1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
4 * All Rights Reserved.
5 */
6 #include "xfs.h"
7 #include "xfs_fs.h"
8 #include "xfs_shared.h"
9 #include "xfs_format.h"
10 #include "xfs_log_format.h"
11 #include "xfs_trans_resv.h"
12 #include "xfs_bit.h"
13 #include "xfs_sb.h"
14 #include "xfs_mount.h"
15 #include "xfs_ialloc.h"
16 #include "xfs_alloc.h"
17 #include "xfs_error.h"
18 #include "xfs_trans.h"
19 #include "xfs_buf_item.h"
20 #include "xfs_bmap_btree.h"
21 #include "xfs_alloc_btree.h"
22 #include "xfs_log.h"
23 #include "xfs_rmap_btree.h"
24 #include "xfs_refcount_btree.h"
25 #include "xfs_da_format.h"
26 #include "xfs_health.h"
27 #include "xfs_ag.h"
28 #include "xfs_rtbitmap.h"
29 #include "xfs_exchrange.h"
30 #include "xfs_rtgroup.h"
31 #include "xfs_rtrmap_btree.h"
32 #include "xfs_rtrefcount_btree.h"
33
34 /*
35 * Physical superblock buffer manipulations. Shared with libxfs in userspace.
36 */
37
38 /*
39 * Check that all the V4 feature bits that the V5 filesystem format requires are
40 * correctly set.
41 */
42 static bool
xfs_sb_validate_v5_features(struct xfs_sb * sbp)43 xfs_sb_validate_v5_features(
44 struct xfs_sb *sbp)
45 {
46 /* We must not have any unknown V4 feature bits set */
47 if (sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS)
48 return false;
49
50 /*
51 * The CRC bit is considered an invalid V4 flag, so we have to add it
52 * manually to the OKBITS mask.
53 */
54 if (sbp->sb_features2 & ~(XFS_SB_VERSION2_OKBITS |
55 XFS_SB_VERSION2_CRCBIT))
56 return false;
57
58 /* Now check all the required V4 feature flags are set. */
59
60 #define V5_VERS_FLAGS (XFS_SB_VERSION_NLINKBIT | \
61 XFS_SB_VERSION_ALIGNBIT | \
62 XFS_SB_VERSION_LOGV2BIT | \
63 XFS_SB_VERSION_EXTFLGBIT | \
64 XFS_SB_VERSION_DIRV2BIT | \
65 XFS_SB_VERSION_MOREBITSBIT)
66
67 #define V5_FEAT_FLAGS (XFS_SB_VERSION2_LAZYSBCOUNTBIT | \
68 XFS_SB_VERSION2_ATTR2BIT | \
69 XFS_SB_VERSION2_PROJID32BIT | \
70 XFS_SB_VERSION2_CRCBIT)
71
72 if ((sbp->sb_versionnum & V5_VERS_FLAGS) != V5_VERS_FLAGS)
73 return false;
74 if ((sbp->sb_features2 & V5_FEAT_FLAGS) != V5_FEAT_FLAGS)
75 return false;
76 return true;
77 }
78
79 /*
80 * We current support XFS v5 formats with known features and v4 superblocks with
81 * at least V2 directories.
82 */
83 bool
xfs_sb_good_version(struct xfs_sb * sbp)84 xfs_sb_good_version(
85 struct xfs_sb *sbp)
86 {
87 /*
88 * All v5 filesystems are supported, but we must check that all the
89 * required v4 feature flags are enabled correctly as the code checks
90 * those flags and not for v5 support.
91 */
92 if (xfs_sb_is_v5(sbp))
93 return xfs_sb_validate_v5_features(sbp);
94
95 /* versions prior to v4 are not supported */
96 if (XFS_SB_VERSION_NUM(sbp) != XFS_SB_VERSION_4)
97 return false;
98
99 /* We must not have any unknown v4 feature bits set */
100 if ((sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS) ||
101 ((sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) &&
102 (sbp->sb_features2 & ~XFS_SB_VERSION2_OKBITS)))
103 return false;
104
105 /* V4 filesystems need v2 directories and unwritten extents */
106 if (!(sbp->sb_versionnum & XFS_SB_VERSION_DIRV2BIT))
107 return false;
108 if (!(sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT))
109 return false;
110
111 /* It's a supported v4 filesystem */
112 return true;
113 }
114
115 uint64_t
xfs_sb_version_to_features(struct xfs_sb * sbp)116 xfs_sb_version_to_features(
117 struct xfs_sb *sbp)
118 {
119 uint64_t features = 0;
120
121 /* optional V4 features */
122 if (sbp->sb_rblocks > 0)
123 features |= XFS_FEAT_REALTIME;
124 if (sbp->sb_versionnum & XFS_SB_VERSION_NLINKBIT)
125 features |= XFS_FEAT_NLINK;
126 if (sbp->sb_versionnum & XFS_SB_VERSION_ATTRBIT)
127 features |= XFS_FEAT_ATTR;
128 if (sbp->sb_versionnum & XFS_SB_VERSION_QUOTABIT)
129 features |= XFS_FEAT_QUOTA;
130 if (sbp->sb_versionnum & XFS_SB_VERSION_ALIGNBIT)
131 features |= XFS_FEAT_ALIGN;
132 if (sbp->sb_versionnum & XFS_SB_VERSION_LOGV2BIT)
133 features |= XFS_FEAT_LOGV2;
134 if (sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT)
135 features |= XFS_FEAT_DALIGN;
136 if (sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT)
137 features |= XFS_FEAT_EXTFLG;
138 if (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT)
139 features |= XFS_FEAT_SECTOR;
140 if (sbp->sb_versionnum & XFS_SB_VERSION_BORGBIT)
141 features |= XFS_FEAT_ASCIICI;
142 if (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) {
143 if (sbp->sb_features2 & XFS_SB_VERSION2_LAZYSBCOUNTBIT)
144 features |= XFS_FEAT_LAZYSBCOUNT;
145 if (sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT)
146 features |= XFS_FEAT_PROJID32;
147 if (sbp->sb_features2 & XFS_SB_VERSION2_FTYPE)
148 features |= XFS_FEAT_FTYPE;
149 }
150
151 if (!xfs_sb_is_v5(sbp))
152 return features;
153
154 /* Always on V5 features */
155 features |= XFS_FEAT_ALIGN | XFS_FEAT_LOGV2 | XFS_FEAT_EXTFLG |
156 XFS_FEAT_LAZYSBCOUNT | XFS_FEAT_PROJID32 |
157 XFS_FEAT_V3INODES | XFS_FEAT_CRC | XFS_FEAT_PQUOTINO;
158
159 /* Optional V5 features */
160 if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT)
161 features |= XFS_FEAT_FINOBT;
162 if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_RMAPBT)
163 features |= XFS_FEAT_RMAPBT;
164 if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_REFLINK)
165 features |= XFS_FEAT_REFLINK;
166 if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_INOBTCNT)
167 features |= XFS_FEAT_INOBTCNT;
168 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_FTYPE)
169 features |= XFS_FEAT_FTYPE;
170 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_SPINODES)
171 features |= XFS_FEAT_SPINODES;
172 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID)
173 features |= XFS_FEAT_META_UUID;
174 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_BIGTIME)
175 features |= XFS_FEAT_BIGTIME;
176 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR)
177 features |= XFS_FEAT_NEEDSREPAIR;
178 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_NREXT64)
179 features |= XFS_FEAT_NREXT64;
180 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_EXCHRANGE)
181 features |= XFS_FEAT_EXCHANGE_RANGE;
182 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_PARENT)
183 features |= XFS_FEAT_PARENT;
184 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR)
185 features |= XFS_FEAT_METADIR;
186 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED)
187 features |= XFS_FEAT_ZONED;
188
189 return features;
190 }
191
192 /* Check all the superblock fields we care about when reading one in. */
193 STATIC int
xfs_validate_sb_read(struct xfs_mount * mp,struct xfs_sb * sbp)194 xfs_validate_sb_read(
195 struct xfs_mount *mp,
196 struct xfs_sb *sbp)
197 {
198 if (!xfs_sb_is_v5(sbp))
199 return 0;
200
201 /*
202 * Version 5 superblock feature mask validation. Reject combinations
203 * the kernel cannot support up front before checking anything else.
