1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
8 #include "xfs_shared.h"
9 #include "xfs_format.h"
10 #include "xfs_log_format.h"
11 #include "xfs_trans_resv.h"
14 #include "xfs_mount.h"
15 #include "xfs_ialloc.h"
16 #include "xfs_alloc.h"
17 #include "xfs_error.h"
18 #include "xfs_trace.h"
19 #include "xfs_trans.h"
20 #include "xfs_buf_item.h"
21 #include "xfs_bmap_btree.h"
22 #include "xfs_alloc_btree.h"
24 #include "xfs_rmap_btree.h"
25 #include "xfs_refcount_btree.h"
26 #include "xfs_da_format.h"
27 #include "xfs_health.h"
30 * Physical superblock buffer manipulations. Shared with libxfs in userspace.
34 * Reference counting access wrappers to the perag structures.
35 * Because we never free per-ag structures, the only thing we
36 * have to protect against changes is the tree structure itself.
43 struct xfs_perag *pag;
47 pag = radix_tree_lookup(&mp->m_perag_tree, agno);
49 ASSERT(atomic_read(&pag->pag_ref) >= 0);
50 ref = atomic_inc_return(&pag->pag_ref);
53 trace_xfs_perag_get(mp, agno, ref, _RET_IP_);
58 * search from @first to find the next perag with the given tag set.
66 struct xfs_perag *pag;
71 found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
72 (void **)&pag, first, 1, tag);
77 ref = atomic_inc_return(&pag->pag_ref);
79 trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_);
85 struct xfs_perag *pag)
89 ASSERT(atomic_read(&pag->pag_ref) > 0);
90 ref = atomic_dec_return(&pag->pag_ref);
91 trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_);
94 /* Check all the superblock fields we care about when reading one in. */
100 if (XFS_SB_VERSION_NUM(sbp) != XFS_SB_VERSION_5)
104 * Version 5 superblock feature mask validation. Reject combinations
105 * the kernel cannot support up front before checking anything else.
107 if (xfs_sb_has_compat_feature(sbp, XFS_SB_FEAT_COMPAT_UNKNOWN)) {
109 "Superblock has unknown compatible features (0x%x) enabled.",
110 (sbp->sb_features_compat & XFS_SB_FEAT_COMPAT_UNKNOWN));
112 "Using a more recent kernel is recommended.");
115 if (xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
117 "Superblock has unknown read-only compatible features (0x%x) enabled.",
118 (sbp->sb_features_ro_compat &
119 XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
120 if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
122 "Attempted to mount read-only compatible filesystem read-write.");
124 "Filesystem can only be safely mounted read only.");
129 if (xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
131 "Superblock has unknown incompatible features (0x%x) enabled.",
132 (sbp->sb_features_incompat &
133 XFS_SB_FEAT_INCOMPAT_UNKNOWN));
135 "Filesystem cannot be safely mounted by this kernel.");
142 /* Check all the superblock fields we care about when writing one out. */
144 xfs_validate_sb_write(
145 struct xfs_mount *mp,
150 * Carry out additional sb summary counter sanity checks when we write
151 * the superblock. We skip this in the read validator because there
152 * could be newer superblocks in the log and if the values are garbage
153 * even after replay we'll recalculate them at the end of log mount.
155 * mkfs has traditionally written zeroed counters to inprogress and
156 * secondary superblocks, so allow this usage to continue because
157 * we never read counters from such superblocks.
159 if (XFS_BUF_ADDR(bp) == XFS_SB_DADDR && !sbp->sb_inprogress &&
160 (sbp->sb_fdblocks > sbp->sb_dblocks ||
161 !xfs_verify_icount(mp, sbp->sb_icount) ||
162 sbp->sb_ifree > sbp->sb_icount)) {
163 xfs_warn(mp, "SB summary counter sanity check failed");
164 return -EFSCORRUPTED;
167 if (XFS_SB_VERSION_NUM(sbp) != XFS_SB_VERSION_5)
171 * Version 5 superblock feature mask validation. Reject combinations
172 * the kernel cannot support since we checked for unsupported bits in
173 * the read verifier, which means that memory is corrupt.
