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"
26 #include "xfs_mount.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_dinode.h"
44 #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
;
80 (xfs_sb_version_hasattr(&mp
->m_sb
) ?
81 XFS_FSOP_GEOM_FLAGS_ATTR
: 0) |
82 (xfs_sb_version_hasnlink(&mp
->m_sb
) ?
83 XFS_FSOP_GEOM_FLAGS_NLINK
: 0) |
84 (xfs_sb_version_hasquota(&mp
->m_sb
) ?
85 XFS_FSOP_GEOM_FLAGS_QUOTA
: 0) |
86 (xfs_sb_version_hasalign(&mp
->m_sb
) ?
87 XFS_FSOP_GEOM_FLAGS_IALIGN
: 0) |
88 (xfs_sb_version_hasdalign(&mp
->m_sb
) ?
89 XFS_FSOP_GEOM_FLAGS_DALIGN
: 0) |
90 (xfs_sb_version_hasshared(&mp
->m_sb
) ?
91 XFS_FSOP_GEOM_FLAGS_SHARED
: 0) |
92 (xfs_sb_version_hasextflgbit(&mp
->m_sb
) ?
93 XFS_FSOP_GEOM_FLAGS_EXTFLG
: 0) |
94 (xfs_sb_version_hasdirv2(&mp
->m_sb
) ?
95 XFS_FSOP_GEOM_FLAGS_DIRV2
: 0) |
96 (xfs_sb_version_hassector(&mp
->m_sb
) ?
97 XFS_FSOP_GEOM_FLAGS_SECTOR
: 0) |
98 (xfs_sb_version_hasasciici(&mp
->m_sb
) ?
99 XFS_FSOP_GEOM_FLAGS_DIRV2CI
: 0) |
100 (xfs_sb_version_haslazysbcount(&mp
->m_sb
) ?
101 XFS_FSOP_GEOM_FLAGS_LAZYSB
: 0) |
102 (xfs_sb_version_hasattr2(&mp
->m_sb
) ?
103 XFS_FSOP_GEOM_FLAGS_ATTR2
: 0) |
104 (xfs_sb_version_hasprojid32bit(&mp
->m_sb
) ?
105 XFS_FSOP_GEOM_FLAGS_PROJID32
: 0) |
106 (xfs_sb_version_hascrc(&mp
->m_sb
) ?
107 XFS_FSOP_GEOM_FLAGS_V5SB
: 0) |
108 (xfs_sb_version_hasftype(&mp
->m_sb
) ?
109 XFS_FSOP_GEOM_FLAGS_FTYPE
: 0) |
110 (xfs_sb_version_hasfinobt(&mp
->m_sb
) ?
111 XFS_FSOP_GEOM_FLAGS_FINOBT
: 0);
112 geo
->logsectsize
= xfs_sb_version_hassector(&mp
->m_sb
) ?
113 mp
->m_sb
.sb_logsectsize
: BBSIZE
;
114 geo
->rtsectsize
= mp
->m_sb
.sb_blocksize
;
115 geo
->dirblocksize
= mp
->m_dir_geo
->blksize
;
117 if (new_version
>= 4) {
119 (xfs_sb_version_haslogv2(&mp
->m_sb
) ?
