2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18 #include "libxfs_priv.h"
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
26 #include "xfs_mount.h"
27 #include "xfs_defer.h"
28 #include "xfs_inode.h"
29 #include "xfs_ialloc.h"
30 #include "xfs_alloc.h"
31 #include "xfs_trace.h"
32 #include "xfs_cksum.h"
33 #include "xfs_trans.h"
34 #include "xfs_bmap_btree.h"
35 #include "xfs_alloc_btree.h"
36 #include "xfs_ialloc_btree.h"
37 #include "xfs_rmap_btree.h"
39 #include "xfs_refcount_btree.h"
42 * Physical superblock buffer manipulations. Shared with libxfs in userspace.
46 * Reference counting access wrappers to the perag structures.
47 * Because we never free per-ag structures, the only thing we
48 * have to protect against changes is the tree structure itself.
55 struct xfs_perag
*pag
;
59 pag
= radix_tree_lookup(&mp
->m_perag_tree
, agno
);
61 ASSERT(atomic_read(&pag
->pag_ref
) >= 0);
62 ref
= atomic_inc_return(&pag
->pag_ref
);
65 trace_xfs_perag_get(mp
, agno
, ref
, _RET_IP_
);
70 * search from @first to find the next perag with the given tag set.
78 struct xfs_perag
*pag
;
83 found
= radix_tree_gang_lookup_tag(&mp
->m_perag_tree
,
84 (void **)&pag
, first
, 1, tag
);
89 ref
= atomic_inc_return(&pag
->pag_ref
);
91 trace_xfs_perag_get_tag(mp
, pag
->pag_agno
, ref
, _RET_IP_
);
97 struct xfs_perag
*pag
)
101 ASSERT(atomic_read(&pag
->pag_ref
) > 0);
102 ref
= atomic_dec_return(&pag
->pag_ref
);
103 trace_xfs_perag_put(pag
->pag_mount
, pag
->pag_agno
, ref
, _RET_IP_
);
107 * Check the validity of the SB found.
110 xfs_mount_validate_sb(
113 bool check_inprogress
,
116 if (sbp
->sb_magicnum
!= XFS_SB_MAGIC
) {
117 xfs_warn(mp
, "bad magic number");
122 if (!xfs_sb_good_version(sbp
)) {
123 xfs_warn(mp
, "bad version");
128 * Version 5 superblock feature mask validation. Reject combinations the
129 * kernel cannot support up front before checking anything else. For
130 * write validation, we don't need to check feature masks.
132 if (check_version
&& XFS_SB_VERSION_NUM(sbp
) == XFS_SB_VERSION_5
) {
133 if (xfs_sb_has_compat_feature(sbp
,
134 XFS_SB_FEAT_COMPAT_UNKNOWN
)) {
136 "Superblock has unknown compatible features (0x%x) enabled.",
137 (sbp
->sb_features_compat
&
138 XFS_SB_FEAT_COMPAT_UNKNOWN
));
140 "Using a more recent kernel is recommended.");
143 if (xfs_sb_has_ro_compat_feature(sbp
,
144 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
)) {
146 "Superblock has unknown read-only compatible features (0x%x) enabled.",
147 (sbp
->sb_features_ro_compat
&
148 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
));
149 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
151 "Attempted to mount read-only compatible filesystem read-write.");
153 "Filesystem can only be safely mounted read only.");
158 if (xfs_sb_has_incompat_feature(sbp
,
159 XFS_SB_FEAT_INCOMPAT_UNKNOWN
)) {
161 "Superblock has unknown incompatible features (0x%x) enabled.",
162 (sbp
->sb_features_incompat
&
163 XFS_SB_FEAT_INCOMPAT_UNKNOWN
));
165 "Filesystem can not be safely mounted by this kernel.");
168 } else if (xfs_sb_version_hascrc(sbp
)) {
170 * We can't read verify the sb LSN because the read verifier is
171 * called before the log is allocated and processed. We know the
172 * log is set up before write verifier (!check_version) calls,
173 * so just check it here.
