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_types.h"
24 #include "xfs_trans.h"
29 #include "xfs_dmapi.h"
30 #include "xfs_mount.h"
31 #include "xfs_bmap_btree.h"
32 #include "xfs_alloc_btree.h"
33 #include "xfs_ialloc_btree.h"
34 #include "xfs_dir_sf.h"
35 #include "xfs_dir2_sf.h"
36 #include "xfs_attr_sf.h"
37 #include "xfs_dinode.h"
38 #include "xfs_inode.h"
39 #include "xfs_inode_item.h"
40 #include "xfs_btree.h"
41 #include "xfs_error.h"
42 #include "xfs_alloc.h"
43 #include "xfs_ialloc.h"
44 #include "xfs_fsops.h"
45 #include "xfs_itable.h"
46 #include "xfs_trans_space.h"
47 #include "xfs_rtalloc.h"
51 * File system operations
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
;
81 (XFS_SB_VERSION_HASATTR(&mp
->m_sb
) ?
82 XFS_FSOP_GEOM_FLAGS_ATTR
: 0) |
83 (XFS_SB_VERSION_HASNLINK(&mp
->m_sb
) ?
84 XFS_FSOP_GEOM_FLAGS_NLINK
: 0) |
85 (XFS_SB_VERSION_HASQUOTA(&mp
->m_sb
) ?
86 XFS_FSOP_GEOM_FLAGS_QUOTA
: 0) |
87 (XFS_SB_VERSION_HASALIGN(&mp
->m_sb
) ?
88 XFS_FSOP_GEOM_FLAGS_IALIGN
: 0) |
89 (XFS_SB_VERSION_HASDALIGN(&mp
->m_sb
) ?
90 XFS_FSOP_GEOM_FLAGS_DALIGN
: 0) |
91 (XFS_SB_VERSION_HASSHARED(&mp
->m_sb
) ?
92 XFS_FSOP_GEOM_FLAGS_SHARED
: 0) |
93 (XFS_SB_VERSION_HASEXTFLGBIT(&mp
->m_sb
) ?
94 XFS_FSOP_GEOM_FLAGS_EXTFLG
: 0) |
95 (XFS_SB_VERSION_HASDIRV2(&mp
->m_sb
) ?
96 XFS_FSOP_GEOM_FLAGS_DIRV2
: 0) |
97 (XFS_SB_VERSION_HASSECTOR(&mp
->m_sb
) ?
98 XFS_FSOP_GEOM_FLAGS_SECTOR
: 0) |
99 (XFS_SB_VERSION_HASATTR2(&mp
->m_sb
) ?
100 XFS_FSOP_GEOM_FLAGS_ATTR2
: 0);
101 geo
->logsectsize
= XFS_SB_VERSION_HASSECTOR(&mp
->m_sb
) ?
102 mp
->m_sb
.sb_logsectsize
: BBSIZE
;
103 geo
->rtsectsize
= mp
->m_sb
.sb_blocksize
;
104 geo
->dirblocksize
= mp
->m_dirblksize
;
106 if (new_version
>= 4) {
108 (XFS_SB_VERSION_HASLOGV2(&mp
->m_sb
) ?