204 */
205 if (xfs_sb_has_compat_feature(sbp, XFS_SB_FEAT_COMPAT_UNKNOWN)) {
206 xfs_warn(mp,
207 "Superblock has unknown compatible features (0x%x) enabled.",
208 (sbp->sb_features_compat & XFS_SB_FEAT_COMPAT_UNKNOWN));
209 xfs_warn(mp,
210 "Using a more recent kernel is recommended.");
211 }
212
213 if (xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
214 xfs_alert(mp,
215 "Superblock has unknown read-only compatible features (0x%x) enabled.",
216 (sbp->sb_features_ro_compat &
217 XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
218 if (!xfs_is_readonly(mp)) {
219 xfs_warn(mp,
220 "Attempted to mount read-only compatible filesystem read-write.");
221 xfs_warn(mp,
222 "Filesystem can only be safely mounted read only.");
223
224 return -EINVAL;
225 }
226 }
227 if (xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
228 xfs_warn(mp,
229 "Superblock has unknown incompatible features (0x%x) enabled.",
230 (sbp->sb_features_incompat &
231 XFS_SB_FEAT_INCOMPAT_UNKNOWN));
232 xfs_warn(mp,
233 "Filesystem cannot be safely mounted by this kernel.");
234 return -EINVAL;
235 }
236
237 return 0;
238 }
239
240 /* Return the number of extents covered by a single rt bitmap file */
241 static xfs_rtbxlen_t
xfs_extents_per_rbm(struct xfs_sb * sbp)242 xfs_extents_per_rbm(
243 struct xfs_sb *sbp)
244 {
245 if (xfs_sb_is_v5(sbp) &&
246 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR))
247 return sbp->sb_rgextents;
248 return sbp->sb_rextents;
249 }
250
251 /*
252 * Return the payload size of a single rt bitmap block (without the metadata
253 * header if any).
254 */
255 static inline unsigned int
xfs_rtbmblock_size(struct xfs_sb * sbp)256 xfs_rtbmblock_size(
257 struct xfs_sb *sbp)
258 {
259 if (xfs_sb_is_v5(sbp) &&
260 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR))
261 return sbp->sb_blocksize - sizeof(struct xfs_rtbuf_blkinfo);
262 return sbp->sb_blocksize;
263 }
264
265 static uint64_t
xfs_expected_rbmblocks(struct xfs_sb * sbp)266 xfs_expected_rbmblocks(
267 struct xfs_sb *sbp)
268 {
269 if (xfs_sb_is_v5(sbp) &&
270 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED))
271 return 0;
272 return howmany_64(xfs_extents_per_rbm(sbp),
273 NBBY * xfs_rtbmblock_size(sbp));
274 }
275
276 /* Validate the realtime geometry */
277 bool
xfs_validate_rt_geometry(struct xfs_sb * sbp)278 xfs_validate_rt_geometry(
279 struct xfs_sb *sbp)
280 {
281 if (xfs_sb_is_v5(sbp) &&
282 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED)) {
283 if (sbp->sb_rextsize != 1)
284 return false;
285 } else {
286 if (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE ||
287 sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE)
288 return false;
289 }
290
291 if (sbp->sb_rblocks == 0) {
292 if (sbp->sb_rextents != 0 || sbp->sb_rbmblocks != 0 ||
293 sbp->sb_rextslog != 0 || sbp->sb_frextents != 0)
294 return false;
295 return true;
296 }
297
298 if (sbp->sb_rextents == 0 ||
299 sbp->sb_rextents != div_u64(sbp->sb_rblocks, sbp->sb_rextsize) ||
300 sbp->sb_rextslog != xfs_compute_rextslog(sbp->sb_rextents) ||
301 sbp->sb_rbmblocks != xfs_expected_rbmblocks(sbp))
302 return false;
303
304 return true;
305 }
306
307 /* Check all the superblock fields we care about when writing one out. */
308 STATIC int
xfs_validate_sb_write(struct xfs_mount * mp,struct xfs_buf * bp,struct xfs_sb * sbp)309 xfs_validate_sb_write(
310 struct xfs_mount *mp,
311 struct xfs_buf *bp,
312 struct xfs_sb *sbp)
313 {
314 /*
315 * Carry out additional sb summary counter sanity checks when we write
316 * the superblock. We skip this in the read validator because there
317 * could be newer superblocks in the log and if the values are garbage
318 * even after replay we'll recalculate them at the end of log mount.
319 *
320 * mkfs has traditionally written zeroed counters to inprogress and
321 * secondary superblocks, so allow this usage to continue because
322 * we never read counters from such superblocks.
323 */
324 if (xfs_buf_daddr(bp) == XFS_SB_DADDR && !sbp->sb_inprogress &&
325 (sbp->sb_fdblocks > sbp->sb_dblocks ||
326 !xfs_verify_icount(mp, sbp->sb_icount) ||
327 sbp->sb_ifree > sbp->sb_icount)) {
328 xfs_warn(mp, "SB summary counter sanity check failed");
329 return -EFSCORRUPTED;
330 }
331
332 if (!xfs_sb_is_v5(sbp))
333 return 0;
334
335 /*
336 * Version 5 superblock feature mask validation. Reject combinations
337 * the kernel cannot support since we checked for unsupported bits in
338 * the read verifier, which means that memory is corrupt.
339 */
340 if (!xfs_is_readonly(mp) &&
341 xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
342 xfs_alert(mp,
343 "Corruption detected in superblock read-only compatible features (0x%x)!",
344 (sbp->sb_features_ro_compat &
345 XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
346 return -EFSCORRUPTED;
347 }
348 if (xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
349 xfs_warn(mp,
350 "Corruption detected in superblock incompatible features (0x%x)!",
351 (sbp->sb_features_incompat &
352 XFS_SB_FEAT_INCOMPAT_UNKNOWN));
353 return -EFSCORRUPTED;
354 }
355 if (xfs_sb_has_incompat_log_feature(sbp,
356 XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN)) {
357 xfs_warn(mp,
358 "Corruption detected in superblock incompatible log features (0x%x)!",
359 (sbp->sb_features_log_incompat &
360 XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN));
361 return -EFSCORRUPTED;
362 }
363
364 /*
365 * We can't read verify the sb LSN because the read verifier is called
366 * before the log is allocated and processed. We know the log is set up
367 * before write verifier calls, so check it here.
368 */
369 if (!xfs_log_check_lsn(mp, sbp->sb_lsn))
370 return -EFSCORRUPTED;
371
372 return 0;
373 }
374
375 int
xfs_compute_rgblklog(xfs_rtxlen_t rgextents,xfs_rgblock_t rextsize)376 xfs_compute_rgblklog(
377 xfs_rtxlen_t rgextents,
378 xfs_rgblock_t rextsize)
379 {
380 uint64_t rgblocks = (uint64_t)rgextents * rextsize;
381
382 return xfs_highbit64(rgblocks - 1) + 1;
383 }
384
385 static int
xfs_validate_sb_rtgroups(struct xfs_mount * mp,struct xfs_sb * sbp)386 xfs_validate_sb_rtgroups(
387 struct xfs_mount *mp,
388 struct xfs_sb *sbp)
389 {
390 uint64_t groups;
391 int rgblklog;
392
393 if (sbp->sb_rextsize == 0) {
394 xfs_warn(mp,
395 "Realtime extent size must not be zero.");
396 return -EINVAL;
397 }
398
399 if (sbp->sb_rgextents > XFS_MAX_RGBLOCKS / sbp->sb_rextsize) {
400 xfs_warn(mp,
401 "Realtime group size (%u) must be less than %u rt extents.",
402 sbp->sb_rgextents,
403 XFS_MAX_RGBLOCKS / sbp->sb_rextsize);
404 return -EINVAL;
405 }
406
407 if (sbp->sb_rgextents < XFS_MIN_RGEXTENTS) {
408 xfs_warn(mp,
409 "Realtime group size (%u) must be at least %u rt extents.",
410 sbp->sb_rgextents, XFS_MIN_RGEXTENTS);
411 return -EINVAL;
412 }
413
414 if (sbp->sb_rgcount > XFS_MAX_RGNUMBER) {
415 xfs_warn(mp,
416 "Realtime groups (%u) must be less than %u.",
417 sbp->sb_rgcount, XFS_MAX_RGNUMBER);
418 return -EINVAL;
419 }
420
421 groups = howmany_64(sbp->sb_rextents, sbp->sb_rgextents);
422 if (groups != sbp->sb_rgcount) {
423 xfs_warn(mp,
424 "Realtime groups (%u) do not cover the entire rt section; need (%llu) groups.",
425 sbp->sb_rgcount, groups);
426 return -EINVAL;
427 }
428
429 /* Exchange-range is required for fsr to work on realtime files */
430 if (!(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_EXCHRANGE)) {
431 xfs_warn(mp,
432 "Realtime groups feature requires exchange-range support.");
433 return -EINVAL;