175 if (!xfs_log_check_lsn(mp
, sbp
->sb_lsn
))
176 return -EFSCORRUPTED
;
179 if (xfs_sb_version_has_pquotino(sbp
)) {
180 if (sbp
->sb_qflags
& (XFS_OQUOTA_ENFD
| XFS_OQUOTA_CHKD
)) {
182 "Version 5 of Super block has XFS_OQUOTA bits.");
183 return -EFSCORRUPTED
;
185 } else if (sbp
->sb_qflags
& (XFS_PQUOTA_ENFD
| XFS_GQUOTA_ENFD
|
186 XFS_PQUOTA_CHKD
| XFS_GQUOTA_CHKD
)) {
188 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.");
189 return -EFSCORRUPTED
;
193 * Full inode chunks must be aligned to inode chunk size when
194 * sparse inodes are enabled to support the sparse chunk
195 * allocation algorithm and prevent overlapping inode records.
197 if (xfs_sb_version_hassparseinodes(sbp
)) {
200 align
= XFS_INODES_PER_CHUNK
* sbp
->sb_inodesize
202 if (sbp
->sb_inoalignmt
!= align
) {
204 "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.",
205 sbp
->sb_inoalignmt
, align
);
211 sbp
->sb_logstart
== 0 && mp
->m_logdev_targp
== mp
->m_ddev_targp
)) {
213 "filesystem is marked as having an external log; "
214 "specify logdev on the mount command line.");
219 sbp
->sb_logstart
!= 0 && mp
->m_logdev_targp
!= mp
->m_ddev_targp
)) {
221 "filesystem is marked as having an internal log; "
222 "do not specify logdev on the mount command line.");
227 * More sanity checking. Most of these were stolen directly from
231 sbp
->sb_agcount
<= 0 ||
232 sbp
->sb_sectsize
< XFS_MIN_SECTORSIZE
||
233 sbp
->sb_sectsize
> XFS_MAX_SECTORSIZE
||
234 sbp
->sb_sectlog
< XFS_MIN_SECTORSIZE_LOG
||
235 sbp
->sb_sectlog
> XFS_MAX_SECTORSIZE_LOG
||
236 sbp
->sb_sectsize
!= (1 << sbp
->sb_sectlog
) ||
237 sbp
->sb_blocksize
< XFS_MIN_BLOCKSIZE
||
238 sbp
->sb_blocksize
> XFS_MAX_BLOCKSIZE
||
239 sbp
->sb_blocklog
< XFS_MIN_BLOCKSIZE_LOG
||
240 sbp
->sb_blocklog
> XFS_MAX_BLOCKSIZE_LOG
||
241 sbp
->sb_blocksize
!= (1 << sbp
->sb_blocklog
) ||
242 sbp
->sb_dirblklog
+ sbp
->sb_blocklog
> XFS_MAX_BLOCKSIZE_LOG
||
243 sbp
->sb_inodesize
< XFS_DINODE_MIN_SIZE
||
244 sbp
->sb_inodesize
> XFS_DINODE_MAX_SIZE
||
245 sbp
->sb_inodelog
< XFS_DINODE_MIN_LOG
||
246 sbp
->sb_inodelog
> XFS_DINODE_MAX_LOG
||
247 sbp
->sb_inodesize
!= (1 << sbp
->sb_inodelog
) ||
248 sbp
->sb_logsunit
> XLOG_MAX_RECORD_BSIZE
||
249 sbp
->sb_inopblock
!= howmany(sbp
->sb_blocksize
,sbp
->sb_inodesize
) ||
250 (sbp
->sb_blocklog
- sbp
->sb_inodelog
!= sbp
->sb_inopblog
) ||
251 (sbp
->sb_rextsize
* sbp
->sb_blocksize
> XFS_MAX_RTEXTSIZE
) ||
252 (sbp
->sb_rextsize
* sbp
->sb_blocksize
< XFS_MIN_RTEXTSIZE
) ||
253 (sbp
->sb_imax_pct
> 100 /* zero sb_imax_pct is valid */) ||
254 sbp
->sb_dblocks
== 0 ||
255 sbp
->sb_dblocks
> XFS_MAX_DBLOCKS(sbp
) ||
256 sbp
->sb_dblocks
< XFS_MIN_DBLOCKS(sbp
) ||
257 sbp
->sb_shared_vn
!= 0)) {
258 xfs_notice(mp
, "SB sanity check failed");
259 return -EFSCORRUPTED
;
262 if (xfs_sb_version_hascrc(&mp
->m_sb
) &&
263 sbp
->sb_blocksize
< XFS_MIN_CRC_BLOCKSIZE
) {
264 xfs_notice(mp
, "v5 SB sanity check failed");
265 return -EFSCORRUPTED
;
269 * Currently only very few inode sizes are supported.
