printf: Remove unused 'bprintf'
[drm/drm-misc.git] / fs / xfs / libxfs / xfs_sb.c
blobe81b240b71583e7940806badb0c0f37a9b708802
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"
33 * Physical superblock buffer manipulations. Shared with libxfs in userspace.
37 * Check that all the V4 feature bits that the V5 filesystem format requires are
38 * correctly set.
40 static bool
41 xfs_sb_validate_v5_features(
42 struct xfs_sb *sbp)
44 /* We must not have any unknown V4 feature bits set */
45 if (sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS)
46 return false;
49 * The CRC bit is considered an invalid V4 flag, so we have to add it
50 * manually to the OKBITS mask.
52 if (sbp->sb_features2 & ~(XFS_SB_VERSION2_OKBITS |
53 XFS_SB_VERSION2_CRCBIT))
54 return false;
56 /* Now check all the required V4 feature flags are set. */
58 #define V5_VERS_FLAGS (XFS_SB_VERSION_NLINKBIT | \
59 XFS_SB_VERSION_ALIGNBIT | \
60 XFS_SB_VERSION_LOGV2BIT | \
61 XFS_SB_VERSION_EXTFLGBIT | \
62 XFS_SB_VERSION_DIRV2BIT | \
63 XFS_SB_VERSION_MOREBITSBIT)
65 #define V5_FEAT_FLAGS (XFS_SB_VERSION2_LAZYSBCOUNTBIT | \
66 XFS_SB_VERSION2_ATTR2BIT | \
67 XFS_SB_VERSION2_PROJID32BIT | \
68 XFS_SB_VERSION2_CRCBIT)
70 if ((sbp->sb_versionnum & V5_VERS_FLAGS) != V5_VERS_FLAGS)
71 return false;
72 if ((sbp->sb_features2 & V5_FEAT_FLAGS) != V5_FEAT_FLAGS)
73 return false;
74 return true;
78 * We current support XFS v5 formats with known features and v4 superblocks with
79 * at least V2 directories.
81 bool
82 xfs_sb_good_version(
83 struct xfs_sb *sbp)
86 * All v5 filesystems are supported, but we must check that all the
87 * required v4 feature flags are enabled correctly as the code checks
88 * those flags and not for v5 support.
90 if (xfs_sb_is_v5(sbp))
91 return xfs_sb_validate_v5_features(sbp);
93 /* versions prior to v4 are not supported */
94 if (XFS_SB_VERSION_NUM(sbp) != XFS_SB_VERSION_4)
95 return false;
97 /* We must not have any unknown v4 feature bits set */
98 if ((sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS) ||
99 ((sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) &&
100 (sbp->sb_features2 & ~XFS_SB_VERSION2_OKBITS)))
101 return false;
103 /* V4 filesystems need v2 directories and unwritten extents */
104 if (!(sbp->sb_versionnum & XFS_SB_VERSION_DIRV2BIT))
105 return false;
106 if (!(sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT))
107 return false;
109 /* It's a supported v4 filesystem */
110 return true;
113 uint64_t
114 xfs_sb_version_to_features(
115 struct xfs_sb *sbp)
117 uint64_t features = 0;
119 /* optional V4 features */
120 if (sbp->sb_rblocks > 0)
121 features |= XFS_FEAT_REALTIME;
122 if (sbp->sb_versionnum & XFS_SB_VERSION_NLINKBIT)
123 features |= XFS_FEAT_NLINK;
124 if (sbp->sb_versionnum & XFS_SB_VERSION_ATTRBIT)
125 features |= XFS_FEAT_ATTR;
126 if (sbp->sb_versionnum & XFS_SB_VERSION_QUOTABIT)
127 features |= XFS_FEAT_QUOTA;
128 if (sbp->sb_versionnum & XFS_SB_VERSION_ALIGNBIT)
129 features |= XFS_FEAT_ALIGN;
130 if (sbp->sb_versionnum & XFS_SB_VERSION_LOGV2BIT)
131 features |= XFS_FEAT_LOGV2;
132 if (sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT)
133 features |= XFS_FEAT_DALIGN;
134 if (sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT)
135 features |= XFS_FEAT_EXTFLG;
136 if (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT)
137 features |= XFS_FEAT_SECTOR;
138 if (sbp->sb_versionnum & XFS_SB_VERSION_BORGBIT)
139 features |= XFS_FEAT_ASCIICI;
140 if (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) {
141 if (sbp->sb_features2 & XFS_SB_VERSION2_LAZYSBCOUNTBIT)
142 features |= XFS_FEAT_LAZYSBCOUNT;
143 if (sbp->sb_features2 & XFS_SB_VERSION2_ATTR2BIT)
144 features |= XFS_FEAT_ATTR2;
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;
151 if (!xfs_sb_is_v5(sbp))
152 return features;
154 /* Always on V5 features */
155 features |= XFS_FEAT_ALIGN | XFS_FEAT_LOGV2 | XFS_FEAT_EXTFLG |
156 XFS_FEAT_LAZYSBCOUNT | XFS_FEAT_ATTR2 | XFS_FEAT_PROJID32 |
157 XFS_FEAT_V3INODES | XFS_FEAT_CRC | XFS_FEAT_PQUOTINO;
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;
187 return features;
190 /* Check all the superblock fields we care about when reading one in. */
191 STATIC int
192 xfs_validate_sb_read(
193 struct xfs_mount *mp,
194 struct xfs_sb *sbp)
196 if (!xfs_sb_is_v5(sbp))
197 return 0;
200 * Version 5 superblock feature mask validation. Reject combinations
201 * the kernel cannot support up front before checking anything else.
203 if (xfs_sb_has_compat_feature(sbp, XFS_SB_FEAT_COMPAT_UNKNOWN)) {
204 xfs_warn(mp,
205 "Superblock has unknown compatible features (0x%x) enabled.",
206 (sbp->sb_features_compat & XFS_SB_FEAT_COMPAT_UNKNOWN));
207 xfs_warn(mp,
208 "Using a more recent kernel is recommended.");
211 if (xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
212 xfs_alert(mp,
213 "Superblock has unknown read-only compatible features (0x%x) enabled.",
214 (sbp->sb_features_ro_compat &
215 XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
216 if (!xfs_is_readonly(mp)) {
217 xfs_warn(mp,
218 "Attempted to mount read-only compatible filesystem read-write.");
219 xfs_warn(mp,
220 "Filesystem can only be safely mounted read only.");
222 return -EINVAL;
225 if (xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
226 xfs_warn(mp,
227 "Superblock has unknown incompatible features (0x%x) enabled.",
228 (sbp->sb_features_incompat &
229 XFS_SB_FEAT_INCOMPAT_UNKNOWN));
230 xfs_warn(mp,
231 "Filesystem cannot be safely mounted by this kernel.");
232 return -EINVAL;
235 return 0;
238 /* Return the number of extents covered by a single rt bitmap file */
239 static xfs_rtbxlen_t
240 xfs_extents_per_rbm(
241 struct xfs_sb *sbp)
243 if (xfs_sb_is_v5(sbp) &&
244 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR))
245 return sbp->sb_rgextents;
246 return sbp->sb_rextents;
250 * Return the payload size of a single rt bitmap block (without the metadata
251 * header if any).