434 }
435
436 rgblklog = xfs_compute_rgblklog(sbp->sb_rgextents, sbp->sb_rextsize);
437 if (sbp->sb_rgblklog != rgblklog) {
438 xfs_warn(mp,
439 "Realtime group log (%d) does not match expected value (%d).",
440 sbp->sb_rgblklog, rgblklog);
441 return -EINVAL;
442 }
443
444 return 0;
445 }
446
447 static int
xfs_validate_sb_zoned(struct xfs_mount * mp,struct xfs_sb * sbp)448 xfs_validate_sb_zoned(
449 struct xfs_mount *mp,
450 struct xfs_sb *sbp)
451 {
452 if (sbp->sb_frextents != 0) {
453 xfs_warn(mp,
454 "sb_frextents must be zero for zoned file systems.");
455 return -EINVAL;
456 }
457
458 if (sbp->sb_rtstart && sbp->sb_rtstart < sbp->sb_dblocks) {
459 xfs_warn(mp,
460 "sb_rtstart (%lld) overlaps sb_dblocks (%lld).",
461 sbp->sb_rtstart, sbp->sb_dblocks);
462 return -EINVAL;
463 }
464
465 if (sbp->sb_rtreserved && sbp->sb_rtreserved >= sbp->sb_rblocks) {
466 xfs_warn(mp,
467 "sb_rtreserved (%lld) larger than sb_rblocks (%lld).",
468 sbp->sb_rtreserved, sbp->sb_rblocks);
469 return -EINVAL;
470 }
471
472 return 0;
473 }
474
475 /* Check the validity of the SB. */
476 STATIC int
xfs_validate_sb_common(struct xfs_mount * mp,struct xfs_buf * bp,struct xfs_sb * sbp)477 xfs_validate_sb_common(
478 struct xfs_mount *mp,
479 struct xfs_buf *bp,
480 struct xfs_sb *sbp)
481 {
482 struct xfs_dsb *dsb = bp->b_addr;
483 uint32_t agcount = 0;
484 uint32_t rem;
485 bool has_dalign;
486 int error;
487
488 if (!xfs_verify_magic(bp, dsb->sb_magicnum)) {
489 xfs_warn(mp,
490 "Superblock has bad magic number 0x%x. Not an XFS filesystem?",
491 be32_to_cpu(dsb->sb_magicnum));
492 return -EWRONGFS;
493 }
494
495 if (!xfs_sb_good_version(sbp)) {
496 xfs_warn(mp,
497 "Superblock has unknown features enabled or corrupted feature masks.");
498 return -EWRONGFS;
499 }
500
501 /*
502 * Validate feature flags and state
503 */
504 if (xfs_sb_is_v5(sbp)) {
505 if (sbp->sb_blocksize < XFS_MIN_CRC_BLOCKSIZE) {
506 xfs_notice(mp,
507 "Block size (%u bytes) too small for Version 5 superblock (minimum %d bytes)",
508 sbp->sb_blocksize, XFS_MIN_CRC_BLOCKSIZE);
509 return -EFSCORRUPTED;
510 }
511
512 /* V5 has a separate project quota inode */
513 if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
514 xfs_notice(mp,
515 "Version 5 of Super block has XFS_OQUOTA bits.");
516 return -EFSCORRUPTED;
517 }
518
519 /*
520 * Full inode chunks must be aligned to inode chunk size when
521 * sparse inodes are enabled to support the sparse chunk
522 * allocation algorithm and prevent overlapping inode records.
523 */
524 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_SPINODES) {
525 uint32_t align;
526
527 align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize
528 >> sbp->sb_blocklog;
529 if (sbp->sb_inoalignmt != align) {
530 xfs_warn(mp,
531 "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.",
532 sbp->sb_inoalignmt, align);
533 return -EINVAL;
534 }
535
536 if (sbp->sb_spino_align &&
537 (sbp->sb_spino_align > sbp->sb_inoalignmt ||
538 (sbp->sb_inoalignmt % sbp->sb_spino_align) != 0)) {
539 xfs_warn(mp,
540 "Sparse inode alignment (%u) is invalid, must be integer factor of (%u).",
541 sbp->sb_spino_align,
542 sbp->sb_inoalignmt);
543 return -EINVAL;
544 }
545 } else if (sbp->sb_spino_align) {
546 xfs_warn(mp,
547 "Sparse inode alignment (%u) should be zero.",
548 sbp->sb_spino_align);
549 return -EINVAL;
550 }
551
552 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR) {
553 if (memchr_inv(sbp->sb_pad, 0, sizeof(sbp->sb_pad))) {
554 xfs_warn(mp,
555 "Metadir superblock padding fields must be zero.");
556 return -EINVAL;
557 }
558
559 error = xfs_validate_sb_rtgroups(mp, sbp);
560 if (error)
561 return error;
562 }
563 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED) {
564 error = xfs_validate_sb_zoned(mp, sbp);
565 if (error)
566 return error;
567 }
568 } else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
569 XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
570 xfs_notice(mp,
571 "Superblock earlier than Version 5 has XFS_{P|G}QUOTA_{ENFD|CHKD} bits.");
572 return -EFSCORRUPTED;
573 }
574
575 if (unlikely(
576 sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
577 xfs_warn(mp,
578 "filesystem is marked as having an external log; "
579 "specify logdev on the mount command line.");
580 return -EINVAL;
581 }
582
583 if (unlikely(
584 sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
585 xfs_warn(mp,
586 "filesystem is marked as having an internal log; "
587 "do not specify logdev on the mount command line.");
588 return -EINVAL;
589 }
590
591 /* Compute agcount for this number of dblocks and agblocks */
592 if (sbp->sb_agblocks) {
593 agcount = div_u64_rem(sbp->sb_dblocks, sbp->sb_agblocks, &rem);
594 if (rem)
595 agcount++;
596 }
597
598 /*
599 * More sanity checking. Most of these were stolen directly from
600 * xfs_repair.
601 */
602 if (unlikely(
603 sbp->sb_agcount <= 0 ||
604 sbp->sb_sectsize < XFS_MIN_SECTORSIZE ||
605 sbp->sb_sectsize > XFS_MAX_SECTORSIZE ||
606 sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG ||
607 sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG ||
608 sbp->sb_sectsize != (1 << sbp->sb_sectlog) ||
609 sbp->sb_blocksize < XFS_MIN_BLOCKSIZE ||
610 sbp->sb_blocksize > XFS_MAX_BLOCKSIZE ||
611 sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG ||
612 sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
613 sbp->sb_blocksize != (1 << sbp->sb_blocklog) ||
614 sbp->sb_dirblklog + sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
615 sbp->sb_inodesize < XFS_DINODE_MIN_SIZE ||
616 sbp->sb_inodesize > XFS_DINODE_MAX_SIZE ||
617 sbp->sb_inodelog < XFS_DINODE_MIN_LOG ||
618 sbp->sb_inodelog > XFS_DINODE_MAX_LOG ||
619 sbp->sb_inodesize != (1 << sbp->sb_inodelog) ||
620 sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
621 XFS_FSB_TO_B(mp, sbp->sb_agblocks) < XFS_MIN_AG_BYTES ||
622 XFS_FSB_TO_B(mp, sbp->sb_agblocks) > XFS_MAX_AG_BYTES ||
623 sbp->sb_agblklog != xfs_highbit32(sbp->sb_agblocks - 1) + 1 ||
624 agcount == 0 || agcount != sbp->sb_agcount ||
625 (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog) ||
626 (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) ||
627 (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) ||
628 (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */) ||
629 sbp->sb_dblocks == 0 ||
630 sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) ||
631 sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp) ||
632 sbp->sb_shared_vn != 0)) {
633 xfs_notice(mp, "SB sanity check failed");
634 return -EFSCORRUPTED;
635 }
636
637 /*
638 * Logs that are too large are not supported at all. Reject them
639 * outright. Logs that are too small are tolerated on v4 filesystems,
640 * but we can only check that when mounting the log. Hence we skip
641 * those checks here.
642 */
643 if (sbp->sb_logblocks > XFS_MAX_LOG_BLOCKS) {
644 xfs_notice(mp,
645 "Log size 0x%x blocks too large, maximum size is 0x%llx blocks",
646 sbp->sb_logblocks, XFS_MAX_LOG_BLOCKS);
647 return -EFSCORRUPTED;
648 }
649
650 if (XFS_FSB_TO_B(mp, sbp->sb_logblocks) > XFS_MAX_LOG_BYTES) {
651 xfs_warn(mp,
652 "log size 0x%llx bytes too large, maximum size is 0x%llx bytes",
653 XFS_FSB_TO_B(mp, sbp->sb_logblocks),
654 XFS_MAX_LOG_BYTES);
655 return -EFSCORRUPTED;
656 }
657
658 /*
659 * Do not allow filesystems with corrupted log sector or stripe units to
660 * be mounted. We cannot safely size the iclogs or write to the log if
661 * the log stripe unit is not valid.