271 switch (sbp
->sb_inodesize
) {
278 xfs_warn(mp
, "inode size of %d bytes not supported",
283 if (xfs_sb_validate_fsb_count(sbp
, sbp
->sb_dblocks
) ||
284 xfs_sb_validate_fsb_count(sbp
, sbp
->sb_rblocks
)) {
286 "file system too large to be mounted on this system.");
294 xfs_sb_quota_from_disk(struct xfs_sb
*sbp
)
297 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
298 * leads to in-core values having two different values for a quota
299 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
302 * Note that this change affect only the in-core values. These
303 * values are not written back to disk unless any quota information
304 * is written to the disk. Even in that case, sb_pquotino field is
305 * not written to disk unless the superblock supports pquotino.
307 if (sbp
->sb_uquotino
== 0)
308 sbp
->sb_uquotino
= NULLFSINO
;
309 if (sbp
->sb_gquotino
== 0)
310 sbp
->sb_gquotino
= NULLFSINO
;
311 if (sbp
->sb_pquotino
== 0)
312 sbp
->sb_pquotino
= NULLFSINO
;
315 * We need to do these manipilations only if we are working
316 * with an older version of on-disk superblock.
318 if (xfs_sb_version_has_pquotino(sbp
))
321 if (sbp
->sb_qflags
& XFS_OQUOTA_ENFD
)
322 sbp
->sb_qflags
|= (sbp
->sb_qflags
& XFS_PQUOTA_ACCT
) ?
323 XFS_PQUOTA_ENFD
: XFS_GQUOTA_ENFD
;
324 if (sbp
->sb_qflags
& XFS_OQUOTA_CHKD
)
325 sbp
->sb_qflags
|= (sbp
->sb_qflags
& XFS_PQUOTA_ACCT
) ?
326 XFS_PQUOTA_CHKD
: XFS_GQUOTA_CHKD
;
327 sbp
->sb_qflags
&= ~(XFS_OQUOTA_ENFD
| XFS_OQUOTA_CHKD
);
329 if (sbp
->sb_qflags
& XFS_PQUOTA_ACCT
&&
330 sbp
->sb_gquotino
!= NULLFSINO
) {
332 * In older version of superblock, on-disk superblock only
333 * has sb_gquotino, and in-core superblock has both sb_gquotino
334 * and sb_pquotino. But, only one of them is supported at any
335 * point of time. So, if PQUOTA is set in disk superblock,
336 * copy over sb_gquotino to sb_pquotino. The NULLFSINO test
337 * above is to make sure we don't do this twice and wipe them
340 sbp
->sb_pquotino
= sbp
->sb_gquotino
;
341 sbp
->sb_gquotino
= NULLFSINO
;
351 to
->sb_magicnum
= be32_to_cpu(from
->sb_magicnum
);
352 to
->sb_blocksize
= be32_to_cpu(from
->sb_blocksize
);
353 to
->sb_dblocks
= be64_to_cpu(from
->sb_dblocks
);
354 to
->sb_rblocks
= be64_to_cpu(from
->sb_rblocks
);
355 to
->sb_rextents
= be64_to_cpu(from
->sb_rextents
);
356 memcpy(&to
->sb_uuid
, &from
->sb_uuid
, sizeof(to
->sb_uuid
));
357 to
->sb_logstart
= be64_to_cpu(from
->sb_logstart
);
358 to
->sb_rootino
= be64_to_cpu(from
->sb_rootino
);
359 to
->sb_rbmino
= be64_to_cpu(from
->sb_rbmino
);
360 to
->sb_rsumino
= be64_to_cpu(from
->sb_rsumino
);
361 to
->sb_rextsize
= be32_to_cpu(from
->sb_rextsize
);
362 to
->sb_agblocks
= be32_to_cpu(from
->sb_agblocks
);
363 to
->sb_agcount
= be32_to_cpu(from
->sb_agcount
);
364 to
->sb_rbmblocks
= be32_to_cpu(from
->sb_rbmblocks
);
365 to
->sb_logblocks
= be32_to_cpu(from
->sb_logblocks
);
366 to
->sb_versionnum
= be16_to_cpu(from
->sb_versionnum
);
367 to
->sb_sectsize
= be16_to_cpu(from
->sb_sectsize
);
368 to
->sb_inodesize
= be16_to_cpu(from
->sb_inodesize
);
369 to
->sb_inopblock
= be16_to_cpu(from
->sb_inopblock
);
370 memcpy(&to
->sb_fname
, &from
->sb_fname
, sizeof(to
->sb_fname
));
371 to
->sb_blocklog
= from
->sb_blocklog
;
372 to
->sb_sectlog
= from
->sb_sectlog
;
373 to
