2 * Copyright (c) 2000-2002,2005 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"
27 #include "xfs_mount.h"
28 #include "xfs_bmap_btree.h"
29 #include "xfs_alloc_btree.h"
30 #include "xfs_ialloc_btree.h"
31 #include "xfs_dinode.h"
32 #include "xfs_inode.h"
33 #include "xfs_btree.h"
34 #include "xfs_ialloc.h"
35 #include "xfs_alloc.h"
36 #include "xfs_rtalloc.h"
37 #include "xfs_error.h"
42 * Allocation group level functions.
45 xfs_ialloc_cluster_alignment(
46 xfs_alloc_arg_t
*args
)
48 if (xfs_sb_version_hasalign(&args
->mp
->m_sb
) &&
49 args
->mp
->m_sb
.sb_inoalignmt
>=
50 XFS_B_TO_FSBT(args
->mp
, XFS_INODE_CLUSTER_SIZE(args
->mp
)))
51 return args
->mp
->m_sb
.sb_inoalignmt
;
56 * Lookup a record by ino in the btree given by cur.
60 struct xfs_btree_cur
*cur
, /* btree cursor */
61 xfs_agino_t ino
, /* starting inode of chunk */
62 xfs_lookup_t dir
, /* <=, >=, == */
63 int *stat
) /* success/failure */
65 cur
->bc_rec
.i
.ir_startino
= ino
;
66 cur
->bc_rec
.i
.ir_freecount
= 0;
67 cur
->bc_rec
.i
.ir_free
= 0;
68 return xfs_btree_lookup(cur
, dir
, stat
);
72 * Update the record referred to by cur to the value given.
73 * This either works (return 0) or gets an EFSCORRUPTED error.
75 STATIC
int /* error */
77 struct xfs_btree_cur
*cur
, /* btree cursor */
78 xfs_inobt_rec_incore_t
*irec
) /* btree record */
80 union xfs_btree_rec rec
;
82 rec
.inobt
.ir_startino
= cpu_to_be32(irec
->ir_startino
);
83 rec
.inobt
.ir_freecount
= cpu_to_be32(irec
->ir_freecount
);
84 rec
.inobt
.ir_free
= cpu_to_be64(irec
->ir_free
);
85 return xfs_btree_update(cur
, &rec
);
89 * Get the data from the pointed-to record.
93 struct xfs_btree_cur
*cur
, /* btree cursor */
94 xfs_inobt_rec_incore_t
*irec
, /* btree record */
95 int *stat
) /* output: success/failure */
97 union xfs_btree_rec
*rec
;
100 error
= xfs_btree_get_rec(cur
, &rec
, stat
);
101 if (!error
&& *stat
== 1) {
102 irec
->ir_startino
= be32_to_cpu(rec
->inobt
.ir_startino
);
103 irec
->ir_freecount
= be32_to_cpu(rec
->inobt
.ir_freecount
);
104 irec
->ir_free
= be64_to_cpu(rec
->inobt
.ir_free
);
110 * Verify that the number of free inodes in the AGI is correct.
114 xfs_check_agi_freecount(
115 struct xfs_btree_cur
*cur
,
118 if (cur
->bc_nlevels
== 1) {
119 xfs_inobt_rec_incore_t rec
;
124 error
= xfs_inobt_lookup(cur
, 0, XFS_LOOKUP_GE
, &i
);
129 error
= xfs_inobt_get_rec(cur
, &rec
, &i
);
134 freecount
+= rec
.ir_freecount
;
135 error
= xfs_btree_increment(cur
, 0, &i
);
141 if (!XFS_FORCED_SHUTDOWN(cur
->bc_mp
))
142 ASSERT(freecount
== be32_to_cpu(agi
->agi_freecount
));
147 #define xfs_check_agi_freecount(cur, agi) 0
151 * Initialise a new set of inodes.
154 xfs_ialloc_inode_init(
155 struct xfs_mount
*mp
,
156 struct xfs_trans
*tp
,
159 xfs_agblock_t length
,
162 struct xfs_buf
*fbuf
;
163 struct xfs_dinode
*free
;
164 int blks_per_cluster
, nbufs
, ninodes
;
170 * Loop over the new block(s), filling in the inodes.
171 * For small block sizes, manipulate the inodes in buffers
172 * which are multiples of the blocks size.
174 if (mp
->m_sb
.sb_blocksize
>= XFS_INODE_CLUSTER_SIZE(mp
)) {
175 blks_per_cluster
= 1;
177 ninodes
= mp
->m_sb
.sb_inopblock
;
179 blks_per_cluster
= XFS_INODE_CLUSTER_SIZE(mp
) /
180 mp
->m_sb
.sb_blocksize
;
181 nbufs
= length
/ blks_per_cluster
;
182 ninodes
= blks_per_cluster
* mp
->m_sb
.sb_inopblock
;
186 * Figure out what version number to use in the inodes we create.
187 * If the superblock version has caught up to the one that supports
188 * the new inode format, then use the new inode version. Otherwise
189 * use the old version so that old kernels will continue to be
190 * able to use the file system.
192 if (xfs_sb_version_hasnlink(&mp
->m_sb
))
197 for (j
= 0; j
< nbufs
; j
++) {
201 d
= XFS_AGB_TO_DADDR(mp
, agno
, agbno
+ (j
* blks_per_cluster
));
202 fbuf
= xfs_trans_get_buf(tp
, mp
->m_ddev_targp
, d
,
203 mp
->m_bsize
* blks_per_cluster
,
206 ASSERT(!XFS_BUF_GETERROR(fbuf
));
208 xfs_biozero(fbuf
, 0, ninodes
<< mp
->m_sb
.sb_inodelog
);
209 for (i
= 0; i
< ninodes
; i
++) {
210 int ioffset
= i
<< mp
->m_sb
.sb_inodelog
;
211 uint isize
= sizeof(struct xfs_dinode
);
213 free
= xfs_make_iptr(mp
, fbuf
, i
);
214 free
->di_magic
= cpu_to_be16(XFS_DINODE_MAGIC
);
215 free
->di_version
= version
;
216 free
->di_gen
= cpu_to_be32(gen
);
217 free
->di_next_unlinked
= cpu_to_be32(NULLAGINO
);
218 xfs_trans_log_buf(tp
, fbuf
, ioffset
, ioffset
+ isize
- 1);
220 xfs_trans_inode_alloc_buf(tp
, fbuf
);
225 * Allocate new inodes in the allocation group specified by agbp.
226 * Return 0 for success, else error code.