253 static inline unsigned int
254 xfs_rtbmblock_size(
255 struct xfs_sb *sbp)
257 if (xfs_sb_is_v5(sbp) &&
258 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR))
259 return sbp->sb_blocksize - sizeof(struct xfs_rtbuf_blkinfo);
260 return sbp->sb_blocksize;
263 static uint64_t
264 xfs_expected_rbmblocks(
265 struct xfs_sb *sbp)
267 return howmany_64(xfs_extents_per_rbm(sbp),
268 NBBY * xfs_rtbmblock_size(sbp));
271 /* Validate the realtime geometry */
272 bool
273 xfs_validate_rt_geometry(
274 struct xfs_sb *sbp)
276 if (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE ||
277 sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE)
278 return false;
280 if (sbp->sb_rblocks == 0) {
281 if (sbp->sb_rextents != 0 || sbp->sb_rbmblocks != 0 ||
282 sbp->sb_rextslog != 0 || sbp->sb_frextents != 0)
283 return false;
284 return true;
287 if (sbp->sb_rextents == 0 ||
288 sbp->sb_rextents != div_u64(sbp->sb_rblocks, sbp->sb_rextsize) ||
289 sbp->sb_rextslog != xfs_compute_rextslog(sbp->sb_rextents) ||
290 sbp->sb_rbmblocks != xfs_expected_rbmblocks(sbp))
291 return false;
293 return true;
296 /* Check all the superblock fields we care about when writing one out. */
297 STATIC int
298 xfs_validate_sb_write(
299 struct xfs_mount *mp,
300 struct xfs_buf *bp,
301 struct xfs_sb *sbp)
304 * Carry out additional sb summary counter sanity checks when we write
305 * the superblock. We skip this in the read validator because there
306 * could be newer superblocks in the log and if the values are garbage
307 * even after replay we'll recalculate them at the end of log mount.
309 * mkfs has traditionally written zeroed counters to inprogress and
310 * secondary superblocks, so allow this usage to continue because
311 * we never read counters from such superblocks.
313 if (xfs_buf_daddr(bp) == XFS_SB_DADDR && !sbp->sb_inprogress &&
314 (sbp->sb_fdblocks > sbp->sb_dblocks ||
315 !xfs_verify_icount(mp, sbp->sb_icount) ||
316 sbp->sb_ifree > sbp->sb_icount)) {
317 xfs_warn(mp, "SB summary counter sanity check failed");
318 return -EFSCORRUPTED;
321 if (!xfs_sb_is_v5(sbp))
322 return 0;
325 * Version 5 superblock feature mask validation. Reject combinations
326 * the kernel cannot support since we checked for unsupported bits in
327 * the read verifier, which means that memory is corrupt.
329 if (xfs_sb_has_compat_feature(sbp, XFS_SB_FEAT_COMPAT_UNKNOWN)) {
330 xfs_warn(mp,
331 "Corruption detected in superblock compatible features (0x%x)!",
332 (sbp->sb_features_compat & XFS_SB_FEAT_COMPAT_UNKNOWN));
333 return -EFSCORRUPTED;
336 if (!xfs_is_readonly(mp) &&
337 xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
338 xfs_alert(mp,
339 "Corruption detected in superblock read-only compatible features (0x%x)!",
340 (sbp->sb_features_ro_compat &
341 XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
342 return -EFSCORRUPTED;
344 if (xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
345 xfs_warn(mp,
346 "Corruption detected in superblock incompatible features (0x%x)!",
347 (sbp->sb_features_incompat &
348 XFS_SB_FEAT_INCOMPAT_UNKNOWN));
349 return -EFSCORRUPTED;
351 if (xfs_sb_has_incompat_log_feature(sbp,
352 XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN)) {
353 xfs_warn(mp,
354 "Corruption detected in superblock incompatible log features (0x%x)!",
355 (sbp->sb_features_log_incompat &
356 XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN));
357 return -EFSCORRUPTED;
361 * We can't read verify the sb LSN because the read verifier is called
362 * before the log is allocated and processed. We know the log is set up
363 * before write verifier calls, so check it here.
365 if (!xfs_log_check_lsn(mp, sbp->sb_lsn))
366 return -EFSCORRUPTED;
368 return 0;
372 xfs_compute_rgblklog(
373 xfs_rtxlen_t rgextents,
374 xfs_rgblock_t rextsize)
376 uint64_t rgblocks = (uint64_t)rgextents * rextsize;
378 return xfs_highbit64(rgblocks - 1) + 1;
381 static int
382 xfs_validate_sb_rtgroups(
383 struct xfs_mount *mp,
384 struct xfs_sb *sbp)
386 uint64_t groups;
387 int rgblklog;
389 if (sbp->sb_rextsize == 0) {
390 xfs_warn(mp,
391 "Realtime extent size must not be zero.");
392 return -EINVAL;
395 if (sbp->sb_rgextents > XFS_MAX_RGBLOCKS / sbp->sb_rextsize) {
396 xfs_warn(mp,
397 "Realtime group size (%u) must be less than %u rt extents.",
398 sbp->sb_rgextents,
399 XFS_MAX_RGBLOCKS / sbp->sb_rextsize);
400 return -EINVAL;
403 if (sbp->sb_rgextents < XFS_MIN_RGEXTENTS) {
404 xfs_warn(mp,
405 "Realtime group size (%u) must be at least %u rt extents.",
406 sbp->sb_rgextents, XFS_MIN_RGEXTENTS);
407 return -EINVAL;
410 if (sbp->sb_rgcount > XFS_MAX_RGNUMBER) {
411 xfs_warn(mp,
412 "Realtime groups (%u) must be less than %u.",
413 sbp->sb_rgcount, XFS_MAX_RGNUMBER);
414 return -EINVAL;
417 groups = howmany_64(sbp->sb_rextents, sbp->sb_rgextents);
418 if (groups != sbp->sb_rgcount) {
419 xfs_warn(mp,
420 "Realtime groups (%u) do not cover the entire rt section; need (%llu) groups.",
421 sbp->sb_rgcount, groups);
422 return -EINVAL;
425 /* Exchange-range is required for fsr to work on realtime files */
426 if (!(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_EXCHRANGE)) {
427 xfs_warn(mp,
428 "Realtime groups feature requires exchange-range support.");
429 return -EINVAL;
432 rgblklog = xfs_compute_rgblklog(sbp->sb_rgextents, sbp->sb_rextsize);
433 if (sbp->sb_rgblklog != rgblklog) {
434 xfs_warn(mp,
435 "Realtime group log (%d) does not match expected value (%d).",
436 sbp->sb_rgblklog, rgblklog);
437 return -EINVAL;
440 return 0;
443 /* Check the validity of the SB. */
444 STATIC int
445 xfs_validate_sb_common(
446 struct xfs_mount *mp,
447 struct xfs_buf *bp,
448 struct xfs_sb *sbp)
450 struct xfs_dsb *dsb = bp->b_addr;
451 uint32_t agcount = 0;
452 uint32_t rem;
453 bool has_dalign;
454 int error;
456 if (!xfs_verify_magic(bp, dsb->sb_magicnum)) {
457 xfs_warn(mp,
458 "Superblock has bad magic number 0x%x. Not an XFS filesystem?",
459 be32_to_cpu(dsb->sb_magicnum));
460 return -EWRONGFS;
463 if (!xfs_sb_good_version(sbp)) {
464 xfs_warn(mp,
465 "Superblock has unknown features enabled or corrupted feature masks.");
466 return -EWRONGFS;
470 * Validate feature flags and state
472 if (xfs_sb_is_v5(sbp)) {
473 if (sbp->sb_blocksize < XFS_MIN_CRC_BLOCKSIZE) {
474 xfs_notice(mp,
475 "Block size (%u bytes) too small for Version 5 superblock (minimum %d bytes)",
476 sbp->sb_blocksize, XFS_MIN_CRC_BLOCKSIZE);
477 return -EFSCORRUPTED;
480 /* V5 has a separate project quota inode */
481 if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
482 xfs_notice(mp,
483 "Version 5 of Super block has XFS_OQUOTA bits.");
484 return -EFSCORRUPTED;
488 * Full inode chunks must be aligned to inode chunk size when
489 * sparse inodes are enabled to support the sparse chunk
490 * allocation algorithm and prevent overlapping inode records.
