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
20 #include "xfs_format.h"
24 #include "xfs_trans.h"
25 #include "xfs_trans_priv.h"
28 #include "xfs_mount.h"
29 #include "xfs_da_btree.h"
30 #include "xfs_dir2_format.h"
32 #include "xfs_bmap_btree.h"
33 #include "xfs_alloc_btree.h"
34 #include "xfs_ialloc_btree.h"
35 #include "xfs_dinode.h"
36 #include "xfs_inode.h"
37 #include "xfs_btree.h"
38 #include "xfs_ialloc.h"
39 #include "xfs_alloc.h"
40 #include "xfs_rtalloc.h"
42 #include "xfs_error.h"
43 #include "xfs_quota.h"
44 #include "xfs_fsops.h"
45 #include "xfs_trace.h"
46 #include "xfs_cksum.h"
47 #include "xfs_buf_item.h"
50 * Physical superblock buffer manipulations. Shared with libxfs in userspace.
55 short type
; /* 0 = integer
56 * 1 = binary / string (no translation)
59 { offsetof(xfs_sb_t
, sb_magicnum
), 0 },
60 { offsetof(xfs_sb_t
, sb_blocksize
), 0 },
61 { offsetof(xfs_sb_t
, sb_dblocks
), 0 },
62 { offsetof(xfs_sb_t
, sb_rblocks
), 0 },
63 { offsetof(xfs_sb_t
, sb_rextents
), 0 },
64 { offsetof(xfs_sb_t
, sb_uuid
), 1 },
65 { offsetof(xfs_sb_t
, sb_logstart
), 0 },
66 { offsetof(xfs_sb_t
, sb_rootino
), 0 },
67 { offsetof(xfs_sb_t
, sb_rbmino
), 0 },
68 { offsetof(xfs_sb_t
, sb_rsumino
), 0 },
69 { offsetof(xfs_sb_t
, sb_rextsize
), 0 },
70 { offsetof(xfs_sb_t
, sb_agblocks
), 0 },
71 { offsetof(xfs_sb_t
, sb_agcount
), 0 },
72 { offsetof(xfs_sb_t
, sb_rbmblocks
), 0 },
73 { offsetof(xfs_sb_t
, sb_logblocks
), 0 },
74 { offsetof(xfs_sb_t
, sb_versionnum
), 0 },
75 { offsetof(xfs_sb_t
, sb_sectsize
), 0 },
76 { offsetof(xfs_sb_t
, sb_inodesize
), 0 },
77 { offsetof(xfs_sb_t
, sb_inopblock
), 0 },
78 { offsetof(xfs_sb_t
, sb_fname
[0]), 1 },
79 { offsetof(xfs_sb_t
, sb_blocklog
), 0 },
80 { offsetof(xfs_sb_t
, sb_sectlog
), 0 },
81 { offsetof(xfs_sb_t
, sb_inodelog
), 0 },
82 { offsetof(xfs_sb_t
, sb_inopblog
), 0 },
83 { offsetof(xfs_sb_t
, sb_agblklog
), 0 },
84 { offsetof(xfs_sb_t
, sb_rextslog
), 0 },
85 { offsetof(xfs_sb_t
, sb_inprogress
), 0 },
86 { offsetof(xfs_sb_t
, sb_imax_pct
), 0 },
87 { offsetof(xfs_sb_t
, sb_icount
), 0 },
88 { offsetof(xfs_sb_t
, sb_ifree
), 0 },
89 { offsetof(xfs_sb_t
, sb_fdblocks
), 0 },
90 { offsetof(xfs_sb_t
, sb_frextents
), 0 },
91 { offsetof(xfs_sb_t
, sb_uquotino
), 0 },
92 { offsetof(xfs_sb_t
, sb_gquotino
), 0 },
93 { offsetof(xfs_sb_t
, sb_qflags
), 0 },
94 { offsetof(xfs_sb_t
, sb_flags
), 0 },
95 { offsetof(xfs_sb_t
, sb_shared_vn
), 0 },
96 { offsetof(xfs_sb_t
, sb_inoalignmt
), 0 },
97 { offsetof(xfs_sb_t
, sb_unit
), 0 },
98 { offsetof(xfs_sb_t
, sb_width
), 0 },
99 { offsetof(xfs_sb_t
, sb_dirblklog
), 0 },
100 { offsetof(xfs_sb_t
, sb_logsectlog
), 0 },
101 { offsetof(xfs_sb_t
, sb_logsectsize
), 0 },
102 { offsetof(xfs_sb_t
, sb_logsunit
), 0 },
103 { offsetof(xfs_sb_t
, sb_features2
), 0 },
104 { offsetof(xfs_sb_t
, sb_bad_features2
), 0 },
105 { offsetof(xfs_sb_t
, sb_features_compat
), 0 },
106 { offsetof(xfs_sb_t
, sb_features_ro_compat
), 0 },
107 { offsetof(xfs_sb_t
, sb_features_incompat
), 0 },
108 { offsetof(xfs_sb_t
, sb_features_log_incompat
), 0 },
109 { offsetof(xfs_sb_t
, sb_crc
), 0 },
110 { offsetof(xfs_sb_t
, sb_pad
), 0 },
111 { offsetof(xfs_sb_t
, sb_pquotino
), 0 },
112 { offsetof(xfs_sb_t
, sb_lsn
), 0 },
113 { sizeof(xfs_sb_t
), 0 }
117 * Reference counting access wrappers to the perag structures.
