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_inode.h"
28 #include "xfs_ialloc.h"
29 #include "xfs_alloc.h"
30 #include "xfs_trace.h"
31 #include "xfs_cksum.h"
32 #include "xfs_trans.h"
33 #include "xfs_bmap_btree.h"
34 #include "xfs_alloc_btree.h"
35 #include "xfs_ialloc_btree.h"
38 * Physical superblock buffer manipulations. Shared with libxfs in userspace.
42 * Reference counting access wrappers to the perag structures.
43 * Because we never free per-ag structures, the only thing we
44 * have to protect against changes is the tree structure itself.
51 struct xfs_perag
*pag
;
55 pag
= radix_tree_lookup(&mp
->m_perag_tree
, agno
);
57 ASSERT(atomic_read(&pag
->pag_ref
) >= 0);
58 ref
= atomic_inc_return(&pag
->pag_ref
);
61 trace_xfs_perag_get(mp
, agno
, ref
, _RET_IP_
);
66 * search from @first to find the next perag with the given tag set.
74 struct xfs_perag
*pag
;
79 found
= radix_tree_gang_lookup_tag(&mp
->m_perag_tree
,
80 (void **)&pag
, first
, 1, tag
);
85 ref
= atomic_inc_return(&pag
->pag_ref
);
87 trace_xfs_perag_get_tag(mp
, pag
->pag_agno
, ref
, _RET_IP_
);
93 struct xfs_perag
*pag
)
97 ASSERT(atomic_read(&pag
->pag_ref
) > 0);
98 ref
= atomic_dec_return(&pag
->pag_ref
);
99 trace_xfs_perag_put(pag
->pag_mount
, pag
->pag_agno
, ref
, _RET_IP_
);
103 * Check the validity of the SB found.
106 xfs_mount_validate_sb(
109 bool check_inprogress
,
112 if (sbp
->sb_magicnum
!= XFS_SB_MAGIC
) {
113 xfs_warn(mp
, "bad magic number");
118 if (!xfs_sb_good_version(sbp
)) {
119 xfs_warn(mp
, "bad version");
124 * Version 5 superblock feature mask validation. Reject combinations the
125 * kernel cannot support up front before checking anything else. For
126 * write validation, we don't need to check feature masks.
128 if (check_version
&& XFS_SB_VERSION_NUM(sbp
) == XFS_SB_VERSION_5
) {
129 if (xfs_sb_has_compat_feature(sbp
,
130 XFS_SB_FEAT_COMPAT_UNKNOWN
)) {
132 "Superblock has unknown compatible features (0x%x) enabled.",
133 (sbp
->sb_features_compat
&
134 XFS_SB_FEAT_COMPAT_UNKNOWN
));
136 "Using a more recent kernel is recommended.");
139 if (xfs_sb_has_ro_compat_feature(sbp
,
140 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
)) {
142 "Superblock has unknown read-only compatible features (0x%x) enabled.",
143 (sbp
->sb_features_ro_compat
&
144 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
));
145 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
147 "Attempted to mount read-only compatible filesystem read-write.");
149 "Filesystem can only be safely mounted read only.");
154 if (xfs_sb_has_incompat_feature(sbp
,
155 XFS_SB_FEAT_INCOMPAT_UNKNOWN
)) {
157 "Superblock has unknown incompatible features (0x%x) enabled.",
158 (sbp
->sb_features_incompat
&
159 XFS_SB_FEAT_INCOMPAT_UNKNOWN
));
161 "Filesystem can not be safely mounted by this kernel.");
164 } else if (xfs_sb_version_hascrc(sbp
)) {
166 * We can't read verify the sb LSN because the read verifier is
167 * called before the log is allocated and processed. We know the
168 * log is set up before write verifier (!check_version) calls,
169 * so just check it here.