175 if (xfs_sb_has_compat_feature(sbp, XFS_SB_FEAT_COMPAT_UNKNOWN)) {
177 "Corruption detected in superblock compatible features (0x%x)!",
178 (sbp->sb_features_compat & XFS_SB_FEAT_COMPAT_UNKNOWN));
179 return -EFSCORRUPTED;
182 if (xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
184 "Corruption detected in superblock read-only compatible features (0x%x)!",
185 (sbp->sb_features_ro_compat &
186 XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
187 return -EFSCORRUPTED;
189 if (xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
191 "Corruption detected in superblock incompatible features (0x%x)!",
192 (sbp->sb_features_incompat &
193 XFS_SB_FEAT_INCOMPAT_UNKNOWN));
194 return -EFSCORRUPTED;
196 if (xfs_sb_has_incompat_log_feature(sbp,
197 XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN)) {
199 "Corruption detected in superblock incompatible log features (0x%x)!",
200 (sbp->sb_features_log_incompat &
201 XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN));
202 return -EFSCORRUPTED;
206 * We can't read verify the sb LSN because the read verifier is called
207 * before the log is allocated and processed. We know the log is set up
208 * before write verifier calls, so check it here.
210 if (!xfs_log_check_lsn(mp, sbp->sb_lsn))
211 return -EFSCORRUPTED;
216 /* Check the validity of the SB. */
218 xfs_validate_sb_common(
219 struct xfs_mount *mp,
223 struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp);
224 uint32_t agcount = 0;
227 if (!xfs_verify_magic(bp, dsb->sb_magicnum)) {
228 xfs_warn(mp, "bad magic number");
232 if (!xfs_sb_good_version(sbp)) {
233 xfs_warn(mp, "bad version");
237 if (xfs_sb_version_has_pquotino(sbp)) {
238 if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
240 "Version 5 of Super block has XFS_OQUOTA bits.");
241 return -EFSCORRUPTED;
243 } else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
244 XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
246 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.");
247 return -EFSCORRUPTED;
251 * Full inode chunks must be aligned to inode chunk size when
252 * sparse inodes are enabled to support the sparse chunk
253 * allocation algorithm and prevent overlapping inode records.
255 if (xfs_sb_version_hassparseinodes(sbp)) {
258 align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize
260 if (sbp->sb_inoalignmt != align) {
262 "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.",
263 sbp->sb_inoalignmt, align);
269 sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
271 "filesystem is marked as having an external log; "
272 "specify logdev on the mount command line.");
277 sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
279 "filesystem is marked as having an internal log; "
280 "do not specify logdev on the mount command line.");
284 /* Compute agcount for this number of dblocks and agblocks */
285 if (sbp->sb_agblocks) {
286 agcount = div_u64_rem(sbp->sb_dblocks, sbp->sb_agblocks, &rem);
292 * More sanity checking. Most of these were stolen directly from
296 sbp->sb_agcount <= 0 ||
297 sbp->sb_sectsize < XFS_MIN_SECTORSIZE ||
298 sbp->sb_sectsize > XFS_MAX_SECTORSIZE ||
299 sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG ||
300 sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG ||
301 sbp->sb_sectsize != (1 << sbp->sb_sectlog) ||
302 sbp->sb_blocksize < XFS_MIN_BLOCKSIZE ||
303 sbp->sb_blocksize > XFS_MAX_BLOCKSIZE ||
304 sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG ||
305 sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
306 sbp->sb_blocksize != (1 << sbp->sb_blocklog) ||
307 sbp->sb_dirblklog + sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
308 sbp->sb_inodesize < XFS_DINODE_MIN_SIZE ||
309 sbp->sb_inodesize > XFS_DINODE_MAX_SIZE ||
310 sbp->sb_inodelog < XFS_DINODE_MIN_LOG ||
311 sbp->sb_inodelog > XFS_DINODE_MAX_LOG ||
312 sbp->sb_inodesize != (1 << sbp->sb_inodelog) ||
313 sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE ||
314 sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
315 XFS_FSB_TO_B(mp, sbp->sb_agblocks) < XFS_MIN_AG_BYTES ||
316 XFS_FSB_TO_B(mp, sbp->sb_agblocks) > XFS_MAX_AG_BYTES ||
317 sbp->sb_agblklog != xfs_highbit32(sbp->sb_agblocks - 1) + 1 ||
318 agcount == 0 || agcount != sbp->sb_agcount ||
319 (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog) ||
320 (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) ||
321 (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) ||
322 (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */) ||
323 sbp->sb_dblocks == 0 ||
324 sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) ||
325 sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp) ||
326 sbp->sb_shared_vn != 0)) {
327 xfs_notice(mp, "SB sanity check failed");
328 return -EFSCORRUPTED;
332 if (!xfs_sb_version_hasdalign(sbp) ||
333 sbp->sb_unit > sbp->sb_width ||
334 (sbp->sb_width % sbp->sb_unit) != 0) {
335 xfs_notice(mp, "SB stripe unit sanity check failed");
336 return -EFSCORRUPTED;
338 } else if (xfs_sb_version_hasdalign(sbp)) {
339 xfs_notice(mp, "SB stripe alignment sanity check failed");
340 return -EFSCORRUPTED;
341 } else if (sbp->sb_width) {
342 xfs_notice(mp, "SB stripe width sanity check failed");
343 return -EFSCORRUPTED;
347 if (xfs_sb_version_hascrc(&mp->m_sb) &&
348 sbp->sb_blocksize < XFS_MIN_CRC_BLOCKSIZE) {
349 xfs_notice(mp, "v5 SB sanity check failed");
350 return -EFSCORRUPTED;
354 * Until this is fixed only page-sized or smaller data blocks work.