->sb_inodelog
= from
->sb_inodelog
;
374 to
->sb_inopblog
= from
->sb_inopblog
;
375 to
->sb_agblklog
= from
->sb_agblklog
;
376 to
->sb_rextslog
= from
->sb_rextslog
;
377 to
->sb_inprogress
= from
->sb_inprogress
;
378 to
->sb_imax_pct
= from
->sb_imax_pct
;
379 to
->sb_icount
= be64_to_cpu(from
->sb_icount
);
380 to
->sb_ifree
= be64_to_cpu(from
->sb_ifree
);
381 to
->sb_fdblocks
= be64_to_cpu(from
->sb_fdblocks
);
382 to
->sb_frextents
= be64_to_cpu(from
->sb_frextents
);
383 to
->sb_uquotino
= be64_to_cpu(from
->sb_uquotino
);
384 to
->sb_gquotino
= be64_to_cpu(from
->sb_gquotino
);
385 to
->sb_qflags
= be16_to_cpu(from
->sb_qflags
);
386 to
->sb_flags
= from
->sb_flags
;
387 to
->sb_shared_vn
= from
->sb_shared_vn
;
388 to
->sb_inoalignmt
= be32_to_cpu(from
->sb_inoalignmt
);
389 to
->sb_unit
= be32_to_cpu(from
->sb_unit
);
390 to
->sb_width
= be32_to_cpu(from
->sb_width
);
391 to
->sb_dirblklog
= from
->sb_dirblklog
;
392 to
->sb_logsectlog
= from
->sb_logsectlog
;
393 to
->sb_logsectsize
= be16_to_cpu(from
->sb_logsectsize
);
394 to
->sb_logsunit
= be32_to_cpu(from
->sb_logsunit
);
395 to
->sb_features2
= be32_to_cpu(from
->sb_features2
);
396 to
->sb_bad_features2
= be32_to_cpu(from
->sb_bad_features2
);
397 to
->sb_features_compat
= be32_to_cpu(from
->sb_features_compat
);
398 to
->sb_features_ro_compat
= be32_to_cpu(from
->sb_features_ro_compat
);
399 to
->sb_features_incompat
= be32_to_cpu(from
->sb_features_incompat
);
400 to
->sb_features_log_incompat
=
401 be32_to_cpu(from
->sb_features_log_incompat
);
402 /* crc is only used on disk, not in memory; just init to 0 here. */
404 to
->sb_spino_align
= be32_to_cpu(from
->sb_spino_align
);
405 to
->sb_pquotino
= be64_to_cpu(from
->sb_pquotino
);
406 to
->sb_lsn
= be64_to_cpu(from
->sb_lsn
);
408 * sb_meta_uuid is only on disk if it differs from sb_uuid and the
409 * feature flag is set; if not set we keep it only in memory.
411 if (xfs_sb_version_hasmetauuid(to
))
412 uuid_copy(&to
->sb_meta_uuid
, &from
->sb_meta_uuid
);
414 uuid_copy(&to
->sb_meta_uuid
, &from
->sb_uuid
);
415 /* Convert on-disk flags to in-memory flags? */
417 xfs_sb_quota_from_disk(to
);
425 __xfs_sb_from_disk(to
, from
, true);
429 xfs_sb_quota_to_disk(
433 uint16_t qflags
= from
->sb_qflags
;
435 to
->sb_uquotino
= cpu_to_be64(from
->sb_uquotino
);
436 if (xfs_sb_version_has_pquotino(from
)) {
437 to
->sb_qflags
= cpu_to_be16(from
->sb_qflags
);
438 to
->sb_gquotino
= cpu_to_be64(from
->sb_gquotino
);
439 to
->sb_pquotino
= cpu_to_be64(from
->sb_pquotino
);
444 * The in-core version of sb_qflags do not have XFS_OQUOTA_*
445 * flags, whereas the on-disk version does. So, convert incore
446 * XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
448 qflags
&= ~(XFS_PQUOTA_ENFD
| XFS_PQUOTA_CHKD
|
449 XFS_GQUOTA_ENFD
| XFS_GQUOTA_CHKD
);
451 if (from
->sb_qflags
&
452 (XFS_PQUOTA_ENFD
| XFS_GQUOTA_ENFD
))
453 qflags
|= XFS_OQUOTA_ENFD
;
454 if (from
->sb_qflags
&
455 (XFS_PQUOTA_CHKD
| XFS_GQUOTA_CHKD
))
456 qflags
|= XFS_OQUOTA_CHKD
;
457 to
->sb_qflags
= cpu_to_be16(qflags
);
460 * GQUOTINO and PQUOTINO cannot be used together in versions
461 * of superblock that do not have pquotino. from->sb_flags
462 * tells us which quota is active and should be copied to
463 * disk. If neither are active, we should NULL the inode.