120 XFS_FSOP_GEOM_FLAGS_LOGV2
: 0);
121 geo
->logsunit
= mp
->m_sb
.sb_logsunit
;
126 static struct xfs_buf
*
127 xfs_growfs_get_hdr_buf(
128 struct xfs_mount
*mp
,
132 const struct xfs_buf_ops
*ops
)
136 bp
= xfs_buf_get_uncached(mp
->m_ddev_targp
, numblks
, flags
);
140 xfs_buf_zero(bp
, 0, BBTOB(bp
->b_length
));
142 bp
->b_maps
[0].bm_bn
= blkno
;
149 xfs_growfs_data_private(
150 xfs_mount_t
*mp
, /* mount point for filesystem */
151 xfs_growfs_data_t
*in
) /* growfs data input struct */
154 struct xfs_agfl
*agfl
;
158 xfs_extlen_t tmpsize
;
159 xfs_alloc_rec_t
*arec
;
163 int error
, saved_error
= 0;
164 xfs_agnumber_t nagcount
;
165 xfs_agnumber_t nagimax
= 0;
166 xfs_rfsblock_t nb
, nb_mod
;
168 xfs_rfsblock_t nfree
;
169 xfs_agnumber_t oagcount
;
175 if (nb
< mp
->m_sb
.sb_dblocks
|| pct
< 0 || pct
> 100)
176 return XFS_ERROR(EINVAL
);
177 if ((error
= xfs_sb_validate_fsb_count(&mp
->m_sb
, nb
)))
179 dpct
= pct
- mp
->m_sb
.sb_imax_pct
;
180 bp
= xfs_buf_read_uncached(mp
->m_ddev_targp
,
181 XFS_FSB_TO_BB(mp
, nb
) - XFS_FSS_TO_BB(mp
, 1),
182 XFS_FSS_TO_BB(mp
, 1), 0, NULL
);
192 new = nb
; /* use new as a temporary here */
193 nb_mod
= do_div(new, mp
->m_sb
.sb_agblocks
);
194 nagcount
= new + (nb_mod
!= 0);
195 if (nb_mod
&& nb_mod
< XFS_MIN_AG_BLOCKS
) {
197 nb
= (xfs_rfsblock_t
)nagcount
* mp
->m_sb
.sb_agblocks
;
198 if (nb
< mp
->m_sb
.sb_dblocks
)
199 return XFS_ERROR(EINVAL
);
201 new = nb
- mp
->m_sb
.sb_dblocks
;
202 oagcount
= mp
->m_sb
.sb_agcount
;
204 /* allocate the new per-ag structures */
205 if (nagcount
> oagcount
) {
206 error
= xfs_initialize_perag(mp
, nagcount
, &nagimax
);
211 tp
= xfs_trans_alloc(mp
, XFS_TRANS_GROWFS
);
212 tp
->t_flags
|= XFS_TRANS_RESERVE
;
213 error
= xfs_trans_reserve(tp
, &M_RES(mp
)->tr_growdata
,
214 XFS_GROWFS_SPACE_RES(mp
), 0);
216 xfs_trans_cancel(tp
, 0);
221 * Write new AG headers to disk. Non-transactional, but written
222 * synchronously so they are completed prior to the growfs transaction
226 for (agno
= nagcount
- 1; agno
>= oagcount
; agno
--, new -= agsize
) {
230 * AG freespace header block
232 bp
= xfs_growfs_get_hdr_buf(mp
,
233 XFS_AG_DADDR(mp
, agno
, XFS_AGF_DADDR(mp
)),
234 XFS_FSS_TO_BB(mp
, 1), 0,
241 agf
= XFS_BUF_TO_AGF(bp
);
242 agf
->agf_magicnum
= cpu_to_be32(XFS_AGF_MAGIC
);
243 agf
->agf_versionnum
= cpu_to_be32(XFS_AGF_VERSION
);
244 agf
->agf_seqno
= cpu_to_be32(agno
);
245 if (agno
== nagcount
- 1)
248 (agno
* (xfs_rfsblock_t
)mp
->m_sb
.sb_agblocks
);
250 agsize
= mp
->m_sb
.sb_agblocks
;
251 agf
->agf_length
= cpu_to_be32(agsize
);
252 agf
->agf_roots
[XFS_BTNUM_BNOi
] = cpu_to_be32(XFS_BNO_BLOCK(mp
));
253 agf