356 if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
358 "File system with blocksize %d bytes. "
359 "Only pagesize (%ld) or less will currently work.",
360 sbp->sb_blocksize, PAGE_SIZE);
365 * Currently only very few inode sizes are supported.
367 switch (sbp->sb_inodesize) {
374 xfs_warn(mp, "inode size of %d bytes not supported",
379 if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) ||
380 xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) {
382 "file system too large to be mounted on this system.");
387 * Don't touch the filesystem if a user tool thinks it owns the primary
388 * superblock. mkfs doesn't clear the flag from secondary supers, so
389 * we don't check them at all.
391 if (XFS_BUF_ADDR(bp) == XFS_SB_DADDR && sbp->sb_inprogress) {
392 xfs_warn(mp, "Offline file system operation in progress!");
393 return -EFSCORRUPTED;
399 xfs_sb_quota_from_disk(struct xfs_sb *sbp)
402 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
403 * leads to in-core values having two different values for a quota
404 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
407 * Note that this change affect only the in-core values. These
408 * values are not written back to disk unless any quota information
409 * is written to the disk. Even in that case, sb_pquotino field is
410 * not written to disk unless the superblock supports pquotino.
412 if (sbp->sb_uquotino == 0)
413 sbp->sb_uquotino = NULLFSINO;
414 if (sbp->sb_gquotino == 0)
415 sbp->sb_gquotino = NULLFSINO;
416 if (sbp->sb_pquotino == 0)
417 sbp->sb_pquotino = NULLFSINO;
420 * We need to do these manipilations only if we are working
421 * with an older version of on-disk superblock.
423 if (xfs_sb_version_has_pquotino(sbp))
426 if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
427 sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
428 XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
429 if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
430 sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
431 XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
432 sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
434 if (sbp->sb_qflags & XFS_PQUOTA_ACCT &&
435 sbp->sb_gquotino != NULLFSINO) {
437 * In older version of superblock, on-disk superblock only
438 * has sb_gquotino, and in-core superblock has both sb_gquotino
439 * and sb_pquotino. But, only one of them is supported at any
440 * point of time. So, if PQUOTA is set in disk superblock,
441 * copy over sb_gquotino to sb_pquotino. The NULLFSINO test
442 * above is to make sure we don't do this twice and wipe them
445 sbp->sb_pquotino = sbp->sb_gquotino;
446 sbp->sb_gquotino = NULLFSINO;
456 to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
457 to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
458 to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
459 to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
460 to->sb_rextents = be64_to_cpu(from->sb_rextents);
461 memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
462 to->sb_logstart = be64_to_cpu(from->sb_logstart);
463 to->sb_rootino = be64_to_cpu(from->sb_rootino);
464 to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
465 to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
466 to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
467 to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
468 to->sb_agcount = be32_to_cpu(from->sb_agcount);
469 to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
470 to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
471 to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
472 to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
473 to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
474 to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
475 memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
476 to->sb_blocklog = from->sb_blocklog;
477 to->sb_sectlog = from->sb_sectlog;
478 to->sb_inodelog = from->sb_inodelog;
479 to->sb_inopblog = from->sb_inopblog;
480 to->sb_agblklog = from->sb_agblklog;
481 to->sb_rextslog = from->sb_rextslog;
482 to->sb_inprogress = from->sb_inprogress;
483 to->sb_imax_pct = from->sb_imax_pct;
484 to->sb_icount = be64_to_cpu(from->sb_icount);
485 to->sb_ifree = be64_to_cpu(from->sb_ifree);
486 to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
487 to->sb_frextents = be64_to_cpu(from->sb_frextents);
488 to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
489 to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
490 to->sb_qflags = be16_to_cpu(from->sb_qflags);
491 to->sb_flags = from->sb_flags;
492 to->sb_shared_vn = from->sb_shared_vn;
493 to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
494 to->sb_unit = be32_to_cpu(from->sb_unit);
495 to->sb_width = be32_to_cpu(from->sb_width);
496 to->sb_dirblklog = from->sb_dirblklog;
497 to->sb_logsectlog = from->sb_logsectlog;
498 to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
499 to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
500 to->sb_features2 = be32_to_cpu(from->sb_features2);
501 to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
502 to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
503 to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
504 to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
505 to->sb_features_log_incompat =
506 be32_to_cpu(from->sb_features_log_incompat);
507 /* crc is only used on disk, not in memory; just init to 0 here. */
509 to->sb_spino_align = be32_to_cpu(from->sb_spino_align);
510 to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
511 to->sb_lsn = be64_to_cpu(from->sb_lsn);
513 * sb_meta_uuid is only on disk if it differs from sb_uuid and the
514 * feature flag is set; if not set we keep it only in memory.