492 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_SPINODES) {
493 uint32_t align;
495 align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize
496 >> sbp->sb_blocklog;
497 if (sbp->sb_inoalignmt != align) {
498 xfs_warn(mp,
499 "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.",
500 sbp->sb_inoalignmt, align);
501 return -EINVAL;
504 if (!sbp->sb_spino_align ||
505 sbp->sb_spino_align > sbp->sb_inoalignmt ||
506 (sbp->sb_inoalignmt % sbp->sb_spino_align) != 0) {
507 xfs_warn(mp,
508 "Sparse inode alignment (%u) is invalid.",
509 sbp->sb_spino_align);
510 return -EINVAL;
512 } else if (sbp->sb_spino_align) {
513 xfs_warn(mp,
514 "Sparse inode alignment (%u) should be zero.",
515 sbp->sb_spino_align);
516 return -EINVAL;
519 if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR) {
520 if (memchr_inv(sbp->sb_pad, 0, sizeof(sbp->sb_pad))) {
521 xfs_warn(mp,
522 "Metadir superblock padding fields must be zero.");
523 return -EINVAL;
526 error = xfs_validate_sb_rtgroups(mp, sbp);
527 if (error)
528 return error;
530 } else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
531 XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
532 xfs_notice(mp,
533 "Superblock earlier than Version 5 has XFS_{P|G}QUOTA_{ENFD|CHKD} bits.");
534 return -EFSCORRUPTED;
537 if (unlikely(
538 sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
539 xfs_warn(mp,
540 "filesystem is marked as having an external log; "
541 "specify logdev on the mount command line.");
542 return -EINVAL;
545 if (unlikely(
546 sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
547 xfs_warn(mp,
548 "filesystem is marked as having an internal log; "
549 "do not specify logdev on the mount command line.");
550 return -EINVAL;
553 /* Compute agcount for this number of dblocks and agblocks */
554 if (sbp->sb_agblocks) {
555 agcount = div_u64_rem(sbp->sb_dblocks, sbp->sb_agblocks, &rem);
556 if (rem)
557 agcount++;
561 * More sanity checking. Most of these were stolen directly from
562 * xfs_repair.
564 if (unlikely(
565 sbp->sb_agcount <= 0 ||
566 sbp->sb_sectsize < XFS_MIN_SECTORSIZE ||
567 sbp->sb_sectsize > XFS_MAX_SECTORSIZE ||
568 sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG ||
569 sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG ||
570 sbp->sb_sectsize != (1 << sbp->sb_sectlog) ||
571 sbp->sb_blocksize < XFS_MIN_BLOCKSIZE ||
572 sbp->sb_blocksize > XFS_MAX_BLOCKSIZE ||
573 sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG ||
574 sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
575 sbp->sb_blocksize != (1 << sbp->sb_blocklog) ||
576 sbp->sb_dirblklog + sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
577 sbp->sb_inodesize < XFS_DINODE_MIN_SIZE ||
578 sbp->sb_inodesize > XFS_DINODE_MAX_SIZE ||
579 sbp->sb_inodelog < XFS_DINODE_MIN_LOG ||
580 sbp->sb_inodelog > XFS_DINODE_MAX_LOG ||
581 sbp->sb_inodesize != (1 << sbp->sb_inodelog) ||
582 sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
583 XFS_FSB_TO_B(mp, sbp->sb_agblocks) < XFS_MIN_AG_BYTES ||
584 XFS_FSB_TO_B(mp, sbp->sb_agblocks) > XFS_MAX_AG_BYTES ||
585 sbp->sb_agblklog != xfs_highbit32(sbp->sb_agblocks - 1) + 1 ||
586 agcount == 0 || agcount != sbp->sb_agcount ||
587 (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog) ||
588 (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) ||
589 (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) ||
590 (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */) ||
591 sbp->sb_dblocks == 0 ||
592 sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) ||
593 sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp) ||
594 sbp->sb_shared_vn != 0)) {
595 xfs_notice(mp, "SB sanity check failed");
596 return -EFSCORRUPTED;
600 * Logs that are too large are not supported at all. Reject them
601 * outright. Logs that are too small are tolerated on v4 filesystems,
602 * but we can only check that when mounting the log. Hence we skip
603 * those checks here.
605 if (sbp->sb_logblocks > XFS_MAX_LOG_BLOCKS) {
606 xfs_notice(mp,
607 "Log size 0x%x blocks too large, maximum size is 0x%llx blocks",
608 sbp->sb_logblocks, XFS_MAX_LOG_BLOCKS);
609 return -EFSCORRUPTED;
612 if (XFS_FSB_TO_B(mp, sbp->sb_logblocks) > XFS_MAX_LOG_BYTES) {
613 xfs_warn(mp,
614 "log size 0x%llx bytes too large, maximum size is 0x%llx bytes",
615 XFS_FSB_TO_B(mp, sbp->sb_logblocks),
616 XFS_MAX_LOG_BYTES);
617 return -EFSCORRUPTED;
621 * Do not allow filesystems with corrupted log sector or stripe units to
622 * be mounted. We cannot safely size the iclogs or write to the log if
623 * the log stripe unit is not valid.
625 if (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT) {
626 if (sbp->sb_logsectsize != (1U << sbp->sb_logsectlog)) {
627 xfs_notice(mp,
628 "log sector size in bytes/log2 (0x%x/0x%x) must match",
629 sbp->sb_logsectsize, 1U << sbp->sb_logsectlog);
630 return -EFSCORRUPTED;
632 } else if (sbp->sb_logsectsize || sbp->sb_logsectlog) {
633 xfs_notice(mp,
634 "log sector size in bytes/log2 (0x%x/0x%x) are not zero",
635 sbp->sb_logsectsize, sbp->sb_logsectlog);
636 return -EFSCORRUPTED;
639 if (sbp->sb_logsunit > 1) {
640 if (sbp->sb_logsunit % sbp->sb_blocksize) {
641 xfs_notice(mp,
642 "log stripe unit 0x%x bytes must be a multiple of block size",
643 sbp->sb_logsunit);
644 return -EFSCORRUPTED;
646 if (sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE) {
647 xfs_notice(mp,
648 "log stripe unit 0x%x bytes over maximum size (0x%x bytes)",
649 sbp->sb_logsunit, XLOG_MAX_RECORD_BSIZE);
650 return -EFSCORRUPTED;
654 if (!xfs_validate_rt_geometry(sbp)) {
655 xfs_notice(mp,
656 "realtime %sgeometry check failed",
657 sbp->sb_rblocks ? "" : "zeroed ");
658 return -EFSCORRUPTED;
662 * Either (sb_unit and !hasdalign) or (!sb_unit and hasdalign)
663 * would imply the image is corrupted.
