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_rmap_btree.h"
37 #include "xfs_ialloc.h"
38 #include "xfs_fsops.h"
39 #include "xfs_itable.h"
40 #include "xfs_trans_space.h"
41 #include "xfs_rtalloc.h"
42 #include "xfs_trace.h"
44 #include "xfs_filestream.h"
48 * File system operations
58 memset(geo
, 0, sizeof(*geo
));
60 geo
->blocksize
= mp
->m_sb
.sb_blocksize
;
61 geo
->rtextsize
= mp
->m_sb
.sb_rextsize
;
62 geo
->agblocks
= mp
->m_sb
.sb_agblocks
;
63 geo
->agcount
= mp
->m_sb
.sb_agcount
;
64 geo
->logblocks
= mp
->m_sb
.sb_logblocks
;
65 geo
->sectsize
= mp
->m_sb
.sb_sectsize
;
66 geo
->inodesize
= mp
->m_sb
.sb_inodesize
;
67 geo
->imaxpct
= mp
->m_sb
.sb_imax_pct
;
68 geo
->datablocks
= mp
->m_sb
.sb_dblocks
;
69 geo
->rtblocks
= mp
->m_sb
.sb_rblocks
;
70 geo
->rtextents
= mp
->m_sb
.sb_rextents
;
71 geo
->logstart
= mp
->m_sb
.sb_logstart
;
72 ASSERT(sizeof(geo
->uuid
)==sizeof(mp
->m_sb
.sb_uuid
));
73 memcpy(geo
->uuid
, &mp
->m_sb
.sb_uuid
, sizeof(mp
->m_sb
.sb_uuid
));
74 if (new_version
>= 2) {
75 geo
->sunit
= mp
->m_sb
.sb_unit
;
76 geo
->swidth
= mp
->m_sb
.sb_width
;
78 if (new_version
>= 3) {
79 geo
->version
= XFS_FSOP_GEOM_VERSION
;
80 geo
->flags
= XFS_FSOP_GEOM_FLAGS_NLINK
|
81 XFS_FSOP_GEOM_FLAGS_DIRV2
|
82 (xfs_sb_version_hasattr(&mp
->m_sb
) ?
83 XFS_FSOP_GEOM_FLAGS_ATTR
: 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_hasextflgbit(&mp
->m_sb
) ?
91 XFS_FSOP_GEOM_FLAGS_EXTFLG
: 0) |
92 (xfs_sb_version_hassector(&mp
->m_sb
) ?
93 XFS_FSOP_GEOM_FLAGS_SECTOR
: 0) |
94 (xfs_sb_version_hasasciici(&mp
->m_sb
) ?
95 XFS_FSOP_GEOM_FLAGS_DIRV2CI
: 0) |
96 (xfs_sb_version_haslazysbcount(&mp
->m_sb
) ?
97 XFS_FSOP_GEOM_FLAGS_LAZYSB
: 0) |
98 (xfs_sb_version_hasattr2(&mp
->m_sb
) ?
99 XFS_FSOP_GEOM_FLAGS_ATTR2
: 0) |
100 (xfs_sb_version_hasprojid32bit(&mp
->m_sb
) ?
101 XFS_FSOP_GEOM_FLAGS_PROJID32
: 0) |
102 (xfs_sb_version_hascrc(&mp
->m_sb
) ?
103 XFS_FSOP_GEOM_FLAGS_V5SB
: 0) |
104 (xfs_sb_version_hasftype(&mp
->m_sb
) ?
105 XFS_FSOP_GEOM_FLAGS_FTYPE
: 0) |
106 (xfs_sb_version_hasfinobt(&mp
->m_sb
) ?
107 XFS_FSOP_GEOM_FLAGS_FINOBT
: 0) |
108 (xfs_sb_version_hassparseinodes(&mp
->m_sb
) ?
109 XFS_FSOP_GEOM_FLAGS_SPINODES
: 0);
110 geo
->logsectsize
= xfs_sb_version_hassector(&mp
->m_sb
) ?