516 if (xfs_sb_version_hasmetauuid(to))
517 uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
519 uuid_copy(&to->sb_meta_uuid, &from->sb_uuid);
520 /* Convert on-disk flags to in-memory flags? */
522 xfs_sb_quota_from_disk(to);
530 __xfs_sb_from_disk(to, from, true);
534 xfs_sb_quota_to_disk(
538 uint16_t qflags = from->sb_qflags;
540 to->sb_uquotino = cpu_to_be64(from->sb_uquotino);
541 if (xfs_sb_version_has_pquotino(from)) {
542 to->sb_qflags = cpu_to_be16(from->sb_qflags);
543 to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
544 to->sb_pquotino = cpu_to_be64(from->sb_pquotino);
549 * The in-core version of sb_qflags do not have XFS_OQUOTA_*
550 * flags, whereas the on-disk version does. So, convert incore
551 * XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
553 qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
554 XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
556 if (from->sb_qflags &
557 (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
558 qflags |= XFS_OQUOTA_ENFD;
559 if (from->sb_qflags &
560 (XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
561 qflags |= XFS_OQUOTA_CHKD;
562 to->sb_qflags = cpu_to_be16(qflags);
565 * GQUOTINO and PQUOTINO cannot be used together in versions
566 * of superblock that do not have pquotino. from->sb_flags
567 * tells us which quota is active and should be copied to
568 * disk. If neither are active, we should NULL the inode.
570 * In all cases, the separate pquotino must remain 0 because it
571 * it beyond the "end" of the valid non-pquotino superblock.
573 if (from->sb_qflags & XFS_GQUOTA_ACCT)
574 to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
575 else if (from->sb_qflags & XFS_PQUOTA_ACCT)
576 to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
579 * We can't rely on just the fields being logged to tell us
580 * that it is safe to write NULLFSINO - we should only do that
581 * if quotas are not actually enabled. Hence only write
582 * NULLFSINO if both in-core quota inodes are NULL.
584 if (from->sb_gquotino == NULLFSINO &&
585 from->sb_pquotino == NULLFSINO)
586 to->sb_gquotino = cpu_to_be64(NULLFSINO);
597 xfs_sb_quota_to_disk(to, from);
599 to->sb_magicnum = cpu_to_be32(from->sb_magicnum);
600 to->sb_blocksize = cpu_to_be32(from->sb_blocksize);
601 to->sb_dblocks = cpu_to_be64(from->sb_dblocks);
602 to->sb_rblocks = cpu_to_be64(from->sb_rblocks);
603 to->sb_rextents = cpu_to_be64(from->sb_rextents);
604 memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
605 to->sb_logstart = cpu_to_be64(from->sb_logstart);
606 to->sb_rootino = cpu_to_be64(from->sb_rootino);
607 to->sb_rbmino = cpu_to_be64(from->sb_rbmino);
608 to->sb_rsumino = cpu_to_be64(from->sb_rsumino);
609 to->sb_rextsize = cpu_to_be32(from->sb_rextsize);
610 to->sb_agblocks = cpu_to_be32(from->sb_agblocks);
611 to->sb_agcount = cpu_to_be32(from->sb_agcount);
612 to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks);
613 to->sb_logblocks = cpu_to_be32(from->sb_logblocks);
614 to->sb_versionnum = cpu_to_be16(from->sb_versionnum);
615 to->sb_sectsize = cpu_to_be16(from->sb_sectsize);
616 to->sb_inodesize = cpu_to_be16(from->sb_inodesize);
617 to->sb_inopblock = cpu_to_be16(from->sb_inopblock);
618 memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
619 to->sb_blocklog = from->sb_blocklog;
620 to->sb_sectlog = from->sb_sectlog;
621 to->sb_inodelog = from->sb_inodelog;
622 to->sb_inopblog = from->sb_inopblog;
623 to->sb_agblklog = from->sb_agblklog;
624 to->sb_rextslog = from->sb_rextslog;
625 to->sb_inprogress = from->sb_inprogress;
626 to->sb_imax_pct = from->sb_imax_pct;
627 to->sb_icount = cpu_to_be64(from->sb_icount);
628 to->sb_ifree = cpu_to_be64(from->sb_ifree);
629 to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks);
630 to->sb_frextents = cpu_to_be64(from->sb_frextents);
632 to->sb_flags = from->sb_flags;
633 to->sb_shared_vn = from->sb_shared_vn;
634 to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt);
635 to->sb_unit = cpu_to_be32(from->sb_unit);
636 to->sb_width = cpu_to_be32(from->sb_width);
637 to->sb_dirblklog = from->sb_dirblklog;
638 to->sb_logsectlog = from->sb_logsectlog;
639 to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize);
640 to->sb_logsunit = cpu_to_be32(from->sb_logsunit);
643 * We need to ensure that bad_features2 always matches features2.