109 XFS_FSOP_GEOM_FLAGS_LOGV2
: 0);
110 geo
->logsunit
= mp
->m_sb
.sb_logsunit
;
116 xfs_growfs_data_private(
117 xfs_mount_t
*mp
, /* mount point for filesystem */
118 xfs_growfs_data_t
*in
) /* growfs data input struct */
124 xfs_extlen_t tmpsize
;
125 xfs_alloc_rec_t
*arec
;
126 xfs_btree_sblock_t
*block
;
131 xfs_agnumber_t nagcount
;
132 xfs_agnumber_t nagimax
= 0;
133 xfs_rfsblock_t nb
, nb_mod
;
135 xfs_rfsblock_t nfree
;
136 xfs_agnumber_t oagcount
;
143 if (nb
< mp
->m_sb
.sb_dblocks
|| pct
< 0 || pct
> 100)
144 return XFS_ERROR(EINVAL
);
145 dpct
= pct
- mp
->m_sb
.sb_imax_pct
;
146 error
= xfs_read_buf(mp
, mp
->m_ddev_targp
,
147 XFS_FSB_TO_BB(mp
, nb
) - XFS_FSS_TO_BB(mp
, 1),
148 XFS_FSS_TO_BB(mp
, 1), 0, &bp
);
154 new = nb
; /* use new as a temporary here */
155 nb_mod
= do_div(new, mp
->m_sb
.sb_agblocks
);
156 nagcount
= new + (nb_mod
!= 0);
157 if (nb_mod
&& nb_mod
< XFS_MIN_AG_BLOCKS
) {
159 nb
= nagcount
* mp
->m_sb
.sb_agblocks
;
160 if (nb
< mp
->m_sb
.sb_dblocks
)
161 return XFS_ERROR(EINVAL
);
163 new = nb
- mp
->m_sb
.sb_dblocks
;
164 oagcount
= mp
->m_sb
.sb_agcount
;
165 if (nagcount
> oagcount
) {
166 down_write(&mp
->m_peraglock
);
167 mp
->m_perag
= kmem_realloc(mp
->m_perag
,
168 sizeof(xfs_perag_t
) * nagcount
,
169 sizeof(xfs_perag_t
) * oagcount
,
171 memset(&mp
->m_perag
[oagcount
], 0,
172 (nagcount
- oagcount
) * sizeof(xfs_perag_t
));
173 mp
->m_flags
|= XFS_MOUNT_32BITINODES
;
174 nagimax
= xfs_initialize_perag(XFS_MTOVFS(mp
), mp
, nagcount
);
175 up_write(&mp
->m_peraglock
);
177 tp
= xfs_trans_alloc(mp
, XFS_TRANS_GROWFS
);
178 if ((error
= xfs_trans_reserve(tp
, XFS_GROWFS_SPACE_RES(mp
),
179 XFS_GROWDATA_LOG_RES(mp
), 0, 0, 0))) {
180 xfs_trans_cancel(tp
, 0);
185 for (agno
= nagcount
- 1; agno
>= oagcount
; agno
--, new -= agsize
) {
187 * AG freelist header block
189 bp
= xfs_buf_get(mp
->m_ddev_targp
,
190 XFS_AG_DADDR(mp
, agno
, XFS_AGF_DADDR(mp
)),
191 XFS_FSS_TO_BB(mp
, 1), 0);
192 agf
= XFS_BUF_TO_AGF(bp
);
193 memset(agf
, 0, mp
->m_sb
.sb_sectsize
);
194 agf
->agf_magicnum
= cpu_to_be32(XFS_AGF_MAGIC
);
195 agf
->agf_versionnum
= cpu_to_be32(XFS_AGF_VERSION
);
196 agf
->agf_seqno
= cpu_to_be32(agno
);
197 if (agno
== nagcount
- 1)
200 (agno
* (xfs_rfsblock_t
)mp
->m_sb
.sb_agblocks
);
202 agsize
= mp
->m_sb
.sb_agblocks
;
203 agf
->agf_length
= cpu_to_be32(agsize
);
204 agf
->agf_roots
[XFS_BTNUM_BNOi
] = cpu_to_be32(XFS_BNO_BLOCK(mp
));
205 agf
->agf_roots
[XFS_BTNUM_CNTi
] = cpu_to_be32(XFS_CNT_BLOCK(mp
));
206 agf
->agf_levels
[XFS_BTNUM_BNOi
] = cpu_to_be32(1);
207 agf
->agf_levels
[XFS_BTNUM_CNTi
] = cpu_to_be32(1);
208 agf
->agf_flfirst
= 0;
209 agf
->agf_fllast
= cpu_to_be32(XFS_AGFL_SIZE(mp
) - 1);
210 agf
->agf_flcount
= 0;
211 tmpsize
= agsize
- XFS_PREALLOC_BLOCKS(mp
);
212 agf
->agf_freeblks
= cpu_to_be32(tmpsize
);
213 agf
->agf_longest
= cpu_to_be32(tmpsize
);
214 error
= xfs_bwrite(mp
, bp
);
