameters are stored as strings for post-parsing after their dependent
710 * options have been resolved (e.g. block size and sector size have been parsed
711 * and validated).
712 *
713 * This allows us to check that values have been set without needing separate
714 * flags for each value, and hence avoids needing to record and check for each
715 * specific option that can set the value later on in the code. In the cases
716 * where we don't have a cli_params structure around, the above cli_opt_set()
717 * function can be used.
718 */
719 struct sb_feat_args {
720 int log_version;
721 int attr_version;
722 int dir_version;
723 bool inode_align; /* XFS_SB_VERSION_ALIGNBIT */
724 bool nci; /* XFS_SB_VERSION_BORGBIT */
725 bool lazy_sb_counters; /* XFS_SB_VERSION2_LAZYSBCOUNTBIT */
726 bool parent_pointers; /* XFS_SB_VERSION2_PARENTBIT */
727 bool projid32bit; /* XFS_SB_VERSION2_PROJID32BIT */
728 bool crcs_enabled; /* XFS_SB_VERSION2_CRCBIT */
729 bool dirftype; /* XFS_SB_VERSION2_FTYPE */
730 bool finobt; /* XFS_SB_FEAT_RO_COMPAT_FINOBT */
731 bool spinodes; /* XFS_SB_FEAT_INCOMPAT_SPINODES */
732 bool rmapbt; /* XFS_SB_FEAT_RO_COMPAT_RMAPBT */
733 bool reflink; /* XFS_SB_FEAT_RO_COMPAT_REFLINK */
734 bool nodalign;
735 bool nortalign;
736 };
737
738 struct cli_params {
739 int sectorsize;
740 int blocksize;
741
742 /* parameters that depend on sector/block size being validated. */
743 char *dsize;
744 char *agsize;
745 char *dsu;
746 char *dirblocksize;
747 char *logsize;
748 char *lsu;
749 char *rtextsize;
750 char *rtsize;
751
752 /* parameters where 0 is a valid CLI value */
753 int dsunit;
754 int dswidth;
755 int dsw;
756 int64_t logagno;
757 int loginternal;
758 int lsunit;
759
760 /* parameters where 0 is not a valid value */
761 int64_t agcount;
762 int inodesize;
763 int inopblock;
764 int imaxpct;
765 int lsectorsize;
766 uuid_t uuid;
767
768 /* feature flags that are set */
769 struct sb_feat_args sb_feat;
770
771 /* root inode characteristics */
772 struct fsxattr fsx;
773
774 /* libxfs device setup */
775 struct libxfs_xinit *xi;
776 };
777
778 /*
779 * Calculated filesystem feature and geometry information.
780 *
781 * This structure contains the information we will use to create the on-disk
782 * filesystem from. The validation and calculation code uses it to store all the
783 * temporary and final config state for the filesystem.
784 *
785 * The information in this structure will contain a mix of validated CLI input
786 * variables, default feature state and calculated values that are needed to
787 * construct the superblock and other on disk features. These are all in one
788 * place so that we don't have to pass handfuls of seemingly arbitrary variables
789 * around to different functions to do the work we need to do.
790 */
791 struct mkfs_params {
792 int blocksize;
793 int blocklog;
794 int sectorsize;
795 int sectorlog;
796 int lsectorsize;
797 int lsectorlog;
798 int dirblocksize;
799 int dirblocklog;
800 int inodesize;
801 int inodelog;
802 int inopblock;
803
804 uint64_t dblocks;
805 uint64_t logblocks;
806 uint64_t rtblocks;
807 uint64_t rtextblocks;
808 uint64_t rtextents;
809 uint64_t rtbmblocks; /* rt bitmap blocks */
810
811 int dsunit; /* in FSBs */
812 int dswidth; /* in FSBs */
813 int lsunit; /* in FSBs */
814
815 uint64_t agsize;
816 uint64_t agcount;
817
818 int imaxpct;
819
820 bool loginternal;
821 uint64_t logstart;
822 uint64_t logagno;
823
824 uuid_t uuid;
825 char *label;
826
827 struct sb_feat_args sb_feat;
828 };
829
830 /*
831 * Default filesystem features and configuration values
832 *
833 * This structure contains the default mkfs values that are to be used when
834 * a user does not specify the option on the command line. We do not use these
835 * values directly - they are inputs to the mkfs geometry validation and
836 * calculations.
837 */
838 struct mkfs_default_params {
839 char *source; /* where the defaults came from */
840
841 int sectorsize;
842 int blocksize;
843
844 /* feature flags that are set */
845 struct sb_feat_args sb_feat;
846
847 /* root inode characteristics */
848 struct fsxattr fsx;
849 };
850
851 static void __attribute__((noreturn))
852 usage( void )
853 {
854 fprintf(stderr, _("Usage: %s\n\
855 /* blocksize */ [-b size=num]\n\
856 /* metadata */ [-m crc=0|1,finobt=0|1,uuid=xxx,rmapbt=0|1,reflink=0|1]\n\
857 /* data subvol */ [-d agcount=n,agsize=n,file,name=xxx,size=num,\n\
858 (sunit=value,swidth=value|su=num,sw=num|noalign),\n\
859 sectsize=num\n\
860 /* force overwrite */ [-f]\n\
861 /* inode size */ [-i log=n|perblock=n|size=num,maxpct=n,attr=0|1|2,\n\
862 projid32bit=0|1,sparse=0|1]\n\
863 /* no discard */ [-K]\n\
864 /* log subvol */ [-l agnum=n,internal,size=num,logdev=xxx,version=n\n\
865 sunit=value|su=num,sectsize=num,lazy-count=0|1]\n\
866 /* label */ [-L label (maximum 12 characters)]\n\
867 /* naming */ [-n size=num,version=2|ci,ftype=0|1]\n\
868 /* no-op info only */ [-N]\n\
869 /* prototype file */ [-p fname]\n\
870 /* quiet */ [-q]\n\
871 /* realtime subvol */ [-r extsize=num,size=num,rtdev=xxx]\n\
872 /* sectorsize */ [-s size=num]\n\
873 /* version */ [-V]\n\
874 devicename\n\
875 <devicename> is required unless -d name=xxx is given.\n\
876 <num> is xxx (bytes), xxxs (sectors), xxxb (fs blocks), xxxk (xxx KiB),\n\
877 xxxm (xxx MiB), xxxg (xxx GiB), xxxt (xxx TiB) or xxxp (xxx PiB).\n\
878 <value> is xxx (512 byte blocks).\n"),
879 progname);
880 exit(1);
881 }
882
883 static void
884 conflict(
885 struct opt_params *opts,
886 int option,
887 struct opt_params *con_opts,
888 int conflict)
889 {
890 fprintf(stderr, _("Cannot specify both -%c %s and -%c %s\n"),
891 con_opts->name, con_opts->subopts[conflict],
892 opts->name, opts->subopts[option]);
893 usage();
894 }
895
896
897 static void
898 illegal(
899 const char *value,
900 const char *opt)
901 {
902 fprintf(stderr, _("Invalid value %s for -%s option\n"), value, opt);
903 usage();
904 }
905
906 static int
907 ispow2(
908 unsigned int i)
909 {
910 return (i & (i - 1)) == 0;
911 }
912
913 static void __attribute__((noreturn))
914 reqval(
915 char opt,
916 const char *tab[],
917 int idx)
918 {
919 fprintf(stderr, _("-%c %s option requires a value\n"), opt, tab[idx]);
920 usage();
921 }
922
923 static void
924 respec(
925 char opt,
926 const char *tab[],
927 int idx)
928 {
929 fprintf(stderr, "-%c ", opt);
930 if (tab)
931 fprintf(stderr, "%s ", tab[idx]);
932 fprintf(stderr, _("option respecified\n"));
933 usage();
934 }
935
936 static void
937 unknown(
938 char opt,
939 char *s)
940 {
941 fprintf(stderr, _("unknown option -%c %s\n"), opt, s);
942 usage();
943 }
944
945 long long
946 cvtnum(
947 unsigned int blksize,
948 unsigned int sectsize,
949 const char *s)
950 {
951 long long i;
952 char *sp;
953 int c;
954
955 i = strtoll(s, &sp, 0);
956 if (i == 0 && sp == s)
957 return -1LL;
958 if (*sp == '\0')
959 return i;
960
961 if (sp[1] != '\0')
962 return -1LL;
963
964 if (*sp == 'b') {
965 if (!blksize) {
966 fprintf(stderr,
967 _("Blocksize must be provided prior to using 'b' suffix.\n"));
968 usage();
969 } else {
970 return i * blksize;
971 }
972 }
973 if (*sp == 's') {
974 if (!sectsize) {
975 fprintf(stderr,
976 _("Sectorsize must be specified prior to using 's' suffix.\n"));
977 usage();
978 } else {
979 return i * sectsize;
980 }
981 }
982
983 c = tolower(*sp);
984 switch (c) {
985 case 'e':
986 i *= 1024LL;
987 /* fall through */
988 case 'p':
989 i *= 1024LL;
990 /* fall through */
991 case 't':
992 i *= 1024LL;
993 /* fall through */
994 case 'g':
995 i *= 1024LL;
996 /* fall through */
997 case 'm':
998 i *= 1024LL;
999 /* fall through */
1000 case 'k':
1001 return i * 1024LL;
1002 default:
1003 break;
1004 }
1005 return -1LL;
1006 }
1007
1008 static void
1009 check_device_type(
1010 const char *name,
1011 int *isfile,
1012 bool no_size,
1013 bool no_name,
1014 int *create,
1015 const char *optname)
1016 {
1017 struct stat statbuf;
1018
1019 if (*isfile && (no_size || no_name)) {
1020 fprintf(stderr,
1021 _("if -%s file then -%s name and -%s size are required\n"),
1022 optname, optname, optname);
1023 usage();
1024 }
1025
1026 if (!name) {
1027 fprintf(stderr, _("No device name specified\n"));
1028 usage();
1029 }
1030
1031 if (stat(name, &statbuf)) {
1032 if (errno == ENOENT && *isfile) {
1033 if (create)
1034 *create = 1;
1035 return;
1036 }
1037
1038 fprintf(stderr,
1039 _("Error accessing specified device %s: %s\n"),
1040 name, strerror(errno));
1041 usage();
1042 return;
1043 }
1044
1045 /*
1046 * We only want to completely truncate and recreate an existing file if
1047 * we were specifically told it was a file. Set the create flag only in
1048 * this case to trigger that behaviour.
1049 */
1050 if (S_ISREG(statbuf.st_mode)) {
1051 if (!*isfile)
1052 *isfile = 1;
1053 else if (create)
1054 *create = 1;
1055 return;
1056 }
1057
1058 if (S_ISBLK(statbuf.st_mode)) {
1059 if (*isfile) {
1060 fprintf(stderr,
1061 _("specified \"-%s file\" on a block device %s\n"),
1062 optname, name);
1063 usage();
1064 }
1065 return;
1066 }
1067
1068 fprintf(stderr,
1069 _("specified device %s not a file or block device\n"),
1070 name);
1071 usage();
1072 }
1073
1074 static void
1075 validate_overwrite(
1076 const char *name,
1077 bool force_overwrite)
1078 {
1079 if (!force_overwrite && check_overwrite(name)) {
1080 fprintf(stderr,
1081 _("%s: Use the -f option to force overwrite.\n"),
1082 progname);
1083 exit(1);
1084 }
1085
1086 }
1087
1088 static void
1089 validate_ag_geometry(
1090 int blocklog,
1091 uint64_t dblocks,
1092 uint64_t agsize,
1093 uint64_t agcount)
1094 {
1095 if (agsize < XFS_AG_MIN_BLOCKS(blocklog)) {
1096 fprintf(stderr,
1097 _("agsize (%lld blocks) too small, need at least %lld blocks\n"),
1098 (long long)agsize,
1099 (long long)XFS_AG_MIN_BLOCKS(blocklog));
1100 usage();
1101 }
1102
1103 if (agsize > XFS_AG_MAX_BLOCKS(blocklog)) {
1104 fprintf(stderr,
1105 _("agsize (%lld blocks) too big, maximum is %lld blocks\n"),
1106 (long long)agsize,
1107 (long long)XFS_AG_MAX_BLOCKS(blocklog));
1108 usage();
1109 }
1110
1111 if (agsize > dblocks) {
1112 fprintf(stderr,
1113 _("agsize (%lld blocks) too big, data area is %lld blocks\n"),
1114 (long long)agsize, (long long)dblocks);
1115 usage();
1116 }
1117
1118 if (agsize < XFS_AG_MIN_BLOCKS(blocklog)) {
1119 fprintf(stderr,
1120 _("too many allocation groups for size = %lld\n"),
1121 (long long)agsize);
1122 fprintf(stderr, _("need at most %lld allocation groups\n"),
1123 (long long)(dblocks / XFS_AG_MIN_BLOCKS(blocklog) +
1124 (dblocks % XFS_AG_MIN_BLOCKS(blocklog) != 0)));
1125 usage();
1126 }
1127
1128 if (agsize > XFS_AG_MAX_BLOCKS(blocklog)) {
1129 fprintf(stderr,
1130 _("too few allocation groups for size = %lld\n"), (long long)agsize);
1131 fprintf(stderr,
1132 _("need at least %lld allocation groups\n"),
1133 (long long)(dblocks / XFS_AG_MAX_BLOCKS(blocklog) +
1134 (dblocks % XFS_AG_MAX_BLOCKS(blocklog) != 0)));
1135 usage();
1136 }
1137
1138 /*
1139 * If the last AG is too small, reduce the filesystem size
1140 * and drop the blocks.
1141 */
1142 if ( dblocks % agsize != 0 &&
1143 (dblocks % agsize < XFS_AG_MIN_BLOCKS(blocklog))) {
1144 fprintf(stderr,
1145 _("last AG size %lld blocks too small, minimum size is %lld blocks\n"),
1146 (long long)(dblocks % agsize),
1147 (long long)XFS_AG_MIN_BLOCKS(blocklog));
1148 usage();
1149 }
1150
1151 /*
1152 * If agcount is too large, make it smaller.
1153 */
1154 if (agcount > XFS_MAX_AGNUMBER + 1) {
1155 fprintf(stderr,
1156 _("%lld allocation groups is too many, maximum is %lld\n"),
1157 (long long)agcount, (long long)XFS_MAX_AGNUMBER + 1);
1158 usage();
1159 }
1160 }
1161
1162 static void
1163 zero_old_xfs_structures(
1164 libxfs_init_t *xi,
1165 xfs_sb_t *new_sb)
1166 {
1167 void *buf;
1168 xfs_sb_t sb;
1169 uint32_t bsize;
1170 int i;
1171 xfs_off_t off;
1172
1173 /*
1174 * We open regular files with O_TRUNC|O_CREAT. Nothing to do here...
1175 */
1176 if (xi->disfile && xi->dcreat)
1177 return;
1178
1179 /*
1180 * read in existing filesystem superblock, use its geometry
1181 * settings and zero the existing secondary superblocks.
1182 */
1183 buf = memalign(libxfs_device_alignment(), new_sb->sb_sectsize);
1184 if (!buf) {
1185 fprintf(stderr,
1186 _("error reading existing superblock -- failed to memalign buffer\n"));
1187 return;
1188 }
1189 memset(buf, 0, new_sb->sb_sectsize);
1190
1191 /*
1192 * If we are creating an image file, it might be of zero length at this
1193 * point in time. Hence reading the existing superblock is going to
1194 * return zero bytes. It's not a failure we need to warn about in this
1195 * case.