118 * Because we never free per-ag structures, the only thing we
119 * have to protect against changes is the tree structure itself.
123 struct xfs_mount
*mp
,
126 struct xfs_perag
*pag
;
130 pag
= radix_tree_lookup(&mp
->m_perag_tree
, agno
);
132 ASSERT(atomic_read(&pag
->pag_ref
) >= 0);
133 ref
= atomic_inc_return(&pag
->pag_ref
);
136 trace_xfs_perag_get(mp
, agno
, ref
, _RET_IP_
);
141 * search from @first to find the next perag with the given tag set.
145 struct xfs_mount
*mp
,
146 xfs_agnumber_t first
,
149 struct xfs_perag
*pag
;
154 found
= radix_tree_gang_lookup_tag(&mp
->m_perag_tree
,
155 (void **)&pag
, first
, 1, tag
);
160 ref
= atomic_inc_return(&pag
->pag_ref
);
162 trace_xfs_perag_get_tag(mp
, pag
->pag_agno
, ref
, _RET_IP_
);
168 struct xfs_perag
*pag
)
172 ASSERT(atomic_read(&pag
->pag_ref
) > 0);
173 ref
= atomic_dec_return(&pag
->pag_ref
);
174 trace_xfs_perag_put(pag
->pag_mount
, pag
->pag_agno
, ref
, _RET_IP_
);
178 * Check the validity of the SB found.
181 xfs_mount_validate_sb(
184 bool check_inprogress
,
189 * If the log device and data device have the
190 * same device number, the log is internal.
191 * Consequently, the sb_logstart should be non-zero. If
192 * we have a zero sb_logstart in this case, we may be trying to mount
193 * a volume filesystem in a non-volume manner.
195 if (sbp
->sb_magicnum
!= XFS_SB_MAGIC
) {
196 xfs_warn(mp
, "bad magic number");
197 return XFS_ERROR(EWRONGFS
);
201 if (!xfs_sb_good_version(sbp
)) {
202 xfs_warn(mp
, "bad version");
203 return XFS_ERROR(EWRONGFS
);
207 * Version 5 superblock feature mask validation. Reject combinations the
208 * kernel cannot support up front before checking anything else. For
209 * write validation, we don't need to check feature masks.
