2 * Copyright (c) 2000-2002 Silicon Graphics, Inc. All Rights Reserved.
4 * This program is free software; you can redistribute it and/or modify it
5 * under the terms of version 2 of the GNU General Public License as
6 * published by the Free Software Foundation.
8 * This program is distributed in the hope that it would be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
12 * Further, this software is distributed without any warranty that it is
13 * free of the rightful claim of any third person regarding infringement
14 * or the like. Any license provided herein, whether implied or
15 * otherwise, applies only to this software file. Patent licenses, if
16 * any, provided herein do not apply to combinations of this program with
17 * other software, or any other product whatsoever.
19 * You should have received a copy of the GNU General Public License along
20 * with this program; if not, write the Free Software Foundation, Inc., 59
21 * Temple Place - Suite 330, Boston MA 02111-1307, USA.
23 * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
24 * Mountain View, CA 94043, or:
28 * For further information regarding this notice, see:
30 * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
34 #include "xfs_macros.h"
35 #include "xfs_types.h"
38 #include "xfs_trans.h"
41 #include "xfs_dmapi.h"
42 #include "xfs_mount.h"
44 #include "xfs_alloc_btree.h"
45 #include "xfs_bmap_btree.h"
46 #include "xfs_ialloc_btree.h"
47 #include "xfs_btree.h"
48 #include "xfs_error.h"
49 #include "xfs_alloc.h"
50 #include "xfs_ialloc.h"
51 #include "xfs_fsops.h"
52 #include "xfs_itable.h"
54 #include "xfs_refcache.h"
55 #include "xfs_trans_space.h"
56 #include "xfs_rtalloc.h"
58 #include "xfs_attr_sf.h"
59 #include "xfs_dir_sf.h"
60 #include "xfs_dir2_sf.h"
61 #include "xfs_dinode.h"
62 #include "xfs_inode.h"
63 #include "xfs_inode_item.h"
66 * File system operations
75 geo
->blocksize
= mp
->m_sb
.sb_blocksize
;
76 geo
->rtextsize
= mp
->m_sb
.sb_rextsize
;
77 geo
->agblocks
= mp
->m_sb
.sb_agblocks
;
78 geo
->agcount
= mp
->m_sb
.sb_agcount
;
79 geo
->logblocks
= mp
->m_sb
.sb_logblocks
;
80 geo
->sectsize
= mp
->m_sb
.sb_sectsize
;
81 geo
->inodesize
= mp
->m_sb
.sb_inodesize
;
82 geo
->imaxpct
= mp
->m_sb
.sb_imax_pct
;
83 geo
->datablocks
= mp
->m_sb
.sb_dblocks
;
84 geo
->rtblocks
= mp
->m_sb
.sb_rblocks
;
85 geo
->rtextents
= mp
->m_sb
.sb_rextents
;
86 geo
->logstart
= mp
->m_sb
.sb_logstart
;
87 ASSERT(sizeof(geo
->uuid
)==sizeof(mp
->m_sb
.sb_uuid
));
88 memcpy(geo
->uuid
, &mp
->m_sb
.sb_uuid
, sizeof(mp
->m_sb
.sb_uuid
));
89 if (new_version
>= 2) {
90 geo
->sunit
= mp
->m_sb
.sb_unit
;
91 geo
->swidth
= mp
->m_sb
.sb_width
;
93 if (new_version
>= 3) {
94 geo
->version
= XFS_FSOP_GEOM_VERSION
;
96 (XFS_SB_VERSION_HASATTR(&mp
->m_sb
) ?
97 XFS_FSOP_GEOM_FLAGS_ATTR
: 0) |
98 (XFS_SB_VERSION_HASNLINK(&mp
->m_sb
) ?
99 XFS_FSOP_GEOM_FLAGS_NLINK
: 0) |
100 (XFS_SB_VERSION_HASQUOTA(&mp
->m_sb
) ?
101 XFS_FSOP_GEOM_FLAGS_QUOTA
: 0) |
102 (XFS_SB_VERSION_HASALIGN(&mp
->m_sb
) ?