644 * Hence we enforce that here rather than having to remember to do it
645 * everywhere else that updates features2.
647 from->sb_bad_features2 = from->sb_features2;
648 to->sb_features2 = cpu_to_be32(from->sb_features2);
649 to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2);
651 if (xfs_sb_version_hascrc(from)) {
652 to->sb_features_compat = cpu_to_be32(from->sb_features_compat);
653 to->sb_features_ro_compat =
654 cpu_to_be32(from->sb_features_ro_compat);
655 to->sb_features_incompat =
656 cpu_to_be32(from->sb_features_incompat);
657 to->sb_features_log_incompat =
658 cpu_to_be32(from->sb_features_log_incompat);
659 to->sb_spino_align = cpu_to_be32(from->sb_spino_align);
660 to->sb_lsn = cpu_to_be64(from->sb_lsn);
661 if (xfs_sb_version_hasmetauuid(from))
662 uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
667 * If the superblock has the CRC feature bit set or the CRC field is non-null,
668 * check that the CRC is valid. We check the CRC field is non-null because a
669 * single bit error could clear the feature bit and unused parts of the
670 * superblock are supposed to be zero. Hence a non-null crc field indicates that
671 * we've potentially lost a feature bit and we should check it anyway.
673 * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
674 * last field in V4 secondary superblocks. So for secondary superblocks,
675 * we are more forgiving, and ignore CRC failures if the primary doesn't
676 * indicate that the fs version is V5.
683 struct xfs_mount *mp = bp->b_mount;
684 struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp);
688 * open code the version check to avoid needing to convert the entire
689 * superblock from disk order just to check the version number
691 if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
692 (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
696 if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
697 /* Only fail bad secondaries on a known V5 filesystem */
698 if (bp->b_bn == XFS_SB_DADDR ||
699 xfs_sb_version_hascrc(&mp->m_sb)) {
707 * Check all the superblock fields. Don't byteswap the xquota flags
708 * because _verify_common checks the on-disk values.
710 __xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false);
711 error = xfs_validate_sb_common(mp, bp, &sb);
714 error = xfs_validate_sb_read(mp, &sb);
717 if (error == -EFSCORRUPTED || error == -EFSBADCRC)
718 xfs_verifier_error(bp, error, __this_address);
720 xfs_buf_ioerror(bp, error);
724 * We may be probed for a filesystem match, so we may not want to emit
725 * messages when the superblock buffer is not actually an XFS superblock.
726 * If we find an XFS superblock, then run a normal, noisy mount because we are
727 * really going to mount it and want to know about errors.
730 xfs_sb_quiet_read_verify(
733 struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp);
735 if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
736 /* XFS filesystem, verify noisily! */
737 xfs_sb_read_verify(bp);
741 xfs_buf_ioerror(bp, -EWRONGFS);
749 struct xfs_mount *mp = bp->b_mount;
750 struct xfs_buf_log_item *bip = bp->b_log_item;
754 * Check all the superblock fields. Don't byteswap the xquota flags
755 * because _verify_common checks the on-disk values.