465 * In all cases, the separate pquotino must remain 0 because it
466 * it beyond the "end" of the valid non-pquotino superblock.
468 if (from
->sb_qflags
& XFS_GQUOTA_ACCT
)
469 to
->sb_gquotino
= cpu_to_be64(from
->sb_gquotino
);
470 else if (from
->sb_qflags
& XFS_PQUOTA_ACCT
)
471 to
->sb_gquotino
= cpu_to_be64(from
->sb_pquotino
);
474 * We can't rely on just the fields being logged to tell us
475 * that it is safe to write NULLFSINO - we should only do that
476 * if quotas are not actually enabled. Hence only write
477 * NULLFSINO if both in-core quota inodes are NULL.
479 if (from
->sb_gquotino
== NULLFSINO
&&
480 from
->sb_pquotino
== NULLFSINO
)
481 to
->sb_gquotino
= cpu_to_be64(NULLFSINO
);
492 xfs_sb_quota_to_disk(to
, from
);
494 to
->sb_magicnum
= cpu_to_be32(from
->sb_magicnum
);
495 to
->sb_blocksize
= cpu_to_be32(from
->sb_blocksize
);
496 to
->sb_dblocks
= cpu_to_be64(from
->sb_dblocks
);
497 to
->sb_rblocks
= cpu_to_be64(from
->sb_rblocks
);
498 to
->sb_rextents
= cpu_to_be64(from
->sb_rextents
);
499 memcpy(&to
->sb_uuid
, &from
->sb_uuid
, sizeof(to
->sb_uuid
));
500 to
->sb_logstart
= cpu_to_be64(from
->sb_logstart
);
501 to
->sb_rootino
= cpu_to_be64(from
->sb_rootino
);
502 to
->sb_rbmino
= cpu_to_be64(from
->sb_rbmino
);
503 to
->sb_rsumino
= cpu_to_be64(from
->sb_rsumino
);
504 to
->sb_rextsize
= cpu_to_be32(from
->sb_rextsize
);
505 to
->sb_agblocks
= cpu_to_be32(from
->sb_agblocks
);
506 to
->sb_agcount
= cpu_to_be32(from
->sb_agcount
);
507 to
->sb_rbmblocks
= cpu_to_be32(from
->sb_rbmblocks
);
508 to
->sb_logblocks
= cpu_to_be32(from
->sb_logblocks
);
509 to
->sb_versionnum
= cpu_to_be16(from
->sb_versionnum
);
510 to
->sb_sectsize
= cpu_to_be16(from
->sb_sectsize
);
511 to
->sb_inodesize
= cpu_to_be16(from
->sb_inodesize
);
512 to
->sb_inopblock
= cpu_to_be16(from
->sb_inopblock
);
513 memcpy(&to
->sb_fname
, &from
->sb_fname
, sizeof(to
->sb_fname
));
514 to
->sb_blocklog
= from
->sb_blocklog
;
515 to
->sb_sectlog
= from
->sb_sectlog
;
516 to
->sb_inodelog
= from
->sb_inodelog
;
517 to
->sb_inopblog
= from
->sb_inopblog
;
518 to
->sb_agblklog
= from
->sb_agblklog
;
519 to
->sb_rextslog
= from
->sb_rextslog
;
520 to
->sb_inprogress
= from
->sb_inprogress
;
521 to
->sb_imax_pct
= from
->sb_imax_pct
;
522 to
->sb_icount
= cpu_to_be64(from
->sb_icount
);
523 to
->sb_ifree
= cpu_to_be64(from
->sb_ifree
);
524 to
->sb_fdblocks
= cpu_to_be64(from
->sb_fdblocks
);
525 to
->sb_frextents
= cpu_to_be64(from
->sb_frextents
);
527 to
->sb_flags
= from
->sb_flags
;
528 to
->sb_shared_vn
= from
->sb_shared_vn
;
529 to
->sb_inoalignmt
= cpu_to_be32(from
->sb_inoalignmt
);
530 to
->sb_unit
= cpu_to_be32(from
->sb_unit
);
531 to
->sb_width
= cpu_to_be32(from
->sb_width
);
532 to
->sb_dirblklog
= from
->sb_dirblklog
;
533 to
->sb_logsectlog
= from
->sb_logsectlog
;
534 to
->sb_logsectsize
= cpu_to_be16(from
->sb_logsectsize
);
535 to
->sb_logsunit
= cpu_to_be32(from
->sb_logsunit
);
538 * We need to ensure that bad_features2 always matches features2.