211 if (check_version
&& XFS_SB_VERSION_NUM(sbp
) == XFS_SB_VERSION_5
) {
213 "Version 5 superblock detected. This kernel has EXPERIMENTAL support enabled!\n"
214 "Use of these features in this kernel is at your own risk!");
216 if (xfs_sb_has_compat_feature(sbp
,
217 XFS_SB_FEAT_COMPAT_UNKNOWN
)) {
219 "Superblock has unknown compatible features (0x%x) enabled.\n"
220 "Using a more recent kernel is recommended.",
221 (sbp
->sb_features_compat
&
222 XFS_SB_FEAT_COMPAT_UNKNOWN
));
225 if (xfs_sb_has_ro_compat_feature(sbp
,
226 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
)) {
228 "Superblock has unknown read-only compatible features (0x%x) enabled.",
229 (sbp
->sb_features_ro_compat
&
230 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
));
231 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
233 "Attempted to mount read-only compatible filesystem read-write.\n"
234 "Filesystem can only be safely mounted read only.");
235 return XFS_ERROR(EINVAL
);
238 if (xfs_sb_has_incompat_feature(sbp
,
239 XFS_SB_FEAT_INCOMPAT_UNKNOWN
)) {
241 "Superblock has unknown incompatible features (0x%x) enabled.\n"
242 "Filesystem can not be safely mounted by this kernel.",
243 (sbp
->sb_features_incompat
&
244 XFS_SB_FEAT_INCOMPAT_UNKNOWN
));
245 return XFS_ERROR(EINVAL
);
249 if (xfs_sb_version_has_pquotino(sbp
)) {
250 if (sbp
->sb_qflags
& (XFS_OQUOTA_ENFD
| XFS_OQUOTA_CHKD
)) {
252 "Version 5 of Super block has XFS_OQUOTA bits.\n");
253 return XFS_ERROR(EFSCORRUPTED
);
255 } else if (sbp
->sb_qflags
& (XFS_PQUOTA_ENFD
| XFS_GQUOTA_ENFD
|
256 XFS_PQUOTA_CHKD
| XFS_GQUOTA_CHKD
)) {
258 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.\n");
259 return XFS_ERROR(EFSCORRUPTED
);
263 sbp
->sb_logstart
== 0 && mp
->m_logdev_targp
== mp
->m_ddev_targp
)) {
265 "filesystem is marked as having an external log; "
266 "specify logdev on the mount command line.");
267 return XFS_ERROR(EINVAL
);
271 sbp
->sb_logstart
!= 0 && mp
->m_logdev_targp
!= mp
->m_ddev_targp
)) {
273 "filesystem is marked as having an internal log; "
274 "do not specify logdev on the mount command line.");
275 return XFS_ERROR(EINVAL
);
279 * More sanity checking. Most of these were stolen directly from
283 sbp
->sb_agcount
<= 0 ||
284 sbp
->sb_sectsize
< XFS_MIN_SECTORSIZE
||
285 sbp
->sb_sectsize
> XFS_MAX_SECTORSIZE
||
286 sbp
->sb_sectlog
< XFS_MIN_SECTORSIZE_LOG
||
287 sbp
->sb_sectlog
> XFS_MAX_SECTORSIZE_LOG
||
288 sbp
->sb_sectsize
!= (1 << sbp
->sb_sectlog
) ||
289 sbp
->sb_blocksize
< XFS_MIN_BLOCKSIZE
||
290 sbp
->sb_blocksize
> XFS_MAX_BLOCKSIZE
||
291 sbp
->sb_blocklog
< XFS_MIN_BLOCKSIZE_LOG
||
292 sbp
->sb_blocklog
> XFS_MAX_BLOCKSIZE_LOG
||
293 sbp
->sb_blocksize
!= (1 << sbp
->sb_blocklog
) ||
294 sbp
->sb_inodesize
< XFS_DINODE_MIN_SIZE
||
295 sbp
->sb_inodesize
> XFS_DINODE_MAX_SIZE
||
296 sbp
->sb_inodelog
< XFS_DINODE_MIN_LOG
||
297 sbp
->sb_inodelog
> XFS_DINODE_MAX_LOG
||
298 sbp
->sb_inodesize
!= (1 << sbp
->sb_inodelog
) ||
299 (sbp
->sb_blocklog
- sbp
->sb_inodelog
!= sbp
->sb_inopblog
) ||
300 (sbp
->sb_rextsize
* sbp
->sb_blocksize
> XFS_MAX_RTEXTSIZE
) ||
301 (sbp
->sb_rextsize
* sbp
->sb_blocksize
< XFS_MIN_RTEXTSIZE
) ||
302 (sbp
->sb_imax_pct
> 100 /* zero sb_imax_pct is valid */) ||
303 sbp
->sb_dblocks
== 0 ||
304 sbp
->sb_dblocks
> XFS_MAX_DBLOCKS(sbp
) ||
305 sbp
->sb_dblocks
< XFS_MIN_DBLOCKS(sbp
))) {
306 XFS_CORRUPTION_ERROR("SB sanity check failed",
307 XFS_ERRLEVEL_LOW
, mp
, sbp
);
308 return XFS_ERROR(EFSCORRUPTED
);
312 * Until this is fixed only page-sized or smaller data blocks work.