111 mp
->m_sb
.sb_logsectsize
: BBSIZE
;
112 geo
->rtsectsize
= mp
->m_sb
.sb_blocksize
;
113 geo
->dirblocksize
= mp
->m_dir_geo
->blksize
;
115 if (new_version
>= 4) {
117 (xfs_sb_version_haslogv2(&mp
->m_sb
) ?
118 XFS_FSOP_GEOM_FLAGS_LOGV2
: 0);
119 geo
->logsunit
= mp
->m_sb
.sb_logsunit
;
124 static struct xfs_buf
*
125 xfs_growfs_get_hdr_buf(
126 struct xfs_mount
*mp
,
130 const struct xfs_buf_ops
*ops
)
134 bp
= xfs_buf_get_uncached(mp
->m_ddev_targp
, numblks
, flags
);
138 xfs_buf_zero(bp
, 0, BBTOB(bp
->b_length
));
140 bp
->b_maps
[0].bm_bn
= blkno
;
147 xfs_growfs_data_private(
148 xfs_mount_t
*mp
, /* mount point for filesystem */
149 xfs_growfs_data_t
*in
) /* growfs data input struct */
152 struct xfs_agfl
*agfl
;
156 xfs_extlen_t tmpsize
;
157 xfs_alloc_rec_t
*arec
;
161 int error
, saved_error
= 0;
162 xfs_agnumber_t nagcount
;
163 xfs_agnumber_t nagimax
= 0;
164 xfs_rfsblock_t nb
, nb_mod
;
166 xfs_rfsblock_t nfree
;
167 xfs_agnumber_t oagcount
;
173 if (nb
< mp
->m_sb
.sb_dblocks
|| pct
< 0 || pct
> 100)
175 if ((error
= xfs_sb_validate_fsb_count(&mp
->m_sb
, nb
)))
177 dpct
= pct
- mp
->m_sb
.sb_imax_pct
;
178 error
= xfs_buf_read_uncached(mp
->m_ddev_targp
,
179 XFS_FSB_TO_BB(mp
, nb
) - XFS_FSS_TO_BB(mp
, 1),
180 XFS_FSS_TO_BB(mp
, 1), 0, &bp
, NULL
);
185 new = nb
; /* use new as a temporary here */
186 nb_mod
= do_div(new, mp
->m_sb
.sb_agblocks
);
187 nagcount
= new + (nb_mod
!= 0);
188 if (nb_mod
&& nb_mod
< XFS_MIN_AG_BLOCKS
) {
190 nb
= (xfs_rfsblock_t
)nagcount
* mp
->m_sb
.sb_agblocks
;
191 if (nb
< mp
->m_sb
.sb_dblocks
)
194 new = nb
- mp
->m_sb
.sb_dblocks
;
195 oagcount
= mp
->m_sb
.sb_agcount
;
197 /* allocate the new per-ag structures */
198 if (nagcount
> oagcount
) {
199 error
= xfs_initialize_perag(mp
, nagcount
, &nagimax
);
204 error
= xfs_trans_alloc(mp
, &M_RES(mp
)->tr_growdata
,
205 XFS_GROWFS_SPACE_RES(mp
), 0, XFS_TRANS_RESERVE
, &tp
);
210 * Write new AG headers to disk. Non-transactional, but written
211 * synchronously so they are completed prior to the growfs transaction
215 for (agno
= nagcount
- 1; agno
>= oagcount
; agno
--, new -= agsize
) {
219 * AG freespace header block
221 bp
= xfs_growfs_get_hdr_buf(mp
,
222 XFS_AG_DADDR(mp
, agno
, XFS_AGF_DADDR(mp
)),
223 XFS_FSS_TO_BB(mp
, 1), 0,
230 agf
= XFS_BUF_TO_AGF(bp
);
231 agf
->agf_magicnum
= cpu_to_be32(XFS_AGF_MAGIC
);
232 agf
->agf_versionnum
= cpu_to_be32(XFS_AGF_VERSION
);
233 agf
->agf_seqno