1196 */
1197 off = pread(xi->dfd, buf, new_sb->sb_sectsize, 0);
1198 if (off != new_sb->sb_sectsize) {
1199 if (!xi->disfile)
1200 fprintf(stderr,
1201 _("error reading existing superblock: %s\n"),
1202 strerror(errno));
1203 goto done;
1204 }
1205 libxfs_sb_from_disk(&sb, buf);
1206
1207 /*
1208 * perform same basic superblock validation to make sure we
1209 * actually zero secondary blocks
1210 */
1211 if (sb.sb_magicnum != XFS_SB_MAGIC || sb.sb_blocksize == 0)
1212 goto done;
1213
1214 for (bsize = 1, i = 0; bsize < sb.sb_blocksize &&
1215 i < sizeof(sb.sb_blocksize) * NBBY; i++)
1216 bsize <<= 1;
1217
1218 if (i < XFS_MIN_BLOCKSIZE_LOG || i > XFS_MAX_BLOCKSIZE_LOG ||
1219 i != sb.sb_blocklog)
1220 goto done;
1221
1222 if (sb.sb_dblocks > ((uint64_t)sb.sb_agcount * sb.sb_agblocks) ||
1223 sb.sb_dblocks < ((uint64_t)(sb.sb_agcount - 1) *
1224 sb.sb_agblocks + XFS_MIN_AG_BLOCKS))
1225 goto done;
1226
1227 /*
1228 * block size and basic geometry seems alright, zero the secondaries.
1229 */
1230 memset(buf, 0, new_sb->sb_sectsize);
1231 off = 0;
1232 for (i = 1; i < sb.sb_agcount; i++) {
1233 off += sb.sb_agblocks;
1234 if (pwrite(xi->dfd, buf, new_sb->sb_sectsize,
1235 off << sb.sb_blocklog) == -1)
1236 break;
1237 }
1238 done:
1239 free(buf);
1240 }
1241
1242 static void
1243 discard_blocks(dev_t dev, uint64_t nsectors)
1244 {
1245 int fd;
1246
1247 /*
1248 * We intentionally ignore errors from the discard ioctl. It is
1249 * not necessary for the mkfs functionality but just an optimization.
1250 */
1251 fd = libxfs_device_to_fd(dev);
1252 if (fd > 0)
1253 platform_discard_blocks(fd, 0, nsectors << 9);
1254 }
1255
1256 static __attribute__((noreturn)) void
1257 illegal_option(
1258 const char *value,
1259 struct opt_params *opts,
1260 int index,
1261 const char *reason)
1262 {
1263 fprintf(stderr,
1264 _("Invalid value %s for -%c %s option. %s\n"),
1265 value, opts->name, opts->subopts[index],
1266 reason);
1267 usage();
1268 }
1269
1270 /*
1271 * Check for conflicts and option respecification.
1272 */
1273 static void
1274 check_opt(
1275 struct opt_params *opts,
1276 int index,
1277 bool str_seen)
1278 {
1279 struct subopt_param *sp = &opts->subopt_params[index];
1280 int i;
1281
1282 if (sp->index != index) {
1283 fprintf(stderr,
1284 _("Developer screwed up option parsing (%d/%d)! Please report!\n"),
1285 sp->index, index);
1286 reqval(opts->name, opts->subopts, index);
1287 }
1288
1289 /*
1290 * Check for respecification of the option. This is more complex than it
1291 * seems because some options are parsed twice - once as a string during
1292 * input parsing, then later the string is passed to getnum for
1293 * conversion into a number and bounds checking. Hence the two variables
1294 * used to track the different uses based on the @str parameter passed
1295 * to us.
1296 */
1297 if (!str_seen) {
1298 if (sp->seen)
1299 respec(opts->name, opts->subopts, index);
1300 sp->seen = true;
1301 } else {
1302 if (sp->str_seen)
1303 respec(opts->name, opts->subopts, index);
1304 sp->str_seen = true;
1305 }
1306
1307 /* check for conflicts with the option */
1308 for (i = 0; i < MAX_CONFLICTS; i++) {
1309 struct _conflict *con = &sp->conflicts[i];
1310
1311 if (con->subopt == LAST_CONFLICT)
1312 break;
1313 if (con->opts->subopt_params[con->subopt].seen ||
1314 con->opts->subopt_params[con->subopt].str_seen)
1315 conflict(opts, index, con->opts, con->subopt);
1316 }
1317 }
1318
1319 static long long
1320 getnum(
1321 const char *str,
1322 struct opt_params *opts,
1323 int index)
1324 {
1325 struct subopt_param *sp = &opts->subopt_params[index];
1326 long long c;
1327
1328 check_opt(opts, index, false);
1329 /* empty strings might just return a default value */
1330 if (!str || *str == '\0') {
1331 if (sp->defaultval == SUBOPT_NEEDS_VAL)
1332 reqval(opts->name, opts->subopts, index);
1333 return sp->defaultval;
1334 }
1335
1336 if (sp->minval == 0 && sp->maxval == 0) {
1337 fprintf(stderr,
1338 _("Option -%c %s has undefined minval/maxval."
1339 "Can't verify value range. This is a bug.\n"),
1340 opts->name, opts->subopts[index]);
1341 exit(1);
1342 }
1343
1344 /*
1345 * Some values are pure numbers, others can have suffixes that define
1346 * the units of the number. Those get passed to cvtnum(), otherwise we
1347 * convert it ourselves to guarantee there is no trailing garbage in the
1348 * number.
1349 */
1350 if (sp->convert)
1351 c = cvtnum(blocksize, sectorsize, str);
1352 else {
1353 char *str_end;
1354
1355 c = strtoll(str, &str_end, 0);
1356 if (c == 0 && str_end == str)
1357 illegal_option(str, opts, index,
1358 _("Value not recognized as number."));
1359 if (*str_end != '\0')
1360 illegal_option(str, opts, index,
1361 _("Unit suffixes are not allowed."));
1362 }
1363
1364 /* Validity check the result. */
1365 if (c < sp->minval)
1366 illegal_option(str, opts, index, _("Value is too small."));
1367 else if (c > sp->maxval)
1368 illegal_option(str, opts, index, _("Value is too large."));
1369 if (sp->is_power_2 && !ispow2(c))
1370 illegal_option(str, opts, index, _("Value must be a power of 2."));
1371 return c;
1372 }
1373
1374 /*
1375 * Option is a string - do all the option table work, and check there
1376 * is actually an option string. Otherwise we don't do anything with the string
1377 * here - validation will be done later when the string is converted to a value
1378 * or used as a file/device path.
1379 */
1380 static char *
1381 getstr(
1382 char *str,
1383 struct opt_params *opts,
1384 int index)
1385 {
1386 check_opt(opts, index, true);
1387
1388 /* empty strings for string options are not valid */
1389 if (!str || *str == '\0')
1390 reqval(opts->name, opts->subopts, index);
1391 return str;
1392 }
1393
1394 static int
1395 block_opts_parser(
1396 struct opt_params *opts,
1397 int subopt,
1398 char *value,
1399 struct cli_params *cli)
1400 {
1401 switch (subopt) {
1402 case B_SIZE:
1403 cli->blocksize = getnum(value, opts, subopt);
1404 break;
1405 default:
1406 return -EINVAL;
1407 }
1408 return 0;
1409 }
1410
1411 static int
1412 data_opts_parser(
1413 struct opt_params *opts,
1414 int subopt,
1415 char *value,
1416 struct cli_params *cli)
1417 {
1418 switch (subopt) {
1419 case D_AGCOUNT:
1420 cli->agcount = getnum(value, opts, subopt);
1421 break;
1422 case D_AGSIZE:
1423 cli->agsize = getstr(value, opts, subopt);
1424 break;
1425 case D_FILE:
1426 cli->xi->disfile = getnum(value, opts, subopt);
1427 break;
1428 case D_NAME:
1429 cli->xi->dname = getstr(value, opts, subopt);
1430 break;
1431 case D_SIZE:
1432 cli->dsize = getstr(value, opts, subopt);
1433 break;
1434 case D_SUNIT:
1435 cli->dsunit = getnum(value, opts, subopt);
1436 break;
1437 case D_SWIDTH:
1438 cli->dswidth = getnum(value, opts, subopt);
1439 break;
1440 case D_SU:
1441 cli->dsu = getstr(value, opts, subopt);
1442 break;
1443 case D_SW:
1444 cli->dsw = getnum(value, opts, subopt);
1445 break;
1446 case D_NOALIGN:
1447 cli->sb_feat.nodalign = getnum(value, opts, subopt);
1448 break;
1449 case D_SECTSIZE:
1450 cli->sectorsize = getnum(value, opts, subopt);
1451 break;
1452 case D_RTINHERIT:
1453 if (getnum(value, opts, subopt))
1454 cli->fsx.fsx_xflags |= XFS_DIFLAG_RTINHERIT;
1455 break;
1456 case D_PROJINHERIT:
1457 cli->fsx.fsx_projid = getnum(value, opts, subopt);
1458 cli->fsx.fsx_xflags |= XFS_DIFLAG_PROJINHERIT;
1459 break;
1460 case D_EXTSZINHERIT:
1461 cli->fsx.fsx_extsize = getnum(value, opts, subopt);
1462 cli->fsx.fsx_xflags |= XFS_DIFLAG_EXTSZINHERIT;
1463 break;
1464 case D_COWEXTSIZE:
1465 cli->fsx.fsx_cowextsize = getnum(value, opts, subopt);
1466 cli->fsx.fsx_xflags |= FS_XFLAG_COWEXTSIZE;
1467 break;
1468 default:
1469 return -EINVAL;
1470 }
1471 return 0;
1472 }
1473
1474 static int
1475 inode_opts_parser(
1476 struct opt_params *opts,
1477 int subopt,
1478 char *value,
1479 struct cli_params *cli)
1480 {
1481 switch (subopt) {
1482 case I_ALIGN:
1483 cli->sb_feat.inode_align = getnum(value, opts, subopt);
1484 break;
1485 case I_MAXPCT:
1486 cli->imaxpct = getnum(value, opts, subopt);
1487 break;
1488 case I_PERBLOCK:
1489 cli->inopblock = getnum(value, opts, subopt);
1490 break;
1491 case I_SIZE:
1492 cli->inodesize = getnum(value, opts, subopt);
1493 break;
1494 case I_ATTR:
1495 cli->sb_feat.attr_version = getnum(value, opts, subopt);
1496 break;
1497 case I_PROJID32BIT:
1498 cli->sb_feat.projid32bit = getnum(value, opts, subopt);
1499 break;
1500 case I_SPINODES:
1501 cli->sb_feat.spinodes = getnum(value, opts, subopt);
1502 break;
1503 default:
1504 return -EINVAL;
1505 }
1506 return 0;
1507 }
1508
1509 static int
1510 log_opts_parser(
1511 struct opt_params *opts,
1512 int subopt,
1513 char *value,
1514 struct cli_params *cli)
1515 {
1516 switch (subopt) {
1517 case L_AGNUM:
1518 cli->logagno = getnum(value, opts, subopt);
1519 break;
1520 case L_FILE:
1521 cli->xi->lisfile = getnum(value, opts, subopt);
1522 break;
1523 case L_INTERNAL:
1524 cli->loginternal = getnum(value, opts, subopt);
1525 break;
1526 case L_SU:
1527 cli->lsu = getstr(value, opts, subopt);
1528 break;
1529 case L_SUNIT:
1530 cli->lsunit = getnum(value, opts, subopt);
1531 break;
1532 case L_NAME:
1533 case L_DEV:
1534 cli->xi->logname = getstr(value, opts, subopt);
1535 cli->loginternal = 0;
1536 break;
1537 case L_VERSION:
1538 cli->sb_feat.log_version = getnum(value, opts, subopt);
1539 break;
1540 case L_SIZE:
1541 cli->logsize = getstr(value, opts, subopt);
1542 break;
1543 case L_SECTSIZE:
1544 cli->lsectorsize = getnum(value, opts, subopt);
1545 break;
1546 case L_LAZYSBCNTR:
1547 cli->sb_feat.lazy_sb_counters = getnum(value, opts, subopt);
1548 break;
1549 default:
1550 return -EINVAL;
1551 }
1552 return 0;
1553 }
1554
1555 static int
1556 meta_opts_parser(
1557 struct opt_params *opts,
1558 int subopt,
1559 char *value,
1560 struct cli_params *cli)
1561 {
1562 switch (subopt) {
1563 case M_CRC:
1564 cli->sb_feat.crcs_enabled = getnum(value, opts, subopt);
1565 if (cli->sb_feat.crcs_enabled)
1566 cli->sb_feat.dirftype = true;
1567 break;
1568 case M_FINOBT:
1569 cli->sb_feat.finobt = getnum(value, opts, subopt);
1570 break;
1571 case M_UUID:
1572 if (!value || *value == '\0')
1573 reqval('m', opts->subopts, subopt);
1574 if (platform_uuid_parse(value, &cli->uuid))
1575 illegal(value, "m uuid");
1576 break;
1577 case M_RMAPBT:
1578 cli->sb_feat.rmapbt = getnum(value, opts, subopt);
1579 break;
1580 case M_REFLINK:
1581 cli->sb_feat.reflink = getnum(value, opts, subopt);
1582 break;
1583 default:
1584 return -EINVAL;
1585 }
1586 return 0;
1587 }
1588
1589 static int
1590 naming_opts_parser(
1591 struct opt_params *opts,
1592 int subopt,
1593 char *value,
1594 struct cli_params *cli)
1595 {
1596 switch (subopt) {
1597 case N_SIZE:
1598 cli->dirblocksize = getstr(value, opts, subopt);
1599 break;
1600 case N_VERSION:
1601 value = getstr(value, &nopts, subopt);
1602 if (!strcasecmp(value, "ci")) {
1603 /* ASCII CI mode */
1604 cli->sb_feat.nci = true;
1605 } else {
1606 cli->sb_feat.dir_version = getnum(value, opts, subopt);
1607 }
1608 break;
1609 case N_FTYPE:
1610 cli->sb_feat.dirftype = getnum(value, opts, subopt);
1611 break;
1612 default:
1613 return -EINVAL;
1614 }
1615 return 0;
1616 }
1617
1618 static int
1619 rtdev_opts_parser(
1620 struct opt_params *opts,
1621 int subopt,
1622 char *value,
1623 struct cli_params *cli)
1624 {
1625 switch (subopt) {
1626 case R_EXTSIZE:
1627 cli->rtextsize = getstr(value, opts, subopt);
1628 break;
1629 case R_FILE:
1630 cli->xi->risfile = getnum(value, opts, subopt);
1631 break;
1632 case R_NAME:
1633 case R_DEV:
1634 cli->xi->rtname = getstr(value, opts, subopt);
1635 break;
1636 case R_SIZE:
1637 cli->rtsize = getstr(value, opts, subopt);
1638 break;
1639 case R_NOALIGN:
1640 cli->sb_feat.nortalign = getnum(value, opts, subopt);
1641 break;
1642 default:
1643 return -EINVAL;
1644 }
1645 return 0;
1646 }
1647
1648 static int
1649 sector_opts_parser(
1650 struct opt_params *opts,
1651 int subopt,
1652 char *value,
1653 struct cli_params *cli)
1654 {
1655 switch (subopt) {
1656 case S_SIZE:
1657 case S_SECTSIZE:
1658 cli->sectorsize = getnum(value, opts, subopt);
1659 cli->lsectorsize = cli->sectorsize;
1660 break;
1661 default:
1662 return -EINVAL;
1663 }
1664 return 0;
1665 }
1666
1667 struct subopts {
1668 char opt;
1669 struct opt_params *opts;
1670 int (*parser)(struct opt_params *opts,
1671 int subopt,
1672 char *value,
1673 struct cli_params *cli);
1674 } subopt_tab[] = {
1675 { 'b', &bopts, block_opts_parser },
1676 { 'd', &dopts, data_opts_parser },
1677 { 'i', &iopts, inode_opts_parser },
1678 { 'l', &lopts, log_opts_parser },
1679 { 'm', &mopts, meta_opts_parser },
1680 { 'n', &nopts, naming_opts_parser },
1681 { 'r', &ropts, rtdev_opts_parser },
1682 { 's', &sopts, sector_opts_parser },
1683 { '\0', NULL, NULL },
1684 };
1685
1686 static void
1687 parse_subopts(
1688 char opt,
1689 char *arg,
1690 struct cli_params *cli)
1691 {
1692 struct subopts *sop = &subopt_tab[0];
1693 char *p;
1694 int ret = 0;
1695
1696 while (sop->opts) {
1697 if (sop->opt == opt)
1698 break;
1699 sop++;
1700 }
1701
1702 /* should never happen */
1703 if (!sop->opts)
1704 return;
1705
1706 p = arg;
1707 while (*p != '\0') {
1708 char **subopts = (char **)sop->opts->subopts;
1709 char *value;
1710 int subopt;
1711
1712 subopt = getsubopt(&p, subopts, &value);
1713
1714 ret = (sop->parser)(sop->opts, subopt, value, cli);
1715 if (ret)
1716 unknown(opt, value);
1717 }
1718 }
1719
1720 static void
1721 validate_sectorsize(
1722 struct mkfs_params *cfg,
1723 struct cli_params *cli,
1724 struct mkfs_default_params *dft,
1725 struct fs_topology *ft,
1726 char *dfile,
1727 int dry_run,
1728 int force_overwrite)
1729 {
1730 /* set configured sector sizes in preparation for checks */
1731 if (!cli->sectorsize) {
1732 cfg->sectorsize = dft->sectorsize;
1733 } else {
1734 cfg->sectorsize = cli->sectorsize;
1735 }
1736 cfg->sectorlog = libxfs_highbit32(cfg->sectorsize);
1737
1738 /*
1739 * Before anything else, verify that we are correctly operating on
1740 * files or block devices and set the control parameters correctly.