314 if (unlikely(sbp
->sb_blocksize
> PAGE_SIZE
)) {
316 "File system with blocksize %d bytes. "
317 "Only pagesize (%ld) or less will currently work.",
318 sbp
->sb_blocksize
, PAGE_SIZE
);
319 return XFS_ERROR(ENOSYS
);
323 * Currently only very few inode sizes are supported.
325 switch (sbp
->sb_inodesize
) {
332 xfs_warn(mp
, "inode size of %d bytes not supported",
334 return XFS_ERROR(ENOSYS
);
337 if (xfs_sb_validate_fsb_count(sbp
, sbp
->sb_dblocks
) ||
338 xfs_sb_validate_fsb_count(sbp
, sbp
->sb_rblocks
)) {
340 "file system too large to be mounted on this system.");
341 return XFS_ERROR(EFBIG
);
344 if (check_inprogress
&& sbp
->sb_inprogress
) {
345 xfs_warn(mp
, "Offline file system operation in progress!");
346 return XFS_ERROR(EFSCORRUPTED
);
350 * Version 1 directory format has never worked on Linux.
352 if (unlikely(!xfs_sb_version_hasdirv2(sbp
))) {
353 xfs_warn(mp
, "file system using version 1 directory format");
354 return XFS_ERROR(ENOSYS
);
361 xfs_sb_quota_from_disk(struct xfs_sb
*sbp
)
364 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
365 * leads to in-core values having two different values for a quota
366 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
369 * Note that this change affect only the in-core values. These
370 * values are not written back to disk unless any quota information
371 * is written to the disk. Even in that case, sb_pquotino field is
372 * not written to disk unless the superblock supports pquotino.
374 if (sbp
->sb_uquotino
== 0)
375 sbp
->sb_uquotino
= NULLFSINO
;
376 if (sbp
->sb_gquotino
== 0)
377 sbp
->sb_gquotino
= NULLFSINO
;
378 if (sbp
->sb_pquotino
== 0)
379 sbp
->sb_pquotino
= NULLFSINO
;
382 * We need to do these manipilations only if we are working
383 * with an older version of on-disk superblock.
385 if (xfs_sb_version_has_pquotino(sbp
))
388 if (sbp
->sb_qflags
& XFS_OQUOTA_ENFD
)
389 sbp
->sb_qflags
|= (sbp
->sb_qflags
& XFS_PQUOTA_ACCT
) ?
390 XFS_PQUOTA_ENFD
: XFS_GQUOTA_ENFD
;
391 if (sbp
->sb_qflags
& XFS_OQUOTA_CHKD
)
392 sbp
->sb_qflags
|= (sbp
->sb_qflags
& XFS_PQUOTA_ACCT
) ?
393 XFS_PQUOTA_CHKD
: XFS_GQUOTA_CHKD
;
394 sbp
->sb_qflags
&= ~(XFS_OQUOTA_ENFD
| XFS_OQUOTA_CHKD
);
396 if (sbp
->sb_qflags
& XFS_PQUOTA_ACCT
) {
398 * In older version of superblock, on-disk superblock only
399 * has sb_gquotino, and in-core superblock has both sb_gquotino
400 * and sb_pquotino. But, only one of them is supported at any
401 * point of time. So, if PQUOTA is set in disk superblock,
402 * copy over sb_gquotino to sb_pquotino.