103 XFS_FSOP_GEOM_FLAGS_IALIGN
: 0) |
104 (XFS_SB_VERSION_HASDALIGN(&mp
->m_sb
) ?
105 XFS_FSOP_GEOM_FLAGS_DALIGN
: 0) |
106 (XFS_SB_VERSION_HASSHARED(&mp
->m_sb
) ?
107 XFS_FSOP_GEOM_FLAGS_SHARED
: 0) |
108 (XFS_SB_VERSION_HASEXTFLGBIT(&mp
->m_sb
) ?
109 XFS_FSOP_GEOM_FLAGS_EXTFLG
: 0) |
110 (XFS_SB_VERSION_HASDIRV2(&mp
->m_sb
) ?
111 XFS_FSOP_GEOM_FLAGS_DIRV2
: 0) |
112 (XFS_SB_VERSION_HASSECTOR(&mp
->m_sb
) ?
113 XFS_FSOP_GEOM_FLAGS_SECTOR
: 0);
114 geo
->logsectsize
= XFS_SB_VERSION_HASSECTOR(&mp
->m_sb
) ?
115 mp
->m_sb
.sb_logsectsize
: BBSIZE
;
116 geo
->rtsectsize
= mp
->m_sb
.sb_blocksize
;
117 geo
->dirblocksize
= mp
->m_dirblksize
;
119 if (new_version
>= 4) {
121 (XFS_SB_VERSION_HASLOGV2(&mp
->m_sb
) ?
122 XFS_FSOP_GEOM_FLAGS_LOGV2
: 0);
123 geo
->logsunit
= mp
->m_sb
.sb_logsunit
;
129 xfs_growfs_data_private(
130 xfs_mount_t
*mp
, /* mount point for filesystem */
131 xfs_growfs_data_t
*in
) /* growfs data input struct */
137 xfs_extlen_t tmpsize
;
138 xfs_alloc_rec_t
*arec
;
139 xfs_btree_sblock_t
*block
;
144 xfs_agnumber_t nagcount
;
145 xfs_agnumber_t nagimax
= 0;
146 xfs_rfsblock_t nb
, nb_mod
;
148 xfs_rfsblock_t nfree
;
149 xfs_agnumber_t oagcount
;
156 if (nb
< mp
->m_sb
.sb_dblocks
|| pct
< 0 || pct
> 100)
157 return XFS_ERROR(EINVAL
);
158 dpct
= pct
- mp
->m_sb
.sb_imax_pct
;
159 error
= xfs_read_buf(mp
, mp
->m_ddev_targp
,
160 XFS_FSB_TO_BB(mp
, nb
) - XFS_FSS_TO_BB(mp
, 1),
161 XFS_FSS_TO_BB(mp
, 1), 0, &bp
);
167 new = nb
; /* use new as a temporary here */
168 nb_mod
= do_div(new, mp
->m_sb
.sb_agblocks
);
169 nagcount
= new + (nb_mod
!= 0);
170 if (nb_mod
&& nb_mod
< XFS_MIN_AG_BLOCKS
) {
172 nb
= nagcount
* mp
->m_sb
.sb_agblocks
;
173 if (nb
< mp
->m_sb
.sb_dblocks
)
174 return XFS_ERROR(EINVAL
);
176 new = nb
- mp
->m_sb
.sb_dblocks
;
177 oagcount
= mp
->m_sb
.sb_agcount
;
178 if (nagcount
> oagcount
) {
179 down_write(&mp
->m_peraglock
);
180 mp
->m_perag
= kmem_realloc(mp
->m_perag
,
181 sizeof(xfs_perag_t
) * nagcount
,
182 sizeof(xfs_perag_t
) * oagcount
,
184 memset(&mp
->m_perag
[oagcount
], 0,
185 (nagcount
- oagcount
) * sizeof(xfs_perag_t
));
186 mp
->m_flags
|= XFS_MOUNT_32BITINODES
;
187 nagimax
= xfs_initialize_perag(mp
, nagcount
);
188 up_write(&mp
->m_peraglock
);
190 tp
= xfs_trans_alloc(mp
, XFS_TRANS_GROWFS