1741 */
1742 check_device_type(dfile, &cli->xi->disfile, !cli->dsize, !dfile,
1743 dry_run ? NULL : &cli->xi->dcreat, "d");
1744 if (!cli->loginternal)
1745 check_device_type(cli->xi->logname, &cli->xi->lisfile,
1746 !cli->logsize, !cli->xi->logname,
1747 dry_run ? NULL : &cli->xi->lcreat, "l");
1748 if (cli->xi->rtname)
1749 check_device_type(cli->xi->rtname, &cli->xi->risfile,
1750 !cli->rtsize, !cli->xi->rtname,
1751 dry_run ? NULL : &cli->xi->rcreat, "r");
1752
1753 /*
1754 * Explicitly disable direct IO for image files so we don't error out on
1755 * sector size mismatches between the new filesystem and the underlying
1756 * host filesystem.
1757 */
1758 if (cli->xi->disfile || cli->xi->lisfile || cli->xi->risfile)
1759 cli->xi->isdirect = 0;
1760
1761 memset(ft, 0, sizeof(*ft));
1762 get_topology(cli->xi, ft, force_overwrite);
1763
1764 if (!cli->sectorsize) {
1765 /*
1766 * Unless specified manually on the command line use the
1767 * advertised sector size of the device. We use the physical
1768 * sector size unless the requested block size is smaller
1769 * than that, then we can use logical, but warn about the
1770 * inefficiency.
1771 *
1772 * Set the topology sectors if they were not probed to the
1773 * minimum supported sector size.
1774 */
1775
1776 if (!ft->lsectorsize)
1777 ft->lsectorsize = XFS_MIN_SECTORSIZE;
1778
1779 /* Older kernels may not have physical/logical distinction */
1780 if (!ft->psectorsize)
1781 ft->psectorsize = ft->lsectorsize;
1782
1783 cfg->sectorsize = ft->psectorsize;
1784 if (cfg->blocksize < cfg->sectorsize &&
1785 cfg->blocksize >= ft->lsectorsize) {
1786 fprintf(stderr,
1787 _("specified blocksize %d is less than device physical sector size %d\n"
1788 "switching to logical sector size %d\n"),
1789 cfg->blocksize, ft->psectorsize,
1790 ft->lsectorsize);
1791 cfg->sectorsize = ft->lsectorsize;
1792 }
1793
1794 cfg->sectorlog = libxfs_highbit32(cfg->sectorsize);
1795 }
1796
1797 /* validate specified/probed sector size */
1798 if (cfg->sectorsize < XFS_MIN_SECTORSIZE ||
1799 cfg->sectorsize > XFS_MAX_SECTORSIZE) {
1800 fprintf(stderr, _("illegal sector size %d\n"), cfg->sectorsize);
1801 usage();
1802 }
1803
1804 if (cfg->blocksize < cfg->sectorsize) {
1805 fprintf(stderr,
1806 _("block size %d cannot be smaller than sector size %d\n"),
1807 cfg->blocksize, cfg->sectorsize);
1808 usage();
1809 }
1810
1811 if (cfg->sectorsize < ft->lsectorsize) {
1812 fprintf(stderr, _("illegal sector size %d; hw sector is %d\n"),
1813 cfg->sectorsize, ft->lsectorsize);
1814 usage();
1815 }
1816 }
1817
1818 static void
1819 validate_blocksize(
1820 struct mkfs_params *cfg,
1821 struct cli_params *cli,
1822 struct mkfs_default_params *dft)
1823 {
1824 /*
1825 * Blocksize and sectorsize first, other things depend on them
1826 * For RAID4/5/6 we want to align sector size and block size,
1827 * so we need to start with the device geometry extraction too.
1828 */
1829 if (!cli->blocksize)
1830 cfg->blocksize = dft->blocksize;
1831 else
1832 cfg->blocksize = cli->blocksize;
1833 cfg->blocklog = libxfs_highbit32(cfg->blocksize);
1834
1835 /* validate block sizes are in range */
1836 if (cfg->blocksize < XFS_MIN_BLOCKSIZE ||
1837 cfg->blocksize > XFS_MAX_BLOCKSIZE) {
1838 fprintf(stderr, _("illegal block size %d\n"), cfg->blocksize);
1839 usage();
1840 }
1841
1842 if (cli->sb_feat.crcs_enabled &&
1843 cfg->blocksize < XFS_MIN_CRC_BLOCKSIZE) {
1844 fprintf(stderr,
1845 _("Minimum block size for CRC enabled filesystems is %d bytes.\n"),
1846 XFS_MIN_CRC_BLOCKSIZE);
1847 usage();
1848 }
1849
1850 }
1851
1852 /*
1853 * Grab log sector size and validate.
1854 *
1855 * XXX: should we probe sector size on external log device rather than using
1856 * the data device sector size?
1857 */
1858 static void
1859 validate_log_sectorsize(
1860 struct mkfs_params *cfg,
1861 struct cli_params *cli,
1862 struct mkfs_default_params *dft)
1863 {
1864
1865 if (cli->loginternal && cli->lsectorsize &&
1866 cli->lsectorsize != cfg->sectorsize) {
1867 fprintf(stderr,
1868 _("Can't change sector size on internal log!\n"));
1869 usage();
1870 }
1871
1872 if (cli->lsectorsize)
1873 cfg->lsectorsize = cli->lsectorsize;
1874 else if (cli->loginternal)
1875 cfg->lsectorsize = cfg->sectorsize;
1876 else
1877 cfg->lsectorsize = dft->sectorsize;
1878 cfg->lsectorlog = libxfs_highbit32(cfg->lsectorsize);
1879
1880 if (cfg->lsectorsize < XFS_MIN_SECTORSIZE ||
1881 cfg->lsectorsize > XFS_MAX_SECTORSIZE ||
1882 cfg->lsectorsize > cfg->blocksize) {
1883 fprintf(stderr, _("illegal log sector size %d\n"),
1884 cfg->lsectorsize);
1885 usage();
1886 }
1887 if (cfg->lsectorsize > XFS_MIN_SECTORSIZE) {
1888 if (cli->sb_feat.log_version < 2) {
1889 /* user specified non-default log version */
1890 fprintf(stderr,
1891 _("Version 1 logs do not support sector size %d\n"),
1892 cfg->lsectorsize);
1893 usage();
1894 }
1895 }
1896
1897 /* if lsu or lsunit was specified, automatically use v2 logs */
1898 if ((cli_opt_set(&lopts, L_SU) || cli_opt_set(&lopts, L_SUNIT)) &&
1899 cli->sb_feat.log_version == 1) {
1900 fprintf(stderr,
1901 _("log stripe unit specified, using v2 logs\n"));
1902 cli->sb_feat.log_version = 2;
1903 }
1904
1905 }
1906
1907 /*
1908 * Check that the incoming features make sense. The CLI structure was
1909 * initialised with the default values before parsing, so we can just
1910 * check it and copy it straight across to the cfg structure if it
1911 * checks out.
1912 */
1913 static void
1914 validate_sb_features(
1915 struct mkfs_params *cfg,
1916 struct cli_params *cli)
1917 {
1918 /*
1919 * Now we have blocks and sector sizes set up, check parameters that are
1920 * no longer optional for CRC enabled filesystems. Catch them up front
1921 * here before doing anything else.
1922 */
1923 if (cli->sb_feat.crcs_enabled) {
1924 /* minimum inode size is 512 bytes, rest checked later */
1925 if (cli->inodesize &&
1926 cli->inodesize < (1 << XFS_DINODE_DFL_CRC_LOG)) {
1927 fprintf(stderr,
1928 _("Minimum inode size for CRCs is %d bytes\n"),
1929 1 << XFS_DINODE_DFL_CRC_LOG);
1930 usage();
1931 }
1932
1933 /* inodes always aligned */
1934 if (!cli->sb_feat.inode_align) {
1935 fprintf(stderr,
1936 _("Inodes always aligned for CRC enabled filesystems\n"));
1937 usage();
1938 }
1939
1940 /* lazy sb counters always on */
1941 if (!cli->sb_feat.lazy_sb_counters) {
1942 fprintf(stderr,
1943 _("Lazy superblock counters always enabled for CRC enabled filesystems\n"));
1944 usage();
1945 }
1946
1947 /* version 2 logs always on */
1948 if (cli->sb_feat.log_version != 2) {
1949 fprintf(stderr,
1950 _("V2 logs always enabled for CRC enabled filesystems\n"));
1951 usage();
1952 }
1953
1954 /* attr2 always on */
1955 if (cli->sb_feat.attr_version != 2) {
1956 fprintf(stderr,
1957 _("V2 attribute format always enabled on CRC enabled filesystems\n"));
1958 usage();
1959 }
1960
1961 /* 32 bit project quota always on */
1962 /* attr2 always on */
1963 if (!cli->sb_feat.projid32bit) {
1964 fprintf(stderr,
1965 _("32 bit Project IDs always enabled on CRC enabled filesystems\n"));
1966 usage();
1967 }
1968
1969 /* ftype always on */
1970 if (!cli->sb_feat.dirftype) {
1971 fprintf(stderr,
1972 _("Directory ftype field always enabled on CRC enabled filesystems\n"));
1973 usage();
1974 }
1975
1976 } else {
1977 /*
1978 * The kernel doesn't currently support crc=0,finobt=1
1979 * filesystems. If crcs are not enabled and the user has not
1980 * explicitly turned finobt on, then silently turn it off to
1981 * avoid an unnecessary warning.
1982 * If the user explicitly tried to use crc=0,finobt=1,
1983 * then issue an error.
1984 * The same is also for sparse inodes.
1985 */
1986 if (cli->sb_feat.finobt && cli_opt_set(&mopts, M_FINOBT)) {
1987 fprintf(stderr,
1988 _("finobt not supported without CRC support\n"));
1989 usage();
1990 }
1991 cli->sb_feat.finobt = false;
1992
1993 if (cli->sb_feat.spinodes && cli_opt_set(&iopts, I_SPINODES)) {
1994 fprintf(stderr,
1995 _("sparse inodes not supported without CRC support\n"));
1996 usage();
1997 }
1998 cli->sb_feat.spinodes = false;
1999
2000 if (cli->sb_feat.rmapbt) {
2001 fprintf(stderr,
2002 _("rmapbt not supported without CRC support\n"));
2003 usage();
2004 }
2005 cli->sb_feat.rmapbt = false;
2006
2007 if (cli->sb_feat.reflink) {
2008 fprintf(stderr,
2009 _("reflink not supported without CRC support\n"));
2010 usage();
2011 }
2012 cli->sb_feat.reflink = false;
2013 }
2014
2015 if ((cli->fsx.fsx_xflags & FS_XFLAG_COWEXTSIZE) &&
2016 !cli->sb_feat.reflink) {
2017 fprintf(stderr,
2018 _("cowextsize not supported without reflink support\n"));
2019 usage();
2020 }
2021
2022 if (cli->sb_feat.reflink && cli->xi->rtname) {
2023 fprintf(stderr,
2024 _("reflink not supported with realtime devices\n"));
2025 usage();
2026 cli->sb_feat.reflink = false;
2027 }
2028
2029 if (cli->sb_feat.rmapbt && cli->xi->rtname) {
2030 fprintf(stderr,
2031 _("rmapbt not supported with realtime devices\n"));
2032 usage();
2033 cli->sb_feat.rmapbt = false;
2034 }
2035
2036 /*
2037 * Copy features across to config structure now.
2038 */
2039 cfg->sb_feat = cli->sb_feat;
2040 if (!platform_uuid_is_null(&cli->uuid))
2041 platform_uuid_copy(&cfg->uuid, &cli->uuid);
2042 }
2043
2044 static void
2045 validate_dirblocksize(
2046 struct mkfs_params *cfg,
2047 struct cli_params *cli)
2048 {
2049
2050 if (cli->dirblocksize)
2051 cfg->dirblocksize = getnum(cli->dirblocksize, &nopts, N_SIZE);
2052
2053 if (cfg->dirblocksize) {
2054 if (cfg->dirblocksize < cfg->blocksize ||
2055 cfg->dirblocksize > XFS_MAX_BLOCKSIZE) {
2056 fprintf(stderr, _("illegal directory block size %d\n"),
2057 cfg->dirblocksize);
2058 usage();
2059 }
2060 cfg->dirblocklog = libxfs_highbit32(cfg->dirblocksize);
2061 return;
2062 }
2063
2064 /* use default size based on current block size */
2065 if (cfg->blocksize < (1 << XFS_MIN_REC_DIRSIZE))
2066 cfg->dirblocklog = XFS_MIN_REC_DIRSIZE;
2067 else
2068 cfg->dirblocklog = cfg->blocklog;
2069 cfg->dirblocksize = 1 << cfg->dirblocklog;
2070 }
2071
2072 static void
2073 validate_inodesize(
2074 struct mkfs_params *cfg,
2075 struct cli_params *cli)
2076 {
2077
2078 if (cli->inopblock)
2079 cfg->inodelog = cfg->blocklog - libxfs_highbit32(cli->inopblock);
2080 else if (cli->inodesize)
2081 cfg->inodelog = libxfs_highbit32(cli->inodesize);
2082 else if (cfg->sb_feat.crcs_enabled)
2083 cfg->inodelog = XFS_DINODE_DFL_CRC_LOG;
2084 else
2085 cfg->inodelog = XFS_DINODE_DFL_LOG;
2086
2087 cfg->inodesize = 1 << cfg->inodelog;
2088 cfg->inopblock = cfg->blocksize / cfg->inodesize;
2089
2090 /* input parsing has already validated non-crc inode size range */
2091 if (cfg->sb_feat.crcs_enabled &&
2092 cfg->inodelog < XFS_DINODE_DFL_CRC_LOG) {
2093 fprintf(stderr,
2094 _("Minimum inode size for CRCs is %d bytes\n"),
2095 1 << XFS_DINODE_DFL_CRC_LOG);
2096 usage();
2097 }
2098
2099 if (cfg->inodesize > cfg->blocksize / XFS_MIN_INODE_PERBLOCK ||
2100 cfg->inopblock < XFS_MIN_INODE_PERBLOCK ||
2101 cfg->inodesize < XFS_DINODE_MIN_SIZE ||
2102 cfg->inodesize > XFS_DINODE_MAX_SIZE) {
2103 int maxsz;
2104
2105 fprintf(stderr, _("illegal inode size %d\n"), cfg->inodesize);
2106 maxsz = min(cfg->blocksize / XFS_MIN_INODE_PERBLOCK,
2107 XFS_DINODE_MAX_SIZE);
2108 if (XFS_DINODE_MIN_SIZE == maxsz)
2109 fprintf(stderr,
2110 _("allowable inode size with %d byte blocks is %d\n"),
2111 cfg->blocksize, XFS_DINODE_MIN_SIZE);
2112 else
2113 fprintf(stderr,
2114 _("allowable inode size with %d byte blocks is between %d and %d\n"),
2115 cfg->blocksize, XFS_DINODE_MIN_SIZE, maxsz);
2116 exit(1);
2117 }
2118 }
2119
2120 static xfs_rfsblock_t
2121 calc_dev_size(
2122 char *size,
2123 struct mkfs_params *cfg,
2124 struct opt_params *opts,
2125 int sizeopt,
2126 char *type)
2127 {
2128 uint64_t dbytes;
2129 xfs_rfsblock_t dblocks;
2130
2131 if (!size)
2132 return 0;
2133
2134 dbytes = getnum(size, opts, sizeopt);
2135 if (dbytes % XFS_MIN_BLOCKSIZE) {
2136 fprintf(stderr,
2137 _("illegal %s length %lld, not a multiple of %d\n"),
2138 type, (long long)dbytes, XFS_MIN_BLOCKSIZE);
2139 usage();
2140 }
2141 dblocks = (xfs_rfsblock_t)(dbytes >> cfg->blocklog);
2142 if (dbytes % cfg->blocksize) {
2143 fprintf(stderr,
2144 _("warning: %s length %lld not a multiple of %d, truncated to %lld\n"),
2145 type, (long long)dbytes, cfg->blocksize,
2146 (long long)(dblocks << cfg->blocklog));
2147 }
2148 return dblocks;
2149 }
2150
2151 static void
2152 validate_rtextsize(
2153 struct mkfs_params *cfg,
2154 struct cli_params *cli,
2155 struct fs_topology *ft)
2156 {
2157 uint64_t rtextbytes;
2158
2159 /*
2160 * If specified, check rt extent size against its constraints.