665 has_dalign = sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT;
666 if (!!sbp->sb_unit ^ has_dalign) {
667 xfs_notice(mp, "SB stripe alignment sanity check failed");
668 return -EFSCORRUPTED;
671 if (!xfs_validate_stripe_geometry(mp, XFS_FSB_TO_B(mp, sbp->sb_unit),
672 XFS_FSB_TO_B(mp, sbp->sb_width), 0,
673 xfs_buf_daddr(bp) == XFS_SB_DADDR, false))
674 return -EFSCORRUPTED;
677 * Currently only very few inode sizes are supported.
679 switch (sbp->sb_inodesize) {
680 case 256:
681 case 512:
682 case 1024:
683 case 2048:
684 break;
685 default:
686 xfs_warn(mp, "inode size of %d bytes not supported",
687 sbp->sb_inodesize);
688 return -ENOSYS;
691 return 0;
694 void
695 xfs_sb_quota_from_disk(struct xfs_sb *sbp)
697 if (xfs_sb_is_v5(sbp) &&
698 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR)) {
699 sbp->sb_uquotino = NULLFSINO;
700 sbp->sb_gquotino = NULLFSINO;
701 sbp->sb_pquotino = NULLFSINO;
702 return;
706 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
707 * leads to in-core values having two different values for a quota
708 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
709 * NULLFSINO.
711 * Note that this change affect only the in-core values. These
712 * values are not written back to disk unless any quota information
713 * is written to the disk. Even in that case, sb_pquotino field is
714 * not written to disk unless the superblock supports pquotino.
716 if (sbp->sb_uquotino == 0)
717 sbp->sb_uquotino = NULLFSINO;
718 if (sbp->sb_gquotino == 0)
719 sbp->sb_gquotino = NULLFSINO;
720 if (sbp->sb_pquotino == 0)
721 sbp->sb_pquotino = NULLFSINO;
724 * We need to do these manipilations only if we are working
725 * with an older version of on-disk superblock.
727 if (xfs_sb_is_v5(sbp))
728 return;
730 if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
731 sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
732 XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
733 if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
734 sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
735 XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
736 sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
738 if (sbp->sb_qflags & XFS_PQUOTA_ACCT &&
739 sbp->sb_gquotino != NULLFSINO) {
741 * In older version of superblock, on-disk superblock only
742 * has sb_gquotino, and in-core superblock has both sb_gquotino
743 * and sb_pquotino. But, only one of them is supported at any
744 * point of time. So, if PQUOTA is set in disk superblock,
745 * copy over sb_gquotino to sb_pquotino. The NULLFSINO test
746 * above is to make sure we don't do this twice and wipe them
747 * both out!
749 sbp->sb_pquotino = sbp->sb_gquotino;
750 sbp->sb_gquotino = NULLFSINO;
754 static void
755 __xfs_sb_from_disk(
756 struct xfs_sb *to,
757 struct xfs_dsb *from,
758 bool convert_xquota)
760 to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
761 to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
762 to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
763 to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
764 to->sb_rextents = be64_to_cpu(from->sb_rextents);
765 memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
766 to->sb_logstart = be64_to_cpu(from->sb_logstart);
767 to->sb_rootino = be64_to_cpu(from->sb_rootino);
768 to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
769 to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
770 to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
771 to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
772 to->sb_agcount = be32_to_cpu(from->sb_agcount);
773 to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
774 to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
775 to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
776 to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
777 to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
778 to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
779 memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
780 to->sb_blocklog = from->sb_blocklog;
781 to->sb_sectlog = from->sb_sectlog;
782 to->sb_inodelog = from->sb_inodelog;
783 to->sb_inopblog = from->sb_inopblog;
784 to->sb_agblklog = from->sb_agblklog;
785 to->sb_rextslog = from->sb_rextslog;
786 to->sb_inprogress = from->sb_inprogress;
787 to->sb_imax_pct = from->sb_imax_pct;
788 to->sb_icount = be64_to_cpu(from->sb_icount);
789 to->sb_ifree = be64_to_cpu(from->sb_ifree);
790 to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
791 to->sb_frextents = be64_to_cpu(from->sb_frextents);
792 to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
793 to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
794 to->sb_qflags = be16_to_cpu(from->sb_qflags);
795 to->sb_flags = from->sb_flags;
796 to->sb_shared_vn = from->sb_shared_vn;
797 to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
798 to->sb_unit = be32_to_cpu(from->sb_unit);
799 to->sb_width = be32_to_cpu(from->sb_width);
800 to->sb_dirblklog = from->sb_dirblklog;
801 to->sb_logsectlog = from->sb_logsectlog;
802 to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
803 to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
804 to->sb_features2 = be32_to_cpu(from->sb_features2);
805 to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
806 to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
807 to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
808 to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
809 to->sb_features_log_incompat =
810 be32_to_cpu(from->sb_features_log_incompat);
811 /* crc is only used on disk, not in memory; just init to 0 here. */
812 to->sb_crc = 0;
813 to->sb_spino_align = be32_to_cpu(from->sb_spino_align);
814 to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
815 to->sb_lsn = be64_to_cpu(from->sb_lsn);
817 * sb_meta_uuid is only on disk if it differs from sb_uuid and the
818 * feature flag is set; if not set we keep it only in memory.
820 if (xfs_sb_is_v5(to) &&
821 (to->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID))
822 uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
823 else
824 uuid_copy(&to->sb_meta_uuid, &from->sb_uuid);
825 /* Convert on-disk flags to in-memory flags? */
826 if (convert_xquota)
827 xfs_sb_quota_from_disk(to);
829 if (to->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR) {
830 to->sb_metadirino = be64_to_cpu(from->sb_metadirino);
831 to->sb_rgblklog = from->sb_rgblklog;
832 memcpy(to->sb_pad, from->sb_pad, sizeof(to->sb_pad));
833 to->sb_rgcount = be32_to_cpu(from->sb_rgcount);
834 to->sb_rgextents = be32_to_cpu(from->sb_rgextents);
835 to->sb_rbmino = NULLFSINO;
836 to->sb_rsumino = NULLFSINO;
837 } else {
838 to->sb_metadirino = NULLFSINO;
839 to->sb_rgcount = 1;
840 to->sb_rgextents = 0;
844 void
845 xfs_sb_from_disk(
846 struct xfs_sb *to,
847 struct xfs_dsb *from)
849 __xfs_sb_from_disk(to, from, true);
852 static void
853 xfs_sb_quota_to_disk(
854 struct xfs_dsb *to,
855 struct xfs_sb *from)
857 uint16_t qflags = from->sb_qflags;
859 if (xfs_sb_is_v5(from) &&
860 (from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR)) {
861 to->sb_qflags = cpu_to_be16(from->sb_qflags);
862 to->sb_uquotino = cpu_to_be64(0);
863 to->sb_gquotino = cpu_to_be64(0);
864 to->sb_pquotino = cpu_to_be64(0);
865 return;
868 to->sb_uquotino = cpu_to_be64(from->sb_uquotino);
871 * The in-memory superblock quota state matches the v5 on-disk format so
872 * just write them out and return
874 if (xfs_sb_is_v5(from)) {
875 to->sb_qflags = cpu_to_be16(from->sb_qflags);
876 to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
877 to->sb_pquotino = cpu_to_be64(from->sb_pquotino);
878 return;
882 * For older superblocks (v4), the in-core version of sb_qflags do not
883 * have XFS_OQUOTA_* flags, whereas the on-disk version does. So,
884 * convert incore XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
886 qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
887 XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
889 if (from->sb_qflags &
890 (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
891 qflags |= XFS_OQUOTA_ENFD;
892 if (from->sb_qflags &
893 (XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
894 qflags |= XFS_OQUOTA_CHKD;
895 to->sb_qflags = cpu_to_be16(qflags);
898 * GQUOTINO and PQUOTINO cannot be used together in versions
899 * of superblock that do not have pquotino. from->sb_flags
900 * tells us which quota is active and should be copied to
901 * disk. If neither are active, we should NULL the inode.