);
191 if ((error
= xfs_trans_reserve(tp
, XFS_GROWFS_SPACE_RES(mp
),
192 XFS_GROWDATA_LOG_RES(mp
), 0, 0, 0))) {
193 xfs_trans_cancel(tp
, 0);
198 for (agno
= nagcount
- 1; agno
>= oagcount
; agno
--, new -= agsize
) {
200 * AG freelist header block
202 bp
= xfs_buf_get(mp
->m_ddev_targp
,
203 XFS_AG_DADDR(mp
, agno
, XFS_AGF_DADDR(mp
)),
204 XFS_FSS_TO_BB(mp
, 1), 0);
205 agf
= XFS_BUF_TO_AGF(bp
);
206 memset(agf
, 0, mp
->m_sb
.sb_sectsize
);
207 INT_SET(agf
->agf_magicnum
, ARCH_CONVERT
, XFS_AGF_MAGIC
);
208 INT_SET(agf
->agf_versionnum
, ARCH_CONVERT
, XFS_AGF_VERSION
);
209 INT_SET(agf
->agf_seqno
, ARCH_CONVERT
, agno
);
210 if (agno
== nagcount
- 1)
213 (agno
* (xfs_rfsblock_t
)mp
->m_sb
.sb_agblocks
);
215 agsize
= mp
->m_sb
.sb_agblocks
;
216 INT_SET(agf
->agf_length
, ARCH_CONVERT
, agsize
);
217 INT_SET(agf
->agf_roots
[XFS_BTNUM_BNOi
], ARCH_CONVERT
,
219 INT_SET(agf
->agf_roots
[XFS_BTNUM_CNTi
], ARCH_CONVERT
,
221 INT_SET(agf
->agf_levels
[XFS_BTNUM_BNOi
], ARCH_CONVERT
, 1);
222 INT_SET(agf
->agf_levels
[XFS_BTNUM_CNTi
], ARCH_CONVERT
, 1);
223 agf
->agf_flfirst
= 0;
224 INT_SET(agf
->agf_fllast
, ARCH_CONVERT
, XFS_AGFL_SIZE(mp
) - 1);
225 agf
->agf_flcount
= 0;
226 tmpsize
= agsize
- XFS_PREALLOC_BLOCKS(mp
);
227 INT_SET(agf
->agf_freeblks
, ARCH_CONVERT
, tmpsize
);
228 INT_SET(agf
->agf_longest
, ARCH_CONVERT
, tmpsize
);
229 error
= xfs_bwrite(mp
, bp
);
234 * AG inode header block
236 bp
= xfs_buf_get(mp
->m_ddev_targp
,
237 XFS_AG_DADDR(mp
, agno
, XFS_AGI_DADDR(mp
)),
238 XFS_FSS_TO_BB(mp
, 1), 0);
239 agi
= XFS_BUF_TO_AGI(bp
);
240 memset(agi
, 0, mp
->m_sb
.sb_sectsize
);
241 INT_SET(agi
->agi_magicnum
, ARCH_CONVERT
, XFS_AGI_MAGIC
);
242 INT_SET(agi
->agi_versionnum
, ARCH_CONVERT
, XFS_AGI_VERSION
);
243 INT_SET(agi
->agi_seqno
, ARCH_CONVERT
, agno
);
244 INT_SET(agi
->agi_length
, ARCH_CONVERT
, agsize
);
246 INT_SET(agi
->agi_root
, ARCH_CONVERT
, XFS_IBT_BLOCK(mp
));
247 INT_SET(agi
->agi_level
, ARCH_CONVERT
, 1);
248 agi
->agi_freecount
= 0;
249 INT_SET(agi
->agi_newino
, ARCH_CONVERT
, NULLAGINO
);
250 INT_SET(agi
->agi_dirino
, ARCH_CONVERT
, NULLAGINO
);
251 for (bucket
= 0; bucket
< XFS_AGI_UNLINKED_BUCKETS
; bucket
++)
252 INT_SET(agi
->agi_unlinked
[bucket
], ARCH_CONVERT
,
254 error
= xfs_bwrite(mp
, bp
);