228 STATIC
int /* error code or 0 */
230 xfs_trans_t
*tp
, /* transaction pointer */
231 xfs_buf_t
*agbp
, /* alloc group buffer */
234 xfs_agi_t
*agi
; /* allocation group header */
235 xfs_alloc_arg_t args
; /* allocation argument structure */
236 xfs_btree_cur_t
*cur
; /* inode btree cursor */
240 xfs_agino_t newino
; /* new first inode's number */
241 xfs_agino_t newlen
; /* new number of inodes */
242 xfs_agino_t thisino
; /* current inode number, for loop */
243 int isaligned
= 0; /* inode allocation at stripe unit */
245 struct xfs_perag
*pag
;
248 args
.mp
= tp
->t_mountp
;
251 * Locking will ensure that we don't have two callers in here
254 newlen
= XFS_IALLOC_INODES(args
.mp
);
255 if (args
.mp
->m_maxicount
&&
256 args
.mp
->m_sb
.sb_icount
+ newlen
> args
.mp
->m_maxicount
)
257 return XFS_ERROR(ENOSPC
);
258 args
.minlen
= args
.maxlen
= XFS_IALLOC_BLOCKS(args
.mp
);
260 * First try to allocate inodes contiguous with the last-allocated
261 * chunk of inodes. If the filesystem is striped, this will fill
262 * an entire stripe unit with inodes.
264 agi
= XFS_BUF_TO_AGI(agbp
);
265 newino
= be32_to_cpu(agi
->agi_newino
);
266 agno
= be32_to_cpu(agi
->agi_seqno
);
267 args
.agbno
= XFS_AGINO_TO_AGBNO(args
.mp
, newino
) +
268 XFS_IALLOC_BLOCKS(args
.mp
);
269 if (likely(newino
!= NULLAGINO
&&
270 (args
.agbno
< be32_to_cpu(agi
->agi_length
)))) {
271 args
.fsbno
= XFS_AGB_TO_FSB(args
.mp
, agno
, args
.agbno
);
272 args
.type
= XFS_ALLOCTYPE_THIS_BNO
;
273 args
.mod
= args
.total
= args
.wasdel
= args
.isfl
=
274 args
.userdata
= args
.minalignslop
= 0;
278 * We need to take into account alignment here to ensure that
279 * we don't modify the free list if we fail to have an exact
280 * block. If we don't have an exact match, and every oher
281 * attempt allocation attempt fails, we'll end up cancelling
282 * a dirty transaction and shutting down.
284 * For an exact allocation, alignment must be 1,
285 * however we need to take cluster alignment into account when
286 * fixing up the freelist. Use the minalignslop field to
287 * indicate that extra blocks might be required for alignment,
288 * but not to use them in the actual exact allocation.
291 args
.minalignslop
= xfs_ialloc_cluster_alignment(&args
) - 1;
293 /* Allow space for the inode btree to split. */
294 args
.minleft
= args
.mp
->m_in_maxlevels
- 1;
295 if ((error
= xfs_alloc_vextent(&args
)))
298 args
.fsbno
= NULLFSBLOCK
;
300 if (unlikely(args
.fsbno
== NULLFSBLOCK
)) {
302 * Set the alignment for the allocation.
303 * If stripe alignment is turned on then align at stripe unit
305 * If the cluster size is smaller than a filesystem block
306 * then we're doing I/O for inodes in filesystem block size
307 * pieces, so don't need alignment anyway.
310 if (args
.mp
->m_sinoalign
) {
311 ASSERT(!(args
.mp
->m_flags
& XFS_MOUNT_NOALIGN
));
312 args
.alignment
= args
.mp
->m_dalign
;
315 args
.alignment
= xfs_ialloc_cluster_alignment(&args
);
317 * Need to figure out where to allocate the inode blocks.
318 * Ideally they should be spaced out through the a.g.
319 * For now, just allocate blocks up front.
321 args
.agbno
= be32_to_cpu(agi
->agi_root
);
322 args
.fsbno
= XFS_AGB_TO_FSB(args
.mp
, agno
, args
.agbno
);
324 * Allocate a fixed-size extent of inodes.
326 args
.type
= XFS_ALLOCTYPE_NEAR_BNO
;
327 args
.mod
= args
.total
= args
.wasdel
= args
.isfl
=
328 args
.userdata
= args
.minalignslop
= 0;
331 * Allow space for the inode btree to split.
333 args
.minleft
= args
.mp
->m_in_maxlevels
- 1;
334 if ((error
= xfs_alloc_vextent(&args
)))
339 * If stripe alignment is turned on, then try again with cluster
342 if (isaligned
&& args
.fsbno
== NULLFSBLOCK
) {
343 args
.type
= XFS_ALLOCTYPE_NEAR_BNO
;
344 args
.agbno
= be32_to_cpu(agi
->agi_root
);
345 args
.fsbno
= XFS_AGB_TO_FSB(args
.mp
, agno
, args
.agbno
);
346 args
.alignment
= xfs_ialloc_cluster_alignment(&args
);
347 if ((error
= xfs_alloc_vextent(&args
)))
351 if (args
.fsbno
== NULLFSBLOCK
) {
355 ASSERT(args
.len
== args
.minlen
);
358 * Stamp and write the inode buffers.
360 * Seed the new inode cluster with a random generation number. This
361 * prevents short-term reuse of generation numbers if a chunk is
362 * freed and then immediately reallocated. We use random numbers
363 * rather than a linear progression to prevent the next generation
364 * number from being easily guessable.
366 xfs_ialloc_inode_init(args
.mp
, tp
, agno
, args
.agbno
, args
.len
,
370 * Convert the results.
372 newino
= XFS_OFFBNO_TO_AGINO(args
.mp
, args
.agbno
, 0);
373 be32_add_cpu(&agi
->agi_count
, newlen
);
374 be32_add_cpu(&agi
->agi_freecount
, newlen
);
375 pag
= xfs_perag_get(args
.mp
, agno
);
376 pag
->pagi_freecount
+= newlen
;
378 agi
->agi_newino
= cpu_to_be32(newino
);
381 * Insert records describing the new inode chunk into the btree.
383 cur
= xfs_inobt_init_cursor(args
.mp
, tp
, agbp
, agno
);
384 for (thisino
= newino
;
385 thisino
< newino
+ newlen
;
386 thisino
+= XFS_INODES_PER_CHUNK
) {
387 cur
->bc_rec
.i
.ir_startino
= thisino
;
388 cur
->bc_rec
.i
.ir_freecount
= XFS_INODES_PER_CHUNK
;
389 cur
->bc_rec
.i
.ir_free
= XFS_INOBT_ALL_FREE
;
390 error
= xfs_btree_lookup(cur
, XFS_LOOKUP_EQ
, &i
);
392 xfs_btree_del_cursor(cur
, XFS_BTREE_ERROR
);
396 error
= xfs_btree_insert(cur
, &i
);
398 xfs_btree_del_cursor(cur
, XFS_BTREE_ERROR
);
403 xfs_btree_del_cursor(cur
, XFS_BTREE_NOERROR
);
405 * Log allocation group header fields
407 xfs_ialloc_log_agi(tp
, agbp
,
408 XFS_AGI_COUNT
| XFS_AGI_FREECOUNT
| XFS_AGI_NEWINO
);
410 * Modify/log superblock values for inode count and inode free count.
