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_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
25 #include "xfs_mount.h"
26 #include "xfs_defer.h"
27 #include "xfs_da_format.h"
28 #include "xfs_da_btree.h"
29 #include "xfs_inode.h"
30 #include "xfs_trans.h"
31 #include "xfs_inode_item.h"
32 #include "xfs_error.h"
33 #include "xfs_btree.h"
34 #include "xfs_alloc_btree.h"
35 #include "xfs_alloc.h"
36 #include "xfs_ialloc.h"
37 #include "xfs_fsops.h"
38 #include "xfs_itable.h"
39 #include "xfs_trans_space.h"
40 #include "xfs_rtalloc.h"
41 #include "xfs_trace.h"
43 #include "xfs_filestream.h"
47 * File system operations
57 memset(geo
, 0, sizeof(*geo
));
59 geo
->blocksize
= mp
->m_sb
.sb_blocksize
;
60 geo
->rtextsize
= mp
->m_sb
.sb_rextsize
;
61 geo
->agblocks
= mp
->m_sb
.sb_agblocks
;
62 geo
->agcount
= mp
->m_sb
.sb_agcount
;
63 geo
->logblocks
= mp
->m_sb
.sb_logblocks
;
64 geo
->sectsize
= mp
->m_sb
.sb_sectsize
;
65 geo
->inodesize
= mp
->m_sb
.sb_inodesize
;
66 geo
->imaxpct
= mp
->m_sb
.sb_imax_pct
;
67 geo
->datablocks
= mp
->m_sb
.sb_dblocks
;
68 geo
->rtblocks
= mp
->m_sb
.sb_rblocks
;
69 geo
->rtextents
= mp
->m_sb
.sb_rextents
;
70 geo
->logstart
= mp
->m_sb
.sb_logstart
;
71 ASSERT(sizeof(geo
->uuid
)==sizeof(mp
->m_sb
.sb_uuid
));
72 memcpy(geo
->uuid
, &mp
->m_sb
.sb_uuid
, sizeof(mp
->m_sb
.sb_uuid
));
73 if (new_version
>= 2) {
74 geo
->sunit
= mp
->m_sb
.sb_unit
;
75 geo
->swidth
= mp
->m_sb
.sb_width
;
77 if (new_version
>= 3) {
78 geo
->version
= XFS_FSOP_GEOM_VERSION
;
79 geo
->flags
= XFS_FSOP_GEOM_FLAGS_NLINK
|
80 XFS_FSOP_GEOM_FLAGS_DIRV2
|
81 (xfs_sb_version_hasattr(&mp
->m_sb
) ?
82 XFS_FSOP_GEOM_FLAGS_ATTR
: 0) |
83 (xfs_sb_version_hasquota(&mp
->m_sb
) ?
84 XFS_FSOP_GEOM_FLAGS_QUOTA
: 0) |
85 (xfs_sb_version_hasalign(&mp
->m_sb
) ?
86 XFS_FSOP_GEOM_FLAGS_IALIGN
: 0) |
87 (xfs_sb_version_hasdalign(&mp
->m_sb
) ?
88 XFS_FSOP_GEOM_FLAGS_DALIGN
: 0) |
89 (xfs_sb_version_hasextflgbit(&mp
->m_sb
) ?
90 XFS_FSOP_GEOM_FLAGS_EXTFLG
: 0) |
91 (xfs_sb_version_hassector(&mp
->m_sb
) ?
92 XFS_FSOP_GEOM_FLAGS_SECTOR
: 0) |
93 (xfs_sb_version_hasasciici(&mp
->m_sb
) ?
94 XFS_FSOP_GEOM_FLAGS_DIRV2CI
: 0) |
95 (xfs_sb_version_haslazysbcount(&mp
->m_sb
) ?
96 XFS_FSOP_GEOM_FLAGS_LAZYSB
: 0) |
97 (xfs_sb_version_hasattr2(&mp
->m_sb
) ?
98 XFS_FSOP_GEOM_FLAGS_ATTR2
: 0) |
99 (xfs_sb_version_hasprojid32bit(&mp
->m_sb
) ?
100 XFS_FSOP_GEOM_FLAGS_PROJID32
: 0) |
101 (xfs_sb_version_hascrc(&mp
->m_sb
) ?