539 * Hence we enforce that here rather than having to remember to do it
540 * everywhere else that updates features2.
542 from
->sb_bad_features2
= from
->sb_features2
;
543 to
->sb_features2
= cpu_to_be32(from
->sb_features2
);
544 to
->sb_bad_features2
= cpu_to_be32(from
->sb_bad_features2
);
546 if (xfs_sb_version_hascrc(from
)) {
547 to
->sb_features_compat
= cpu_to_be32(from
->sb_features_compat
);
548 to
->sb_features_ro_compat
=
549 cpu_to_be32(from
->sb_features_ro_compat
);
550 to
->sb_features_incompat
=
551 cpu_to_be32(from
->sb_features_incompat
);
552 to
->sb_features_log_incompat
=
553 cpu_to_be32(from
->sb_features_log_incompat
);
554 to
->sb_spino_align
= cpu_to_be32(from
->sb_spino_align
);
555 to
->sb_lsn
= cpu_to_be64(from
->sb_lsn
);
556 if (xfs_sb_version_hasmetauuid(from
))
557 uuid_copy(&to
->sb_meta_uuid
, &from
->sb_meta_uuid
);
566 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
570 * Use call variant which doesn't convert quota flags from disk
571 * format, because xfs_mount_validate_sb checks the on-disk flags.
573 __xfs_sb_from_disk(&sb
, XFS_BUF_TO_SBP(bp
), false);
576 * Only check the in progress field for the primary superblock as
577 * mkfs.xfs doesn't clear it from secondary superblocks.
579 return xfs_mount_validate_sb(mp
, &sb
,
580 bp
->b_maps
[0].bm_bn
== XFS_SB_DADDR
,
585 * If the superblock has the CRC feature bit set or the CRC field is non-null,
586 * check that the CRC is valid. We check the CRC field is non-null because a
587 * single bit error could clear the feature bit and unused parts of the
588 * superblock are supposed to be zero. Hence a non-null crc field indicates that
589 * we've potentially lost a feature bit and we should check it anyway.
591 * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
592 * last field in V4 secondary superblocks. So for secondary superblocks,
593 * we are more forgiving, and ignore CRC failures if the primary doesn't
594 * indicate that the fs version is V5.
600 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
601 struct xfs_dsb
*dsb
= XFS_BUF_TO_SBP(bp
);
605 * open code the version check to avoid needing to convert the entire
606 * superblock from disk order just to check the version number
608 if (dsb
->sb_magicnum
== cpu_to_be32(XFS_SB_MAGIC
) &&
609 (((be16_to_cpu(dsb
->sb_versionnum
) & XFS_SB_VERSION_NUMBITS
) ==
613 if (!xfs_buf_verify_cksum(bp
, XFS_SB_CRC_OFF
)) {
614 /* Only fail bad secondaries on a known V5 filesystem */
615 if (bp
->b_bn
== XFS_SB_DADDR
||
616 xfs_sb_version_hascrc(&mp
->m_sb
)) {
622 error
= xfs_sb_verify(bp
, true);
626 xfs_buf_ioerror(bp
, error
);
627 if (error
== -EFSCORRUPTED
|| error
== -EFSBADCRC
)
628 xfs_verifier_error(bp
);
633 * We may be probed for a filesystem match, so we may not want to emit
634 * messages when the superblock buffer is not actually an XFS superblock.
635 * If we find an XFS superblock, then run a normal, noisy mount because we are
636 * really going to mount it and want to know about errors.