= cpu_to_be32(agno
);
234 if (agno
== nagcount
- 1)
237 (agno
* (xfs_rfsblock_t
)mp
->m_sb
.sb_agblocks
);
239 agsize
= mp
->m_sb
.sb_agblocks
;
240 agf
->agf_length
= cpu_to_be32(agsize
);
241 agf
->agf_roots
[XFS_BTNUM_BNOi
] = cpu_to_be32(XFS_BNO_BLOCK(mp
));
242 agf
->agf_roots
[XFS_BTNUM_CNTi
] = cpu_to_be32(XFS_CNT_BLOCK(mp
));
243 agf
->agf_levels
[XFS_BTNUM_BNOi
] = cpu_to_be32(1);
244 agf
->agf_levels
[XFS_BTNUM_CNTi
] = cpu_to_be32(1);
245 if (xfs_sb_version_hasrmapbt(&mp
->m_sb
)) {
246 agf
->agf_roots
[XFS_BTNUM_RMAPi
] =
247 cpu_to_be32(XFS_RMAP_BLOCK(mp
));
248 agf
->agf_levels
[XFS_BTNUM_RMAPi
] = cpu_to_be32(1);
251 agf
->agf_flfirst
= cpu_to_be32(1);
253 agf
->agf_flcount
= 0;
254 tmpsize
= agsize
- mp
->m_ag_prealloc_blocks
;
255 agf
->agf_freeblks
= cpu_to_be32(tmpsize
);
256 agf
->agf_longest
= cpu_to_be32(tmpsize
);
257 if (xfs_sb_version_hascrc(&mp
->m_sb
))
258 uuid_copy(&agf
->agf_uuid
, &mp
->m_sb
.sb_meta_uuid
);
260 error
= xfs_bwrite(bp
);
266 * AG freelist header block
268 bp
= xfs_growfs_get_hdr_buf(mp
,
269 XFS_AG_DADDR(mp
, agno
, XFS_AGFL_DADDR(mp
)),
270 XFS_FSS_TO_BB(mp
, 1), 0,
277 agfl
= XFS_BUF_TO_AGFL(bp
);
278 if (xfs_sb_version_hascrc(&mp
->m_sb
)) {
279 agfl
->agfl_magicnum
= cpu_to_be32(XFS_AGFL_MAGIC
);
280 agfl
->agfl_seqno
= cpu_to_be32(agno
);
281 uuid_copy(&agfl
->agfl_uuid
, &mp
->m_sb
.sb_meta_uuid
);
284 agfl_bno
= XFS_BUF_TO_AGFL_BNO(mp
, bp
);
285 for (bucket
= 0; bucket
< XFS_AGFL_SIZE(mp
); bucket
++)
286 agfl_bno
[bucket
] = cpu_to_be32(NULLAGBLOCK
);
288 error
= xfs_bwrite(bp
);
294 * AG inode header block
296 bp
= xfs_growfs_get_hdr_buf(mp
,
297 XFS_AG_DADDR(mp
, agno
, XFS_AGI_DADDR(mp
)),
298 XFS_FSS_TO_BB(mp
, 1), 0,
305 agi
= XFS_BUF_TO_AGI(bp
);
306 agi
->agi_magicnum
= cpu_to_be32(XFS_AGI_MAGIC
);
307 agi
->agi_versionnum
= cpu_to_be32(XFS_AGI_VERSION
);
308 agi
->agi_seqno
= cpu_to_be32(agno
);
309 agi
->agi_length
= cpu_to_be32(agsize
);
311 agi
->agi_root
= cpu_to_be32(XFS_IBT_BLOCK(mp
));
312 agi
->agi_level
= cpu_to_be32(1);
313 agi
->agi_freecount
= 0;
314 agi
->agi_newino
= cpu_to_be32(NULLAGINO
);
315 agi
->agi_dirino
= cpu_to_be32(NULLAGINO
);
316 if (xfs_sb_version_hascrc(&mp
->m_sb
))
317 uuid_copy(&agi
->agi_uuid
, &mp
->m_sb
.sb_meta_uuid
);
318 if (xfs_sb_version_hasfinobt(&mp
->m_sb
)) {
319 agi
->agi_free_root
= cpu_to_be32(XFS_FIBT_BLOCK(mp
));
320 agi
->agi_free_level
= cpu_to_be32(1);
322 for (bucket
= 0; bucket