->agf_roots
[XFS_BTNUM_CNTi
] = cpu_to_be32(XFS_CNT_BLOCK(mp
));
254 agf
->agf_levels
[XFS_BTNUM_BNOi
] = cpu_to_be32(1);
255 agf
->agf_levels
[XFS_BTNUM_CNTi
] = cpu_to_be32(1);
256 agf
->agf_flfirst
= 0;
257 agf
->agf_fllast
= cpu_to_be32(XFS_AGFL_SIZE(mp
) - 1);
258 agf
->agf_flcount
= 0;
259 tmpsize
= agsize
- XFS_PREALLOC_BLOCKS(mp
);
260 agf
->agf_freeblks
= cpu_to_be32(tmpsize
);
261 agf
->agf_longest
= cpu_to_be32(tmpsize
);
262 if (xfs_sb_version_hascrc(&mp
->m_sb
))
263 uuid_copy(&agf
->agf_uuid
, &mp
->m_sb
.sb_uuid
);
265 error
= xfs_bwrite(bp
);
271 * AG freelist header block
273 bp
= xfs_growfs_get_hdr_buf(mp
,
274 XFS_AG_DADDR(mp
, agno
, XFS_AGFL_DADDR(mp
)),
275 XFS_FSS_TO_BB(mp
, 1), 0,
282 agfl
= XFS_BUF_TO_AGFL(bp
);
283 if (xfs_sb_version_hascrc(&mp
->m_sb
)) {
284 agfl
->agfl_magicnum
= cpu_to_be32(XFS_AGFL_MAGIC
);
285 agfl
->agfl_seqno
= cpu_to_be32(agno
);
286 uuid_copy(&agfl
->agfl_uuid
, &mp
->m_sb
.sb_uuid
);
289 agfl_bno
= XFS_BUF_TO_AGFL_BNO(mp
, bp
);
290 for (bucket
= 0; bucket
< XFS_AGFL_SIZE(mp
); bucket
++)
291 agfl_bno
[bucket
] = cpu_to_be32(NULLAGBLOCK
);
293 error
= xfs_bwrite(bp
);
299 * AG inode header block
301 bp
= xfs_growfs_get_hdr_buf(mp
,
302 XFS_AG_DADDR(mp
, agno
, XFS_AGI_DADDR(mp
)),
303 XFS_FSS_TO_BB(mp
, 1), 0,
310 agi
= XFS_BUF_TO_AGI(bp
);
311 agi
->agi_magicnum
= cpu_to_be32(XFS_AGI_MAGIC
);
312 agi
->agi_versionnum
= cpu_to_be32(XFS_AGI_VERSION
);
313 agi
->agi_seqno
= cpu_to_be32(agno
);
314 agi
->agi_length
= cpu_to_be32(agsize
);
316 agi
->agi_root
= cpu_to_be32(XFS_IBT_BLOCK(mp
));
317 agi
->agi_level
= cpu_to_be32(1);
318 agi
->agi_freecount
= 0;
319 agi
->agi_newino
= cpu_to_be32(NULLAGINO
);
320 agi
->agi_dirino
= cpu_to_be32(NULLAGINO
);
321 if (xfs_sb_version_hascrc(&mp
->m_sb
))
322 uuid_copy(&agi
->agi_uuid
, &mp
->m_sb
.sb_uuid
);
323 if (xfs_sb_version_hasfinobt(&mp
->m_sb
)) {
324 agi
->agi_free_root
= cpu_to_be32(XFS_FIBT_BLOCK(mp
));
325 agi
->agi_free_level
= cpu_to_be32(1);
327 for (bucket
= 0; bucket
< XFS_AGI_UNLINKED_BUCKETS
; bucket
++)
328 agi
->agi_unlinked
[bucket
] = cpu_to_be32(NULLAGINO
);
330 error
= xfs_bwrite(bp
);
336 * BNO btree root block
338 bp
= xfs_growfs_get_hdr_buf(mp
,
339 XFS_AGB_TO_DADDR(mp
, agno
, XFS_BNO_BLOCK(mp
)),
340 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
341 &xfs_allocbt_buf_ops
);
348 if (xfs_sb_version_hascrc(&mp
->m_sb
))
349 xfs_btree_init_block(mp
, bp
, XFS_ABTB_CRC_MAGIC
, 0, 1,
350 agno
, XFS_BTREE_CRC_BLOCKS
);
352 xfs_btree_init_block(mp
, bp
, XFS_ABTB_MAGIC
, 0, 1,
355 arec
= XFS_ALLOC_REC_ADDR(mp
, XFS_BUF_TO_BLOCK(bp
), 1);
356 arec
->ar_startblock
= cpu_to_be32(XFS_PREALLOC_BLOCKS(mp