171 if (!xfs_log_check_lsn(mp
, sbp
->sb_lsn
))
172 return -EFSCORRUPTED
;
175 if (xfs_sb_version_has_pquotino(sbp
)) {
176 if (sbp
->sb_qflags
& (XFS_OQUOTA_ENFD
| XFS_OQUOTA_CHKD
)) {
178 "Version 5 of Super block has XFS_OQUOTA bits.");
179 return -EFSCORRUPTED
;
181 } else if (sbp
->sb_qflags
& (XFS_PQUOTA_ENFD
| XFS_GQUOTA_ENFD
|
182 XFS_PQUOTA_CHKD
| XFS_GQUOTA_CHKD
)) {
184 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.");
185 return -EFSCORRUPTED
;
189 * Full inode chunks must be aligned to inode chunk size when
190 * sparse inodes are enabled to support the sparse chunk
191 * allocation algorithm and prevent overlapping inode records.
193 if (xfs_sb_version_hassparseinodes(sbp
)) {
196 align
= XFS_INODES_PER_CHUNK
* sbp
->sb_inodesize
198 if (sbp
->sb_inoalignmt
!= align
) {
200 "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.",
201 sbp
->sb_inoalignmt
, align
);
207 sbp
->sb_logstart
== 0 && mp
->m_logdev_targp
== mp
->m_ddev_targp
)) {
209 "filesystem is marked as having an external log; "
210 "specify logdev on the mount command line.");
215 sbp
->sb_logstart
!= 0 && mp
->m_logdev_targp
!= mp
->m_ddev_targp
)) {
217 "filesystem is marked as having an internal log; "
218 "do not specify logdev on the mount command line.");
223 * More sanity checking. Most of these were stolen directly from
227 sbp
->sb_agcount
<= 0 ||
228 sbp
->sb_sectsize
< XFS_MIN_SECTORSIZE
||
229 sbp
->sb_sectsize
> XFS_MAX_SECTORSIZE
||
230 sbp
->sb_sectlog
< XFS_MIN_SECTORSIZE_LOG
||
231 sbp
->sb_sectlog
> XFS_MAX_SECTORSIZE_LOG
||
232 sbp
->sb_sectsize
!= (1 << sbp
->sb_sectlog
) ||
233 sbp
->sb_blocksize
< XFS_MIN_BLOCKSIZE
||
234 sbp
->sb_blocksize
> XFS_MAX_BLOCKSIZE
||
235 sbp
->sb_blocklog
< XFS_MIN_BLOCKSIZE_LOG
||
236 sbp
->sb_blocklog
> XFS_MAX_BLOCKSIZE_LOG
||
237 sbp
->sb_blocksize
!= (1 << sbp
->sb_blocklog
) ||
238 sbp
->sb_dirblklog
> XFS_MAX_BLOCKSIZE_LOG
||
239 sbp
->sb_inodesize
< XFS_DINODE_MIN_SIZE
||
240 sbp
->sb_inodesize
> XFS_DINODE_MAX_SIZE
||
241 sbp
->sb_inodelog
< XFS_DINODE_MIN_LOG
||
242 sbp
->sb_inodelog
> XFS_DINODE_MAX_LOG
||
243 sbp
->sb_inodesize
!= (1 << sbp
->sb_inodelog
) ||
244 sbp
->sb_logsunit
> XLOG_MAX_RECORD_BSIZE
||
245 sbp
->sb_inopblock
!= howmany(sbp
->sb_blocksize
,sbp
->sb_inodesize
) ||
246 (sbp
->sb_blocklog
- sbp
->sb_inodelog
!= sbp
->sb_inopblog
) ||
247 (sbp
->sb_rextsize
* sbp
->sb_blocksize
> XFS_MAX_RTEXTSIZE
) ||
248 (sbp
->sb_rextsize
* sbp
->sb_blocksize
< XFS_MIN_RTEXTSIZE
) ||
249 (sbp
->sb_imax_pct
> 100 /* zero sb_imax_pct is valid */) ||
250 sbp
->sb_dblocks
== 0 ||
251 sbp
->sb_dblocks
> XFS_MAX_DBLOCKS(sbp
) ||
252 sbp
->sb_dblocks
< XFS_MIN_DBLOCKS(sbp
) ||
253 sbp
->sb_shared_vn
!= 0)) {
254 xfs_notice(mp
, "SB sanity check failed");
255 return -EFSCORRUPTED
;
259 * Currently only very few inode sizes are supported.