< XFS_AGI_UNLINKED_BUCKETS
; bucket
++)
323 agi
->agi_unlinked
[bucket
] = cpu_to_be32(NULLAGINO
);
325 error
= xfs_bwrite(bp
);
331 * BNO btree root block
333 bp
= xfs_growfs_get_hdr_buf(mp
,
334 XFS_AGB_TO_DADDR(mp
, agno
, XFS_BNO_BLOCK(mp
)),
335 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
336 &xfs_allocbt_buf_ops
);
343 if (xfs_sb_version_hascrc(&mp
->m_sb
))
344 xfs_btree_init_block(mp
, bp
, XFS_ABTB_CRC_MAGIC
, 0, 1,
345 agno
, XFS_BTREE_CRC_BLOCKS
);
347 xfs_btree_init_block(mp
, bp
, XFS_ABTB_MAGIC
, 0, 1,
350 arec
= XFS_ALLOC_REC_ADDR(mp
, XFS_BUF_TO_BLOCK(bp
), 1);
351 arec
->ar_startblock
= cpu_to_be32(mp
->m_ag_prealloc_blocks
);
352 arec
->ar_blockcount
= cpu_to_be32(
353 agsize
- be32_to_cpu(arec
->ar_startblock
));
355 error
= xfs_bwrite(bp
);
361 * CNT btree root block
363 bp
= xfs_growfs_get_hdr_buf(mp
,
364 XFS_AGB_TO_DADDR(mp
, agno
, XFS_CNT_BLOCK(mp
)),
365 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
366 &xfs_allocbt_buf_ops
);
372 if (xfs_sb_version_hascrc(&mp
->m_sb
))
373 xfs_btree_init_block(mp
, bp
, XFS_ABTC_CRC_MAGIC
, 0, 1,
374 agno
, XFS_BTREE_CRC_BLOCKS
);
376 xfs_btree_init_block(mp
, bp
, XFS_ABTC_MAGIC
, 0, 1,
379 arec
= XFS_ALLOC_REC_ADDR(mp
, XFS_BUF_TO_BLOCK(bp
), 1);
380 arec
->ar_startblock
= cpu_to_be32(mp
->m_ag_prealloc_blocks
);
381 arec
->ar_blockcount
= cpu_to_be32(
382 agsize
- be32_to_cpu(arec
->ar_startblock
));
383 nfree
+= be32_to_cpu(arec
->ar_blockcount
);
385 error
= xfs_bwrite(bp
);
390 /* RMAP btree root block */
391 if (xfs_sb_version_hasrmapbt(&mp
->m_sb
)) {
392 struct xfs_rmap_rec
*rrec
;
393 struct xfs_btree_block
*block
;
395 bp
= xfs_growfs_get_hdr_buf(mp
,
396 XFS_AGB_TO_DADDR(mp
, agno
, XFS_RMAP_BLOCK(mp
)),
397 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
398 &xfs_rmapbt_buf_ops
);
404 xfs_btree_init_block(mp
, bp
, XFS_RMAP_CRC_MAGIC
, 0, 0,
405 agno
, XFS_BTREE_CRC_BLOCKS
);
406 block
= XFS_BUF_TO_BLOCK(bp
);
410 * mark the AG header regions as static metadata The BNO
411 * btree block is the first block after the headers, so
412 * it's location defines the size of region the static
415 * Note: unlike mkfs, we never have to account for log
416 * space when growing the data regions
418 rrec
= XFS_RMAP_REC_ADDR(block
, 1);
419 rrec
->rm_startblock
= 0;
420 rrec
->rm_blockcount
= cpu_to_be32(XFS_BNO_BLOCK(mp
));
421 rrec
->rm_owner
= cpu_to_be64(XFS_RMAP_OWN_FS
);
423 be16_add_cpu(&block
->bb_numrecs
, 1);