102 XFS_FSOP_GEOM_FLAGS_V5SB
: 0) |
103 (xfs_sb_version_hasftype(&mp
->m_sb
) ?
104 XFS_FSOP_GEOM_FLAGS_FTYPE
: 0) |
105 (xfs_sb_version_hasfinobt(&mp
->m_sb
) ?
106 XFS_FSOP_GEOM_FLAGS_FINOBT
: 0) |
107 (xfs_sb_version_hassparseinodes(&mp
->m_sb
) ?
108 XFS_FSOP_GEOM_FLAGS_SPINODES
: 0);
109 geo
->logsectsize
= xfs_sb_version_hassector(&mp
->m_sb
) ?
110 mp
->m_sb
.sb_logsectsize
: BBSIZE
;
111 geo
->rtsectsize
= mp
->m_sb
.sb_blocksize
;
112 geo
->dirblocksize
= mp
->m_dir_geo
->blksize
;
114 if (new_version
>= 4) {
116 (xfs_sb_version_haslogv2(&mp
->m_sb
) ?
117 XFS_FSOP_GEOM_FLAGS_LOGV2
: 0);
118 geo
->logsunit
= mp
->m_sb
.sb_logsunit
;
123 static struct xfs_buf
*
124 xfs_growfs_get_hdr_buf(
125 struct xfs_mount
*mp
,
129 const struct xfs_buf_ops
*ops
)
133 bp
= xfs_buf_get_uncached(mp
->m_ddev_targp
, numblks
, flags
);
137 xfs_buf_zero(bp
, 0, BBTOB(bp
->b_length
));
139 bp
->b_maps
[0].bm_bn
= blkno
;
146 xfs_growfs_data_private(
147 xfs_mount_t
*mp
, /* mount point for filesystem */
148 xfs_growfs_data_t
*in
) /* growfs data input struct */
151 struct xfs_agfl
*agfl
;
155 xfs_extlen_t tmpsize
;
156 xfs_alloc_rec_t
*arec
;
160 int error
, saved_error
= 0;
161 xfs_agnumber_t nagcount
;
162 xfs_agnumber_t nagimax
= 0;
163 xfs_rfsblock_t nb
, nb_mod
;
165 xfs_rfsblock_t nfree
;
166 xfs_agnumber_t oagcount
;
172 if (nb
< mp
->m_sb
.sb_dblocks
|| pct
< 0 || pct
> 100)
174 if ((error
= xfs_sb_validate_fsb_count(&mp
->m_sb
, nb
)))
176 dpct
= pct
- mp
->m_sb
.sb_imax_pct
;
177 error
= xfs_buf_read_uncached(mp
->m_ddev_targp
,
178 XFS_FSB_TO_BB(mp
, nb
) - XFS_FSS_TO_BB(mp
, 1),
179 XFS_FSS_TO_BB(mp
, 1), 0, &bp
, NULL
);
184 new = nb
; /* use new as a temporary here */
185 nb_mod
= do_div(new, mp
->m_sb
.sb_agblocks
);
186 nagcount
= new + (nb_mod
!= 0);
187 if (nb_mod
&& nb_mod
< XFS_MIN_AG_BLOCKS
) {
189 nb
= (xfs_rfsblock_t
)nagcount
* mp
->m_sb
.sb_agblocks
;
190 if (nb
< mp
->m_sb
.sb_dblocks
)
193 new = nb
- mp
->m_sb
.sb_dblocks
;
194 oagcount
= mp
->m_sb
.sb_agcount
;
196 /* allocate the new per-ag structures */
197 if (nagcount
> oagcount
) {
198 error
= xfs_initialize_perag(mp
, nagcount
, &nagimax
);
203 error
= xfs_trans_alloc(mp
, &M_RES(mp
)->tr_growdata
,
204 XFS_GROWFS_SPACE_RES(mp
), 0, XFS_TRANS_RESERVE
, &tp
);
209 * Write new AG headers to disk. Non-transactional, but written
210 * synchronously so they are completed prior to the growfs transaction
214 for (agno
= nagcount
- 1; agno
>= oagcount
; agno
--, new -= agsize
) {