662 */
663 if (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT) {
664 if (sbp->sb_logsectsize != (1U << sbp->sb_logsectlog)) {
665 xfs_notice(mp,
666 "log sector size in bytes/log2 (0x%x/0x%x) must match",
667 sbp->sb_logsectsize, 1U << sbp->sb_logsectlog);
668 return -EFSCORRUPTED;
669 }
670 } else if (sbp->sb_logsectsize || sbp->sb_logsectlog) {
671 xfs_notice(mp,
672 "log sector size in bytes/log2 (0x%x/0x%x) are not zero",
673 sbp->sb_logsectsize, sbp->sb_logsectlog);
674 return -EFSCORRUPTED;
675 }
676
677 if (sbp->sb_logsunit > 1) {
678 if (sbp->sb_logsunit % sbp->sb_blocksize) {
679 xfs_notice(mp,
680 "log stripe unit 0x%x bytes must be a multiple of block size",
681 sbp->sb_logsunit);
682 return -EFSCORRUPTED;
683 }
684 if (sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE) {
685 xfs_notice(mp,
686 "log stripe unit 0x%x bytes over maximum size (0x%x bytes)",
687 sbp->sb_logsunit, XLOG_MAX_RECORD_BSIZE);
688 return -EFSCORRUPTED;
689 }
690 }
691
692 if (!xfs_validate_rt_geometry(sbp)) {
693 xfs_notice(mp,
694 "realtime %sgeometry check failed",
695 sbp->sb_rblocks ? "" : "zeroed ");
696 return -EFSCORRUPTED;
697 }
698
699 /*
700 * Either (sb_unit and !hasdalign) or (!sb_unit and hasdalign)
701 * would imply the image is corrupted.
702 */
703 has_dalign = sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT;
704 if (!!sbp->sb_unit ^ has_dalign) {
705 xfs_notice(mp, "SB stripe alignment sanity check failed");
706 return -EFSCORRUPTED;
707 }
708
709 if (!xfs_validate_stripe_geometry(mp, XFS_FSB_TO_B(mp, sbp->sb_unit),
710 XFS_FSB_TO_B(mp, sbp->sb_width), 0,
711 xfs_buf_daddr(bp) == XFS_SB_DADDR, false))
712 return -EFSCORRUPTED;
713
714 /*
715 * Currently only very few inode sizes are supported.
716 */
717 switch (sbp->sb_inodesize) {
718 case 256:
719 case 512:
720 case 1024:
721 case 2048:
722 break;
723 default:
724 xfs_warn(mp, "inode size of %d bytes not supported",
725 sbp->sb_inodesize);
726 return -ENOSYS;
727 }
728
729 return 0;
730 }
731
732 void
xfs_sb_quota_from_disk(struct xfs_sb * sbp)733 xfs_sb_quota_from_disk(struct xfs_sb *sbp)
734 {
735 if (xfs_sb_is_v5(sbp) &&
736 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR)) {
737 sbp->sb_uquotino = NULLFSINO;
738 sbp->sb_gquotino = NULLFSINO;
739 sbp->sb_pquotino = NULLFSINO;
740 return;
741 }
742
743 /*
744 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
745 * leads to in-core values having two different values for a quota
746 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
747 * NULLFSINO.
748 *
749 * Note that this change affect only the in-core values. These
750 * values are not written back to disk unless any quota information
751 * is written to the disk. Even in that case, sb_pquotino field is
752 * not written to disk unless the superblock supports pquotino.
753 */
754 if (sbp->sb_uquotino == 0)
755 sbp->sb_uquotino = NULLFSINO;
756 if (sbp->sb_gquotino == 0)
757 sbp->sb_gquotino = NULLFSINO;
758 if (sbp->sb_pquotino == 0)
759 sbp->sb_pquotino = NULLFSINO;
760
761 /*
762 * We need to do these manipilations only if we are working
763 * with an older version of on-disk superblock.
764 */
765 if (xfs_sb_is_v5(sbp))
766 return;
767
768 if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
769 sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
770 XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
771 if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
772 sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
773 XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
774 sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
775
776 if (sbp->sb_qflags & XFS_PQUOTA_ACCT &&
777 sbp->sb_gquotino != NULLFSINO) {
778 /*
779 * In older version of superblock, on-disk superblock only
780 * has sb_gquotino, and in-core superblock has both sb_gquotino
781 * and sb_pquotino. But, only one of them is supported at any
782 * point of time. So, if PQUOTA is set in disk superblock,
783 * copy over sb_gquotino to sb_pquotino. The NULLFSINO test
784 * above is to make sure we don't do this twice and wipe them
785 * both out!
786 */
787 sbp->sb_pquotino = sbp->sb_gquotino;
788 sbp->sb_gquotino = NULLFSINO;
789 }
790 }
791
792 static void
__xfs_sb_from_disk(struct xfs_sb * to,struct xfs_dsb * from,bool convert_xquota)793 __xfs_sb_from_disk(
794 struct xfs_sb *to,
795 struct xfs_dsb *from,
796 bool convert_xquota)
797 {
798 to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
799 to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
800 to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
801 to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
802 to->sb_rextents = be64_to_cpu(from->sb_rextents);
803 memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
804 to->sb_logstart = be64_to_cpu(from->sb_logstart);
805 to->sb_rootino = be64_to_cpu(from->sb_rootino);
806 to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
807 to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
808 to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
809 to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
810 to->sb_agcount = be32_to_cpu(from->sb_agcount);
811 to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
812 to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
813 to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
814 to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
815 to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
816 to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
817 memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
818 to->sb_blocklog = from->sb_blocklog;
819 to->sb_sectlog = from->sb_sectlog;
820 to->sb_inodelog = from->sb_inodelog;
821 to->sb_inopblog = from->sb_inopblog;
822 to->sb_agblklog = from->sb_agblklog;
823 to->sb_rextslog = from->sb_rextslog;
824 to->sb_inprogress = from->sb_inprogress;
825 to->sb_imax_pct = from->sb_imax_pct;
826 to->sb_icount = be64_to_cpu(from->sb_icount);
827 to->sb_ifree = be64_to_cpu(from->sb_ifree);
828 to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
829 to->sb_frextents = be64_to_cpu(from->sb_frextents);
830 to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
831 to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
832 to->sb_qflags = be16_to_cpu(from->sb_qflags);
833 to->sb_flags = from->sb_flags;
834 to->sb_shared_vn = from->sb_shared_vn;
835 to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
836 to->sb_unit = be32_to_cpu(from->sb_unit);
837 to->sb_width = be32_to_cpu(from->sb_width);
838 to->sb_dirblklog = from->sb_dirblklog;
839 to->sb_logsectlog = from->sb_logsectlog;
840 to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
841 to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
842 to->sb_features2 = be32_to_cpu(from->sb_features2);
843 to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
844 to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
845 to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
846 to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
847 to->sb_features_log_incompat =
848 be32_to_cpu(from->sb_features_log_incompat);
849 /* crc is only used on disk, not in memory; just init to 0 here. */
850 to->sb_crc = 0;
851 to->sb_spino_align = be32_to_cpu(from->sb_spino_align);
852 to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
853 to->sb_lsn = be64_to_cpu(from->sb_lsn);
854 /*
855 * sb_meta_uuid is only on disk if it differs from sb_uuid and the
856 * feature flag is set; if not set we keep it only in memory.
857 */
858 if (xfs_sb_is_v5(to) &&
859 (to->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID))
860 uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
861 else
862 uuid_copy(&to->sb_meta_uuid, &from->sb_uuid);
863 /* Convert on-disk flags to in-memory flags? */
864 if (convert_xquota)
865 xfs_sb_quota_from_disk(to);
866
867 if (to->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR) {
868 to->sb_metadirino = be64_to_cpu(from->sb_metadirino);
869 to->sb_rgblklog = from->sb_rgblklog;
870 memcpy(to->sb_pad, from->sb_pad, sizeof(to->sb_pad));
871 to->sb_rgcount = be32_to_cpu(from->sb_rgcount);
872 to->sb_rgextents = be32_to_cpu(from->sb_rgextents);
873 to->sb_rbmino = NULLFSINO;
874 to->sb_rsumino = NULLFSINO;
875 } else {
876 to->sb_metadirino = NULLFSINO;
877 to->sb_rgcount = 1;
878 to->sb_rgextents = 0;
879 }
880
881 if (to->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED) {
882 to->sb_rtstart = be64_to_cpu(from->sb_rtstart);
883 to->sb_rtreserved = be64_to_cpu(from->sb_rtreserved);
884 } else {
885 to->sb_rtstart = 0;
886 to->sb_rtreserved = 0;
887 }
888 }
889
890 void
xfs_sb_from_disk(struct xfs_sb * to,struct xfs_dsb * from)891 xfs_sb_from_disk(
892 struct xfs_sb *to,
893 struct xfs_dsb *from)
894 {
895 __xfs_sb_from_disk(to, from, true);
896 }
897
898 static void
xfs_sb_quota_to_disk(struct xfs_dsb * to,struct xfs_sb * from)899 xfs_sb_quota_to_disk(
900 struct xfs_dsb *to,
901 struct xfs_sb *from)
902 {
903 uint16_t qflags = from->sb_qflags;
904
905 if (xfs_sb_is_v5(from) &&
906 (from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR)) {
907 to->sb_qflags = cpu_to_be16(from->sb_qflags);
908 to->sb_uquotino = cpu_to_be64(0);
909 to->sb_gquotino = cpu_to_be64(0);
910 to->sb_pquotino = cpu_to_be64(0);
911 return;
912 }
913
914 to->sb_uquotino = cpu_to_be64(from->sb_uquotino);
915
916 /*
917 * The in-memory superblock quota state matches the v5 on-disk format so
918 * just write them out and return
919 */
920 if (xfs_sb_is_v5(from)) {
921 to->sb_qflags = cpu_to_be16(from->sb_qflags);
922 to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
923 to->sb_pquotino = cpu_to_be64(from->sb_pquotino);
924 return;
925 }
926
927 /*
928 * For older superblocks (v4), the in-core version of sb_qflags do not
929 * have XFS_OQUOTA_* flags, whereas the on-disk version does. So,
930 * convert incore XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
931 */
932 qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
933 XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
934
935 if (from->sb_qflags &
936 (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
937 qflags |= XFS_OQUOTA_ENFD;
938 if (from->sb_qflags &
939 (XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
940 qflags |= XFS_OQUOTA_CHKD;
941 to->sb_qflags = cpu_to_be16(qflags);
942
943 /*
944 * GQUOTINO and PQUOTINO cannot be used together in versions
945 * of superblock that do not have pquotino. from->sb_flags
946 * tells us which quota is active and should be copied to
947 * disk. If neither are active, we should NULL the inode.