425 /* account freespace btree root blocks */
426 rrec
= XFS_RMAP_REC_ADDR(block
, 2);
427 rrec
->rm_startblock
= cpu_to_be32(XFS_BNO_BLOCK(mp
));
428 rrec
->rm_blockcount
= cpu_to_be32(2);
429 rrec
->rm_owner
= cpu_to_be64(XFS_RMAP_OWN_AG
);
431 be16_add_cpu(&block
->bb_numrecs
, 1);
433 /* account inode btree root blocks */
434 rrec
= XFS_RMAP_REC_ADDR(block
, 3);
435 rrec
->rm_startblock
= cpu_to_be32(XFS_IBT_BLOCK(mp
));
436 rrec
->rm_blockcount
= cpu_to_be32(XFS_RMAP_BLOCK(mp
) -
438 rrec
->rm_owner
= cpu_to_be64(XFS_RMAP_OWN_INOBT
);
440 be16_add_cpu(&block
->bb_numrecs
, 1);
442 /* account for rmap btree root */
443 rrec
= XFS_RMAP_REC_ADDR(block
, 4);
444 rrec
->rm_startblock
= cpu_to_be32(XFS_RMAP_BLOCK(mp
));
445 rrec
->rm_blockcount
= cpu_to_be32(1);
446 rrec
->rm_owner
= cpu_to_be64(XFS_RMAP_OWN_AG
);
448 be16_add_cpu(&block
->bb_numrecs
, 1);
450 error
= xfs_bwrite(bp
);
457 * INO btree root block
459 bp
= xfs_growfs_get_hdr_buf(mp
,
460 XFS_AGB_TO_DADDR(mp
, agno
, XFS_IBT_BLOCK(mp
)),
461 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
468 if (xfs_sb_version_hascrc(&mp
->m_sb
))
469 xfs_btree_init_block(mp
, bp
, XFS_IBT_CRC_MAGIC
, 0, 0,
470 agno
, XFS_BTREE_CRC_BLOCKS
);
472 xfs_btree_init_block(mp
, bp
, XFS_IBT_MAGIC
, 0, 0,
475 error
= xfs_bwrite(bp
);
481 * FINO btree root block
483 if (xfs_sb_version_hasfinobt(&mp
->m_sb
)) {
484 bp
= xfs_growfs_get_hdr_buf(mp
,
485 XFS_AGB_TO_DADDR(mp
, agno
, XFS_FIBT_BLOCK(mp
)),
486 BTOBB(mp
->m_sb
.sb_blocksize
), 0,
493 if (xfs_sb_version_hascrc(&mp
->m_sb
))
494 xfs_btree_init_block(mp
, bp
, XFS_FIBT_CRC_MAGIC
,
496 XFS_BTREE_CRC_BLOCKS
);
498 xfs_btree_init_block(mp
, bp
, XFS_FIBT_MAGIC
, 0,
501 error
= xfs_bwrite(bp
);
508 xfs_trans_agblocks_delta(tp
, nfree
);
510 * There are new blocks in the old last a.g.
513 struct xfs_owner_info oinfo
;
516 * Change the agi length.
518 error
= xfs_ialloc_read_agi(mp
, tp
, agno
, &bp
);
523 agi
= XFS_BUF_TO_AGI(bp
);
524 be32_add_cpu(&agi
->agi_length
, new);
525 ASSERT(nagcount
== oagcount
||
526 be32_to_cpu(agi
->agi_length
) == mp
->m_sb
.sb_agblocks
);
527 xfs_ialloc_log_agi(tp
, bp
, XFS_AGI_LENGTH
);
531 error
= xfs_alloc_read_agf(mp
, tp
, agno
, 0, &bp
);
536 agf
= XFS_BUF_TO_AGF(bp
);
537 be32_add_cpu(&agf
->agf_length
, new);
538 ASSERT(be32_to_cpu(agf
->agf_length
) ==
539 be32_to_cpu(agi
->agi_length
));
541 xfs_alloc_log_agf(tp
, bp
, XFS_AGF_LENGTH
);
544 * Free the new space.