218 * AG freespace header block
220 bp
= xfs_growfs_get_hdr_buf(mp
,
221 XFS_AG_DADDR(mp
, agno
, XFS_AGF_DADDR(mp
)),
222 XFS_FSS_TO_BB(mp
, 1), 0,
229 agf
= XFS_BUF_TO_AGF(bp
);
230 agf
->agf_magicnum
= cpu_to_be32(XFS_AGF_MAGIC
);
231 agf
->agf_versionnum
= cpu_to_be32(XFS_AGF_VERSION
);
232 agf
->agf_seqno
= cpu_to_be32(agno
);
233 if (agno
== nagcount
- 1)
236 (agno
* (xfs_rfsblock_t
)mp
->m_sb
.sb_agblocks
);
238 agsize
= mp
->m_sb
.sb_agblocks
;
239 agf
->agf_length
= cpu_to_be32(agsize
);
240 agf
->agf_roots
[XFS_BTNUM_BNOi
] = cpu_to_be32(XFS_BNO_BLOCK(mp
));
241 agf
->agf_roots
[XFS_BTNUM_CNTi
] = cpu_to_be32(XFS_CNT_BLOCK(mp
));
242 agf
->agf_levels
[XFS_BTNUM_BNOi
] = cpu_to_be32(1);
243 agf
->agf_levels
[XFS_BTNUM_CNTi
] = cpu_to_be32(1);
244 agf
->agf_flfirst
= cpu_to_be32(1);
246 agf
->agf_flcount
= 0;
247 tmpsize
= agsize
- mp
->m_ag_prealloc_blocks
;
248 agf
->agf_freeblks
= cpu_to_be32(tmpsize
);
249 agf
->agf_longest
= cpu_to_be32(tmpsize
);
250 if (xfs_sb_version_hascrc(&mp
->m_sb
))
251 uuid_copy(&agf
->agf_uuid
, &mp
->m_sb
.sb_meta_uuid
);
253 error
= xfs_bwrite(bp
);
259 * AG freelist header block
261 bp
= xfs_growfs_get_hdr_buf(mp
,
262 XFS_AG_DADDR(mp
, agno
, XFS_AGFL_DADDR(mp
)),
263 XFS_FSS_TO_BB(mp
, 1), 0,
270 agfl
= XFS_BUF_TO_AGFL(bp
);
271 if (xfs_sb_version_hascrc(&mp
->m_sb
)) {
272 agfl
->agfl_magicnum
= cpu_to_be32(XFS_AGFL_MAGIC
);
273 agfl
->agfl_seqno
= cpu_to_be32(agno
);
274 uuid_copy(&agfl
->agfl_uuid
, &mp
->m_sb
.sb_meta_uuid
);
277 agfl_bno
= XFS_BUF_TO_AGFL_BNO(mp
, bp
);
278 for (bucket
= 0; bucket
< XFS_AGFL_SIZE(mp
); bucket
++)
279 agfl_bno
[bucket
] = cpu_to_be32(NULLAGBLOCK
);
281 error
= xfs_bwrite(bp
);
287 * AG inode header block
289 bp
= xfs_growfs_get_hdr_buf(mp
,
290 XFS_AG_DADDR(mp
, agno
, XFS_AGI_DADDR(mp
)),
291 XFS_FSS_TO_BB(mp
, 1), 0,
298 agi
= XFS_BUF_TO_AGI(bp
);
299 agi
->agi_magicnum
= cpu_to_be32(XFS_AGI_MAGIC
);
300 agi
->agi_versionnum
= cpu_to_be32(XFS_AGI_VERSION
);
301 agi
->agi_seqno
= cpu_to_be32(agno
);
302 agi
->agi_length
= cpu_to_be32(agsize
);
304 agi
->agi_root
= cpu_to_be32(XFS_IBT_BLOCK(mp
));
305 agi
->agi_level
= cpu_to_be32(1);
306 agi
->agi_freecount
= 0;
307 agi
->agi_newino
= cpu_to_be32(NULLAGINO
);
308 agi
->agi_dirino
= cpu_to_be32(NULLAGINO
);
309 if (xfs_sb_version_hascrc(&mp
->m_sb
))
310 uuid_copy(&agi
->agi_uuid
, &mp
->m_sb
.sb_meta_uuid
);
311 if (xfs_sb_version_hasfinobt(&mp
->m_sb
)) {
312 agi
->agi_free_root
= cpu_to_be32(XFS_FIBT_BLOCK(mp