261 switch (sbp
->sb_inodesize
) {
268 xfs_warn(mp
, "inode size of %d bytes not supported",
273 if (xfs_sb_validate_fsb_count(sbp
, sbp
->sb_dblocks
) ||
274 xfs_sb_validate_fsb_count(sbp
, sbp
->sb_rblocks
)) {
276 "file system too large to be mounted on this system.");
284 xfs_sb_quota_from_disk(struct xfs_sb
*sbp
)
287 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
288 * leads to in-core values having two different values for a quota
289 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
292 * Note that this change affect only the in-core values. These
293 * values are not written back to disk unless any quota information
294 * is written to the disk. Even in that case, sb_pquotino field is
295 * not written to disk unless the superblock supports pquotino.
297 if (sbp
->sb_uquotino
== 0)
298 sbp
->sb_uquotino
= NULLFSINO
;
299 if (sbp
->sb_gquotino
== 0)
300 sbp
->sb_gquotino
= NULLFSINO
;
301 if (sbp
->sb_pquotino
== 0)
302 sbp
->sb_pquotino
= NULLFSINO
;
305 * We need to do these manipilations only if we are working
306 * with an older version of on-disk superblock.
308 if (xfs_sb_version_has_pquotino(sbp
))
311 if (sbp
->sb_qflags
& XFS_OQUOTA_ENFD
)
312 sbp
->sb_qflags
|= (sbp
->sb_qflags
& XFS_PQUOTA_ACCT
) ?
313 XFS_PQUOTA_ENFD
: XFS_GQUOTA_ENFD
;
314 if (sbp
->sb_qflags
& XFS_OQUOTA_CHKD
)
315 sbp
->sb_qflags
|= (sbp
->sb_qflags
& XFS_PQUOTA_ACCT
) ?
316 XFS_PQUOTA_CHKD
: XFS_GQUOTA_CHKD
;
317 sbp
->sb_qflags
&= ~(XFS_OQUOTA_ENFD
| XFS_OQUOTA_CHKD
);
319 if (sbp
->sb_qflags
& XFS_PQUOTA_ACCT
) {
321 * In older version of superblock, on-disk superblock only
322 * has sb_gquotino, and in-core superblock has both sb_gquotino
323 * and sb_pquotino. But, only one of them is supported at any
324 * point of time. So, if PQUOTA is set in disk superblock,
325 * copy over sb_gquotino to sb_pquotino.
327 sbp
->sb_pquotino
= sbp
->sb_gquotino
;
328 sbp
->sb_gquotino
= NULLFSINO
;
338 to
->sb_magicnum
= be32_to_cpu(from
->sb_magicnum
);
339 to
->sb_blocksize
= be32_to_cpu(from
->sb_blocksize
);
340 to
->sb_dblocks
= be64_to_cpu(from
->sb_dblocks
);
341 to
->sb_rblocks
= be64_to_cpu(from
->sb_rblocks
);
342 to
->sb_rextents
= be64_to_cpu(from
->sb_rextents
);
343 memcpy(&to
->sb_uuid
, &from
->sb_uuid
, sizeof(to
->sb_uuid
));
344 to
->sb_logstart
= be64_to_cpu(from
->sb_logstart
);
345 to
->sb_rootino
= be64_to_cpu(from
->sb_rootino
);
346 to
->sb_rbmino
= be64_to_cpu(from
->sb_rbmino
);
347 to
->sb_rsumino
= be64_to_cpu(from
->sb_rsumino
);
348 to
->sb_rextsize
= be32_to_cpu(from
->sb_rextsize
);
349 to
->sb_agblocks
= be32_to_cpu(from
->sb_agblocks
);
350 to
->sb_agcount
= be32_to_cpu(from
->sb_agcount
);
351 to
->sb_rbmblocks
= be32_to_cpu(from
->sb_rbmblocks
);
352 to
->sb_logblocks
= be32_to_cpu(from
->sb_logblocks
);
353 to
->sb_versionnum
= be16_to_cpu(from
->sb_versionnum
);
354 to
->sb_sectsize
= be16_to_cpu(from
->sb_sectsize
);
355 to
->sb_inodesize
= be16_to_cpu(from
->sb_inodesize
);
356 to
->sb_inopblock
= be16_to_cpu(from
->sb_inopblock
);
357 memcpy(&to
->sb_fname
, &from
->sb_fname
, sizeof(to
->sb_fname
));
358 to
->sb_blocklog
= from
->sb_blocklog
;
359 to
->sb_sectlog
= from
->sb_sectlog
;
360 to
->sb_inodelog
= from
->sb_inodelog
;
361 to
->sb_inopblog
= from