546 * XFS_RMAP_OWN_NULL is used here to tell the rmap btree that
547 * this doesn't actually exist in the rmap btree.
549 xfs_rmap_ag_owner(&oinfo
, XFS_RMAP_OWN_NULL
);
550 error
= xfs_free_extent(tp
,
551 XFS_AGB_TO_FSB(mp
, agno
,
552 be32_to_cpu(agf
->agf_length
) - new),
559 * Update changed superblock fields transactionally. These are not
560 * seen by the rest of the world until the transaction commit applies
561 * them atomically to the superblock.
563 if (nagcount
> oagcount
)
564 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_AGCOUNT
, nagcount
- oagcount
);
565 if (nb
> mp
->m_sb
.sb_dblocks
)
566 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_DBLOCKS
,
567 nb
- mp
->m_sb
.sb_dblocks
);
569 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_FDBLOCKS
, nfree
);
571 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_IMAXPCT
, dpct
);
572 xfs_trans_set_sync(tp
);
573 error
= xfs_trans_commit(tp
);
577 /* New allocation groups fully initialized, so update mount struct */
579 mp
->m_maxagi
= nagimax
;
580 if (mp
->m_sb
.sb_imax_pct
) {
581 __uint64_t icount
= mp
->m_sb
.sb_dblocks
* mp
->m_sb
.sb_imax_pct
;
583 mp
->m_maxicount
= icount
<< mp
->m_sb
.sb_inopblog
;
586 xfs_set_low_space_thresholds(mp
);
587 mp
->m_alloc_set_aside
= xfs_alloc_set_aside(mp
);
589 /* update secondary superblocks. */
590 for (agno
= 1; agno
< nagcount
; agno
++) {
593 * new secondary superblocks need to be zeroed, not read from
594 * disk as the contents of the new area we are growing into is
595 * completely unknown.
597 if (agno
< oagcount
) {
598 error
= xfs_trans_read_buf(mp
, NULL
, mp
->m_ddev_targp
,
599 XFS_AGB_TO_DADDR(mp
, agno
, XFS_SB_BLOCK(mp
)),
600 XFS_FSS_TO_BB(mp
, 1), 0, &bp
,
603 bp
= xfs_trans_get_buf(NULL
, mp
->m_ddev_targp
,
604 XFS_AGB_TO_DADDR(mp
, agno
, XFS_SB_BLOCK(mp
)),
605 XFS_FSS_TO_BB(mp
, 1), 0);
607 bp
->b_ops
= &xfs_sb_buf_ops
;
608 xfs_buf_zero(bp
, 0, BBTOB(bp
->b_length
));
614 * If we get an error reading or writing alternate superblocks,
615 * continue. xfs_repair chooses the "best" superblock based
616 * on most matches; if we break early, we'll leave more
617 * superblocks un-updated than updated, and xfs_repair may
618 * pick them over the properly-updated primary.
622 "error %d reading secondary superblock for ag %d",
627 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp
), &mp
->m_sb
);
629 error
= xfs_bwrite(bp
);
633 "write error %d updating secondary superblock for ag %d",
639 return saved_error
? saved_error
: error
;
642 xfs_trans_cancel(tp
);
647 xfs_growfs_log_private(
648 xfs_mount_t
*mp
, /* mount point for filesystem */
649 xfs_growfs_log_t
*in
) /* growfs log input struct */
654 if (nb
< XFS_MIN_LOG_BLOCKS
|| nb
< XFS_B_TO_FSB(mp
, XFS_MIN_LOG_BYTES
))
656 if (nb
== mp
->m_sb
.sb_logblocks
&&
657 in
->isint
== (mp
->m_sb
.sb_logstart
!= 0))
660 * Moving the log is hard, need new interfaces to sync
661 * the log first, hold off all activity while moving it.