219 * AG inode header block
221 bp
= xfs_buf_get(mp
->m_ddev_targp
,
222 XFS_AG_DADDR(mp
, agno
, XFS_AGI_DADDR(mp
)),
223 XFS_FSS_TO_BB(mp
, 1), 0);
224 agi
= XFS_BUF_TO_AGI(bp
);
225 memset(agi
, 0, mp
->m_sb
.sb_sectsize
);
226 agi
->agi_magicnum
= cpu_to_be32(XFS_AGI_MAGIC
);
227 agi
->agi_versionnum
= cpu_to_be32(XFS_AGI_VERSION
);
228 agi
->agi_seqno
= cpu_to_be32(agno
);
229 agi
->agi_length
= cpu_to_be32(agsize
);
231 agi
->agi_root
= cpu_to_be32(XFS_IBT_BLOCK(mp
));
232 agi
->agi_level
= cpu_to_be32(1);
233 agi
->agi_freecount
= 0;
234 agi
->agi_newino
= cpu_to_be32(NULLAGINO
);
235 agi
->agi_dirino
= cpu_to_be32(NULLAGINO
);
236 for (bucket
= 0; bucket
< XFS_AGI_UNLINKED_BUCKETS
; bucket
++)
237 agi
->agi_unlinked
[bucket
] = cpu_to_be32(NULLAGINO
);
238 error
= xfs_bwrite(mp
, bp
);
243 * BNO btree root block
245 bp
= xfs_buf_get(mp
->m_ddev_targp
,
246 XFS_AGB_TO_DADDR(mp
, agno
, XFS_BNO_BLOCK(mp
)),
247 BTOBB(mp
->m_sb
.sb_blocksize
), 0);
248 block
= XFS_BUF_TO_SBLOCK(bp
);
249 memset(block
, 0, mp
->m_sb
.sb_blocksize
);
250 block
->bb_magic
= cpu_to_be32(XFS_ABTB_MAGIC
);
252 block
->bb_numrecs
= cpu_to_be16(1);
253 block
->bb_leftsib
= cpu_to_be32(NULLAGBLOCK
);
254 block
->bb_rightsib
= cpu_to_be32(NULLAGBLOCK
);
255 arec
= XFS_BTREE_REC_ADDR(mp
->m_sb
.sb_blocksize
, xfs_alloc
,
256 block
, 1, mp
->m_alloc_mxr
[0]);
257 arec
->ar_startblock
= cpu_to_be32(XFS_PREALLOC_BLOCKS(mp
));
258 arec
->ar_blockcount
= cpu_to_be32(
259 agsize
- be32_to_cpu(arec
->ar_startblock
));
260 error
= xfs_bwrite(mp
, bp
);
265 * CNT btree root block
267 bp
= xfs_buf_get(mp
->m_ddev_targp
,
268 XFS_AGB_TO_DADDR(mp
, agno
, XFS_CNT_BLOCK(mp
)),
269 BTOBB(mp
->m_sb
.sb_blocksize
), 0);
270 block
= XFS_BUF_TO_SBLOCK(bp
);
271 memset(block
, 0, mp
->m_sb
.sb_blocksize
);
272 block
->bb_magic
= cpu_to_be32(XFS_ABTC_MAGIC
);
274 block
->bb_numrecs
= cpu_to_be16(1);
275 block
->bb_leftsib
= cpu_to_be32(NULLAGBLOCK
);
276 block
->bb_rightsib
= cpu_to_be32(NULLAGBLOCK
);
277 arec
= XFS_BTREE_REC_ADDR(mp
->m_sb
.sb_blocksize
, xfs_alloc
,
278 block
, 1, mp
->m_alloc_mxr
[0]);
279 arec
->ar_startblock
= cpu_to_be32(XFS_PREALLOC_BLOCKS(mp
));
280 arec
->ar_blockcount
= cpu_to_be32(
281 agsize
- be32_to_cpu(arec
->ar_startblock
));
282 nfree
+= be32_to_cpu(arec
->ar_blockcount
);
283 error
= xfs_bwrite(mp
, bp
);
288 * INO btree root block
290 bp
= xfs_buf_get(mp
->m_ddev_targp
,
291 XFS_AGB_TO_DADDR(mp
, agno
, XFS_IBT_BLOCK(mp
)),
292 BTOBB(mp
->m_sb
.sb_blocksize
), 0);
293 block
= XFS_BUF_TO_SBLOCK(bp
);
294 memset(block
, 0, mp
->m_sb
.sb_blocksize
);
295 block
->bb_magic
= cpu_to_be32(XFS_IBT_MAGIC
);
297 block
->bb_numrecs
= 0;
298 block
->bb_leftsib
= cpu_to_be32(NULLAGBLOCK
);
299 block
->bb_rightsib
= cpu_to_be32(NULLAGBLOCK
);
300 error
= xfs_bwrite(mp
, bp
);
305 xfs_trans_agblocks_delta(tp
, nfree
);
307 * There are new blocks in the old last a.g.
311 * Change the agi length.