2161 */
2162 if (cli->rtextsize) {
2163
2164 rtextbytes = getnum(cli->rtextsize, &ropts, R_EXTSIZE);
2165 if (rtextbytes % cfg->blocksize) {
2166 fprintf(stderr,
2167 _("illegal rt extent size %lld, not a multiple of %d\n"),
2168 (long long)rtextbytes, cfg->blocksize);
2169 usage();
2170 }
2171 cfg->rtextblocks = (xfs_extlen_t)(rtextbytes >> cfg->blocklog);
2172 } else {
2173 /*
2174 * If realtime extsize has not been specified by the user,
2175 * and the underlying volume is striped, then set rtextblocks
2176 * to the stripe width.
2177 */
2178 uint64_t rswidth;
2179
2180 if (!cfg->sb_feat.nortalign && !cli->xi->risfile &&
2181 !(!cli->rtsize && cli->xi->disfile))
2182 rswidth = ft->rtswidth;
2183 else
2184 rswidth = 0;
2185
2186 /* check that rswidth is a multiple of fs blocksize */
2187 if (!cfg->sb_feat.nortalign && rswidth &&
2188 !(BBTOB(rswidth) % cfg->blocksize)) {
2189 rswidth = DTOBT(rswidth, cfg->blocklog);
2190 rtextbytes = rswidth << cfg->blocklog;
2191 if (rtextbytes > XFS_MIN_RTEXTSIZE &&
2192 rtextbytes <= XFS_MAX_RTEXTSIZE) {
2193 cfg->rtextblocks = rswidth;
2194 }
2195 }
2196 if (!cfg->rtextblocks) {
2197 cfg->rtextblocks = (cfg->blocksize < XFS_MIN_RTEXTSIZE)
2198 ? XFS_MIN_RTEXTSIZE >> cfg->blocklog
2199 : 1;
2200 }
2201 }
2202 ASSERT(cfg->rtextblocks);
2203 }
2204
2205 /*
2206 * Validate the configured stripe geometry, or is none is specified, pull
2207 * the configuration from the underlying device.
2208 *
2209 * CLI parameters come in as different units, go out as filesystem blocks.
2210 */
2211 static void
2212 calc_stripe_factors(
2213 struct mkfs_params *cfg,
2214 struct cli_params *cli,
2215 struct fs_topology *ft)
2216 {
2217 long long int big_dswidth;
2218 int dsunit = 0;
2219 int dswidth = 0;
2220 int lsunit = 0;
2221 int dsu = 0;
2222 int dsw = 0;
2223 int lsu = 0;
2224 bool use_dev = false;
2225
2226 if (cli_opt_set(&dopts, D_SUNIT))
2227 dsunit = cli->dsunit;
2228 if (cli_opt_set(&dopts, D_SWIDTH))
2229 dswidth = cli->dswidth;
2230
2231 if (cli_opt_set(&dopts, D_SU))
2232 dsu = getnum(cli->dsu, &dopts, D_SU);
2233 if (cli_opt_set(&dopts, D_SW))
2234 dsw = cli->dsw;
2235
2236 /* data sunit/swidth options */
2237 if (cli_opt_set(&dopts, D_SUNIT) != cli_opt_set(&dopts, D_SWIDTH)) {
2238 fprintf(stderr,
2239 _("both data sunit and data swidth options must be specified\n"));
2240 usage();
2241 }
2242
2243 /* convert dsu/dsw to dsunit/dswidth and use them from now on */
2244 if (dsu || dsw) {
2245 if (cli_opt_set(&dopts, D_SU) != cli_opt_set(&dopts, D_SW)) {
2246 fprintf(stderr,
2247 _("both data su and data sw options must be specified\n"));
2248 usage();
2249 }
2250
2251 if (dsu % cfg->sectorsize) {
2252 fprintf(stderr,
2253 _("data su must be a multiple of the sector size (%d)\n"), cfg->sectorsize);
2254 usage();
2255 }
2256
2257 dsunit = (int)BTOBBT(dsu);
2258 big_dswidth = (long long int)dsunit * dsw;
2259 if (big_dswidth > INT_MAX) {
2260 fprintf(stderr,
2261 _("data stripe width (%lld) is too large of a multiple of the data stripe unit (%d)\n"),
2262 big_dswidth, dsunit);
2263 usage();
2264 }
2265 dswidth = big_dswidth;
2266 }
2267
2268 if ((dsunit && !dswidth) || (!dsunit && dswidth) ||
2269 (dsunit && (dswidth % dsunit != 0))) {
2270 fprintf(stderr,
2271 _("data stripe width (%d) must be a multiple of the data stripe unit (%d)\n"),
2272 dswidth, dsunit);
2273 usage();
2274 }
2275
2276 /* If sunit & swidth were manually specified as 0, same as noalign */
2277 if ((cli_opt_set(&dopts, D_SUNIT) || cli_opt_set(&dopts, D_SU)) &&
2278 !dsunit && !dswidth)
2279 cfg->sb_feat.nodalign = true;
2280
2281 /* if we are not using alignment, don't apply device defaults */
2282 if (cfg->sb_feat.nodalign) {
2283 cfg->dsunit = 0;
2284 cfg->dswidth = 0;
2285 goto check_lsunit;
2286 }
2287
2288 /* if no stripe config set, use the device default */
2289 if (!dsunit) {
2290 /* Ignore nonsense from device. XXX add more validation */
2291 if (ft->dsunit && ft->dswidth == 0) {
2292 fprintf(stderr,
2293 _("%s: Volume reports stripe unit of %d bytes and stripe width of 0, ignoring.\n"),
2294 progname, BBTOB(ft->dsunit));
2295 ft->dsunit = 0;
2296 ft->dswidth = 0;
2297 } else {
2298 dsunit = ft->dsunit;
2299 dswidth = ft->dswidth;
2300 use_dev = true;
2301 }
2302 } else {
2303 /* check and warn if user-specified alignment is sub-optimal */
2304 if (ft->dsunit && ft->dsunit != dsunit) {
2305 fprintf(stderr,
2306 _("%s: Specified data stripe unit %d is not the same as the volume stripe unit %d\n"),
2307 progname, dsunit, ft->dsunit);
2308 }
2309 if (ft->dswidth && ft->dswidth != dswidth) {
2310 fprintf(stderr,
2311 _("%s: Specified data stripe width %d is not the same as the volume stripe width %d\n"),
2312 progname, dswidth, ft->dswidth);
2313 }
2314 }
2315
2316 /*
2317 * now we have our stripe config, check it's a multiple of block
2318 * size.
2319 */
2320 if ((BBTOB(dsunit) % cfg->blocksize) ||
2321 (BBTOB(dswidth) % cfg->blocksize)) {
2322 /*
2323 * If we are using device defaults, just clear them and we're
2324 * good to go. Otherwise bail out with an error.
2325 */
2326 if (!use_dev) {
2327 fprintf(stderr,
2328 _("%s: Stripe unit(%d) or stripe width(%d) is not a multiple of the block size(%d)\n"),
2329 progname, BBTOB(dsunit), BBTOB(dswidth),
2330 cfg->blocksize);
2331 exit(1);
2332 }
2333 dsunit = 0;
2334 dswidth = 0;
2335 cfg->sb_feat.nodalign = true;
2336 }
2337
2338 /* convert from 512 byte blocks to fs blocksize */
2339 cfg->dsunit = DTOBT(dsunit, cfg->blocklog);
2340 cfg->dswidth = DTOBT(dswidth, cfg->blocklog);
2341
2342 check_lsunit:
2343 /* log sunit options */
2344 if (cli_opt_set(&lopts, L_SUNIT))
2345 lsunit = cli->lsunit;
2346 else if (cli_opt_set(&lopts, L_SU))
2347 lsu = getnum(cli->lsu, &lopts, L_SU);
2348 else if (cfg->lsectorsize > XLOG_HEADER_SIZE)
2349 lsu = cfg->blocksize; /* lsunit matches filesystem block size */
2350
2351 if (lsu) {
2352 /* verify if lsu is a multiple block size */
2353 if (lsu % cfg->blocksize != 0) {
2354 fprintf(stderr,
2355 _("log stripe unit (%d) must be a multiple of the block size (%d)\n"),
2356 lsu, cfg->blocksize);
2357 usage();
2358 }
2359 lsunit = (int)BTOBBT(lsu);
2360 }
2361 if (BBTOB(lsunit) % cfg->blocksize != 0) {
2362 fprintf(stderr,
2363 _("log stripe unit (%d) must be a multiple of the block size (%d)\n"),
2364 BBTOB(lsunit), cfg->blocksize);
2365 usage();
2366 }
2367
2368 /*
2369 * check that log sunit is modulo fsblksize or default it to dsunit.
2370 */
2371 if (lsunit) {
2372 /* convert from 512 byte blocks to fs blocks */
2373 cfg->lsunit = DTOBT(lsunit, cfg->blocklog);
2374 } else if (cfg->sb_feat.log_version == 2 &&
2375 cfg->loginternal && cfg->dsunit) {
2376 /* lsunit and dsunit now in fs blocks */
2377 cfg->lsunit = cfg->dsunit;
2378 }
2379
2380 if (cfg->sb_feat.log_version == 2 &&
2381 cfg->lsunit * cfg->blocksize > 256 * 1024) {
2382 /* Warn only if specified on commandline */
2383 if (cli->lsu || cli->lsunit != -1) {
2384 fprintf(stderr,
2385 _("log stripe unit (%d bytes) is too large (maximum is 256KiB)\n"
2386 "log stripe unit adjusted to 32KiB\n"),
2387 (cfg->lsunit * cfg->blocksize));
2388 }
2389 /* XXX: 64k block size? */
2390 cfg->lsunit = (32 * 1024) / cfg->blocksize;
2391 }
2392
2393 }
2394
2395 static void
2396 open_devices(
2397 struct mkfs_params *cfg,
2398 struct libxfs_xinit *xi)
2399 {
2400 uint64_t sector_mask;
2401
2402 /*
2403 * Initialize. This will open the log and rt devices as well.
2404 */
2405 xi->setblksize = cfg->sectorsize;
2406 if (!libxfs_init(xi))
2407 usage();
2408 if (!xi->ddev) {
2409 fprintf(stderr, _("no device name given in argument list\n"));
2410 usage();
2411 }
2412
2413 /*
2414 * Ok, Linux only has a 1024-byte resolution on device _size_,
2415 * and the sizes below are in basic 512-byte blocks,
2416 * so if we have (size % 2), on any partition, we can't get
2417 * to the last 512 bytes. The same issue exists for larger
2418 * sector sizes - we cannot write past the last sector.
2419 *
2420 * So, we reduce the size (in basic blocks) to a perfect
2421 * multiple of the sector size, or 1024, whichever is larger.
2422 */
2423 sector_mask = (uint64_t)-1 << (max(cfg->sectorlog, 10) - BBSHIFT);
2424 xi->dsize &= sector_mask;
2425 xi->rtsize &= sector_mask;
2426 xi->logBBsize &= (uint64_t)-1 << (max(cfg->lsectorlog, 10) - BBSHIFT);
2427 }
2428
2429 static void
2430 discard_devices(
2431 struct libxfs_xinit *xi)
2432 {
2433 /*
2434 * This function has to be called after libxfs has been initialized.
2435 */
2436
2437 if (!xi->disfile)
2438 discard_blocks(xi->ddev, xi->dsize);
2439 if (xi->rtdev && !xi->risfile)
2440 discard_blocks(xi->rtdev, xi->rtsize);
2441 if (xi->logdev && xi->logdev != xi->ddev && !xi->lisfile)
2442 discard_blocks(xi->logdev, xi->logBBsize);
2443 }
2444
2445 static void
2446 validate_datadev(
2447 struct mkfs_params *cfg,
2448 struct cli_params *cli)
2449 {
2450 struct libxfs_xinit *xi = cli->xi;
2451
2452 if (!xi->dsize) {
2453 /*
2454 * if the device is a file, we can't validate the size here.
2455 * Instead, the file will be truncated to the correct length
2456 * later on. if it's not a file, we've got a dud device.
2457 */
2458 if (!xi->disfile) {
2459 fprintf(stderr, _("can't get size of data subvolume\n"));
2460 usage();
2461 }
2462 ASSERT(cfg->dblocks);
2463 } else if (cfg->dblocks) {
2464 /* check the size fits into the underlying device */
2465 if (cfg->dblocks > DTOBT(xi->dsize, cfg->blocklog)) {
2466 fprintf(stderr,
2467 _("size %s specified for data subvolume is too large, maximum is %lld blocks\n"),
2468 cli->dsize,
2469 (long long)DTOBT(xi->dsize, cfg->blocklog));
2470 usage();
2471 }
2472 } else {
2473 /* no user size, so use the full block device */
2474 cfg->dblocks = DTOBT(xi->dsize, cfg->blocklog);
2475 }
2476
2477 if (cfg->dblocks < XFS_MIN_DATA_BLOCKS) {
2478 fprintf(stderr,
2479 _("size %lld of data subvolume is too small, minimum %d blocks\n"),
2480 (long long)cfg->dblocks, XFS_MIN_DATA_BLOCKS);
2481 usage();
2482 }
2483
2484 if (xi->dbsize > cfg->sectorsize) {
2485 fprintf(stderr, _(
2486 "Warning: the data subvolume sector size %u is less than the sector size \n\
2487 reported by the device (%u).\n"),
2488 cfg->sectorsize, xi->dbsize);
2489 }
2490 }
2491
2492 /*
2493 * This is more complex than it needs to be because we still support volume
2494 * based external logs. They are only discovered *after* the devices have been
2495 * opened, hence the crazy "is this really an internal log" checks here.
2496 */
2497 static void
2498 validate_logdev(
2499 struct mkfs_params *cfg,
2500 struct cli_params *cli,
2501 char **devname)
2502 {
2503 struct libxfs_xinit *xi = cli->xi;
2504
2505 *devname = NULL;
2506
2507 /* check for volume log first */
2508 if (cli->loginternal && xi->volname && xi->logdev) {
2509 *devname = _("volume log");
2510 cfg->loginternal = false;
2511 } else
2512 cfg->loginternal = cli->loginternal;
2513
2514 /* now run device checks */
2515 if (cfg->loginternal) {
2516 if (xi->logdev) {
2517 fprintf(stderr,
2518 _("can't have both external and internal logs\n"));
2519 usage();
2520 }
2521
2522 /*
2523 * if no sector size has been specified on the command line,
2524 * use what has been configured and validated for the data
2525 * device.