404 sbp
->sb_pquotino
= sbp
->sb_gquotino
;
405 sbp
->sb_gquotino
= NULLFSINO
;
414 to
->sb_magicnum
= be32_to_cpu(from
->sb_magicnum
);
415 to
->sb_blocksize
= be32_to_cpu(from
->sb_blocksize
);
416 to
->sb_dblocks
= be64_to_cpu(from
->sb_dblocks
);
417 to
->sb_rblocks
= be64_to_cpu(from
->sb_rblocks
);
418 to
->sb_rextents
= be64_to_cpu(from
->sb_rextents
);
419 memcpy(&to
->sb_uuid
, &from
->sb_uuid
, sizeof(to
->sb_uuid
));
420 to
->sb_logstart
= be64_to_cpu(from
->sb_logstart
);
421 to
->sb_rootino
= be64_to_cpu(from
->sb_rootino
);
422 to
->sb_rbmino
= be64_to_cpu(from
->sb_rbmino
);
423 to
->sb_rsumino
= be64_to_cpu(from
->sb_rsumino
);
424 to
->sb_rextsize
= be32_to_cpu(from
->sb_rextsize
);
425 to
->sb_agblocks
= be32_to_cpu(from
->sb_agblocks
);
426 to
->sb_agcount
= be32_to_cpu(from
->sb_agcount
);
427 to
->sb_rbmblocks
= be32_to_cpu(from
->sb_rbmblocks
);
428 to
->sb_logblocks
= be32_to_cpu(from
->sb_logblocks
);
429 to
->sb_versionnum
= be16_to_cpu(from
->sb_versionnum
);
430 to
->sb_sectsize
= be16_to_cpu(from
->sb_sectsize
);
431 to
->sb_inodesize
= be16_to_cpu(from
->sb_inodesize
);
432 to
->sb_inopblock
= be16_to_cpu(from
->sb_inopblock
);
433 memcpy(&to
->sb_fname
, &from
->sb_fname
, sizeof(to
->sb_fname
));
434 to
->sb_blocklog
= from
->sb_blocklog
;
435 to
->sb_sectlog
= from
->sb_sectlog
;
436 to
->sb_inodelog
= from
->sb_inodelog
;
437 to
->sb_inopblog
= from
->sb_inopblog
;
438 to
->sb_agblklog
= from
->sb_agblklog
;
439 to
->sb_rextslog
= from
->sb_rextslog
;
440 to
->sb_inprogress
= from
->sb_inprogress
;
441 to
->sb_imax_pct
= from
->sb_imax_pct
;
442 to
->sb_icount
= be64_to_cpu(from
->sb_icount
);
443 to
->sb_ifree
= be64_to_cpu(from
->sb_ifree
);
444 to
->sb_fdblocks
= be64_to_cpu(from
->sb_fdblocks
);
445 to
->sb_frextents
= be64_to_cpu(from
->sb_frextents
);
446 to
->sb_uquotino
= be64_to_cpu(from
->sb_uquotino
);
447 to
->sb_gquotino
= be64_to_cpu(from
->sb_gquotino
);
448 to
->sb_qflags
= be16_to_cpu(from
->sb_qflags
);
449 to
->sb_flags
= from
->sb_flags
;
450 to
->sb_shared_vn
= from
->sb_shared_vn
;
451 to
->sb_inoalignmt
= be32_to_cpu(from
->sb_inoalignmt
);
452 to
->sb_unit
= be32_to_cpu(from
->sb_unit
);
453 to
->sb_width
= be32_to_cpu(from
->sb_width
);
454 to
->sb_dirblklog
= from
->sb_dirblklog
;
455 to
->sb_logsectlog
= from
->sb_logsectlog
;
456 to
->sb_logsectsize
= be16_to_cpu(from
->sb_logsectsize
);
457 to
->sb_logsunit
= be32_to_cpu(from
->sb_logsunit
);
458 to
->sb_features2
= be32_to_cpu(from
->sb_features2
);
459 to
->sb_bad_features2
= be32_to_cpu(from
->sb_bad_features2
);
460 to
->sb_features_compat
= be32_to_cpu(from
->sb_features_compat
);
461 to
->sb_features_ro_compat
= be32_to_cpu(from
->sb_features_ro_compat
);
462 to
->sb_features_incompat
= be32_to_cpu(from
->sb_features_incompat
);
463 to
->sb_features_log_incompat
=
464 be32_to_cpu(from
->sb_features_log_incompat
);
466 to
->sb_pquotino
= be64_to_cpu(from
->sb_pquotino
);
467 to
->sb_lsn
= be64_to_cpu(from
->sb_lsn
);
471 xfs_sb_quota_to_disk(
476 __uint16_t qflags
= from
->sb_qflags
;
479 * We need to do these manipilations only if we are working
480 * with an older version of on-disk superblock.