412 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_ICOUNT
, (long)newlen
);
413 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_IFREE
, (long)newlen
);
418 STATIC xfs_agnumber_t
424 spin_lock(&mp
->m_agirotor_lock
);
425 agno
= mp
->m_agirotor
;
426 if (++mp
->m_agirotor
== mp
->m_maxagi
)
428 spin_unlock(&mp
->m_agirotor_lock
);
434 * Select an allocation group to look for a free inode in, based on the parent
435 * inode and then mode. Return the allocation group buffer.
437 STATIC xfs_buf_t
* /* allocation group buffer */
438 xfs_ialloc_ag_select(
439 xfs_trans_t
*tp
, /* transaction pointer */
440 xfs_ino_t parent
, /* parent directory inode number */
441 mode_t mode
, /* bits set to indicate file type */
442 int okalloc
) /* ok to allocate more space */
444 xfs_buf_t
*agbp
; /* allocation group header buffer */
445 xfs_agnumber_t agcount
; /* number of ag's in the filesystem */
446 xfs_agnumber_t agno
; /* current ag number */
447 int flags
; /* alloc buffer locking flags */
448 xfs_extlen_t ineed
; /* blocks needed for inode allocation */
449 xfs_extlen_t longest
= 0; /* longest extent available */
450 xfs_mount_t
*mp
; /* mount point structure */
451 int needspace
; /* file mode implies space allocated */
452 xfs_perag_t
*pag
; /* per allocation group data */
453 xfs_agnumber_t pagno
; /* parent (starting) ag number */
456 * Files of these types need at least one block if length > 0
457 * (and they won't fit in the inode, but that's hard to figure out).
459 needspace
= S_ISDIR(mode
) || S_ISREG(mode
) || S_ISLNK(mode
);
461 agcount
= mp
->m_maxagi
;
463 pagno
= xfs_ialloc_next_ag(mp
);
465 pagno
= XFS_INO_TO_AGNO(mp
, parent
);
466 if (pagno
>= agcount
)
469 ASSERT(pagno
< agcount
);
471 * Loop through allocation groups, looking for one with a little
472 * free space in it. Note we don't look for free inodes, exactly.
473 * Instead, we include whether there is a need to allocate inodes
474 * to mean that blocks must be allocated for them,
475 * if none are currently free.
478 flags
= XFS_ALLOC_FLAG_TRYLOCK
;
480 pag
= xfs_perag_get(mp
, agno
);
481 if (!pag
->pagi_init
) {
482 if (xfs_ialloc_read_agi(mp
, tp
, agno
, &agbp
)) {
489 if (!pag
->pagi_inodeok
) {
490 xfs_ialloc_next_ag(mp
);
495 * Is there enough free space for the file plus a block
496 * of inodes (if we need to allocate some)?
498 ineed
= pag
->pagi_freecount
? 0 : XFS_IALLOC_BLOCKS(mp
);
499 if (ineed
&& !pag
->pagf_init
) {
501 xfs_ialloc_read_agi(mp
, tp
, agno
, &agbp
)) {
505 (void)xfs_alloc_pagf_init(mp
, tp
, agno
, flags
);
507 if (!ineed
|| pag
->pagf_init
) {
508 if (ineed
&& !(longest
= pag
->pagf_longest
))
509 longest
= pag
->pagf_flcount
> 0;
511 (pag
->pagf_freeblks
>= needspace
+ ineed
&&
515 xfs_ialloc_read_agi(mp
, tp
, agno
, &agbp
)) {
525 xfs_trans_brelse(tp
, agbp
);
529 * No point in iterating over the rest, if we're shutting
532 if (XFS_FORCED_SHUTDOWN(mp
))
546 * Try to retrieve the next record to the left/right from the current one.
550 struct xfs_btree_cur
*cur
,
551 xfs_inobt_rec_incore_t
*rec
,
559 error
= xfs_btree_decrement(cur
, 0, &i
);
561 error
= xfs_btree_increment(cur
, 0, &i
);
567 error
= xfs_inobt_get_rec(cur
, rec
, &i
);
570 XFS_WANT_CORRUPTED_RETURN(i
== 1);
578 struct xfs_btree_cur
*cur
,
580 xfs_inobt_rec_incore_t
*rec
,
587 error
= xfs_inobt_lookup(cur
, agino
, XFS_LOOKUP_EQ
, &i
);
592 error
= xfs_inobt_get_rec(cur
, rec
, &i
);
595 XFS_WANT_CORRUPTED_RETURN(i
== 1);
602 * Visible inode allocation functions.
606 * Allocate an inode on disk.
607 * Mode is used to tell whether the new inode will need space, and whether
610 * The arguments IO_agbp and alloc_done are defined to work within
611 * the constraint of one allocation per transaction.
612 * xfs_dialloc() is designed to be called twice if it has to do an
613 * allocation to make more free inodes. On the first call,
614 * IO_agbp should be set to NULL. If an inode is available,
615 * i.e., xfs_dialloc() did not need to do an allocation, an inode
616 * number is returned. In this case, IO_agbp would be set to the
617 * current ag_buf and alloc_done set to false.
618 * If an allocation needed to be done, xfs_dialloc would return
619 * the current ag_buf in IO_agbp and set alloc_done to true.
620 * The caller should then commit the current transaction, allocate a new
621 * transaction, and call xfs_dialloc() again, passing in the previous
622 * value of IO_agbp. IO_agbp should be held across the transactions.
623 * Since the agbp is locked across the two calls, the second call is
624 * guaranteed to have a free inode available.
626 * Once we successfully pick an inode its number is returned and the
627 * on-disk data structures are updated. The inode itself is not read
628 * in, since doing so would break ordering constraints with xfs_reclaim.