));
357 arec
->ar_blockcount
= cpu_to_be32(
358 agsize
- be32_to_cpu(arec
->ar_startblock
));
360 error
= xfs_bwrite(bp
);
366 * CNT btree root block
368 bp
= xfs_growfs_get_hdr_buf(mp
,
369 XFS_AGB_TO_DADDR(mp
, agno
, XFS_CNT_BLOCK(mp
)),
370 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
371 &xfs_allocbt_buf_ops
);
377 if (xfs_sb_version_hascrc(&mp
->m_sb
))
378 xfs_btree_init_block(mp
, bp
, XFS_ABTC_CRC_MAGIC
, 0, 1,
379 agno
, XFS_BTREE_CRC_BLOCKS
);
381 xfs_btree_init_block(mp
, bp
, XFS_ABTC_MAGIC
, 0, 1,
384 arec
= XFS_ALLOC_REC_ADDR(mp
, XFS_BUF_TO_BLOCK(bp
), 1);
385 arec
->ar_startblock
= cpu_to_be32(XFS_PREALLOC_BLOCKS(mp
));
386 arec
->ar_blockcount
= cpu_to_be32(
387 agsize
- be32_to_cpu(arec
->ar_startblock
));
388 nfree
+= be32_to_cpu(arec
->ar_blockcount
);
390 error
= xfs_bwrite(bp
);
396 * INO btree root block
398 bp
= xfs_growfs_get_hdr_buf(mp
,
399 XFS_AGB_TO_DADDR(mp
, agno
, XFS_IBT_BLOCK(mp
)),
400 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
407 if (xfs_sb_version_hascrc(&mp
->m_sb
))
408 xfs_btree_init_block(mp
, bp
, XFS_IBT_CRC_MAGIC
, 0, 0,
409 agno
, XFS_BTREE_CRC_BLOCKS
);
411 xfs_btree_init_block(mp
, bp
, XFS_IBT_MAGIC
, 0, 0,
414 error
= xfs_bwrite(bp
);
420 * FINO btree root block
422 if (xfs_sb_version_hasfinobt(&mp
->m_sb
)) {
423 bp
= xfs_growfs_get_hdr_buf(mp
,
424 XFS_AGB_TO_DADDR(mp
, agno
, XFS_FIBT_BLOCK(mp
)),
425 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
432 if (xfs_sb_version_hascrc(&mp
->m_sb
))
433 xfs_btree_init_block(mp
, bp
, XFS_FIBT_CRC_MAGIC
,
435 XFS_BTREE_CRC_BLOCKS
);
437 xfs_btree_init_block(mp
, bp
, XFS_FIBT_MAGIC
, 0,
440 error
= xfs_bwrite(bp
);
447 xfs_trans_agblocks_delta(tp
, nfree
);
449 * There are new blocks in the old last a.g.
453 * Change the agi length.
455 error
= xfs_ialloc_read_agi(mp
, tp
, agno
, &bp
);
460 agi
= XFS_BUF_TO_AGI(bp
);
461 be32_add_cpu(&agi
->agi_length
, new);
462 ASSERT(nagcount
== oagcount
||
463 be32_to_cpu(agi
->agi_length
) == mp
->m_sb
.sb_agblocks
);
464 xfs_ialloc_log_agi(tp
, bp
, XFS_AGI_LENGTH
);
468 error
= xfs_alloc_read_agf(mp
, tp
, agno
, 0, &bp
);
473 agf
= XFS_BUF_TO_AGF(bp
);
474 be32_add_cpu(&agf
->agf_length
, new);
475 ASSERT(be32_to_cpu(agf
->agf_length
) ==
476 be32_to_cpu(agi
->agi_length
));
478 xfs_alloc_log_agf(tp
, bp
, XFS_AGF_LENGTH
);
480 * Free the new space.
482 error
= xfs_free_extent(tp
, XFS_AGB_TO_FSB(mp
, agno
,
483 be32_to_cpu(agf
->agf_length
) - new), new);
490 * Update changed superblock fields transactionally. These are not
491 * seen by the rest of the world until the transaction commit applies
492 * them atomically to the superblock.