->sb_inopblog
;
362 to
->sb_agblklog
= from
->sb_agblklog
;
363 to
->sb_rextslog
= from
->sb_rextslog
;
364 to
->sb_inprogress
= from
->sb_inprogress
;
365 to
->sb_imax_pct
= from
->sb_imax_pct
;
366 to
->sb_icount
= be64_to_cpu(from
->sb_icount
);
367 to
->sb_ifree
= be64_to_cpu(from
->sb_ifree
);
368 to
->sb_fdblocks
= be64_to_cpu(from
->sb_fdblocks
);
369 to
->sb_frextents
= be64_to_cpu(from
->sb_frextents
);
370 to
->sb_uquotino
= be64_to_cpu(from
->sb_uquotino
);
371 to
->sb_gquotino
= be64_to_cpu(from
->sb_gquotino
);
372 to
->sb_qflags
= be16_to_cpu(from
->sb_qflags
);
373 to
->sb_flags
= from
->sb_flags
;
374 to
->sb_shared_vn
= from
->sb_shared_vn
;
375 to
->sb_inoalignmt
= be32_to_cpu(from
->sb_inoalignmt
);
376 to
->sb_unit
= be32_to_cpu(from
->sb_unit
);
377 to
->sb_width
= be32_to_cpu(from
->sb_width
);
378 to
->sb_dirblklog
= from
->sb_dirblklog
;
379 to
->sb_logsectlog
= from
->sb_logsectlog
;
380 to
->sb_logsectsize
= be16_to_cpu(from
->sb_logsectsize
);
381 to
->sb_logsunit
= be32_to_cpu(from
->sb_logsunit
);
382 to
->sb_features2
= be32_to_cpu(from
->sb_features2
);
383 to
->sb_bad_features2
= be32_to_cpu(from
->sb_bad_features2
);
384 to
->sb_features_compat
= be32_to_cpu(from
->sb_features_compat
);
385 to
->sb_features_ro_compat
= be32_to_cpu(from
->sb_features_ro_compat
);
386 to
->sb_features_incompat
= be32_to_cpu(from
->sb_features_incompat
);
387 to
->sb_features_log_incompat
=
388 be32_to_cpu(from
->sb_features_log_incompat
);
389 /* crc is only used on disk, not in memory; just init to 0 here. */
391 to
->sb_spino_align
= be32_to_cpu(from
->sb_spino_align
);
392 to
->sb_pquotino
= be64_to_cpu(from
->sb_pquotino
);
393 to
->sb_lsn
= be64_to_cpu(from
->sb_lsn
);
395 * sb_meta_uuid is only on disk if it differs from sb_uuid and the
396 * feature flag is set; if not set we keep it only in memory.
398 if (xfs_sb_version_hasmetauuid(to
))
399 uuid_copy(&to
->sb_meta_uuid
, &from
->sb_meta_uuid
);
401 uuid_copy(&to
->sb_meta_uuid
, &from
->sb_uuid
);
402 /* Convert on-disk flags to in-memory flags? */
404 xfs_sb_quota_from_disk(to
);
412 __xfs_sb_from_disk(to
, from
, true);
416 xfs_sb_quota_to_disk(
420 __uint16_t qflags
= from
->sb_qflags
;
422 to
->sb_uquotino
= cpu_to_be64(from
->sb_uquotino
);
423 if (xfs_sb_version_has_pquotino(from
)) {
424 to
->sb_qflags
= cpu_to_be16(from
->sb_qflags
);
425 to
->sb_gquotino
= cpu_to_be64(from
->sb_gquotino
);
426 to
->sb_pquotino
= cpu_to_be64(from
->sb_pquotino
);
431 * The in-core version of sb_qflags do not have XFS_OQUOTA_*
432 * flags, whereas the on-disk version does. So, convert incore
433 * XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
435 qflags
&= ~(XFS_PQUOTA_ENFD
| XFS_PQUOTA_CHKD
|
436 XFS_GQUOTA_ENFD
| XFS_GQUOTA_CHKD
);
438 if (from
->sb_qflags
&
439 (XFS_PQUOTA_ENFD
| XFS_GQUOTA_ENFD
))
440 qflags
|= XFS_OQUOTA_ENFD
;
441 if (from
->sb_qflags
&
442 (XFS_PQUOTA_CHKD
| XFS_GQUOTA_CHKD
))
443 qflags
|= XFS_OQUOTA_CHKD
;
444 to
->sb_qflags
= cpu_to_be16(qflags
);
447 * GQUOTINO and PQUOTINO cannot be used together in versions
448 * of superblock that do not have pquotino. from->sb_flags
449 * tells us which quota is active and should be copied to
450 * disk. If neither are active, we should NULL the inode.