903 * In all cases, the separate pquotino must remain 0 because it
904 * is beyond the "end" of the valid non-pquotino superblock.
906 if (from->sb_qflags & XFS_GQUOTA_ACCT)
907 to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
908 else if (from->sb_qflags & XFS_PQUOTA_ACCT)
909 to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
910 else {
912 * We can't rely on just the fields being logged to tell us
913 * that it is safe to write NULLFSINO - we should only do that
914 * if quotas are not actually enabled. Hence only write
915 * NULLFSINO if both in-core quota inodes are NULL.
917 if (from->sb_gquotino == NULLFSINO &&
918 from->sb_pquotino == NULLFSINO)
919 to->sb_gquotino = cpu_to_be64(NULLFSINO);
922 to->sb_pquotino = 0;
925 void
926 xfs_sb_to_disk(
927 struct xfs_dsb *to,
928 struct xfs_sb *from)
930 xfs_sb_quota_to_disk(to, from);
932 to->sb_magicnum = cpu_to_be32(from->sb_magicnum);
933 to->sb_blocksize = cpu_to_be32(from->sb_blocksize);
934 to->sb_dblocks = cpu_to_be64(from->sb_dblocks);
935 to->sb_rblocks = cpu_to_be64(from->sb_rblocks);
936 to->sb_rextents = cpu_to_be64(from->sb_rextents);
937 memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
938 to->sb_logstart = cpu_to_be64(from->sb_logstart);
939 to->sb_rootino = cpu_to_be64(from->sb_rootino);
940 to->sb_rbmino = cpu_to_be64(from->sb_rbmino);
941 to->sb_rsumino = cpu_to_be64(from->sb_rsumino);
942 to->sb_rextsize = cpu_to_be32(from->sb_rextsize);
943 to->sb_agblocks = cpu_to_be32(from->sb_agblocks);
944 to->sb_agcount = cpu_to_be32(from->sb_agcount);
945 to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks);
946 to->sb_logblocks = cpu_to_be32(from->sb_logblocks);
947 to->sb_versionnum = cpu_to_be16(from->sb_versionnum);
948 to->sb_sectsize = cpu_to_be16(from->sb_sectsize);
949 to->sb_inodesize = cpu_to_be16(from->sb_inodesize);
950 to->sb_inopblock = cpu_to_be16(from->sb_inopblock);
951 memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
952 to->sb_blocklog = from->sb_blocklog;
953 to->sb_sectlog = from->sb_sectlog;
954 to->sb_inodelog = from->sb_inodelog;
955 to->sb_inopblog = from->sb_inopblog;
956 to->sb_agblklog = from->sb_agblklog;
957 to->sb_rextslog = from->sb_rextslog;
958 to->sb_inprogress = from->sb_inprogress;
959 to->sb_imax_pct = from->sb_imax_pct;
960 to->sb_icount = cpu_to_be64(from->sb_icount);
961 to->sb_ifree = cpu_to_be64(from->sb_ifree);
962 to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks);
963 to->sb_frextents = cpu_to_be64(from->sb_frextents);
965 to->sb_flags = from->sb_flags;
966 to->sb_shared_vn = from->sb_shared_vn;
967 to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt);
968 to->sb_unit = cpu_to_be32(from->sb_unit);
969 to->sb_width = cpu_to_be32(from->sb_width);
970 to->sb_dirblklog = from->sb_dirblklog;
971 to->sb_logsectlog = from->sb_logsectlog;
972 to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize);
973 to->sb_logsunit = cpu_to_be32(from->sb_logsunit);
976 * We need to ensure that bad_features2 always matches features2.
977 * Hence we enforce that here rather than having to remember to do it
978 * everywhere else that updates features2.
980 from->sb_bad_features2 = from->sb_features2;
981 to->sb_features2 = cpu_to_be32(from->sb_features2);
982 to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2);
984 if (!xfs_sb_is_v5(from))
985 return;
987 to->sb_features_compat = cpu_to_be32(from->sb_features_compat);
988 to->sb_features_ro_compat =
989 cpu_to_be32(from->sb_features_ro_compat);
990 to->sb_features_incompat =
991 cpu_to_be32(from->sb_features_incompat);
992 to->sb_features_log_incompat =
993 cpu_to_be32(from->sb_features_log_incompat);
994 to->sb_spino_align = cpu_to_be32(from->sb_spino_align);
995 to->sb_lsn = cpu_to_be64(from->sb_lsn);
996 if (from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID)
997 uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
999 if (from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR) {
1000 to->sb_metadirino = cpu_to_be64(from->sb_metadirino);
1001 to->sb_rgblklog = from->sb_rgblklog;
1002 memset(to->sb_pad, 0, sizeof(to->sb_pad));
1003 to->sb_rgcount = cpu_to_be32(from->sb_rgcount);
1004 to->sb_rgextents = cpu_to_be32(from->sb_rgextents);
1005 to->sb_rbmino = cpu_to_be64(0);
1006 to->sb_rsumino = cpu_to_be64(0);
1011 * If the superblock has the CRC feature bit set or the CRC field is non-null,
1012 * check that the CRC is valid. We check the CRC field is non-null because a
1013 * single bit error could clear the feature bit and unused parts of the
1014 * superblock are supposed to be zero. Hence a non-null crc field indicates that
1015 * we've potentially lost a feature bit and we should check it anyway.
1017 * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
1018 * last field in V4 secondary superblocks. So for secondary superblocks,
1019 * we are more forgiving, and ignore CRC failures if the primary doesn't
1020 * indicate that the fs version is V5.
1022 static void
1023 xfs_sb_read_verify(
1024 struct xfs_buf *bp)
1026 struct xfs_sb sb;
1027 struct xfs_mount *mp = bp->b_mount;
1028 struct xfs_dsb *dsb = bp->b_addr;
1029 int error;
1032 * open code the version check to avoid needing to convert the entire
1033 * superblock from disk order just to check the version number
1035 if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
1036 (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
1037 XFS_SB_VERSION_5) ||
1038 dsb->sb_crc != 0)) {
1040 if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
1041 /* Only fail bad secondaries on a known V5 filesystem */
1042 if (xfs_buf_daddr(bp) == XFS_SB_DADDR ||
1043 xfs_has_crc(mp)) {
1044 error = -EFSBADCRC;
1045 goto out_error;
1051 * Check all the superblock fields. Don't byteswap the xquota flags
1052 * because _verify_common checks the on-disk values.
1054 __xfs_sb_from_disk(&sb, dsb, false);
1055 error = xfs_validate_sb_common(mp, bp, &sb);
1056 if (error)
1057 goto out_error;
1058 error = xfs_validate_sb_read(mp, &sb);
1060 out_error:
1061 if (error == -EFSCORRUPTED || error == -EFSBADCRC)
1062 xfs_verifier_error(bp, error, __this_address);
1063 else if (error)
1064 xfs_buf_ioerror(bp, error);
1068 * We may be probed for a filesystem match, so we may not want to emit
1069 * messages when the superblock buffer is not actually an XFS superblock.