494 if (nagcount
> oagcount
)
495 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_AGCOUNT
, nagcount
- oagcount
);
496 if (nb
> mp
->m_sb
.sb_dblocks
)
497 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_DBLOCKS
,
498 nb
- mp
->m_sb
.sb_dblocks
);
500 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_FDBLOCKS
, nfree
);
502 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_IMAXPCT
, dpct
);
503 error
= xfs_trans_commit(tp
, 0);
507 /* New allocation groups fully initialized, so update mount struct */
509 mp
->m_maxagi
= nagimax
;
510 if (mp
->m_sb
.sb_imax_pct
) {
511 __uint64_t icount
= mp
->m_sb
.sb_dblocks
* mp
->m_sb
.sb_imax_pct
;
513 mp
->m_maxicount
= icount
<< mp
->m_sb
.sb_inopblog
;
516 xfs_set_low_space_thresholds(mp
);
518 /* update secondary superblocks. */
519 for (agno
= 1; agno
< nagcount
; agno
++) {
522 * new secondary superblocks need to be zeroed, not read from
523 * disk as the contents of the new area we are growing into is
524 * completely unknown.
526 if (agno
< oagcount
) {
527 error
= xfs_trans_read_buf(mp
, NULL
, mp
->m_ddev_targp
,
528 XFS_AGB_TO_DADDR(mp
, agno
, XFS_SB_BLOCK(mp
)),
529 XFS_FSS_TO_BB(mp
, 1), 0, &bp
,
532 bp
= xfs_trans_get_buf(NULL
, mp
->m_ddev_targp
,
533 XFS_AGB_TO_DADDR(mp
, agno
, XFS_SB_BLOCK(mp
)),
534 XFS_FSS_TO_BB(mp
, 1), 0);
536 bp
->b_ops
= &xfs_sb_buf_ops
;
537 xfs_buf_zero(bp
, 0, BBTOB(bp
->b_length
));
543 * If we get an error reading or writing alternate superblocks,
544 * continue. xfs_repair chooses the "best" superblock based
545 * on most matches; if we break early, we'll leave more
546 * superblocks un-updated than updated, and xfs_repair may
547 * pick them over the properly-updated primary.
551 "error %d reading secondary superblock for ag %d",
556 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp
), &mp
->m_sb
, XFS_SB_ALL_BITS
);
558 error
= xfs_bwrite(bp
);
562 "write error %d updating secondary superblock for ag %d",
568 return saved_error
? saved_error
: error
;
571 xfs_trans_cancel(tp
, XFS_TRANS_ABORT
);
576 xfs_growfs_log_private(
577 xfs_mount_t
*mp
, /* mount point for filesystem */
578 xfs_growfs_log_t
*in
) /* growfs log input struct */
583 if (nb
< XFS_MIN_LOG_BLOCKS
|| nb
< XFS_B_TO_FSB(mp
, XFS_MIN_LOG_BYTES
))
584 return XFS_ERROR(EINVAL
);
585 if (nb
== mp
->m_sb
.sb_logblocks
&&
586 in
->isint
== (mp
->m_sb
.sb_logstart
!= 0))
587 return XFS_ERROR(EINVAL
);
589 * Moving the log is hard, need new interfaces to sync
590 * the log first, hold off all activity while moving it.
591 * Can have shorter or longer log in the same space,
592 * or transform internal to external log or vice versa.