452 * In all cases, the separate pquotino must remain 0 because it
453 * it beyond the "end" of the valid non-pquotino superblock.
455 if (from
->sb_qflags
& XFS_GQUOTA_ACCT
)
456 to
->sb_gquotino
= cpu_to_be64(from
->sb_gquotino
);
457 else if (from
->sb_qflags
& XFS_PQUOTA_ACCT
)
458 to
->sb_gquotino
= cpu_to_be64(from
->sb_pquotino
);
461 * We can't rely on just the fields being logged to tell us
462 * that it is safe to write NULLFSINO - we should only do that
463 * if quotas are not actually enabled. Hence only write
464 * NULLFSINO if both in-core quota inodes are NULL.
466 if (from
->sb_gquotino
== NULLFSINO
&&
467 from
->sb_pquotino
== NULLFSINO
)
468 to
->sb_gquotino
= cpu_to_be64(NULLFSINO
);
479 xfs_sb_quota_to_disk(to
, from
);
481 to
->sb_magicnum
= cpu_to_be32(from
->sb_magicnum
);
482 to
->sb_blocksize
= cpu_to_be32(from
->sb_blocksize
);
483 to
->sb_dblocks
= cpu_to_be64(from
->sb_dblocks
);
484 to
->sb_rblocks
= cpu_to_be64(from
->sb_rblocks
);
485 to
->sb_rextents
= cpu_to_be64(from
->sb_rextents
);
486 memcpy(&to
->sb_uuid
, &from
->sb_uuid
, sizeof(to
->sb_uuid
));
487 to
->sb_logstart
= cpu_to_be64(from
->sb_logstart
);
488 to
->sb_rootino
= cpu_to_be64(from
->sb_rootino
);
489 to
->sb_rbmino
= cpu_to_be64(from
->sb_rbmino
);
490 to
->sb_rsumino
= cpu_to_be64(from
->sb_rsumino
);
491 to
->sb_rextsize
= cpu_to_be32(from
->sb_rextsize
);
492 to
->sb_agblocks
= cpu_to_be32(from
->sb_agblocks
);
493 to
->sb_agcount
= cpu_to_be32(from
->sb_agcount
);
494 to
->sb_rbmblocks
= cpu_to_be32(from
->sb_rbmblocks
);
495 to
->sb_logblocks
= cpu_to_be32(from
->sb_logblocks
);
496 to
->sb_versionnum
= cpu_to_be16(from
->sb_versionnum
);
497 to
->sb_sectsize
= cpu_to_be16(from
->sb_sectsize
);
498 to
->sb_inodesize
= cpu_to_be16(from
->sb_inodesize
);
499 to
->sb_inopblock
= cpu_to_be16(from
->sb_inopblock
);
500 memcpy(&to
->sb_fname
, &from
->sb_fname
, sizeof(to
->sb_fname
));
501 to
->sb_blocklog
= from
->sb_blocklog
;
502 to
->sb_sectlog
= from
->sb_sectlog
;
503 to
->sb_inodelog
= from
->sb_inodelog
;
504 to
->sb_inopblog
= from
->sb_inopblog
;
505 to
->sb_agblklog
= from
->sb_agblklog
;
506 to
->sb_rextslog
= from
->sb_rextslog
;
507 to
->sb_inprogress
= from
->sb_inprogress
;
508 to
->sb_imax_pct
= from
->sb_imax_pct
;
509 to
->sb_icount
= cpu_to_be64(from
->sb_icount
);
510 to
->sb_ifree
= cpu_to_be64(from
->sb_ifree
);
511 to
->sb_fdblocks
= cpu_to_be64(from
->sb_fdblocks
);
512 to
->sb_frextents
= cpu_to_be64(from
->sb_frextents
);
514 to
->sb_flags
= from
->sb_flags
;
515 to
->sb_shared_vn
= from
->sb_shared_vn
;
516 to
->sb_inoalignmt
= cpu_to_be32(from
->sb_inoalignmt
);
517 to
->sb_unit
= cpu_to_be32(from
->sb_unit
);
518 to
->sb_width
= cpu_to_be32(from
->sb_width
);
519 to
->sb_dirblklog
= from
->sb_dirblklog
;
520 to
->sb_logsectlog
= from
->sb_logsectlog
;
521 to
->sb_logsectsize
= cpu_to_be16(from
->sb_logsectsize
);
522 to
->sb_logsunit
= cpu_to_be32(from
->sb_logsunit
);
525 * We need to ensure that bad_features2 always matches features2.