632 xfs_trans_t
*tp
, /* transaction pointer */
633 xfs_ino_t parent
, /* parent inode (directory) */
634 mode_t mode
, /* mode bits for new inode */
635 int okalloc
, /* ok to allocate more space */
636 xfs_buf_t
**IO_agbp
, /* in/out ag header's buffer */
637 boolean_t
*alloc_done
, /* true if we needed to replenish
639 xfs_ino_t
*inop
) /* inode number allocated */
641 xfs_agnumber_t agcount
; /* number of allocation groups */
642 xfs_buf_t
*agbp
; /* allocation group header's buffer */
643 xfs_agnumber_t agno
; /* allocation group number */
644 xfs_agi_t
*agi
; /* allocation group header structure */
645 xfs_btree_cur_t
*cur
; /* inode allocation btree cursor */
646 int error
; /* error return value */
647 int i
; /* result code */
648 int ialloced
; /* inode allocation status */
649 int noroom
= 0; /* no space for inode blk allocation */
650 xfs_ino_t ino
; /* fs-relative inode to be returned */
652 int j
; /* result code */
653 xfs_mount_t
*mp
; /* file system mount structure */
654 int offset
; /* index of inode in chunk */
655 xfs_agino_t pagino
; /* parent's AG relative inode # */
656 xfs_agnumber_t pagno
; /* parent's AG number */
657 xfs_inobt_rec_incore_t rec
; /* inode allocation record */
658 xfs_agnumber_t tagno
; /* testing allocation group number */
659 xfs_btree_cur_t
*tcur
; /* temp cursor */
660 xfs_inobt_rec_incore_t trec
; /* temp inode allocation record */
661 struct xfs_perag
*pag
;
664 if (*IO_agbp
== NULL
) {
666 * We do not have an agbp, so select an initial allocation
667 * group for inode allocation.
669 agbp
= xfs_ialloc_ag_select(tp
, parent
, mode
, okalloc
);
671 * Couldn't find an allocation group satisfying the
678 agi
= XFS_BUF_TO_AGI(agbp
);
679 ASSERT(be32_to_cpu(agi
->agi_magicnum
) == XFS_AGI_MAGIC
);
682 * Continue where we left off before. In this case, we
683 * know that the allocation group has free inodes.
686 agi
= XFS_BUF_TO_AGI(agbp
);
687 ASSERT(be32_to_cpu(agi
->agi_magicnum
) == XFS_AGI_MAGIC
);
688 ASSERT(be32_to_cpu(agi
->agi_freecount
) > 0);
691 agcount
= mp
->m_sb
.sb_agcount
;
692 agno
= be32_to_cpu(agi
->agi_seqno
);
694 pagno
= XFS_INO_TO_AGNO(mp
, parent
);
695 pagino
= XFS_INO_TO_AGINO(mp
, parent
);
698 * If we have already hit the ceiling of inode blocks then clear
699 * okalloc so we scan all available agi structures for a free
703 if (mp
->m_maxicount
&&
704 mp
->m_sb
.sb_icount
+ XFS_IALLOC_INODES(mp
) > mp
->m_maxicount
) {
710 * Loop until we find an allocation group that either has free inodes
711 * or in which we can allocate some inodes. Iterate through the
712 * allocation groups upward, wrapping at the end.
714 *alloc_done
= B_FALSE
;
715 while (!agi
->agi_freecount
) {
717 * Don't do anything if we're not supposed to allocate
718 * any blocks, just go on to the next ag.
722 * Try to allocate some new inodes in the allocation
725 if ((error
= xfs_ialloc_ag_alloc(tp
, agbp
, &ialloced
))) {
726 xfs_trans_brelse(tp
, agbp
);
727 if (error
== ENOSPC
) {
735 * We successfully allocated some inodes, return
736 * the current context to the caller so that it
737 * can commit the current transaction and call
738 * us again where we left off.
740 ASSERT(be32_to_cpu(agi
->agi_freecount
) > 0);
741 *alloc_done
= B_TRUE
;
748 * If it failed, give up on this ag.
750 xfs_trans_brelse(tp
, agbp
);
752 * Go on to the next ag: get its ag header.
755 if (++tagno
== agcount
)
759 return noroom
? ENOSPC
: 0;
761 pag
= xfs_perag_get(mp
, tagno
);
762 if (pag
->pagi_inodeok
== 0) {
766 error
= xfs_ialloc_read_agi(mp
, tp
, tagno
, &agbp
);
770 agi
= XFS_BUF_TO_AGI(agbp
);
771 ASSERT(be32_to_cpu(agi
->agi_magicnum
) == XFS_AGI_MAGIC
);
774 * Here with an allocation group that has a free inode.
775 * Reset agno since we may have chosen a new ag in the
780 pag
= xfs_perag_get(mp
, agno
);
783 cur
= xfs_inobt_init_cursor(mp
, tp
, agbp
, be32_to_cpu(agi
->agi_seqno
));
785 * If pagino is 0 (this is the root inode allocation) use newino.
786 * This must work because we've just allocated some.
789 pagino
= be32_to_cpu(agi
->agi_newino
);
791 error
= xfs_check_agi_freecount(cur
, agi
);
796 * If in the same AG as the parent, try to get near the parent.
799 int doneleft
; /* done, to the left */
800 int doneright
; /* done, to the right */
801 int searchdistance
= 10;
803 error
= xfs_inobt_lookup(cur
, pagino
, XFS_LOOKUP_LE
, &i
);
806 XFS_WANT_CORRUPTED_GOTO(i
== 1, error0
);
808 error
= xfs_inobt_get_rec(cur
, &rec
, &j
);
811 XFS_WANT_CORRUPTED_GOTO(i
== 1, error0
);
813 if (rec
.ir_freecount
> 0) {
815 * Found a free inode in the same chunk
816 * as the parent, done.
823 * In the same AG as parent, but parent's chunk is full.
826 /* duplicate the cursor, search left & right simultaneously */
827 error
= xfs_btree_dup_cursor(cur
, &tcur
);
832 * Skip to last blocks looked up if same parent inode.
834 if (pagino
!= NULLAGINO
&&
835 pag
->pagl_pagino
== pagino
&&
836 pag
->pagl_leftrec
!= NULLAGINO
&&
837 pag
->pagl_rightrec
!= NULLAGINO
) {
838 error
= xfs_ialloc_get_rec(tcur
, pag
->pagl_leftrec
,
839 &trec
, &doneleft
, 1);
843 error
= xfs_ialloc_get_rec(cur
, pag
->pagl_rightrec
,
844 &rec
, &doneright
, 0);
848 /* search left with tcur, back up 1 record */
849 error
= xfs_ialloc_next_rec(tcur
, &trec
, &doneleft
, 1);
853 /* search right with cur, go forward 1 record. */
854 error
= xfs_ialloc_next_rec(cur
, &rec
, &doneright
, 0);
860 * Loop until we find an inode chunk with a free inode.