1070 * If we find an XFS superblock, then run a normal, noisy mount because we are
1071 * really going to mount it and want to know about errors.
1073 static void
1074 xfs_sb_quiet_read_verify(
1075 struct xfs_buf *bp)
1077 struct xfs_dsb *dsb = bp->b_addr;
1079 if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
1080 /* XFS filesystem, verify noisily! */
1081 xfs_sb_read_verify(bp);
1082 return;
1084 /* quietly fail */
1085 xfs_buf_ioerror(bp, -EWRONGFS);
1088 static void
1089 xfs_sb_write_verify(
1090 struct xfs_buf *bp)
1092 struct xfs_sb sb;
1093 struct xfs_mount *mp = bp->b_mount;
1094 struct xfs_buf_log_item *bip = bp->b_log_item;
1095 struct xfs_dsb *dsb = bp->b_addr;
1096 int error;
1099 * Check all the superblock fields. Don't byteswap the xquota flags
1100 * because _verify_common checks the on-disk values.
1102 __xfs_sb_from_disk(&sb, dsb, false);
1103 error = xfs_validate_sb_common(mp, bp, &sb);
1104 if (error)
1105 goto out_error;
1106 error = xfs_validate_sb_write(mp, bp, &sb);
1107 if (error)
1108 goto out_error;
1110 if (!xfs_sb_is_v5(&sb))
1111 return;
1113 if (bip)
1114 dsb->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
1116 xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
1117 return;
1119 out_error:
1120 xfs_verifier_error(bp, error, __this_address);
1123 const struct xfs_buf_ops xfs_sb_buf_ops = {
1124 .name = "xfs_sb",
1125 .magic = { cpu_to_be32(XFS_SB_MAGIC), cpu_to_be32(XFS_SB_MAGIC) },
1126 .verify_read = xfs_sb_read_verify,
1127 .verify_write = xfs_sb_write_verify,
1130 const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
1131 .name = "xfs_sb_quiet",
1132 .magic = { cpu_to_be32(XFS_SB_MAGIC), cpu_to_be32(XFS_SB_MAGIC) },
1133 .verify_read = xfs_sb_quiet_read_verify,
1134 .verify_write = xfs_sb_write_verify,
1137 /* Compute cached rt geometry from the incore sb. */
1138 void
1139 xfs_sb_mount_rextsize(
1140 struct xfs_mount *mp,
1141 struct xfs_sb *sbp)
1143 struct xfs_groups *rgs = &mp->m_groups[XG_TYPE_RTG];
1145 mp->m_rtxblklog = log2_if_power2(sbp->sb_rextsize);
1146 mp->m_rtxblkmask = mask64_if_power2(sbp->sb_rextsize);
1148 if (xfs_sb_is_v5(sbp) &&
1149 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR)) {
1150 rgs->blocks = sbp->sb_rgextents * sbp->sb_rextsize;
1151 rgs->blklog = mp->m_sb.sb_rgblklog;
1152 rgs->blkmask = xfs_mask32lo(mp->m_sb.sb_rgblklog);
1153 } else {
1154 rgs->blocks = 0;
1155 rgs->blklog = 0;
1156 rgs->blkmask = (uint64_t)-1;
1160 /* Update incore sb rt extent size, then recompute the cached rt geometry. */
1161 void
1162 xfs_mount_sb_set_rextsize(
1163 struct xfs_mount *mp,
1164 struct xfs_sb *sbp,
1165 xfs_agblock_t rextsize)
1167 sbp->sb_rextsize = rextsize;
1168 if (xfs_sb_is_v5(sbp) &&
1169 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR))
1170 sbp->sb_rgblklog = xfs_compute_rgblklog(sbp->sb_rgextents,
1171 rextsize);
1173 xfs_sb_mount_rextsize(mp, sbp);
1177 * xfs_mount_common
1179 * Mount initialization code establishing various mount
1180 * fields from the superblock associated with the given
1181 * mount structure.
1183 * Inode geometry are calculated in xfs_ialloc_setup_geometry.
1185 void
1186 xfs_sb_mount_common(
1187 struct xfs_mount *mp,
1188 struct xfs_sb *sbp)
1190 struct xfs_groups *ags = &mp->m_groups[XG_TYPE_AG];
1192 mp->m_agfrotor = 0;
1193 atomic_set(&mp->m_agirotor, 0);
1194 mp->m_maxagi = mp->m_sb.sb_agcount;
1195 mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
1196 mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
1197 mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
1198 mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
1199 mp->m_blockmask = sbp->sb_blocksize - 1;
1200 mp->m_blockwsize = xfs_rtbmblock_size(sbp) >> XFS_WORDLOG;
1201 mp->m_rtx_per_rbmblock = mp->m_blockwsize << XFS_NBWORDLOG;
1203 ags->blocks = mp->m_sb.sb_agblocks;
1204 ags->blklog = mp->m_sb.sb_agblklog;
1205 ags->blkmask = xfs_mask32lo(mp->m_sb.sb_agblklog);
1207 xfs_sb_mount_rextsize(mp, sbp);
1209 mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, true);
1210 mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, false);
1211 mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
1212 mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
1214 mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, true);
1215 mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, false);
1216 mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
1217 mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
1219 mp->m_rmap_mxr[0] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, true);
1220 mp->m_rmap_mxr[1] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, false);
1221 mp->m_rmap_mnr[0] = mp->m_rmap_mxr[0] / 2;
1222 mp->m_rmap_mnr[1] = mp->m_rmap_mxr[1] / 2;
1224 mp->m_refc_mxr[0] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize, true);
1225 mp->m_refc_mxr[1] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize, false);
1226 mp->m_refc_mnr[0] = mp->m_refc_mxr[0] / 2;
1227 mp->m_refc_mnr[1] = mp->m_refc_mxr[1] / 2;
1229 mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
1230 mp->m_alloc_set_aside = xfs_alloc_set_aside(mp);
1231 mp->m_ag_max_usable = xfs_alloc_ag_max_usable(mp);
1235 * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
1236 * into the superblock buffer to be logged. It does not provide the higher
1237 * level of locking that is needed to protect the in-core superblock from
1238 * concurrent access.
1240 void
1241 xfs_log_sb(
1242 struct xfs_trans *tp)
1244 struct xfs_mount *mp = tp->t_mountp;
1245 struct xfs_buf *bp = xfs_trans_getsb(tp);
1248 * Lazy sb counters don't update the in-core superblock so do that now.
1249 * If this is at unmount, the counters will be exactly correct, but at
1250 * any other time they will only be ballpark correct because of
1251 * reservations that have been taken out percpu counters. If we have an
1252 * unclean shutdown, this will be corrected by log recovery rebuilding
1253 * the counters from the AGF block counts.
1255 if (xfs_has_lazysbcount(mp)) {
1256 mp->m_sb.sb_icount = percpu_counter_sum_positive(&mp->m_icount);
1257 mp->m_sb.sb_ifree = min_t(uint64_t,
1258 percpu_counter_sum_positive(&mp->m_ifree),
1259 mp->m_sb.sb_icount);
1260 mp->m_sb.sb_fdblocks =
1261 percpu_counter_sum_positive(&mp->m_fdblocks);
1265 * sb_frextents was added to the lazy sb counters when the rt groups
1266 * feature was introduced. This counter can go negative due to the way
1267 * we handle nearly-lockless reservations, so we must use the _positive
1268 * variant here to avoid writing out nonsense frextents.