));
313 agi
->agi_free_level
= cpu_to_be32(1);
315 for (bucket
= 0; bucket
< XFS_AGI_UNLINKED_BUCKETS
; bucket
++)
316 agi
->agi_unlinked
[bucket
] = cpu_to_be32(NULLAGINO
);
318 error
= xfs_bwrite(bp
);
324 * BNO btree root block
326 bp
= xfs_growfs_get_hdr_buf(mp
,
327 XFS_AGB_TO_DADDR(mp
, agno
, XFS_BNO_BLOCK(mp
)),
328 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
329 &xfs_allocbt_buf_ops
);
336 if (xfs_sb_version_hascrc(&mp
->m_sb
))
337 xfs_btree_init_block(mp
, bp
, XFS_ABTB_CRC_MAGIC
, 0, 1,
338 agno
, XFS_BTREE_CRC_BLOCKS
);
340 xfs_btree_init_block(mp
, bp
, XFS_ABTB_MAGIC
, 0, 1,
343 arec
= XFS_ALLOC_REC_ADDR(mp
, XFS_BUF_TO_BLOCK(bp
), 1);
344 arec
->ar_startblock
= cpu_to_be32(mp
->m_ag_prealloc_blocks
);
345 arec
->ar_blockcount
= cpu_to_be32(
346 agsize
- be32_to_cpu(arec
->ar_startblock
));
348 error
= xfs_bwrite(bp
);
354 * CNT btree root block
356 bp
= xfs_growfs_get_hdr_buf(mp
,
357 XFS_AGB_TO_DADDR(mp
, agno
, XFS_CNT_BLOCK(mp
)),
358 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
359 &xfs_allocbt_buf_ops
);
365 if (xfs_sb_version_hascrc(&mp
->m_sb
))
366 xfs_btree_init_block(mp
, bp
, XFS_ABTC_CRC_MAGIC
, 0, 1,
367 agno
, XFS_BTREE_CRC_BLOCKS
);
369 xfs_btree_init_block(mp
, bp
, XFS_ABTC_MAGIC
, 0, 1,
372 arec
= XFS_ALLOC_REC_ADDR(mp
, XFS_BUF_TO_BLOCK(bp
), 1);
373 arec
->ar_startblock
= cpu_to_be32(mp
->m_ag_prealloc_blocks
);
374 arec
->ar_blockcount
= cpu_to_be32(
375 agsize
- be32_to_cpu(arec
->ar_startblock
));
376 nfree
+= be32_to_cpu(arec
->ar_blockcount
);
378 error
= xfs_bwrite(bp
);
384 * INO btree root block
386 bp
= xfs_growfs_get_hdr_buf(mp
,
387 XFS_AGB_TO_DADDR(mp
, agno
, XFS_IBT_BLOCK(mp
)),
388 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
395 if (xfs_sb_version_hascrc(&mp
->m_sb
))
396 xfs_btree_init_block(mp
, bp
, XFS_IBT_CRC_MAGIC
, 0, 0,
397 agno
, XFS_BTREE_CRC_BLOCKS
);
399 xfs_btree_init_block(mp
, bp
, XFS_IBT_MAGIC
, 0, 0,
402 error
= xfs_bwrite(bp
);
408 * FINO btree root block
410 if (xfs_sb_version_hasfinobt(&mp
->m_sb
)) {
411 bp
= xfs_growfs_get_hdr_buf(mp
,
412 XFS_AGB_TO_DADDR(mp
, agno
, XFS_FIBT_BLOCK(mp
)),
413 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
420 if (xfs_sb_version_hascrc(&mp
->m_sb
))
421 xfs_btree_init_block(mp
, bp
, XFS_FIBT_CRC_MAGIC
,
423 XFS_BTREE_CRC_BLOCKS
);
425 xfs_btree_init_block(mp
, bp
, XFS_FIBT_MAGIC
, 0,
428 error
= xfs_bwrite(bp
);
435 xfs_trans_agblocks_delta(tp
, nfree
);
437 * There are new blocks in the old last a.g.