594 return XFS_ERROR(ENOSYS
);
598 * protected versions of growfs function acquire and release locks on the mount
599 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
607 xfs_growfs_data_t
*in
)
611 if (!capable(CAP_SYS_ADMIN
))
612 return XFS_ERROR(EPERM
);
613 if (!mutex_trylock(&mp
->m_growlock
))
614 return XFS_ERROR(EWOULDBLOCK
);
615 error
= xfs_growfs_data_private(mp
, in
);
616 mutex_unlock(&mp
->m_growlock
);
623 xfs_growfs_log_t
*in
)
627 if (!capable(CAP_SYS_ADMIN
))
628 return XFS_ERROR(EPERM
);
629 if (!mutex_trylock(&mp
->m_growlock
))
630 return XFS_ERROR(EWOULDBLOCK
);
631 error
= xfs_growfs_log_private(mp
, in
);
632 mutex_unlock(&mp
->m_growlock
);
637 * exported through ioctl XFS_IOC_FSCOUNTS
643 xfs_fsop_counts_t
*cnt
)
645 xfs_icsb_sync_counters(mp
, XFS_ICSB_LAZY_COUNT
);
646 spin_lock(&mp
->m_sb_lock
);
647 cnt
->freedata
= mp
->m_sb
.sb_fdblocks
- XFS_ALLOC_SET_ASIDE(mp
);
648 cnt
->freertx
= mp
->m_sb
.sb_frextents
;
649 cnt
->freeino
= mp
->m_sb
.sb_ifree
;
650 cnt
->allocino
= mp
->m_sb
.sb_icount
;
651 spin_unlock(&mp
->m_sb_lock
);
656 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
658 * xfs_reserve_blocks is called to set m_resblks
659 * in the in-core mount table. The number of unused reserved blocks
660 * is kept in m_resblks_avail.
662 * Reserve the requested number of blocks if available. Otherwise return
663 * as many as possible to satisfy the request. The actual number
664 * reserved are returned in outval
666 * A null inval pointer indicates that only the current reserved blocks
667 * available should be returned no settings are changed.
674 xfs_fsop_resblks_t
*outval
)
676 __int64_t lcounter
, delta
, fdblks_delta
;
679 /* If inval is null, report current values and return */
680 if (inval
== (__uint64_t
*)NULL
) {
683 outval
->resblks
= mp
->m_resblks
;
684 outval
->resblks_avail
= mp
->m_resblks_avail
;
691 * With per-cpu counters, this becomes an interesting
692 * problem. we needto work out if we are freeing or allocation
693 * blocks first, then we can do the modification as necessary.
695 * We do this under the m_sb_lock so that if we are near
696 * ENOSPC, we will hold out any changes while we work out
697 * what to do. This means that the amount of free space can
698 * change while we do this, so we need to retry if we end up
699 * trying to reserve more space than is available.
701 * We also use the xfs_mod_incore_sb() interface so that we
702 * don't have to care about whether per cpu counter are
703 * enabled, disabled or even compiled in....
706 spin_lock(&mp
->m_sb_lock
);
707 xfs_icsb_sync_counters_locked(mp
, 0);
710 * If our previous reservation was larger than the current value,
711 * then move any unused blocks back to the free pool.
714 if (mp
->m_resblks
> request
) {
715 lcounter
= mp
->m_resblks_avail
- request
;
716 if (lcounter
> 0) { /* release unused blocks */
717 fdblks_delta
= lcounter
;
718 mp
->m_resblks_avail
-= lcounter
;
720 mp
->m_resblks
= request
;
724 free
= mp
->m_sb
.sb_fdblocks
- XFS_ALLOC_SET_ASIDE(mp
);
726 goto out
; /* ENOSPC and fdblks_delta = 0 */
728 delta
= request
- mp
->m_resblks
;
729 lcounter
= free
- delta
;
731 /* We can't satisfy the request, just get what we can */
732 mp
->m_resblks
+= free
;
733 mp
->m_resblks_avail
+= free
;
734 fdblks_delta
= -free
;
736 fdblks_delta
= -delta
;
737 mp
->m_resblks
= request
;
738 mp
->m_resblks_avail
+= delta
;
743 outval
->resblks
= mp
->m_resblks
;
744 outval
->resblks_avail
= mp
->m_resblks_avail
;
746 spin_unlock(&mp
->m_sb_lock
);
750 * If we are putting blocks back here, m_resblks_avail is
751 * already at its max so this will put it in the free pool.
753 * If we need space, we'll either succeed in getting it
754 * from the free block count or we'll get an enospc. If
755 * we get a ENOSPC, it means things changed while we were
756 * calculating fdblks_delta and so we should try again to
757 * see if there is anything left to reserve.