948 *
949 * In all cases, the separate pquotino must remain 0 because it
950 * is beyond the "end" of the valid non-pquotino superblock.
951 */
952 if (from->sb_qflags & XFS_GQUOTA_ACCT)
953 to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
954 else if (from->sb_qflags & XFS_PQUOTA_ACCT)
955 to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
956 else {
957 /*
958 * We can't rely on just the fields being logged to tell us
959 * that it is safe to write NULLFSINO - we should only do that
960 * if quotas are not actually enabled. Hence only write
961 * NULLFSINO if both in-core quota inodes are NULL.
962 */
963 if (from->sb_gquotino == NULLFSINO &&
964 from->sb_pquotino == NULLFSINO)
965 to->sb_gquotino = cpu_to_be64(NULLFSINO);
966 }
967
968 to->sb_pquotino = 0;
969 }
970
971 void
xfs_sb_to_disk(struct xfs_dsb * to,struct xfs_sb * from)972 xfs_sb_to_disk(
973 struct xfs_dsb *to,
974 struct xfs_sb *from)
975 {
976 xfs_sb_quota_to_disk(to, from);
977
978 to->sb_magicnum = cpu_to_be32(from->sb_magicnum);
979 to->sb_blocksize = cpu_to_be32(from->sb_blocksize);
980 to->sb_dblocks = cpu_to_be64(from->sb_dblocks);
981 to->sb_rblocks = cpu_to_be64(from->sb_rblocks);
982 to->sb_rextents = cpu_to_be64(from->sb_rextents);
983 memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
984 to->sb_logstart = cpu_to_be64(from->sb_logstart);
985 to->sb_rootino = cpu_to_be64(from->sb_rootino);
986 to->sb_rbmino = cpu_to_be64(from->sb_rbmino);
987 to->sb_rsumino = cpu_to_be64(from->sb_rsumino);
988 to->sb_rextsize = cpu_to_be32(from->sb_rextsize);
989 to->sb_agblocks = cpu_to_be32(from->sb_agblocks);
990 to->sb_agcount = cpu_to_be32(from->sb_agcount);
991 to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks);
992 to->sb_logblocks = cpu_to_be32(from->sb_logblocks);
993 to->sb_versionnum = cpu_to_be16(from->sb_versionnum);
994 to->sb_sectsize = cpu_to_be16(from->sb_sectsize);
995 to->sb_inodesize = cpu_to_be16(from->sb_inodesize);
996 to->sb_inopblock = cpu_to_be16(from->sb_inopblock);
997 memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
998 to->sb_blocklog = from->sb_blocklog;
999 to->sb_sectlog = from->sb_sectlog;
1000 to->sb_inodelog = from->sb_inodelog;
1001 to->sb_inopblog = from->sb_inopblog;
1002 to->sb_agblklog = from->sb_agblklog;
1003 to->sb_rextslog = from->sb_rextslog;
1004 to->sb_inprogress = from->sb_inprogress;
1005 to->sb_imax_pct = from->sb_imax_pct;
1006 to->sb_icount = cpu_to_be64(from->sb_icount);
1007 to->sb_ifree = cpu_to_be64(from->sb_ifree);
1008 to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks);
1009 to->sb_frextents = cpu_to_be64(from->sb_frextents);
1010
1011 to->sb_flags = from->sb_flags;
1012 to->sb_shared_vn = from->sb_shared_vn;
1013 to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt);
1014 to->sb_unit = cpu_to_be32(from->sb_unit);
1015 to->sb_width = cpu_to_be32(from->sb_width);
1016 to->sb_dirblklog = from->sb_dirblklog;
1017 to->sb_logsectlog = from->sb_logsectlog;
1018 to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize);
1019 to->sb_logsunit = cpu_to_be32(from->sb_logsunit);
1020
1021 /*
1022 * We need to ensure that bad_features2 always matches features2.
1023 * Hence we enforce that here rather than having to remember to do it
1024 * everywhere else that updates features2.
1025 */
1026 from->sb_bad_features2 = from->sb_features2;
1027 to->sb_features2 = cpu_to_be32(from->sb_features2);
1028 to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2);
1029
1030 if (!xfs_sb_is_v5(from))
1031 return;
1032
1033 to->sb_features_compat = cpu_to_be32(from->sb_features_compat);
1034 to->sb_features_ro_compat =
1035 cpu_to_be32(from->sb_features_ro_compat);
1036 to->sb_features_incompat =
1037 cpu_to_be32(from->sb_features_incompat);
1038 to->sb_features_log_incompat =
1039 cpu_to_be32(from->sb_features_log_incompat);
1040 to->sb_spino_align = cpu_to_be32(from->sb_spino_align);
1041 to->sb_lsn = cpu_to_be64(from->sb_lsn);
1042 if (from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID)
1043 uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
1044
1045 if (from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR) {
1046 to->sb_metadirino = cpu_to_be64(from->sb_metadirino);
1047 to->sb_rgblklog = from->sb_rgblklog;
1048 memset(to->sb_pad, 0, sizeof(to->sb_pad));
1049 to->sb_rgcount = cpu_to_be32(from->sb_rgcount);
1050 to->sb_rgextents = cpu_to_be32(from->sb_rgextents);
1051 to->sb_rbmino = cpu_to_be64(0);
1052 to->sb_rsumino = cpu_to_be64(0);
1053 }
1054
1055 if (from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED) {
1056 to->sb_rtstart = cpu_to_be64(from->sb_rtstart);
1057 to->sb_rtreserved = cpu_to_be64(from->sb_rtreserved);
1058 }
1059 }
1060
1061 /*
1062 * If the superblock has the CRC feature bit set or the CRC field is non-null,
1063 * check that the CRC is valid. We check the CRC field is non-null because a
1064 * single bit error could clear the feature bit and unused parts of the
1065 * superblock are supposed to be zero. Hence a non-null crc field indicates that
1066 * we've potentially lost a feature bit and we should check it anyway.
1067 *
1068 * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
1069 * last field in V4 secondary superblocks. So for secondary superblocks,
1070 * we are more forgiving, and ignore CRC failures if the primary doesn't
1071 * indicate that the fs version is V5.
1072 */
1073 static void
xfs_sb_read_verify(struct xfs_buf * bp)1074 xfs_sb_read_verify(
1075 struct xfs_buf *bp)
1076 {
1077 struct xfs_sb sb;
1078 struct xfs_mount *mp = bp->b_mount;
1079 struct xfs_dsb *dsb = bp->b_addr;
1080 int error;
1081
1082 /*
1083 * open code the version check to avoid needing to convert the entire
1084 * superblock from disk order just to check the version number
1085 */
1086 if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
1087 (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
1088 XFS_SB_VERSION_5) ||
1089 dsb->sb_crc != 0)) {
1090
1091 if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
1092 /* Only fail bad secondaries on a known V5 filesystem */
1093 if (xfs_buf_daddr(bp) == XFS_SB_DADDR ||
1094 xfs_has_crc(mp)) {
1095 error = -EFSBADCRC;
1096 goto out_error;
1097 }
1098 }
1099 }
1100
1101 /*
1102 * Check all the superblock fields. Don't byteswap the xquota flags
1103 * because _verify_common checks the on-disk values.
1104 */
1105 __xfs_sb_from_disk(&sb, dsb, false);
1106 error = xfs_validate_sb_common(mp, bp, &sb);
1107 if (error)
1108 goto out_error;
1109 error = xfs_validate_sb_read(mp, &sb);
1110
1111 out_error:
1112 if (error == -EFSCORRUPTED || error == -EFSBADCRC)
1113 xfs_verifier_error(bp, error, __this_address);
1114 else if (error)
1115 xfs_buf_ioerror(bp, error);
1116 }
1117
1118 /*
1119 * We may be probed for a filesystem match, so we may not want to emit
1120 * messages when the superblock buffer is not actually an XFS superblock.