482 if (xfs_sb_version_has_pquotino(from
))
485 if (*fields
& XFS_SB_QFLAGS
) {
487 * The in-core version of sb_qflags do not have
488 * XFS_OQUOTA_* flags, whereas the on-disk version
489 * does. So, convert incore XFS_{PG}QUOTA_* flags
490 * to on-disk XFS_OQUOTA_* flags.
492 qflags
&= ~(XFS_PQUOTA_ENFD
| XFS_PQUOTA_CHKD
|
493 XFS_GQUOTA_ENFD
| XFS_GQUOTA_CHKD
);
495 if (from
->sb_qflags
&
496 (XFS_PQUOTA_ENFD
| XFS_GQUOTA_ENFD
))
497 qflags
|= XFS_OQUOTA_ENFD
;
498 if (from
->sb_qflags
&
499 (XFS_PQUOTA_CHKD
| XFS_GQUOTA_CHKD
))
500 qflags
|= XFS_OQUOTA_CHKD
;
501 to
->sb_qflags
= cpu_to_be16(qflags
);
502 *fields
&= ~XFS_SB_QFLAGS
;
506 * GQUOTINO and PQUOTINO cannot be used together in versions
507 * of superblock that do not have pquotino. from->sb_flags
508 * tells us which quota is active and should be copied to
511 if ((*fields
& XFS_SB_GQUOTINO
) &&
512 (from
->sb_qflags
& XFS_GQUOTA_ACCT
))
513 to
->sb_gquotino
= cpu_to_be64(from
->sb_gquotino
);
514 else if ((*fields
& XFS_SB_PQUOTINO
) &&
515 (from
->sb_qflags
& XFS_PQUOTA_ACCT
))
516 to
->sb_gquotino
= cpu_to_be64(from
->sb_pquotino
);
518 *fields
&= ~(XFS_SB_PQUOTINO
| XFS_SB_GQUOTINO
);
522 * Copy in core superblock to ondisk one.
524 * The fields argument is mask of superblock fields to copy.
532 xfs_caddr_t to_ptr
= (xfs_caddr_t
)to
;
533 xfs_caddr_t from_ptr
= (xfs_caddr_t
)from
;
542 xfs_sb_quota_to_disk(to
, from
, &fields
);
544 f
= (xfs_sb_field_t
)xfs_lowbit64((__uint64_t
)fields
);
545 first
= xfs_sb_info
[f
].offset
;
546 size
= xfs_sb_info
[f
+ 1].offset
- first
;
548 ASSERT(xfs_sb_info
[f
].type
== 0 || xfs_sb_info
[f
].type
== 1);
550 if (size
== 1 || xfs_sb_info
[f
].type
== 1) {
551 memcpy(to_ptr
+ first
, from_ptr
+ first
, size
);
555 *(__be16
*)(to_ptr
+ first
) =
556 cpu_to_be16(*(__u16
*)(from_ptr
+ first
));
559 *(__be32
*)(to_ptr
+ first
) =
560 cpu_to_be32(*(__u32
*)(from_ptr
+ first
));
563 *(__be64
*)(to_ptr
+ first
) =
564 cpu_to_be64(*(__u64
*)(from_ptr
+ first
));
571 fields
&= ~(1LL << f
);
580 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
583 xfs_sb_from_disk(&sb
, XFS_BUF_TO_SBP(bp
));
586 * Only check the in progress field for the primary superblock as
587 * mkfs.xfs doesn't clear it from secondary superblocks.
589 return xfs_mount_validate_sb(mp
, &sb
, bp
->b_bn
== XFS_SB_DADDR
,
594 * If the superblock has the CRC feature bit set or the CRC field is non-null,
595 * check that the CRC is valid. We check the CRC field is non-null because a
596 * single bit error could clear the feature bit and unused parts of the
597 * superblock are supposed to be zero. Hence a non-null crc field indicates that
598 * we've potentially lost a feature bit and we should check it anyway.
600 * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
601 * last field in V4 secondary superblocks. So for secondary superblocks,
602 * we are more forgiving, and ignore CRC failures if the primary doesn't
603 * indicate that the fs version is V5.