759 * Don't set the reserved flag here - we don't want to reserve
760 * the extra reserve blocks from the reserve.....
763 error
= xfs_icsb_modify_counters(mp
, XFS_SBS_FDBLOCKS
,
772 * Dump a transaction into the log that contains no real change. This is needed
773 * to be able to make the log dirty or stamp the current tail LSN into the log
774 * during the covering operation.
776 * We cannot use an inode here for this - that will push dirty state back up
777 * into the VFS and then periodic inode flushing will prevent log covering from
778 * making progress. Hence we log a field in the superblock instead and use a
779 * synchronous transaction to ensure the superblock is immediately unpinned
780 * and can be written back.
789 tp
= _xfs_trans_alloc(mp
, XFS_TRANS_DUMMY1
, KM_SLEEP
);
790 error
= xfs_trans_reserve(tp
, &M_RES(mp
)->tr_sb
, 0, 0);
792 xfs_trans_cancel(tp
, 0);
796 /* log the UUID because it is an unchanging field */
797 xfs_mod_sb(tp
, XFS_SB_UUID
);
798 xfs_trans_set_sync(tp
);
799 return xfs_trans_commit(tp
, 0);
808 case XFS_FSOP_GOING_FLAGS_DEFAULT
: {
809 struct super_block
*sb
= freeze_bdev(mp
->m_super
->s_bdev
);
811 if (sb
&& !IS_ERR(sb
)) {
812 xfs_force_shutdown(mp
, SHUTDOWN_FORCE_UMOUNT
);
813 thaw_bdev(sb
->s_bdev
, sb
);
818 case XFS_FSOP_GOING_FLAGS_LOGFLUSH
:
819 xfs_force_shutdown(mp
, SHUTDOWN_FORCE_UMOUNT
);
821 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH
:
822 xfs_force_shutdown(mp
,
823 SHUTDOWN_FORCE_UMOUNT
| SHUTDOWN_LOG_IO_ERROR
);
826 return XFS_ERROR(EINVAL
);
833 * Force a shutdown of the filesystem instantly while keeping the filesystem
834 * consistent. We don't do an unmount here; just shutdown the shop, make sure
835 * that absolutely nothing persistent happens to this filesystem after this
839 xfs_do_force_shutdown(
847 logerror
= flags
& SHUTDOWN_LOG_IO_ERROR
;
849 if (!(flags
& SHUTDOWN_FORCE_UMOUNT
)) {
851 "%s(0x%x) called from line %d of file %s. Return address = 0x%p",
852 __func__
, flags
, lnnum
, fname
, __return_address
);
855 * No need to duplicate efforts.
857 if (XFS_FORCED_SHUTDOWN(mp
) && !logerror
)
861 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
862 * queue up anybody new on the log reservations, and wakes up
863 * everybody who's sleeping on log reservations to tell them
866 if (xfs_log_force_umount(mp
, logerror
))
869 if (flags
& SHUTDOWN_CORRUPT_INCORE
) {
870 xfs_alert_tag(mp
, XFS_PTAG_SHUTDOWN_CORRUPT
,
871 "Corruption of in-memory data detected. Shutting down filesystem");
872 if (XFS_ERRLEVEL_HIGH
<= xfs_error_level
)
874 } else if (!(flags
& SHUTDOWN_FORCE_UMOUNT
)) {
876 xfs_alert_tag(mp
, XFS_PTAG_SHUTDOWN_LOGERROR
,
877 "Log I/O Error Detected. Shutting down filesystem");
878 } else if (flags
& SHUTDOWN_DEVICE_REQ
) {
879 xfs_alert_tag(mp
, XFS_PTAG_SHUTDOWN_IOERROR
,
880 "All device paths lost. Shutting down filesystem");
881 } else if (!(flags
& SHUTDOWN_REMOTE_REQ
)) {
882 xfs_alert_tag(mp
, XFS_PTAG_SHUTDOWN_IOERROR
,
883 "I/O Error Detected. Shutting down filesystem");
886 if (!(flags
& SHUTDOWN_FORCE_UMOUNT
)) {
888 "Please umount the filesystem and rectify the problem(s)");