662 * Can have shorter or longer log in the same space,
663 * or transform internal to external log or vice versa.
669 * protected versions of growfs function acquire and release locks on the mount
670 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
678 xfs_growfs_data_t
*in
)
682 if (!capable(CAP_SYS_ADMIN
))
684 if (!mutex_trylock(&mp
->m_growlock
))
686 error
= xfs_growfs_data_private(mp
, in
);
688 * Increment the generation unconditionally, the error could be from
689 * updating the secondary superblocks, in which case the new size
693 mutex_unlock(&mp
->m_growlock
);
700 xfs_growfs_log_t
*in
)
704 if (!capable(CAP_SYS_ADMIN
))
706 if (!mutex_trylock(&mp
->m_growlock
))
708 error
= xfs_growfs_log_private(mp
, in
);
709 mutex_unlock(&mp
->m_growlock
);
714 * exported through ioctl XFS_IOC_FSCOUNTS
720 xfs_fsop_counts_t
*cnt
)
722 cnt
->allocino
= percpu_counter_read_positive(&mp
->m_icount
);
723 cnt
->freeino
= percpu_counter_read_positive(&mp
->m_ifree
);
724 cnt
->freedata
= percpu_counter_read_positive(&mp
->m_fdblocks
) -
725 mp
->m_alloc_set_aside
;
727 spin_lock(&mp
->m_sb_lock
);
728 cnt
->freertx
= mp
->m_sb
.sb_frextents
;
729 spin_unlock(&mp
->m_sb_lock
);
734 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
736 * xfs_reserve_blocks is called to set m_resblks
737 * in the in-core mount table. The number of unused reserved blocks
738 * is kept in m_resblks_avail.
740 * Reserve the requested number of blocks if available. Otherwise return
741 * as many as possible to satisfy the request. The actual number
742 * reserved are returned in outval
744 * A null inval pointer indicates that only the current reserved blocks
745 * available should be returned no settings are changed.
752 xfs_fsop_resblks_t
*outval
)
754 __int64_t lcounter
, delta
;
755 __int64_t fdblks_delta
= 0;
760 /* If inval is null, report current values and return */
761 if (inval
== (__uint64_t
*)NULL
) {
764 outval
->resblks
= mp
->m_resblks
;
765 outval
->resblks_avail
= mp
->m_resblks_avail
;
772 * With per-cpu counters, this becomes an interesting problem. we need
773 * to work out if we are freeing or allocation blocks first, then we can
774 * do the modification as necessary.
776 * We do this under the m_sb_lock so that if we are near ENOSPC, we will
777 * hold out any changes while we work out what to do. This means that
778 * the amount of free space can change while we do this, so we need to
779 * retry if we end up trying to reserve more space than is available.
781 spin_lock(&mp
->m_sb_lock
);
784 * If our previous reservation was larger than the current value,
785 * then move any unused blocks back to the free pool. Modify the resblks
786 * counters directly since we shouldn't have any problems unreserving
789 if (mp
->m_resblks
> request
) {
790 lcounter
= mp
->m_resblks_avail
- request
;
791 if (lcounter
> 0) { /* release unused blocks */
792 fdblks_delta
= lcounter
;
793 mp
->m_resblks_avail
-= lcounter
;
795 mp
->m_resblks
= request
;
797 spin_unlock(&mp
->m_sb_lock
);
798 error
= xfs_mod_fdblocks(mp
, fdblks_delta
, 0);
799 spin_lock(&mp
->m_sb_lock
);
806 * If the request is larger than the current reservation, reserve the
807 * blocks before we update the reserve counters. Sample m_fdblocks and
808 * perform a partial reservation if the request exceeds free space.