862 while (!doneleft
|| !doneright
) {
863 int useleft
; /* using left inode chunk this time */
865 if (!--searchdistance
) {
867 * Not in range - save last search
868 * location and allocate a new inode
870 xfs_btree_del_cursor(tcur
, XFS_BTREE_NOERROR
);
871 pag
->pagl_leftrec
= trec
.ir_startino
;
872 pag
->pagl_rightrec
= rec
.ir_startino
;
873 pag
->pagl_pagino
= pagino
;
877 /* figure out the closer block if both are valid. */
878 if (!doneleft
&& !doneright
) {
880 (trec
.ir_startino
+ XFS_INODES_PER_CHUNK
- 1) <
881 rec
.ir_startino
- pagino
;
886 /* free inodes to the left? */
887 if (useleft
&& trec
.ir_freecount
) {
889 xfs_btree_del_cursor(cur
, XFS_BTREE_NOERROR
);
892 pag
->pagl_leftrec
= trec
.ir_startino
;
893 pag
->pagl_rightrec
= rec
.ir_startino
;
894 pag
->pagl_pagino
= pagino
;
898 /* free inodes to the right? */
899 if (!useleft
&& rec
.ir_freecount
) {
900 xfs_btree_del_cursor(tcur
, XFS_BTREE_NOERROR
);
902 pag
->pagl_leftrec
= trec
.ir_startino
;
903 pag
->pagl_rightrec
= rec
.ir_startino
;
904 pag
->pagl_pagino
= pagino
;
908 /* get next record to check */
910 error
= xfs_ialloc_next_rec(tcur
, &trec
,
913 error
= xfs_ialloc_next_rec(cur
, &rec
,
921 * We've reached the end of the btree. because
922 * we are only searching a small chunk of the
923 * btree each search, there is obviously free
924 * inodes closer to the parent inode than we
925 * are now. restart the search again.
927 pag
->pagl_pagino
= NULLAGINO
;
928 pag
->pagl_leftrec
= NULLAGINO
;
929 pag
->pagl_rightrec
= NULLAGINO
;
930 xfs_btree_del_cursor(tcur
, XFS_BTREE_NOERROR
);
931 xfs_btree_del_cursor(cur
, XFS_BTREE_NOERROR
);
936 * In a different AG from the parent.
937 * See if the most recently allocated block has any free.
940 if (be32_to_cpu(agi
->agi_newino
) != NULLAGINO
) {
941 error
= xfs_inobt_lookup(cur
, be32_to_cpu(agi
->agi_newino
),
947 error
= xfs_inobt_get_rec(cur
, &rec
, &j
);
951 if (j
== 1 && rec
.ir_freecount
> 0) {
953 * The last chunk allocated in the group
954 * still has a free inode.
962 * None left in the last group, search the whole AG
964 error
= xfs_inobt_lookup(cur
, 0, XFS_LOOKUP_GE
, &i
);
967 XFS_WANT_CORRUPTED_GOTO(i
== 1, error0
);
970 error
= xfs_inobt_get_rec(cur
, &rec
, &i
);
973 XFS_WANT_CORRUPTED_GOTO(i
== 1, error0
);
974 if (rec
.ir_freecount
> 0)
976 error
= xfs_btree_increment(cur
, 0, &i
);
979 XFS_WANT_CORRUPTED_GOTO(i
== 1, error0
);
983 offset
= xfs_ialloc_find_free(&rec
.ir_free
);
985 ASSERT(offset
< XFS_INODES_PER_CHUNK
);
986 ASSERT((XFS_AGINO_TO_OFFSET(mp
, rec
.ir_startino
) %
987 XFS_INODES_PER_CHUNK
) == 0);
988 ino
= XFS_AGINO_TO_INO(mp
, agno
, rec
.ir_startino
+ offset
);
989 rec
.ir_free
&= ~XFS_INOBT_MASK(offset
);
991 error
= xfs_inobt_update(cur
, &rec
);
994 be32_add_cpu(&agi
->agi_freecount
, -1);
995 xfs_ialloc_log_agi(tp
, agbp
, XFS_AGI_FREECOUNT
);
996 pag
->pagi_freecount
--;
998 error
= xfs_check_agi_freecount(cur
, agi
);
1002 xfs_btree_del_cursor(cur
, XFS_BTREE_NOERROR
);
1003 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_IFREE
, -1);
1008 xfs_btree_del_cursor(tcur
, XFS_BTREE_ERROR
);
1010 xfs_btree_del_cursor(cur
, XFS_BTREE_ERROR
);
1016 * Free disk inode. Carefully avoids touching the incore inode, all
1017 * manipulations incore are the caller's responsibility.
1018 * The on-disk inode is not changed by this operation, only the
1019 * btree (free inode mask) is changed.
1023 xfs_trans_t
*tp
, /* transaction pointer */
1024 xfs_ino_t inode
, /* inode to be freed */
1025 xfs_bmap_free_t
*flist
, /* extents to free */
1026 int *delete, /* set if inode cluster was deleted */
1027 xfs_ino_t
*first_ino
) /* first inode in deleted cluster */
1030 xfs_agblock_t agbno
; /* block number containing inode */
1031 xfs_buf_t
*agbp
; /* buffer containing allocation group header */
1032 xfs_agino_t agino
; /* inode number relative to allocation group */
1033 xfs_agnumber_t agno
; /* allocation group number */
1034 xfs_agi_t
*agi
; /* allocation group header */
1035 xfs_btree_cur_t
*cur
; /* inode btree cursor */
1036 int error
; /* error return value */
1037 int i
; /* result code */
1038 int ilen
; /* inodes in an inode cluster */
1039 xfs_mount_t
*mp
; /* mount structure for filesystem */
1040 int off
; /* offset of inode in inode chunk */
1041 xfs_inobt_rec_incore_t rec
; /* btree record */
1042 struct xfs_perag
*pag
;
1047 * Break up inode number into its components.
1049 agno
= XFS_INO_TO_AGNO(mp
, inode
);
1050 if (agno
>= mp
->m_sb
.sb_agcount
) {
1052 "xfs_difree: agno >= mp->m_sb.sb_agcount (%d >= %d) on %s. Returning EINVAL.",
1053 agno
, mp
->m_sb
.sb_agcount
, mp
->m_fsname
);
1055 return XFS_ERROR(EINVAL
);
1057 agino
= XFS_INO_TO_AGINO(mp
, inode
);
1058 if (inode
!= XFS_AGINO_TO_INO(mp
, agno
, agino
)) {
1060 "xfs_difree: inode != XFS_AGINO_TO_INO() "
1061 "(%llu != %llu) on %s. Returning EINVAL.",
1062 (unsigned long long)inode
,
1063 (unsigned long long)XFS_AGINO_TO_INO(mp
, agno
, agino
),
1066 return XFS_ERROR(EINVAL
);
1068 agbno
= XFS_AGINO_TO_AGBNO(mp
, agino
);
1069 if (agbno
>= mp
->m_sb
.sb_agblocks
) {
1071 "xfs_difree: agbno >= mp->m_sb.sb_agblocks (%d >= %d) on %s. Returning EINVAL.",
1072 agbno
, mp
->m_sb
.sb_agblocks
, mp
->m_fsname
);
1074 return XFS_ERROR(EINVAL
);
1077 * Get the allocation group header.