526 * Hence we enforce that here rather than having to remember to do it
527 * everywhere else that updates features2.
529 from
->sb_bad_features2
= from
->sb_features2
;
530 to
->sb_features2
= cpu_to_be32(from
->sb_features2
);
531 to
->sb_bad_features2
= cpu_to_be32(from
->sb_bad_features2
);
533 if (xfs_sb_version_hascrc(from
)) {
534 to
->sb_features_compat
= cpu_to_be32(from
->sb_features_compat
);
535 to
->sb_features_ro_compat
=
536 cpu_to_be32(from
->sb_features_ro_compat
);
537 to
->sb_features_incompat
=
538 cpu_to_be32(from
->sb_features_incompat
);
539 to
->sb_features_log_incompat
=
540 cpu_to_be32(from
->sb_features_log_incompat
);
541 to
->sb_spino_align
= cpu_to_be32(from
->sb_spino_align
);
542 to
->sb_lsn
= cpu_to_be64(from
->sb_lsn
);
543 if (xfs_sb_version_hasmetauuid(from
))
544 uuid_copy(&to
->sb_meta_uuid
, &from
->sb_meta_uuid
);
553 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
557 * Use call variant which doesn't convert quota flags from disk
558 * format, because xfs_mount_validate_sb checks the on-disk flags.
560 __xfs_sb_from_disk(&sb
, XFS_BUF_TO_SBP(bp
), false);
563 * Only check the in progress field for the primary superblock as
564 * mkfs.xfs doesn't clear it from secondary superblocks.
566 return xfs_mount_validate_sb(mp
, &sb
, bp
->b_bn
== XFS_SB_DADDR
,
571 * If the superblock has the CRC feature bit set or the CRC field is non-null,
572 * check that the CRC is valid. We check the CRC field is non-null because a
573 * single bit error could clear the feature bit and unused parts of the
574 * superblock are supposed to be zero. Hence a non-null crc field indicates that
575 * we've potentially lost a feature bit and we should check it anyway.
577 * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
578 * last field in V4 secondary superblocks. So for secondary superblocks,
579 * we are more forgiving, and ignore CRC failures if the primary doesn't
580 * indicate that the fs version is V5.
586 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
587 struct xfs_dsb
*dsb
= XFS_BUF_TO_SBP(bp
);
591 * open code the version check to avoid needing to convert the entire
592 * superblock from disk order just to check the version number
594 if (dsb
->sb_magicnum
== cpu_to_be32(XFS_SB_MAGIC
) &&
595 (((be16_to_cpu(dsb
->sb_versionnum
) & XFS_SB_VERSION_NUMBITS
) ==
599 if (!xfs_buf_verify_cksum(bp
, XFS_SB_CRC_OFF
)) {
600 /* Only fail bad secondaries on a known V5 filesystem */
601 if (bp
->b_bn
== XFS_SB_DADDR
||
602 xfs_sb_version_hascrc(&mp
->m_sb
)) {
608 error
= xfs_sb_verify(bp
, true);
612 xfs_buf_ioerror(bp
, error
);
613 if (error
== -EFSCORRUPTED
|| error
== -EFSBADCRC
)
614 xfs_verifier_error(bp
);
619 * We may be probed for a filesystem match, so we may not want to emit
620 * messages when the superblock buffer is not actually an XFS superblock.