440 struct xfs_owner_info oinfo
;
443 * Change the agi length.
445 error
= xfs_ialloc_read_agi(mp
, tp
, agno
, &bp
);
450 agi
= XFS_BUF_TO_AGI(bp
);
451 be32_add_cpu(&agi
->agi_length
, new);
452 ASSERT(nagcount
== oagcount
||
453 be32_to_cpu(agi
->agi_length
) == mp
->m_sb
.sb_agblocks
);
454 xfs_ialloc_log_agi(tp
, bp
, XFS_AGI_LENGTH
);
458 error
= xfs_alloc_read_agf(mp
, tp
, agno
, 0, &bp
);
463 agf
= XFS_BUF_TO_AGF(bp
);
464 be32_add_cpu(&agf
->agf_length
, new);
465 ASSERT(be32_to_cpu(agf
->agf_length
) ==
466 be32_to_cpu(agi
->agi_length
));
468 xfs_alloc_log_agf(tp
, bp
, XFS_AGF_LENGTH
);
471 * Free the new space.
473 * XFS_RMAP_OWN_NULL is used here to tell the rmap btree that
474 * this doesn't actually exist in the rmap btree.
476 xfs_rmap_ag_owner(&oinfo
, XFS_RMAP_OWN_NULL
);
477 error
= xfs_free_extent(tp
,
478 XFS_AGB_TO_FSB(mp
, agno
,
479 be32_to_cpu(agf
->agf_length
) - new),
486 * Update changed superblock fields transactionally. These are not
487 * seen by the rest of the world until the transaction commit applies
488 * them atomically to the superblock.
490 if (nagcount
> oagcount
)
491 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_AGCOUNT
, nagcount
- oagcount
);
492 if (nb
> mp
->m_sb
.sb_dblocks
)
493 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_DBLOCKS
,
494 nb
- mp
->m_sb
.sb_dblocks
);
496 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_FDBLOCKS
, nfree
);
498 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_IMAXPCT
, dpct
);
499 xfs_trans_set_sync(tp
);
500 error
= xfs_trans_commit(tp
);
504 /* New allocation groups fully initialized, so update mount struct */
506 mp
->m_maxagi
= nagimax
;
507 if (mp
->m_sb
.sb_imax_pct
) {
508 __uint64_t icount
= mp
->m_sb
.sb_dblocks
* mp
->m_sb
.sb_imax_pct
;
510 mp
->m_maxicount
= icount
<< mp
->m_sb
.sb_inopblog
;
513 xfs_set_low_space_thresholds(mp
);
515 /* update secondary superblocks. */
516 for (agno
= 1; agno
< nagcount
; agno
++) {
519 * new secondary superblocks need to be zeroed, not read from
520 * disk as the contents of the new area we are growing into is
521 * completely unknown.
523 if (agno
< oagcount
) {
524 error
= xfs_trans_read_buf(mp
, NULL
, mp
->m_ddev_targp
,
525 XFS_AGB_TO_DADDR(mp
, agno
, XFS_SB_BLOCK(mp
)),
526 XFS_FSS_TO_BB(mp
, 1), 0, &bp
,
529 bp
= xfs_trans_get_buf(NULL
, mp
->m_ddev_targp
,
530 XFS_AGB_TO_DADDR(mp
, agno
, XFS_SB_BLOCK(mp
)),
531 XFS_FSS_TO_BB(mp
, 1), 0);
533 bp
->b_ops
= &xfs_sb_buf_ops
;
534 xfs_buf_zero(bp
, 0, BBTOB(bp
->b_length
));
540 * If we get an error reading or writing alternate superblocks,
541 * continue. xfs_repair chooses the "best" superblock based
542 * on most matches; if we break early, we'll leave more
543 * superblocks un-updated than updated, and xfs_repair may
544 * pick them over the properly-updated primary.