1079 error
= xfs_ialloc_read_agi(mp
, tp
, agno
, &agbp
);
1082 "xfs_difree: xfs_ialloc_read_agi() returned an error %d on %s. Returning error.",
1083 error
, mp
->m_fsname
);
1086 agi
= XFS_BUF_TO_AGI(agbp
);
1087 ASSERT(be32_to_cpu(agi
->agi_magicnum
) == XFS_AGI_MAGIC
);
1088 ASSERT(agbno
< be32_to_cpu(agi
->agi_length
));
1090 * Initialize the cursor.
1092 cur
= xfs_inobt_init_cursor(mp
, tp
, agbp
, agno
);
1094 error
= xfs_check_agi_freecount(cur
, agi
);
1099 * Look for the entry describing this inode.
1101 if ((error
= xfs_inobt_lookup(cur
, agino
, XFS_LOOKUP_LE
, &i
))) {
1103 "xfs_difree: xfs_inobt_lookup returned() an error %d on %s. Returning error.",
1104 error
, mp
->m_fsname
);
1107 XFS_WANT_CORRUPTED_GOTO(i
== 1, error0
);
1108 error
= xfs_inobt_get_rec(cur
, &rec
, &i
);
1111 "xfs_difree: xfs_inobt_get_rec() returned an error %d on %s. Returning error.",
1112 error
, mp
->m_fsname
);
1115 XFS_WANT_CORRUPTED_GOTO(i
== 1, error0
);
1117 * Get the offset in the inode chunk.
1119 off
= agino
- rec
.ir_startino
;
1120 ASSERT(off
>= 0 && off
< XFS_INODES_PER_CHUNK
);
1121 ASSERT(!(rec
.ir_free
& XFS_INOBT_MASK(off
)));
1123 * Mark the inode free & increment the count.
1125 rec
.ir_free
|= XFS_INOBT_MASK(off
);
1129 * When an inode cluster is free, it becomes eligible for removal
1131 if (!(mp
->m_flags
& XFS_MOUNT_IKEEP
) &&
1132 (rec
.ir_freecount
== XFS_IALLOC_INODES(mp
))) {
1135 *first_ino
= XFS_AGINO_TO_INO(mp
, agno
, rec
.ir_startino
);
1138 * Remove the inode cluster from the AGI B+Tree, adjust the
1139 * AGI and Superblock inode counts, and mark the disk space
1140 * to be freed when the transaction is committed.
1142 ilen
= XFS_IALLOC_INODES(mp
);
1143 be32_add_cpu(&agi
->agi_count
, -ilen
);
1144 be32_add_cpu(&agi
->agi_freecount
, -(ilen
- 1));
1145 xfs_ialloc_log_agi(tp
, agbp
, XFS_AGI_COUNT
| XFS_AGI_FREECOUNT
);
1146 pag
= xfs_perag_get(mp
, agno
);
1147 pag
->pagi_freecount
-= ilen
- 1;
1149 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_ICOUNT
, -ilen
);
1150 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_IFREE
, -(ilen
- 1));
1152 if ((error
= xfs_btree_delete(cur
, &i
))) {
1153 cmn_err(CE_WARN
, "xfs_difree: xfs_btree_delete returned an error %d on %s.\n",
1154 error
, mp
->m_fsname
);
1158 xfs_bmap_add_free(XFS_AGB_TO_FSB(mp
,
1159 agno
, XFS_INO_TO_AGBNO(mp
,rec
.ir_startino
)),
1160 XFS_IALLOC_BLOCKS(mp
), flist
, mp
);
1164 error
= xfs_inobt_update(cur
, &rec
);
1167 "xfs_difree: xfs_inobt_update returned an error %d on %s.",
1168 error
, mp
->m_fsname
);
1173 * Change the inode free counts and log the ag/sb changes.
1175 be32_add_cpu(&agi
->agi_freecount
, 1);
1176 xfs_ialloc_log_agi(tp
, agbp
, XFS_AGI_FREECOUNT
);
1177 pag
= xfs_perag_get(mp
, agno
);
1178 pag
->pagi_freecount
++;
1180 xfs_trans_mod_sb(tp
, XFS_TRANS_SB_IFREE
, 1);
1183 error
= xfs_check_agi_freecount(cur
, agi
);
1187 xfs_btree_del_cursor(cur
, XFS_BTREE_NOERROR
);
1191 xfs_btree_del_cursor(cur
, XFS_BTREE_ERROR
);
1197 struct xfs_mount
*mp
,
1198 struct xfs_trans
*tp
,
1199 xfs_agnumber_t agno
,
1201 xfs_agblock_t agbno
,
1202 xfs_agblock_t
*chunk_agbno
,
1203 xfs_agblock_t
*offset_agbno
,
1206 struct xfs_inobt_rec_incore rec
;
1207 struct xfs_btree_cur
*cur
;
1208 struct xfs_buf
*agbp
;
1212 error
= xfs_ialloc_read_agi(mp
, tp
, agno
, &agbp
);
1214 xfs_fs_cmn_err(CE_ALERT
, mp
, "xfs_imap: "
1215 "xfs_ialloc_read_agi() returned "
1216 "error %d, agno %d",
1222 * Lookup the inode record for the given agino. If the record cannot be
1223 * found, then it's an invalid inode number and we should abort. Once
1224 * we have a record, we need to ensure it contains the inode number
1225 * we are looking up.
1227 cur
= xfs_inobt_init_cursor(mp
, tp
, agbp
, agno
);
1228 error
= xfs_inobt_lookup(cur
, agino
, XFS_LOOKUP_LE
, &i
);
1231 error
= xfs_inobt_get_rec(cur
, &rec
, &i
);
1232 if (!error
&& i
== 0)
1236 xfs_trans_brelse(tp
, agbp
);
1237 xfs_btree_del_cursor(cur
, XFS_BTREE_NOERROR
);
1241 /* check that the returned record contains the required inode */
1242 if (rec
.ir_startino
> agino
||
1243 rec
.ir_startino
+ XFS_IALLOC_INODES(mp
) <= agino
)
1246 /* for untrusted inodes check it is allocated first */
1247 if ((flags
& XFS_IGET_UNTRUSTED
) &&
1248 (rec
.ir_free
& XFS_INOBT_MASK(agino
- rec
.ir_startino
)))
1251 *chunk_agbno
= XFS_AGINO_TO_AGBNO(mp
, rec
.ir_startino
);
1252 *offset_agbno
= agbno
- *chunk_agbno
;
1257 * Return the location of the inode in imap, for mapping it into a buffer.