621 * If we find an XFS superblock, then run a normal, noisy mount because we are
622 * really going to mount it and want to know about errors.
625 xfs_sb_quiet_read_verify(
628 struct xfs_dsb
*dsb
= XFS_BUF_TO_SBP(bp
);
630 if (dsb
->sb_magicnum
== cpu_to_be32(XFS_SB_MAGIC
)) {
631 /* XFS filesystem, verify noisily! */
632 xfs_sb_read_verify(bp
);
636 xfs_buf_ioerror(bp
, -EWRONGFS
);
643 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
644 struct xfs_buf_log_item
*bip
= bp
->b_fspriv
;
647 error
= xfs_sb_verify(bp
, false);
649 xfs_buf_ioerror(bp
, error
);
650 xfs_verifier_error(bp
);
654 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
658 XFS_BUF_TO_SBP(bp
)->sb_lsn
= cpu_to_be64(bip
->bli_item
.li_lsn
);
660 xfs_buf_update_cksum(bp
, XFS_SB_CRC_OFF
);
663 const struct xfs_buf_ops xfs_sb_buf_ops
= {
665 .verify_read
= xfs_sb_read_verify
,
666 .verify_write
= xfs_sb_write_verify
,
669 const struct xfs_buf_ops xfs_sb_quiet_buf_ops
= {
670 .name
= "xfs_sb_quiet",
671 .verify_read
= xfs_sb_quiet_read_verify
,
672 .verify_write
= xfs_sb_write_verify
,
678 * Mount initialization code establishing various mount
679 * fields from the superblock associated with the given
684 struct xfs_mount
*mp
,
687 mp
->m_agfrotor
= mp
->m_agirotor
= 0;
688 spin_lock_init(&mp
->m_agirotor_lock
);
689 mp
->m_maxagi
= mp
->m_sb
.sb_agcount
;
690 mp
->m_blkbit_log
= sbp
->sb_blocklog
+ XFS_NBBYLOG
;
691 mp
->m_blkbb_log
= sbp
->sb_blocklog
- BBSHIFT
;
692 mp
->m_sectbb_log
= sbp
->sb_sectlog
- BBSHIFT
;
693 mp
->m_agno_log
= xfs_highbit32(sbp
->sb_agcount
- 1) + 1;
694 mp
->m_agino_log
= sbp
->sb_inopblog
+ sbp
->sb_agblklog
;
695 mp
->m_blockmask
= sbp
->sb_blocksize
- 1;
696 mp
->m_blockwsize
= sbp
->sb_blocksize
>> XFS_WORDLOG
;
697 mp
->m_blockwmask
= mp
->m_blockwsize
- 1;
699 mp
->m_alloc_mxr
[0] = xfs_allocbt_maxrecs(mp
, sbp
->sb_blocksize
, 1);
700 mp
->m_alloc_mxr
[1] = xfs_allocbt_maxrecs(mp
, sbp
->sb_blocksize
, 0);
701 mp
->m_alloc_mnr
[0] = mp
->m_alloc_mxr
[0] / 2;
702 mp
->m_alloc_mnr
[1] = mp
->m_alloc_mxr
[1] / 2;
704 mp
->m_inobt_mxr
[0] = xfs_inobt_maxrecs(mp
, sbp
->sb_blocksize
, 1);
705 mp
->m_inobt_mxr
[1] = xfs_inobt_maxrecs(mp
, sbp
->sb_blocksize
, 0);
706 mp
->m_inobt_mnr
[0] = mp
->m_inobt_mxr
[0] / 2;
707 mp
->m_inobt_mnr
[1] = mp
->m_inobt_mxr
[1] / 2;
709 mp
->m_bmap_dmxr
[0] = xfs_bmbt_maxrecs(mp
, sbp
->sb_blocksize
, 1);
710 mp
->m_bmap_dmxr
[1] = xfs_bmbt_maxrecs(mp
, sbp
->sb_blocksize
, 0);
711 mp
->m_bmap_dmnr
[0] = mp
->m_bmap_dmxr
[0] / 2;
712 mp
->m_bmap_dmnr
[1] = mp
->m_bmap_dmxr
[1] / 2;
714 mp
->m_bsize
= XFS_FSB_TO_BB(mp
, 1);
715 mp
->m_ialloc_inos
= (int)MAX((__uint16_t
)XFS_INODES_PER_CHUNK
,
717 mp
->m_ialloc_blks
= mp
->m_ialloc_inos
>> sbp
->sb_inopblog
;
719 if (sbp
->sb_spino_align
)
720 mp
->m_ialloc_min_blks
= sbp
->sb_spino_align
;
722 mp
->m_ialloc_min_blks
= mp
->m_ialloc_blks
;
726 * xfs_initialize_perag_data
728 * Read in each per-ag structure so we can count up the number of
729 * allocated inodes, free inodes and used filesystem blocks as this
730 * information is no longer persistent in the superblock. Once we have
731 * this information, write it into the in-core superblock structure.