639 xfs_sb_quiet_read_verify(
642 struct xfs_dsb
*dsb
= XFS_BUF_TO_SBP(bp
);
644 if (dsb
->sb_magicnum
== cpu_to_be32(XFS_SB_MAGIC
)) {
645 /* XFS filesystem, verify noisily! */
646 xfs_sb_read_verify(bp
);
650 xfs_buf_ioerror(bp
, -EWRONGFS
);
657 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
658 struct xfs_buf_log_item
*bip
= bp
->b_fspriv
;
661 error
= xfs_sb_verify(bp
, false);
663 xfs_buf_ioerror(bp
, error
);
664 xfs_verifier_error(bp
);
668 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
672 XFS_BUF_TO_SBP(bp
)->sb_lsn
= cpu_to_be64(bip
->bli_item
.li_lsn
);
674 xfs_buf_update_cksum(bp
, XFS_SB_CRC_OFF
);
677 const struct xfs_buf_ops xfs_sb_buf_ops
= {
679 .verify_read
= xfs_sb_read_verify
,
680 .verify_write
= xfs_sb_write_verify
,
683 const struct xfs_buf_ops xfs_sb_quiet_buf_ops
= {
684 .name
= "xfs_sb_quiet",
685 .verify_read
= xfs_sb_quiet_read_verify
,
686 .verify_write
= xfs_sb_write_verify
,
692 * Mount initialization code establishing various mount
693 * fields from the superblock associated with the given
698 struct xfs_mount
*mp
,
701 mp
->m_agfrotor
= mp
->m_agirotor
= 0;
702 spin_lock_init(&mp
->m_agirotor_lock
);
703 mp
->m_maxagi
= mp
->m_sb
.sb_agcount
;
704 mp
->m_blkbit_log
= sbp
->sb_blocklog
+ XFS_NBBYLOG
;
705 mp
->m_blkbb_log
= sbp
->sb_blocklog
- BBSHIFT
;
706 mp
->m_sectbb_log
= sbp
->sb_sectlog
- BBSHIFT
;
707 mp
->m_agno_log
= xfs_highbit32(sbp
->sb_agcount
- 1) + 1;
708 mp
->m_agino_log
= sbp
->sb_inopblog
+ sbp
->sb_agblklog
;
709 mp
->m_blockmask
= sbp
->sb_blocksize
- 1;
710 mp
->m_blockwsize
= sbp
->sb_blocksize
>> XFS_WORDLOG
;
711 mp
->m_blockwmask
= mp
->m_blockwsize
- 1;
713 mp
->m_alloc_mxr
[0] = xfs_allocbt_maxrecs(mp
, sbp
->sb_blocksize
, 1);
714 mp
->m_alloc_mxr
[1] = xfs_allocbt_maxrecs(mp
, sbp
->sb_blocksize
, 0);
715 mp
->m_alloc_mnr
[0] = mp
->m_alloc_mxr
[0] / 2;
716 mp
->m_alloc_mnr
[1] = mp
->m_alloc_mxr
[1] / 2;
718 mp
->m_inobt_mxr
[0] = xfs_inobt_maxrecs(mp
, sbp
->sb_blocksize
, 1);
719 mp
->m_inobt_mxr
[1] = xfs_inobt_maxrecs(mp
, sbp
->sb_blocksize
, 0);
720 mp
->m_inobt_mnr
[0] = mp
->m_inobt_mxr
[0] / 2;
721 mp
->m_inobt_mnr
[1] = mp
->m_inobt_mxr
[1] / 2;
723 mp
->m_bmap_dmxr
[0] = xfs_bmbt_maxrecs(mp
, sbp
->sb_blocksize
, 1);
724 mp
->m_bmap_dmxr
[1] = xfs_bmbt_maxrecs(mp
, sbp
->sb_blocksize
, 0);
725 mp
->m_bmap_dmnr
[0] = mp
->m_bmap_dmxr
[0] / 2;
726 mp
->m_bmap_dmnr
[1] = mp
->m_bmap_dmxr
[1] / 2;
728 mp
->m_rmap_mxr
[0] = xfs_rmapbt_maxrecs(mp
, sbp
->sb_blocksize
, 1);
729 mp
->m_rmap_mxr
[1] = xfs_rmapbt_maxrecs(mp
, sbp
->sb_blocksize
, 0);
730 mp
->m_rmap_mnr
[0] = mp
->m_rmap_mxr
[0] / 2;
731 mp
->m_rmap_mnr
[1] = mp
->m_rmap_mxr
[1] / 2;
733 mp
->m_refc_mxr
[0] = xfs_refcountbt_maxrecs(mp
, sbp
->sb_blocksize
,
735 mp
->m_refc_mxr
[1] = xfs_refcountbt_maxrecs(mp
, sbp
->sb_blocksize
,
737 mp
->m_refc_mnr
[0] = mp
->m_refc_mxr
[0] / 2;
738 mp
->m_refc_mnr
[1] = mp
->m_refc_mxr
[1] / 2;
740 mp
->m_bsize
= XFS_FSB_TO_BB(mp
, 1);
741 mp