259 * BNO btree root block
261 bp
= xfs_buf_get(mp
->m_ddev_targp
,
262 XFS_AGB_TO_DADDR(mp
, agno
, XFS_BNO_BLOCK(mp
)),
263 BTOBB(mp
->m_sb
.sb_blocksize
), 0);
264 block
= XFS_BUF_TO_SBLOCK(bp
);
265 memset(block
, 0, mp
->m_sb
.sb_blocksize
);
266 INT_SET(block
->bb_magic
, ARCH_CONVERT
, XFS_ABTB_MAGIC
);
268 INT_SET(block
->bb_numrecs
, ARCH_CONVERT
, 1);
269 INT_SET(block
->bb_leftsib
, ARCH_CONVERT
, NULLAGBLOCK
);
270 INT_SET(block
->bb_rightsib
, ARCH_CONVERT
, NULLAGBLOCK
);
271 arec
= XFS_BTREE_REC_ADDR(mp
->m_sb
.sb_blocksize
, xfs_alloc
,
272 block
, 1, mp
->m_alloc_mxr
[0]);
273 INT_SET(arec
->ar_startblock
, ARCH_CONVERT
,
274 XFS_PREALLOC_BLOCKS(mp
));
275 INT_SET(arec
->ar_blockcount
, ARCH_CONVERT
,
276 agsize
- INT_GET(arec
->ar_startblock
, ARCH_CONVERT
));
277 error
= xfs_bwrite(mp
, bp
);
282 * CNT btree root block
284 bp
= xfs_buf_get(mp
->m_ddev_targp
,
285 XFS_AGB_TO_DADDR(mp
, agno
, XFS_CNT_BLOCK(mp
)),
286 BTOBB(mp
->m_sb
.sb_blocksize
), 0);
287 block
= XFS_BUF_TO_SBLOCK(bp
);
288 memset(block
, 0, mp
->m_sb
.sb_blocksize
);
289 INT_SET(block
->bb_magic
, ARCH_CONVERT
, XFS_ABTC_MAGIC
);
291 INT_SET(block
->bb_numrecs
, ARCH_CONVERT
, 1);
292 INT_SET(block
->bb_leftsib
, ARCH_CONVERT
, NULLAGBLOCK
);
293 INT_SET(block
->bb_rightsib
, ARCH_CONVERT
, NULLAGBLOCK
);
294 arec
= XFS_BTREE_REC_ADDR(mp
->m_sb
.sb_blocksize
, xfs_alloc
,
295 block
, 1, mp
->m_alloc_mxr
[0]);
296 INT_SET(arec
->ar_startblock
, ARCH_CONVERT
,
297 XFS_PREALLOC_BLOCKS(mp
));
298 INT_SET(arec
->ar_blockcount
, ARCH_CONVERT
,
299 agsize
- INT_GET(arec
->ar_startblock
, ARCH_CONVERT
));
300 nfree
+= INT_GET(arec
->ar_blockcount
, ARCH_CONVERT
);
301 error
= xfs_bwrite(mp
, bp
);
306 * INO btree root block
308 bp
= xfs_buf_get(mp
->m_ddev_targp
,
309 XFS_AGB_TO_DADDR(mp
, agno
, XFS_IBT_BLOCK(mp
)),
310 BTOBB(mp
->m_sb
.sb_blocksize
), 0);
311 block
= XFS_BUF_TO_SBLOCK(bp
);
312 memset(block
, 0, mp
->m_sb
.sb_blocksize
);
313 INT_SET(block
->bb_magic
, ARCH_CONVERT
, XFS_IBT_MAGIC
);
315 block
->bb_numrecs
= 0;
316 INT_SET(block
->bb_leftsib
, ARCH_CONVERT
, NULLAGBLOCK
);
317 INT_SET(block
->bb_rightsib
, ARCH_CONVERT
, NULLAGBLOCK
);
318 error
= xfs_bwrite(mp
, bp
);
323 xfs_trans_agblocks_delta(tp
, nfree
);
325 * There are new blocks in the old last a.g.