757 __xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false);
758 error = xfs_validate_sb_common(mp, bp, &sb);
761 error = xfs_validate_sb_write(mp, bp, &sb);
765 if (!xfs_sb_version_hascrc(&mp->m_sb))
769 XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
771 xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
775 xfs_verifier_error(bp, error, __this_address);
778 const struct xfs_buf_ops xfs_sb_buf_ops = {
780 .magic = { cpu_to_be32(XFS_SB_MAGIC), cpu_to_be32(XFS_SB_MAGIC) },
781 .verify_read = xfs_sb_read_verify,
782 .verify_write = xfs_sb_write_verify,
785 const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
786 .name = "xfs_sb_quiet",
787 .magic = { cpu_to_be32(XFS_SB_MAGIC), cpu_to_be32(XFS_SB_MAGIC) },
788 .verify_read = xfs_sb_quiet_read_verify,
789 .verify_write = xfs_sb_write_verify,
795 * Mount initialization code establishing various mount
796 * fields from the superblock associated with the given
799 * Inode geometry are calculated in xfs_ialloc_setup_geometry.
803 struct xfs_mount *mp,
806 mp->m_agfrotor = mp->m_agirotor = 0;
807 mp->m_maxagi = mp->m_sb.sb_agcount;
808 mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
809 mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
810 mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
811 mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
812 mp->m_blockmask = sbp->sb_blocksize - 1;
813 mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
814 mp->m_blockwmask = mp->m_blockwsize - 1;
816 mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1);
817 mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0);
818 mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
819 mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
821 mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1);
822 mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0);
823 mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
824 mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
826 mp->m_rmap_mxr[0] = xfs_rmapbt_maxrecs(sbp->sb_blocksize, 1);
827 mp->m_rmap_mxr[1] = xfs_rmapbt_maxrecs(sbp->sb_blocksize, 0);
828 mp->m_rmap_mnr[0] = mp->m_rmap_mxr[0] / 2;
829 mp->m_rmap_mnr[1] = mp->m_rmap_mxr[1] / 2;
831 mp->m_refc_mxr[0] = xfs_refcountbt_maxrecs(sbp->sb_blocksize, true);
832 mp->m_refc_mxr[1] = xfs_refcountbt_maxrecs(sbp->sb_blocksize, false);
833 mp->m_refc_mnr[0] = mp->m_refc_mxr[0] / 2;
834 mp->m_refc_mnr[1] = mp->m_refc_mxr[1] / 2;
836 mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
837 mp->m_alloc_set_aside = xfs_alloc_set_aside(mp);
838 mp->m_ag_max_usable = xfs_alloc_ag_max_usable(mp);
842 * xfs_initialize_perag_data
844 * Read in each per-ag structure so we can count up the number of
845 * allocated inodes, free inodes and used filesystem blocks as this
846 * information is no longer persistent in the superblock. Once we have
847 * this information, write it into the in-core superblock structure.
850 xfs_initialize_perag_data(
851 struct xfs_mount *mp,
852 xfs_agnumber_t agcount)
854 xfs_agnumber_t index;
856 xfs_sb_t *sbp = &mp->m_sb;
860 uint64_t bfreelst = 0;
865 for (index = 0; index < agcount; index++) {
867 * read the agf, then the agi. This gets us
868 * all the information we need and populates the
869 * per-ag structures for us.
871 error = xfs_alloc_pagf_init(mp, NULL, index, 0);
875 error = xfs_ialloc_pagi_init(mp, NULL, index);
878 pag = xfs_perag_get(mp, index);
879 ifree += pag->pagi_freecount;
880 ialloc += pag->pagi_count;
881 bfree += pag->pagf_freeblks;
882 bfreelst += pag->pagf_flcount;
883 btree += pag->pagf_btreeblks;
886 fdblocks = bfree + bfreelst + btree;
889 * If the new summary counts are obviously incorrect, fail the
890 * mount operation because that implies the AGFs are also corrupt.
891 * Clear FS_COUNTERS so that we don't unmount with a dirty log, which
892 * will prevent xfs_repair from fixing anything.
894 if (fdblocks > sbp->sb_dblocks || ifree > ialloc) {
895 xfs_alert(mp, "AGF corruption. Please run xfs_repair.");
896 error = -EFSCORRUPTED;
900 /* Overwrite incore superblock counters with just-read data */
901 spin_lock(&mp->m_sb_lock);
902 sbp->sb_ifree = ifree;
903 sbp->sb_icount = ialloc;
904 sbp->sb_fdblocks = fdblocks;
905 spin_unlock(&mp->m_sb_lock);
907 xfs_reinit_percpu_counters(mp);
909 xfs_fs_mark_healthy(mp, XFS_SICK_FS_COUNTERS);
914 * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
915 * into the superblock buffer to be logged. It does not provide the higher
916 * level of locking that is needed to protect the in-core superblock from
921 struct xfs_trans *tp)
923 struct xfs_mount *mp = tp->t_mountp;
924 struct xfs_buf *bp = xfs_trans_getsb(tp, mp);
926 mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount);
927 mp->m_sb.sb_ifree = percpu_counter_sum(&mp->m_ifree);
928 mp->m_sb.sb_fdblocks = percpu_counter_sum(&mp->m_fdblocks);
930 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb);
931 xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
932 xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb) - 1);
938 * Sync the superblock to disk.