1261 xfs_mount_t
*mp
, /* file system mount structure */
1262 xfs_trans_t
*tp
, /* transaction pointer */
1263 xfs_ino_t ino
, /* inode to locate */
1264 struct xfs_imap
*imap
, /* location map structure */
1265 uint flags
) /* flags for inode btree lookup */
1267 xfs_agblock_t agbno
; /* block number of inode in the alloc group */
1268 xfs_agino_t agino
; /* inode number within alloc group */
1269 xfs_agnumber_t agno
; /* allocation group number */
1270 int blks_per_cluster
; /* num blocks per inode cluster */
1271 xfs_agblock_t chunk_agbno
; /* first block in inode chunk */
1272 xfs_agblock_t cluster_agbno
; /* first block in inode cluster */
1273 int error
; /* error code */
1274 int offset
; /* index of inode in its buffer */
1275 int offset_agbno
; /* blks from chunk start to inode */
1277 ASSERT(ino
!= NULLFSINO
);
1280 * Split up the inode number into its parts.
1282 agno
= XFS_INO_TO_AGNO(mp
, ino
);
1283 agino
= XFS_INO_TO_AGINO(mp
, ino
);
1284 agbno
= XFS_AGINO_TO_AGBNO(mp
, agino
);
1285 if (agno
>= mp
->m_sb
.sb_agcount
|| agbno
>= mp
->m_sb
.sb_agblocks
||
1286 ino
!= XFS_AGINO_TO_INO(mp
, agno
, agino
)) {
1289 * Don't output diagnostic information for untrusted inodes
1290 * as they can be invalid without implying corruption.
1292 if (flags
& XFS_IGET_UNTRUSTED
)
1293 return XFS_ERROR(EINVAL
);
1294 if (agno
>= mp
->m_sb
.sb_agcount
) {
1295 xfs_fs_cmn_err(CE_ALERT
, mp
,
1296 "xfs_imap: agno (%d) >= "
1297 "mp->m_sb.sb_agcount (%d)",
1298 agno
, mp
->m_sb
.sb_agcount
);
1300 if (agbno
>= mp
->m_sb
.sb_agblocks
) {
1301 xfs_fs_cmn_err(CE_ALERT
, mp
,
1302 "xfs_imap: agbno (0x%llx) >= "
1303 "mp->m_sb.sb_agblocks (0x%lx)",
1304 (unsigned long long) agbno
,
1305 (unsigned long) mp
->m_sb
.sb_agblocks
);
1307 if (ino
!= XFS_AGINO_TO_INO(mp
, agno
, agino
)) {
1308 xfs_fs_cmn_err(CE_ALERT
, mp
,
1309 "xfs_imap: ino (0x%llx) != "
1310 "XFS_AGINO_TO_INO(mp, agno, agino) "
1312 ino
, XFS_AGINO_TO_INO(mp
, agno
, agino
));
1316 return XFS_ERROR(EINVAL
);
1319 blks_per_cluster
= XFS_INODE_CLUSTER_SIZE(mp
) >> mp
->m_sb
.sb_blocklog
;
1322 * For bulkstat and handle lookups, we have an untrusted inode number
1323 * that we have to verify is valid. We cannot do this just by reading
1324 * the inode buffer as it may have been unlinked and removed leaving
1325 * inodes in stale state on disk. Hence we have to do a btree lookup
1326 * in all cases where an untrusted inode number is passed.
1328 if (flags
& XFS_IGET_UNTRUSTED
) {
1329 error
= xfs_imap_lookup(mp
, tp
, agno
, agino
, agbno
,
1330 &chunk_agbno
, &offset_agbno
, flags
);
1337 * If the inode cluster size is the same as the blocksize or
1338 * smaller we get to the buffer by simple arithmetics.
1340 if (XFS_INODE_CLUSTER_SIZE(mp
) <= mp
->m_sb
.sb_blocksize
) {
1341 offset
= XFS_INO_TO_OFFSET(mp
, ino
);
1342 ASSERT(offset
< mp
->m_sb
.sb_inopblock
);
1344 imap
->im_blkno
= XFS_AGB_TO_DADDR(mp
, agno
, agbno
);
1345 imap
->im_len
= XFS_FSB_TO_BB(mp
, 1);
1346 imap
->im_boffset
= (ushort
)(offset
<< mp
->m_sb
.sb_inodelog
);
1351 * If the inode chunks are aligned then use simple maths to
1352 * find the location. Otherwise we have to do a btree
1353 * lookup to find the location.
1355 if (mp
->m_inoalign_mask
) {
1356 offset_agbno
= agbno
& mp
->m_inoalign_mask
;
1357 chunk_agbno
= agbno
- offset_agbno
;
1359 error
= xfs_imap_lookup(mp
, tp
, agno
, agino
, agbno
,
1360 &chunk_agbno
, &offset_agbno
, flags
);
1366 ASSERT(agbno
>= chunk_agbno
);
1367 cluster_agbno
= chunk_agbno
+
1368 ((offset_agbno
/ blks_per_cluster
) * blks_per_cluster
);
1369 offset
= ((agbno
- cluster_agbno
) * mp
->m_sb
.sb_inopblock
) +
1370 XFS_INO_TO_OFFSET(mp
, ino
);
1372 imap
->im_blkno
= XFS_AGB_TO_DADDR(mp
, agno
, cluster_agbno
);
1373 imap
->im_len
= XFS_FSB_TO_BB(mp
, blks_per_cluster
);
1374 imap
->im_boffset
= (ushort
)(offset
<< mp
->m_sb
.sb_inodelog
);
1377 * If the inode number maps to a block outside the bounds
1378 * of the file system then return NULL rather than calling
1379 * read_buf and panicing when we get an error from the
1382 if ((imap
->im_blkno
+ imap
->im_len
) >
1383 XFS_FSB_TO_BB(mp
, mp
->m_sb
.sb_dblocks
)) {
1384 xfs_fs_cmn_err(CE_ALERT
, mp
, "xfs_imap: "
1385 "(imap->im_blkno (0x%llx) + imap->im_len (0x%llx)) > "
1386 " XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks) (0x%llx)",
1387 (unsigned long long) imap
->im_blkno
,
1388 (unsigned long long) imap
->im_len
,
1389 XFS_FSB_TO_BB(mp
, mp
->m_sb
.sb_dblocks
));
1390 return XFS_ERROR(EINVAL
);
1396 * Compute and fill in value of m_in_maxlevels.