313 error
= xfs_ialloc_read_agi(mp
, tp
, agno
, &bp
);
318 agi
= XFS_BUF_TO_AGI(bp
);
319 be32_add(&agi
->agi_length
, new);
320 ASSERT(nagcount
== oagcount
||
321 be32_to_cpu(agi
->agi_length
) == mp
->m_sb
.sb_agblocks
);
322 xfs_ialloc_log_agi(tp
, bp
, XFS_AGI_LENGTH
);
326 error
= xfs_alloc_read_agf(mp
, tp
, agno
, 0, &bp
);
331 agf
= XFS_BUF_TO_AGF(bp
);
332 be32_add(&agf
->agf_length
, new);
333 ASSERT(be32_to_cpu(agf
->agf_length
) ==
334 be32_to_cpu(agi
->agi_length
));
336 * Free the new space.
338 error
= xfs_free_extent(tp
, XFS_AGB_TO_FSB(mp
, agno
,
339 be32_to_cpu(agf
->agf_length
) - new), new);
344 if (nagcount
> oagcount
)
345 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_AGCOUNT
, nagcount
- oagcount
);
346 if (nb
> mp
->m_sb
.sb_dblocks
)
347 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_DBLOCKS
,
348 nb
- mp
->m_sb
.sb_dblocks
);
350 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_FDBLOCKS
, nfree
);
352 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_IMAXPCT
, dpct
);
353 error
= xfs_trans_commit(tp
, 0, NULL
);
357 /* New allocation groups fully initialized, so update mount struct */
359 mp
->m_maxagi
= nagimax
;
360 if (mp
->m_sb
.sb_imax_pct
) {
361 __uint64_t icount
= mp
->m_sb
.sb_dblocks
* mp
->m_sb
.sb_imax_pct
;
363 mp
->m_maxicount
= icount
<< mp
->m_sb
.sb_inopblog
;
366 for (agno
= 1; agno
< nagcount
; agno
++) {
367 error
= xfs_read_buf(mp
, mp
->m_ddev_targp
,
368 XFS_AGB_TO_DADDR(mp
, agno
, XFS_SB_BLOCK(mp
)),
369 XFS_FSS_TO_BB(mp
, 1), 0, &bp
);
371 xfs_fs_cmn_err(CE_WARN
, mp
,
372 "error %d reading secondary superblock for ag %d",
376 sbp
= XFS_BUF_TO_SBP(bp
);
377 xfs_xlatesb(sbp
, &mp
->m_sb
, -1, XFS_SB_ALL_BITS
);
379 * If we get an error writing out the alternate superblocks,
380 * just issue a warning and continue. The real work is
381 * already done and committed.
383 if (!(error
= xfs_bwrite(mp
, bp
))) {
386 xfs_fs_cmn_err(CE_WARN
, mp
,
387 "write error %d updating secondary superblock for ag %d",
389 break; /* no point in continuing */
395 xfs_trans_cancel(tp
, XFS_TRANS_ABORT
);
400 xfs_growfs_log_private(
401 xfs_mount_t
*mp
, /* mount point for filesystem */
402 xfs_growfs_log_t
*in
) /* growfs log input struct */
407 if (nb
< XFS_MIN_LOG_BLOCKS
|| nb
< XFS_B_TO_FSB(mp
, XFS_MIN_LOG_BYTES
))
408 return XFS_ERROR(EINVAL
);
409 if (nb
== mp
->m_sb
.sb_logblocks
&&
410 in
->isint
== (mp
->m_sb
.sb_logstart
!= 0))
411 return XFS_ERROR(EINVAL
);
413 * Moving the log is hard, need new interfaces to sync
414 * the log first, hold off all activity while moving it.
415 * Can have shorter or longer log in the same space,
416 * or transform internal to external log or vice versa.