940 * Note that the caller is responsible for checking the frozen state of the
941 * filesystem. This procedure uses the non-blocking transaction allocator and
942 * thus will allow modifications to a frozen fs. This is required because this
943 * code can be called during the process of freezing where use of the high-level
944 * allocator would deadlock.
948 struct xfs_mount *mp,
951 struct xfs_trans *tp;
954 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0,
955 XFS_TRANS_NO_WRITECOUNT, &tp);
961 xfs_trans_set_sync(tp);
962 return xfs_trans_commit(tp);
966 * Update all the secondary superblocks to match the new state of the primary.
967 * Because we are completely overwriting all the existing fields in the
968 * secondary superblock buffers, there is no need to read them in from disk.
969 * Just get a new buffer, stamp it and write it.
971 * The sb buffers need to be cached here so that we serialise against other
972 * operations that access the secondary superblocks, but we don't want to keep
973 * them in memory once it is written so we mark it as a one-shot buffer.
976 xfs_update_secondary_sbs(
977 struct xfs_mount *mp)
982 LIST_HEAD (buffer_list);
984 /* update secondary superblocks. */
985 for (agno = 1; agno < mp->m_sb.sb_agcount; agno++) {
988 error = xfs_buf_get(mp->m_ddev_targp,
989 XFS_AG_DADDR(mp, agno, XFS_SB_DADDR),
990 XFS_FSS_TO_BB(mp, 1), &bp);
992 * If we get an error reading or writing alternate superblocks,
993 * continue. xfs_repair chooses the "best" superblock based
994 * on most matches; if we break early, we'll leave more
995 * superblocks un-updated than updated, and xfs_repair may
996 * pick them over the properly-updated primary.
1000 "error allocating secondary superblock for ag %d",
1003 saved_error = error;
1007 bp->b_ops = &xfs_sb_buf_ops;
1008 xfs_buf_oneshot(bp);
1009 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
1010 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb);
1011 xfs_buf_delwri_queue(bp, &buffer_list);
1014 /* don't hold too many buffers at once */
1018 error = xfs_buf_delwri_submit(&buffer_list);
1021 "write error %d updating a secondary superblock near ag %d",
1024 saved_error = error;
1028 error = xfs_buf_delwri_submit(&buffer_list);
1031 "write error %d updating a secondary superblock near ag %d",
1035 return saved_error ? saved_error : error;
1039 * Same behavior as xfs_sync_sb, except that it is always synchronous and it
1040 * also writes the superblock buffer to disk sector 0 immediately.
1044 struct xfs_mount *mp)
1046 struct xfs_trans *tp;
1050 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0, 0, &tp);
1054 bp = xfs_trans_getsb(tp, mp);
1056 xfs_trans_bhold(tp, bp);
1057 xfs_trans_set_sync(tp);
1058 error = xfs_trans_commit(tp);
1062 * write out the sb buffer to get the changes to disk
1064 error = xfs_bwrite(bp);
1073 struct xfs_fsop_geom *geo,
1076 memset(geo, 0, sizeof(struct xfs_fsop_geom));
1078 geo->blocksize = sbp->sb_blocksize;
1079 geo->rtextsize = sbp->sb_rextsize;
1080 geo->agblocks = sbp->sb_agblocks;
1081 geo->agcount = sbp->sb_agcount;
1082 geo->logblocks = sbp->sb_logblocks;
1083 geo->sectsize = sbp->sb_sectsize;
1084 geo->inodesize = sbp->sb_inodesize;
1085 geo->imaxpct = sbp->sb_imax_pct;