812 free
= percpu_counter_sum(&mp
->m_fdblocks
) -
813 mp
->m_alloc_set_aside
;
817 delta
= request
- mp
->m_resblks
;
818 lcounter
= free
- delta
;
820 /* We can't satisfy the request, just get what we can */
823 fdblks_delta
= delta
;
826 * We'll either succeed in getting space from the free block
827 * count or we'll get an ENOSPC. If we get a ENOSPC, it means
828 * things changed while we were calculating fdblks_delta and so
829 * we should try again to see if there is anything left to
832 * Don't set the reserved flag here - we don't want to reserve
833 * the extra reserve blocks from the reserve.....
835 spin_unlock(&mp
->m_sb_lock
);
836 error
= xfs_mod_fdblocks(mp
, -fdblks_delta
, 0);
837 spin_lock(&mp
->m_sb_lock
);
838 } while (error
== -ENOSPC
);
841 * Update the reserve counters if blocks have been successfully
844 if (!error
&& fdblks_delta
) {
845 mp
->m_resblks
+= fdblks_delta
;
846 mp
->m_resblks_avail
+= fdblks_delta
;
851 outval
->resblks
= mp
->m_resblks
;
852 outval
->resblks_avail
= mp
->m_resblks_avail
;
855 spin_unlock(&mp
->m_sb_lock
);
865 case XFS_FSOP_GOING_FLAGS_DEFAULT
: {
866 struct super_block
*sb
= freeze_bdev(mp
->m_super
->s_bdev
);
868 if (sb
&& !IS_ERR(sb
)) {
869 xfs_force_shutdown(mp
, SHUTDOWN_FORCE_UMOUNT
);
870 thaw_bdev(sb
->s_bdev
, sb
);
875 case XFS_FSOP_GOING_FLAGS_LOGFLUSH
:
876 xfs_force_shutdown(mp
, SHUTDOWN_FORCE_UMOUNT
);
878 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH
:
879 xfs_force_shutdown(mp
,
880 SHUTDOWN_FORCE_UMOUNT
| SHUTDOWN_LOG_IO_ERROR
);
890 * Force a shutdown of the filesystem instantly while keeping the filesystem
891 * consistent. We don't do an unmount here; just shutdown the shop, make sure
892 * that absolutely nothing persistent happens to this filesystem after this
896 xfs_do_force_shutdown(
904 logerror
= flags
& SHUTDOWN_LOG_IO_ERROR
;
906 if (!(flags
& SHUTDOWN_FORCE_UMOUNT
)) {
908 "%s(0x%x) called from line %d of file %s. Return address = 0x%p",
909 __func__
, flags
, lnnum
, fname
, __return_address
);
912 * No need to duplicate efforts.
914 if (XFS_FORCED_SHUTDOWN(mp
) && !logerror
)
918 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
919 * queue up anybody new on the log reservations, and wakes up
920 * everybody who's sleeping on log reservations to tell them
923 if (xfs_log_force_umount(mp
, logerror
))
926 if (flags
& SHUTDOWN_CORRUPT_INCORE
) {
927 xfs_alert_tag(mp
, XFS_PTAG_SHUTDOWN_CORRUPT
,
928 "Corruption of in-memory data detected. Shutting down filesystem");
929 if (XFS_ERRLEVEL_HIGH
<= xfs_error_level
)
931 } else if (!(flags
& SHUTDOWN_FORCE_UMOUNT
)) {
933 xfs_alert_tag(mp
, XFS_PTAG_SHUTDOWN_LOGERROR
,
934 "Log I/O Error Detected. Shutting down filesystem");
935 } else if (flags
& SHUTDOWN_DEVICE_REQ
) {
936 xfs_alert_tag(mp
, XFS_PTAG_SHUTDOWN_IOERROR
,
937 "All device paths lost. Shutting down filesystem");
938 } else if (!(flags
& SHUTDOWN_REMOTE_REQ
)) {
939 xfs_alert_tag(mp
, XFS_PTAG_SHUTDOWN_IOERROR
,
940 "I/O Error Detected. Shutting down filesystem");
943 if (!(flags
& SHUTDOWN_FORCE_UMOUNT
)) {
945 "Please umount the filesystem and rectify the problem(s)");