329 * Change the agi length.
331 error
= xfs_ialloc_read_agi(mp
, tp
, agno
, &bp
);
336 agi
= XFS_BUF_TO_AGI(bp
);
337 INT_MOD(agi
->agi_length
, ARCH_CONVERT
, new);
338 ASSERT(nagcount
== oagcount
||
339 INT_GET(agi
->agi_length
, ARCH_CONVERT
) ==
340 mp
->m_sb
.sb_agblocks
);
341 xfs_ialloc_log_agi(tp
, bp
, XFS_AGI_LENGTH
);
345 error
= xfs_alloc_read_agf(mp
, tp
, agno
, 0, &bp
);
350 agf
= XFS_BUF_TO_AGF(bp
);
351 INT_MOD(agf
->agf_length
, ARCH_CONVERT
, new);
352 ASSERT(INT_GET(agf
->agf_length
, ARCH_CONVERT
) ==
353 INT_GET(agi
->agi_length
, ARCH_CONVERT
));
355 * Free the new space.
357 error
= xfs_free_extent(tp
, XFS_AGB_TO_FSB(mp
, agno
,
358 INT_GET(agf
->agf_length
, ARCH_CONVERT
) - new), new);
363 if (nagcount
> oagcount
)
364 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_AGCOUNT
, nagcount
- oagcount
);
365 if (nb
> mp
->m_sb
.sb_dblocks
)
366 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_DBLOCKS
,
367 nb
- mp
->m_sb
.sb_dblocks
);
369 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_FDBLOCKS
, nfree
);
371 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_IMAXPCT
, dpct
);
372 error
= xfs_trans_commit(tp
, 0, NULL
);
376 /* New allocation groups fully initialized, so update mount struct */
378 mp
->m_maxagi
= nagimax
;
379 if (mp
->m_sb
.sb_imax_pct
) {
380 __uint64_t icount
= mp
->m_sb
.sb_dblocks
* mp
->m_sb
.sb_imax_pct
;
382 mp
->m_maxicount
= icount
<< mp
->m_sb
.sb_inopblog
;
385 for (agno
= 1; agno
< nagcount
; agno
++) {
386 error
= xfs_read_buf(mp
, mp
->m_ddev_targp
,
387 XFS_AGB_TO_DADDR(mp
, agno
, XFS_SB_BLOCK(mp
)),
388 XFS_FSS_TO_BB(mp
, 1), 0, &bp
);
390 xfs_fs_cmn_err(CE_WARN
, mp
,
391 "error %d reading secondary superblock for ag %d",
395 sbp
= XFS_BUF_TO_SBP(bp
);
396 xfs_xlatesb(sbp
, &mp
->m_sb
, -1, XFS_SB_ALL_BITS
);
398 * If we get an error writing out the alternate superblocks,
399 * just issue a warning and continue. The real work is
400 * already done and committed.
402 if (!(error
= xfs_bwrite(mp
, bp
))) {
405 xfs_fs_cmn_err(CE_WARN
, mp
,
406 "write error %d updating secondary superblock for ag %d",
408 break; /* no point in continuing */
414 xfs_trans_cancel(tp
, XFS_TRANS_ABORT
);
419 xfs_growfs_log_private(
420 xfs_mount_t
*mp
, /* mount point for filesystem */
421 xfs_growfs_log_t
*in
) /* growfs log input struct */
426 if (nb
< XFS_MIN_LOG_BLOCKS
|| nb
< XFS_B_TO_FSB(mp
, XFS_MIN_LOG_BYTES
))
427 return XFS_ERROR(EINVAL
);
428 if (nb
== mp
->m_sb
.sb_logblocks
&&
429 in
->isint
== (mp
->m_sb
.sb_logstart
!= 0))
430 return XFS_ERROR(EINVAL
);
432 * Moving the log is hard, need new interfaces to sync
433 * the log first, hold off all activity while moving it.