2526 */
2527 if (!cli->lsectorsize) {
2528 cfg->lsectorsize = cfg->sectorsize;
2529 cfg->lsectorlog = cfg->sectorlog;
2530 }
2531
2532 if (cfg->sectorsize != cfg->lsectorsize) {
2533 fprintf(stderr,
2534 _("data and log sector sizes must be equal for internal logs\n"));
2535 usage();
2536 }
2537 if (cli->logsize && cfg->logblocks >= cfg->dblocks) {
2538 fprintf(stderr,
2539 _("log size %lld too large for internal log\n"),
2540 (long long)cfg->logblocks);
2541 usage();
2542 }
2543 *devname = _("internal log");
2544 return;
2545 }
2546
2547 /* External/log subvolume checks */
2548 if (xi->logname)
2549 *devname = xi->logname;
2550 if (!*devname || !xi->logdev) {
2551 fprintf(stderr, _("no log subvolume or external log.\n"));
2552 usage();
2553 }
2554
2555 if (!cfg->logblocks) {
2556 if (xi->logBBsize == 0) {
2557 fprintf(stderr,
2558 _("unable to get size of the log subvolume.\n"));
2559 usage();
2560 }
2561 cfg->logblocks = DTOBT(xi->logBBsize, cfg->blocklog);
2562 } else if (cfg->logblocks > DTOBT(xi->logBBsize, cfg->blocklog)) {
2563 fprintf(stderr,
2564 _("size %s specified for log subvolume is too large, maximum is %lld blocks\n"),
2565 cli->logsize,
2566 (long long)DTOBT(xi->logBBsize, cfg->blocklog));
2567 usage();
2568 }
2569
2570 if (xi->lbsize > cfg->lsectorsize) {
2571 fprintf(stderr, _(
2572 "Warning: the log subvolume sector size %u is less than the sector size\n\
2573 reported by the device (%u).\n"),
2574 cfg->lsectorsize, xi->lbsize);
2575 }
2576 }
2577
2578 static void
2579 validate_rtdev(
2580 struct mkfs_params *cfg,
2581 struct cli_params *cli,
2582 char **devname)
2583 {
2584 struct libxfs_xinit *xi = cli->xi;
2585
2586 *devname = NULL;
2587
2588 if (!xi->rtdev) {
2589 if (cli->rtsize) {
2590 fprintf(stderr,
2591 _("size specified for non-existent rt subvolume\n"));
2592 usage();
2593 }
2594
2595 *devname = _("none");
2596 cfg->rtblocks = 0;
2597 cfg->rtextents = 0;
2598 cfg->rtbmblocks = 0;
2599 return;
2600 }
2601 if (!xi->rtsize) {
2602 fprintf(stderr, _("Invalid zero length rt subvolume found\n"));
2603 usage();
2604 }
2605
2606 /* volume rtdev */
2607 if (xi->volname)
2608 *devname = _("volume rt");
2609 else
2610 *devname = xi->rtname;
2611
2612 if (cli->rtsize) {
2613 if (cfg->rtblocks > DTOBT(xi->rtsize, cfg->blocklog)) {
2614 fprintf(stderr,
2615 _("size %s specified for rt subvolume is too large, maxi->um is %lld blocks\n"),
2616 cli->rtsize,
2617 (long long)DTOBT(xi->rtsize, cfg->blocklog));
2618 usage();
2619 }
2620 if (xi->rtbsize > cfg->sectorsize) {
2621 fprintf(stderr, _(
2622 "Warning: the realtime subvolume sector size %u is less than the sector size\n\
2623 reported by the device (%u).\n"),
2624 cfg->sectorsize, xi->rtbsize);
2625 }
2626 } else {
2627 /* grab volume size */
2628 cfg->rtblocks = DTOBT(xi->rtsize, cfg->blocklog);
2629 }
2630
2631 cfg->rtextents = cfg->rtblocks / cfg->rtextblocks;
2632 cfg->rtbmblocks = (xfs_extlen_t)howmany(cfg->rtextents,
2633 NBBY * cfg->blocksize);
2634 }
2635
2636 static void
2637 calculate_initial_ag_geometry(
2638 struct mkfs_params *cfg,
2639 struct cli_params *cli)
2640 {
2641 if (cli->agsize) { /* User-specified AG size */
2642 cfg->agsize = getnum(cli->agsize, &dopts, D_AGSIZE);
2643
2644 /*
2645 * Check specified agsize is a multiple of blocksize.
2646 */
2647 if (cfg->agsize % cfg->blocksize) {
2648 fprintf(stderr,
2649 _("agsize (%s) not a multiple of fs blk size (%d)\n"),
2650 cli->agsize, cfg->blocksize);
2651 usage();
2652 }
2653 cfg->agsize /= cfg->blocksize;
2654 cfg->agcount = cfg->dblocks / cfg->agsize +
2655 (cfg->dblocks % cfg->agsize != 0);
2656
2657 } else if (cli->agcount) { /* User-specified AG count */
2658 cfg->agcount = cli->agcount;
2659 cfg->agsize = cfg->dblocks / cfg->agcount +
2660 (cfg->dblocks % cfg->agcount != 0);
2661 } else {
2662 calc_default_ag_geometry(cfg->blocklog, cfg->dblocks,
2663 cfg->dsunit, &cfg->agsize,
2664 &cfg->agcount);
2665 }
2666 }
2667
2668 /*
2669 * Align the AG size to stripe geometry. If this fails and we are using
2670 * discovered stripe geometry, tell the caller to clear the stripe geometry.
2671 * Otherwise, set the aligned geometry (valid or invalid!) so that the
2672 * validation call will fail and exit.
2673 */
2674 static void
2675 align_ag_geometry(
2676 struct mkfs_params *cfg)
2677 {
2678 uint64_t tmp_agsize;
2679 int dsunit = cfg->dsunit;
2680
2681 if (!dsunit)
2682 goto validate;
2683
2684 /*
2685 * agsize is not a multiple of dsunit
2686 */
2687 if ((cfg->agsize % dsunit) != 0) {
2688 /*
2689 * Round up to stripe unit boundary. Also make sure
2690 * that agsize is still larger than
2691 * XFS_AG_MIN_BLOCKS(blocklog)
2692 */
2693 tmp_agsize = ((cfg->agsize + dsunit - 1) / dsunit) * dsunit;
2694 /*
2695 * Round down to stripe unit boundary if rounding up
2696 * created an AG size that is larger than the AG max.
2697 */
2698 if (tmp_agsize > XFS_AG_MAX_BLOCKS(cfg->blocklog))
2699 tmp_agsize = (cfg->agsize / dsunit) * dsunit;
2700
2701 if (tmp_agsize < XFS_AG_MIN_BLOCKS(cfg->blocklog) &&
2702 tmp_agsize > XFS_AG_MAX_BLOCKS(cfg->blocklog)) {
2703
2704 /*
2705 * If the AG size is invalid and we are using device
2706 * probed stripe alignment, just clear the alignment
2707 * and continue on.
2708 */
2709 if (!cli_opt_set(&dopts, D_SUNIT) &&
2710 !cli_opt_set(&dopts, D_SU)) {
2711 cfg->dsunit = 0;
2712 cfg->dswidth = 0;
2713 goto validate;
2714 }
2715 /*
2716 * set the agsize to the invalid value so the following
2717 * validation of the ag will fail and print a nice error
2718 * and exit.
2719 */
2720 cfg->agsize = tmp_agsize;
2721 goto validate;
2722 }
2723
2724 /* update geometry to be stripe unit aligned */
2725 cfg->agsize = tmp_agsize;
2726 if (!cli_opt_set(&dopts, D_AGCOUNT))
2727 cfg->agcount = cfg->dblocks / cfg->agsize +
2728 (cfg->dblocks % cfg->agsize != 0);
2729 if (cli_opt_set(&dopts, D_AGSIZE))
2730 fprintf(stderr,
2731 _("agsize rounded to %lld, sunit = %d\n"),
2732 (long long)cfg->agsize, dsunit);
2733 }
2734
2735 if ((cfg->agsize % cfg->dswidth) == 0 &&
2736 cfg->dswidth != cfg->dsunit &&
2737 cfg->agcount > 1) {
2738
2739 if (cli_opt_set(&dopts, D_AGCOUNT) ||
2740 cli_opt_set(&dopts, D_AGSIZE)) {
2741 fprintf(stderr, _(
2742 "Warning: AG size is a multiple of stripe width. This can cause performance\n\
2743 problems by aligning all AGs on the same disk. To avoid this, run mkfs with\n\
2744 an AG size that is one stripe unit smaller or larger, for example %llu.\n"),
2745 (unsigned long long)cfg->agsize - dsunit);
2746 goto validate;
2747 }
2748
2749 /*
2750 * This is a non-optimal configuration because all AGs start on
2751 * the same disk in the stripe. Changing the AG size by one
2752 * sunit will guarantee that this does not happen.
2753 */
2754 tmp_agsize = cfg->agsize - dsunit;
2755 if (tmp_agsize < XFS_AG_MIN_BLOCKS(cfg->blocklog)) {
2756 tmp_agsize = cfg->agsize + dsunit;
2757 if (cfg->dblocks < cfg->agsize) {
2758 /* oh well, nothing to do */
2759 tmp_agsize = cfg->agsize;
2760 }
2761 }
2762
2763 cfg->agsize = tmp_agsize;
2764 cfg->agcount = cfg->dblocks / cfg->agsize +
2765 (cfg->dblocks % cfg->agsize != 0);
2766 }
2767
2768 validate:
2769 /*
2770 * If the last AG is too small, reduce the filesystem size
2771 * and drop the blocks.
2772 */
2773 if (cfg->dblocks % cfg->agsize != 0 &&
2774 (cfg->dblocks % cfg->agsize < XFS_AG_MIN_BLOCKS(cfg->blocklog))) {
2775 ASSERT(!cli_opt_set(&dopts, D_AGCOUNT));
2776 cfg->dblocks = (xfs_rfsblock_t)((cfg->agcount - 1) * cfg->agsize);
2777 cfg->agcount--;
2778 ASSERT(cfg->agcount != 0);
2779 }
2780
2781 validate_ag_geometry(cfg->blocklog, cfg->dblocks,
2782 cfg->agsize, cfg->agcount);
2783 }
2784
2785 static void
2786 calculate_imaxpct(
2787 struct mkfs_params *cfg,
2788 struct cli_params *cli)
2789 {
2790 cfg->imaxpct = cli->imaxpct;
2791 if (cfg->imaxpct)
2792 return;
2793
2794 /*
2795 * This returns the % of the disk space that is used for
2796 * inodes, it changes relatively to the FS size:
2797 * - over 50 TB, use 1%,
2798 * - 1TB - 50 TB, use 5%,
2799 * - under 1 TB, use XFS_DFL_IMAXIMUM_PCT (25%).
2800 */
2801
2802 if (cfg->dblocks < TERABYTES(1, cfg->blocklog))
2803 cfg->imaxpct = XFS_DFL_IMAXIMUM_PCT;
2804 else if (cfg->dblocks < TERABYTES(50, cfg->blocklog))
2805 cfg->imaxpct = 5;
2806 else
2807 cfg->imaxpct = 1;
2808 }
2809
2810 /*
2811 * Set up the initial state of the superblock so we can start using the
2812 * libxfs geometry macros.
2813 */
2814 static void
2815 sb_set_features(
2816 struct mkfs_params *cfg,
2817 struct xfs_sb *sbp)
2818 {
2819 struct sb_feat_args *fp = &cfg->sb_feat;
2820
2821 sbp->sb_versionnum = XFS_DFL_SB_VERSION_BITS;
2822 if (fp->crcs_enabled)
2823 sbp->sb_versionnum |= XFS_SB_VERSION_5;
2824 else
2825 sbp->sb_versionnum |= XFS_SB_VERSION_4;
2826
2827 if (fp->inode_align) {
2828 int cluster_size = XFS_INODE_BIG_CLUSTER_SIZE;
2829
2830 sbp->sb_versionnum |= XFS_SB_VERSION_ALIGNBIT;
2831 if (cfg->sb_feat.crcs_enabled)
2832 cluster_size *= cfg->inodesize / XFS_DINODE_MIN_SIZE;
2833 sbp->sb_inoalignmt = cluster_size >> cfg->blocklog;
2834 } else
2835 sbp->sb_inoalignmt = 0;
2836
2837 if (cfg->dsunit)
2838 sbp->sb_versionnum |= XFS_SB_VERSION_DALIGNBIT;
2839 if (fp->log_version == 2)
2840 sbp->sb_versionnum |= XFS_SB_VERSION_LOGV2BIT;
2841 if (fp->attr_version == 1)
2842 sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
2843 if (fp->nci)
2844 sbp->sb_versionnum |= XFS_SB_VERSION_BORGBIT;
2845
2846 if (cfg->sectorsize > BBSIZE || cfg->lsectorsize > BBSIZE) {
2847 sbp->sb_versionnum |= XFS_SB_VERSION_SECTORBIT;
2848 sbp->sb_logsectlog = (uint8_t)cfg->lsectorlog;
2849 sbp->sb_logsectsize = (uint16_t)cfg->lsectorsize;
2850 } else {
2851 sbp->sb_logsectlog = 0;
2852 sbp->sb_logsectsize = 0;
2853 }
2854
2855 sbp->sb_features2 = 0;
2856 if (fp->lazy_sb_counters)
2857 sbp->sb_features2 |= XFS_SB_VERSION2_LAZYSBCOUNTBIT;
2858 if (fp->projid32bit)
2859 sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT;
2860 if (fp->parent_pointers)
2861 sbp->sb_features2 |= XFS_SB_VERSION2_PARENTBIT;
2862 if (fp->crcs_enabled)
2863 sbp->sb_features2 |= XFS_SB_VERSION2_CRCBIT;
2864 if (fp->attr_version == 2)
2865 sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT;
2866
2867 /* v5 superblocks have their own feature bit for dirftype */
2868 if (fp->dirftype && !fp->crcs_enabled)
2869 sbp->sb_features2 |= XFS_SB_VERSION2_FTYPE;
2870
2871 /* update whether extended features are in use */
2872 if (sbp->sb_features2 != 0)
2873 sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
2874
2875 /*
2876 * Due to a structure alignment issue, sb_features2 ended up in one
2877 * of two locations, the second "incorrect" location represented by
2878 * the sb_bad_features2 field. To avoid older kernels mounting
2879 * filesystems they shouldn't, set both field to the same value.
2880 */
2881 sbp->sb_bad_features2 = sbp->sb_features2;
2882
2883 if (!fp->crcs_enabled)
2884 return;
2885
2886 /* default features for v5 filesystems */
2887 sbp->sb_features_compat = 0;
2888 sbp->sb_features_ro_compat = 0;
2889 sbp->sb_features_incompat = XFS_SB_FEAT_INCOMPAT_FTYPE;
2890 sbp->sb_features_log_incompat = 0;
2891
2892 if (fp->finobt)
2893 sbp->sb_features_ro_compat = XFS_SB_FEAT_RO_COMPAT_FINOBT;
2894 if (fp->rmapbt)
2895 sbp->sb_features_ro_compat |= XFS_SB_FEAT_RO_COMPAT_RMAPBT;
2896 if (fp->reflink)
2897 sbp->sb_features_ro_compat |= XFS_SB_FEAT_RO_COMPAT_REFLINK;
2898
2899 /*
2900 * Sparse inode chunk support has two main inode alignment requirements.
2901 * First, sparse chunk alignment must match the cluster size. Second,
2902 * full chunk alignment must match the inode chunk size.
2903 *
2904 * Copy the already calculated/scaled inoalignmt to spino_align and
2905 * update the former to the full inode chunk size.