548 "error %d reading secondary superblock for ag %d",
553 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp
), &mp
->m_sb
);
555 error
= xfs_bwrite(bp
);
559 "write error %d updating secondary superblock for ag %d",
565 return saved_error
? saved_error
: error
;
568 xfs_trans_cancel(tp
);
573 xfs_growfs_log_private(
574 xfs_mount_t
*mp
, /* mount point for filesystem */
575 xfs_growfs_log_t
*in
) /* growfs log input struct */
580 if (nb
< XFS_MIN_LOG_BLOCKS
|| nb
< XFS_B_TO_FSB(mp
, XFS_MIN_LOG_BYTES
))
582 if (nb
== mp
->m_sb
.sb_logblocks
&&
583 in
->isint
== (mp
->m_sb
.sb_logstart
!= 0))
586 * Moving the log is hard, need new interfaces to sync
587 * the log first, hold off all activity while moving it.
588 * Can have shorter or longer log in the same space,
589 * or transform internal to external log or vice versa.
595 * protected versions of growfs function acquire and release locks on the mount
596 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
604 xfs_growfs_data_t
*in
)
608 if (!capable(CAP_SYS_ADMIN
))
610 if (!mutex_trylock(&mp
->m_growlock
))
612 error
= xfs_growfs_data_private(mp
, in
);
614 * Increment the generation unconditionally, the error could be from
615 * updating the secondary superblocks, in which case the new size
619 mutex_unlock(&mp
->m_growlock
);
626 xfs_growfs_log_t
*in
)
630 if (!capable(CAP_SYS_ADMIN
))
632 if (!mutex_trylock(&mp
->m_growlock
))
634 error
= xfs_growfs_log_private(mp
, in
);
635 mutex_unlock(&mp
->m_growlock
);
640 * exported through ioctl XFS_IOC_FSCOUNTS
646 xfs_fsop_counts_t
*cnt
)
648 cnt
->allocino
= percpu_counter_read_positive(&mp
->m_icount
);
649 cnt
->freeino
= percpu_counter_read_positive(&mp
->m_ifree
);
650 cnt
->freedata
= percpu_counter_read_positive(&mp
->m_fdblocks
) -
651 XFS_ALLOC_SET_ASIDE(mp
);
653 spin_lock(&mp
->m_sb_lock
);
654 cnt
->freertx
= mp
->m_sb
.sb_frextents
;
655 spin_unlock(&mp
->m_sb_lock
);
660 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
662 * xfs_reserve_blocks is called to set m_resblks
663 * in the in-core mount table. The number of unused reserved blocks
664 * is kept in m_resblks_avail.
666 * Reserve the requested number of blocks if available. Otherwise return
667 * as many as possible to satisfy the request. The actual number
668 * reserved are returned in outval
670 * A null inval pointer indicates that only the current reserved blocks
671 * available should be returned no settings are changed.
678 xfs_fsop_resblks_t
*outval
)
680 __int64_t lcounter
, delta
;
681 __int64_t fdblks_delta
= 0;
686 /* If inval is null, report current values and return */
687 if (inval
== (__uint64_t
*)NULL
) {
690 outval
->resblks
= mp
->m_resblks
;
691 outval
->resblks_avail
= mp
->m_resblks_avail
;
698 * With per-cpu counters, this becomes an interesting problem. we need
699 * to work out if we are freeing or allocation blocks first, then we can
700 * do the modification as necessary.
702 * We do this under the m_sb_lock so that if we are near ENOSPC, we will
703 * hold out any changes while we work out what to do. This means that
704 * the amount of free space can change while we do this, so we need to
705 * retry if we end up trying to reserve more space than is available.
707 spin_lock(&mp
->m_sb_lock
);
710 * If our previous reservation was larger than the current value,
711 * then move any unused blocks back to the free pool. Modify the resblks
712 * counters directly since we shouldn't have any problems unreserving
715 if (mp
->m_resblks
> request
) {
716 lcounter
= mp
->m_resblks_avail
- request
;
717 if (lcounter
> 0) { /* release unused blocks */
718 fdblks_delta
= lcounter
;
719 mp
->m_resblks_avail
-= lcounter
;
721 mp
->m_resblks
= request
;
723 spin_unlock(&mp
->m_sb_lock
);
724 error
= xfs_mod_fdblocks(mp
, fdblks_delta
, 0);
725 spin_lock(&mp
->m_sb_lock
);
732 * If the request is larger than the current reservation, reserve the
733 * blocks before we update the reserve counters. Sample m_fdblocks and
734 * perform a partial reservation if the request exceeds free space.