1121 * If we find an XFS superblock, then run a normal, noisy mount because we are
1122 * really going to mount it and want to know about errors.
1123 */
1124 static void
xfs_sb_quiet_read_verify(struct xfs_buf * bp)1125 xfs_sb_quiet_read_verify(
1126 struct xfs_buf *bp)
1127 {
1128 struct xfs_dsb *dsb = bp->b_addr;
1129
1130 if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
1131 /* XFS filesystem, verify noisily! */
1132 xfs_sb_read_verify(bp);
1133 return;
1134 }
1135 /* quietly fail */
1136 xfs_buf_ioerror(bp, -EWRONGFS);
1137 }
1138
1139 static void
xfs_sb_write_verify(struct xfs_buf * bp)1140 xfs_sb_write_verify(
1141 struct xfs_buf *bp)
1142 {
1143 struct xfs_sb sb;
1144 struct xfs_mount *mp = bp->b_mount;
1145 struct xfs_buf_log_item *bip = bp->b_log_item;
1146 struct xfs_dsb *dsb = bp->b_addr;
1147 int error;
1148
1149 /*
1150 * Check all the superblock fields. Don't byteswap the xquota flags
1151 * because _verify_common checks the on-disk values.
1152 */
1153 __xfs_sb_from_disk(&sb, dsb, false);
1154 error = xfs_validate_sb_common(mp, bp, &sb);
1155 if (error)
1156 goto out_error;
1157 error = xfs_validate_sb_write(mp, bp, &sb);
1158 if (error)
1159 goto out_error;
1160
1161 if (!xfs_sb_is_v5(&sb))
1162 return;
1163
1164 if (bip)
1165 dsb->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
1166
1167 xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
1168 return;
1169
1170 out_error:
1171 xfs_verifier_error(bp, error, __this_address);
1172 }
1173
1174 const struct xfs_buf_ops xfs_sb_buf_ops = {
1175 .name = "xfs_sb",
1176 .magic = { cpu_to_be32(XFS_SB_MAGIC), cpu_to_be32(XFS_SB_MAGIC) },
1177 .verify_read = xfs_sb_read_verify,
1178 .verify_write = xfs_sb_write_verify,
1179 };
1180
1181 const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
1182 .name = "xfs_sb_quiet",
1183 .magic = { cpu_to_be32(XFS_SB_MAGIC), cpu_to_be32(XFS_SB_MAGIC) },
1184 .verify_read = xfs_sb_quiet_read_verify,
1185 .verify_write = xfs_sb_write_verify,
1186 };
1187
1188 /* Compute cached rt geometry from the incore sb. */
1189 void
xfs_sb_mount_rextsize(struct xfs_mount * mp,struct xfs_sb * sbp)1190 xfs_sb_mount_rextsize(
1191 struct xfs_mount *mp,
1192 struct xfs_sb *sbp)
1193 {
1194 struct xfs_groups *rgs = &mp->m_groups[XG_TYPE_RTG];
1195
1196 mp->m_rtxblklog = log2_if_power2(sbp->sb_rextsize);
1197 mp->m_rtxblkmask = mask64_if_power2(sbp->sb_rextsize);
1198
1199 if (xfs_sb_is_v5(sbp) &&
1200 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR)) {
1201 rgs->blocks = sbp->sb_rgextents * sbp->sb_rextsize;
1202 rgs->blklog = mp->m_sb.sb_rgblklog;
1203 rgs->blkmask = xfs_mask32lo(mp->m_sb.sb_rgblklog);
1204 rgs->start_fsb = mp->m_sb.sb_rtstart;
1205 if (xfs_sb_has_incompat_feature(sbp,
1206 XFS_SB_FEAT_INCOMPAT_ZONE_GAPS))
1207 rgs->has_daddr_gaps = true;
1208 } else {
1209 rgs->blocks = 0;
1210 rgs->blklog = 0;
1211 rgs->blkmask = (uint64_t)-1;
1212 }
1213 }
1214
1215 /* Update incore sb rt extent size, then recompute the cached rt geometry. */
1216 void
xfs_mount_sb_set_rextsize(struct xfs_mount * mp,struct xfs_sb * sbp,xfs_agblock_t rextsize)1217 xfs_mount_sb_set_rextsize(
1218 struct xfs_mount *mp,
1219 struct xfs_sb *sbp,
1220 xfs_agblock_t rextsize)
1221 {
1222 sbp->sb_rextsize = rextsize;
1223 if (xfs_sb_is_v5(sbp) &&
1224 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR))
1225 sbp->sb_rgblklog = xfs_compute_rgblklog(sbp->sb_rgextents,
1226 rextsize);
1227
1228 xfs_sb_mount_rextsize(mp, sbp);
1229 }
1230
1231 /*
1232 * xfs_mount_common
1233 *
1234 * Mount initialization code establishing various mount
1235 * fields from the superblock associated with the given
1236 * mount structure.
1237 *
1238 * Inode geometry are calculated in xfs_ialloc_setup_geometry.
1239 */
1240 void
xfs_sb_mount_common(struct xfs_mount * mp,struct xfs_sb * sbp)1241 xfs_sb_mount_common(
1242 struct xfs_mount *mp,
1243 struct xfs_sb *sbp)
1244 {
1245 struct xfs_groups *ags = &mp->m_groups[XG_TYPE_AG];
1246
1247 mp->m_agfrotor = 0;
1248 atomic_set(&mp->m_agirotor, 0);
1249 mp->m_maxagi = mp->m_sb.sb_agcount;
1250 mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
1251 mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
1252 mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
1253 mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
1254 mp->m_blockmask = sbp->sb_blocksize - 1;
1255 mp->m_blockwsize = xfs_rtbmblock_size(sbp) >> XFS_WORDLOG;
1256 mp->m_rtx_per_rbmblock = mp->m_blockwsize << XFS_NBWORDLOG;
1257
1258 ags->blocks = mp->m_sb.sb_agblocks;
1259 ags->blklog = mp->m_sb.sb_agblklog;
1260 ags->blkmask = xfs_mask32lo(mp->m_sb.sb_agblklog);
1261
1262 xfs_sb_mount_rextsize(mp, sbp);
1263
1264 mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, true);
1265 mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, false);
1266 mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
1267 mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
1268
1269 mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, true);
1270 mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, false);
1271 mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
1272 mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
1273
1274 mp->m_rmap_mxr[0] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, true);
1275 mp->m_rmap_mxr[1] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, false);
1276 mp->m_rmap_mnr[0] = mp->m_rmap_mxr[0] / 2;
1277 mp->m_rmap_mnr[1] = mp->m_rmap_mxr[1] / 2;
1278
1279 mp->m_rtrmap_mxr[0] = xfs_rtrmapbt_maxrecs(mp, sbp->sb_blocksize, true);
1280 mp->m_rtrmap_mxr[1] = xfs_rtrmapbt_maxrecs(mp, sbp->sb_blocksize, false);
1281 mp->m_rtrmap_mnr[0] = mp->m_rtrmap_mxr[0] / 2;
1282 mp->m_rtrmap_mnr[1] = mp->m_rtrmap_mxr[1] / 2;
1283
1284 mp->m_refc_mxr[0] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize, true);
1285 mp->m_refc_mxr[1] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize, false);
1286 mp->m_refc_mnr[0] = mp->m_refc_mxr[0] / 2;
1287 mp->m_refc_mnr[1] = mp->m_refc_mxr[1] / 2;
1288
1289 mp->m_rtrefc_mxr[0] = xfs_rtrefcountbt_maxrecs(mp, sbp->sb_blocksize,
1290 true);
1291 mp->m_rtrefc_mxr[1] = xfs_rtrefcountbt_maxrecs(mp, sbp->sb_blocksize,
1292 false);
1293 mp->m_rtrefc_mnr[0] = mp->m_rtrefc_mxr[0] / 2;
1294 mp->m_rtrefc_mnr[1] = mp->m_rtrefc_mxr[1] / 2;
1295
1296 mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
1297 mp->m_alloc_set_aside = xfs_alloc_set_aside(mp);
1298 mp->m_ag_max_usable = xfs_alloc_ag_max_usable(mp);
1299 }
1300
1301 /*
1302 * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
1303 * into the superblock buffer to be logged. It does not provide the higher
1304 * level of locking that is needed to protect the in-core superblock from
1305 * concurrent access.
1306 */
1307 void
xfs_log_sb(struct xfs_trans * tp)1308 xfs_log_sb(
1309 struct xfs_trans *tp)
1310 {
1311 struct xfs_mount *mp = tp->t_mountp;
1312 struct xfs_buf *bp = xfs_trans_getsb(tp);
1313
1314 /*
1315 * Lazy sb counters don't update the in-core superblock so do that now.
1316 * If this is at unmount, the counters will be exactly correct, but at
1317 * any other time they will only be ballpark correct because of
1318 * reservations that have been taken out percpu counters. If we have an
1319 * unclean shutdown, this will be corrected by log recovery rebuilding
1320 * the counters from the AGF block counts.