2906 */
2907 if (fp->spinodes) {
2908 sbp->sb_spino_align = sbp->sb_inoalignmt;
2909 sbp->sb_inoalignmt = XFS_INODES_PER_CHUNK *
2910 cfg->inodesize >> cfg->blocklog;
2911 sbp->sb_features_incompat |= XFS_SB_FEAT_INCOMPAT_SPINODES;
2912 }
2913
2914 }
2915
2916 /*
2917 * Make sure that the log size is a multiple of the stripe unit
2918 */
2919 static void
2920 align_log_size(
2921 struct mkfs_params *cfg,
2922 int sunit)
2923 {
2924 uint64_t tmp_logblocks;
2925
2926 /* nothing to do if it's already aligned. */
2927 if ((cfg->logblocks % sunit) == 0)
2928 return;
2929
2930 if (cli_opt_set(&lopts, L_SIZE)) {
2931 fprintf(stderr,
2932 _("log size %lld is not a multiple of the log stripe unit %d\n"),
2933 (long long) cfg->logblocks, sunit);
2934 usage();
2935 }
2936
2937 tmp_logblocks = ((cfg->logblocks + (sunit - 1)) / sunit) * sunit;
2938
2939 /* If the log is too large, round down instead of round up */
2940 if ((tmp_logblocks > XFS_MAX_LOG_BLOCKS) ||
2941 ((tmp_logblocks << cfg->blocklog) > XFS_MAX_LOG_BYTES)) {
2942 tmp_logblocks = (cfg->logblocks / sunit) * sunit;
2943 }
2944 cfg->logblocks = tmp_logblocks;
2945 }
2946
2947 /*
2948 * Make sure that the internal log is correctly aligned to the specified
2949 * stripe unit.
2950 */
2951 static void
2952 align_internal_log(
2953 struct mkfs_params *cfg,
2954 struct xfs_mount *mp,
2955 int sunit)
2956 {
2957 /* round up log start if necessary */
2958 if ((cfg->logstart % sunit) != 0)
2959 cfg->logstart = ((cfg->logstart + (sunit - 1)) / sunit) * sunit;
2960
2961 /* round up/down the log size now */
2962 align_log_size(cfg, sunit);
2963
2964 /* check the aligned log still fits in an AG. */
2965 if (cfg->logblocks > cfg->agsize - XFS_FSB_TO_AGBNO(mp, cfg->logstart)) {
2966 fprintf(stderr,
2967 _("Due to stripe alignment, the internal log size (%lld) is too large.\n"
2968 "Must fit within an allocation group.\n"),
2969 (long long) cfg->logblocks);
2970 usage();
2971 }
2972 }
2973
2974 void
2975 validate_log_size(uint64_t logblocks, int blocklog, int min_logblocks)
2976 {
2977 if (logblocks < min_logblocks) {
2978 fprintf(stderr,
2979 _("log size %lld blocks too small, minimum size is %d blocks\n"),
2980 (long long)logblocks, min_logblocks);
2981 usage();
2982 }
2983 if (logblocks > XFS_MAX_LOG_BLOCKS) {
2984 fprintf(stderr,
2985 _("log size %lld blocks too large, maximum size is %lld blocks\n"),
2986 (long long)logblocks, XFS_MAX_LOG_BLOCKS);
2987 usage();
2988 }
2989 if ((logblocks << blocklog) > XFS_MAX_LOG_BYTES) {
2990 fprintf(stderr,
2991 _("log size %lld bytes too large, maximum size is %lld bytes\n"),
2992 (long long)(logblocks << blocklog), XFS_MAX_LOG_BYTES);
2993 usage();
2994 }
2995 }
2996
2997 static void
2998 calculate_log_size(
2999 struct mkfs_params *cfg,
3000 struct cli_params *cli,
3001 struct xfs_mount *mp)
3002 {
3003 struct xfs_sb *sbp = &mp->m_sb;
3004 int min_logblocks;
3005 struct xfs_mount mount;
3006
3007 /* we need a temporary mount to calculate the minimum log size. */
3008 memset(&mount, 0, sizeof(mount));
3009 mount.m_sb = *sbp;
3010 libxfs_mount(&mount, &mp->m_sb, 0, 0, 0, 0);
3011 min_logblocks = libxfs_log_calc_minimum_size(&mount);
3012 libxfs_umount(&mount);
3013
3014 ASSERT(min_logblocks);
3015 min_logblocks = max(XFS_MIN_LOG_BLOCKS, min_logblocks);
3016
3017 /* if we have lots of blocks, check against XFS_MIN_LOG_BYTES, too */
3018 if (!cli->logsize &&
3019 cfg->dblocks >= (1024*1024*1024) >> cfg->blocklog)
3020 min_logblocks = max(min_logblocks,
3021 XFS_MIN_LOG_BYTES >> cfg->blocklog);
3022
3023 /*
3024 * external logs will have a device and size by now, so all we have
3025 * to do is validate it against minimum size and align it.
3026 */
3027 if (!cfg->loginternal) {
3028 if (min_logblocks > cfg->logblocks) {
3029 fprintf(stderr,
3030 _("external log device %lld too small, must be at least %lld blocks\n"),
3031 (long long)cfg->logblocks,
3032 (long long)min_logblocks);
3033 usage();
3034 }
3035 cfg->logstart = 0;
3036 cfg->logagno = 0;
3037 if (cfg->lsunit)
3038 align_log_size(cfg, cfg->lsunit);
3039
3040 validate_log_size(cfg->logblocks, cfg->blocklog, min_logblocks);
3041 return;
3042 }
3043
3044 /* internal log - if no size specified, calculate automatically */
3045 if (!cfg->logblocks) {
3046 if (cfg->dblocks < GIGABYTES(1, cfg->blocklog)) {
3047 /* tiny filesystems get minimum sized logs. */
3048 cfg->logblocks = min_logblocks;
3049 } else if (cfg->dblocks < GIGABYTES(16, cfg->blocklog)) {
3050
3051 /*
3052 * For small filesystems, we want to use the
3053 * XFS_MIN_LOG_BYTES for filesystems smaller than 16G if
3054 * at all possible, ramping up to 128MB at 256GB.
3055 */
3056 cfg->logblocks = min(XFS_MIN_LOG_BYTES >> cfg->blocklog,
3057 min_logblocks * XFS_DFL_LOG_FACTOR);
3058 } else {
3059 /*
3060 * With a 2GB max log size, default to maximum size
3061 * at 4TB. This keeps the same ratio from the older
3062 * max log size of 128M at 256GB fs size. IOWs,
3063 * the ratio of fs size to log size is 2048:1.
3064 */
3065 cfg->logblocks = (cfg->dblocks << cfg->blocklog) / 2048;
3066 cfg->logblocks = cfg->logblocks >> cfg->blocklog;
3067 }
3068
3069 /* Ensure the chosen size meets minimum log size requirements */
3070 cfg->logblocks = max(min_logblocks, cfg->logblocks);
3071
3072 /*
3073 * Make sure the log fits wholly within an AG
3074 *
3075 * XXX: If agf->freeblks ends up as 0 because the log uses all
3076 * the free space, it causes the kernel all sorts of problems
3077 * with per-ag reservations. Right now just back it off one
3078 * block, but there's a whole can of worms here that needs to be
3079 * opened to decide what is the valid maximum size of a log in
3080 * an AG.
3081 */
3082 cfg->logblocks = min(cfg->logblocks,
3083 libxfs_alloc_ag_max_usable(mp) - 1);
3084
3085 /* and now clamp the size to the maximum supported size */
3086 cfg->logblocks = min(cfg->logblocks, XFS_MAX_LOG_BLOCKS);
3087 if ((cfg->logblocks << cfg->blocklog) > XFS_MAX_LOG_BYTES)
3088 cfg->logblocks = XFS_MAX_LOG_BYTES >> cfg->blocklog;
3089
3090 validate_log_size(cfg->logblocks, cfg->blocklog, min_logblocks);
3091 }
3092
3093 if (cfg->logblocks > sbp->sb_agblocks - libxfs_prealloc_blocks(mp)) {
3094 fprintf(stderr,
3095 _("internal log size %lld too large, must fit in allocation group\n"),
3096 (long long)cfg->logblocks);
3097 usage();
3098 }
3099
3100 if (cli_opt_set(&lopts, L_AGNUM)) {
3101 if (cli->logagno >= sbp->sb_agcount) {
3102 fprintf(stderr,
3103 _("log ag number %lld too large, must be less than %lld\n"),
3104 (long long)cli->logagno,
3105 (long long)sbp->sb_agcount);
3106 usage();
3107 }
3108 cfg->logagno = cli->logagno;
3109 } else
3110 cfg->logagno = (xfs_agnumber_t)(sbp->sb_agcount / 2);
3111
3112 cfg->logstart = XFS_AGB_TO_FSB(mp, cfg->logagno,
3113 libxfs_prealloc_blocks(mp));
3114
3115 /*
3116 * Align the logstart at stripe unit boundary.
3117 */
3118 if (cfg->lsunit) {
3119 align_internal_log(cfg, mp, cfg->lsunit);
3120 } else if (cfg->dsunit) {
3121 align_internal_log(cfg, mp, cfg->dsunit);
3122 }
3123 validate_log_size(cfg->logblocks, cfg->blocklog, min_logblocks);
3124 }
3125
3126 /*
3127 * Set up superblock with the minimum parameters required for
3128 * the libxfs macros needed by the log sizing code to run successfully.
3129 * This includes a minimum log size calculation, so we need everything
3130 * that goes into that calculation to be setup here including feature
3131 * flags.
3132 */
3133 static void
3134 start_superblock_setup(
3135 struct mkfs_params *cfg,
3136 struct xfs_mount *mp,
3137 struct xfs_sb *sbp)
3138 {
3139 sbp->sb_magicnum = XFS_SB_MAGIC;
3140 sbp->sb_sectsize = (uint16_t)cfg->sectorsize;
3141 sbp->sb_sectlog = (uint8_t)cfg->sectorlog;
3142 sbp->sb_blocksize = cfg->blocksize;
3143 sbp->sb_blocklog = (uint8_t)cfg->blocklog;
3144
3145 sbp->sb_agblocks = (xfs_agblock_t)cfg->agsize;
3146 sbp->sb_agblklog = (uint8_t)log2_roundup(cfg->agsize);
3147 sbp->sb_agcount = (xfs_agnumber_t)cfg->agcount;
3148
3149 sbp->sb_inodesize = (uint16_t)cfg->inodesize;
3150 sbp->sb_inodelog = (uint8_t)cfg->inodelog;
3151 sbp->sb_inopblock = (uint16_t)(cfg->blocksize / cfg->inodesize);
3152 sbp->sb_inopblog = (uint8_t)(cfg->blocklog - cfg->inodelog);
3153
3154 sbp->sb_dirblklog = cfg->dirblocklog - cfg->blocklog;
3155
3156 sb_set_features(cfg, sbp);
3157
3158 /*
3159 * log stripe unit is stored in bytes on disk and cannot be zero
3160 * for v2 logs.
3161 */
3162 if (cfg->sb_feat.log_version == 2) {
3163 if (cfg->lsunit)
3164 sbp->sb_logsunit = XFS_FSB_TO_B(mp, cfg->lsunit);
3165 else
3166 sbp->sb_logsunit = 1;
3167 } else
3168 sbp->sb_logsunit = 0;
3169
3170 }
3171
3172 static void
3173 initialise_mount(
3174 struct mkfs_params *cfg,
3175 struct xfs_mount *mp,
3176 struct xfs_sb *sbp)
3177 {
3178 /* Minimum needed for libxfs_prealloc_blocks() */
3179 mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
3180 mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
3181 }
3182
3183 /*
3184 * Format everything from the generated config into the superblock that
3185 * will be used to initialise the on-disk superblock. This is the in-memory
3186 * copy, so no need to care about endian swapping here.
3187 */
3188 static void
3189 finish_superblock_setup(
3190 struct mkfs_params *cfg,
3191 struct xfs_mount *mp,
3192 struct xfs_sb *sbp)
3193 {
3194 if (cfg->label)
3195 strncpy(sbp->sb_fname, cfg->label, sizeof(sbp->sb_fname));
3196
3197 sbp->sb_dblocks = cfg->dblocks;
3198 sbp->sb_rblocks = cfg->rtblocks;
3199 sbp->sb_rextents = cfg->rtextents;
3200 platform_uuid_copy(&sbp->sb_uuid, &cfg->uuid);
3201 /* Only in memory; libxfs expects this as if read from disk */
3202 platform_uuid_copy(&sbp->sb_meta_uuid, &cfg->uuid);
3203 sbp->sb_logstart = cfg->logstart;
3204 sbp->sb_rootino = sbp->sb_rbmino = sbp->sb_rsumino = NULLFSINO;
3205 sbp->sb_rextsize = cfg->rtextblocks;
3206 sbp->sb_agcount = (xfs_agnumber_t)cfg->agcount;
3207 sbp->sb_rbmblocks = cfg->rtbmblocks;
3208 sbp->sb_logblocks = (xfs_extlen_t)cfg->logblocks;
3209 sbp->sb_rextslog = (uint8_t)(cfg->rtextents ?
3210 libxfs_highbit32((unsigned int)cfg->rtextents) : 0);
3211 sbp->sb_inprogress = 1; /* mkfs is in progress */
3212 sbp->sb_imax_pct = cfg->imaxpct;
3213 sbp->sb_icount = 0;
3214 sbp->sb_ifree = 0;
3215 sbp->sb_fdblocks = cfg->dblocks -
3216 cfg->agcount * libxfs_prealloc_blocks(mp) -
3217 (cfg->loginternal ? cfg->logblocks : 0);
3218 sbp->sb_frextents = 0; /* will do a free later */
3219 sbp->sb_uquotino = sbp->sb_gquotino = sbp->sb_pquotino = 0;
3220 sbp->sb_qflags = 0;
3221 sbp->sb_unit = cfg->dsunit;
3222 sbp->sb_width = cfg->dswidth;
3223
3224 }
3225
3226 /*
3227 * Sanitise the data and log devices and prepare them so libxfs can mount the
3228 * device successfully. Also check we can access the rt device if configured.
3229 */
3230 static void
3231 prepare_devices(
3232 struct mkfs_params *cfg,
3233 struct libxfs_xinit *xi,
3234 struct xfs_mount *mp,
3235 struct xfs_sb *sbp,
3236 bool clear_stale)
3237 {
3238 struct xfs_buf *buf;
3239 int whack_blks = BTOBB(WHACK_SIZE);
3240 int lsunit;
3241
3242 /*
3243 * If there's an old XFS filesystem on the device with enough intact
3244 * information that we can parse the superblock, there's enough
3245 * information on disk to confuse a future xfs_repair call. To avoid
3246 * this, whack all the old secondary superblocks that we can find.
3247 */
3248 if (clear_stale)
3249 zero_old_xfs_structures(xi, sbp);
3250
3251 /*
3252 * If the data device is a file, grow out the file to its final size if
3253 * needed so that the reads for the end of the device in the mount code
3254 * will succeed.
3255 */
3256 if (xi->disfile &&
3257 xi->dsize * xi->dbsize < cfg->dblocks * cfg->blocksize) {
3258 if (ftruncate(xi->dfd, cfg->dblocks * cfg->blocksize) < 0) {
3259 fprintf(stderr,
3260 _("%s: Growing the data section failed\n"),
3261 progname);
3262 exit(1);
3263 }
3264
3265 /* update size to be able to whack blocks correctly */
3266 xi->dsize = BTOBB(cfg->dblocks * cfg->blocksize);
3267 }
3268
3269 /*
3270 * Zero out the end to obliterate any old MD RAID (or other) metadata at
3271 * the end of the device. (MD sb is ~64k from the end, take out a wider
3272 * swath to be sure)
3273 */
3274 buf = libxfs_getbuf(mp->m_ddev_targp, (xi->dsize - whack_blks),
3275 whack_blks);
3276 memset(buf->b_addr, 0, WHACK_SIZE);
3277 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3278 libxfs_purgebuf(buf);
3279
3280 /*
3281 * Now zero out the beginning of the device, to obliterate any old
3282 * filesystem signatures out there. This should take care of
3283 * swap (somewhere around the page size), jfs (32k),
3284 * ext[2,3] and reiserfs (64k) - and hopefully all else.