1270 if (xfs_has_rtgroups(mp))
1271 mp->m_sb.sb_frextents =
1272 percpu_counter_sum_positive(&mp->m_frextents);
1274 xfs_sb_to_disk(bp->b_addr, &mp->m_sb);
1275 xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
1276 xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb) - 1);
1280 * xfs_sync_sb
1282 * Sync the superblock to disk.
1284 * Note that the caller is responsible for checking the frozen state of the
1285 * filesystem. This procedure uses the non-blocking transaction allocator and
1286 * thus will allow modifications to a frozen fs. This is required because this
1287 * code can be called during the process of freezing where use of the high-level
1288 * allocator would deadlock.
1291 xfs_sync_sb(
1292 struct xfs_mount *mp,
1293 bool wait)
1295 struct xfs_trans *tp;
1296 int error;
1298 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0,
1299 XFS_TRANS_NO_WRITECOUNT, &tp);
1300 if (error)
1301 return error;
1303 xfs_log_sb(tp);
1304 if (wait)
1305 xfs_trans_set_sync(tp);
1306 return xfs_trans_commit(tp);
1310 * Update all the secondary superblocks to match the new state of the primary.
1311 * Because we are completely overwriting all the existing fields in the
1312 * secondary superblock buffers, there is no need to read them in from disk.
1313 * Just get a new buffer, stamp it and write it.
1315 * The sb buffers need to be cached here so that we serialise against other
1316 * operations that access the secondary superblocks, but we don't want to keep
1317 * them in memory once it is written so we mark it as a one-shot buffer.
1320 xfs_update_secondary_sbs(
1321 struct xfs_mount *mp)
1323 struct xfs_perag *pag = NULL;
1324 int saved_error = 0;
1325 int error = 0;
1326 LIST_HEAD (buffer_list);
1328 /* update secondary superblocks. */
1329 while ((pag = xfs_perag_next_from(mp, pag, 1))) {
1330 struct xfs_buf *bp;
1332 error = xfs_buf_get(mp->m_ddev_targp,
1333 XFS_AG_DADDR(mp, pag_agno(pag), XFS_SB_DADDR),
1334 XFS_FSS_TO_BB(mp, 1), &bp);
1336 * If we get an error reading or writing alternate superblocks,
1337 * continue. xfs_repair chooses the "best" superblock based
1338 * on most matches; if we break early, we'll leave more
1339 * superblocks un-updated than updated, and xfs_repair may
1340 * pick them over the properly-updated primary.
1342 if (error) {
1343 xfs_warn(mp,
1344 "error allocating secondary superblock for ag %d",
1345 pag_agno(pag));
1346 if (!saved_error)
1347 saved_error = error;
1348 continue;
1351 bp->b_ops = &xfs_sb_buf_ops;
1352 xfs_buf_oneshot(bp);
1353 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
1354 xfs_sb_to_disk(bp->b_addr, &mp->m_sb);
1355 xfs_buf_delwri_queue(bp, &buffer_list);
1356 xfs_buf_relse(bp);
1358 /* don't hold too many buffers at once */
1359 if (pag_agno(pag) % 16)
1360 continue;
1362 error = xfs_buf_delwri_submit(&buffer_list);
1363 if (error) {
1364 xfs_warn(mp,
1365 "write error %d updating a secondary superblock near ag %d",
1366 error, pag_agno(pag));
1367 if (!saved_error)
1368 saved_error = error;
1369 continue;
1372 error = xfs_buf_delwri_submit(&buffer_list);
1373 if (error)
1374 xfs_warn(mp, "error %d writing secondary superblocks", error);
1375 return saved_error ? saved_error : error;
1379 * Same behavior as xfs_sync_sb, except that it is always synchronous and it
1380 * also writes the superblock buffer to disk sector 0 immediately.
1383 xfs_sync_sb_buf(
1384 struct xfs_mount *mp,
1385 bool update_rtsb)
1387 struct xfs_trans *tp;
1388 struct xfs_buf *bp;
1389 struct xfs_buf *rtsb_bp = NULL;
1390 int error;
1392 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0, 0, &tp);
1393 if (error)
1394 return error;
1396 bp = xfs_trans_getsb(tp);
1397 xfs_log_sb(tp);
1398 xfs_trans_bhold(tp, bp);
1399 if (update_rtsb) {
1400 rtsb_bp = xfs_log_rtsb(tp, bp);
1401 if (rtsb_bp)
1402 xfs_trans_bhold(tp, rtsb_bp);
1404 xfs_trans_set_sync(tp);
1405 error = xfs_trans_commit(tp);
1406 if (error)
1407 goto out;
1409 * write out the sb buffer to get the changes to disk
1411 error = xfs_bwrite(bp);
1412 if (!error && rtsb_bp)
1413 error = xfs_bwrite(rtsb_bp);
1414 out:
1415 if (rtsb_bp)
1416 xfs_buf_relse(rtsb_bp);
1417 xfs_buf_relse(bp);
1418 return error;
1421 void
1422 xfs_fs_geometry(
1423 struct xfs_mount *mp,
1424 struct xfs_fsop_geom *geo,
1425 int struct_version)
1427 struct xfs_sb *sbp = &mp->m_sb;
1429 memset(geo, 0, sizeof(struct xfs_fsop_geom));
1431 geo->blocksize = sbp->sb_blocksize;
1432 geo->rtextsize = sbp->sb_rextsize;
1433 geo->agblocks = sbp->sb_agblocks;
1434 geo->agcount = sbp->sb_agcount;
1435 geo->logblocks = sbp->sb_logblocks;
1436 geo->sectsize = sbp->sb_sectsize;
1437 geo->inodesize = sbp->sb_inodesize;
1438 geo->imaxpct = sbp->sb_imax_pct;
1439 geo->datablocks = sbp->sb_dblocks;
1440 geo->rtblocks = sbp->sb_rblocks;
1441 geo->rtextents = sbp->sb_rextents;
1442 geo->logstart = sbp->sb_logstart;
1443 BUILD_BUG_ON(sizeof(geo->uuid) != sizeof(sbp->sb_uuid));
1444 memcpy(geo->uuid, &sbp->sb_uuid, sizeof(sbp->sb_uuid));
1446 if (struct_version < 2)
1447 return;
1449 geo->sunit = sbp->sb_unit;
1450 geo->swidth = sbp->sb_width;
1452 if (struct_version < 3)
1453 return;
1455 geo->version = XFS_FSOP_GEOM_VERSION;
1456 geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK |
1457 XFS_FSOP_GEOM_FLAGS_DIRV2 |
1458 XFS_FSOP_GEOM_FLAGS_EXTFLG;
1459 if (xfs_has_attr(mp))
1460 geo->flags |= XFS_FSOP_GEOM_FLAGS_ATTR;
1461 if (xfs_has_quota(mp))
1462 geo->flags |= XFS_FSOP_GEOM_FLAGS_QUOTA;
1463 if (xfs_has_align(mp))
1464 geo->flags |= XFS_FSOP_GEOM_FLAGS_IALIGN;
1465 if (xfs_has_dalign(mp))
1466 geo->flags |= XFS_FSOP_GEOM_FLAGS_DALIGN;