434 * Can have shorter or longer log in the same space,
435 * or transform internal to external log or vice versa.
437 return XFS_ERROR(ENOSYS
);
441 * protected versions of growfs function acquire and release locks on the mount
442 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
450 xfs_growfs_data_t
*in
)
453 if (!cpsema(&mp
->m_growlock
))
454 return XFS_ERROR(EWOULDBLOCK
);
455 error
= xfs_growfs_data_private(mp
, in
);
456 vsema(&mp
->m_growlock
);
463 xfs_growfs_log_t
*in
)
466 if (!cpsema(&mp
->m_growlock
))
467 return XFS_ERROR(EWOULDBLOCK
);
468 error
= xfs_growfs_log_private(mp
, in
);
469 vsema(&mp
->m_growlock
);
474 * exported through ioctl XFS_IOC_FSCOUNTS
480 xfs_fsop_counts_t
*cnt
)
485 cnt
->freedata
= mp
->m_sb
.sb_fdblocks
;
486 cnt
->freertx
= mp
->m_sb
.sb_frextents
;
487 cnt
->freeino
= mp
->m_sb
.sb_ifree
;
488 cnt
->allocino
= mp
->m_sb
.sb_icount
;
489 XFS_SB_UNLOCK(mp
, s
);
494 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
496 * xfs_reserve_blocks is called to set m_resblks
497 * in the in-core mount table. The number of unused reserved blocks
498 * is kept in m_resbls_avail.
500 * Reserve the requested number of blocks if available. Otherwise return
501 * as many as possible to satisfy the request. The actual number
502 * reserved are returned in outval
504 * A null inval pointer indicates that only the current reserved blocks
505 * available should be returned no settings are changed.
512 xfs_fsop_resblks_t
*outval
)
514 __int64_t lcounter
, delta
;
518 /* If inval is null, report current values and return */
520 if (inval
== (__uint64_t
*)NULL
) {
521 outval
->resblks
= mp
->m_resblks
;
522 outval
->resblks_avail
= mp
->m_resblks_avail
;
530 * If our previous reservation was larger than the current value,
531 * then move any unused blocks back to the free pool.
534 if (mp
->m_resblks
> request
) {
535 lcounter
= mp
->m_resblks_avail
- request
;
536 if (lcounter
> 0) { /* release unused blocks */
537 mp
->m_sb
.sb_fdblocks
+= lcounter
;
538 mp
->m_resblks_avail
-= lcounter
;
540 mp
->m_resblks
= request
;
542 delta
= request
- mp
->m_resblks
;
543 lcounter
= mp
->m_sb
.sb_fdblocks
- delta
;
545 /* We can't satisfy the request, just get what we can */
546 mp
->m_resblks
+= mp
->m_sb
.sb_fdblocks
;
547 mp
->m_resblks_avail
+= mp
->m_sb
.sb_fdblocks
;
548 mp
->m_sb
.sb_fdblocks
= 0;
550 mp
->m_sb
.sb_fdblocks
= lcounter
;
551 mp
->m_resblks
= request
;
552 mp
->m_resblks_avail
+= delta
;
556 outval
->resblks
= mp
->m_resblks
;
557 outval
->resblks_avail
= mp
->m_resblks_avail
;
558 XFS_SB_UNLOCK(mp
, s
);
568 case XFS_FSOP_GOING_FLAGS_DEFAULT
: {
569 struct vfs
*vfsp
= XFS_MTOVFS(mp
);
570 struct super_block
*sb
= freeze_bdev(vfsp
->vfs_super
->s_bdev
);
573 xfs_force_shutdown(mp
, XFS_FORCE_UMOUNT
);
574 thaw_bdev(sb
->s_bdev
, sb
);
579 case XFS_FSOP_GOING_FLAGS_LOGFLUSH
:
580 xfs_force_shutdown(mp
, XFS_FORCE_UMOUNT
);
582 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH
:
583 xfs_force_shutdown(mp
, XFS_FORCE_UMOUNT
|XFS_LOG_IO_ERROR
);
586 return XFS_ERROR(EINVAL
);