3285 */
3286 buf = libxfs_getbuf(mp->m_ddev_targp, 0, whack_blks);
3287 memset(buf->b_addr, 0, WHACK_SIZE);
3288 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3289 libxfs_purgebuf(buf);
3290
3291 /* OK, now write the superblock... */
3292 buf = libxfs_getbuf(mp->m_ddev_targp, XFS_SB_DADDR, XFS_FSS_TO_BB(mp, 1));
3293 buf->b_ops = &xfs_sb_buf_ops;
3294 memset(buf->b_addr, 0, cfg->sectorsize);
3295 libxfs_sb_to_disk(buf->b_addr, sbp);
3296 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3297 libxfs_purgebuf(buf);
3298
3299 /* ...and zero the log.... */
3300 lsunit = sbp->sb_logsunit;
3301 if (lsunit == 1)
3302 lsunit = sbp->sb_logsectsize;
3303
3304 libxfs_log_clear(mp->m_logdev_targp, NULL,
3305 XFS_FSB_TO_DADDR(mp, cfg->logstart),
3306 (xfs_extlen_t)XFS_FSB_TO_BB(mp, cfg->logblocks),
3307 &sbp->sb_uuid, cfg->sb_feat.log_version,
3308 lsunit, XLOG_FMT, XLOG_INIT_CYCLE, false);
3309
3310 /* finally, check we can write the last block in the realtime area */
3311 if (mp->m_rtdev_targp->dev && cfg->rtblocks > 0) {
3312 buf = libxfs_getbuf(mp->m_rtdev_targp,
3313 XFS_FSB_TO_BB(mp, cfg->rtblocks - 1LL),
3314 BTOBB(cfg->blocksize));
3315 memset(buf->b_addr, 0, cfg->blocksize);
3316 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3317 libxfs_purgebuf(buf);
3318 }
3319
3320 }
3321
3322 /*
3323 * XXX: this code is mostly common with the kernel growfs code.
3324 * These initialisations should be pulled into libxfs to keep the
3325 * kernel/userspace header initialisation code the same.
3326 */
3327 static void
3328 initialise_ag_headers(
3329 struct mkfs_params *cfg,
3330 struct xfs_mount *mp,
3331 struct xfs_sb *sbp,
3332 xfs_agnumber_t agno,
3333 int *worst_freelist)
3334 {
3335 struct xfs_perag *pag = libxfs_perag_get(mp, agno);
3336 struct xfs_agfl *agfl;
3337 struct xfs_agf *agf;
3338 struct xfs_agi *agi;
3339 struct xfs_buf *buf;
3340 struct xfs_btree_block *block;
3341 struct xfs_alloc_rec *arec;
3342 struct xfs_alloc_rec *nrec;
3343 int bucket;
3344 uint64_t agsize = cfg->agsize;
3345 xfs_agblock_t agblocks;
3346 bool is_log_ag = false;
3347 int c;
3348
3349 if (cfg->loginternal && agno == cfg->logagno)
3350 is_log_ag = true;
3351
3352 /*
3353 * Superblock.
3354 */
3355 buf = libxfs_getbuf(mp->m_ddev_targp,
3356 XFS_AG_DADDR(mp, agno, XFS_SB_DADDR),
3357 XFS_FSS_TO_BB(mp, 1));
3358 buf->b_ops = &xfs_sb_buf_ops;
3359 memset(buf->b_addr, 0, cfg->sectorsize);
3360 libxfs_sb_to_disk(buf->b_addr, sbp);
3361 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3362
3363 /*
3364 * AG header block: freespace
3365 */
3366 buf = libxfs_getbuf(mp->m_ddev_targp,
3367 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
3368 XFS_FSS_TO_BB(mp, 1));
3369 buf->b_ops = &xfs_agf_buf_ops;
3370 agf = XFS_BUF_TO_AGF(buf);
3371 memset(agf, 0, cfg->sectorsize);
3372 if (agno == cfg->agcount - 1)
3373 agsize = cfg->dblocks - (xfs_rfsblock_t)(agno * agsize);
3374 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
3375 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
3376 agf->agf_seqno = cpu_to_be32(agno);
3377 agf->agf_length = cpu_to_be32(agsize);
3378 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
3379 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
3380 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
3381 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
3382 pag->pagf_levels[XFS_BTNUM_BNOi] = 1;
3383 pag->pagf_levels[XFS_BTNUM_CNTi] = 1;
3384
3385 if (xfs_sb_version_hasrmapbt(sbp)) {
3386 agf->agf_roots[XFS_BTNUM_RMAPi] = cpu_to_be32(XFS_RMAP_BLOCK(mp));
3387 agf->agf_levels[XFS_BTNUM_RMAPi] = cpu_to_be32(1);
3388 agf->agf_rmap_blocks = cpu_to_be32(1);
3389 }
3390
3391 if (xfs_sb_version_hasreflink(sbp)) {
3392 agf->agf_refcount_root = cpu_to_be32(libxfs_refc_block(mp));
3393 agf->agf_refcount_level = cpu_to_be32(1);
3394 agf->agf_refcount_blocks = cpu_to_be32(1);
3395 }
3396
3397 agf->agf_flfirst = 0;
3398 agf->agf_fllast = cpu_to_be32(libxfs_agfl_size(mp) - 1);
3399 agf->agf_flcount = 0;
3400 agblocks = (xfs_agblock_t)(agsize - libxfs_prealloc_blocks(mp));
3401 agf->agf_freeblks = cpu_to_be32(agblocks);
3402 agf->agf_longest = cpu_to_be32(agblocks);
3403
3404 if (xfs_sb_version_hascrc(sbp))
3405 platform_uuid_copy(&agf->agf_uuid, &sbp->sb_uuid);
3406
3407 if (is_log_ag) {
3408 be32_add_cpu(&agf->agf_freeblks, -(int64_t)cfg->logblocks);
3409 agf->agf_longest = cpu_to_be32(agsize -
3410 XFS_FSB_TO_AGBNO(mp, cfg->logstart) - cfg->logblocks);
3411 }
3412 if (libxfs_alloc_min_freelist(mp, pag) > *worst_freelist)
3413 *worst_freelist = libxfs_alloc_min_freelist(mp, pag);
3414 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3415
3416 /*
3417 * AG freelist header block
3418 */
3419 buf = libxfs_getbuf(mp->m_ddev_targp,
3420 XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
3421 XFS_FSS_TO_BB(mp, 1));
3422 buf->b_ops = &xfs_agfl_buf_ops;
3423 agfl = XFS_BUF_TO_AGFL(buf);
3424 /* setting to 0xff results in initialisation to NULLAGBLOCK */
3425 memset(agfl, 0xff, cfg->sectorsize);
3426 if (xfs_sb_version_hascrc(sbp)) {
3427 agfl->agfl_magicnum = cpu_to_be32(XFS_AGFL_MAGIC);
3428 agfl->agfl_seqno = cpu_to_be32(agno);
3429 platform_uuid_copy(&agfl->agfl_uuid, &sbp->sb_uuid);
3430 for (bucket = 0; bucket < libxfs_agfl_size(mp); bucket++)
3431 agfl->agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK);
3432 }
3433
3434 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3435
3436 /*
3437 * AG header block: inodes
3438 */
3439 buf = libxfs_getbuf(mp->m_ddev_targp,
3440 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
3441 XFS_FSS_TO_BB(mp, 1));
3442 agi = XFS_BUF_TO_AGI(buf);
3443 buf->b_ops = &xfs_agi_buf_ops;
3444 memset(agi, 0, cfg->sectorsize);
3445 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
3446 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
3447 agi->agi_seqno = cpu_to_be32(agno);
3448 agi->agi_length = cpu_to_be32(agsize);
3449 agi->agi_count = 0;
3450 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
3451 agi->agi_level = cpu_to_be32(1);
3452 if (xfs_sb_version_hasfinobt(sbp)) {
3453 agi->agi_free_root = cpu_to_be32(XFS_FIBT_BLOCK(mp));
3454 agi->agi_free_level = cpu_to_be32(1);
3455 }
3456 agi->agi_freecount = 0;
3457 agi->agi_newino = cpu_to_be32(NULLAGINO);
3458 agi->agi_dirino = cpu_to_be32(NULLAGINO);
3459 if (xfs_sb_version_hascrc(sbp))
3460 platform_uuid_copy(&agi->agi_uuid, &sbp->sb_uuid);
3461 for (c = 0; c < XFS_AGI_UNLINKED_BUCKETS; c++)
3462 agi->agi_unlinked[c] = cpu_to_be32(NULLAGINO);
3463 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3464
3465 /*
3466 * BNO btree root block
3467 */
3468 buf = libxfs_getbuf(mp->m_ddev_targp,
3469 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
3470 BTOBB(cfg->blocksize));
3471 buf->b_ops = &xfs_allocbt_buf_ops;
3472 block = XFS_BUF_TO_BLOCK(buf);
3473 memset(block, 0, cfg->blocksize);
3474 libxfs_btree_init_block(mp, buf, XFS_BTNUM_BNO, 0, 1, agno, 0);
3475
3476 arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
3477 arec->ar_startblock = cpu_to_be32(libxfs_prealloc_blocks(mp));
3478 if (is_log_ag) {
3479 xfs_agblock_t start = XFS_FSB_TO_AGBNO(mp, cfg->logstart);
3480
3481 ASSERT(start >= libxfs_prealloc_blocks(mp));
3482 if (start != libxfs_prealloc_blocks(mp)) {
3483 /*
3484 * Modify first record to pad stripe align of log
3485 */
3486 arec->ar_blockcount = cpu_to_be32(start -
3487 libxfs_prealloc_blocks(mp));
3488 nrec = arec + 1;
3489 /*
3490 * Insert second record at start of internal log
3491 * which then gets trimmed.
3492 */
3493 nrec->ar_startblock = cpu_to_be32(
3494 be32_to_cpu(arec->ar_startblock) +
3495 be32_to_cpu(arec->ar_blockcount));
3496 arec = nrec;
3497 be16_add_cpu(&block->bb_numrecs, 1);
3498 }
3499 /*
3500 * Change record start to after the internal log
3501 */
3502 be32_add_cpu(&arec->ar_startblock, cfg->logblocks);
3503 }
3504 /*
3505 * Calculate the record block count and check for the case where
3506 * the log might have consumed all available space in the AG. If
3507 * so, reset the record count to 0 to avoid exposure of an invalid
3508 * record start block.
3509 */
3510 arec->ar_blockcount = cpu_to_be32(agsize -
3511 be32_to_cpu(arec->ar_startblock));
3512 if (!arec->ar_blockcount)
3513 block->bb_numrecs = 0;
3514
3515 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3516
3517 /*
3518 * CNT btree root block
3519 */
3520 buf = libxfs_getbuf(mp->m_ddev_targp,
3521 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
3522 BTOBB(cfg->blocksize));
3523 buf->b_ops = &xfs_allocbt_buf_ops;
3524 block = XFS_BUF_TO_BLOCK(buf);
3525 memset(block, 0, cfg->blocksize);
3526 libxfs_btree_init_block(mp, buf, XFS_BTNUM_CNT, 0, 1, agno, 0);
3527
3528 arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
3529 arec->ar_startblock = cpu_to_be32(libxfs_prealloc_blocks(mp));
3530 if (is_log_ag) {
3531 xfs_agblock_t start = XFS_FSB_TO_AGBNO(mp, cfg->logstart);
3532
3533 ASSERT(start >= libxfs_prealloc_blocks(mp));
3534 if (start != libxfs_prealloc_blocks(mp)) {
3535 arec->ar_blockcount = cpu_to_be32(start -
3536 libxfs_prealloc_blocks(mp));
3537 nrec = arec + 1;
3538 nrec->ar_startblock = cpu_to_be32(
3539 be32_to_cpu(arec->ar_startblock) +
3540 be32_to_cpu(arec->ar_blockcount));
3541 arec = nrec;
3542 be16_add_cpu(&block->bb_numrecs, 1);
3543 }
3544 be32_add_cpu(&arec->ar_startblock, cfg->logblocks);
3545 }
3546 /*
3547 * Calculate the record block count and check for the case where
3548 * the log might have consumed all available space in the AG. If
3549 * so, reset the record count to 0 to avoid exposure of an invalid
3550 * record start block.
3551 */
3552 arec->ar_blockcount = cpu_to_be32(agsize -
3553 be32_to_cpu(arec->ar_startblock));
3554 if (!arec->ar_blockcount)
3555 block->bb_numrecs = 0;
3556
3557 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3558
3559 /*
3560 * refcount btree root block
3561 */
3562 if (xfs_sb_version_hasreflink(sbp)) {
3563 buf = libxfs_getbuf(mp->m_ddev_targp,
3564 XFS_AGB_TO_DADDR(mp, agno, libxfs_refc_block(mp)),
3565 BTOBB(cfg->blocksize));
3566 buf->b_ops = &xfs_refcountbt_buf_ops;
3567
3568 block = XFS_BUF_TO_BLOCK(buf);
3569 memset(block, 0, cfg->blocksize);
3570 libxfs_btree_init_block(mp, buf, XFS_BTNUM_REFC, 0, 0, agno, 0);
3571 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3572 }
3573
3574 /*
3575 * INO btree root block
3576 */
3577 buf = libxfs_getbuf(mp->m_ddev_targp,
3578 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
3579 BTOBB(cfg->blocksize));
3580 buf->b_ops = &xfs_inobt_buf_ops;
3581 block = XFS_BUF_TO_BLOCK(buf);
3582 memset(block, 0, cfg->blocksize);
3583 libxfs_btree_init_block(mp, buf, XFS_BTNUM_INO, 0, 0, agno, 0);
3584 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3585
3586 /*
3587 * Free INO btree root block
3588 */
3589 if (xfs_sb_version_hasfinobt(sbp)) {
3590 buf = libxfs_getbuf(mp->m_ddev_targp,
3591 XFS_AGB_TO_DADDR(mp, agno, XFS_FIBT_BLOCK(mp)),
3592 BTOBB(cfg->blocksize));
3593 buf->b_ops = &xfs_inobt_buf_ops;
3594 block = XFS_BUF_TO_BLOCK(buf);
3595 memset(block, 0, cfg->blocksize);
3596 libxfs_btree_init_block(mp, buf, XFS_BTNUM_FINO, 0, 0, agno, 0);
3597 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3598 }
3599
3600 /* RMAP btree root block */
3601 if (xfs_sb_version_hasrmapbt(sbp)) {
3602 struct xfs_rmap_rec *rrec;
3603
3604 buf = libxfs_getbuf(mp->m_ddev_targp,
3605 XFS_AGB_TO_DADDR(mp, agno, XFS_RMAP_BLOCK(mp)),
3606 BTOBB(cfg->blocksize));
3607 buf->b_ops = &xfs_rmapbt_buf_ops;
3608 block = XFS_BUF_TO_BLOCK(buf);
3609 memset(block, 0, cfg->blocksize);
3610
3611 libxfs_btree_init_block(mp, buf, XFS_BTNUM_RMAP, 0, 0, agno, 0);
3612
3613 /*
3614 * mark the AG header regions as static metadata
3615 * The BNO btree block is the first block after the
3616 * headers, so it's location defines the size of region
3617 * the static metadata consumes.