1399 xfs_ialloc_compute_maxlevels(
1400 xfs_mount_t
*mp
) /* file system mount structure */
1408 maxleafents
= (1LL << XFS_INO_AGINO_BITS(mp
)) >>
1409 XFS_INODES_PER_CHUNK_LOG
;
1410 minleafrecs
= mp
->m_alloc_mnr
[0];
1411 minnoderecs
= mp
->m_alloc_mnr
[1];
1412 maxblocks
= (maxleafents
+ minleafrecs
- 1) / minleafrecs
;
1413 for (level
= 1; maxblocks
> 1; level
++)
1414 maxblocks
= (maxblocks
+ minnoderecs
- 1) / minnoderecs
;
1415 mp
->m_in_maxlevels
= level
;
1419 * Log specified fields for the ag hdr (inode section)
1423 xfs_trans_t
*tp
, /* transaction pointer */
1424 xfs_buf_t
*bp
, /* allocation group header buffer */
1425 int fields
) /* bitmask of fields to log */
1427 int first
; /* first byte number */
1428 int last
; /* last byte number */
1429 static const short offsets
[] = { /* field starting offsets */
1430 /* keep in sync with bit definitions */
1431 offsetof(xfs_agi_t
, agi_magicnum
),
1432 offsetof(xfs_agi_t
, agi_versionnum
),
1433 offsetof(xfs_agi_t
, agi_seqno
),
1434 offsetof(xfs_agi_t
, agi_length
),
1435 offsetof(xfs_agi_t
, agi_count
),
1436 offsetof(xfs_agi_t
, agi_root
),
1437 offsetof(xfs_agi_t
, agi_level
),
1438 offsetof(xfs_agi_t
, agi_freecount
),
1439 offsetof(xfs_agi_t
, agi_newino
),
1440 offsetof(xfs_agi_t
, agi_dirino
),
1441 offsetof(xfs_agi_t
, agi_unlinked
),
1445 xfs_agi_t
*agi
; /* allocation group header */
1447 agi
= XFS_BUF_TO_AGI(bp
);
1448 ASSERT(be32_to_cpu(agi
->agi_magicnum
) == XFS_AGI_MAGIC
);
1451 * Compute byte offsets for the first and last fields.
1453 xfs_btree_offsets(fields
, offsets
, XFS_AGI_NUM_BITS
, &first
, &last
);
1455 * Log the allocation group inode header buffer.
1457 xfs_trans_log_buf(tp
, bp
, first
, last
);
1462 xfs_check_agi_unlinked(
1463 struct xfs_agi
*agi
)
1467 for (i
= 0; i
< XFS_AGI_UNLINKED_BUCKETS
; i
++)
1468 ASSERT(agi
->agi_unlinked
[i
]);
1471 #define xfs_check_agi_unlinked(agi)
1475 * Read in the allocation group header (inode allocation section)
1479 struct xfs_mount
*mp
, /* file system mount structure */
1480 struct xfs_trans
*tp
, /* transaction pointer */
1481 xfs_agnumber_t agno
, /* allocation group number */
1482 struct xfs_buf
**bpp
) /* allocation group hdr buf */
1484 struct xfs_agi
*agi
; /* allocation group header */
1485 int agi_ok
; /* agi is consistent */
1488 ASSERT(agno
!= NULLAGNUMBER
);
1490 error
= xfs_trans_read_buf(mp
, tp
, mp
->m_ddev_targp
,
1491 XFS_AG_DADDR(mp
, agno
, XFS_AGI_DADDR(mp
)),
1492 XFS_FSS_TO_BB(mp
, 1), 0, bpp
);
1496 ASSERT(*bpp
&& !XFS_BUF_GETERROR(*bpp
));
1497 agi
= XFS_BUF_TO_AGI(*bpp
);
1500 * Validate the magic number of the agi block.
1502 agi_ok
= be32_to_cpu(agi
->agi_magicnum
) == XFS_AGI_MAGIC
&&
1503 XFS_AGI_GOOD_VERSION(be32_to_cpu(agi
->agi_versionnum
)) &&
1504 be32_to_cpu(agi
->agi_seqno
) == agno
;
1505 if (unlikely(XFS_TEST_ERROR(!agi_ok
, mp
, XFS_ERRTAG_IALLOC_READ_AGI
,
1506 XFS_RANDOM_IALLOC_READ_AGI
))) {
1507 XFS_CORRUPTION_ERROR("xfs_read_agi", XFS_ERRLEVEL_LOW
,
1509 xfs_trans_brelse(tp
, *bpp
);
1510 return XFS_ERROR(EFSCORRUPTED
);
1513 XFS_BUF_SET_VTYPE_REF(*bpp
, B_FS_AGI
, XFS_AGI_REF
);
1515 xfs_check_agi_unlinked(agi
);
1520 xfs_ialloc_read_agi(
1521 struct xfs_mount
*mp
, /* file system mount structure */
1522 struct xfs_trans
*tp
, /* transaction pointer */
1523 xfs_agnumber_t agno
, /* allocation group number */
1524 struct xfs_buf
**bpp
) /* allocation group hdr buf */
1526 struct xfs_agi
*agi
; /* allocation group header */
1527 struct xfs_perag
*pag
; /* per allocation group data */
1530 error
= xfs_read_agi(mp
, tp
, agno
, bpp
);
1534 agi
= XFS_BUF_TO_AGI(*bpp
);
1535 pag
= xfs_perag_get(mp
, agno
);
1536 if (!pag
->pagi_init
) {
1537 pag
->pagi_freecount
= be32_to_cpu(agi
->agi_freecount
);
1538 pag
->pagi_count
= be32_to_cpu(agi
->agi_count
);
1543 * It's possible for these to be out of sync if
1544 * we are in the middle of a forced shutdown.
1546 ASSERT(pag
->pagi_freecount
== be32_to_cpu(agi
->agi_freecount
) ||
1547 XFS_FORCED_SHUTDOWN(mp
));
1553 * Read in the agi to initialise the per-ag data in the mount structure
1556 xfs_ialloc_pagi_init(
1557 xfs_mount_t
*mp
, /* file system mount structure */
1558 xfs_trans_t
*tp
, /* transaction pointer */
1559 xfs_agnumber_t agno
) /* allocation group number */
1561 xfs_buf_t
*bp
= NULL
;
1564 error
= xfs_ialloc_read_agi(mp
, tp
, agno
, &bp
);
1568 xfs_trans_brelse(tp
, bp
);