1321 */
1322 if (xfs_has_lazysbcount(mp)) {
1323 mp->m_sb.sb_icount = percpu_counter_sum_positive(&mp->m_icount);
1324 mp->m_sb.sb_ifree = min_t(uint64_t,
1325 percpu_counter_sum_positive(&mp->m_ifree),
1326 mp->m_sb.sb_icount);
1327 mp->m_sb.sb_fdblocks = xfs_sum_freecounter(mp, XC_FREE_BLOCKS);
1328 }
1329
1330 /*
1331 * sb_frextents was added to the lazy sb counters when the rt groups
1332 * feature was introduced. This counter can go negative due to the way
1333 * we handle nearly-lockless reservations, so we must use the _positive
1334 * variant here to avoid writing out nonsense frextents.
1335 */
1336 if (xfs_has_rtgroups(mp) && !xfs_has_zoned(mp)) {
1337 mp->m_sb.sb_frextents =
1338 xfs_sum_freecounter(mp, XC_FREE_RTEXTENTS);
1339 }
1340
1341 xfs_sb_to_disk(bp->b_addr, &mp->m_sb);
1342 xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
1343 xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb) - 1);
1344 }
1345
1346 /*
1347 * xfs_sync_sb
1348 *
1349 * Sync the superblock to disk.
1350 *
1351 * Note that the caller is responsible for checking the frozen state of the
1352 * filesystem. This procedure uses the non-blocking transaction allocator and
1353 * thus will allow modifications to a frozen fs. This is required because this
1354 * code can be called during the process of freezing where use of the high-level
1355 * allocator would deadlock.
1356 */
1357 int
xfs_sync_sb(struct xfs_mount * mp,bool wait)1358 xfs_sync_sb(
1359 struct xfs_mount *mp,
1360 bool wait)
1361 {
1362 struct xfs_trans *tp;
1363 int error;
1364
1365 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0,
1366 XFS_TRANS_NO_WRITECOUNT, &tp);
1367 if (error)
1368 return error;
1369
1370 xfs_log_sb(tp);
1371 if (wait)
1372 xfs_trans_set_sync(tp);
1373 return xfs_trans_commit(tp);
1374 }
1375
1376 /*
1377 * Update all the secondary superblocks to match the new state of the primary.
1378 * Because we are completely overwriting all the existing fields in the
1379 * secondary superblock buffers, there is no need to read them in from disk.
1380 * Just get a new buffer, stamp it and write it.
1381 *
1382 * The sb buffers need to be cached here so that we serialise against other
1383 * operations that access the secondary superblocks, but we don't want to keep
1384 * them in memory once it is written so we mark it as a one-shot buffer.
1385 */
1386 int
xfs_update_secondary_sbs(struct xfs_mount * mp)1387 xfs_update_secondary_sbs(
1388 struct xfs_mount *mp)
1389 {
1390 struct xfs_perag *pag = NULL;
1391 int saved_error = 0;
1392 int error = 0;
1393 LIST_HEAD (buffer_list);
1394
1395 /* update secondary superblocks. */
1396 while ((pag = xfs_perag_next_from(mp, pag, 1))) {
1397 struct xfs_buf *bp;
1398
1399 error = xfs_buf_get(mp->m_ddev_targp,
1400 XFS_AG_DADDR(mp, pag_agno(pag), XFS_SB_DADDR),
1401 XFS_FSS_TO_BB(mp, 1), &bp);
1402 /*
1403 * If we get an error reading or writing alternate superblocks,
1404 * continue. xfs_repair chooses the "best" superblock based
1405 * on most matches; if we break early, we'll leave more
1406 * superblocks un-updated than updated, and xfs_repair may
1407 * pick them over the properly-updated primary.
1408 */
1409 if (error) {
1410 xfs_warn(mp,
1411 "error allocating secondary superblock for ag %d",
1412 pag_agno(pag));
1413 if (!saved_error)
1414 saved_error = error;
1415 continue;
1416 }
1417
1418 bp->b_ops = &xfs_sb_buf_ops;
1419 xfs_buf_oneshot(bp);
1420 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
1421 xfs_sb_to_disk(bp->b_addr, &mp->m_sb);
1422 xfs_buf_delwri_queue(bp, &buffer_list);
1423 xfs_buf_relse(bp);
1424
1425 /* don't hold too many buffers at once */
1426 if (pag_agno(pag) % 16)
1427 continue;
1428
1429 error = xfs_buf_delwri_submit(&buffer_list);
1430 if (error) {
1431 xfs_warn(mp,
1432 "write error %d updating a secondary superblock near ag %d",
1433 error, pag_agno(pag));
1434 if (!saved_error)
1435 saved_error = error;
1436 continue;
1437 }
1438 }
1439 error = xfs_buf_delwri_submit(&buffer_list);
1440 if (error)
1441 xfs_warn(mp, "error %d writing secondary superblocks", error);
1442 return saved_error ? saved_error : error;
1443 }
1444
1445 /*
1446 * Same behavior as xfs_sync_sb, except that it is always synchronous and it
1447 * also writes the superblock buffer to disk sector 0 immediately.
1448 */
1449 int
xfs_sync_sb_buf(struct xfs_mount * mp,bool update_rtsb)1450 xfs_sync_sb_buf(
1451 struct xfs_mount *mp,
1452 bool update_rtsb)
1453 {
1454 struct xfs_trans *tp;
1455 struct xfs_buf *bp;
1456 struct xfs_buf *rtsb_bp = NULL;
1457 int error;
1458
1459 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0, 0, &tp);
1460 if (error)
1461 return error;
1462
1463 bp = xfs_trans_getsb(tp);
1464 xfs_log_sb(tp);
1465 xfs_trans_bhold(tp, bp);
1466 if (update_rtsb) {
1467 rtsb_bp = xfs_log_rtsb(tp, bp);
1468 if (rtsb_bp)
1469 xfs_trans_bhold(tp, rtsb_bp);
1470 }
1471 xfs_trans_set_sync(tp);
1472 error = xfs_trans_commit(tp);
1473 if (error)
1474 goto out;
1475 /*
1476 * write out the sb buffer to get the changes to disk
1477 */
1478 error = xfs_bwrite(bp);
1479 if (!error && rtsb_bp)
1480 error = xfs_bwrite(rtsb_bp);
1481 out:
1482 if (rtsb_bp)
1483 xfs_buf_relse(rtsb_bp);
1484 xfs_buf_relse(bp);
1485 return error;
1486 }
1487
1488 void
xfs_fs_geometry(struct xfs_mount * mp,struct xfs_fsop_geom * geo,int struct_version)1489 xfs_fs_geometry(
1490 struct xfs_mount *mp,
1491 struct xfs_fsop_geom *geo,
1492 int struct_version)
1493 {
1494 struct xfs_sb *sbp = &mp->m_sb;
1495
1496 memset(geo, 0, sizeof(struct xfs_fsop_geom));
1497
1498 geo->blocksize = sbp->sb_blocksize;
1499 geo->rtextsize = sbp->sb_rextsize;
1500 geo->agblocks = sbp->sb_agblocks;
1501 geo->agcount = sbp->sb_agcount;
1502 geo->logblocks = sbp->sb_logblocks;
1503 geo->sectsize = sbp->sb_sectsize;
1504 geo->inodesize = sbp->sb_inodesize;
1505 geo->imaxpct = sbp->sb_imax_pct;
1506 geo->datablocks = sbp->sb_dblocks;
1507 geo->rtblocks = sbp->sb_rblocks;
1508 geo->rtextents = sbp->sb_rextents;
1509 geo->logstart = sbp->sb_logstart;
1510 BUILD_BUG_ON(sizeof(geo->uuid) != sizeof(sbp->sb_uuid));
1511 memcpy(geo->uuid, &sbp->sb_uuid, sizeof(sbp->sb_uuid));
1512
1513 if (struct_version < 2)
1514 return;
1515
1516 geo->sunit = sbp->sb_unit;
1517 geo->swidth = sbp->sb_width;
1518
1519 if (struct_version < 3)
1520 return;
1521
1522 geo->version = XFS_FSOP_GEOM_VERSION;
1523 geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK |
1524 XFS_FSOP_GEOM_FLAGS_DIRV2 |
1525 XFS_FSOP_GEOM_FLAGS_EXTFLG |
1526 XFS_FSOP_GEOM_FLAGS_ATTR2;
1527 if (xfs_has_attr(mp))
1528 geo->flags |= XFS_FSOP_GEOM_FLAGS_ATTR;
1529 if (xfs_has_quota(mp))