3618 */
3619 rrec = XFS_RMAP_REC_ADDR(block, 1);
3620 rrec->rm_startblock = 0;
3621 rrec->rm_blockcount = cpu_to_be32(XFS_BNO_BLOCK(mp));
3622 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_FS);
3623 rrec->rm_offset = 0;
3624 be16_add_cpu(&block->bb_numrecs, 1);
3625
3626 /* account freespace btree root blocks */
3627 rrec = XFS_RMAP_REC_ADDR(block, 2);
3628 rrec->rm_startblock = cpu_to_be32(XFS_BNO_BLOCK(mp));
3629 rrec->rm_blockcount = cpu_to_be32(2);
3630 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_AG);
3631 rrec->rm_offset = 0;
3632 be16_add_cpu(&block->bb_numrecs, 1);
3633
3634 /* account inode btree root blocks */
3635 rrec = XFS_RMAP_REC_ADDR(block, 3);
3636 rrec->rm_startblock = cpu_to_be32(XFS_IBT_BLOCK(mp));
3637 rrec->rm_blockcount = cpu_to_be32(XFS_RMAP_BLOCK(mp) -
3638 XFS_IBT_BLOCK(mp));
3639 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_INOBT);
3640 rrec->rm_offset = 0;
3641 be16_add_cpu(&block->bb_numrecs, 1);
3642
3643 /* account for rmap btree root */
3644 rrec = XFS_RMAP_REC_ADDR(block, 4);
3645 rrec->rm_startblock = cpu_to_be32(XFS_RMAP_BLOCK(mp));
3646 rrec->rm_blockcount = cpu_to_be32(1);
3647 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_AG);
3648 rrec->rm_offset = 0;
3649 be16_add_cpu(&block->bb_numrecs, 1);
3650
3651 /* account for refcount btree root */
3652 if (xfs_sb_version_hasreflink(sbp)) {
3653 rrec = XFS_RMAP_REC_ADDR(block, 5);
3654 rrec->rm_startblock = cpu_to_be32(libxfs_refc_block(mp));
3655 rrec->rm_blockcount = cpu_to_be32(1);
3656 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_REFC);
3657 rrec->rm_offset = 0;
3658 be16_add_cpu(&block->bb_numrecs, 1);
3659 }
3660
3661 /* account for the log space */
3662 if (is_log_ag) {
3663 rrec = XFS_RMAP_REC_ADDR(block,
3664 be16_to_cpu(block->bb_numrecs) + 1);
3665 rrec->rm_startblock = cpu_to_be32(
3666 XFS_FSB_TO_AGBNO(mp, cfg->logstart));
3667 rrec->rm_blockcount = cpu_to_be32(cfg->logblocks);
3668 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_LOG);
3669 rrec->rm_offset = 0;
3670 be16_add_cpu(&block->bb_numrecs, 1);
3671 }
3672
3673 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3674 }
3675
3676 libxfs_perag_put(pag);
3677 }
3678
3679 static void
3680 initialise_ag_freespace(
3681 struct xfs_mount *mp,
3682 xfs_agnumber_t agno,
3683 int worst_freelist)
3684 {
3685 struct xfs_alloc_arg args;
3686 struct xfs_trans *tp;
3687 int c;
3688
3689 c = -libxfs_trans_alloc_rollable(mp, worst_freelist, &tp);
3690 if (c)
3691 res_failed(c);
3692
3693 memset(&args, 0, sizeof(args));
3694 args.tp = tp;
3695 args.mp = mp;
3696 args.agno = agno;
3697 args.alignment = 1;
3698 args.pag = libxfs_perag_get(mp, agno);
3699
3700 libxfs_alloc_fix_freelist(&args, 0);
3701 libxfs_perag_put(args.pag);
3702 c = -libxfs_trans_commit(tp);
3703 if (c)
3704 res_failed(c);
3705 }
3706
3707 /*
3708 * rewrite several secondary superblocks with the root inode number filled out.
3709 * This can help repair recovery from a trashed primary superblock without
3710 * losing the root inode.
3711 */
3712 static void
3713 rewrite_secondary_superblocks(
3714 struct xfs_mount *mp)
3715 {
3716 struct xfs_buf *buf;
3717
3718 /* rewrite the last superblock */
3719 buf = libxfs_readbuf(mp->m_dev,
3720 XFS_AGB_TO_DADDR(mp, mp->m_sb.sb_agcount - 1,
3721 XFS_SB_DADDR),
3722 XFS_FSS_TO_BB(mp, 1),
3723 LIBXFS_EXIT_ON_FAILURE, &xfs_sb_buf_ops);
3724 XFS_BUF_TO_SBP(buf)->sb_rootino = cpu_to_be64(mp->m_sb.sb_rootino);
3725 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3726
3727 /* and one in the middle for luck if there's enough AGs for that */
3728 if (mp->m_sb.sb_agcount <= 2)
3729 return;
3730
3731 buf = libxfs_readbuf(mp->m_dev,
3732 XFS_AGB_TO_DADDR(mp, (mp->m_sb.sb_agcount - 1) / 2,
3733 XFS_SB_DADDR),
3734 XFS_FSS_TO_BB(mp, 1),
3735 LIBXFS_EXIT_ON_FAILURE, &xfs_sb_buf_ops);
3736 XFS_BUF_TO_SBP(buf)->sb_rootino = cpu_to_be64(mp->m_sb.sb_rootino);
3737 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
3738 }
3739
3740 int
3741 main(
3742 int argc,
3743 char **argv)
3744 {
3745 xfs_agnumber_t agno;
3746 xfs_buf_t *buf;
3747 int c;
3748 char *dfile = NULL;
3749 char *logfile = NULL;
3750 char *rtfile = NULL;
3751 int dry_run = 0;
3752 int discard = 1;
3753 int force_overwrite = 0;
3754 int quiet = 0;
3755 char *protofile = NULL;
3756 char *protostring = NULL;
3757 int worst_freelist = 0;
3758
3759 struct libxfs_xinit xi = {
3760 .isdirect = LIBXFS_DIRECT,
3761 .isreadonly = LIBXFS_EXCLUSIVELY,
3762 };
3763 struct xfs_mount mbuf = {};
3764 struct xfs_mount *mp = &mbuf;
3765 struct xfs_sb *sbp = &mp->m_sb;
3766 struct fs_topology ft = {};
3767 struct cli_params cli = {
3768 .xi = &xi,
3769 .loginternal = 1,
3770 };
3771 struct mkfs_params cfg = {};
3772
3773 /* build time defaults */
3774 struct mkfs_default_params dft = {
3775 .source = _("package build definitions"),
3776 .sectorsize = XFS_MIN_SECTORSIZE,
3777 .blocksize = 1 << XFS_DFL_BLOCKSIZE_LOG,
3778 .sb_feat = {
3779 .log_version = 2,
3780 .attr_version = 2,
3781 .dir_version = 2,
3782 .inode_align = true,
3783 .nci = false,
3784 .lazy_sb_counters = true,
3785 .projid32bit = true,
3786 .crcs_enabled = true,
3787 .dirftype = true,
3788 .finobt = true,
3789 .spinodes = true,
3790 .rmapbt = false,
3791 .reflink = false,
3792 .parent_pointers = false,
3793 .nodalign = false,
3794 .nortalign = false,
3795 },
3796 };
3797
3798 platform_uuid_generate(&cli.uuid);
3799 progname = basename(argv[0]);
3800 setlocale(LC_ALL, "");
3801 bindtextdomain(PACKAGE, LOCALEDIR);
3802 textdomain(PACKAGE);
3803
3804 /*
3805 * TODO: Sourcing defaults from a config file
3806 *
3807 * Before anything else, see if there's a config file with different
3808 * defaults. If a file exists in <package location>, read in the new
3809 * default values and overwrite them in the &dft structure. This way the
3810 * new defaults will apply before we parse the CLI, and the CLI will
3811 * still be able to override them. When more than one source is
3812 * implemented, emit a message to indicate where the defaults being
3813 * used came from.
3814 *
3815 * printf(_("Default configuration sourced from %s\n"), dft.source);
3816 */
3817
3818 /* copy new defaults into CLI parsing structure */
3819 memcpy(&cli.sb_feat, &dft.sb_feat, sizeof(cli.sb_feat));
3820 memcpy(&cli.fsx, &dft.fsx, sizeof(cli.fsx));
3821
3822 while ((c = getopt(argc, argv, "b:d:i:l:L:m:n:KNp:qr:s:CfV")) != EOF) {
3823 switch (c) {
3824 case 'C':
3825 case 'f':
3826 force_overwrite = 1;
3827 break;
3828 case 'b':
3829 case 'd':
3830 case 'i':
3831 case 'l':
3832 case 'm':
3833 case 'n':
3834 case 'r':
3835 case 's':
3836 parse_subopts(c, optarg, &cli);
3837 break;
3838 case 'L':
3839 if (strlen(optarg) > sizeof(sbp->sb_fname))
3840 illegal(optarg, "L");
3841 cfg.label = optarg;
3842 break;
3843 case 'N':
3844 dry_run = 1;
3845 break;
3846 case 'K':
3847 discard = 0;
3848 break;
3849 case 'p':
3850 if (protofile)
3851 respec('p', NULL, 0);
3852 protofile = optarg;
3853 break;
3854 case 'q':
3855 quiet = 1;
3856 break;
3857 case 'V':
3858 printf(_("%s version %s\n"), progname, VERSION);
3859 exit(0);
3860 case '?':
3861 unknown(optopt, "");
3862 }
3863 }
3864 if (argc - optind > 1) {
3865 fprintf(stderr, _("extra arguments\n"));
3866 usage();
3867 } else if (argc - optind == 1) {
3868 dfile = xi.volname = getstr(argv[optind], &dopts, D_NAME);
3869 } else
3870 dfile = xi.dname;
3871
3872 protostring = setup_proto(protofile);
3873
3874 /*
3875 * Extract as much of the valid config as we can from the CLI input
3876 * before opening the libxfs devices.
3877 */
3878 validate_blocksize(&cfg, &cli, &dft);
3879 validate_sectorsize(&cfg, &cli, &dft, &ft, dfile, dry_run,
3880 force_overwrite);
3881
3882 /*
3883 * XXX: we still need to set block size and sector size global variables
3884 * so that getnum/cvtnum works correctly
3885 */
3886 blocksize = cfg.blocksize;
3887 sectorsize = cfg.sectorsize;
3888
3889 validate_log_sectorsize(&cfg, &cli, &dft);
3890 validate_sb_features(&cfg, &cli);
3891
3892 /*
3893 * we've now completed basic validation of the features, sector and
3894 * block sizes, so from this point onwards we use the values found in
3895 * the cfg structure for them, not the command line structure.
3896 */
3897 validate_dirblocksize(&cfg, &cli);
3898 validate_inodesize(&cfg, &cli);
3899
3900 /*
3901 * if the device size was specified convert it to a block count
3902 * now we have a valid block size. These will be set to zero if
3903 * nothing was specified, indicating we should use the full device.
3904 */
3905 cfg.dblocks = calc_dev_size(cli.dsize, &cfg, &dopts, D_SIZE, "data");
3906 cfg.logblocks = calc_dev_size(cli.logsize, &cfg, &lopts, L_SIZE, "log");
3907 cfg.rtblocks = calc_dev_size(cli.rtsize, &cfg, &ropts, R_SIZE, "rt");
3908
3909 validate_rtextsize(&cfg, &cli, &ft);
3910 calc_stripe_factors(&cfg, &cli, &ft);
3911
3912 /*
3913 * Open and validate the device configurations
3914 */
3915 open_devices(&cfg, &xi);
3916 validate_overwrite(dfile, force_overwrite);
3917 validate_datadev(&cfg, &cli);
3918 validate_logdev(&cfg, &cli, &logfile);
3919 validate_rtdev(&cfg, &cli, &rtfile);
3920
3921 /*
3922 * At this point when know exactly what size all the devices are,
3923 * so we can start validating and calculating layout options that are
3924 * dependent on device sizes. Once calculated, make sure everything
3925 * aligns to device geometry correctly.
3926 */
3927 calculate_initial_ag_geometry(&cfg, &cli);
3928 align_ag_geometry(&cfg);
3929
3930 calculate_imaxpct(&cfg, &cli);
3931
3932 /*
3933 * Set up the basic superblock parameters now so that we can use
3934 * the geometry information we've already validated in libxfs
3935 * provided functions to determine on-disk format information.
3936 */
3937 start_superblock_setup(&cfg, mp, sbp);
3938 initialise_mount(&cfg, mp, sbp);
3939
3940 /*
3941 * With the mount set up, we can finally calculate the log size
3942 * constraints and do default size calculations and final validation
3943 */
3944 calculate_log_size(&cfg, &cli, mp);
3945
3946 finish_superblock_setup(&cfg, mp, sbp);
3947
3948 /* Print the intended geometry of the fs. */
3949 if (!quiet || dry_run) {
3950 struct xfs_fsop_geom geo;
3951 int error;
3952
3953 error = -libxfs_fs_geometry(sbp, &geo,
3954 XFS_FS_GEOM_MAX_STRUCT_VER);
3955 if (error) {
3956 fprintf(stderr,
3957 _("%s: failed to generate filesystem geometry\n"),
3958 progname);
3959 exit(1);
3960 }
3961
3962 xfs_report_geom(&geo, dfile, logfile, rtfile);
3963 if (dry_run)
3964 exit(0);
3965 }
3966
3967 /*
3968 * All values have been validated, discard the old device layout.
3969 */
3970 if (discard && !dry_run)
3971 discard_devices(&xi);
3972
3973 /*
3974 * we need the libxfs buffer cache from here on in.
3975 */
3976 libxfs_buftarg_init(mp, xi.ddev, xi.logdev, xi.rtdev);
3977
3978 /*
3979 * Before we mount the filesystem we need to make sure the devices have
3980 * enough of the filesystem structure on them that allows libxfs to
3981 * mount.
3982 */
3983 prepare_devices(&cfg, &xi, mp, sbp, force_overwrite);
3984 mp = libxfs_mount(mp, sbp, xi.ddev, xi.logdev, xi.rtdev, 0);
3985 if (mp == NULL) {
3986 fprintf(stderr, _("%s: filesystem failed to initialize\n"),
3987 progname);
3988 exit(1);
3989 }
3990
3991 /*
3992 * Initialise all the static on disk metadata.
3993 */
3994 for (agno = 0; agno < cfg.agcount; agno++)
3995 initialise_ag_headers(&cfg, mp, sbp, agno, &worst_freelist);
3996
3997 /*
3998 * Initialise the freespace freelists (i.e. AGFLs) in each AG.
3999 */
4000 for (agno = 0; agno < cfg.agcount; agno++)
4001 initialise_ag_freespace(mp, agno, worst_freelist);
4002
4003 /*
4004 * Allocate the root inode and anything else in the proto file.
4005 */
4006 parse_proto(mp, &cli.fsx, &protostring);
4007
4008 /*
4009 * Protect ourselves against possible stupidity
4010 */
4011 if (XFS_INO_TO_AGNO(mp, mp->m_sb.sb_rootino) != 0) {
4012 fprintf(stderr,
4013 _("%s: root inode created in AG %u, not AG 0\n"),
4014 progname, XFS_INO_TO_AGNO(mp, mp->m_sb.sb_rootino));
4015 exit(1);
4016 }
4017
4018 /*
4019 * Re-write multiple secondary superblocks with rootinode field set
4020 */
4021 if (mp->m_sb.sb_agcount > 1)
4022 rewrite_secondary_superblocks(mp);
4023
4024 /*
4025 * Dump all inodes and buffers before marking us all done.
4026 * Need to drop references to inodes we still hold, first.
4027 */
4028 libxfs_rtmount_destroy(mp);
4029 libxfs_bcache_purge();
4030
4031 /*
4032 * Mark the filesystem ok.
4033 */
4034 buf = libxfs_getsb(mp, LIBXFS_EXIT_ON_FAILURE);
4035 (XFS_BUF_TO_SBP(buf))->sb_inprogress = 0;
4036 libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
4037
4038 libxfs_umount(mp);
4039 if (xi.rtdev)
4040 libxfs_device_close(xi.rtdev);
4041 if (xi.logdev && xi.logdev != xi.ddev)
4042 libxfs_device_close(xi.logdev);
4043 libxfs_device_close(xi.ddev);
4044 libxfs_destroy();
4045
4046 return 0;
4047 }