1086 geo->datablocks = sbp->sb_dblocks;
1087 geo->rtblocks = sbp->sb_rblocks;
1088 geo->rtextents = sbp->sb_rextents;
1089 geo->logstart = sbp->sb_logstart;
1090 BUILD_BUG_ON(sizeof(geo->uuid) != sizeof(sbp->sb_uuid));
1091 memcpy(geo->uuid, &sbp->sb_uuid, sizeof(sbp->sb_uuid));
1093 if (struct_version < 2)
1096 geo->sunit = sbp->sb_unit;
1097 geo->swidth = sbp->sb_width;
1099 if (struct_version < 3)
1102 geo->version = XFS_FSOP_GEOM_VERSION;
1103 geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK |
1104 XFS_FSOP_GEOM_FLAGS_DIRV2 |
1105 XFS_FSOP_GEOM_FLAGS_EXTFLG;
1106 if (xfs_sb_version_hasattr(sbp))
1107 geo->flags |= XFS_FSOP_GEOM_FLAGS_ATTR;
1108 if (xfs_sb_version_hasquota(sbp))
1109 geo->flags |= XFS_FSOP_GEOM_FLAGS_QUOTA;
1110 if (xfs_sb_version_hasalign(sbp))
1111 geo->flags |= XFS_FSOP_GEOM_FLAGS_IALIGN;
1112 if (xfs_sb_version_hasdalign(sbp))
1113 geo->flags |= XFS_FSOP_GEOM_FLAGS_DALIGN;
1114 if (xfs_sb_version_hassector(sbp))
1115 geo->flags |= XFS_FSOP_GEOM_FLAGS_SECTOR;
1116 if (xfs_sb_version_hasasciici(sbp))
1117 geo->flags |= XFS_FSOP_GEOM_FLAGS_DIRV2CI;
1118 if (xfs_sb_version_haslazysbcount(sbp))
1119 geo->flags |= XFS_FSOP_GEOM_FLAGS_LAZYSB;
1120 if (xfs_sb_version_hasattr2(sbp))
1121 geo->flags |= XFS_FSOP_GEOM_FLAGS_ATTR2;
1122 if (xfs_sb_version_hasprojid32bit(sbp))
1123 geo->flags |= XFS_FSOP_GEOM_FLAGS_PROJID32;
1124 if (xfs_sb_version_hascrc(sbp))
1125 geo->flags |= XFS_FSOP_GEOM_FLAGS_V5SB;
1126 if (xfs_sb_version_hasftype(sbp))
1127 geo->flags |= XFS_FSOP_GEOM_FLAGS_FTYPE;
1128 if (xfs_sb_version_hasfinobt(sbp))
1129 geo->flags |= XFS_FSOP_GEOM_FLAGS_FINOBT;
1130 if (xfs_sb_version_hassparseinodes(sbp))
1131 geo->flags |= XFS_FSOP_GEOM_FLAGS_SPINODES;
1132 if (xfs_sb_version_hasrmapbt(sbp))
1133 geo->flags |= XFS_FSOP_GEOM_FLAGS_RMAPBT;
1134 if (xfs_sb_version_hasreflink(sbp))
1135 geo->flags |= XFS_FSOP_GEOM_FLAGS_REFLINK;
1136 if (xfs_sb_version_hassector(sbp))
1137 geo->logsectsize = sbp->sb_logsectsize;
1139 geo->logsectsize = BBSIZE;
1140 geo->rtsectsize = sbp->sb_blocksize;
1141 geo->dirblocksize = xfs_dir2_dirblock_bytes(sbp);
1143 if (struct_version < 4)
1146 if (xfs_sb_version_haslogv2(sbp))
1147 geo->flags |= XFS_FSOP_GEOM_FLAGS_LOGV2;
1149 geo->logsunit = sbp->sb_logsunit;
1151 if (struct_version < 5)
1154 geo->version = XFS_FSOP_GEOM_VERSION_V5;
1157 /* Read a secondary superblock. */
1159 xfs_sb_read_secondary(
1160 struct xfs_mount *mp,
1161 struct xfs_trans *tp,
1162 xfs_agnumber_t agno,
1163 struct xfs_buf **bpp)
1168 ASSERT(agno != 0 && agno != NULLAGNUMBER);
1169 error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
1170 XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
1171 XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_sb_buf_ops);
1174 xfs_buf_set_ref(bp, XFS_SSB_REF);
1179 /* Get an uninitialised secondary superblock buffer. */
1181 xfs_sb_get_secondary(
1182 struct xfs_mount *mp,
1183 struct xfs_trans *tp,
1184 xfs_agnumber_t agno,
1185 struct xfs_buf **bpp)
1190 ASSERT(agno != 0 && agno != NULLAGNUMBER);
1191 error = xfs_trans_get_buf(tp, mp->m_ddev_targp,
1192 XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
1193 XFS_FSS_TO_BB(mp, 1), 0, &bp);
1196 bp->b_ops = &xfs_sb_buf_ops;
1197 xfs_buf_oneshot(bp);