609 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
610 struct xfs_dsb
*dsb
= XFS_BUF_TO_SBP(bp
);
614 * open code the version check to avoid needing to convert the entire
615 * superblock from disk order just to check the version number
617 if (dsb
->sb_magicnum
== cpu_to_be32(XFS_SB_MAGIC
) &&
618 (((be16_to_cpu(dsb
->sb_versionnum
) & XFS_SB_VERSION_NUMBITS
) ==
622 if (!xfs_verify_cksum(bp
->b_addr
, be16_to_cpu(dsb
->sb_sectsize
),
623 offsetof(struct xfs_sb
, sb_crc
))) {
624 /* Only fail bad secondaries on a known V5 filesystem */
625 if (bp
->b_bn
!= XFS_SB_DADDR
&&
626 xfs_sb_version_hascrc(&mp
->m_sb
)) {
627 error
= EFSCORRUPTED
;
632 error
= xfs_sb_verify(bp
, true);
636 if (error
!= EWRONGFS
)
637 XFS_CORRUPTION_ERROR(__func__
, XFS_ERRLEVEL_LOW
,
639 xfs_buf_ioerror(bp
, error
);
644 * We may be probed for a filesystem match, so we may not want to emit
645 * messages when the superblock buffer is not actually an XFS superblock.
646 * If we find an XFS superblock, then run a normal, noisy mount because we are
647 * really going to mount it and want to know about errors.
650 xfs_sb_quiet_read_verify(
653 struct xfs_dsb
*dsb
= XFS_BUF_TO_SBP(bp
);
656 if (dsb
->sb_magicnum
== cpu_to_be32(XFS_SB_MAGIC
)) {
657 /* XFS filesystem, verify noisily! */
658 xfs_sb_read_verify(bp
);
662 xfs_buf_ioerror(bp
, EWRONGFS
);
669 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
670 struct xfs_buf_log_item
*bip
= bp
->b_fspriv
;
673 error
= xfs_sb_verify(bp
, false);
675 XFS_CORRUPTION_ERROR(__func__
, XFS_ERRLEVEL_LOW
,
677 xfs_buf_ioerror(bp
, error
);
681 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
685 XFS_BUF_TO_SBP(bp
)->sb_lsn
= cpu_to_be64(bip
->bli_item
.li_lsn
);
687 xfs_update_cksum(bp
->b_addr
, BBTOB(bp
->b_length
),
688 offsetof(struct xfs_sb
, sb_crc
));
691 const struct xfs_buf_ops xfs_sb_buf_ops
= {
692 .verify_read
= xfs_sb_read_verify
,
693 .verify_write
= xfs_sb_write_verify
,
696 const struct xfs_buf_ops xfs_sb_quiet_buf_ops
= {
697 .verify_read
= xfs_sb_quiet_read_verify
,
698 .verify_write
= xfs_sb_write_verify
,
704 * Mount initialization code establishing various mount
705 * fields from the superblock associated with the given
710 struct xfs_mount
*mp
,
713 mp
->m_agfrotor
= mp
->m_agirotor
= 0;
714 spin_lock_init(&mp
->m_agirotor_lock
);
715 mp
->m_maxagi
= mp
->m_sb
.sb_agcount
;
716 mp
->m_blkbit_log
= sbp
->sb_blocklog
+ XFS_NBBYLOG
;
717 mp
->m_blkbb_log
= sbp
->sb_blocklog
- BBSHIFT
;
718 mp
->m_sectbb_log
= sbp
->sb_sectlog
- BBSHIFT
;
719 mp
->m_agno_log
= xfs_highbit32(sbp
->sb_agcount
- 1) + 1;
720 mp
->m_agino_log
= sbp
->sb_inopblog
+ sbp
->sb_agblklog
;
721 mp
->m_blockmask
= sbp
->sb_blocksize
- 1;
722 mp
->m_blockwsize
= sbp
->sb_blocksize
>> XFS_WORDLOG
;
723 mp
->m_blockwmask
= mp
->m_blockwsize
- 1;
725 mp
->m_alloc_mxr
[0] = xfs_allocbt_maxrecs(mp
, sbp
->sb_blocksize
, 1);
726 mp
->m_alloc_mxr