734 xfs_initialize_perag_data(
735 struct xfs_mount
*mp
,
736 xfs_agnumber_t agcount
)
738 xfs_agnumber_t index
;
740 xfs_sb_t
*sbp
= &mp
->m_sb
;
744 uint64_t bfreelst
= 0;
748 for (index
= 0; index
< agcount
; index
++) {
750 * read the agf, then the agi. This gets us
751 * all the information we need and populates the
752 * per-ag structures for us.
754 error
= xfs_alloc_pagf_init(mp
, NULL
, index
, 0);
758 error
= xfs_ialloc_pagi_init(mp
, NULL
, index
);
761 pag
= xfs_perag_get(mp
, index
);
762 ifree
+= pag
->pagi_freecount
;
763 ialloc
+= pag
->pagi_count
;
764 bfree
+= pag
->pagf_freeblks
;
765 bfreelst
+= pag
->pagf_flcount
;
766 btree
+= pag
->pagf_btreeblks
;
770 /* Overwrite incore superblock counters with just-read data */
771 spin_lock(&mp
->m_sb_lock
);
772 sbp
->sb_ifree
= ifree
;
773 sbp
->sb_icount
= ialloc
;
774 sbp
->sb_fdblocks
= bfree
+ bfreelst
+ btree
;
775 spin_unlock(&mp
->m_sb_lock
);
777 xfs_reinit_percpu_counters(mp
);
783 * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
784 * into the superblock buffer to be logged. It does not provide the higher
785 * level of locking that is needed to protect the in-core superblock from
790 struct xfs_trans
*tp
)
792 struct xfs_mount
*mp
= tp
->t_mountp
;
793 struct xfs_buf
*bp
= xfs_trans_getsb(tp
, mp
, 0);
795 mp
->m_sb
.sb_icount
= percpu_counter_sum(&mp
->m_icount
);
796 mp
->m_sb
.sb_ifree
= percpu_counter_sum(&mp
->m_ifree
);
797 mp
->m_sb
.sb_fdblocks
= percpu_counter_sum(&mp
->m_fdblocks
);
799 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp
), &mp
->m_sb
);
800 xfs_trans_buf_set_type(tp
, bp
, XFS_BLFT_SB_BUF
);
801 xfs_trans_log_buf(tp
, bp
, 0, sizeof(struct xfs_dsb
));
807 * Sync the superblock to disk.
809 * Note that the caller is responsible for checking the frozen state of the
810 * filesystem. This procedure uses the non-blocking transaction allocator and
811 * thus will allow modifications to a frozen fs. This is required because this
812 * code can be called during the process of freezing where use of the high-level
813 * allocator would deadlock.
817 struct xfs_mount
*mp
,
820 struct xfs_trans
*tp
;
823 tp
= _xfs_trans_alloc(mp
, XFS_TRANS_SB_CHANGE
, KM_SLEEP
);
824 error
= xfs_trans_reserve(tp
, &M_RES(mp
)->tr_sb
, 0, 0);
826 xfs_trans_cancel(tp
);
832 xfs_trans_set_sync(tp
);
833 return xfs_trans_commit(tp
);