418 return XFS_ERROR(ENOSYS
);
422 * protected versions of growfs function acquire and release locks on the mount
423 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
431 xfs_growfs_data_t
*in
)
434 if (!cpsema(&mp
->m_growlock
))
435 return XFS_ERROR(EWOULDBLOCK
);
436 error
= xfs_growfs_data_private(mp
, in
);
437 vsema(&mp
->m_growlock
);
444 xfs_growfs_log_t
*in
)
447 if (!cpsema(&mp
->m_growlock
))
448 return XFS_ERROR(EWOULDBLOCK
);
449 error
= xfs_growfs_log_private(mp
, in
);
450 vsema(&mp
->m_growlock
);
455 * exported through ioctl XFS_IOC_FSCOUNTS
461 xfs_fsop_counts_t
*cnt
)
465 xfs_icsb_sync_counters_lazy(mp
);
467 cnt
->freedata
= mp
->m_sb
.sb_fdblocks
;
468 cnt
->freertx
= mp
->m_sb
.sb_frextents
;
469 cnt
->freeino
= mp
->m_sb
.sb_ifree
;
470 cnt
->allocino
= mp
->m_sb
.sb_icount
;
471 XFS_SB_UNLOCK(mp
, s
);
476 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
478 * xfs_reserve_blocks is called to set m_resblks
479 * in the in-core mount table. The number of unused reserved blocks
480 * is kept in m_resblks_avail.
482 * Reserve the requested number of blocks if available. Otherwise return
483 * as many as possible to satisfy the request. The actual number
484 * reserved are returned in outval
486 * A null inval pointer indicates that only the current reserved blocks
487 * available should be returned no settings are changed.
494 xfs_fsop_resblks_t
*outval
)
496 __int64_t lcounter
, delta
;
500 /* If inval is null, report current values and return */
502 if (inval
== (__uint64_t
*)NULL
) {
503 outval
->resblks
= mp
->m_resblks
;
504 outval
->resblks_avail
= mp
->m_resblks_avail
;
512 * If our previous reservation was larger than the current value,
513 * then move any unused blocks back to the free pool.
516 if (mp
->m_resblks
> request
) {
517 lcounter
= mp
->m_resblks_avail
- request
;
518 if (lcounter
> 0) { /* release unused blocks */
519 mp
->m_sb
.sb_fdblocks
+= lcounter
;
520 mp
->m_resblks_avail
-= lcounter
;
522 mp
->m_resblks
= request
;
524 delta
= request
- mp
->m_resblks
;
525 lcounter
= mp
->m_sb
.sb_fdblocks
- delta
;
527 /* We can't satisfy the request, just get what we can */
528 mp
->m_resblks
+= mp
->m_sb
.sb_fdblocks
;
529 mp
->m_resblks_avail
+= mp
->m_sb
.sb_fdblocks
;
530 mp
->m_sb
.sb_fdblocks
= 0;
532 mp
->m_sb
.sb_fdblocks
= lcounter
;
533 mp
->m_resblks
= request
;
534 mp
->m_resblks_avail
+= delta
;
538 outval
->resblks
= mp
->m_resblks
;
539 outval
->resblks_avail
= mp
->m_resblks_avail
;
540 XFS_SB_UNLOCK(mp
, s
);
545 xfs_fs_log_dummy(xfs_mount_t
*mp
)
551 tp
= _xfs_trans_alloc(mp
, XFS_TRANS_DUMMY1
);
552 atomic_inc(&mp
->m_active_trans
);
553 if (xfs_trans_reserve(tp
, 0, XFS_ICHANGE_LOG_RES(mp
), 0, 0, 0)) {
554 xfs_trans_cancel(tp
, 0);
559 xfs_ilock(ip
, XFS_ILOCK_EXCL
);
561 xfs_trans_ijoin(tp
, ip
, XFS_ILOCK_EXCL
);
562 xfs_trans_ihold(tp
, ip
);
563 xfs_trans_log_inode(tp
, ip
, XFS_ILOG_CORE
);
564 xfs_trans_set_sync(tp
);
565 xfs_trans_commit(tp
, 0, NULL
);
567 xfs_iunlock(ip
, XFS_ILOCK_EXCL
);
576 case XFS_FSOP_GOING_FLAGS_DEFAULT
: {
577 struct vfs
*vfsp
= XFS_MTOVFS(mp
);
578 struct super_block
*sb
= freeze_bdev(vfsp
->vfs_super
->s_bdev
);
580 if (sb
&& !IS_ERR(sb
)) {
581 xfs_force_shutdown(mp
, XFS_FORCE_UMOUNT
);
582 thaw_bdev(sb
->s_bdev
, sb
);
587 case XFS_FSOP_GOING_FLAGS_LOGFLUSH
:
588 xfs_force_shutdown(mp
, XFS_FORCE_UMOUNT
);
590 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH
:
591 xfs_force_shutdown(mp
, XFS_FORCE_UMOUNT
|XFS_LOG_IO_ERROR
);
594 return XFS_ERROR(EINVAL
);