[1] = xfs_allocbt_maxrecs(mp
, sbp
->sb_blocksize
, 0);
727 mp
->m_alloc_mnr
[0] = mp
->m_alloc_mxr
[0] / 2;
728 mp
->m_alloc_mnr
[1] = mp
->m_alloc_mxr
[1] / 2;
730 mp
->m_inobt_mxr
[0] = xfs_inobt_maxrecs(mp
, sbp
->sb_blocksize
, 1);
731 mp
->m_inobt_mxr
[1] = xfs_inobt_maxrecs(mp
, sbp
->sb_blocksize
, 0);
732 mp
->m_inobt_mnr
[0] = mp
->m_inobt_mxr
[0] / 2;
733 mp
->m_inobt_mnr
[1] = mp
->m_inobt_mxr
[1] / 2;
735 mp
->m_bmap_dmxr
[0] = xfs_bmbt_maxrecs(mp
, sbp
->sb_blocksize
, 1);
736 mp
->m_bmap_dmxr
[1] = xfs_bmbt_maxrecs(mp
, sbp
->sb_blocksize
, 0);
737 mp
->m_bmap_dmnr
[0] = mp
->m_bmap_dmxr
[0] / 2;
738 mp
->m_bmap_dmnr
[1] = mp
->m_bmap_dmxr
[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
;
747 * xfs_initialize_perag_data
749 * Read in each per-ag structure so we can count up the number of
750 * allocated inodes, free inodes and used filesystem blocks as this
751 * information is no longer persistent in the superblock. Once we have
752 * this information, write it into the in-core superblock structure.
755 xfs_initialize_perag_data(
756 struct xfs_mount
*mp
,
757 xfs_agnumber_t agcount
)
759 xfs_agnumber_t index
;
761 xfs_sb_t
*sbp
= &mp
->m_sb
;
765 uint64_t bfreelst
= 0;
769 for (index
= 0; index
< agcount
; index
++) {
771 * read the agf, then the agi. This gets us
772 * all the information we need and populates the
773 * per-ag structures for us.
775 error
= xfs_alloc_pagf_init(mp
, NULL
, index
, 0);
779 error
= xfs_ialloc_pagi_init(mp
, NULL
, index
);
782 pag
= xfs_perag_get(mp
, index
);
783 ifree
+= pag
->pagi_freecount
;
784 ialloc
+= pag
->pagi_count
;
785 bfree
+= pag
->pagf_freeblks
;
786 bfreelst
+= pag
->pagf_flcount
;
787 btree
+= pag
->pagf_btreeblks
;
791 * Overwrite incore superblock counters with just-read data
793 spin_lock(&mp
->m_sb_lock
);
794 sbp
->sb_ifree
= ifree
;
795 sbp
->sb_icount
= ialloc
;
796 sbp
->sb_fdblocks
= bfree
+ bfreelst
+ btree
;
797 spin_unlock(&mp
->m_sb_lock
);
799 /* Fixup the per-cpu counters as well. */
800 xfs_icsb_reinit_counters(mp
);
806 * xfs_mod_sb() can be used to copy arbitrary changes to the
807 * in-core superblock into the superblock buffer to be logged.
808 * It does not provide the higher level of locking that is
809 * needed to protect the in-core superblock from concurrent
813 xfs_mod_sb(xfs_trans_t
*tp
, __int64_t fields
)
825 bp
= xfs_trans_getsb(tp
, mp
, 0);
826 first
= sizeof(xfs_sb_t
);
831 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp
), &mp
->m_sb
, fields
);
833 /* find modified range */
834 f
= (xfs_sb_field_t
)xfs_highbit64((__uint64_t
)fields
);
835 ASSERT((1LL << f
) & XFS_SB_MOD_BITS
);
836 last
= xfs_sb_info
[f
+ 1].offset
- 1;
838 f
= (xfs_sb_field_t
)xfs_lowbit64((__uint64_t
)fields
);
839 ASSERT((1LL << f
) & XFS_SB_MOD_BITS
);
840 first
= xfs_sb_info
[f
].offset
;
842 xfs_trans_buf_set_type(tp
, bp
, XFS_BLFT_SB_BUF
);
843 xfs_trans_log_buf(tp
, bp
, first
, last
);