1530 geo->flags |= XFS_FSOP_GEOM_FLAGS_QUOTA;
1531 if (xfs_has_align(mp))
1532 geo->flags |= XFS_FSOP_GEOM_FLAGS_IALIGN;
1533 if (xfs_has_dalign(mp))
1534 geo->flags |= XFS_FSOP_GEOM_FLAGS_DALIGN;
1535 if (xfs_has_asciici(mp))
1536 geo->flags |= XFS_FSOP_GEOM_FLAGS_DIRV2CI;
1537 if (xfs_has_lazysbcount(mp))
1538 geo->flags |= XFS_FSOP_GEOM_FLAGS_LAZYSB;
1539 if (xfs_has_projid32(mp))
1540 geo->flags |= XFS_FSOP_GEOM_FLAGS_PROJID32;
1541 if (xfs_has_crc(mp))
1542 geo->flags |= XFS_FSOP_GEOM_FLAGS_V5SB;
1543 if (xfs_has_ftype(mp))
1544 geo->flags |= XFS_FSOP_GEOM_FLAGS_FTYPE;
1545 if (xfs_has_finobt(mp))
1546 geo->flags |= XFS_FSOP_GEOM_FLAGS_FINOBT;
1547 if (xfs_has_sparseinodes(mp))
1548 geo->flags |= XFS_FSOP_GEOM_FLAGS_SPINODES;
1549 if (xfs_has_rmapbt(mp))
1550 geo->flags |= XFS_FSOP_GEOM_FLAGS_RMAPBT;
1551 if (xfs_has_reflink(mp))
1552 geo->flags |= XFS_FSOP_GEOM_FLAGS_REFLINK;
1553 if (xfs_has_bigtime(mp))
1554 geo->flags |= XFS_FSOP_GEOM_FLAGS_BIGTIME;
1555 if (xfs_has_inobtcounts(mp))
1556 geo->flags |= XFS_FSOP_GEOM_FLAGS_INOBTCNT;
1557 if (xfs_has_parent(mp))
1558 geo->flags |= XFS_FSOP_GEOM_FLAGS_PARENT;
1559 if (xfs_has_sector(mp)) {
1560 geo->flags |= XFS_FSOP_GEOM_FLAGS_SECTOR;
1561 geo->logsectsize = sbp->sb_logsectsize;
1562 } else {
1563 geo->logsectsize = BBSIZE;
1564 }
1565 if (xfs_has_large_extent_counts(mp))
1566 geo->flags |= XFS_FSOP_GEOM_FLAGS_NREXT64;
1567 if (xfs_has_exchange_range(mp))
1568 geo->flags |= XFS_FSOP_GEOM_FLAGS_EXCHANGE_RANGE;
1569 if (xfs_has_metadir(mp))
1570 geo->flags |= XFS_FSOP_GEOM_FLAGS_METADIR;
1571 if (xfs_has_zoned(mp))
1572 geo->flags |= XFS_FSOP_GEOM_FLAGS_ZONED;
1573 geo->rtsectsize = sbp->sb_blocksize;
1574 geo->dirblocksize = xfs_dir2_dirblock_bytes(sbp);
1575
1576 if (struct_version < 4)
1577 return;
1578
1579 if (xfs_has_logv2(mp))
1580 geo->flags |= XFS_FSOP_GEOM_FLAGS_LOGV2;
1581
1582 geo->logsunit = sbp->sb_logsunit;
1583
1584 if (struct_version < 5)
1585 return;
1586
1587 geo->version = XFS_FSOP_GEOM_VERSION_V5;
1588
1589 if (xfs_has_rtgroups(mp)) {
1590 geo->rgcount = sbp->sb_rgcount;
1591 geo->rgextents = sbp->sb_rgextents;
1592 }
1593 if (xfs_has_zoned(mp)) {
1594 geo->rtstart = sbp->sb_rtstart;
1595 geo->rtreserved = sbp->sb_rtreserved;
1596 }
1597 }
1598
1599 /* Read a secondary superblock. */
1600 int
xfs_sb_read_secondary(struct xfs_mount * mp,struct xfs_trans * tp,xfs_agnumber_t agno,struct xfs_buf ** bpp)1601 xfs_sb_read_secondary(
1602 struct xfs_mount *mp,
1603 struct xfs_trans *tp,
1604 xfs_agnumber_t agno,
1605 struct xfs_buf **bpp)
1606 {
1607 struct xfs_buf *bp;
1608 int error;
1609
1610 ASSERT(agno != 0 && agno != NULLAGNUMBER);
1611 error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
1612 XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
1613 XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_sb_buf_ops);
1614 if (xfs_metadata_is_sick(error))
1615 xfs_agno_mark_sick(mp, agno, XFS_SICK_AG_SB);
1616 if (error)
1617 return error;
1618 xfs_buf_set_ref(bp, XFS_SSB_REF);
1619 *bpp = bp;
1620 return 0;
1621 }
1622
1623 /* Get an uninitialised secondary superblock buffer. */
1624 int
xfs_sb_get_secondary(struct xfs_mount * mp,struct xfs_trans * tp,xfs_agnumber_t agno,struct xfs_buf ** bpp)1625 xfs_sb_get_secondary(
1626 struct xfs_mount *mp,
1627 struct xfs_trans *tp,
1628 xfs_agnumber_t agno,
1629 struct xfs_buf **bpp)
1630 {
1631 struct xfs_buf *bp;
1632 int error;
1633
1634 ASSERT(agno != 0 && agno != NULLAGNUMBER);
1635 error = xfs_trans_get_buf(tp, mp->m_ddev_targp,
1636 XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
1637 XFS_FSS_TO_BB(mp, 1), 0, &bp);
1638 if (error)
1639 return error;
1640 bp->b_ops = &xfs_sb_buf_ops;
1641 xfs_buf_oneshot(bp);
1642 *bpp = bp;
1643 return 0;
1644 }
1645
1646 /*
1647 * sunit, swidth, sectorsize(optional with 0) should be all in bytes, so users
1648 * won't be confused by values in error messages. This function returns false
1649 * if the stripe geometry is invalid and the caller is unable to repair the
1650 * stripe configuration later in the mount process.
1651 */
1652 bool
xfs_validate_stripe_geometry(struct xfs_mount * mp,__s64 sunit,__s64 swidth,int sectorsize,bool may_repair,bool silent)1653 xfs_validate_stripe_geometry(
1654 struct xfs_mount *mp,
1655 __s64 sunit,
1656 __s64 swidth,
1657 int sectorsize,
1658 bool may_repair,
1659 bool silent)
1660 {
1661 if (swidth > INT_MAX) {
1662 if (!silent)
1663 xfs_notice(mp,
1664 "stripe width (%lld) is too large", swidth);
1665 goto check_override;
1666 }
1667
1668 if (sunit > swidth) {
1669 if (!silent)
1670 xfs_notice(mp,
1671 "stripe unit (%lld) is larger than the stripe width (%lld)", sunit, swidth);
1672 goto check_override;
1673 }
1674
1675 if (sectorsize && (int)sunit % sectorsize) {
1676 if (!silent)
1677 xfs_notice(mp,
1678 "stripe unit (%lld) must be a multiple of the sector size (%d)",
1679 sunit, sectorsize);
1680 goto check_override;
1681 }
1682
1683 if (sunit && !swidth) {
1684 if (!silent)
1685 xfs_notice(mp,
1686 "invalid stripe unit (%lld) and stripe width of 0", sunit);
1687 goto check_override;
1688 }
1689
1690 if (!sunit && swidth) {
1691 if (!silent)
1692 xfs_notice(mp,
1693 "invalid stripe width (%lld) and stripe unit of 0", swidth);
1694 goto check_override;
1695 }
1696
1697 if (sunit && (int)swidth % (int)sunit) {
1698 if (!silent)
1699 xfs_notice(mp,
1700 "stripe width (%lld) must be a multiple of the stripe unit (%lld)",
1701 swidth, sunit);
1702 goto check_override;
1703 }
1704 return true;
1705
1706 check_override:
1707 if (!may_repair)
1708 return false;
1709 /*
1710 * During mount, mp->m_dalign will not be set unless the sunit mount
1711 * option was set. If it was set, ignore the bad stripe alignment values
1712 * and allow the validation and overwrite later in the mount process to
1713 * attempt to overwrite the bad stripe alignment values with the values
1714 * supplied by mount options.
1715 */
1716 if (!mp->m_dalign)
1717 return false;
1718 if (!silent)
1719 xfs_notice(mp,
1720 "Will try to correct with specified mount options sunit (%d) and swidth (%d)",
1721 BBTOB(mp->m_dalign), BBTOB(mp->m_swidth));
1722 return true;
1723 }
1724
1725 /*
1726 * Compute the maximum level number of the realtime summary file, as defined by
1727 * mkfs. The historic use of highbit32 on a 64-bit quantity prohibited correct
1728 * use of rt volumes with more than 2^32 extents.
1729 */
1730 uint8_t
xfs_compute_rextslog(xfs_rtbxlen_t rtextents)1731 xfs_compute_rextslog(
1732 xfs_rtbxlen_t rtextents)
1733 {
1734 if (!rtextents)
1735 return 0;
1736 return xfs_highbit64(rtextents);
1737 }
1738