738 free
= percpu_counter_sum(&mp
->m_fdblocks
) -
739 XFS_ALLOC_SET_ASIDE(mp
);
743 delta
= request
- mp
->m_resblks
;
744 lcounter
= free
- delta
;
746 /* We can't satisfy the request, just get what we can */
749 fdblks_delta
= delta
;
752 * We'll either succeed in getting space from the free block
753 * count or we'll get an ENOSPC. If we get a ENOSPC, it means
754 * things changed while we were calculating fdblks_delta and so
755 * we should try again to see if there is anything left to
758 * Don't set the reserved flag here - we don't want to reserve
759 * the extra reserve blocks from the reserve.....
761 spin_unlock(&mp
->m_sb_lock
);
762 error
= xfs_mod_fdblocks(mp
, -fdblks_delta
, 0);
763 spin_lock(&mp
->m_sb_lock
);
764 } while (error
== -ENOSPC
);
767 * Update the reserve counters if blocks have been successfully
770 if (!error
&& fdblks_delta
) {
771 mp
->m_resblks
+= fdblks_delta
;
772 mp
->m_resblks_avail
+= fdblks_delta
;
777 outval
->resblks
= mp
->m_resblks
;
778 outval
->resblks_avail
= mp
->m_resblks_avail
;
781 spin_unlock(&mp
->m_sb_lock
);
791 case XFS_FSOP_GOING_FLAGS_DEFAULT
: {
792 struct super_block
*sb
= freeze_bdev(mp
->m_super
->s_bdev
);
794 if (sb
&& !IS_ERR(sb
)) {
795 xfs_force_shutdown(mp
, SHUTDOWN_FORCE_UMOUNT
);
796 thaw_bdev(sb
->s_bdev
, sb
);
801 case XFS_FSOP_GOING_FLAGS_LOGFLUSH
:
802 xfs_force_shutdown(mp
, SHUTDOWN_FORCE_UMOUNT
);
804 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH
:
805 xfs_force_shutdown(mp
,
806 SHUTDOWN_FORCE_UMOUNT
| SHUTDOWN_LOG_IO_ERROR
);
816 * Force a shutdown of the filesystem instantly while keeping the filesystem
817 * consistent. We don't do an unmount here; just shutdown the shop, make sure
818 * that absolutely nothing persistent happens to this filesystem after this
822 xfs_do_force_shutdown(
830 logerror
= flags
& SHUTDOWN_LOG_IO_ERROR
;
832 if (!(flags
& SHUTDOWN_FORCE_UMOUNT
)) {
834 "%s(0x%x) called from line %d of file %s. Return address = 0x%p",
835 __func__
, flags
, lnnum
, fname
, __return_address
);
838 * No need to duplicate efforts.
840 if (XFS_FORCED_SHUTDOWN(mp
) && !logerror
)
844 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
845 * queue up anybody new on the log reservations, and wakes up
846 * everybody who's sleeping on log reservations to tell them
849 if (xfs_log_force_umount(mp
, logerror
))
852 if (flags
& SHUTDOWN_CORRUPT_INCORE
) {
853 xfs_alert_tag(mp
, XFS_PTAG_SHUTDOWN_CORRUPT
,
854 "Corruption of in-memory data detected. Shutting down filesystem");
855 if (XFS_ERRLEVEL_HIGH
<= xfs_error_level
)
857 } else if (!(flags
& SHUTDOWN_FORCE_UMOUNT
)) {
859 xfs_alert_tag(mp
, XFS_PTAG_SHUTDOWN_LOGERROR
,
860 "Log I/O Error Detected. Shutting down filesystem");
861 } else if (flags
& SHUTDOWN_DEVICE_REQ
) {
862 xfs_alert_tag(mp
, XFS_PTAG_SHUTDOWN_IOERROR
,
863 "All device paths lost. Shutting down filesystem");
864 } else if (!(flags
& SHUTDOWN_REMOTE_REQ
)) {
865 xfs_alert_tag(mp
, XFS_PTAG_SHUTDOWN_IOERROR
,
866 "I/O Error Detected. Shutting down filesystem");
869 if (!(flags
& SHUTDOWN_FORCE_UMOUNT
)) {
871 "Please umount the filesystem and rectify the problem(s)");