1467 if (xfs_has_asciici(mp))
1468 geo->flags |= XFS_FSOP_GEOM_FLAGS_DIRV2CI;
1469 if (xfs_has_lazysbcount(mp))
1470 geo->flags |= XFS_FSOP_GEOM_FLAGS_LAZYSB;
1471 if (xfs_has_attr2(mp))
1472 geo->flags |= XFS_FSOP_GEOM_FLAGS_ATTR2;
1473 if (xfs_has_projid32(mp))
1474 geo->flags |= XFS_FSOP_GEOM_FLAGS_PROJID32;
1475 if (xfs_has_crc(mp))
1476 geo->flags |= XFS_FSOP_GEOM_FLAGS_V5SB;
1477 if (xfs_has_ftype(mp))
1478 geo->flags |= XFS_FSOP_GEOM_FLAGS_FTYPE;
1479 if (xfs_has_finobt(mp))
1480 geo->flags |= XFS_FSOP_GEOM_FLAGS_FINOBT;
1481 if (xfs_has_sparseinodes(mp))
1482 geo->flags |= XFS_FSOP_GEOM_FLAGS_SPINODES;
1483 if (xfs_has_rmapbt(mp))
1484 geo->flags |= XFS_FSOP_GEOM_FLAGS_RMAPBT;
1485 if (xfs_has_reflink(mp))
1486 geo->flags |= XFS_FSOP_GEOM_FLAGS_REFLINK;
1487 if (xfs_has_bigtime(mp))
1488 geo->flags |= XFS_FSOP_GEOM_FLAGS_BIGTIME;
1489 if (xfs_has_inobtcounts(mp))
1490 geo->flags |= XFS_FSOP_GEOM_FLAGS_INOBTCNT;
1491 if (xfs_has_parent(mp))
1492 geo->flags |= XFS_FSOP_GEOM_FLAGS_PARENT;
1493 if (xfs_has_sector(mp)) {
1494 geo->flags |= XFS_FSOP_GEOM_FLAGS_SECTOR;
1495 geo->logsectsize = sbp->sb_logsectsize;
1496 } else {
1497 geo->logsectsize = BBSIZE;
1499 if (xfs_has_large_extent_counts(mp))
1500 geo->flags |= XFS_FSOP_GEOM_FLAGS_NREXT64;
1501 if (xfs_has_exchange_range(mp))
1502 geo->flags |= XFS_FSOP_GEOM_FLAGS_EXCHANGE_RANGE;
1503 if (xfs_has_metadir(mp))
1504 geo->flags |= XFS_FSOP_GEOM_FLAGS_METADIR;
1505 geo->rtsectsize = sbp->sb_blocksize;
1506 geo->dirblocksize = xfs_dir2_dirblock_bytes(sbp);
1508 if (struct_version < 4)
1509 return;
1511 if (xfs_has_logv2(mp))
1512 geo->flags |= XFS_FSOP_GEOM_FLAGS_LOGV2;
1514 geo->logsunit = sbp->sb_logsunit;
1516 if (struct_version < 5)
1517 return;
1519 geo->version = XFS_FSOP_GEOM_VERSION_V5;
1521 if (xfs_has_rtgroups(mp)) {
1522 geo->rgcount = sbp->sb_rgcount;
1523 geo->rgextents = sbp->sb_rgextents;
1527 /* Read a secondary superblock. */
1529 xfs_sb_read_secondary(
1530 struct xfs_mount *mp,
1531 struct xfs_trans *tp,
1532 xfs_agnumber_t agno,
1533 struct xfs_buf **bpp)
1535 struct xfs_buf *bp;
1536 int error;
1538 ASSERT(agno != 0 && agno != NULLAGNUMBER);
1539 error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
1540 XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
1541 XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_sb_buf_ops);
1542 if (xfs_metadata_is_sick(error))
1543 xfs_agno_mark_sick(mp, agno, XFS_SICK_AG_SB);
1544 if (error)
1545 return error;
1546 xfs_buf_set_ref(bp, XFS_SSB_REF);
1547 *bpp = bp;
1548 return 0;
1551 /* Get an uninitialised secondary superblock buffer. */
1553 xfs_sb_get_secondary(
1554 struct xfs_mount *mp,
1555 struct xfs_trans *tp,
1556 xfs_agnumber_t agno,
1557 struct xfs_buf **bpp)
1559 struct xfs_buf *bp;
1560 int error;
1562 ASSERT(agno != 0 && agno != NULLAGNUMBER);
1563 error = xfs_trans_get_buf(tp, mp->m_ddev_targp,
1564 XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
1565 XFS_FSS_TO_BB(mp, 1), 0, &bp);
1566 if (error)
1567 return error;
1568 bp->b_ops = &xfs_sb_buf_ops;
1569 xfs_buf_oneshot(bp);
1570 *bpp = bp;
1571 return 0;
1575 * sunit, swidth, sectorsize(optional with 0) should be all in bytes, so users
1576 * won't be confused by values in error messages. This function returns false
1577 * if the stripe geometry is invalid and the caller is unable to repair the
1578 * stripe configuration later in the mount process.
1580 bool
1581 xfs_validate_stripe_geometry(
1582 struct xfs_mount *mp,
1583 __s64 sunit,
1584 __s64 swidth,
1585 int sectorsize,
1586 bool may_repair,
1587 bool silent)
1589 if (swidth > INT_MAX) {
1590 if (!silent)
1591 xfs_notice(mp,
1592 "stripe width (%lld) is too large", swidth);
1593 goto check_override;
1596 if (sunit > swidth) {
1597 if (!silent)
1598 xfs_notice(mp,
1599 "stripe unit (%lld) is larger than the stripe width (%lld)", sunit, swidth);
1600 goto check_override;
1603 if (sectorsize && (int)sunit % sectorsize) {
1604 if (!silent)
1605 xfs_notice(mp,
1606 "stripe unit (%lld) must be a multiple of the sector size (%d)",
1607 sunit, sectorsize);
1608 goto check_override;
1611 if (sunit && !swidth) {
1612 if (!silent)
1613 xfs_notice(mp,
1614 "invalid stripe unit (%lld) and stripe width of 0", sunit);
1615 goto check_override;
1618 if (!sunit && swidth) {
1619 if (!silent)
1620 xfs_notice(mp,
1621 "invalid stripe width (%lld) and stripe unit of 0", swidth);
1622 goto check_override;
1625 if (sunit && (int)swidth % (int)sunit) {
1626 if (!silent)
1627 xfs_notice(mp,
1628 "stripe width (%lld) must be a multiple of the stripe unit (%lld)",
1629 swidth, sunit);
1630 goto check_override;
1632 return true;
1634 check_override:
1635 if (!may_repair)
1636 return false;
1638 * During mount, mp->m_dalign will not be set unless the sunit mount
1639 * option was set. If it was set, ignore the bad stripe alignment values
1640 * and allow the validation and overwrite later in the mount process to
1641 * attempt to overwrite the bad stripe alignment values with the values
1642 * supplied by mount options.
1644 if (!mp->m_dalign)
1645 return false;
1646 if (!silent)
1647 xfs_notice(mp,
1648 "Will try to correct with specified mount options sunit (%d) and swidth (%d)",
1649 BBTOB(mp->m_dalign), BBTOB(mp->m_swidth));
1650 return true;
1654 * Compute the maximum level number of the realtime summary file, as defined by
1655 * mkfs. The historic use of highbit32 on a 64-bit quantity prohibited correct
1656 * use of rt volumes with more than 2^32 extents.
1658 uint8_t
1659 xfs_compute_rextslog(
1660 xfs_rtbxlen_t rtextents)
1662 if (!rtextents)
1663 return 0;
1664 return xfs_highbit64(rtextents);