->m_ialloc_inos
= (int)MAX((uint16_t)XFS_INODES_PER_CHUNK
,
743 mp
->m_ialloc_blks
= mp
->m_ialloc_inos
>> sbp
->sb_inopblog
;
745 if (sbp
->sb_spino_align
)
746 mp
->m_ialloc_min_blks
= sbp
->sb_spino_align
;
748 mp
->m_ialloc_min_blks
= mp
->m_ialloc_blks
;
749 mp
->m_alloc_set_aside
= xfs_alloc_set_aside(mp
);
750 mp
->m_ag_max_usable
= xfs_alloc_ag_max_usable(mp
);
754 * xfs_initialize_perag_data
756 * Read in each per-ag structure so we can count up the number of
757 * allocated inodes, free inodes and used filesystem blocks as this
758 * information is no longer persistent in the superblock. Once we have
759 * this information, write it into the in-core superblock structure.
762 xfs_initialize_perag_data(
763 struct xfs_mount
*mp
,
764 xfs_agnumber_t agcount
)
766 xfs_agnumber_t index
;
768 xfs_sb_t
*sbp
= &mp
->m_sb
;
772 uint64_t bfreelst
= 0;
776 for (index
= 0; index
< agcount
; index
++) {
778 * read the agf, then the agi. This gets us
779 * all the information we need and populates the
780 * per-ag structures for us.
782 error
= xfs_alloc_pagf_init(mp
, NULL
, index
, 0);
786 error
= xfs_ialloc_pagi_init(mp
, NULL
, index
);
789 pag
= xfs_perag_get(mp
, index
);
790 ifree
+= pag
->pagi_freecount
;
791 ialloc
+= pag
->pagi_count
;
792 bfree
+= pag
->pagf_freeblks
;
793 bfreelst
+= pag
->pagf_flcount
;
794 btree
+= pag
->pagf_btreeblks
;
798 /* Overwrite incore superblock counters with just-read data */
799 spin_lock(&mp
->m_sb_lock
);
800 sbp
->sb_ifree
= ifree
;
801 sbp
->sb_icount
= ialloc
;
802 sbp
->sb_fdblocks
= bfree
+ bfreelst
+ btree
;
803 spin_unlock(&mp
->m_sb_lock
);
805 xfs_reinit_percpu_counters(mp
);
811 * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
812 * into the superblock buffer to be logged. It does not provide the higher
813 * level of locking that is needed to protect the in-core superblock from
818 struct xfs_trans
*tp
)
820 struct xfs_mount
*mp
= tp
->t_mountp
;
821 struct xfs_buf
*bp
= xfs_trans_getsb(tp
, mp
, 0);
823 mp
->m_sb
.sb_icount
= percpu_counter_sum(&mp
->m_icount
);
824 mp
->m_sb
.sb_ifree
= percpu_counter_sum(&mp
->m_ifree
);
825 mp
->m_sb
.sb_fdblocks
= percpu_counter_sum(&mp
->m_fdblocks
);
827 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp
), &mp
->m_sb
);
828 xfs_trans_buf_set_type(tp
, bp
, XFS_BLFT_SB_BUF
);
829 xfs_trans_log_buf(tp
, bp
, 0, sizeof(struct xfs_dsb
));
835 * Sync the superblock to disk.
837 * Note that the caller is responsible for checking the frozen state of the
838 * filesystem. This procedure uses the non-blocking transaction allocator and
839 * thus will allow modifications to a frozen fs. This is required because this
840 * code can be called during the process of freezing where use of the high-level
841 * allocator would deadlock.
845 struct xfs_mount
*mp
,
848 struct xfs_trans
*tp
;
851 error
= xfs_trans_alloc(mp
, &M_RES(mp
)->tr_sb
, 0, 0,
852 XFS_TRANS_NO_WRITECOUNT
, &tp
);
858 xfs_trans_set_sync(tp
);
859 return xfs_trans_commit(tp
);